diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,128619 @@
+{
+  "best_metric": 1.35616612,
+  "best_model_checkpoint": "/home/ubuntu/s2/output_qwen1_5BI_2_epochs/qwen2_5-1_5b-instruct/v0-20250320-120700/checkpoint-64000",
+  "epoch": 1.6235413495687467,
+  "eval_steps": 1000,
+  "global_step": 64000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "acc": 0.6747824,
+      "epoch": 2.536783358701167e-05,
+      "grad_norm": 5.40625,
+      "learning_rate": 2.536783358701167e-09,
+      "loss": 1.51009965,
+      "memory(GiB)": 19.52,
+      "step": 1,
+      "train_speed(iter/s)": 0.030017
+    },
+    {
+      "acc": 0.64590144,
+      "epoch": 0.00012683916793505834,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.2683916793505834e-08,
+      "loss": 1.68001819,
+      "memory(GiB)": 21.44,
+      "step": 5,
+      "train_speed(iter/s)": 0.138413
+    },
+    {
+      "acc": 0.62538185,
+      "epoch": 0.0002536783358701167,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.536783358701167e-08,
+      "loss": 1.76752777,
+      "memory(GiB)": 21.46,
+      "step": 10,
+      "train_speed(iter/s)": 0.252917
+    },
+    {
+      "acc": 0.61482439,
+      "epoch": 0.000380517503805175,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.80517503805175e-08,
+      "loss": 1.84412689,
+      "memory(GiB)": 24.24,
+      "step": 15,
+      "train_speed(iter/s)": 0.349251
+    },
+    {
+      "acc": 0.62413063,
+      "epoch": 0.0005073566717402334,
+      "grad_norm": 5.53125,
+      "learning_rate": 5.073566717402334e-08,
+      "loss": 1.7661972,
+      "memory(GiB)": 24.24,
+      "step": 20,
+      "train_speed(iter/s)": 0.430891
+    },
+    {
+      "acc": 0.61876965,
+      "epoch": 0.0006341958396752917,
+      "grad_norm": 4.75,
+      "learning_rate": 6.341958396752917e-08,
+      "loss": 1.82740479,
+      "memory(GiB)": 24.24,
+      "step": 25,
+      "train_speed(iter/s)": 0.502129
+    },
+    {
+      "acc": 0.6343482,
+      "epoch": 0.00076103500761035,
+      "grad_norm": 8.4375,
+      "learning_rate": 7.6103500761035e-08,
+      "loss": 1.80051899,
+      "memory(GiB)": 24.24,
+      "step": 30,
+      "train_speed(iter/s)": 0.564111
+    },
+    {
+      "acc": 0.62518497,
+      "epoch": 0.0008878741755454084,
+      "grad_norm": 4.90625,
+      "learning_rate": 8.878741755454084e-08,
+      "loss": 1.7874094,
+      "memory(GiB)": 24.24,
+      "step": 35,
+      "train_speed(iter/s)": 0.618774
+    },
+    {
+      "acc": 0.64369349,
+      "epoch": 0.0010147133434804667,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.0147133434804667e-07,
+      "loss": 1.7368103,
+      "memory(GiB)": 24.24,
+      "step": 40,
+      "train_speed(iter/s)": 0.666756
+    },
+    {
+      "acc": 0.63235931,
+      "epoch": 0.001141552511415525,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.1415525114155251e-07,
+      "loss": 1.74555855,
+      "memory(GiB)": 24.24,
+      "step": 45,
+      "train_speed(iter/s)": 0.710076
+    },
+    {
+      "acc": 0.64212036,
+      "epoch": 0.0012683916793505834,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.2683916793505834e-07,
+      "loss": 1.75473633,
+      "memory(GiB)": 24.24,
+      "step": 50,
+      "train_speed(iter/s)": 0.748838
+    },
+    {
+      "acc": 0.63963547,
+      "epoch": 0.0013952308472856417,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.3952308472856418e-07,
+      "loss": 1.85870934,
+      "memory(GiB)": 24.28,
+      "step": 55,
+      "train_speed(iter/s)": 0.783796
+    },
+    {
+      "acc": 0.61117225,
+      "epoch": 0.0015220700152207,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.5220700152207e-07,
+      "loss": 1.83271122,
+      "memory(GiB)": 24.28,
+      "step": 60,
+      "train_speed(iter/s)": 0.815837
+    },
+    {
+      "acc": 0.6290905,
+      "epoch": 0.0016489091831557584,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.6489091831557585e-07,
+      "loss": 1.74984589,
+      "memory(GiB)": 24.28,
+      "step": 65,
+      "train_speed(iter/s)": 0.844864
+    },
+    {
+      "acc": 0.63701944,
+      "epoch": 0.0017757483510908167,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.7757483510908168e-07,
+      "loss": 1.74415321,
+      "memory(GiB)": 24.28,
+      "step": 70,
+      "train_speed(iter/s)": 0.871559
+    },
+    {
+      "acc": 0.61787643,
+      "epoch": 0.001902587519025875,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.9025875190258752e-07,
+      "loss": 1.83969574,
+      "memory(GiB)": 24.28,
+      "step": 75,
+      "train_speed(iter/s)": 0.895966
+    },
+    {
+      "acc": 0.64101725,
+      "epoch": 0.0020294266869609334,
+      "grad_norm": 6.28125,
+      "learning_rate": 2.0294266869609335e-07,
+      "loss": 1.70301571,
+      "memory(GiB)": 24.28,
+      "step": 80,
+      "train_speed(iter/s)": 0.918288
+    },
+    {
+      "acc": 0.60716228,
+      "epoch": 0.0021562658548959918,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.1562658548959918e-07,
+      "loss": 1.87644176,
+      "memory(GiB)": 30.16,
+      "step": 85,
+      "train_speed(iter/s)": 0.938929
+    },
+    {
+      "acc": 0.62798767,
+      "epoch": 0.00228310502283105,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.2831050228310502e-07,
+      "loss": 1.74083977,
+      "memory(GiB)": 30.16,
+      "step": 90,
+      "train_speed(iter/s)": 0.958227
+    },
+    {
+      "acc": 0.61846747,
+      "epoch": 0.0024099441907661084,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.409944190766109e-07,
+      "loss": 1.85337906,
+      "memory(GiB)": 30.16,
+      "step": 95,
+      "train_speed(iter/s)": 0.976303
+    },
+    {
+      "acc": 0.62845416,
+      "epoch": 0.0025367833587011668,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.536783358701167e-07,
+      "loss": 1.77967491,
+      "memory(GiB)": 30.16,
+      "step": 100,
+      "train_speed(iter/s)": 0.992892
+    },
+    {
+      "acc": 0.61112614,
+      "epoch": 0.002663622526636225,
+      "grad_norm": 4.875,
+      "learning_rate": 2.6636225266362255e-07,
+      "loss": 1.8328434,
+      "memory(GiB)": 30.19,
+      "step": 105,
+      "train_speed(iter/s)": 1.00849
+    },
+    {
+      "acc": 0.64181776,
+      "epoch": 0.0027904616945712835,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.7904616945712836e-07,
+      "loss": 1.72338181,
+      "memory(GiB)": 30.19,
+      "step": 110,
+      "train_speed(iter/s)": 1.023161
+    },
+    {
+      "acc": 0.63324089,
+      "epoch": 0.002917300862506342,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.917300862506342e-07,
+      "loss": 1.81151314,
+      "memory(GiB)": 30.19,
+      "step": 115,
+      "train_speed(iter/s)": 1.037018
+    },
+    {
+      "acc": 0.62986326,
+      "epoch": 0.0030441400304414,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.0441400304414e-07,
+      "loss": 1.77894821,
+      "memory(GiB)": 30.19,
+      "step": 120,
+      "train_speed(iter/s)": 1.049938
+    },
+    {
+      "acc": 0.63567829,
+      "epoch": 0.0031709791983764585,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.170979198376459e-07,
+      "loss": 1.71367035,
+      "memory(GiB)": 30.19,
+      "step": 125,
+      "train_speed(iter/s)": 1.062194
+    },
+    {
+      "acc": 0.62116871,
+      "epoch": 0.003297818366311517,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.297818366311517e-07,
+      "loss": 1.76715736,
+      "memory(GiB)": 30.19,
+      "step": 130,
+      "train_speed(iter/s)": 1.073772
+    },
+    {
+      "acc": 0.63459816,
+      "epoch": 0.003424657534246575,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.4246575342465755e-07,
+      "loss": 1.737883,
+      "memory(GiB)": 30.19,
+      "step": 135,
+      "train_speed(iter/s)": 1.084678
+    },
+    {
+      "acc": 0.62940059,
+      "epoch": 0.0035514967021816335,
+      "grad_norm": 6.125,
+      "learning_rate": 3.5514967021816336e-07,
+      "loss": 1.73199425,
+      "memory(GiB)": 30.19,
+      "step": 140,
+      "train_speed(iter/s)": 1.095094
+    },
+    {
+      "acc": 0.62991939,
+      "epoch": 0.003678335870116692,
+      "grad_norm": 6.03125,
+      "learning_rate": 3.678335870116692e-07,
+      "loss": 1.78003654,
+      "memory(GiB)": 30.19,
+      "step": 145,
+      "train_speed(iter/s)": 1.104635
+    },
+    {
+      "acc": 0.63566003,
+      "epoch": 0.00380517503805175,
+      "grad_norm": 5.65625,
+      "learning_rate": 3.8051750380517503e-07,
+      "loss": 1.72479763,
+      "memory(GiB)": 30.19,
+      "step": 150,
+      "train_speed(iter/s)": 1.113902
+    },
+    {
+      "acc": 0.63350668,
+      "epoch": 0.0039320142059868085,
+      "grad_norm": 6.59375,
+      "learning_rate": 3.932014205986809e-07,
+      "loss": 1.74458656,
+      "memory(GiB)": 30.19,
+      "step": 155,
+      "train_speed(iter/s)": 1.123015
+    },
+    {
+      "acc": 0.63794727,
+      "epoch": 0.004058853373921867,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.058853373921867e-07,
+      "loss": 1.75714798,
+      "memory(GiB)": 30.19,
+      "step": 160,
+      "train_speed(iter/s)": 1.131445
+    },
+    {
+      "acc": 0.62013307,
+      "epoch": 0.004185692541856925,
+      "grad_norm": 5.8125,
+      "learning_rate": 4.1856925418569256e-07,
+      "loss": 1.83087234,
+      "memory(GiB)": 30.19,
+      "step": 165,
+      "train_speed(iter/s)": 1.138978
+    },
+    {
+      "acc": 0.61516752,
+      "epoch": 0.0043125317097919835,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.3125317097919837e-07,
+      "loss": 1.82721634,
+      "memory(GiB)": 30.19,
+      "step": 170,
+      "train_speed(iter/s)": 1.146652
+    },
+    {
+      "acc": 0.62418947,
+      "epoch": 0.004439370877727042,
+      "grad_norm": 6.59375,
+      "learning_rate": 4.4393708777270423e-07,
+      "loss": 1.84877129,
+      "memory(GiB)": 30.19,
+      "step": 175,
+      "train_speed(iter/s)": 1.15406
+    },
+    {
+      "acc": 0.62058082,
+      "epoch": 0.0045662100456621,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.5662100456621004e-07,
+      "loss": 1.86236076,
+      "memory(GiB)": 30.19,
+      "step": 180,
+      "train_speed(iter/s)": 1.160974
+    },
+    {
+      "acc": 0.64504991,
+      "epoch": 0.0046930492135971585,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.693049213597159e-07,
+      "loss": 1.75609474,
+      "memory(GiB)": 30.19,
+      "step": 185,
+      "train_speed(iter/s)": 1.167849
+    },
+    {
+      "acc": 0.62283936,
+      "epoch": 0.004819888381532217,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.819888381532218e-07,
+      "loss": 1.76074944,
+      "memory(GiB)": 30.19,
+      "step": 190,
+      "train_speed(iter/s)": 1.174213
+    },
+    {
+      "acc": 0.63673005,
+      "epoch": 0.004946727549467275,
+      "grad_norm": 4.625,
+      "learning_rate": 4.946727549467275e-07,
+      "loss": 1.79760208,
+      "memory(GiB)": 30.19,
+      "step": 195,
+      "train_speed(iter/s)": 1.180345
+    },
+    {
+      "acc": 0.62707038,
+      "epoch": 0.0050735667174023336,
+      "grad_norm": 4.71875,
+      "learning_rate": 5.073566717402334e-07,
+      "loss": 1.79353294,
+      "memory(GiB)": 30.19,
+      "step": 200,
+      "train_speed(iter/s)": 1.186149
+    },
+    {
+      "acc": 0.62641711,
+      "epoch": 0.005200405885337392,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.200405885337392e-07,
+      "loss": 1.81665859,
+      "memory(GiB)": 30.19,
+      "step": 205,
+      "train_speed(iter/s)": 1.191778
+    },
+    {
+      "acc": 0.6297471,
+      "epoch": 0.00532724505327245,
+      "grad_norm": 4.625,
+      "learning_rate": 5.327245053272451e-07,
+      "loss": 1.78159313,
+      "memory(GiB)": 30.19,
+      "step": 210,
+      "train_speed(iter/s)": 1.19715
+    },
+    {
+      "acc": 0.61491327,
+      "epoch": 0.005454084221207509,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.454084221207509e-07,
+      "loss": 1.8028698,
+      "memory(GiB)": 30.19,
+      "step": 215,
+      "train_speed(iter/s)": 1.202502
+    },
+    {
+      "acc": 0.63494024,
+      "epoch": 0.005580923389142567,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.580923389142567e-07,
+      "loss": 1.70176086,
+      "memory(GiB)": 30.19,
+      "step": 220,
+      "train_speed(iter/s)": 1.207442
+    },
+    {
+      "acc": 0.63262463,
+      "epoch": 0.005707762557077625,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.707762557077626e-07,
+      "loss": 1.74788551,
+      "memory(GiB)": 30.19,
+      "step": 225,
+      "train_speed(iter/s)": 1.21221
+    },
+    {
+      "acc": 0.62658453,
+      "epoch": 0.005834601725012684,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.834601725012684e-07,
+      "loss": 1.78664398,
+      "memory(GiB)": 30.19,
+      "step": 230,
+      "train_speed(iter/s)": 1.216819
+    },
+    {
+      "acc": 0.63818951,
+      "epoch": 0.005961440892947742,
+      "grad_norm": 5.21875,
+      "learning_rate": 5.961440892947743e-07,
+      "loss": 1.71231804,
+      "memory(GiB)": 30.19,
+      "step": 235,
+      "train_speed(iter/s)": 1.221464
+    },
+    {
+      "acc": 0.62040734,
+      "epoch": 0.0060882800608828,
+      "grad_norm": 4.125,
+      "learning_rate": 6.0882800608828e-07,
+      "loss": 1.83989029,
+      "memory(GiB)": 30.19,
+      "step": 240,
+      "train_speed(iter/s)": 1.225886
+    },
+    {
+      "acc": 0.63575258,
+      "epoch": 0.006215119228817859,
+      "grad_norm": 5.4375,
+      "learning_rate": 6.215119228817859e-07,
+      "loss": 1.76745968,
+      "memory(GiB)": 30.19,
+      "step": 245,
+      "train_speed(iter/s)": 1.230167
+    },
+    {
+      "acc": 0.62307863,
+      "epoch": 0.006341958396752917,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.341958396752918e-07,
+      "loss": 1.89265423,
+      "memory(GiB)": 30.19,
+      "step": 250,
+      "train_speed(iter/s)": 1.234425
+    },
+    {
+      "acc": 0.60853238,
+      "epoch": 0.006468797564687975,
+      "grad_norm": 5.375,
+      "learning_rate": 6.468797564687976e-07,
+      "loss": 1.83797398,
+      "memory(GiB)": 30.19,
+      "step": 255,
+      "train_speed(iter/s)": 1.238475
+    },
+    {
+      "acc": 0.62592688,
+      "epoch": 0.006595636732623034,
+      "grad_norm": 4.8125,
+      "learning_rate": 6.595636732623034e-07,
+      "loss": 1.75319939,
+      "memory(GiB)": 30.19,
+      "step": 260,
+      "train_speed(iter/s)": 1.242389
+    },
+    {
+      "acc": 0.64444304,
+      "epoch": 0.006722475900558092,
+      "grad_norm": 4.59375,
+      "learning_rate": 6.722475900558092e-07,
+      "loss": 1.76238384,
+      "memory(GiB)": 30.19,
+      "step": 265,
+      "train_speed(iter/s)": 1.246329
+    },
+    {
+      "acc": 0.64357762,
+      "epoch": 0.00684931506849315,
+      "grad_norm": 5.28125,
+      "learning_rate": 6.849315068493151e-07,
+      "loss": 1.7203249,
+      "memory(GiB)": 30.19,
+      "step": 270,
+      "train_speed(iter/s)": 1.250052
+    },
+    {
+      "acc": 0.62204256,
+      "epoch": 0.006976154236428209,
+      "grad_norm": 4.5,
+      "learning_rate": 6.97615423642821e-07,
+      "loss": 1.76874084,
+      "memory(GiB)": 30.19,
+      "step": 275,
+      "train_speed(iter/s)": 1.253579
+    },
+    {
+      "acc": 0.62043686,
+      "epoch": 0.007102993404363267,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.102993404363267e-07,
+      "loss": 1.79279518,
+      "memory(GiB)": 30.19,
+      "step": 280,
+      "train_speed(iter/s)": 1.257002
+    },
+    {
+      "acc": 0.64211121,
+      "epoch": 0.007229832572298325,
+      "grad_norm": 4.65625,
+      "learning_rate": 7.229832572298326e-07,
+      "loss": 1.68224335,
+      "memory(GiB)": 30.19,
+      "step": 285,
+      "train_speed(iter/s)": 1.260351
+    },
+    {
+      "acc": 0.65111508,
+      "epoch": 0.007356671740233384,
+      "grad_norm": 4.78125,
+      "learning_rate": 7.356671740233384e-07,
+      "loss": 1.72094669,
+      "memory(GiB)": 30.19,
+      "step": 290,
+      "train_speed(iter/s)": 1.263579
+    },
+    {
+      "acc": 0.63957605,
+      "epoch": 0.007483510908168442,
+      "grad_norm": 5.40625,
+      "learning_rate": 7.483510908168443e-07,
+      "loss": 1.76124382,
+      "memory(GiB)": 30.19,
+      "step": 295,
+      "train_speed(iter/s)": 1.266708
+    },
+    {
+      "acc": 0.6364728,
+      "epoch": 0.0076103500761035,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.610350076103501e-07,
+      "loss": 1.71820126,
+      "memory(GiB)": 36.93,
+      "step": 300,
+      "train_speed(iter/s)": 1.269623
+    },
+    {
+      "acc": 0.63716264,
+      "epoch": 0.007737189244038559,
+      "grad_norm": 5.0625,
+      "learning_rate": 7.737189244038559e-07,
+      "loss": 1.77769775,
+      "memory(GiB)": 36.93,
+      "step": 305,
+      "train_speed(iter/s)": 1.272652
+    },
+    {
+      "acc": 0.63538837,
+      "epoch": 0.007864028411973617,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.864028411973618e-07,
+      "loss": 1.69095211,
+      "memory(GiB)": 36.93,
+      "step": 310,
+      "train_speed(iter/s)": 1.275511
+    },
+    {
+      "acc": 0.62574291,
+      "epoch": 0.007990867579908675,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.990867579908676e-07,
+      "loss": 1.83862724,
+      "memory(GiB)": 36.93,
+      "step": 315,
+      "train_speed(iter/s)": 1.278314
+    },
+    {
+      "acc": 0.63489847,
+      "epoch": 0.008117706747843734,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.117706747843734e-07,
+      "loss": 1.77034588,
+      "memory(GiB)": 36.93,
+      "step": 320,
+      "train_speed(iter/s)": 1.280912
+    },
+    {
+      "acc": 0.61174889,
+      "epoch": 0.008244545915778792,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.244545915778793e-07,
+      "loss": 1.81628819,
+      "memory(GiB)": 36.93,
+      "step": 325,
+      "train_speed(iter/s)": 1.283475
+    },
+    {
+      "acc": 0.62442942,
+      "epoch": 0.00837138508371385,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.371385083713851e-07,
+      "loss": 1.80433617,
+      "memory(GiB)": 36.93,
+      "step": 330,
+      "train_speed(iter/s)": 1.286062
+    },
+    {
+      "acc": 0.62974072,
+      "epoch": 0.008498224251648909,
+      "grad_norm": 4.90625,
+      "learning_rate": 8.49822425164891e-07,
+      "loss": 1.82213898,
+      "memory(GiB)": 36.93,
+      "step": 335,
+      "train_speed(iter/s)": 1.288646
+    },
+    {
+      "acc": 0.62853198,
+      "epoch": 0.008625063419583967,
+      "grad_norm": 4.625,
+      "learning_rate": 8.625063419583967e-07,
+      "loss": 1.75286217,
+      "memory(GiB)": 36.93,
+      "step": 340,
+      "train_speed(iter/s)": 1.291097
+    },
+    {
+      "acc": 0.63615813,
+      "epoch": 0.008751902587519025,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.751902587519026e-07,
+      "loss": 1.79281406,
+      "memory(GiB)": 36.93,
+      "step": 345,
+      "train_speed(iter/s)": 1.293496
+    },
+    {
+      "acc": 0.61542606,
+      "epoch": 0.008878741755454084,
+      "grad_norm": 4.71875,
+      "learning_rate": 8.878741755454085e-07,
+      "loss": 1.77391472,
+      "memory(GiB)": 36.93,
+      "step": 350,
+      "train_speed(iter/s)": 1.295938
+    },
+    {
+      "acc": 0.63571787,
+      "epoch": 0.009005580923389142,
+      "grad_norm": 5.9375,
+      "learning_rate": 9.005580923389143e-07,
+      "loss": 1.77171326,
+      "memory(GiB)": 36.93,
+      "step": 355,
+      "train_speed(iter/s)": 1.298248
+    },
+    {
+      "acc": 0.63448629,
+      "epoch": 0.0091324200913242,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.132420091324201e-07,
+      "loss": 1.68704281,
+      "memory(GiB)": 36.93,
+      "step": 360,
+      "train_speed(iter/s)": 1.30007
+    },
+    {
+      "acc": 0.63694887,
+      "epoch": 0.009259259259259259,
+      "grad_norm": 5.46875,
+      "learning_rate": 9.259259259259259e-07,
+      "loss": 1.7710228,
+      "memory(GiB)": 36.93,
+      "step": 365,
+      "train_speed(iter/s)": 1.302247
+    },
+    {
+      "acc": 0.62909341,
+      "epoch": 0.009386098427194317,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.386098427194318e-07,
+      "loss": 1.77352295,
+      "memory(GiB)": 36.93,
+      "step": 370,
+      "train_speed(iter/s)": 1.304369
+    },
+    {
+      "acc": 0.62529068,
+      "epoch": 0.009512937595129375,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.512937595129377e-07,
+      "loss": 1.85217628,
+      "memory(GiB)": 36.93,
+      "step": 375,
+      "train_speed(iter/s)": 1.306485
+    },
+    {
+      "acc": 0.64148026,
+      "epoch": 0.009639776763064434,
+      "grad_norm": 5.4375,
+      "learning_rate": 9.639776763064435e-07,
+      "loss": 1.71427536,
+      "memory(GiB)": 36.93,
+      "step": 380,
+      "train_speed(iter/s)": 1.308158
+    },
+    {
+      "acc": 0.63076034,
+      "epoch": 0.009766615930999492,
+      "grad_norm": 5.625,
+      "learning_rate": 9.766615930999493e-07,
+      "loss": 1.76640377,
+      "memory(GiB)": 36.93,
+      "step": 385,
+      "train_speed(iter/s)": 1.310142
+    },
+    {
+      "acc": 0.63195171,
+      "epoch": 0.00989345509893455,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.89345509893455e-07,
+      "loss": 1.7636879,
+      "memory(GiB)": 36.93,
+      "step": 390,
+      "train_speed(iter/s)": 1.312113
+    },
+    {
+      "acc": 0.6355206,
+      "epoch": 0.010020294266869609,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.002029426686961e-06,
+      "loss": 1.76655579,
+      "memory(GiB)": 36.93,
+      "step": 395,
+      "train_speed(iter/s)": 1.314042
+    },
+    {
+      "acc": 0.62666392,
+      "epoch": 0.010147133434804667,
+      "grad_norm": 6.75,
+      "learning_rate": 1.0147133434804667e-06,
+      "loss": 1.79499435,
+      "memory(GiB)": 36.93,
+      "step": 400,
+      "train_speed(iter/s)": 1.315829
+    },
+    {
+      "acc": 0.62950916,
+      "epoch": 0.010273972602739725,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.0273972602739727e-06,
+      "loss": 1.77388325,
+      "memory(GiB)": 36.93,
+      "step": 405,
+      "train_speed(iter/s)": 1.317435
+    },
+    {
+      "acc": 0.62940755,
+      "epoch": 0.010400811770674784,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.0400811770674785e-06,
+      "loss": 1.80233841,
+      "memory(GiB)": 36.93,
+      "step": 410,
+      "train_speed(iter/s)": 1.319109
+    },
+    {
+      "acc": 0.64196663,
+      "epoch": 0.010527650938609842,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.0527650938609842e-06,
+      "loss": 1.69066124,
+      "memory(GiB)": 36.93,
+      "step": 415,
+      "train_speed(iter/s)": 1.320871
+    },
+    {
+      "acc": 0.64050994,
+      "epoch": 0.0106544901065449,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.0654490106544902e-06,
+      "loss": 1.76365681,
+      "memory(GiB)": 36.95,
+      "step": 420,
+      "train_speed(iter/s)": 1.322521
+    },
+    {
+      "acc": 0.62674227,
+      "epoch": 0.010781329274479959,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.078132927447996e-06,
+      "loss": 1.7416523,
+      "memory(GiB)": 36.95,
+      "step": 425,
+      "train_speed(iter/s)": 1.324143
+    },
+    {
+      "acc": 0.62747736,
+      "epoch": 0.010908168442415017,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.0908168442415017e-06,
+      "loss": 1.7589962,
+      "memory(GiB)": 36.95,
+      "step": 430,
+      "train_speed(iter/s)": 1.325863
+    },
+    {
+      "acc": 0.625597,
+      "epoch": 0.011035007610350075,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.1035007610350077e-06,
+      "loss": 1.77078362,
+      "memory(GiB)": 36.95,
+      "step": 435,
+      "train_speed(iter/s)": 1.327555
+    },
+    {
+      "acc": 0.64102373,
+      "epoch": 0.011161846778285134,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.1161846778285134e-06,
+      "loss": 1.68609219,
+      "memory(GiB)": 36.95,
+      "step": 440,
+      "train_speed(iter/s)": 1.329164
+    },
+    {
+      "acc": 0.63641434,
+      "epoch": 0.011288685946220192,
+      "grad_norm": 4.0,
+      "learning_rate": 1.1288685946220194e-06,
+      "loss": 1.70438938,
+      "memory(GiB)": 36.95,
+      "step": 445,
+      "train_speed(iter/s)": 1.330705
+    },
+    {
+      "acc": 0.62400947,
+      "epoch": 0.01141552511415525,
+      "grad_norm": 4.125,
+      "learning_rate": 1.1415525114155251e-06,
+      "loss": 1.74982567,
+      "memory(GiB)": 36.95,
+      "step": 450,
+      "train_speed(iter/s)": 1.332234
+    },
+    {
+      "acc": 0.63102932,
+      "epoch": 0.011542364282090309,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.154236428209031e-06,
+      "loss": 1.76664982,
+      "memory(GiB)": 36.95,
+      "step": 455,
+      "train_speed(iter/s)": 1.3338
+    },
+    {
+      "acc": 0.63861089,
+      "epoch": 0.011669203450025367,
+      "grad_norm": 3.625,
+      "learning_rate": 1.1669203450025369e-06,
+      "loss": 1.71299839,
+      "memory(GiB)": 36.95,
+      "step": 460,
+      "train_speed(iter/s)": 1.335168
+    },
+    {
+      "acc": 0.63232622,
+      "epoch": 0.011796042617960426,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.1796042617960426e-06,
+      "loss": 1.80091763,
+      "memory(GiB)": 36.95,
+      "step": 465,
+      "train_speed(iter/s)": 1.336696
+    },
+    {
+      "acc": 0.61827192,
+      "epoch": 0.011922881785895484,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.1922881785895486e-06,
+      "loss": 1.85165253,
+      "memory(GiB)": 36.95,
+      "step": 470,
+      "train_speed(iter/s)": 1.338147
+    },
+    {
+      "acc": 0.64919372,
+      "epoch": 0.012049720953830542,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.2049720953830543e-06,
+      "loss": 1.68111877,
+      "memory(GiB)": 36.95,
+      "step": 475,
+      "train_speed(iter/s)": 1.339614
+    },
+    {
+      "acc": 0.62371912,
+      "epoch": 0.0121765601217656,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.21765601217656e-06,
+      "loss": 1.7628418,
+      "memory(GiB)": 36.95,
+      "step": 480,
+      "train_speed(iter/s)": 1.340973
+    },
+    {
+      "acc": 0.63077726,
+      "epoch": 0.012303399289700659,
+      "grad_norm": 3.0,
+      "learning_rate": 1.230339928970066e-06,
+      "loss": 1.73622475,
+      "memory(GiB)": 36.95,
+      "step": 485,
+      "train_speed(iter/s)": 1.342362
+    },
+    {
+      "acc": 0.64916968,
+      "epoch": 0.012430238457635717,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.2430238457635718e-06,
+      "loss": 1.69453964,
+      "memory(GiB)": 36.95,
+      "step": 490,
+      "train_speed(iter/s)": 1.343741
+    },
+    {
+      "acc": 0.63912706,
+      "epoch": 0.012557077625570776,
+      "grad_norm": 4.625,
+      "learning_rate": 1.2557077625570776e-06,
+      "loss": 1.77241173,
+      "memory(GiB)": 36.95,
+      "step": 495,
+      "train_speed(iter/s)": 1.345035
+    },
+    {
+      "acc": 0.63171277,
+      "epoch": 0.012683916793505834,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.2683916793505835e-06,
+      "loss": 1.73084717,
+      "memory(GiB)": 36.95,
+      "step": 500,
+      "train_speed(iter/s)": 1.346353
+    },
+    {
+      "acc": 0.61240368,
+      "epoch": 0.012810755961440892,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.2810755961440893e-06,
+      "loss": 1.75978584,
+      "memory(GiB)": 37.01,
+      "step": 505,
+      "train_speed(iter/s)": 1.347526
+    },
+    {
+      "acc": 0.63663435,
+      "epoch": 0.01293759512937595,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.2937595129375953e-06,
+      "loss": 1.78165112,
+      "memory(GiB)": 37.01,
+      "step": 510,
+      "train_speed(iter/s)": 1.348699
+    },
+    {
+      "acc": 0.62653027,
+      "epoch": 0.013064434297311009,
+      "grad_norm": 4.125,
+      "learning_rate": 1.306443429731101e-06,
+      "loss": 1.79871216,
+      "memory(GiB)": 37.01,
+      "step": 515,
+      "train_speed(iter/s)": 1.349908
+    },
+    {
+      "acc": 0.64694982,
+      "epoch": 0.013191273465246067,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.3191273465246068e-06,
+      "loss": 1.68398361,
+      "memory(GiB)": 37.01,
+      "step": 520,
+      "train_speed(iter/s)": 1.350971
+    },
+    {
+      "acc": 0.63293924,
+      "epoch": 0.013318112633181126,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.3318112633181127e-06,
+      "loss": 1.71520386,
+      "memory(GiB)": 37.01,
+      "step": 525,
+      "train_speed(iter/s)": 1.352172
+    },
+    {
+      "acc": 0.65352502,
+      "epoch": 0.013444951801116184,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.3444951801116185e-06,
+      "loss": 1.67401428,
+      "memory(GiB)": 37.01,
+      "step": 530,
+      "train_speed(iter/s)": 1.353393
+    },
+    {
+      "acc": 0.64030972,
+      "epoch": 0.013571790969051242,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.3571790969051243e-06,
+      "loss": 1.69462986,
+      "memory(GiB)": 37.01,
+      "step": 535,
+      "train_speed(iter/s)": 1.35458
+    },
+    {
+      "acc": 0.6395617,
+      "epoch": 0.0136986301369863,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.3698630136986302e-06,
+      "loss": 1.72002048,
+      "memory(GiB)": 37.01,
+      "step": 540,
+      "train_speed(iter/s)": 1.355768
+    },
+    {
+      "acc": 0.62865381,
+      "epoch": 0.013825469304921359,
+      "grad_norm": 4.125,
+      "learning_rate": 1.382546930492136e-06,
+      "loss": 1.79340057,
+      "memory(GiB)": 37.01,
+      "step": 545,
+      "train_speed(iter/s)": 1.356956
+    },
+    {
+      "acc": 0.63034687,
+      "epoch": 0.013952308472856417,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.395230847285642e-06,
+      "loss": 1.80240707,
+      "memory(GiB)": 37.01,
+      "step": 550,
+      "train_speed(iter/s)": 1.357907
+    },
+    {
+      "acc": 0.63328171,
+      "epoch": 0.014079147640791476,
+      "grad_norm": 4.5,
+      "learning_rate": 1.4079147640791477e-06,
+      "loss": 1.70638123,
+      "memory(GiB)": 37.01,
+      "step": 555,
+      "train_speed(iter/s)": 1.359063
+    },
+    {
+      "acc": 0.64076757,
+      "epoch": 0.014205986808726534,
+      "grad_norm": 4.375,
+      "learning_rate": 1.4205986808726534e-06,
+      "loss": 1.67918434,
+      "memory(GiB)": 37.01,
+      "step": 560,
+      "train_speed(iter/s)": 1.360159
+    },
+    {
+      "acc": 0.63672018,
+      "epoch": 0.014332825976661592,
+      "grad_norm": 5.625,
+      "learning_rate": 1.4332825976661594e-06,
+      "loss": 1.74802628,
+      "memory(GiB)": 37.01,
+      "step": 565,
+      "train_speed(iter/s)": 1.361248
+    },
+    {
+      "acc": 0.64573216,
+      "epoch": 0.01445966514459665,
+      "grad_norm": 5.28125,
+      "learning_rate": 1.4459665144596652e-06,
+      "loss": 1.68051853,
+      "memory(GiB)": 37.01,
+      "step": 570,
+      "train_speed(iter/s)": 1.362297
+    },
+    {
+      "acc": 0.65378971,
+      "epoch": 0.014586504312531709,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.458650431253171e-06,
+      "loss": 1.61417923,
+      "memory(GiB)": 45.01,
+      "step": 575,
+      "train_speed(iter/s)": 1.362929
+    },
+    {
+      "acc": 0.63332748,
+      "epoch": 0.014713343480466767,
+      "grad_norm": 3.875,
+      "learning_rate": 1.4713343480466769e-06,
+      "loss": 1.8174839,
+      "memory(GiB)": 45.01,
+      "step": 580,
+      "train_speed(iter/s)": 1.363936
+    },
+    {
+      "acc": 0.63740749,
+      "epoch": 0.014840182648401826,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.4840182648401826e-06,
+      "loss": 1.79940567,
+      "memory(GiB)": 45.01,
+      "step": 585,
+      "train_speed(iter/s)": 1.364977
+    },
+    {
+      "acc": 0.61923985,
+      "epoch": 0.014967021816336884,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.4967021816336886e-06,
+      "loss": 1.71579437,
+      "memory(GiB)": 45.01,
+      "step": 590,
+      "train_speed(iter/s)": 1.365939
+    },
+    {
+      "acc": 0.62790217,
+      "epoch": 0.015093860984271942,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.5093860984271944e-06,
+      "loss": 1.75112877,
+      "memory(GiB)": 45.01,
+      "step": 595,
+      "train_speed(iter/s)": 1.366962
+    },
+    {
+      "acc": 0.64629989,
+      "epoch": 0.015220700152207,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.5220700152207001e-06,
+      "loss": 1.64116745,
+      "memory(GiB)": 45.01,
+      "step": 600,
+      "train_speed(iter/s)": 1.367975
+    },
+    {
+      "acc": 0.6507812,
+      "epoch": 0.015347539320142059,
+      "grad_norm": 3.625,
+      "learning_rate": 1.534753932014206e-06,
+      "loss": 1.64193115,
+      "memory(GiB)": 45.01,
+      "step": 605,
+      "train_speed(iter/s)": 1.368937
+    },
+    {
+      "acc": 0.63652549,
+      "epoch": 0.015474378488077117,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.5474378488077118e-06,
+      "loss": 1.73050804,
+      "memory(GiB)": 45.01,
+      "step": 610,
+      "train_speed(iter/s)": 1.369879
+    },
+    {
+      "acc": 0.63809624,
+      "epoch": 0.015601217656012176,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.5601217656012176e-06,
+      "loss": 1.69491158,
+      "memory(GiB)": 45.01,
+      "step": 615,
+      "train_speed(iter/s)": 1.370803
+    },
+    {
+      "acc": 0.64809666,
+      "epoch": 0.015728056823947234,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.5728056823947236e-06,
+      "loss": 1.66685333,
+      "memory(GiB)": 45.01,
+      "step": 620,
+      "train_speed(iter/s)": 1.371729
+    },
+    {
+      "acc": 0.62958708,
+      "epoch": 0.015854895991882292,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.5854895991882293e-06,
+      "loss": 1.75772095,
+      "memory(GiB)": 45.01,
+      "step": 625,
+      "train_speed(iter/s)": 1.372603
+    },
+    {
+      "acc": 0.6332787,
+      "epoch": 0.01598173515981735,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.5981735159817353e-06,
+      "loss": 1.75663109,
+      "memory(GiB)": 45.01,
+      "step": 630,
+      "train_speed(iter/s)": 1.373503
+    },
+    {
+      "acc": 0.64622841,
+      "epoch": 0.01610857432775241,
+      "grad_norm": 6.75,
+      "learning_rate": 1.610857432775241e-06,
+      "loss": 1.73431549,
+      "memory(GiB)": 45.01,
+      "step": 635,
+      "train_speed(iter/s)": 1.374422
+    },
+    {
+      "acc": 0.62687569,
+      "epoch": 0.016235413495687467,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.6235413495687468e-06,
+      "loss": 1.72301006,
+      "memory(GiB)": 45.01,
+      "step": 640,
+      "train_speed(iter/s)": 1.375277
+    },
+    {
+      "acc": 0.64029913,
+      "epoch": 0.016362252663622526,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.6362252663622528e-06,
+      "loss": 1.71870728,
+      "memory(GiB)": 45.01,
+      "step": 645,
+      "train_speed(iter/s)": 1.376024
+    },
+    {
+      "acc": 0.62625251,
+      "epoch": 0.016489091831557584,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.6489091831557585e-06,
+      "loss": 1.72709732,
+      "memory(GiB)": 45.01,
+      "step": 650,
+      "train_speed(iter/s)": 1.37688
+    },
+    {
+      "acc": 0.63835745,
+      "epoch": 0.016615930999492642,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.6615930999492643e-06,
+      "loss": 1.7311779,
+      "memory(GiB)": 45.01,
+      "step": 655,
+      "train_speed(iter/s)": 1.377721
+    },
+    {
+      "acc": 0.62800922,
+      "epoch": 0.0167427701674277,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.6742770167427702e-06,
+      "loss": 1.7049078,
+      "memory(GiB)": 45.01,
+      "step": 660,
+      "train_speed(iter/s)": 1.37857
+    },
+    {
+      "acc": 0.63748555,
+      "epoch": 0.01686960933536276,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.686960933536276e-06,
+      "loss": 1.63019791,
+      "memory(GiB)": 45.01,
+      "step": 665,
+      "train_speed(iter/s)": 1.379345
+    },
+    {
+      "acc": 0.64247742,
+      "epoch": 0.016996448503297817,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.699644850329782e-06,
+      "loss": 1.70050468,
+      "memory(GiB)": 45.01,
+      "step": 670,
+      "train_speed(iter/s)": 1.380092
+    },
+    {
+      "acc": 0.64492102,
+      "epoch": 0.017123287671232876,
+      "grad_norm": 4.0,
+      "learning_rate": 1.7123287671232877e-06,
+      "loss": 1.67414513,
+      "memory(GiB)": 45.01,
+      "step": 675,
+      "train_speed(iter/s)": 1.380864
+    },
+    {
+      "acc": 0.62417326,
+      "epoch": 0.017250126839167934,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.7250126839167935e-06,
+      "loss": 1.81239758,
+      "memory(GiB)": 45.01,
+      "step": 680,
+      "train_speed(iter/s)": 1.381629
+    },
+    {
+      "acc": 0.63681097,
+      "epoch": 0.017376966007102992,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.7376966007102994e-06,
+      "loss": 1.72909145,
+      "memory(GiB)": 45.01,
+      "step": 685,
+      "train_speed(iter/s)": 1.382382
+    },
+    {
+      "acc": 0.64099436,
+      "epoch": 0.01750380517503805,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.7503805175038052e-06,
+      "loss": 1.72369499,
+      "memory(GiB)": 45.01,
+      "step": 690,
+      "train_speed(iter/s)": 1.383134
+    },
+    {
+      "acc": 0.63194942,
+      "epoch": 0.01763064434297311,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.7630644342973112e-06,
+      "loss": 1.70109596,
+      "memory(GiB)": 45.01,
+      "step": 695,
+      "train_speed(iter/s)": 1.383875
+    },
+    {
+      "acc": 0.63319511,
+      "epoch": 0.017757483510908167,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.775748351090817e-06,
+      "loss": 1.67285252,
+      "memory(GiB)": 45.01,
+      "step": 700,
+      "train_speed(iter/s)": 1.384545
+    },
+    {
+      "acc": 0.628338,
+      "epoch": 0.017884322678843226,
+      "grad_norm": 3.5,
+      "learning_rate": 1.7884322678843227e-06,
+      "loss": 1.68817577,
+      "memory(GiB)": 45.01,
+      "step": 705,
+      "train_speed(iter/s)": 1.385235
+    },
+    {
+      "acc": 0.62527866,
+      "epoch": 0.018011161846778284,
+      "grad_norm": 4.0,
+      "learning_rate": 1.8011161846778286e-06,
+      "loss": 1.72637711,
+      "memory(GiB)": 45.01,
+      "step": 710,
+      "train_speed(iter/s)": 1.385961
+    },
+    {
+      "acc": 0.61779947,
+      "epoch": 0.018138001014713342,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.8138001014713344e-06,
+      "loss": 1.77308121,
+      "memory(GiB)": 45.01,
+      "step": 715,
+      "train_speed(iter/s)": 1.386647
+    },
+    {
+      "acc": 0.64129033,
+      "epoch": 0.0182648401826484,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.8264840182648401e-06,
+      "loss": 1.67753258,
+      "memory(GiB)": 45.01,
+      "step": 720,
+      "train_speed(iter/s)": 1.387335
+    },
+    {
+      "acc": 0.62164221,
+      "epoch": 0.01839167935058346,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.8391679350583461e-06,
+      "loss": 1.78700333,
+      "memory(GiB)": 45.01,
+      "step": 725,
+      "train_speed(iter/s)": 1.388054
+    },
+    {
+      "acc": 0.64857349,
+      "epoch": 0.018518518518518517,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.8518518518518519e-06,
+      "loss": 1.62372417,
+      "memory(GiB)": 45.01,
+      "step": 730,
+      "train_speed(iter/s)": 1.388727
+    },
+    {
+      "acc": 0.62063317,
+      "epoch": 0.018645357686453576,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.8645357686453578e-06,
+      "loss": 1.7536665,
+      "memory(GiB)": 45.01,
+      "step": 735,
+      "train_speed(iter/s)": 1.389373
+    },
+    {
+      "acc": 0.63571129,
+      "epoch": 0.018772196854388634,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.8772196854388636e-06,
+      "loss": 1.71347599,
+      "memory(GiB)": 45.01,
+      "step": 740,
+      "train_speed(iter/s)": 1.390002
+    },
+    {
+      "acc": 0.63882947,
+      "epoch": 0.018899036022323693,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.8899036022323693e-06,
+      "loss": 1.6454586,
+      "memory(GiB)": 45.01,
+      "step": 745,
+      "train_speed(iter/s)": 1.390674
+    },
+    {
+      "acc": 0.63996449,
+      "epoch": 0.01902587519025875,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.9025875190258753e-06,
+      "loss": 1.69129181,
+      "memory(GiB)": 45.01,
+      "step": 750,
+      "train_speed(iter/s)": 1.391255
+    },
+    {
+      "acc": 0.64385433,
+      "epoch": 0.01915271435819381,
+      "grad_norm": 3.5,
+      "learning_rate": 1.915271435819381e-06,
+      "loss": 1.69550323,
+      "memory(GiB)": 45.01,
+      "step": 755,
+      "train_speed(iter/s)": 1.391934
+    },
+    {
+      "acc": 0.62322774,
+      "epoch": 0.019279553526128868,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.927955352612887e-06,
+      "loss": 1.712603,
+      "memory(GiB)": 45.01,
+      "step": 760,
+      "train_speed(iter/s)": 1.39254
+    },
+    {
+      "acc": 0.64458151,
+      "epoch": 0.019406392694063926,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.9406392694063926e-06,
+      "loss": 1.71406231,
+      "memory(GiB)": 45.01,
+      "step": 765,
+      "train_speed(iter/s)": 1.393171
+    },
+    {
+      "acc": 0.63997755,
+      "epoch": 0.019533231861998984,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.9533231861998985e-06,
+      "loss": 1.69813423,
+      "memory(GiB)": 45.01,
+      "step": 770,
+      "train_speed(iter/s)": 1.393867
+    },
+    {
+      "acc": 0.6471992,
+      "epoch": 0.019660071029934043,
+      "grad_norm": 3.375,
+      "learning_rate": 1.9660071029934045e-06,
+      "loss": 1.71063156,
+      "memory(GiB)": 45.01,
+      "step": 775,
+      "train_speed(iter/s)": 1.394467
+    },
+    {
+      "acc": 0.60916667,
+      "epoch": 0.0197869101978691,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.97869101978691e-06,
+      "loss": 1.77427654,
+      "memory(GiB)": 45.01,
+      "step": 780,
+      "train_speed(iter/s)": 1.395035
+    },
+    {
+      "acc": 0.64961767,
+      "epoch": 0.01991374936580416,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.991374936580416e-06,
+      "loss": 1.59676456,
+      "memory(GiB)": 45.01,
+      "step": 785,
+      "train_speed(iter/s)": 1.395628
+    },
+    {
+      "acc": 0.64263582,
+      "epoch": 0.020040588533739218,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.004058853373922e-06,
+      "loss": 1.69100952,
+      "memory(GiB)": 45.01,
+      "step": 790,
+      "train_speed(iter/s)": 1.396191
+    },
+    {
+      "acc": 0.64090466,
+      "epoch": 0.020167427701674276,
+      "grad_norm": 4.625,
+      "learning_rate": 2.016742770167428e-06,
+      "loss": 1.63766785,
+      "memory(GiB)": 45.01,
+      "step": 795,
+      "train_speed(iter/s)": 1.396737
+    },
+    {
+      "acc": 0.63896894,
+      "epoch": 0.020294266869609334,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.0294266869609335e-06,
+      "loss": 1.63476715,
+      "memory(GiB)": 45.01,
+      "step": 800,
+      "train_speed(iter/s)": 1.397316
+    },
+    {
+      "acc": 0.63009114,
+      "epoch": 0.020421106037544393,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.0421106037544395e-06,
+      "loss": 1.72911873,
+      "memory(GiB)": 45.01,
+      "step": 805,
+      "train_speed(iter/s)": 1.39789
+    },
+    {
+      "acc": 0.65348992,
+      "epoch": 0.02054794520547945,
+      "grad_norm": 4.125,
+      "learning_rate": 2.0547945205479454e-06,
+      "loss": 1.63602047,
+      "memory(GiB)": 45.01,
+      "step": 810,
+      "train_speed(iter/s)": 1.398453
+    },
+    {
+      "acc": 0.65063987,
+      "epoch": 0.02067478437341451,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.067478437341451e-06,
+      "loss": 1.71620979,
+      "memory(GiB)": 45.01,
+      "step": 815,
+      "train_speed(iter/s)": 1.399027
+    },
+    {
+      "acc": 0.64004316,
+      "epoch": 0.020801623541349568,
+      "grad_norm": 4.0,
+      "learning_rate": 2.080162354134957e-06,
+      "loss": 1.69944115,
+      "memory(GiB)": 45.01,
+      "step": 820,
+      "train_speed(iter/s)": 1.399491
+    },
+    {
+      "acc": 0.62459793,
+      "epoch": 0.020928462709284626,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.092846270928463e-06,
+      "loss": 1.72067833,
+      "memory(GiB)": 45.01,
+      "step": 825,
+      "train_speed(iter/s)": 1.40003
+    },
+    {
+      "acc": 0.6210865,
+      "epoch": 0.021055301877219684,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.1055301877219685e-06,
+      "loss": 1.78516617,
+      "memory(GiB)": 45.01,
+      "step": 830,
+      "train_speed(iter/s)": 1.400611
+    },
+    {
+      "acc": 0.64477558,
+      "epoch": 0.021182141045154743,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.1182141045154744e-06,
+      "loss": 1.65948753,
+      "memory(GiB)": 45.01,
+      "step": 835,
+      "train_speed(iter/s)": 1.401129
+    },
+    {
+      "acc": 0.63925657,
+      "epoch": 0.0213089802130898,
+      "grad_norm": 4.125,
+      "learning_rate": 2.1308980213089804e-06,
+      "loss": 1.64069977,
+      "memory(GiB)": 53.47,
+      "step": 840,
+      "train_speed(iter/s)": 1.401436
+    },
+    {
+      "acc": 0.63804779,
+      "epoch": 0.02143581938102486,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.143581938102486e-06,
+      "loss": 1.64016342,
+      "memory(GiB)": 53.47,
+      "step": 845,
+      "train_speed(iter/s)": 1.401984
+    },
+    {
+      "acc": 0.65818758,
+      "epoch": 0.021562658548959918,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.156265854895992e-06,
+      "loss": 1.66067429,
+      "memory(GiB)": 53.47,
+      "step": 850,
+      "train_speed(iter/s)": 1.402516
+    },
+    {
+      "acc": 0.62126503,
+      "epoch": 0.021689497716894976,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.168949771689498e-06,
+      "loss": 1.72171803,
+      "memory(GiB)": 53.47,
+      "step": 855,
+      "train_speed(iter/s)": 1.403014
+    },
+    {
+      "acc": 0.65038614,
+      "epoch": 0.021816336884830034,
+      "grad_norm": 3.875,
+      "learning_rate": 2.1816336884830034e-06,
+      "loss": 1.70088463,
+      "memory(GiB)": 53.47,
+      "step": 860,
+      "train_speed(iter/s)": 1.403518
+    },
+    {
+      "acc": 0.61944046,
+      "epoch": 0.021943176052765093,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.1943176052765094e-06,
+      "loss": 1.67334938,
+      "memory(GiB)": 53.47,
+      "step": 865,
+      "train_speed(iter/s)": 1.404039
+    },
+    {
+      "acc": 0.63039103,
+      "epoch": 0.02207001522070015,
+      "grad_norm": 4.125,
+      "learning_rate": 2.2070015220700153e-06,
+      "loss": 1.76165428,
+      "memory(GiB)": 53.47,
+      "step": 870,
+      "train_speed(iter/s)": 1.404547
+    },
+    {
+      "acc": 0.63281131,
+      "epoch": 0.02219685438863521,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.2196854388635213e-06,
+      "loss": 1.70353374,
+      "memory(GiB)": 53.47,
+      "step": 875,
+      "train_speed(iter/s)": 1.404961
+    },
+    {
+      "acc": 0.64554691,
+      "epoch": 0.022323693556570268,
+      "grad_norm": 4.375,
+      "learning_rate": 2.232369355657027e-06,
+      "loss": 1.65355568,
+      "memory(GiB)": 53.47,
+      "step": 880,
+      "train_speed(iter/s)": 1.405367
+    },
+    {
+      "acc": 0.63497276,
+      "epoch": 0.022450532724505326,
+      "grad_norm": 4.375,
+      "learning_rate": 2.245053272450533e-06,
+      "loss": 1.75669498,
+      "memory(GiB)": 53.47,
+      "step": 885,
+      "train_speed(iter/s)": 1.405719
+    },
+    {
+      "acc": 0.6362464,
+      "epoch": 0.022577371892440384,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.2577371892440388e-06,
+      "loss": 1.70398941,
+      "memory(GiB)": 53.47,
+      "step": 890,
+      "train_speed(iter/s)": 1.406144
+    },
+    {
+      "acc": 0.65096693,
+      "epoch": 0.022704211060375443,
+      "grad_norm": 2.875,
+      "learning_rate": 2.2704211060375443e-06,
+      "loss": 1.69155407,
+      "memory(GiB)": 53.47,
+      "step": 895,
+      "train_speed(iter/s)": 1.40656
+    },
+    {
+      "acc": 0.64690557,
+      "epoch": 0.0228310502283105,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.2831050228310503e-06,
+      "loss": 1.69622116,
+      "memory(GiB)": 53.47,
+      "step": 900,
+      "train_speed(iter/s)": 1.406968
+    },
+    {
+      "acc": 0.63019361,
+      "epoch": 0.02295788939624556,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.2957889396245563e-06,
+      "loss": 1.70496197,
+      "memory(GiB)": 53.47,
+      "step": 905,
+      "train_speed(iter/s)": 1.407338
+    },
+    {
+      "acc": 0.62627239,
+      "epoch": 0.023084728564180618,
+      "grad_norm": 3.375,
+      "learning_rate": 2.308472856418062e-06,
+      "loss": 1.69036865,
+      "memory(GiB)": 53.47,
+      "step": 910,
+      "train_speed(iter/s)": 1.407737
+    },
+    {
+      "acc": 0.64377556,
+      "epoch": 0.023211567732115676,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.3211567732115678e-06,
+      "loss": 1.66639156,
+      "memory(GiB)": 53.47,
+      "step": 915,
+      "train_speed(iter/s)": 1.408085
+    },
+    {
+      "acc": 0.66818056,
+      "epoch": 0.023338406900050734,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.3338406900050737e-06,
+      "loss": 1.61898594,
+      "memory(GiB)": 53.47,
+      "step": 920,
+      "train_speed(iter/s)": 1.408458
+    },
+    {
+      "acc": 0.6329905,
+      "epoch": 0.023465246067985793,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.3465246067985793e-06,
+      "loss": 1.71271706,
+      "memory(GiB)": 53.47,
+      "step": 925,
+      "train_speed(iter/s)": 1.408842
+    },
+    {
+      "acc": 0.6438798,
+      "epoch": 0.02359208523592085,
+      "grad_norm": 5.84375,
+      "learning_rate": 2.3592085235920852e-06,
+      "loss": 1.67928391,
+      "memory(GiB)": 53.47,
+      "step": 930,
+      "train_speed(iter/s)": 1.409272
+    },
+    {
+      "acc": 0.64368467,
+      "epoch": 0.02371892440385591,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.371892440385591e-06,
+      "loss": 1.60682392,
+      "memory(GiB)": 53.47,
+      "step": 935,
+      "train_speed(iter/s)": 1.409634
+    },
+    {
+      "acc": 0.64377885,
+      "epoch": 0.023845763571790968,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.384576357179097e-06,
+      "loss": 1.67088547,
+      "memory(GiB)": 53.47,
+      "step": 940,
+      "train_speed(iter/s)": 1.410031
+    },
+    {
+      "acc": 0.64703579,
+      "epoch": 0.023972602739726026,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.3972602739726027e-06,
+      "loss": 1.67412148,
+      "memory(GiB)": 53.47,
+      "step": 945,
+      "train_speed(iter/s)": 1.410402
+    },
+    {
+      "acc": 0.63242822,
+      "epoch": 0.024099441907661084,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.4099441907661087e-06,
+      "loss": 1.69469814,
+      "memory(GiB)": 53.47,
+      "step": 950,
+      "train_speed(iter/s)": 1.410756
+    },
+    {
+      "acc": 0.64623909,
+      "epoch": 0.024226281075596143,
+      "grad_norm": 8.0,
+      "learning_rate": 2.4226281075596147e-06,
+      "loss": 1.66908131,
+      "memory(GiB)": 53.47,
+      "step": 955,
+      "train_speed(iter/s)": 1.411101
+    },
+    {
+      "acc": 0.64056196,
+      "epoch": 0.0243531202435312,
+      "grad_norm": 6.375,
+      "learning_rate": 2.43531202435312e-06,
+      "loss": 1.72677078,
+      "memory(GiB)": 53.47,
+      "step": 960,
+      "train_speed(iter/s)": 1.411456
+    },
+    {
+      "acc": 0.63446217,
+      "epoch": 0.02447995941146626,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.447995941146626e-06,
+      "loss": 1.73170357,
+      "memory(GiB)": 53.47,
+      "step": 965,
+      "train_speed(iter/s)": 1.411768
+    },
+    {
+      "acc": 0.63047285,
+      "epoch": 0.024606798579401318,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.460679857940132e-06,
+      "loss": 1.72913952,
+      "memory(GiB)": 53.47,
+      "step": 970,
+      "train_speed(iter/s)": 1.412092
+    },
+    {
+      "acc": 0.6292933,
+      "epoch": 0.024733637747336376,
+      "grad_norm": 3.5,
+      "learning_rate": 2.4733637747336377e-06,
+      "loss": 1.72143803,
+      "memory(GiB)": 53.47,
+      "step": 975,
+      "train_speed(iter/s)": 1.412476
+    },
+    {
+      "acc": 0.66250944,
+      "epoch": 0.024860476915271434,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.4860476915271436e-06,
+      "loss": 1.62486076,
+      "memory(GiB)": 53.47,
+      "step": 980,
+      "train_speed(iter/s)": 1.412822
+    },
+    {
+      "acc": 0.63465538,
+      "epoch": 0.024987316083206493,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.4987316083206496e-06,
+      "loss": 1.71249695,
+      "memory(GiB)": 53.47,
+      "step": 985,
+      "train_speed(iter/s)": 1.413191
+    },
+    {
+      "acc": 0.63313537,
+      "epoch": 0.02511415525114155,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.511415525114155e-06,
+      "loss": 1.76477814,
+      "memory(GiB)": 53.47,
+      "step": 990,
+      "train_speed(iter/s)": 1.413563
+    },
+    {
+      "acc": 0.6291739,
+      "epoch": 0.02524099441907661,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.5240994419076615e-06,
+      "loss": 1.67669907,
+      "memory(GiB)": 53.47,
+      "step": 995,
+      "train_speed(iter/s)": 1.413925
+    },
+    {
+      "acc": 0.65271072,
+      "epoch": 0.025367833587011668,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.536783358701167e-06,
+      "loss": 1.6057827,
+      "memory(GiB)": 53.47,
+      "step": 1000,
+      "train_speed(iter/s)": 1.41428
+    },
+    {
+      "epoch": 0.025367833587011668,
+      "eval_acc": 0.6283174308822884,
+      "eval_loss": 1.6503840684890747,
+      "eval_runtime": 69.4354,
+      "eval_samples_per_second": 91.74,
+      "eval_steps_per_second": 22.942,
+      "step": 1000
+    },
+    {
+      "acc": 0.63742304,
+      "epoch": 0.025494672754946726,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.549467275494673e-06,
+      "loss": 1.70446739,
+      "memory(GiB)": 59.39,
+      "step": 1005,
+      "train_speed(iter/s)": 1.269725
+    },
+    {
+      "acc": 0.63823767,
+      "epoch": 0.025621511922881784,
+      "grad_norm": 4.625,
+      "learning_rate": 2.5621511922881786e-06,
+      "loss": 1.7666338,
+      "memory(GiB)": 59.39,
+      "step": 1010,
+      "train_speed(iter/s)": 1.270624
+    },
+    {
+      "acc": 0.63278737,
+      "epoch": 0.025748351090816843,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.5748351090816846e-06,
+      "loss": 1.71410866,
+      "memory(GiB)": 59.39,
+      "step": 1015,
+      "train_speed(iter/s)": 1.271474
+    },
+    {
+      "acc": 0.64791527,
+      "epoch": 0.0258751902587519,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.5875190258751905e-06,
+      "loss": 1.6745903,
+      "memory(GiB)": 59.39,
+      "step": 1020,
+      "train_speed(iter/s)": 1.272369
+    },
+    {
+      "acc": 0.65812359,
+      "epoch": 0.02600202942668696,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.6002029426686965e-06,
+      "loss": 1.67661438,
+      "memory(GiB)": 59.39,
+      "step": 1025,
+      "train_speed(iter/s)": 1.273213
+    },
+    {
+      "acc": 0.64736786,
+      "epoch": 0.026128868594622018,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.612886859462202e-06,
+      "loss": 1.66765709,
+      "memory(GiB)": 59.39,
+      "step": 1030,
+      "train_speed(iter/s)": 1.274067
+    },
+    {
+      "acc": 0.65381947,
+      "epoch": 0.026255707762557076,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.625570776255708e-06,
+      "loss": 1.66184578,
+      "memory(GiB)": 59.39,
+      "step": 1035,
+      "train_speed(iter/s)": 1.27487
+    },
+    {
+      "acc": 0.64077425,
+      "epoch": 0.026382546930492135,
+      "grad_norm": 5.0,
+      "learning_rate": 2.6382546930492135e-06,
+      "loss": 1.66555138,
+      "memory(GiB)": 59.39,
+      "step": 1040,
+      "train_speed(iter/s)": 1.275703
+    },
+    {
+      "acc": 0.6440217,
+      "epoch": 0.026509386098427193,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.65093860984272e-06,
+      "loss": 1.7258316,
+      "memory(GiB)": 59.39,
+      "step": 1045,
+      "train_speed(iter/s)": 1.276512
+    },
+    {
+      "acc": 0.65024109,
+      "epoch": 0.02663622526636225,
+      "grad_norm": 3.625,
+      "learning_rate": 2.6636225266362255e-06,
+      "loss": 1.63880119,
+      "memory(GiB)": 59.39,
+      "step": 1050,
+      "train_speed(iter/s)": 1.277328
+    },
+    {
+      "acc": 0.64218669,
+      "epoch": 0.02676306443429731,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.6763064434297314e-06,
+      "loss": 1.63450451,
+      "memory(GiB)": 59.39,
+      "step": 1055,
+      "train_speed(iter/s)": 1.278154
+    },
+    {
+      "acc": 0.63855743,
+      "epoch": 0.026889903602232368,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.688990360223237e-06,
+      "loss": 1.6540905,
+      "memory(GiB)": 59.39,
+      "step": 1060,
+      "train_speed(iter/s)": 1.278953
+    },
+    {
+      "acc": 0.64281015,
+      "epoch": 0.027016742770167426,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.701674277016743e-06,
+      "loss": 1.65806465,
+      "memory(GiB)": 59.39,
+      "step": 1065,
+      "train_speed(iter/s)": 1.279779
+    },
+    {
+      "acc": 0.64511309,
+      "epoch": 0.027143581938102485,
+      "grad_norm": 3.625,
+      "learning_rate": 2.7143581938102485e-06,
+      "loss": 1.70993042,
+      "memory(GiB)": 59.39,
+      "step": 1070,
+      "train_speed(iter/s)": 1.280534
+    },
+    {
+      "acc": 0.63697157,
+      "epoch": 0.027270421106037543,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.727042110603755e-06,
+      "loss": 1.67856884,
+      "memory(GiB)": 59.39,
+      "step": 1075,
+      "train_speed(iter/s)": 1.281272
+    },
+    {
+      "acc": 0.62660632,
+      "epoch": 0.0273972602739726,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.7397260273972604e-06,
+      "loss": 1.76990585,
+      "memory(GiB)": 59.39,
+      "step": 1080,
+      "train_speed(iter/s)": 1.282048
+    },
+    {
+      "acc": 0.65651855,
+      "epoch": 0.02752409944190766,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.7524099441907664e-06,
+      "loss": 1.63692837,
+      "memory(GiB)": 59.39,
+      "step": 1085,
+      "train_speed(iter/s)": 1.282821
+    },
+    {
+      "acc": 0.65734825,
+      "epoch": 0.027650938609842718,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.765093860984272e-06,
+      "loss": 1.63038979,
+      "memory(GiB)": 59.39,
+      "step": 1090,
+      "train_speed(iter/s)": 1.283588
+    },
+    {
+      "acc": 0.63047404,
+      "epoch": 0.027777777777777776,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.7777777777777783e-06,
+      "loss": 1.67619934,
+      "memory(GiB)": 59.39,
+      "step": 1095,
+      "train_speed(iter/s)": 1.284379
+    },
+    {
+      "acc": 0.65068808,
+      "epoch": 0.027904616945712835,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.790461694571284e-06,
+      "loss": 1.61234093,
+      "memory(GiB)": 59.39,
+      "step": 1100,
+      "train_speed(iter/s)": 1.285144
+    },
+    {
+      "acc": 0.64407463,
+      "epoch": 0.028031456113647893,
+      "grad_norm": 3.75,
+      "learning_rate": 2.80314561136479e-06,
+      "loss": 1.67489204,
+      "memory(GiB)": 59.39,
+      "step": 1105,
+      "train_speed(iter/s)": 1.285867
+    },
+    {
+      "acc": 0.64778299,
+      "epoch": 0.02815829528158295,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.8158295281582954e-06,
+      "loss": 1.63532848,
+      "memory(GiB)": 59.39,
+      "step": 1110,
+      "train_speed(iter/s)": 1.28661
+    },
+    {
+      "acc": 0.63811235,
+      "epoch": 0.02828513444951801,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.8285134449518014e-06,
+      "loss": 1.72589359,
+      "memory(GiB)": 59.39,
+      "step": 1115,
+      "train_speed(iter/s)": 1.287338
+    },
+    {
+      "acc": 0.635993,
+      "epoch": 0.028411973617453068,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.841197361745307e-06,
+      "loss": 1.70236244,
+      "memory(GiB)": 59.39,
+      "step": 1120,
+      "train_speed(iter/s)": 1.28805
+    },
+    {
+      "acc": 0.62237973,
+      "epoch": 0.028538812785388126,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.8538812785388133e-06,
+      "loss": 1.66360703,
+      "memory(GiB)": 59.39,
+      "step": 1125,
+      "train_speed(iter/s)": 1.288769
+    },
+    {
+      "acc": 0.64396572,
+      "epoch": 0.028665651953323185,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.866565195332319e-06,
+      "loss": 1.61986885,
+      "memory(GiB)": 59.39,
+      "step": 1130,
+      "train_speed(iter/s)": 1.289452
+    },
+    {
+      "acc": 0.65134344,
+      "epoch": 0.028792491121258243,
+      "grad_norm": 3.375,
+      "learning_rate": 2.879249112125825e-06,
+      "loss": 1.62761765,
+      "memory(GiB)": 59.39,
+      "step": 1135,
+      "train_speed(iter/s)": 1.290138
+    },
+    {
+      "acc": 0.63852029,
+      "epoch": 0.0289193302891933,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.8919330289193303e-06,
+      "loss": 1.73116646,
+      "memory(GiB)": 59.39,
+      "step": 1140,
+      "train_speed(iter/s)": 1.29086
+    },
+    {
+      "acc": 0.63659286,
+      "epoch": 0.02904616945712836,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.9046169457128363e-06,
+      "loss": 1.6636858,
+      "memory(GiB)": 59.39,
+      "step": 1145,
+      "train_speed(iter/s)": 1.291546
+    },
+    {
+      "acc": 0.63266573,
+      "epoch": 0.029173008625063418,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.917300862506342e-06,
+      "loss": 1.68827972,
+      "memory(GiB)": 59.39,
+      "step": 1150,
+      "train_speed(iter/s)": 1.292249
+    },
+    {
+      "acc": 0.64466581,
+      "epoch": 0.029299847792998476,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.9299847792998482e-06,
+      "loss": 1.64395714,
+      "memory(GiB)": 59.39,
+      "step": 1155,
+      "train_speed(iter/s)": 1.29296
+    },
+    {
+      "acc": 0.6505599,
+      "epoch": 0.029426686960933535,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.9426686960933538e-06,
+      "loss": 1.66964951,
+      "memory(GiB)": 59.39,
+      "step": 1160,
+      "train_speed(iter/s)": 1.2934
+    },
+    {
+      "acc": 0.64059134,
+      "epoch": 0.029553526128868593,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.9553526128868598e-06,
+      "loss": 1.63648853,
+      "memory(GiB)": 59.39,
+      "step": 1165,
+      "train_speed(iter/s)": 1.294066
+    },
+    {
+      "acc": 0.63618231,
+      "epoch": 0.02968036529680365,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.9680365296803653e-06,
+      "loss": 1.70097141,
+      "memory(GiB)": 59.39,
+      "step": 1170,
+      "train_speed(iter/s)": 1.294722
+    },
+    {
+      "acc": 0.64414325,
+      "epoch": 0.02980720446473871,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.9807204464738717e-06,
+      "loss": 1.65998268,
+      "memory(GiB)": 59.39,
+      "step": 1175,
+      "train_speed(iter/s)": 1.29538
+    },
+    {
+      "acc": 0.65539246,
+      "epoch": 0.029934043632673768,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.9934043632673772e-06,
+      "loss": 1.58997688,
+      "memory(GiB)": 59.39,
+      "step": 1180,
+      "train_speed(iter/s)": 1.296082
+    },
+    {
+      "acc": 0.63417101,
+      "epoch": 0.030060882800608826,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.006088280060883e-06,
+      "loss": 1.65651283,
+      "memory(GiB)": 59.39,
+      "step": 1185,
+      "train_speed(iter/s)": 1.296747
+    },
+    {
+      "acc": 0.63447237,
+      "epoch": 0.030187721968543885,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.0187721968543887e-06,
+      "loss": 1.69890423,
+      "memory(GiB)": 62.29,
+      "step": 1190,
+      "train_speed(iter/s)": 1.297323
+    },
+    {
+      "acc": 0.62934361,
+      "epoch": 0.030314561136478943,
+      "grad_norm": 3.5,
+      "learning_rate": 3.0314561136478947e-06,
+      "loss": 1.70647316,
+      "memory(GiB)": 62.29,
+      "step": 1195,
+      "train_speed(iter/s)": 1.297996
+    },
+    {
+      "acc": 0.67064466,
+      "epoch": 0.030441400304414,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.0441400304414002e-06,
+      "loss": 1.61545811,
+      "memory(GiB)": 62.29,
+      "step": 1200,
+      "train_speed(iter/s)": 1.298666
+    },
+    {
+      "acc": 0.63656082,
+      "epoch": 0.03056823947234906,
+      "grad_norm": 7.15625,
+      "learning_rate": 3.0568239472349066e-06,
+      "loss": 1.75801296,
+      "memory(GiB)": 62.29,
+      "step": 1205,
+      "train_speed(iter/s)": 1.299349
+    },
+    {
+      "acc": 0.62868333,
+      "epoch": 0.030695078640284118,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.069507864028412e-06,
+      "loss": 1.63936501,
+      "memory(GiB)": 62.29,
+      "step": 1210,
+      "train_speed(iter/s)": 1.299987
+    },
+    {
+      "acc": 0.64091196,
+      "epoch": 0.030821917808219176,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.082191780821918e-06,
+      "loss": 1.68087807,
+      "memory(GiB)": 62.29,
+      "step": 1215,
+      "train_speed(iter/s)": 1.300648
+    },
+    {
+      "acc": 0.63661027,
+      "epoch": 0.030948756976154235,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.0948756976154237e-06,
+      "loss": 1.66307144,
+      "memory(GiB)": 62.29,
+      "step": 1220,
+      "train_speed(iter/s)": 1.301211
+    },
+    {
+      "acc": 0.65306468,
+      "epoch": 0.031075596144089293,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.1075596144089297e-06,
+      "loss": 1.66524849,
+      "memory(GiB)": 62.29,
+      "step": 1225,
+      "train_speed(iter/s)": 1.301859
+    },
+    {
+      "acc": 0.65047703,
+      "epoch": 0.03120243531202435,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.120243531202435e-06,
+      "loss": 1.68411713,
+      "memory(GiB)": 62.29,
+      "step": 1230,
+      "train_speed(iter/s)": 1.302505
+    },
+    {
+      "acc": 0.63698506,
+      "epoch": 0.03132927447995941,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.1329274479959416e-06,
+      "loss": 1.73069839,
+      "memory(GiB)": 62.29,
+      "step": 1235,
+      "train_speed(iter/s)": 1.303115
+    },
+    {
+      "acc": 0.64726858,
+      "epoch": 0.03145611364789447,
+      "grad_norm": 3.375,
+      "learning_rate": 3.145611364789447e-06,
+      "loss": 1.67007256,
+      "memory(GiB)": 62.29,
+      "step": 1240,
+      "train_speed(iter/s)": 1.303732
+    },
+    {
+      "acc": 0.64171963,
+      "epoch": 0.031582952815829526,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.158295281582953e-06,
+      "loss": 1.72825241,
+      "memory(GiB)": 62.29,
+      "step": 1245,
+      "train_speed(iter/s)": 1.304338
+    },
+    {
+      "acc": 0.63829417,
+      "epoch": 0.031709791983764585,
+      "grad_norm": 3.875,
+      "learning_rate": 3.1709791983764586e-06,
+      "loss": 1.6622324,
+      "memory(GiB)": 62.29,
+      "step": 1250,
+      "train_speed(iter/s)": 1.304945
+    },
+    {
+      "acc": 0.64901104,
+      "epoch": 0.03183663115169964,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.183663115169965e-06,
+      "loss": 1.66797295,
+      "memory(GiB)": 62.29,
+      "step": 1255,
+      "train_speed(iter/s)": 1.30555
+    },
+    {
+      "acc": 0.6494319,
+      "epoch": 0.0319634703196347,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.1963470319634706e-06,
+      "loss": 1.69942513,
+      "memory(GiB)": 62.29,
+      "step": 1260,
+      "train_speed(iter/s)": 1.30616
+    },
+    {
+      "acc": 0.64505329,
+      "epoch": 0.03209030948756976,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.2090309487569765e-06,
+      "loss": 1.63549442,
+      "memory(GiB)": 62.29,
+      "step": 1265,
+      "train_speed(iter/s)": 1.306763
+    },
+    {
+      "acc": 0.64524407,
+      "epoch": 0.03221714865550482,
+      "grad_norm": 3.5,
+      "learning_rate": 3.221714865550482e-06,
+      "loss": 1.695784,
+      "memory(GiB)": 62.29,
+      "step": 1270,
+      "train_speed(iter/s)": 1.307342
+    },
+    {
+      "acc": 0.6422287,
+      "epoch": 0.032343987823439876,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.234398782343988e-06,
+      "loss": 1.7078434,
+      "memory(GiB)": 62.29,
+      "step": 1275,
+      "train_speed(iter/s)": 1.307884
+    },
+    {
+      "acc": 0.6460825,
+      "epoch": 0.032470826991374935,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.2470826991374936e-06,
+      "loss": 1.65249424,
+      "memory(GiB)": 62.29,
+      "step": 1280,
+      "train_speed(iter/s)": 1.308492
+    },
+    {
+      "acc": 0.64259653,
+      "epoch": 0.03259766615930999,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.259766615931e-06,
+      "loss": 1.66420708,
+      "memory(GiB)": 62.29,
+      "step": 1285,
+      "train_speed(iter/s)": 1.30907
+    },
+    {
+      "acc": 0.64894681,
+      "epoch": 0.03272450532724505,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.2724505327245055e-06,
+      "loss": 1.62121811,
+      "memory(GiB)": 62.29,
+      "step": 1290,
+      "train_speed(iter/s)": 1.309594
+    },
+    {
+      "acc": 0.63853254,
+      "epoch": 0.03285134449518011,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.2851344495180115e-06,
+      "loss": 1.68569965,
+      "memory(GiB)": 62.29,
+      "step": 1295,
+      "train_speed(iter/s)": 1.310173
+    },
+    {
+      "acc": 0.63972149,
+      "epoch": 0.03297818366311517,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.297818366311517e-06,
+      "loss": 1.73914413,
+      "memory(GiB)": 62.29,
+      "step": 1300,
+      "train_speed(iter/s)": 1.310727
+    },
+    {
+      "acc": 0.65501318,
+      "epoch": 0.033105022831050226,
+      "grad_norm": 4.96875,
+      "learning_rate": 3.310502283105023e-06,
+      "loss": 1.61578751,
+      "memory(GiB)": 62.29,
+      "step": 1305,
+      "train_speed(iter/s)": 1.311309
+    },
+    {
+      "acc": 0.64804101,
+      "epoch": 0.033231861998985285,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.3231861998985286e-06,
+      "loss": 1.64674454,
+      "memory(GiB)": 62.29,
+      "step": 1310,
+      "train_speed(iter/s)": 1.311889
+    },
+    {
+      "acc": 0.65997138,
+      "epoch": 0.03335870116692034,
+      "grad_norm": 3.125,
+      "learning_rate": 3.335870116692035e-06,
+      "loss": 1.53894463,
+      "memory(GiB)": 62.29,
+      "step": 1315,
+      "train_speed(iter/s)": 1.312453
+    },
+    {
+      "acc": 0.63640947,
+      "epoch": 0.0334855403348554,
+      "grad_norm": 3.375,
+      "learning_rate": 3.3485540334855405e-06,
+      "loss": 1.6555481,
+      "memory(GiB)": 62.29,
+      "step": 1320,
+      "train_speed(iter/s)": 1.313005
+    },
+    {
+      "acc": 0.64955487,
+      "epoch": 0.03361237950279046,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.3612379502790465e-06,
+      "loss": 1.53392143,
+      "memory(GiB)": 62.29,
+      "step": 1325,
+      "train_speed(iter/s)": 1.313564
+    },
+    {
+      "acc": 0.65349407,
+      "epoch": 0.03373921867072552,
+      "grad_norm": 4.375,
+      "learning_rate": 3.373921867072552e-06,
+      "loss": 1.65668259,
+      "memory(GiB)": 62.29,
+      "step": 1330,
+      "train_speed(iter/s)": 1.314098
+    },
+    {
+      "acc": 0.64496074,
+      "epoch": 0.033866057838660577,
+      "grad_norm": 3.0,
+      "learning_rate": 3.3866057838660584e-06,
+      "loss": 1.71247177,
+      "memory(GiB)": 62.29,
+      "step": 1335,
+      "train_speed(iter/s)": 1.314639
+    },
+    {
+      "acc": 0.65064521,
+      "epoch": 0.033992897006595635,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.399289700659564e-06,
+      "loss": 1.62430382,
+      "memory(GiB)": 62.29,
+      "step": 1340,
+      "train_speed(iter/s)": 1.315167
+    },
+    {
+      "acc": 0.66331811,
+      "epoch": 0.03411973617453069,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.41197361745307e-06,
+      "loss": 1.61776047,
+      "memory(GiB)": 62.29,
+      "step": 1345,
+      "train_speed(iter/s)": 1.3157
+    },
+    {
+      "acc": 0.63273315,
+      "epoch": 0.03424657534246575,
+      "grad_norm": 2.875,
+      "learning_rate": 3.4246575342465754e-06,
+      "loss": 1.63156338,
+      "memory(GiB)": 62.29,
+      "step": 1350,
+      "train_speed(iter/s)": 1.316207
+    },
+    {
+      "acc": 0.66386833,
+      "epoch": 0.03437341451040081,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.4373414510400814e-06,
+      "loss": 1.59157219,
+      "memory(GiB)": 62.29,
+      "step": 1355,
+      "train_speed(iter/s)": 1.31675
+    },
+    {
+      "acc": 0.65150366,
+      "epoch": 0.03450025367833587,
+      "grad_norm": 3.984375,
+      "learning_rate": 3.450025367833587e-06,
+      "loss": 1.64722157,
+      "memory(GiB)": 62.29,
+      "step": 1360,
+      "train_speed(iter/s)": 1.317279
+    },
+    {
+      "acc": 0.64297547,
+      "epoch": 0.03462709284627093,
+      "grad_norm": 3.375,
+      "learning_rate": 3.4627092846270933e-06,
+      "loss": 1.67039509,
+      "memory(GiB)": 62.29,
+      "step": 1365,
+      "train_speed(iter/s)": 1.317807
+    },
+    {
+      "acc": 0.63903985,
+      "epoch": 0.034753932014205985,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.475393201420599e-06,
+      "loss": 1.6654911,
+      "memory(GiB)": 62.29,
+      "step": 1370,
+      "train_speed(iter/s)": 1.31834
+    },
+    {
+      "acc": 0.64927444,
+      "epoch": 0.03488077118214104,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.488077118214105e-06,
+      "loss": 1.61070328,
+      "memory(GiB)": 62.29,
+      "step": 1375,
+      "train_speed(iter/s)": 1.318856
+    },
+    {
+      "acc": 0.64951477,
+      "epoch": 0.0350076103500761,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.5007610350076104e-06,
+      "loss": 1.63521729,
+      "memory(GiB)": 62.29,
+      "step": 1380,
+      "train_speed(iter/s)": 1.319368
+    },
+    {
+      "acc": 0.65427847,
+      "epoch": 0.03513444951801116,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.5134449518011164e-06,
+      "loss": 1.62253036,
+      "memory(GiB)": 62.29,
+      "step": 1385,
+      "train_speed(iter/s)": 1.319878
+    },
+    {
+      "acc": 0.6365026,
+      "epoch": 0.03526128868594622,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.5261288685946223e-06,
+      "loss": 1.6813221,
+      "memory(GiB)": 62.29,
+      "step": 1390,
+      "train_speed(iter/s)": 1.320391
+    },
+    {
+      "acc": 0.63040714,
+      "epoch": 0.03538812785388128,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.5388127853881283e-06,
+      "loss": 1.7316309,
+      "memory(GiB)": 62.29,
+      "step": 1395,
+      "train_speed(iter/s)": 1.320871
+    },
+    {
+      "acc": 0.63493519,
+      "epoch": 0.035514967021816335,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.551496702181634e-06,
+      "loss": 1.62761917,
+      "memory(GiB)": 62.29,
+      "step": 1400,
+      "train_speed(iter/s)": 1.321382
+    },
+    {
+      "acc": 0.64467125,
+      "epoch": 0.03564180618975139,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.56418061897514e-06,
+      "loss": 1.72089157,
+      "memory(GiB)": 62.29,
+      "step": 1405,
+      "train_speed(iter/s)": 1.321878
+    },
+    {
+      "acc": 0.63973837,
+      "epoch": 0.03576864535768645,
+      "grad_norm": 3.5,
+      "learning_rate": 3.5768645357686453e-06,
+      "loss": 1.71157265,
+      "memory(GiB)": 62.29,
+      "step": 1410,
+      "train_speed(iter/s)": 1.322378
+    },
+    {
+      "acc": 0.64853959,
+      "epoch": 0.03589548452562151,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.5895484525621517e-06,
+      "loss": 1.67436295,
+      "memory(GiB)": 62.29,
+      "step": 1415,
+      "train_speed(iter/s)": 1.322864
+    },
+    {
+      "acc": 0.6661211,
+      "epoch": 0.03602232369355657,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.6022323693556573e-06,
+      "loss": 1.60737591,
+      "memory(GiB)": 62.29,
+      "step": 1420,
+      "train_speed(iter/s)": 1.323344
+    },
+    {
+      "acc": 0.64450564,
+      "epoch": 0.03614916286149163,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.6149162861491632e-06,
+      "loss": 1.68028774,
+      "memory(GiB)": 62.29,
+      "step": 1425,
+      "train_speed(iter/s)": 1.323826
+    },
+    {
+      "acc": 0.66723723,
+      "epoch": 0.036276002029426685,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.6276002029426688e-06,
+      "loss": 1.58899145,
+      "memory(GiB)": 62.29,
+      "step": 1430,
+      "train_speed(iter/s)": 1.324278
+    },
+    {
+      "acc": 0.62619119,
+      "epoch": 0.03640284119736174,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.6402841197361748e-06,
+      "loss": 1.68298836,
+      "memory(GiB)": 62.29,
+      "step": 1435,
+      "train_speed(iter/s)": 1.324748
+    },
+    {
+      "acc": 0.6524663,
+      "epoch": 0.0365296803652968,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.6529680365296803e-06,
+      "loss": 1.67757225,
+      "memory(GiB)": 62.29,
+      "step": 1440,
+      "train_speed(iter/s)": 1.325203
+    },
+    {
+      "acc": 0.66538763,
+      "epoch": 0.03665651953323186,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.6656519533231867e-06,
+      "loss": 1.62664604,
+      "memory(GiB)": 62.29,
+      "step": 1445,
+      "train_speed(iter/s)": 1.325646
+    },
+    {
+      "acc": 0.64185009,
+      "epoch": 0.03678335870116692,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.6783358701166922e-06,
+      "loss": 1.73330822,
+      "memory(GiB)": 62.29,
+      "step": 1450,
+      "train_speed(iter/s)": 1.326122
+    },
+    {
+      "acc": 0.65312219,
+      "epoch": 0.03691019786910198,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.691019786910198e-06,
+      "loss": 1.65412788,
+      "memory(GiB)": 62.29,
+      "step": 1455,
+      "train_speed(iter/s)": 1.326589
+    },
+    {
+      "acc": 0.65847569,
+      "epoch": 0.037037037037037035,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 1.54699841,
+      "memory(GiB)": 62.29,
+      "step": 1460,
+      "train_speed(iter/s)": 1.327064
+    },
+    {
+      "acc": 0.63990932,
+      "epoch": 0.03716387620497209,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.7163876204972097e-06,
+      "loss": 1.67625294,
+      "memory(GiB)": 62.29,
+      "step": 1465,
+      "train_speed(iter/s)": 1.327514
+    },
+    {
+      "acc": 0.64200163,
+      "epoch": 0.03729071537290715,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.7290715372907157e-06,
+      "loss": 1.65879059,
+      "memory(GiB)": 62.29,
+      "step": 1470,
+      "train_speed(iter/s)": 1.327954
+    },
+    {
+      "acc": 0.63913779,
+      "epoch": 0.03741755454084221,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.7417554540842216e-06,
+      "loss": 1.65850525,
+      "memory(GiB)": 62.29,
+      "step": 1475,
+      "train_speed(iter/s)": 1.328406
+    },
+    {
+      "acc": 0.66351271,
+      "epoch": 0.03754439370877727,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.754439370877727e-06,
+      "loss": 1.58990593,
+      "memory(GiB)": 62.29,
+      "step": 1480,
+      "train_speed(iter/s)": 1.328848
+    },
+    {
+      "acc": 0.64551296,
+      "epoch": 0.03767123287671233,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.767123287671233e-06,
+      "loss": 1.60236645,
+      "memory(GiB)": 62.29,
+      "step": 1485,
+      "train_speed(iter/s)": 1.329244
+    },
+    {
+      "acc": 0.6429606,
+      "epoch": 0.037798072044647385,
+      "grad_norm": 3.375,
+      "learning_rate": 3.7798072044647387e-06,
+      "loss": 1.66035156,
+      "memory(GiB)": 62.29,
+      "step": 1490,
+      "train_speed(iter/s)": 1.329688
+    },
+    {
+      "acc": 0.62678628,
+      "epoch": 0.03792491121258244,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.792491121258245e-06,
+      "loss": 1.7513979,
+      "memory(GiB)": 62.29,
+      "step": 1495,
+      "train_speed(iter/s)": 1.33014
+    },
+    {
+      "acc": 0.64224467,
+      "epoch": 0.0380517503805175,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.8051750380517506e-06,
+      "loss": 1.66251469,
+      "memory(GiB)": 62.29,
+      "step": 1500,
+      "train_speed(iter/s)": 1.330543
+    },
+    {
+      "acc": 0.65083981,
+      "epoch": 0.03817858954845256,
+      "grad_norm": 3.0,
+      "learning_rate": 3.817858954845256e-06,
+      "loss": 1.64320831,
+      "memory(GiB)": 62.29,
+      "step": 1505,
+      "train_speed(iter/s)": 1.330972
+    },
+    {
+      "acc": 0.64139471,
+      "epoch": 0.03830542871638762,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.830542871638762e-06,
+      "loss": 1.64989395,
+      "memory(GiB)": 62.29,
+      "step": 1510,
+      "train_speed(iter/s)": 1.3314
+    },
+    {
+      "acc": 0.65496244,
+      "epoch": 0.03843226788432268,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.843226788432268e-06,
+      "loss": 1.59442596,
+      "memory(GiB)": 62.29,
+      "step": 1515,
+      "train_speed(iter/s)": 1.331796
+    },
+    {
+      "acc": 0.65103788,
+      "epoch": 0.038559107052257735,
+      "grad_norm": 5.0,
+      "learning_rate": 3.855910705225774e-06,
+      "loss": 1.62994747,
+      "memory(GiB)": 62.29,
+      "step": 1520,
+      "train_speed(iter/s)": 1.332242
+    },
+    {
+      "acc": 0.64011264,
+      "epoch": 0.03868594622019279,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.86859462201928e-06,
+      "loss": 1.64594784,
+      "memory(GiB)": 62.29,
+      "step": 1525,
+      "train_speed(iter/s)": 1.332645
+    },
+    {
+      "acc": 0.63593369,
+      "epoch": 0.03881278538812785,
+      "grad_norm": 3.0,
+      "learning_rate": 3.881278538812785e-06,
+      "loss": 1.65550404,
+      "memory(GiB)": 62.29,
+      "step": 1530,
+      "train_speed(iter/s)": 1.333042
+    },
+    {
+      "acc": 0.65057888,
+      "epoch": 0.03893962455606291,
+      "grad_norm": 3.125,
+      "learning_rate": 3.893962455606292e-06,
+      "loss": 1.62578201,
+      "memory(GiB)": 62.29,
+      "step": 1535,
+      "train_speed(iter/s)": 1.333466
+    },
+    {
+      "acc": 0.66227565,
+      "epoch": 0.03906646372399797,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.906646372399797e-06,
+      "loss": 1.56446037,
+      "memory(GiB)": 62.29,
+      "step": 1540,
+      "train_speed(iter/s)": 1.333871
+    },
+    {
+      "acc": 0.63116436,
+      "epoch": 0.03919330289193303,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.919330289193303e-06,
+      "loss": 1.68866596,
+      "memory(GiB)": 62.29,
+      "step": 1545,
+      "train_speed(iter/s)": 1.334294
+    },
+    {
+      "acc": 0.62386227,
+      "epoch": 0.039320142059868085,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.932014205986809e-06,
+      "loss": 1.65193653,
+      "memory(GiB)": 62.29,
+      "step": 1550,
+      "train_speed(iter/s)": 1.33473
+    },
+    {
+      "acc": 0.65840616,
+      "epoch": 0.03944698122780314,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.944698122780315e-06,
+      "loss": 1.61437874,
+      "memory(GiB)": 62.29,
+      "step": 1555,
+      "train_speed(iter/s)": 1.335146
+    },
+    {
+      "acc": 0.6439508,
+      "epoch": 0.0395738203957382,
+      "grad_norm": 5.40625,
+      "learning_rate": 3.95738203957382e-06,
+      "loss": 1.6724699,
+      "memory(GiB)": 62.29,
+      "step": 1560,
+      "train_speed(iter/s)": 1.335555
+    },
+    {
+      "acc": 0.65688076,
+      "epoch": 0.03970065956367326,
+      "grad_norm": 4.25,
+      "learning_rate": 3.970065956367327e-06,
+      "loss": 1.62284126,
+      "memory(GiB)": 62.29,
+      "step": 1565,
+      "train_speed(iter/s)": 1.335958
+    },
+    {
+      "acc": 0.64529905,
+      "epoch": 0.03982749873160832,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.982749873160832e-06,
+      "loss": 1.65006008,
+      "memory(GiB)": 62.29,
+      "step": 1570,
+      "train_speed(iter/s)": 1.336379
+    },
+    {
+      "acc": 0.63047934,
+      "epoch": 0.03995433789954338,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.995433789954338e-06,
+      "loss": 1.71474609,
+      "memory(GiB)": 62.29,
+      "step": 1575,
+      "train_speed(iter/s)": 1.336783
+    },
+    {
+      "acc": 0.65674028,
+      "epoch": 0.040081177067478435,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.008117706747844e-06,
+      "loss": 1.6654892,
+      "memory(GiB)": 62.29,
+      "step": 1580,
+      "train_speed(iter/s)": 1.337189
+    },
+    {
+      "acc": 0.64690928,
+      "epoch": 0.04020801623541349,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.02080162354135e-06,
+      "loss": 1.60259933,
+      "memory(GiB)": 62.29,
+      "step": 1585,
+      "train_speed(iter/s)": 1.33762
+    },
+    {
+      "acc": 0.65887251,
+      "epoch": 0.04033485540334855,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.033485540334856e-06,
+      "loss": 1.60056458,
+      "memory(GiB)": 62.29,
+      "step": 1590,
+      "train_speed(iter/s)": 1.337981
+    },
+    {
+      "acc": 0.64425178,
+      "epoch": 0.04046169457128361,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.046169457128362e-06,
+      "loss": 1.66846752,
+      "memory(GiB)": 62.29,
+      "step": 1595,
+      "train_speed(iter/s)": 1.338357
+    },
+    {
+      "acc": 0.634095,
+      "epoch": 0.04058853373921867,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.058853373921867e-06,
+      "loss": 1.69305305,
+      "memory(GiB)": 62.29,
+      "step": 1600,
+      "train_speed(iter/s)": 1.338744
+    },
+    {
+      "acc": 0.66320486,
+      "epoch": 0.04071537290715373,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.071537290715373e-06,
+      "loss": 1.53615093,
+      "memory(GiB)": 62.29,
+      "step": 1605,
+      "train_speed(iter/s)": 1.339151
+    },
+    {
+      "acc": 0.65475626,
+      "epoch": 0.040842212075088785,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.084221207508879e-06,
+      "loss": 1.63777752,
+      "memory(GiB)": 62.29,
+      "step": 1610,
+      "train_speed(iter/s)": 1.339532
+    },
+    {
+      "acc": 0.64464264,
+      "epoch": 0.040969051243023843,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.096905124302385e-06,
+      "loss": 1.65603294,
+      "memory(GiB)": 62.29,
+      "step": 1615,
+      "train_speed(iter/s)": 1.339913
+    },
+    {
+      "acc": 0.65369339,
+      "epoch": 0.0410958904109589,
+      "grad_norm": 3.125,
+      "learning_rate": 4.109589041095891e-06,
+      "loss": 1.57735062,
+      "memory(GiB)": 62.29,
+      "step": 1620,
+      "train_speed(iter/s)": 1.340296
+    },
+    {
+      "acc": 0.64069161,
+      "epoch": 0.04122272957889396,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.122272957889397e-06,
+      "loss": 1.74971123,
+      "memory(GiB)": 62.29,
+      "step": 1625,
+      "train_speed(iter/s)": 1.34068
+    },
+    {
+      "acc": 0.63130312,
+      "epoch": 0.04134956874682902,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.134956874682902e-06,
+      "loss": 1.64302883,
+      "memory(GiB)": 62.29,
+      "step": 1630,
+      "train_speed(iter/s)": 1.341075
+    },
+    {
+      "acc": 0.66777143,
+      "epoch": 0.04147640791476408,
+      "grad_norm": 3.5,
+      "learning_rate": 4.147640791476408e-06,
+      "loss": 1.54624062,
+      "memory(GiB)": 62.29,
+      "step": 1635,
+      "train_speed(iter/s)": 1.341445
+    },
+    {
+      "acc": 0.64123778,
+      "epoch": 0.041603247082699135,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.160324708269914e-06,
+      "loss": 1.68955002,
+      "memory(GiB)": 62.29,
+      "step": 1640,
+      "train_speed(iter/s)": 1.341828
+    },
+    {
+      "acc": 0.66578627,
+      "epoch": 0.041730086250634194,
+      "grad_norm": 3.0,
+      "learning_rate": 4.17300862506342e-06,
+      "loss": 1.54427567,
+      "memory(GiB)": 62.29,
+      "step": 1645,
+      "train_speed(iter/s)": 1.342202
+    },
+    {
+      "acc": 0.65751562,
+      "epoch": 0.04185692541856925,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.185692541856926e-06,
+      "loss": 1.55636806,
+      "memory(GiB)": 62.29,
+      "step": 1650,
+      "train_speed(iter/s)": 1.342565
+    },
+    {
+      "acc": 0.65154347,
+      "epoch": 0.04198376458650431,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.198376458650432e-06,
+      "loss": 1.60093498,
+      "memory(GiB)": 62.29,
+      "step": 1655,
+      "train_speed(iter/s)": 1.342951
+    },
+    {
+      "acc": 0.65908852,
+      "epoch": 0.04211060375443937,
+      "grad_norm": 3.5,
+      "learning_rate": 4.211060375443937e-06,
+      "loss": 1.5715085,
+      "memory(GiB)": 62.29,
+      "step": 1660,
+      "train_speed(iter/s)": 1.343334
+    },
+    {
+      "acc": 0.63329716,
+      "epoch": 0.04223744292237443,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.223744292237444e-06,
+      "loss": 1.73878059,
+      "memory(GiB)": 62.29,
+      "step": 1665,
+      "train_speed(iter/s)": 1.343707
+    },
+    {
+      "acc": 0.65058899,
+      "epoch": 0.042364282090309485,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.236428209030949e-06,
+      "loss": 1.63434868,
+      "memory(GiB)": 62.29,
+      "step": 1670,
+      "train_speed(iter/s)": 1.344065
+    },
+    {
+      "acc": 0.63934331,
+      "epoch": 0.042491121258244544,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.249112125824455e-06,
+      "loss": 1.6823967,
+      "memory(GiB)": 62.29,
+      "step": 1675,
+      "train_speed(iter/s)": 1.344445
+    },
+    {
+      "acc": 0.65536323,
+      "epoch": 0.0426179604261796,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.261796042617961e-06,
+      "loss": 1.61363392,
+      "memory(GiB)": 62.29,
+      "step": 1680,
+      "train_speed(iter/s)": 1.344791
+    },
+    {
+      "acc": 0.66715498,
+      "epoch": 0.04274479959411466,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.274479959411467e-06,
+      "loss": 1.55411472,
+      "memory(GiB)": 62.29,
+      "step": 1685,
+      "train_speed(iter/s)": 1.345155
+    },
+    {
+      "acc": 0.65792012,
+      "epoch": 0.04287163876204972,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.287163876204972e-06,
+      "loss": 1.53218994,
+      "memory(GiB)": 62.29,
+      "step": 1690,
+      "train_speed(iter/s)": 1.345524
+    },
+    {
+      "acc": 0.64581223,
+      "epoch": 0.04299847792998478,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.299847792998479e-06,
+      "loss": 1.58741951,
+      "memory(GiB)": 62.29,
+      "step": 1695,
+      "train_speed(iter/s)": 1.345879
+    },
+    {
+      "acc": 0.65164189,
+      "epoch": 0.043125317097919835,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.312531709791984e-06,
+      "loss": 1.61671333,
+      "memory(GiB)": 62.29,
+      "step": 1700,
+      "train_speed(iter/s)": 1.346236
+    },
+    {
+      "acc": 0.63419347,
+      "epoch": 0.043252156265854894,
+      "grad_norm": 4.125,
+      "learning_rate": 4.32521562658549e-06,
+      "loss": 1.62930412,
+      "memory(GiB)": 62.29,
+      "step": 1705,
+      "train_speed(iter/s)": 1.346583
+    },
+    {
+      "acc": 0.65617952,
+      "epoch": 0.04337899543378995,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.337899543378996e-06,
+      "loss": 1.65208225,
+      "memory(GiB)": 62.29,
+      "step": 1710,
+      "train_speed(iter/s)": 1.346926
+    },
+    {
+      "acc": 0.62819157,
+      "epoch": 0.04350583460172501,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.350583460172502e-06,
+      "loss": 1.65590172,
+      "memory(GiB)": 62.29,
+      "step": 1715,
+      "train_speed(iter/s)": 1.347278
+    },
+    {
+      "acc": 0.66678252,
+      "epoch": 0.04363267376966007,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.363267376966007e-06,
+      "loss": 1.58529463,
+      "memory(GiB)": 62.29,
+      "step": 1720,
+      "train_speed(iter/s)": 1.347616
+    },
+    {
+      "acc": 0.64499016,
+      "epoch": 0.04375951293759513,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.375951293759514e-06,
+      "loss": 1.65288849,
+      "memory(GiB)": 62.29,
+      "step": 1725,
+      "train_speed(iter/s)": 1.347959
+    },
+    {
+      "acc": 0.65887966,
+      "epoch": 0.043886352105530185,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.388635210553019e-06,
+      "loss": 1.61358681,
+      "memory(GiB)": 62.29,
+      "step": 1730,
+      "train_speed(iter/s)": 1.348314
+    },
+    {
+      "acc": 0.65165248,
+      "epoch": 0.044013191273465244,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.401319127346525e-06,
+      "loss": 1.61355247,
+      "memory(GiB)": 62.29,
+      "step": 1735,
+      "train_speed(iter/s)": 1.348642
+    },
+    {
+      "acc": 0.65684047,
+      "epoch": 0.0441400304414003,
+      "grad_norm": 4.625,
+      "learning_rate": 4.414003044140031e-06,
+      "loss": 1.57958202,
+      "memory(GiB)": 62.29,
+      "step": 1740,
+      "train_speed(iter/s)": 1.348958
+    },
+    {
+      "acc": 0.65687799,
+      "epoch": 0.04426686960933536,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.426686960933537e-06,
+      "loss": 1.58574247,
+      "memory(GiB)": 62.29,
+      "step": 1745,
+      "train_speed(iter/s)": 1.349293
+    },
+    {
+      "acc": 0.64764376,
+      "epoch": 0.04439370877727042,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.439370877727043e-06,
+      "loss": 1.66606216,
+      "memory(GiB)": 62.29,
+      "step": 1750,
+      "train_speed(iter/s)": 1.349647
+    },
+    {
+      "acc": 0.64270811,
+      "epoch": 0.04452054794520548,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.4520547945205486e-06,
+      "loss": 1.63729286,
+      "memory(GiB)": 62.29,
+      "step": 1755,
+      "train_speed(iter/s)": 1.349915
+    },
+    {
+      "acc": 0.66679058,
+      "epoch": 0.044647387113140535,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.464738711314054e-06,
+      "loss": 1.57432194,
+      "memory(GiB)": 62.29,
+      "step": 1760,
+      "train_speed(iter/s)": 1.350245
+    },
+    {
+      "acc": 0.65143909,
+      "epoch": 0.044774226281075594,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.47742262810756e-06,
+      "loss": 1.65249367,
+      "memory(GiB)": 62.29,
+      "step": 1765,
+      "train_speed(iter/s)": 1.350564
+    },
+    {
+      "acc": 0.65241065,
+      "epoch": 0.04490106544901065,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.490106544901066e-06,
+      "loss": 1.56639442,
+      "memory(GiB)": 62.29,
+      "step": 1770,
+      "train_speed(iter/s)": 1.350889
+    },
+    {
+      "acc": 0.64612379,
+      "epoch": 0.04502790461694571,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.502790461694572e-06,
+      "loss": 1.66514816,
+      "memory(GiB)": 62.29,
+      "step": 1775,
+      "train_speed(iter/s)": 1.351209
+    },
+    {
+      "acc": 0.65178108,
+      "epoch": 0.04515474378488077,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.5154743784880776e-06,
+      "loss": 1.69128838,
+      "memory(GiB)": 62.29,
+      "step": 1780,
+      "train_speed(iter/s)": 1.351537
+    },
+    {
+      "acc": 0.65956163,
+      "epoch": 0.04528158295281583,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.5281582952815835e-06,
+      "loss": 1.59968872,
+      "memory(GiB)": 62.29,
+      "step": 1785,
+      "train_speed(iter/s)": 1.351842
+    },
+    {
+      "acc": 0.65635548,
+      "epoch": 0.045408422120750885,
+      "grad_norm": 3.625,
+      "learning_rate": 4.540842212075089e-06,
+      "loss": 1.5815239,
+      "memory(GiB)": 62.29,
+      "step": 1790,
+      "train_speed(iter/s)": 1.352151
+    },
+    {
+      "acc": 0.64939752,
+      "epoch": 0.045535261288685944,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.553526128868595e-06,
+      "loss": 1.64519768,
+      "memory(GiB)": 62.29,
+      "step": 1795,
+      "train_speed(iter/s)": 1.352483
+    },
+    {
+      "acc": 0.64341202,
+      "epoch": 0.045662100456621,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.566210045662101e-06,
+      "loss": 1.63342018,
+      "memory(GiB)": 62.29,
+      "step": 1800,
+      "train_speed(iter/s)": 1.35276
+    },
+    {
+      "acc": 0.67389755,
+      "epoch": 0.04578893962455606,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.5788939624556065e-06,
+      "loss": 1.48183546,
+      "memory(GiB)": 62.29,
+      "step": 1805,
+      "train_speed(iter/s)": 1.353089
+    },
+    {
+      "acc": 0.64614205,
+      "epoch": 0.04591577879249112,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.5915778792491125e-06,
+      "loss": 1.6152071,
+      "memory(GiB)": 62.29,
+      "step": 1810,
+      "train_speed(iter/s)": 1.353392
+    },
+    {
+      "acc": 0.62274632,
+      "epoch": 0.04604261796042618,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.6042617960426185e-06,
+      "loss": 1.73814812,
+      "memory(GiB)": 62.29,
+      "step": 1815,
+      "train_speed(iter/s)": 1.353704
+    },
+    {
+      "acc": 0.64962049,
+      "epoch": 0.046169457128361235,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.616945712836124e-06,
+      "loss": 1.61187973,
+      "memory(GiB)": 62.29,
+      "step": 1820,
+      "train_speed(iter/s)": 1.35401
+    },
+    {
+      "acc": 0.6687407,
+      "epoch": 0.046296296296296294,
+      "grad_norm": 3.625,
+      "learning_rate": 4.62962962962963e-06,
+      "loss": 1.58052406,
+      "memory(GiB)": 62.29,
+      "step": 1825,
+      "train_speed(iter/s)": 1.354306
+    },
+    {
+      "acc": 0.65757475,
+      "epoch": 0.04642313546423135,
+      "grad_norm": 4.0,
+      "learning_rate": 4.6423135464231355e-06,
+      "loss": 1.67497234,
+      "memory(GiB)": 62.29,
+      "step": 1830,
+      "train_speed(iter/s)": 1.354614
+    },
+    {
+      "acc": 0.67271771,
+      "epoch": 0.04654997463216641,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.6549974632166415e-06,
+      "loss": 1.61847343,
+      "memory(GiB)": 62.29,
+      "step": 1835,
+      "train_speed(iter/s)": 1.354931
+    },
+    {
+      "acc": 0.66951165,
+      "epoch": 0.04667681380010147,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.6676813800101475e-06,
+      "loss": 1.55287247,
+      "memory(GiB)": 62.29,
+      "step": 1840,
+      "train_speed(iter/s)": 1.355215
+    },
+    {
+      "acc": 0.65074606,
+      "epoch": 0.04680365296803653,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.6803652968036534e-06,
+      "loss": 1.6099123,
+      "memory(GiB)": 62.29,
+      "step": 1845,
+      "train_speed(iter/s)": 1.355521
+    },
+    {
+      "acc": 0.66185575,
+      "epoch": 0.046930492135971585,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.6930492135971586e-06,
+      "loss": 1.58433199,
+      "memory(GiB)": 62.29,
+      "step": 1850,
+      "train_speed(iter/s)": 1.35582
+    },
+    {
+      "acc": 0.65254774,
+      "epoch": 0.047057331303906644,
+      "grad_norm": 4.5,
+      "learning_rate": 4.705733130390665e-06,
+      "loss": 1.58976088,
+      "memory(GiB)": 62.29,
+      "step": 1855,
+      "train_speed(iter/s)": 1.356132
+    },
+    {
+      "acc": 0.63246574,
+      "epoch": 0.0471841704718417,
+      "grad_norm": 2.75,
+      "learning_rate": 4.7184170471841705e-06,
+      "loss": 1.62776909,
+      "memory(GiB)": 62.29,
+      "step": 1860,
+      "train_speed(iter/s)": 1.356446
+    },
+    {
+      "acc": 0.64561615,
+      "epoch": 0.04731100963977676,
+      "grad_norm": 3.125,
+      "learning_rate": 4.7311009639776765e-06,
+      "loss": 1.64066486,
+      "memory(GiB)": 62.29,
+      "step": 1865,
+      "train_speed(iter/s)": 1.356737
+    },
+    {
+      "acc": 0.62996073,
+      "epoch": 0.04743784880771182,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.743784880771182e-06,
+      "loss": 1.65244713,
+      "memory(GiB)": 62.29,
+      "step": 1870,
+      "train_speed(iter/s)": 1.357032
+    },
+    {
+      "acc": 0.64916382,
+      "epoch": 0.04756468797564688,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.756468797564688e-06,
+      "loss": 1.58223915,
+      "memory(GiB)": 62.29,
+      "step": 1875,
+      "train_speed(iter/s)": 1.357345
+    },
+    {
+      "acc": 0.6596211,
+      "epoch": 0.047691527143581935,
+      "grad_norm": 3.25,
+      "learning_rate": 4.769152714358194e-06,
+      "loss": 1.65836258,
+      "memory(GiB)": 62.29,
+      "step": 1880,
+      "train_speed(iter/s)": 1.357634
+    },
+    {
+      "acc": 0.66093788,
+      "epoch": 0.047818366311516994,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.7818366311517e-06,
+      "loss": 1.6000494,
+      "memory(GiB)": 62.29,
+      "step": 1885,
+      "train_speed(iter/s)": 1.357906
+    },
+    {
+      "acc": 0.65292997,
+      "epoch": 0.04794520547945205,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.7945205479452054e-06,
+      "loss": 1.62620144,
+      "memory(GiB)": 62.29,
+      "step": 1890,
+      "train_speed(iter/s)": 1.358202
+    },
+    {
+      "acc": 0.67180133,
+      "epoch": 0.04807204464738711,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.807204464738711e-06,
+      "loss": 1.5632576,
+      "memory(GiB)": 62.29,
+      "step": 1895,
+      "train_speed(iter/s)": 1.358484
+    },
+    {
+      "acc": 0.65027585,
+      "epoch": 0.04819888381532217,
+      "grad_norm": 5.96875,
+      "learning_rate": 4.819888381532217e-06,
+      "loss": 1.58923206,
+      "memory(GiB)": 62.29,
+      "step": 1900,
+      "train_speed(iter/s)": 1.358774
+    },
+    {
+      "acc": 0.65751038,
+      "epoch": 0.04832572298325723,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.832572298325723e-06,
+      "loss": 1.57541943,
+      "memory(GiB)": 62.29,
+      "step": 1905,
+      "train_speed(iter/s)": 1.359049
+    },
+    {
+      "acc": 0.64976282,
+      "epoch": 0.048452562151192285,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.845256215119229e-06,
+      "loss": 1.56489792,
+      "memory(GiB)": 62.29,
+      "step": 1910,
+      "train_speed(iter/s)": 1.359315
+    },
+    {
+      "acc": 0.6523912,
+      "epoch": 0.048579401319127344,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.857940131912735e-06,
+      "loss": 1.59550543,
+      "memory(GiB)": 62.29,
+      "step": 1915,
+      "train_speed(iter/s)": 1.35957
+    },
+    {
+      "acc": 0.66911092,
+      "epoch": 0.0487062404870624,
+      "grad_norm": 3.0,
+      "learning_rate": 4.87062404870624e-06,
+      "loss": 1.49573107,
+      "memory(GiB)": 62.29,
+      "step": 1920,
+      "train_speed(iter/s)": 1.359841
+    },
+    {
+      "acc": 0.64037271,
+      "epoch": 0.04883307965499746,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.883307965499746e-06,
+      "loss": 1.62893085,
+      "memory(GiB)": 62.29,
+      "step": 1925,
+      "train_speed(iter/s)": 1.360121
+    },
+    {
+      "acc": 0.63526869,
+      "epoch": 0.04895991882293252,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.895991882293252e-06,
+      "loss": 1.69589691,
+      "memory(GiB)": 62.29,
+      "step": 1930,
+      "train_speed(iter/s)": 1.36041
+    },
+    {
+      "acc": 0.65211701,
+      "epoch": 0.04908675799086758,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.908675799086758e-06,
+      "loss": 1.5645649,
+      "memory(GiB)": 62.29,
+      "step": 1935,
+      "train_speed(iter/s)": 1.360693
+    },
+    {
+      "acc": 0.65179381,
+      "epoch": 0.049213597158802636,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.921359715880264e-06,
+      "loss": 1.62254639,
+      "memory(GiB)": 62.29,
+      "step": 1940,
+      "train_speed(iter/s)": 1.360977
+    },
+    {
+      "acc": 0.65758214,
+      "epoch": 0.049340436326737694,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.93404363267377e-06,
+      "loss": 1.58042936,
+      "memory(GiB)": 62.29,
+      "step": 1945,
+      "train_speed(iter/s)": 1.361263
+    },
+    {
+      "acc": 0.67224245,
+      "epoch": 0.04946727549467275,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.946727549467275e-06,
+      "loss": 1.54983883,
+      "memory(GiB)": 62.29,
+      "step": 1950,
+      "train_speed(iter/s)": 1.361557
+    },
+    {
+      "acc": 0.66362619,
+      "epoch": 0.04959411466260781,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.959411466260781e-06,
+      "loss": 1.52440815,
+      "memory(GiB)": 62.29,
+      "step": 1955,
+      "train_speed(iter/s)": 1.361843
+    },
+    {
+      "acc": 0.6575089,
+      "epoch": 0.04972095383054287,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.972095383054287e-06,
+      "loss": 1.605653,
+      "memory(GiB)": 62.29,
+      "step": 1960,
+      "train_speed(iter/s)": 1.362137
+    },
+    {
+      "acc": 0.66059923,
+      "epoch": 0.04984779299847793,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.984779299847793e-06,
+      "loss": 1.58050718,
+      "memory(GiB)": 62.29,
+      "step": 1965,
+      "train_speed(iter/s)": 1.362398
+    },
+    {
+      "acc": 0.65961232,
+      "epoch": 0.049974632166412986,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.997463216641299e-06,
+      "loss": 1.62399025,
+      "memory(GiB)": 62.29,
+      "step": 1970,
+      "train_speed(iter/s)": 1.36267
+    },
+    {
+      "acc": 0.66079779,
+      "epoch": 0.050101471334348044,
+      "grad_norm": 3.875,
+      "learning_rate": 5.010147133434805e-06,
+      "loss": 1.57688446,
+      "memory(GiB)": 62.29,
+      "step": 1975,
+      "train_speed(iter/s)": 1.362951
+    },
+    {
+      "acc": 0.6643002,
+      "epoch": 0.0502283105022831,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.02283105022831e-06,
+      "loss": 1.52977695,
+      "memory(GiB)": 62.29,
+      "step": 1980,
+      "train_speed(iter/s)": 1.363219
+    },
+    {
+      "acc": 0.63591394,
+      "epoch": 0.05035514967021816,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.035514967021817e-06,
+      "loss": 1.67462387,
+      "memory(GiB)": 62.29,
+      "step": 1985,
+      "train_speed(iter/s)": 1.363489
+    },
+    {
+      "acc": 0.65609035,
+      "epoch": 0.05048198883815322,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.048198883815323e-06,
+      "loss": 1.59412899,
+      "memory(GiB)": 62.29,
+      "step": 1990,
+      "train_speed(iter/s)": 1.363761
+    },
+    {
+      "acc": 0.63516135,
+      "epoch": 0.05060882800608828,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.060882800608828e-06,
+      "loss": 1.66560287,
+      "memory(GiB)": 62.29,
+      "step": 1995,
+      "train_speed(iter/s)": 1.364029
+    },
+    {
+      "acc": 0.65413752,
+      "epoch": 0.050735667174023336,
+      "grad_norm": 3.625,
+      "learning_rate": 5.073566717402334e-06,
+      "loss": 1.58797359,
+      "memory(GiB)": 62.29,
+      "step": 2000,
+      "train_speed(iter/s)": 1.364308
+    },
+    {
+      "epoch": 0.050735667174023336,
+      "eval_acc": 0.6400587231678768,
+      "eval_loss": 1.5562922954559326,
+      "eval_runtime": 69.5552,
+      "eval_samples_per_second": 91.582,
+      "eval_steps_per_second": 22.903,
+      "step": 2000
+    },
+    {
+      "acc": 0.65165806,
+      "epoch": 0.050862506341958394,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.086250634195841e-06,
+      "loss": 1.58740358,
+      "memory(GiB)": 62.29,
+      "step": 2005,
+      "train_speed(iter/s)": 1.293545
+    },
+    {
+      "acc": 0.64081531,
+      "epoch": 0.05098934550989345,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.098934550989346e-06,
+      "loss": 1.57650518,
+      "memory(GiB)": 62.29,
+      "step": 2010,
+      "train_speed(iter/s)": 1.293977
+    },
+    {
+      "acc": 0.66946039,
+      "epoch": 0.05111618467782851,
+      "grad_norm": 4.34375,
+      "learning_rate": 5.111618467782852e-06,
+      "loss": 1.54907646,
+      "memory(GiB)": 62.29,
+      "step": 2015,
+      "train_speed(iter/s)": 1.29439
+    },
+    {
+      "acc": 0.65549111,
+      "epoch": 0.05124302384576357,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.124302384576357e-06,
+      "loss": 1.54122009,
+      "memory(GiB)": 62.29,
+      "step": 2020,
+      "train_speed(iter/s)": 1.294798
+    },
+    {
+      "acc": 0.65703692,
+      "epoch": 0.05136986301369863,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.136986301369864e-06,
+      "loss": 1.55630302,
+      "memory(GiB)": 62.29,
+      "step": 2025,
+      "train_speed(iter/s)": 1.295208
+    },
+    {
+      "acc": 0.65966311,
+      "epoch": 0.051496702181633686,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.149670218163369e-06,
+      "loss": 1.55264187,
+      "memory(GiB)": 62.29,
+      "step": 2030,
+      "train_speed(iter/s)": 1.295618
+    },
+    {
+      "acc": 0.65735292,
+      "epoch": 0.051623541349568744,
+      "grad_norm": 4.5,
+      "learning_rate": 5.162354134956875e-06,
+      "loss": 1.61286392,
+      "memory(GiB)": 62.29,
+      "step": 2035,
+      "train_speed(iter/s)": 1.296035
+    },
+    {
+      "acc": 0.6676672,
+      "epoch": 0.0517503805175038,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.175038051750381e-06,
+      "loss": 1.57464294,
+      "memory(GiB)": 62.35,
+      "step": 2040,
+      "train_speed(iter/s)": 1.296416
+    },
+    {
+      "acc": 0.65663013,
+      "epoch": 0.05187721968543886,
+      "grad_norm": 5.71875,
+      "learning_rate": 5.187721968543887e-06,
+      "loss": 1.60071945,
+      "memory(GiB)": 62.35,
+      "step": 2045,
+      "train_speed(iter/s)": 1.296809
+    },
+    {
+      "acc": 0.65472651,
+      "epoch": 0.05200405885337392,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.200405885337393e-06,
+      "loss": 1.53031654,
+      "memory(GiB)": 65.3,
+      "step": 2050,
+      "train_speed(iter/s)": 1.297125
+    },
+    {
+      "acc": 0.65955148,
+      "epoch": 0.05213089802130898,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.213089802130898e-06,
+      "loss": 1.59295483,
+      "memory(GiB)": 65.3,
+      "step": 2055,
+      "train_speed(iter/s)": 1.297498
+    },
+    {
+      "acc": 0.65325069,
+      "epoch": 0.052257737189244036,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.225773718924404e-06,
+      "loss": 1.55541401,
+      "memory(GiB)": 65.3,
+      "step": 2060,
+      "train_speed(iter/s)": 1.297853
+    },
+    {
+      "acc": 0.62840824,
+      "epoch": 0.052384576357179094,
+      "grad_norm": 3.875,
+      "learning_rate": 5.238457635717911e-06,
+      "loss": 1.70797157,
+      "memory(GiB)": 65.3,
+      "step": 2065,
+      "train_speed(iter/s)": 1.298045
+    },
+    {
+      "acc": 0.6588542,
+      "epoch": 0.05251141552511415,
+      "grad_norm": 4.0,
+      "learning_rate": 5.251141552511416e-06,
+      "loss": 1.60576591,
+      "memory(GiB)": 65.3,
+      "step": 2070,
+      "train_speed(iter/s)": 1.298423
+    },
+    {
+      "acc": 0.64133282,
+      "epoch": 0.05263825469304921,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.263825469304922e-06,
+      "loss": 1.65439034,
+      "memory(GiB)": 65.3,
+      "step": 2075,
+      "train_speed(iter/s)": 1.29881
+    },
+    {
+      "acc": 0.63073826,
+      "epoch": 0.05276509386098427,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.276509386098427e-06,
+      "loss": 1.66823902,
+      "memory(GiB)": 65.3,
+      "step": 2080,
+      "train_speed(iter/s)": 1.299193
+    },
+    {
+      "acc": 0.65113077,
+      "epoch": 0.05289193302891933,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.289193302891934e-06,
+      "loss": 1.58725929,
+      "memory(GiB)": 65.3,
+      "step": 2085,
+      "train_speed(iter/s)": 1.299585
+    },
+    {
+      "acc": 0.66097269,
+      "epoch": 0.053018772196854386,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.30187721968544e-06,
+      "loss": 1.56407948,
+      "memory(GiB)": 65.3,
+      "step": 2090,
+      "train_speed(iter/s)": 1.299975
+    },
+    {
+      "acc": 0.64419079,
+      "epoch": 0.053145611364789444,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.314561136478945e-06,
+      "loss": 1.62042198,
+      "memory(GiB)": 65.3,
+      "step": 2095,
+      "train_speed(iter/s)": 1.300361
+    },
+    {
+      "acc": 0.65401111,
+      "epoch": 0.0532724505327245,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.327245053272451e-06,
+      "loss": 1.61569386,
+      "memory(GiB)": 65.3,
+      "step": 2100,
+      "train_speed(iter/s)": 1.300748
+    },
+    {
+      "acc": 0.67162933,
+      "epoch": 0.05339928970065956,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.339928970065957e-06,
+      "loss": 1.49093542,
+      "memory(GiB)": 65.3,
+      "step": 2105,
+      "train_speed(iter/s)": 1.30115
+    },
+    {
+      "acc": 0.65725889,
+      "epoch": 0.05352612886859462,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.352612886859463e-06,
+      "loss": 1.62426777,
+      "memory(GiB)": 65.3,
+      "step": 2110,
+      "train_speed(iter/s)": 1.301534
+    },
+    {
+      "acc": 0.66067286,
+      "epoch": 0.05365296803652968,
+      "grad_norm": 3.96875,
+      "learning_rate": 5.365296803652969e-06,
+      "loss": 1.56055431,
+      "memory(GiB)": 65.3,
+      "step": 2115,
+      "train_speed(iter/s)": 1.301922
+    },
+    {
+      "acc": 0.65378132,
+      "epoch": 0.053779807204464736,
+      "grad_norm": 4.0,
+      "learning_rate": 5.377980720446474e-06,
+      "loss": 1.56040936,
+      "memory(GiB)": 65.3,
+      "step": 2120,
+      "train_speed(iter/s)": 1.302282
+    },
+    {
+      "acc": 0.64370894,
+      "epoch": 0.053906646372399794,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.390664637239981e-06,
+      "loss": 1.64348717,
+      "memory(GiB)": 65.3,
+      "step": 2125,
+      "train_speed(iter/s)": 1.302661
+    },
+    {
+      "acc": 0.65047874,
+      "epoch": 0.05403348554033485,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.403348554033486e-06,
+      "loss": 1.58415813,
+      "memory(GiB)": 65.3,
+      "step": 2130,
+      "train_speed(iter/s)": 1.303046
+    },
+    {
+      "acc": 0.66259513,
+      "epoch": 0.05416032470826991,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.416032470826992e-06,
+      "loss": 1.53733559,
+      "memory(GiB)": 65.3,
+      "step": 2135,
+      "train_speed(iter/s)": 1.303421
+    },
+    {
+      "acc": 0.66641726,
+      "epoch": 0.05428716387620497,
+      "grad_norm": 3.5,
+      "learning_rate": 5.428716387620497e-06,
+      "loss": 1.54684029,
+      "memory(GiB)": 65.3,
+      "step": 2140,
+      "train_speed(iter/s)": 1.303784
+    },
+    {
+      "acc": 0.6591898,
+      "epoch": 0.05441400304414003,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.441400304414004e-06,
+      "loss": 1.58084946,
+      "memory(GiB)": 65.3,
+      "step": 2145,
+      "train_speed(iter/s)": 1.304164
+    },
+    {
+      "acc": 0.65812712,
+      "epoch": 0.054540842212075086,
+      "grad_norm": 3.0,
+      "learning_rate": 5.45408422120751e-06,
+      "loss": 1.59659767,
+      "memory(GiB)": 65.3,
+      "step": 2150,
+      "train_speed(iter/s)": 1.304526
+    },
+    {
+      "acc": 0.65014801,
+      "epoch": 0.054667681380010144,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.466768138001015e-06,
+      "loss": 1.65574589,
+      "memory(GiB)": 65.3,
+      "step": 2155,
+      "train_speed(iter/s)": 1.304877
+    },
+    {
+      "acc": 0.66178131,
+      "epoch": 0.0547945205479452,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.479452054794521e-06,
+      "loss": 1.64032707,
+      "memory(GiB)": 65.3,
+      "step": 2160,
+      "train_speed(iter/s)": 1.305216
+    },
+    {
+      "acc": 0.64569559,
+      "epoch": 0.05492135971588026,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.492135971588028e-06,
+      "loss": 1.58107204,
+      "memory(GiB)": 65.3,
+      "step": 2165,
+      "train_speed(iter/s)": 1.305557
+    },
+    {
+      "acc": 0.63121347,
+      "epoch": 0.05504819888381532,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.504819888381533e-06,
+      "loss": 1.69110661,
+      "memory(GiB)": 65.3,
+      "step": 2170,
+      "train_speed(iter/s)": 1.305892
+    },
+    {
+      "acc": 0.65939097,
+      "epoch": 0.05517503805175038,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.517503805175039e-06,
+      "loss": 1.56009369,
+      "memory(GiB)": 65.3,
+      "step": 2175,
+      "train_speed(iter/s)": 1.306232
+    },
+    {
+      "acc": 0.65011153,
+      "epoch": 0.055301877219685436,
+      "grad_norm": 3.703125,
+      "learning_rate": 5.530187721968544e-06,
+      "loss": 1.5799902,
+      "memory(GiB)": 65.3,
+      "step": 2180,
+      "train_speed(iter/s)": 1.306571
+    },
+    {
+      "acc": 0.65205693,
+      "epoch": 0.055428716387620494,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.542871638762051e-06,
+      "loss": 1.61694984,
+      "memory(GiB)": 65.3,
+      "step": 2185,
+      "train_speed(iter/s)": 1.306898
+    },
+    {
+      "acc": 0.6467248,
+      "epoch": 0.05555555555555555,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.555555555555557e-06,
+      "loss": 1.65056572,
+      "memory(GiB)": 65.3,
+      "step": 2190,
+      "train_speed(iter/s)": 1.307241
+    },
+    {
+      "acc": 0.6616868,
+      "epoch": 0.05568239472349061,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.568239472349062e-06,
+      "loss": 1.56460657,
+      "memory(GiB)": 65.3,
+      "step": 2195,
+      "train_speed(iter/s)": 1.307596
+    },
+    {
+      "acc": 0.65939751,
+      "epoch": 0.05580923389142567,
+      "grad_norm": 4.25,
+      "learning_rate": 5.580923389142568e-06,
+      "loss": 1.61524944,
+      "memory(GiB)": 65.3,
+      "step": 2200,
+      "train_speed(iter/s)": 1.307958
+    },
+    {
+      "acc": 0.64898477,
+      "epoch": 0.05593607305936073,
+      "grad_norm": 3.859375,
+      "learning_rate": 5.593607305936074e-06,
+      "loss": 1.60141792,
+      "memory(GiB)": 65.3,
+      "step": 2205,
+      "train_speed(iter/s)": 1.308304
+    },
+    {
+      "acc": 0.66017475,
+      "epoch": 0.056062912227295786,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.60629122272958e-06,
+      "loss": 1.55888796,
+      "memory(GiB)": 65.3,
+      "step": 2210,
+      "train_speed(iter/s)": 1.308664
+    },
+    {
+      "acc": 0.65991173,
+      "epoch": 0.056189751395230844,
+      "grad_norm": 4.125,
+      "learning_rate": 5.618975139523085e-06,
+      "loss": 1.61729031,
+      "memory(GiB)": 65.3,
+      "step": 2215,
+      "train_speed(iter/s)": 1.309032
+    },
+    {
+      "acc": 0.65649776,
+      "epoch": 0.0563165905631659,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.631659056316591e-06,
+      "loss": 1.53391962,
+      "memory(GiB)": 65.3,
+      "step": 2220,
+      "train_speed(iter/s)": 1.309392
+    },
+    {
+      "acc": 0.66045427,
+      "epoch": 0.05644342973110096,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.644342973110098e-06,
+      "loss": 1.5829339,
+      "memory(GiB)": 65.3,
+      "step": 2225,
+      "train_speed(iter/s)": 1.30975
+    },
+    {
+      "acc": 0.66125312,
+      "epoch": 0.05657026889903602,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.657026889903603e-06,
+      "loss": 1.54624424,
+      "memory(GiB)": 65.3,
+      "step": 2230,
+      "train_speed(iter/s)": 1.310063
+    },
+    {
+      "acc": 0.64767761,
+      "epoch": 0.05669710806697108,
+      "grad_norm": 3.125,
+      "learning_rate": 5.669710806697109e-06,
+      "loss": 1.64266586,
+      "memory(GiB)": 65.3,
+      "step": 2235,
+      "train_speed(iter/s)": 1.310409
+    },
+    {
+      "acc": 0.65191269,
+      "epoch": 0.056823947234906136,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.682394723490614e-06,
+      "loss": 1.5869235,
+      "memory(GiB)": 65.3,
+      "step": 2240,
+      "train_speed(iter/s)": 1.310768
+    },
+    {
+      "acc": 0.64592085,
+      "epoch": 0.056950786402841194,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.695078640284121e-06,
+      "loss": 1.60880661,
+      "memory(GiB)": 65.3,
+      "step": 2245,
+      "train_speed(iter/s)": 1.311128
+    },
+    {
+      "acc": 0.6472518,
+      "epoch": 0.05707762557077625,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.7077625570776266e-06,
+      "loss": 1.56454811,
+      "memory(GiB)": 65.3,
+      "step": 2250,
+      "train_speed(iter/s)": 1.311486
+    },
+    {
+      "acc": 0.64903326,
+      "epoch": 0.05720446473871131,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.720446473871132e-06,
+      "loss": 1.60309772,
+      "memory(GiB)": 65.3,
+      "step": 2255,
+      "train_speed(iter/s)": 1.311848
+    },
+    {
+      "acc": 0.67201586,
+      "epoch": 0.05733130390664637,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.733130390664638e-06,
+      "loss": 1.49630013,
+      "memory(GiB)": 65.3,
+      "step": 2260,
+      "train_speed(iter/s)": 1.312197
+    },
+    {
+      "acc": 0.65223184,
+      "epoch": 0.05745814307458143,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.7458143074581445e-06,
+      "loss": 1.60007362,
+      "memory(GiB)": 65.3,
+      "step": 2265,
+      "train_speed(iter/s)": 1.312536
+    },
+    {
+      "acc": 0.64231167,
+      "epoch": 0.057584982242516486,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.75849822425165e-06,
+      "loss": 1.62872143,
+      "memory(GiB)": 65.3,
+      "step": 2270,
+      "train_speed(iter/s)": 1.312885
+    },
+    {
+      "acc": 0.64150224,
+      "epoch": 0.057711821410451544,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.7711821410451556e-06,
+      "loss": 1.66193371,
+      "memory(GiB)": 65.3,
+      "step": 2275,
+      "train_speed(iter/s)": 1.313208
+    },
+    {
+      "acc": 0.65749187,
+      "epoch": 0.0578386605783866,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.783866057838661e-06,
+      "loss": 1.53441715,
+      "memory(GiB)": 65.3,
+      "step": 2280,
+      "train_speed(iter/s)": 1.313547
+    },
+    {
+      "acc": 0.6376668,
+      "epoch": 0.05796549974632166,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.7965499746321675e-06,
+      "loss": 1.67353325,
+      "memory(GiB)": 65.3,
+      "step": 2285,
+      "train_speed(iter/s)": 1.313886
+    },
+    {
+      "acc": 0.65504532,
+      "epoch": 0.05809233891425672,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.809233891425673e-06,
+      "loss": 1.58919859,
+      "memory(GiB)": 65.3,
+      "step": 2290,
+      "train_speed(iter/s)": 1.314172
+    },
+    {
+      "acc": 0.66213779,
+      "epoch": 0.05821917808219178,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.821917808219179e-06,
+      "loss": 1.54741592,
+      "memory(GiB)": 65.3,
+      "step": 2295,
+      "train_speed(iter/s)": 1.314508
+    },
+    {
+      "acc": 0.65790019,
+      "epoch": 0.058346017250126836,
+      "grad_norm": 3.859375,
+      "learning_rate": 5.834601725012684e-06,
+      "loss": 1.56761265,
+      "memory(GiB)": 65.3,
+      "step": 2300,
+      "train_speed(iter/s)": 1.31479
+    },
+    {
+      "acc": 0.65352702,
+      "epoch": 0.058472856418061894,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.8472856418061905e-06,
+      "loss": 1.60069771,
+      "memory(GiB)": 65.3,
+      "step": 2305,
+      "train_speed(iter/s)": 1.31512
+    },
+    {
+      "acc": 0.65462809,
+      "epoch": 0.05859969558599695,
+      "grad_norm": 3.5,
+      "learning_rate": 5.8599695585996965e-06,
+      "loss": 1.5595892,
+      "memory(GiB)": 65.3,
+      "step": 2310,
+      "train_speed(iter/s)": 1.315456
+    },
+    {
+      "acc": 0.66006002,
+      "epoch": 0.05872653475393201,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.872653475393202e-06,
+      "loss": 1.51939678,
+      "memory(GiB)": 65.3,
+      "step": 2315,
+      "train_speed(iter/s)": 1.315781
+    },
+    {
+      "acc": 0.66119986,
+      "epoch": 0.05885337392186707,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.8853373921867076e-06,
+      "loss": 1.63669281,
+      "memory(GiB)": 65.3,
+      "step": 2320,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.67331576,
+      "epoch": 0.05898021308980213,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.898021308980214e-06,
+      "loss": 1.51746941,
+      "memory(GiB)": 65.3,
+      "step": 2325,
+      "train_speed(iter/s)": 1.31643
+    },
+    {
+      "acc": 0.63880301,
+      "epoch": 0.059107052257737186,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.9107052257737195e-06,
+      "loss": 1.70703297,
+      "memory(GiB)": 65.3,
+      "step": 2330,
+      "train_speed(iter/s)": 1.316731
+    },
+    {
+      "acc": 0.65290232,
+      "epoch": 0.059233891425672244,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.9233891425672255e-06,
+      "loss": 1.64044933,
+      "memory(GiB)": 65.3,
+      "step": 2335,
+      "train_speed(iter/s)": 1.317025
+    },
+    {
+      "acc": 0.66641102,
+      "epoch": 0.0593607305936073,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.936073059360731e-06,
+      "loss": 1.50991259,
+      "memory(GiB)": 65.3,
+      "step": 2340,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.66091061,
+      "epoch": 0.05948756976154236,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.948756976154237e-06,
+      "loss": 1.59336023,
+      "memory(GiB)": 65.3,
+      "step": 2345,
+      "train_speed(iter/s)": 1.317604
+    },
+    {
+      "acc": 0.66509733,
+      "epoch": 0.05961440892947742,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.961440892947743e-06,
+      "loss": 1.59805222,
+      "memory(GiB)": 65.3,
+      "step": 2350,
+      "train_speed(iter/s)": 1.317902
+    },
+    {
+      "acc": 0.66100645,
+      "epoch": 0.05974124809741248,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.9741248097412485e-06,
+      "loss": 1.5082634,
+      "memory(GiB)": 65.3,
+      "step": 2355,
+      "train_speed(iter/s)": 1.318215
+    },
+    {
+      "acc": 0.65622702,
+      "epoch": 0.059868087265347536,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.9868087265347545e-06,
+      "loss": 1.56978521,
+      "memory(GiB)": 65.3,
+      "step": 2360,
+      "train_speed(iter/s)": 1.318527
+    },
+    {
+      "acc": 0.65847769,
+      "epoch": 0.059994926433282594,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.99949264332826e-06,
+      "loss": 1.53776636,
+      "memory(GiB)": 65.3,
+      "step": 2365,
+      "train_speed(iter/s)": 1.318848
+    },
+    {
+      "acc": 0.66226592,
+      "epoch": 0.06012176560121765,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.012176560121766e-06,
+      "loss": 1.61820927,
+      "memory(GiB)": 65.3,
+      "step": 2370,
+      "train_speed(iter/s)": 1.319165
+    },
+    {
+      "acc": 0.66783409,
+      "epoch": 0.06024860476915271,
+      "grad_norm": 4.125,
+      "learning_rate": 6.0248604769152715e-06,
+      "loss": 1.55879936,
+      "memory(GiB)": 65.3,
+      "step": 2375,
+      "train_speed(iter/s)": 1.319489
+    },
+    {
+      "acc": 0.66569734,
+      "epoch": 0.06037544393708777,
+      "grad_norm": 3.875,
+      "learning_rate": 6.0375443937087775e-06,
+      "loss": 1.49723511,
+      "memory(GiB)": 65.3,
+      "step": 2380,
+      "train_speed(iter/s)": 1.319818
+    },
+    {
+      "acc": 0.65687566,
+      "epoch": 0.06050228310502283,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.050228310502284e-06,
+      "loss": 1.66681175,
+      "memory(GiB)": 65.3,
+      "step": 2385,
+      "train_speed(iter/s)": 1.320102
+    },
+    {
+      "acc": 0.65734057,
+      "epoch": 0.060629122272957886,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.062912227295789e-06,
+      "loss": 1.56518583,
+      "memory(GiB)": 65.3,
+      "step": 2390,
+      "train_speed(iter/s)": 1.320408
+    },
+    {
+      "acc": 0.65311246,
+      "epoch": 0.060755961440892944,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.075596144089295e-06,
+      "loss": 1.5265089,
+      "memory(GiB)": 65.3,
+      "step": 2395,
+      "train_speed(iter/s)": 1.320717
+    },
+    {
+      "acc": 0.65005002,
+      "epoch": 0.060882800608828,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.0882800608828005e-06,
+      "loss": 1.6343544,
+      "memory(GiB)": 65.3,
+      "step": 2400,
+      "train_speed(iter/s)": 1.321025
+    },
+    {
+      "acc": 0.64703693,
+      "epoch": 0.06100963977676306,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.100963977676307e-06,
+      "loss": 1.62248993,
+      "memory(GiB)": 65.3,
+      "step": 2405,
+      "train_speed(iter/s)": 1.321348
+    },
+    {
+      "acc": 0.64315124,
+      "epoch": 0.06113647894469812,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.113647894469813e-06,
+      "loss": 1.60003014,
+      "memory(GiB)": 65.3,
+      "step": 2410,
+      "train_speed(iter/s)": 1.321658
+    },
+    {
+      "acc": 0.66619282,
+      "epoch": 0.06126331811263318,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.126331811263318e-06,
+      "loss": 1.51605787,
+      "memory(GiB)": 65.3,
+      "step": 2415,
+      "train_speed(iter/s)": 1.321925
+    },
+    {
+      "acc": 0.67183599,
+      "epoch": 0.061390157280568236,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.139015728056824e-06,
+      "loss": 1.51507788,
+      "memory(GiB)": 65.3,
+      "step": 2420,
+      "train_speed(iter/s)": 1.32224
+    },
+    {
+      "acc": 0.65276022,
+      "epoch": 0.061516996448503294,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.151699644850331e-06,
+      "loss": 1.60067978,
+      "memory(GiB)": 65.3,
+      "step": 2425,
+      "train_speed(iter/s)": 1.32255
+    },
+    {
+      "acc": 0.6697998,
+      "epoch": 0.06164383561643835,
+      "grad_norm": 2.75,
+      "learning_rate": 6.164383561643836e-06,
+      "loss": 1.50167065,
+      "memory(GiB)": 65.3,
+      "step": 2430,
+      "train_speed(iter/s)": 1.322847
+    },
+    {
+      "acc": 0.64442101,
+      "epoch": 0.06177067478437341,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.177067478437342e-06,
+      "loss": 1.6072361,
+      "memory(GiB)": 65.3,
+      "step": 2435,
+      "train_speed(iter/s)": 1.323091
+    },
+    {
+      "acc": 0.64915524,
+      "epoch": 0.06189751395230847,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.189751395230847e-06,
+      "loss": 1.60281067,
+      "memory(GiB)": 65.3,
+      "step": 2440,
+      "train_speed(iter/s)": 1.323373
+    },
+    {
+      "acc": 0.65049796,
+      "epoch": 0.06202435312024353,
+      "grad_norm": 5.65625,
+      "learning_rate": 6.202435312024354e-06,
+      "loss": 1.62260628,
+      "memory(GiB)": 65.3,
+      "step": 2445,
+      "train_speed(iter/s)": 1.323665
+    },
+    {
+      "acc": 0.67454443,
+      "epoch": 0.062151192288178586,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.215119228817859e-06,
+      "loss": 1.48319702,
+      "memory(GiB)": 65.3,
+      "step": 2450,
+      "train_speed(iter/s)": 1.323963
+    },
+    {
+      "acc": 0.65109062,
+      "epoch": 0.062278031456113644,
+      "grad_norm": 3.0,
+      "learning_rate": 6.227803145611365e-06,
+      "loss": 1.62369385,
+      "memory(GiB)": 65.3,
+      "step": 2455,
+      "train_speed(iter/s)": 1.324263
+    },
+    {
+      "acc": 0.64107704,
+      "epoch": 0.0624048706240487,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.24048706240487e-06,
+      "loss": 1.63214016,
+      "memory(GiB)": 65.3,
+      "step": 2460,
+      "train_speed(iter/s)": 1.324567
+    },
+    {
+      "acc": 0.66712132,
+      "epoch": 0.06253170979198376,
+      "grad_norm": 4.8125,
+      "learning_rate": 6.253170979198377e-06,
+      "loss": 1.55880089,
+      "memory(GiB)": 65.3,
+      "step": 2465,
+      "train_speed(iter/s)": 1.324856
+    },
+    {
+      "acc": 0.6580163,
+      "epoch": 0.06265854895991882,
+      "grad_norm": 3.375,
+      "learning_rate": 6.265854895991883e-06,
+      "loss": 1.57331734,
+      "memory(GiB)": 65.3,
+      "step": 2470,
+      "train_speed(iter/s)": 1.325155
+    },
+    {
+      "acc": 0.66523843,
+      "epoch": 0.06278538812785388,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.278538812785388e-06,
+      "loss": 1.58345985,
+      "memory(GiB)": 65.3,
+      "step": 2475,
+      "train_speed(iter/s)": 1.325459
+    },
+    {
+      "acc": 0.65903344,
+      "epoch": 0.06291222729578894,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.291222729578894e-06,
+      "loss": 1.56365404,
+      "memory(GiB)": 65.3,
+      "step": 2480,
+      "train_speed(iter/s)": 1.325734
+    },
+    {
+      "acc": 0.65830026,
+      "epoch": 0.063039066463724,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.303906646372401e-06,
+      "loss": 1.63926105,
+      "memory(GiB)": 65.3,
+      "step": 2485,
+      "train_speed(iter/s)": 1.325999
+    },
+    {
+      "acc": 0.66300197,
+      "epoch": 0.06316590563165905,
+      "grad_norm": 4.125,
+      "learning_rate": 6.316590563165906e-06,
+      "loss": 1.55130138,
+      "memory(GiB)": 65.3,
+      "step": 2490,
+      "train_speed(iter/s)": 1.326264
+    },
+    {
+      "acc": 0.64749212,
+      "epoch": 0.06329274479959411,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.329274479959412e-06,
+      "loss": 1.57835913,
+      "memory(GiB)": 65.3,
+      "step": 2495,
+      "train_speed(iter/s)": 1.326539
+    },
+    {
+      "acc": 0.64781647,
+      "epoch": 0.06341958396752917,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.341958396752917e-06,
+      "loss": 1.63324432,
+      "memory(GiB)": 65.3,
+      "step": 2500,
+      "train_speed(iter/s)": 1.326804
+    },
+    {
+      "acc": 0.66330729,
+      "epoch": 0.06354642313546423,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.354642313546424e-06,
+      "loss": 1.59984846,
+      "memory(GiB)": 65.3,
+      "step": 2505,
+      "train_speed(iter/s)": 1.327086
+    },
+    {
+      "acc": 0.65390286,
+      "epoch": 0.06367326230339929,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.36732623033993e-06,
+      "loss": 1.59474373,
+      "memory(GiB)": 65.3,
+      "step": 2510,
+      "train_speed(iter/s)": 1.327343
+    },
+    {
+      "acc": 0.66617031,
+      "epoch": 0.06380010147133434,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.380010147133435e-06,
+      "loss": 1.48961353,
+      "memory(GiB)": 65.3,
+      "step": 2515,
+      "train_speed(iter/s)": 1.327597
+    },
+    {
+      "acc": 0.66569109,
+      "epoch": 0.0639269406392694,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.392694063926941e-06,
+      "loss": 1.57572346,
+      "memory(GiB)": 65.3,
+      "step": 2520,
+      "train_speed(iter/s)": 1.327895
+    },
+    {
+      "acc": 0.65236936,
+      "epoch": 0.06405377980720446,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.405377980720447e-06,
+      "loss": 1.61169014,
+      "memory(GiB)": 65.3,
+      "step": 2525,
+      "train_speed(iter/s)": 1.328172
+    },
+    {
+      "acc": 0.67474947,
+      "epoch": 0.06418061897513952,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.418061897513953e-06,
+      "loss": 1.48528004,
+      "memory(GiB)": 65.3,
+      "step": 2530,
+      "train_speed(iter/s)": 1.328432
+    },
+    {
+      "acc": 0.65107365,
+      "epoch": 0.06430745814307458,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.430745814307458e-06,
+      "loss": 1.54318333,
+      "memory(GiB)": 65.3,
+      "step": 2535,
+      "train_speed(iter/s)": 1.328722
+    },
+    {
+      "acc": 0.65303926,
+      "epoch": 0.06443429731100964,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.443429731100964e-06,
+      "loss": 1.58706656,
+      "memory(GiB)": 65.3,
+      "step": 2540,
+      "train_speed(iter/s)": 1.329006
+    },
+    {
+      "acc": 0.68046885,
+      "epoch": 0.0645611364789447,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.456113647894471e-06,
+      "loss": 1.49837341,
+      "memory(GiB)": 65.3,
+      "step": 2545,
+      "train_speed(iter/s)": 1.329291
+    },
+    {
+      "acc": 0.65848417,
+      "epoch": 0.06468797564687975,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.468797564687976e-06,
+      "loss": 1.51504955,
+      "memory(GiB)": 65.3,
+      "step": 2550,
+      "train_speed(iter/s)": 1.329553
+    },
+    {
+      "acc": 0.67981606,
+      "epoch": 0.06481481481481481,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.481481481481482e-06,
+      "loss": 1.4530714,
+      "memory(GiB)": 65.3,
+      "step": 2555,
+      "train_speed(iter/s)": 1.329839
+    },
+    {
+      "acc": 0.66266384,
+      "epoch": 0.06494165398274987,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.494165398274987e-06,
+      "loss": 1.56371422,
+      "memory(GiB)": 65.3,
+      "step": 2560,
+      "train_speed(iter/s)": 1.330117
+    },
+    {
+      "acc": 0.64530859,
+      "epoch": 0.06506849315068493,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.506849315068494e-06,
+      "loss": 1.58655777,
+      "memory(GiB)": 65.3,
+      "step": 2565,
+      "train_speed(iter/s)": 1.330384
+    },
+    {
+      "acc": 0.66285019,
+      "epoch": 0.06519533231861999,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.519533231862e-06,
+      "loss": 1.60598755,
+      "memory(GiB)": 65.3,
+      "step": 2570,
+      "train_speed(iter/s)": 1.330672
+    },
+    {
+      "acc": 0.67049885,
+      "epoch": 0.06532217148655504,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.532217148655505e-06,
+      "loss": 1.51078377,
+      "memory(GiB)": 65.3,
+      "step": 2575,
+      "train_speed(iter/s)": 1.330955
+    },
+    {
+      "acc": 0.65210819,
+      "epoch": 0.0654490106544901,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.544901065449011e-06,
+      "loss": 1.71366386,
+      "memory(GiB)": 65.3,
+      "step": 2580,
+      "train_speed(iter/s)": 1.331237
+    },
+    {
+      "acc": 0.64366999,
+      "epoch": 0.06557584982242516,
+      "grad_norm": 5.6875,
+      "learning_rate": 6.557584982242518e-06,
+      "loss": 1.62807255,
+      "memory(GiB)": 65.3,
+      "step": 2585,
+      "train_speed(iter/s)": 1.331516
+    },
+    {
+      "acc": 0.65762553,
+      "epoch": 0.06570268899036022,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.570268899036023e-06,
+      "loss": 1.59431581,
+      "memory(GiB)": 65.3,
+      "step": 2590,
+      "train_speed(iter/s)": 1.331787
+    },
+    {
+      "acc": 0.65858231,
+      "epoch": 0.06582952815829528,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.582952815829529e-06,
+      "loss": 1.6003355,
+      "memory(GiB)": 65.3,
+      "step": 2595,
+      "train_speed(iter/s)": 1.332055
+    },
+    {
+      "acc": 0.66081228,
+      "epoch": 0.06595636732623034,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.595636732623034e-06,
+      "loss": 1.64874325,
+      "memory(GiB)": 65.3,
+      "step": 2600,
+      "train_speed(iter/s)": 1.332328
+    },
+    {
+      "acc": 0.66371684,
+      "epoch": 0.0660832064941654,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.608320649416541e-06,
+      "loss": 1.51299763,
+      "memory(GiB)": 65.3,
+      "step": 2605,
+      "train_speed(iter/s)": 1.332605
+    },
+    {
+      "acc": 0.65721502,
+      "epoch": 0.06621004566210045,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.621004566210046e-06,
+      "loss": 1.59188881,
+      "memory(GiB)": 65.3,
+      "step": 2610,
+      "train_speed(iter/s)": 1.332885
+    },
+    {
+      "acc": 0.65728092,
+      "epoch": 0.06633688483003551,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.633688483003552e-06,
+      "loss": 1.56768761,
+      "memory(GiB)": 65.3,
+      "step": 2615,
+      "train_speed(iter/s)": 1.333135
+    },
+    {
+      "acc": 0.6568872,
+      "epoch": 0.06646372399797057,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.646372399797057e-06,
+      "loss": 1.5732214,
+      "memory(GiB)": 65.3,
+      "step": 2620,
+      "train_speed(iter/s)": 1.333406
+    },
+    {
+      "acc": 0.66148758,
+      "epoch": 0.06659056316590563,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.659056316590564e-06,
+      "loss": 1.5748107,
+      "memory(GiB)": 65.3,
+      "step": 2625,
+      "train_speed(iter/s)": 1.333665
+    },
+    {
+      "acc": 0.65928268,
+      "epoch": 0.06671740233384069,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.67174023338407e-06,
+      "loss": 1.57870922,
+      "memory(GiB)": 65.3,
+      "step": 2630,
+      "train_speed(iter/s)": 1.33395
+    },
+    {
+      "acc": 0.65914955,
+      "epoch": 0.06684424150177574,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.684424150177575e-06,
+      "loss": 1.58941746,
+      "memory(GiB)": 65.3,
+      "step": 2635,
+      "train_speed(iter/s)": 1.334197
+    },
+    {
+      "acc": 0.66338844,
+      "epoch": 0.0669710806697108,
+      "grad_norm": 3.78125,
+      "learning_rate": 6.697108066971081e-06,
+      "loss": 1.62485352,
+      "memory(GiB)": 65.3,
+      "step": 2640,
+      "train_speed(iter/s)": 1.334476
+    },
+    {
+      "acc": 0.66377935,
+      "epoch": 0.06709791983764586,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.709791983764588e-06,
+      "loss": 1.52430553,
+      "memory(GiB)": 65.3,
+      "step": 2645,
+      "train_speed(iter/s)": 1.334742
+    },
+    {
+      "acc": 0.65637236,
+      "epoch": 0.06722475900558092,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.722475900558093e-06,
+      "loss": 1.57229214,
+      "memory(GiB)": 65.3,
+      "step": 2650,
+      "train_speed(iter/s)": 1.334999
+    },
+    {
+      "acc": 0.67140255,
+      "epoch": 0.06735159817351598,
+      "grad_norm": 4.875,
+      "learning_rate": 6.735159817351599e-06,
+      "loss": 1.51575451,
+      "memory(GiB)": 65.3,
+      "step": 2655,
+      "train_speed(iter/s)": 1.335248
+    },
+    {
+      "acc": 0.66453533,
+      "epoch": 0.06747843734145104,
+      "grad_norm": 3.953125,
+      "learning_rate": 6.747843734145104e-06,
+      "loss": 1.54321079,
+      "memory(GiB)": 65.3,
+      "step": 2660,
+      "train_speed(iter/s)": 1.335496
+    },
+    {
+      "acc": 0.65690384,
+      "epoch": 0.0676052765093861,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.760527650938611e-06,
+      "loss": 1.50556831,
+      "memory(GiB)": 65.3,
+      "step": 2665,
+      "train_speed(iter/s)": 1.33571
+    },
+    {
+      "acc": 0.65272622,
+      "epoch": 0.06773211567732115,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.773211567732117e-06,
+      "loss": 1.572363,
+      "memory(GiB)": 65.3,
+      "step": 2670,
+      "train_speed(iter/s)": 1.33596
+    },
+    {
+      "acc": 0.66410699,
+      "epoch": 0.06785895484525621,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.785895484525622e-06,
+      "loss": 1.62515526,
+      "memory(GiB)": 65.3,
+      "step": 2675,
+      "train_speed(iter/s)": 1.336197
+    },
+    {
+      "acc": 0.67080054,
+      "epoch": 0.06798579401319127,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.798579401319128e-06,
+      "loss": 1.51394062,
+      "memory(GiB)": 65.3,
+      "step": 2680,
+      "train_speed(iter/s)": 1.336436
+    },
+    {
+      "acc": 0.65647745,
+      "epoch": 0.06811263318112633,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.811263318112634e-06,
+      "loss": 1.60272293,
+      "memory(GiB)": 65.3,
+      "step": 2685,
+      "train_speed(iter/s)": 1.336676
+    },
+    {
+      "acc": 0.64313736,
+      "epoch": 0.06823947234906139,
+      "grad_norm": 3.75,
+      "learning_rate": 6.82394723490614e-06,
+      "loss": 1.58883419,
+      "memory(GiB)": 65.3,
+      "step": 2690,
+      "train_speed(iter/s)": 1.336929
+    },
+    {
+      "acc": 0.66347036,
+      "epoch": 0.06836631151699644,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.836631151699645e-06,
+      "loss": 1.58548546,
+      "memory(GiB)": 65.3,
+      "step": 2695,
+      "train_speed(iter/s)": 1.33718
+    },
+    {
+      "acc": 0.64199762,
+      "epoch": 0.0684931506849315,
+      "grad_norm": 4.0,
+      "learning_rate": 6.849315068493151e-06,
+      "loss": 1.64280167,
+      "memory(GiB)": 65.3,
+      "step": 2700,
+      "train_speed(iter/s)": 1.337418
+    },
+    {
+      "acc": 0.65263944,
+      "epoch": 0.06861998985286656,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.861998985286658e-06,
+      "loss": 1.62033043,
+      "memory(GiB)": 65.3,
+      "step": 2705,
+      "train_speed(iter/s)": 1.337669
+    },
+    {
+      "acc": 0.65723596,
+      "epoch": 0.06874682902080162,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.874682902080163e-06,
+      "loss": 1.67069283,
+      "memory(GiB)": 65.3,
+      "step": 2710,
+      "train_speed(iter/s)": 1.33793
+    },
+    {
+      "acc": 0.65880198,
+      "epoch": 0.06887366818873668,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.887366818873669e-06,
+      "loss": 1.52783737,
+      "memory(GiB)": 65.3,
+      "step": 2715,
+      "train_speed(iter/s)": 1.338178
+    },
+    {
+      "acc": 0.65984445,
+      "epoch": 0.06900050735667174,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.900050735667174e-06,
+      "loss": 1.56292906,
+      "memory(GiB)": 65.3,
+      "step": 2720,
+      "train_speed(iter/s)": 1.338428
+    },
+    {
+      "acc": 0.6618494,
+      "epoch": 0.0691273465246068,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.912734652460681e-06,
+      "loss": 1.54205666,
+      "memory(GiB)": 65.3,
+      "step": 2725,
+      "train_speed(iter/s)": 1.33868
+    },
+    {
+      "acc": 0.66311941,
+      "epoch": 0.06925418569254185,
+      "grad_norm": 4.875,
+      "learning_rate": 6.925418569254187e-06,
+      "loss": 1.51663141,
+      "memory(GiB)": 65.3,
+      "step": 2730,
+      "train_speed(iter/s)": 1.338945
+    },
+    {
+      "acc": 0.66315069,
+      "epoch": 0.06938102486047691,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.938102486047692e-06,
+      "loss": 1.51185608,
+      "memory(GiB)": 65.3,
+      "step": 2735,
+      "train_speed(iter/s)": 1.339197
+    },
+    {
+      "acc": 0.65801954,
+      "epoch": 0.06950786402841197,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.950786402841198e-06,
+      "loss": 1.54594097,
+      "memory(GiB)": 65.3,
+      "step": 2740,
+      "train_speed(iter/s)": 1.339451
+    },
+    {
+      "acc": 0.66891026,
+      "epoch": 0.06963470319634703,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.9634703196347046e-06,
+      "loss": 1.51121778,
+      "memory(GiB)": 65.3,
+      "step": 2745,
+      "train_speed(iter/s)": 1.339701
+    },
+    {
+      "acc": 0.66920109,
+      "epoch": 0.06976154236428209,
+      "grad_norm": 3.125,
+      "learning_rate": 6.97615423642821e-06,
+      "loss": 1.58297043,
+      "memory(GiB)": 65.3,
+      "step": 2750,
+      "train_speed(iter/s)": 1.339953
+    },
+    {
+      "acc": 0.65807948,
+      "epoch": 0.06988838153221714,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.988838153221716e-06,
+      "loss": 1.59932699,
+      "memory(GiB)": 65.3,
+      "step": 2755,
+      "train_speed(iter/s)": 1.340199
+    },
+    {
+      "acc": 0.65131001,
+      "epoch": 0.0700152207001522,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.001522070015221e-06,
+      "loss": 1.57210846,
+      "memory(GiB)": 65.3,
+      "step": 2760,
+      "train_speed(iter/s)": 1.340439
+    },
+    {
+      "acc": 0.66332998,
+      "epoch": 0.07014205986808726,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.014205986808728e-06,
+      "loss": 1.64633179,
+      "memory(GiB)": 65.3,
+      "step": 2765,
+      "train_speed(iter/s)": 1.340685
+    },
+    {
+      "acc": 0.66192002,
+      "epoch": 0.07026889903602232,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.026889903602233e-06,
+      "loss": 1.57126532,
+      "memory(GiB)": 65.3,
+      "step": 2770,
+      "train_speed(iter/s)": 1.34095
+    },
+    {
+      "acc": 0.66644707,
+      "epoch": 0.07039573820395738,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.039573820395739e-06,
+      "loss": 1.5706377,
+      "memory(GiB)": 65.3,
+      "step": 2775,
+      "train_speed(iter/s)": 1.341197
+    },
+    {
+      "acc": 0.64233942,
+      "epoch": 0.07052257737189244,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.052257737189245e-06,
+      "loss": 1.60709648,
+      "memory(GiB)": 65.3,
+      "step": 2780,
+      "train_speed(iter/s)": 1.341442
+    },
+    {
+      "acc": 0.66491318,
+      "epoch": 0.0706494165398275,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.064941653982751e-06,
+      "loss": 1.56615925,
+      "memory(GiB)": 65.3,
+      "step": 2785,
+      "train_speed(iter/s)": 1.34165
+    },
+    {
+      "acc": 0.65612674,
+      "epoch": 0.07077625570776255,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.077625570776257e-06,
+      "loss": 1.620368,
+      "memory(GiB)": 65.3,
+      "step": 2790,
+      "train_speed(iter/s)": 1.34188
+    },
+    {
+      "acc": 0.66871338,
+      "epoch": 0.07090309487569761,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.090309487569762e-06,
+      "loss": 1.49709892,
+      "memory(GiB)": 65.3,
+      "step": 2795,
+      "train_speed(iter/s)": 1.342108
+    },
+    {
+      "acc": 0.68628492,
+      "epoch": 0.07102993404363267,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.102993404363268e-06,
+      "loss": 1.47783432,
+      "memory(GiB)": 65.3,
+      "step": 2800,
+      "train_speed(iter/s)": 1.342338
+    },
+    {
+      "acc": 0.67426662,
+      "epoch": 0.07115677321156773,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.1156773211567745e-06,
+      "loss": 1.55702438,
+      "memory(GiB)": 65.3,
+      "step": 2805,
+      "train_speed(iter/s)": 1.342547
+    },
+    {
+      "acc": 0.67657957,
+      "epoch": 0.07128361237950279,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.12836123795028e-06,
+      "loss": 1.57122726,
+      "memory(GiB)": 65.3,
+      "step": 2810,
+      "train_speed(iter/s)": 1.342783
+    },
+    {
+      "acc": 0.64241905,
+      "epoch": 0.07141045154743784,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.1410451547437856e-06,
+      "loss": 1.5799408,
+      "memory(GiB)": 65.3,
+      "step": 2815,
+      "train_speed(iter/s)": 1.342995
+    },
+    {
+      "acc": 0.65822248,
+      "epoch": 0.0715372907153729,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.153729071537291e-06,
+      "loss": 1.50241089,
+      "memory(GiB)": 65.3,
+      "step": 2820,
+      "train_speed(iter/s)": 1.343223
+    },
+    {
+      "acc": 0.66720753,
+      "epoch": 0.07166412988330796,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.1664129883307975e-06,
+      "loss": 1.52185326,
+      "memory(GiB)": 65.3,
+      "step": 2825,
+      "train_speed(iter/s)": 1.343449
+    },
+    {
+      "acc": 0.67360511,
+      "epoch": 0.07179096905124302,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.1790969051243035e-06,
+      "loss": 1.5067234,
+      "memory(GiB)": 65.3,
+      "step": 2830,
+      "train_speed(iter/s)": 1.343669
+    },
+    {
+      "acc": 0.64081078,
+      "epoch": 0.07191780821917808,
+      "grad_norm": 4.25,
+      "learning_rate": 7.191780821917809e-06,
+      "loss": 1.65083389,
+      "memory(GiB)": 65.3,
+      "step": 2835,
+      "train_speed(iter/s)": 1.34389
+    },
+    {
+      "acc": 0.65902405,
+      "epoch": 0.07204464738711314,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.2044647387113146e-06,
+      "loss": 1.59290199,
+      "memory(GiB)": 65.3,
+      "step": 2840,
+      "train_speed(iter/s)": 1.344115
+    },
+    {
+      "acc": 0.66350031,
+      "epoch": 0.0721714865550482,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.2171486555048205e-06,
+      "loss": 1.51631413,
+      "memory(GiB)": 65.3,
+      "step": 2845,
+      "train_speed(iter/s)": 1.344334
+    },
+    {
+      "acc": 0.65698433,
+      "epoch": 0.07229832572298325,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.2298325722983265e-06,
+      "loss": 1.52344294,
+      "memory(GiB)": 65.3,
+      "step": 2850,
+      "train_speed(iter/s)": 1.344559
+    },
+    {
+      "acc": 0.67196779,
+      "epoch": 0.07242516489091831,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.242516489091832e-06,
+      "loss": 1.56369267,
+      "memory(GiB)": 65.3,
+      "step": 2855,
+      "train_speed(iter/s)": 1.344783
+    },
+    {
+      "acc": 0.66713853,
+      "epoch": 0.07255200405885337,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.2552004058853376e-06,
+      "loss": 1.52495785,
+      "memory(GiB)": 65.3,
+      "step": 2860,
+      "train_speed(iter/s)": 1.344999
+    },
+    {
+      "acc": 0.66688309,
+      "epoch": 0.07267884322678843,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.267884322678844e-06,
+      "loss": 1.6337162,
+      "memory(GiB)": 65.3,
+      "step": 2865,
+      "train_speed(iter/s)": 1.345215
+    },
+    {
+      "acc": 0.66266022,
+      "epoch": 0.07280568239472349,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.2805682394723495e-06,
+      "loss": 1.59557819,
+      "memory(GiB)": 65.3,
+      "step": 2870,
+      "train_speed(iter/s)": 1.34543
+    },
+    {
+      "acc": 0.65171471,
+      "epoch": 0.07293252156265854,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.2932521562658555e-06,
+      "loss": 1.5697504,
+      "memory(GiB)": 65.3,
+      "step": 2875,
+      "train_speed(iter/s)": 1.345646
+    },
+    {
+      "acc": 0.65115833,
+      "epoch": 0.0730593607305936,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.305936073059361e-06,
+      "loss": 1.54873371,
+      "memory(GiB)": 65.3,
+      "step": 2880,
+      "train_speed(iter/s)": 1.345863
+    },
+    {
+      "acc": 0.67409487,
+      "epoch": 0.07318619989852866,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.318619989852867e-06,
+      "loss": 1.56681042,
+      "memory(GiB)": 65.3,
+      "step": 2885,
+      "train_speed(iter/s)": 1.346076
+    },
+    {
+      "acc": 0.65286818,
+      "epoch": 0.07331303906646372,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.331303906646373e-06,
+      "loss": 1.55388889,
+      "memory(GiB)": 65.3,
+      "step": 2890,
+      "train_speed(iter/s)": 1.346294
+    },
+    {
+      "acc": 0.67122183,
+      "epoch": 0.07343987823439878,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.3439878234398785e-06,
+      "loss": 1.53680077,
+      "memory(GiB)": 65.3,
+      "step": 2895,
+      "train_speed(iter/s)": 1.346499
+    },
+    {
+      "acc": 0.64327116,
+      "epoch": 0.07356671740233384,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.3566717402333845e-06,
+      "loss": 1.69338226,
+      "memory(GiB)": 65.3,
+      "step": 2900,
+      "train_speed(iter/s)": 1.346707
+    },
+    {
+      "acc": 0.66030059,
+      "epoch": 0.0736935565702689,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.369355657026891e-06,
+      "loss": 1.56061726,
+      "memory(GiB)": 65.3,
+      "step": 2905,
+      "train_speed(iter/s)": 1.346926
+    },
+    {
+      "acc": 0.64346976,
+      "epoch": 0.07382039573820395,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.382039573820396e-06,
+      "loss": 1.67072868,
+      "memory(GiB)": 65.3,
+      "step": 2910,
+      "train_speed(iter/s)": 1.347144
+    },
+    {
+      "acc": 0.66429,
+      "epoch": 0.07394723490613901,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.394723490613902e-06,
+      "loss": 1.58394241,
+      "memory(GiB)": 65.3,
+      "step": 2915,
+      "train_speed(iter/s)": 1.347349
+    },
+    {
+      "acc": 0.65402145,
+      "epoch": 0.07407407407407407,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 1.61457596,
+      "memory(GiB)": 65.3,
+      "step": 2920,
+      "train_speed(iter/s)": 1.347554
+    },
+    {
+      "acc": 0.65261149,
+      "epoch": 0.07420091324200913,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.420091324200914e-06,
+      "loss": 1.60792179,
+      "memory(GiB)": 65.3,
+      "step": 2925,
+      "train_speed(iter/s)": 1.34777
+    },
+    {
+      "acc": 0.65551581,
+      "epoch": 0.07432775240994419,
+      "grad_norm": 4.125,
+      "learning_rate": 7.432775240994419e-06,
+      "loss": 1.6263834,
+      "memory(GiB)": 65.3,
+      "step": 2930,
+      "train_speed(iter/s)": 1.347976
+    },
+    {
+      "acc": 0.66847196,
+      "epoch": 0.07445459157787924,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.445459157787925e-06,
+      "loss": 1.4831953,
+      "memory(GiB)": 65.3,
+      "step": 2935,
+      "train_speed(iter/s)": 1.348187
+    },
+    {
+      "acc": 0.66482582,
+      "epoch": 0.0745814307458143,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.458143074581431e-06,
+      "loss": 1.53827705,
+      "memory(GiB)": 65.3,
+      "step": 2940,
+      "train_speed(iter/s)": 1.348382
+    },
+    {
+      "acc": 0.68546643,
+      "epoch": 0.07470826991374936,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.470826991374937e-06,
+      "loss": 1.46167736,
+      "memory(GiB)": 65.3,
+      "step": 2945,
+      "train_speed(iter/s)": 1.34859
+    },
+    {
+      "acc": 0.64873452,
+      "epoch": 0.07483510908168442,
+      "grad_norm": 4.96875,
+      "learning_rate": 7.483510908168443e-06,
+      "loss": 1.65704231,
+      "memory(GiB)": 65.3,
+      "step": 2950,
+      "train_speed(iter/s)": 1.348805
+    },
+    {
+      "acc": 0.67599297,
+      "epoch": 0.07496194824961948,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.496194824961948e-06,
+      "loss": 1.50733633,
+      "memory(GiB)": 65.3,
+      "step": 2955,
+      "train_speed(iter/s)": 1.348974
+    },
+    {
+      "acc": 0.66898355,
+      "epoch": 0.07508878741755454,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.508878741755454e-06,
+      "loss": 1.50898438,
+      "memory(GiB)": 65.3,
+      "step": 2960,
+      "train_speed(iter/s)": 1.349165
+    },
+    {
+      "acc": 0.66061306,
+      "epoch": 0.0752156265854896,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.521562658548961e-06,
+      "loss": 1.54237051,
+      "memory(GiB)": 65.3,
+      "step": 2965,
+      "train_speed(iter/s)": 1.349357
+    },
+    {
+      "acc": 0.66990566,
+      "epoch": 0.07534246575342465,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.534246575342466e-06,
+      "loss": 1.50994358,
+      "memory(GiB)": 65.3,
+      "step": 2970,
+      "train_speed(iter/s)": 1.349559
+    },
+    {
+      "acc": 0.67281275,
+      "epoch": 0.07546930492135971,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.546930492135972e-06,
+      "loss": 1.50099583,
+      "memory(GiB)": 65.3,
+      "step": 2975,
+      "train_speed(iter/s)": 1.349772
+    },
+    {
+      "acc": 0.6726862,
+      "epoch": 0.07559614408929477,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.559614408929477e-06,
+      "loss": 1.49847002,
+      "memory(GiB)": 65.3,
+      "step": 2980,
+      "train_speed(iter/s)": 1.349982
+    },
+    {
+      "acc": 0.65588055,
+      "epoch": 0.07572298325722983,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.572298325722984e-06,
+      "loss": 1.554632,
+      "memory(GiB)": 65.3,
+      "step": 2985,
+      "train_speed(iter/s)": 1.350175
+    },
+    {
+      "acc": 0.67545295,
+      "epoch": 0.07584982242516489,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.58498224251649e-06,
+      "loss": 1.52101507,
+      "memory(GiB)": 65.3,
+      "step": 2990,
+      "train_speed(iter/s)": 1.350371
+    },
+    {
+      "acc": 0.65174761,
+      "epoch": 0.07597666159309995,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.597666159309995e-06,
+      "loss": 1.57794533,
+      "memory(GiB)": 65.3,
+      "step": 2995,
+      "train_speed(iter/s)": 1.350566
+    },
+    {
+      "acc": 0.6619348,
+      "epoch": 0.076103500761035,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.610350076103501e-06,
+      "loss": 1.57010136,
+      "memory(GiB)": 65.3,
+      "step": 3000,
+      "train_speed(iter/s)": 1.350766
+    },
+    {
+      "epoch": 0.076103500761035,
+      "eval_acc": 0.6482461361116157,
+      "eval_loss": 1.5092693567276,
+      "eval_runtime": 69.9643,
+      "eval_samples_per_second": 91.046,
+      "eval_steps_per_second": 22.769,
+      "step": 3000
+    },
+    {
+      "acc": 0.67718191,
+      "epoch": 0.07623033992897006,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.623033992897007e-06,
+      "loss": 1.46876869,
+      "memory(GiB)": 65.3,
+      "step": 3005,
+      "train_speed(iter/s)": 1.302603
+    },
+    {
+      "acc": 0.66245594,
+      "epoch": 0.07635717909690512,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.635717909690512e-06,
+      "loss": 1.56401815,
+      "memory(GiB)": 65.3,
+      "step": 3010,
+      "train_speed(iter/s)": 1.302865
+    },
+    {
+      "acc": 0.66500835,
+      "epoch": 0.07648401826484018,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.648401826484018e-06,
+      "loss": 1.5301116,
+      "memory(GiB)": 65.3,
+      "step": 3015,
+      "train_speed(iter/s)": 1.303123
+    },
+    {
+      "acc": 0.67836361,
+      "epoch": 0.07661085743277524,
+      "grad_norm": 3.375,
+      "learning_rate": 7.661085743277524e-06,
+      "loss": 1.51409073,
+      "memory(GiB)": 65.3,
+      "step": 3020,
+      "train_speed(iter/s)": 1.303379
+    },
+    {
+      "acc": 0.66712055,
+      "epoch": 0.0767376966007103,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.67376966007103e-06,
+      "loss": 1.5500289,
+      "memory(GiB)": 65.3,
+      "step": 3025,
+      "train_speed(iter/s)": 1.303628
+    },
+    {
+      "acc": 0.66832862,
+      "epoch": 0.07686453576864535,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.686453576864536e-06,
+      "loss": 1.49035339,
+      "memory(GiB)": 65.3,
+      "step": 3030,
+      "train_speed(iter/s)": 1.303846
+    },
+    {
+      "acc": 0.65114079,
+      "epoch": 0.07699137493658041,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.699137493658042e-06,
+      "loss": 1.56710043,
+      "memory(GiB)": 65.3,
+      "step": 3035,
+      "train_speed(iter/s)": 1.30409
+    },
+    {
+      "acc": 0.66619115,
+      "epoch": 0.07711821410451547,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.711821410451548e-06,
+      "loss": 1.50997429,
+      "memory(GiB)": 65.3,
+      "step": 3040,
+      "train_speed(iter/s)": 1.304326
+    },
+    {
+      "acc": 0.65673313,
+      "epoch": 0.07724505327245053,
+      "grad_norm": 4.375,
+      "learning_rate": 7.724505327245054e-06,
+      "loss": 1.53991661,
+      "memory(GiB)": 65.3,
+      "step": 3045,
+      "train_speed(iter/s)": 1.30459
+    },
+    {
+      "acc": 0.66372089,
+      "epoch": 0.07737189244038559,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.73718924403856e-06,
+      "loss": 1.53505182,
+      "memory(GiB)": 65.3,
+      "step": 3050,
+      "train_speed(iter/s)": 1.304841
+    },
+    {
+      "acc": 0.66646423,
+      "epoch": 0.07749873160832065,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.749873160832066e-06,
+      "loss": 1.55615387,
+      "memory(GiB)": 65.3,
+      "step": 3055,
+      "train_speed(iter/s)": 1.305017
+    },
+    {
+      "acc": 0.66912131,
+      "epoch": 0.0776255707762557,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.76255707762557e-06,
+      "loss": 1.50954962,
+      "memory(GiB)": 65.3,
+      "step": 3060,
+      "train_speed(iter/s)": 1.305271
+    },
+    {
+      "acc": 0.67344694,
+      "epoch": 0.07775240994419076,
+      "grad_norm": 3.25,
+      "learning_rate": 7.775240994419078e-06,
+      "loss": 1.49735632,
+      "memory(GiB)": 65.3,
+      "step": 3065,
+      "train_speed(iter/s)": 1.305512
+    },
+    {
+      "acc": 0.66310639,
+      "epoch": 0.07787924911212582,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.787924911212584e-06,
+      "loss": 1.56331577,
+      "memory(GiB)": 65.3,
+      "step": 3070,
+      "train_speed(iter/s)": 1.305756
+    },
+    {
+      "acc": 0.66457691,
+      "epoch": 0.07800608828006088,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.800608828006088e-06,
+      "loss": 1.57486439,
+      "memory(GiB)": 65.3,
+      "step": 3075,
+      "train_speed(iter/s)": 1.306009
+    },
+    {
+      "acc": 0.66052728,
+      "epoch": 0.07813292744799594,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.813292744799594e-06,
+      "loss": 1.53125153,
+      "memory(GiB)": 65.3,
+      "step": 3080,
+      "train_speed(iter/s)": 1.306253
+    },
+    {
+      "acc": 0.65855999,
+      "epoch": 0.078259766615931,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.8259766615931e-06,
+      "loss": 1.50385323,
+      "memory(GiB)": 65.3,
+      "step": 3085,
+      "train_speed(iter/s)": 1.306457
+    },
+    {
+      "acc": 0.66015034,
+      "epoch": 0.07838660578386605,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.838660578386606e-06,
+      "loss": 1.60684624,
+      "memory(GiB)": 65.3,
+      "step": 3090,
+      "train_speed(iter/s)": 1.306705
+    },
+    {
+      "acc": 0.660923,
+      "epoch": 0.07851344495180111,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.851344495180112e-06,
+      "loss": 1.57269697,
+      "memory(GiB)": 65.3,
+      "step": 3095,
+      "train_speed(iter/s)": 1.306944
+    },
+    {
+      "acc": 0.65816965,
+      "epoch": 0.07864028411973617,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.864028411973618e-06,
+      "loss": 1.55334253,
+      "memory(GiB)": 65.3,
+      "step": 3100,
+      "train_speed(iter/s)": 1.307188
+    },
+    {
+      "acc": 0.66016974,
+      "epoch": 0.07876712328767123,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.876712328767124e-06,
+      "loss": 1.52711058,
+      "memory(GiB)": 65.3,
+      "step": 3105,
+      "train_speed(iter/s)": 1.307426
+    },
+    {
+      "acc": 0.65916953,
+      "epoch": 0.07889396245560629,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.88939624556063e-06,
+      "loss": 1.58501749,
+      "memory(GiB)": 65.3,
+      "step": 3110,
+      "train_speed(iter/s)": 1.307673
+    },
+    {
+      "acc": 0.67352681,
+      "epoch": 0.07902080162354135,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.902080162354136e-06,
+      "loss": 1.45718966,
+      "memory(GiB)": 65.3,
+      "step": 3115,
+      "train_speed(iter/s)": 1.307921
+    },
+    {
+      "acc": 0.68964229,
+      "epoch": 0.0791476407914764,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.91476407914764e-06,
+      "loss": 1.42193689,
+      "memory(GiB)": 65.3,
+      "step": 3120,
+      "train_speed(iter/s)": 1.308165
+    },
+    {
+      "acc": 0.67338662,
+      "epoch": 0.07927447995941146,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.927447995941148e-06,
+      "loss": 1.50761118,
+      "memory(GiB)": 65.3,
+      "step": 3125,
+      "train_speed(iter/s)": 1.308414
+    },
+    {
+      "acc": 0.65976334,
+      "epoch": 0.07940131912734652,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.940131912734654e-06,
+      "loss": 1.56206245,
+      "memory(GiB)": 65.3,
+      "step": 3130,
+      "train_speed(iter/s)": 1.308655
+    },
+    {
+      "acc": 0.6809823,
+      "epoch": 0.07952815829528158,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.952815829528158e-06,
+      "loss": 1.47105618,
+      "memory(GiB)": 65.3,
+      "step": 3135,
+      "train_speed(iter/s)": 1.308901
+    },
+    {
+      "acc": 0.66550741,
+      "epoch": 0.07965499746321664,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.965499746321664e-06,
+      "loss": 1.57180796,
+      "memory(GiB)": 65.3,
+      "step": 3140,
+      "train_speed(iter/s)": 1.309139
+    },
+    {
+      "acc": 0.65534344,
+      "epoch": 0.0797818366311517,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.978183663115172e-06,
+      "loss": 1.63891582,
+      "memory(GiB)": 65.3,
+      "step": 3145,
+      "train_speed(iter/s)": 1.309381
+    },
+    {
+      "acc": 0.66317525,
+      "epoch": 0.07990867579908675,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.990867579908676e-06,
+      "loss": 1.55719013,
+      "memory(GiB)": 65.3,
+      "step": 3150,
+      "train_speed(iter/s)": 1.309612
+    },
+    {
+      "acc": 0.65544844,
+      "epoch": 0.08003551496702181,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.003551496702182e-06,
+      "loss": 1.58515301,
+      "memory(GiB)": 65.3,
+      "step": 3155,
+      "train_speed(iter/s)": 1.309849
+    },
+    {
+      "acc": 0.66259995,
+      "epoch": 0.08016235413495687,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.016235413495688e-06,
+      "loss": 1.53917866,
+      "memory(GiB)": 65.3,
+      "step": 3160,
+      "train_speed(iter/s)": 1.310091
+    },
+    {
+      "acc": 0.6647809,
+      "epoch": 0.08028919330289193,
+      "grad_norm": 4.0,
+      "learning_rate": 8.028919330289194e-06,
+      "loss": 1.62228622,
+      "memory(GiB)": 65.3,
+      "step": 3165,
+      "train_speed(iter/s)": 1.310312
+    },
+    {
+      "acc": 0.67679825,
+      "epoch": 0.08041603247082699,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.0416032470827e-06,
+      "loss": 1.51662216,
+      "memory(GiB)": 65.3,
+      "step": 3170,
+      "train_speed(iter/s)": 1.310547
+    },
+    {
+      "acc": 0.65416427,
+      "epoch": 0.08054287163876205,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.054287163876206e-06,
+      "loss": 1.56458702,
+      "memory(GiB)": 65.3,
+      "step": 3175,
+      "train_speed(iter/s)": 1.310778
+    },
+    {
+      "acc": 0.66479406,
+      "epoch": 0.0806697108066971,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.066971080669712e-06,
+      "loss": 1.52299728,
+      "memory(GiB)": 65.3,
+      "step": 3180,
+      "train_speed(iter/s)": 1.311017
+    },
+    {
+      "acc": 0.6505208,
+      "epoch": 0.08079654997463216,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.079654997463218e-06,
+      "loss": 1.62747574,
+      "memory(GiB)": 65.3,
+      "step": 3185,
+      "train_speed(iter/s)": 1.311252
+    },
+    {
+      "acc": 0.66148396,
+      "epoch": 0.08092338914256722,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.092338914256724e-06,
+      "loss": 1.50477543,
+      "memory(GiB)": 65.3,
+      "step": 3190,
+      "train_speed(iter/s)": 1.311491
+    },
+    {
+      "acc": 0.65674348,
+      "epoch": 0.08105022831050228,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.105022831050228e-06,
+      "loss": 1.50485544,
+      "memory(GiB)": 65.3,
+      "step": 3195,
+      "train_speed(iter/s)": 1.311723
+    },
+    {
+      "acc": 0.67173543,
+      "epoch": 0.08117706747843734,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.117706747843734e-06,
+      "loss": 1.50813217,
+      "memory(GiB)": 65.3,
+      "step": 3200,
+      "train_speed(iter/s)": 1.311941
+    },
+    {
+      "acc": 0.65597725,
+      "epoch": 0.0813039066463724,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.130390664637242e-06,
+      "loss": 1.52000256,
+      "memory(GiB)": 65.3,
+      "step": 3205,
+      "train_speed(iter/s)": 1.312165
+    },
+    {
+      "acc": 0.67485495,
+      "epoch": 0.08143074581430745,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.143074581430746e-06,
+      "loss": 1.49169874,
+      "memory(GiB)": 65.3,
+      "step": 3210,
+      "train_speed(iter/s)": 1.312395
+    },
+    {
+      "acc": 0.67648191,
+      "epoch": 0.08155758498224251,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.155758498224252e-06,
+      "loss": 1.52982292,
+      "memory(GiB)": 65.3,
+      "step": 3215,
+      "train_speed(iter/s)": 1.312634
+    },
+    {
+      "acc": 0.67186146,
+      "epoch": 0.08168442415017757,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.168442415017758e-06,
+      "loss": 1.51150417,
+      "memory(GiB)": 65.3,
+      "step": 3220,
+      "train_speed(iter/s)": 1.312858
+    },
+    {
+      "acc": 0.65313845,
+      "epoch": 0.08181126331811263,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.181126331811264e-06,
+      "loss": 1.57638979,
+      "memory(GiB)": 65.3,
+      "step": 3225,
+      "train_speed(iter/s)": 1.313093
+    },
+    {
+      "acc": 0.66062784,
+      "epoch": 0.08193810248604769,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.19381024860477e-06,
+      "loss": 1.48826981,
+      "memory(GiB)": 65.3,
+      "step": 3230,
+      "train_speed(iter/s)": 1.313335
+    },
+    {
+      "acc": 0.66614485,
+      "epoch": 0.08206494165398275,
+      "grad_norm": 3.0,
+      "learning_rate": 8.206494165398276e-06,
+      "loss": 1.44442101,
+      "memory(GiB)": 65.3,
+      "step": 3235,
+      "train_speed(iter/s)": 1.313575
+    },
+    {
+      "acc": 0.66042957,
+      "epoch": 0.0821917808219178,
+      "grad_norm": 3.5,
+      "learning_rate": 8.219178082191782e-06,
+      "loss": 1.53278618,
+      "memory(GiB)": 65.3,
+      "step": 3240,
+      "train_speed(iter/s)": 1.313713
+    },
+    {
+      "acc": 0.66526575,
+      "epoch": 0.08231861998985286,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.231861998985288e-06,
+      "loss": 1.51396379,
+      "memory(GiB)": 65.3,
+      "step": 3245,
+      "train_speed(iter/s)": 1.313946
+    },
+    {
+      "acc": 0.68036833,
+      "epoch": 0.08244545915778792,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.244545915778794e-06,
+      "loss": 1.44050837,
+      "memory(GiB)": 65.3,
+      "step": 3250,
+      "train_speed(iter/s)": 1.314177
+    },
+    {
+      "acc": 0.6551137,
+      "epoch": 0.08257229832572298,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.2572298325723e-06,
+      "loss": 1.54863358,
+      "memory(GiB)": 65.3,
+      "step": 3255,
+      "train_speed(iter/s)": 1.314409
+    },
+    {
+      "acc": 0.66978264,
+      "epoch": 0.08269913749365804,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.269913749365804e-06,
+      "loss": 1.47577858,
+      "memory(GiB)": 65.3,
+      "step": 3260,
+      "train_speed(iter/s)": 1.31464
+    },
+    {
+      "acc": 0.66793442,
+      "epoch": 0.0828259766615931,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.282597666159312e-06,
+      "loss": 1.500494,
+      "memory(GiB)": 65.3,
+      "step": 3265,
+      "train_speed(iter/s)": 1.314873
+    },
+    {
+      "acc": 0.65391178,
+      "epoch": 0.08295281582952815,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.295281582952816e-06,
+      "loss": 1.565767,
+      "memory(GiB)": 65.3,
+      "step": 3270,
+      "train_speed(iter/s)": 1.315099
+    },
+    {
+      "acc": 0.6615808,
+      "epoch": 0.08307965499746321,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.307965499746322e-06,
+      "loss": 1.51421814,
+      "memory(GiB)": 65.3,
+      "step": 3275,
+      "train_speed(iter/s)": 1.315279
+    },
+    {
+      "acc": 0.65394034,
+      "epoch": 0.08320649416539827,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.320649416539828e-06,
+      "loss": 1.62124329,
+      "memory(GiB)": 65.3,
+      "step": 3280,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.67140474,
+      "epoch": 0.08333333333333333,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.44480553,
+      "memory(GiB)": 65.3,
+      "step": 3285,
+      "train_speed(iter/s)": 1.315737
+    },
+    {
+      "acc": 0.66942234,
+      "epoch": 0.08346017250126839,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.34601725012684e-06,
+      "loss": 1.43203421,
+      "memory(GiB)": 65.3,
+      "step": 3290,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.66832962,
+      "epoch": 0.08358701166920345,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.358701166920346e-06,
+      "loss": 1.55665979,
+      "memory(GiB)": 65.3,
+      "step": 3295,
+      "train_speed(iter/s)": 1.316202
+    },
+    {
+      "acc": 0.64665184,
+      "epoch": 0.0837138508371385,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.371385083713852e-06,
+      "loss": 1.62825584,
+      "memory(GiB)": 65.3,
+      "step": 3300,
+      "train_speed(iter/s)": 1.316426
+    },
+    {
+      "acc": 0.65611477,
+      "epoch": 0.08384069000507356,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.384069000507358e-06,
+      "loss": 1.57160406,
+      "memory(GiB)": 65.3,
+      "step": 3305,
+      "train_speed(iter/s)": 1.316656
+    },
+    {
+      "acc": 0.65356302,
+      "epoch": 0.08396752917300862,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.396752917300864e-06,
+      "loss": 1.56488991,
+      "memory(GiB)": 65.3,
+      "step": 3310,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.65649614,
+      "epoch": 0.08409436834094368,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.40943683409437e-06,
+      "loss": 1.51842918,
+      "memory(GiB)": 65.3,
+      "step": 3315,
+      "train_speed(iter/s)": 1.317115
+    },
+    {
+      "acc": 0.66174383,
+      "epoch": 0.08422120750887874,
+      "grad_norm": 3.25,
+      "learning_rate": 8.422120750887874e-06,
+      "loss": 1.55940685,
+      "memory(GiB)": 65.3,
+      "step": 3320,
+      "train_speed(iter/s)": 1.317299
+    },
+    {
+      "acc": 0.6707531,
+      "epoch": 0.0843480466768138,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.434804667681381e-06,
+      "loss": 1.49227934,
+      "memory(GiB)": 65.3,
+      "step": 3325,
+      "train_speed(iter/s)": 1.317528
+    },
+    {
+      "acc": 0.66107769,
+      "epoch": 0.08447488584474885,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.447488584474887e-06,
+      "loss": 1.54295139,
+      "memory(GiB)": 65.3,
+      "step": 3330,
+      "train_speed(iter/s)": 1.317742
+    },
+    {
+      "acc": 0.66617174,
+      "epoch": 0.08460172501268391,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.460172501268392e-06,
+      "loss": 1.54569511,
+      "memory(GiB)": 65.3,
+      "step": 3335,
+      "train_speed(iter/s)": 1.317959
+    },
+    {
+      "acc": 0.66476665,
+      "epoch": 0.08472856418061897,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.472856418061898e-06,
+      "loss": 1.5107707,
+      "memory(GiB)": 65.3,
+      "step": 3340,
+      "train_speed(iter/s)": 1.318177
+    },
+    {
+      "acc": 0.68265266,
+      "epoch": 0.08485540334855403,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.485540334855404e-06,
+      "loss": 1.47964668,
+      "memory(GiB)": 65.3,
+      "step": 3345,
+      "train_speed(iter/s)": 1.318353
+    },
+    {
+      "acc": 0.68146586,
+      "epoch": 0.08498224251648909,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.49822425164891e-06,
+      "loss": 1.46113787,
+      "memory(GiB)": 65.3,
+      "step": 3350,
+      "train_speed(iter/s)": 1.318578
+    },
+    {
+      "acc": 0.65920849,
+      "epoch": 0.08510908168442415,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.510908168442416e-06,
+      "loss": 1.54507923,
+      "memory(GiB)": 65.3,
+      "step": 3355,
+      "train_speed(iter/s)": 1.318786
+    },
+    {
+      "acc": 0.65395689,
+      "epoch": 0.0852359208523592,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.523592085235922e-06,
+      "loss": 1.59547195,
+      "memory(GiB)": 65.3,
+      "step": 3360,
+      "train_speed(iter/s)": 1.319002
+    },
+    {
+      "acc": 0.65166888,
+      "epoch": 0.08536276002029426,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.536276002029428e-06,
+      "loss": 1.51567192,
+      "memory(GiB)": 65.3,
+      "step": 3365,
+      "train_speed(iter/s)": 1.319215
+    },
+    {
+      "acc": 0.66460829,
+      "epoch": 0.08548959918822932,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.548959918822933e-06,
+      "loss": 1.59184971,
+      "memory(GiB)": 65.3,
+      "step": 3370,
+      "train_speed(iter/s)": 1.319429
+    },
+    {
+      "acc": 0.65875635,
+      "epoch": 0.08561643835616438,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.56164383561644e-06,
+      "loss": 1.55908957,
+      "memory(GiB)": 65.3,
+      "step": 3375,
+      "train_speed(iter/s)": 1.31964
+    },
+    {
+      "acc": 0.67906437,
+      "epoch": 0.08574327752409944,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.574327752409944e-06,
+      "loss": 1.48357334,
+      "memory(GiB)": 65.3,
+      "step": 3380,
+      "train_speed(iter/s)": 1.319848
+    },
+    {
+      "acc": 0.65471849,
+      "epoch": 0.0858701166920345,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.587011669203451e-06,
+      "loss": 1.53287926,
+      "memory(GiB)": 65.3,
+      "step": 3385,
+      "train_speed(iter/s)": 1.32006
+    },
+    {
+      "acc": 0.68565421,
+      "epoch": 0.08599695585996955,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.599695585996957e-06,
+      "loss": 1.43632488,
+      "memory(GiB)": 65.3,
+      "step": 3390,
+      "train_speed(iter/s)": 1.320281
+    },
+    {
+      "acc": 0.65494289,
+      "epoch": 0.08612379502790461,
+      "grad_norm": 3.5,
+      "learning_rate": 8.612379502790462e-06,
+      "loss": 1.58551083,
+      "memory(GiB)": 65.3,
+      "step": 3395,
+      "train_speed(iter/s)": 1.320499
+    },
+    {
+      "acc": 0.66714568,
+      "epoch": 0.08625063419583967,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.625063419583968e-06,
+      "loss": 1.53723717,
+      "memory(GiB)": 65.3,
+      "step": 3400,
+      "train_speed(iter/s)": 1.320718
+    },
+    {
+      "acc": 0.67122126,
+      "epoch": 0.08637747336377473,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.637747336377475e-06,
+      "loss": 1.54823627,
+      "memory(GiB)": 65.3,
+      "step": 3405,
+      "train_speed(iter/s)": 1.320934
+    },
+    {
+      "acc": 0.66668468,
+      "epoch": 0.08650431253170979,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.65043125317098e-06,
+      "loss": 1.55419693,
+      "memory(GiB)": 65.3,
+      "step": 3410,
+      "train_speed(iter/s)": 1.321157
+    },
+    {
+      "acc": 0.65261588,
+      "epoch": 0.08663115169964485,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.663115169964485e-06,
+      "loss": 1.49662304,
+      "memory(GiB)": 65.3,
+      "step": 3415,
+      "train_speed(iter/s)": 1.321383
+    },
+    {
+      "acc": 0.63786554,
+      "epoch": 0.0867579908675799,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.675799086757991e-06,
+      "loss": 1.60366993,
+      "memory(GiB)": 65.3,
+      "step": 3420,
+      "train_speed(iter/s)": 1.321604
+    },
+    {
+      "acc": 0.66515427,
+      "epoch": 0.08688483003551496,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.688483003551497e-06,
+      "loss": 1.57847519,
+      "memory(GiB)": 65.3,
+      "step": 3425,
+      "train_speed(iter/s)": 1.321814
+    },
+    {
+      "acc": 0.67251825,
+      "epoch": 0.08701166920345002,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.701166920345003e-06,
+      "loss": 1.51860209,
+      "memory(GiB)": 65.3,
+      "step": 3430,
+      "train_speed(iter/s)": 1.32202
+    },
+    {
+      "acc": 0.66646852,
+      "epoch": 0.08713850837138508,
+      "grad_norm": 3.125,
+      "learning_rate": 8.71385083713851e-06,
+      "loss": 1.55235415,
+      "memory(GiB)": 74.41,
+      "step": 3435,
+      "train_speed(iter/s)": 1.322193
+    },
+    {
+      "acc": 0.65193567,
+      "epoch": 0.08726534753932014,
+      "grad_norm": 3.25,
+      "learning_rate": 8.726534753932014e-06,
+      "loss": 1.64208374,
+      "memory(GiB)": 74.41,
+      "step": 3440,
+      "train_speed(iter/s)": 1.322418
+    },
+    {
+      "acc": 0.66331654,
+      "epoch": 0.0873921867072552,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.739218670725521e-06,
+      "loss": 1.50640383,
+      "memory(GiB)": 74.41,
+      "step": 3445,
+      "train_speed(iter/s)": 1.32262
+    },
+    {
+      "acc": 0.66524172,
+      "epoch": 0.08751902587519025,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.751902587519027e-06,
+      "loss": 1.48821526,
+      "memory(GiB)": 74.41,
+      "step": 3450,
+      "train_speed(iter/s)": 1.322826
+    },
+    {
+      "acc": 0.66061969,
+      "epoch": 0.08764586504312531,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.764586504312532e-06,
+      "loss": 1.60944424,
+      "memory(GiB)": 74.41,
+      "step": 3455,
+      "train_speed(iter/s)": 1.323039
+    },
+    {
+      "acc": 0.67344332,
+      "epoch": 0.08777270421106037,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.777270421106037e-06,
+      "loss": 1.57980337,
+      "memory(GiB)": 74.41,
+      "step": 3460,
+      "train_speed(iter/s)": 1.32324
+    },
+    {
+      "acc": 0.67432103,
+      "epoch": 0.08789954337899543,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.789954337899545e-06,
+      "loss": 1.49575481,
+      "memory(GiB)": 74.41,
+      "step": 3465,
+      "train_speed(iter/s)": 1.32345
+    },
+    {
+      "acc": 0.66299639,
+      "epoch": 0.08802638254693049,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.80263825469305e-06,
+      "loss": 1.52421551,
+      "memory(GiB)": 74.41,
+      "step": 3470,
+      "train_speed(iter/s)": 1.323655
+    },
+    {
+      "acc": 0.67513227,
+      "epoch": 0.08815322171486555,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.815322171486555e-06,
+      "loss": 1.45028915,
+      "memory(GiB)": 74.41,
+      "step": 3475,
+      "train_speed(iter/s)": 1.323862
+    },
+    {
+      "acc": 0.67439537,
+      "epoch": 0.0882800608828006,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.828006088280061e-06,
+      "loss": 1.529778,
+      "memory(GiB)": 74.41,
+      "step": 3480,
+      "train_speed(iter/s)": 1.324068
+    },
+    {
+      "acc": 0.66506815,
+      "epoch": 0.08840690005073566,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.840690005073567e-06,
+      "loss": 1.51793194,
+      "memory(GiB)": 74.41,
+      "step": 3485,
+      "train_speed(iter/s)": 1.324279
+    },
+    {
+      "acc": 0.65643959,
+      "epoch": 0.08853373921867072,
+      "grad_norm": 3.125,
+      "learning_rate": 8.853373921867073e-06,
+      "loss": 1.49946203,
+      "memory(GiB)": 74.41,
+      "step": 3490,
+      "train_speed(iter/s)": 1.324488
+    },
+    {
+      "acc": 0.67318249,
+      "epoch": 0.08866057838660578,
+      "grad_norm": 2.875,
+      "learning_rate": 8.86605783866058e-06,
+      "loss": 1.54618835,
+      "memory(GiB)": 74.41,
+      "step": 3495,
+      "train_speed(iter/s)": 1.324698
+    },
+    {
+      "acc": 0.67332439,
+      "epoch": 0.08878741755454084,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.878741755454085e-06,
+      "loss": 1.46800117,
+      "memory(GiB)": 74.41,
+      "step": 3500,
+      "train_speed(iter/s)": 1.324899
+    },
+    {
+      "acc": 0.66910982,
+      "epoch": 0.0889142567224759,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.891425672247591e-06,
+      "loss": 1.47805748,
+      "memory(GiB)": 74.41,
+      "step": 3505,
+      "train_speed(iter/s)": 1.325094
+    },
+    {
+      "acc": 0.6641326,
+      "epoch": 0.08904109589041095,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.904109589041097e-06,
+      "loss": 1.53107672,
+      "memory(GiB)": 74.41,
+      "step": 3510,
+      "train_speed(iter/s)": 1.325301
+    },
+    {
+      "acc": 0.65677257,
+      "epoch": 0.08916793505834601,
+      "grad_norm": 4.78125,
+      "learning_rate": 8.916793505834601e-06,
+      "loss": 1.57765217,
+      "memory(GiB)": 74.41,
+      "step": 3515,
+      "train_speed(iter/s)": 1.325506
+    },
+    {
+      "acc": 0.66284585,
+      "epoch": 0.08929477422628107,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.929477422628107e-06,
+      "loss": 1.5355175,
+      "memory(GiB)": 74.41,
+      "step": 3520,
+      "train_speed(iter/s)": 1.325707
+    },
+    {
+      "acc": 0.66858358,
+      "epoch": 0.08942161339421613,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.942161339421615e-06,
+      "loss": 1.53078232,
+      "memory(GiB)": 74.41,
+      "step": 3525,
+      "train_speed(iter/s)": 1.325914
+    },
+    {
+      "acc": 0.66917782,
+      "epoch": 0.08954845256215119,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.95484525621512e-06,
+      "loss": 1.51279097,
+      "memory(GiB)": 74.41,
+      "step": 3530,
+      "train_speed(iter/s)": 1.326117
+    },
+    {
+      "acc": 0.65159645,
+      "epoch": 0.08967529173008625,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.967529173008625e-06,
+      "loss": 1.53735695,
+      "memory(GiB)": 74.41,
+      "step": 3535,
+      "train_speed(iter/s)": 1.326322
+    },
+    {
+      "acc": 0.66727839,
+      "epoch": 0.0898021308980213,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.980213089802131e-06,
+      "loss": 1.46436453,
+      "memory(GiB)": 74.41,
+      "step": 3540,
+      "train_speed(iter/s)": 1.326526
+    },
+    {
+      "acc": 0.67384758,
+      "epoch": 0.08992897006595636,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.992897006595637e-06,
+      "loss": 1.45831451,
+      "memory(GiB)": 74.41,
+      "step": 3545,
+      "train_speed(iter/s)": 1.32672
+    },
+    {
+      "acc": 0.66437664,
+      "epoch": 0.09005580923389142,
+      "grad_norm": 3.25,
+      "learning_rate": 9.005580923389143e-06,
+      "loss": 1.48599138,
+      "memory(GiB)": 74.41,
+      "step": 3550,
+      "train_speed(iter/s)": 1.326917
+    },
+    {
+      "acc": 0.64966602,
+      "epoch": 0.09018264840182648,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.01826484018265e-06,
+      "loss": 1.55634804,
+      "memory(GiB)": 74.41,
+      "step": 3555,
+      "train_speed(iter/s)": 1.327121
+    },
+    {
+      "acc": 0.65081644,
+      "epoch": 0.09030948756976154,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.030948756976155e-06,
+      "loss": 1.60264187,
+      "memory(GiB)": 74.41,
+      "step": 3560,
+      "train_speed(iter/s)": 1.327315
+    },
+    {
+      "acc": 0.66532011,
+      "epoch": 0.0904363267376966,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.043632673769661e-06,
+      "loss": 1.54236355,
+      "memory(GiB)": 74.41,
+      "step": 3565,
+      "train_speed(iter/s)": 1.327513
+    },
+    {
+      "acc": 0.66407948,
+      "epoch": 0.09056316590563165,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.056316590563167e-06,
+      "loss": 1.47736444,
+      "memory(GiB)": 74.41,
+      "step": 3570,
+      "train_speed(iter/s)": 1.327706
+    },
+    {
+      "acc": 0.66446366,
+      "epoch": 0.09069000507356671,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.069000507356673e-06,
+      "loss": 1.47629395,
+      "memory(GiB)": 74.41,
+      "step": 3575,
+      "train_speed(iter/s)": 1.327901
+    },
+    {
+      "acc": 0.67155051,
+      "epoch": 0.09081684424150177,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.081684424150177e-06,
+      "loss": 1.51526012,
+      "memory(GiB)": 74.41,
+      "step": 3580,
+      "train_speed(iter/s)": 1.328109
+    },
+    {
+      "acc": 0.66694584,
+      "epoch": 0.09094368340943683,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.094368340943685e-06,
+      "loss": 1.4509306,
+      "memory(GiB)": 74.41,
+      "step": 3585,
+      "train_speed(iter/s)": 1.328299
+    },
+    {
+      "acc": 0.67407446,
+      "epoch": 0.09107052257737189,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.10705225773719e-06,
+      "loss": 1.43186512,
+      "memory(GiB)": 74.41,
+      "step": 3590,
+      "train_speed(iter/s)": 1.328494
+    },
+    {
+      "acc": 0.6653615,
+      "epoch": 0.09119736174530695,
+      "grad_norm": 3.25,
+      "learning_rate": 9.119736174530695e-06,
+      "loss": 1.56705227,
+      "memory(GiB)": 74.41,
+      "step": 3595,
+      "train_speed(iter/s)": 1.328694
+    },
+    {
+      "acc": 0.66768484,
+      "epoch": 0.091324200913242,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.132420091324201e-06,
+      "loss": 1.52656651,
+      "memory(GiB)": 74.41,
+      "step": 3600,
+      "train_speed(iter/s)": 1.328883
+    },
+    {
+      "acc": 0.68541799,
+      "epoch": 0.09145104008117706,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.145104008117707e-06,
+      "loss": 1.47243023,
+      "memory(GiB)": 74.41,
+      "step": 3605,
+      "train_speed(iter/s)": 1.329051
+    },
+    {
+      "acc": 0.65506544,
+      "epoch": 0.09157787924911212,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.157787924911213e-06,
+      "loss": 1.58556414,
+      "memory(GiB)": 74.41,
+      "step": 3610,
+      "train_speed(iter/s)": 1.32924
+    },
+    {
+      "acc": 0.66567764,
+      "epoch": 0.09170471841704718,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.170471841704719e-06,
+      "loss": 1.56283741,
+      "memory(GiB)": 74.41,
+      "step": 3615,
+      "train_speed(iter/s)": 1.32943
+    },
+    {
+      "acc": 0.65707932,
+      "epoch": 0.09183155758498224,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.183155758498225e-06,
+      "loss": 1.54942493,
+      "memory(GiB)": 74.41,
+      "step": 3620,
+      "train_speed(iter/s)": 1.329612
+    },
+    {
+      "acc": 0.65892382,
+      "epoch": 0.0919583967529173,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.195839675291731e-06,
+      "loss": 1.57049046,
+      "memory(GiB)": 74.41,
+      "step": 3625,
+      "train_speed(iter/s)": 1.329804
+    },
+    {
+      "acc": 0.66895761,
+      "epoch": 0.09208523592085235,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.208523592085237e-06,
+      "loss": 1.50717945,
+      "memory(GiB)": 74.41,
+      "step": 3630,
+      "train_speed(iter/s)": 1.329996
+    },
+    {
+      "acc": 0.67409964,
+      "epoch": 0.09221207508878741,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.221207508878743e-06,
+      "loss": 1.45556412,
+      "memory(GiB)": 74.41,
+      "step": 3635,
+      "train_speed(iter/s)": 1.330183
+    },
+    {
+      "acc": 0.66178808,
+      "epoch": 0.09233891425672247,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.233891425672247e-06,
+      "loss": 1.56270533,
+      "memory(GiB)": 74.41,
+      "step": 3640,
+      "train_speed(iter/s)": 1.330377
+    },
+    {
+      "acc": 0.66897726,
+      "epoch": 0.09246575342465753,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.246575342465755e-06,
+      "loss": 1.54498138,
+      "memory(GiB)": 74.41,
+      "step": 3645,
+      "train_speed(iter/s)": 1.330565
+    },
+    {
+      "acc": 0.66894259,
+      "epoch": 0.09259259259259259,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.25925925925926e-06,
+      "loss": 1.53167191,
+      "memory(GiB)": 74.41,
+      "step": 3650,
+      "train_speed(iter/s)": 1.330729
+    },
+    {
+      "acc": 0.66786261,
+      "epoch": 0.09271943176052765,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.271943176052765e-06,
+      "loss": 1.47729168,
+      "memory(GiB)": 74.41,
+      "step": 3655,
+      "train_speed(iter/s)": 1.330917
+    },
+    {
+      "acc": 0.68489676,
+      "epoch": 0.0928462709284627,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.284627092846271e-06,
+      "loss": 1.45590048,
+      "memory(GiB)": 74.41,
+      "step": 3660,
+      "train_speed(iter/s)": 1.331111
+    },
+    {
+      "acc": 0.6770668,
+      "epoch": 0.09297311009639776,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.297311009639777e-06,
+      "loss": 1.53475056,
+      "memory(GiB)": 74.41,
+      "step": 3665,
+      "train_speed(iter/s)": 1.331298
+    },
+    {
+      "acc": 0.66462941,
+      "epoch": 0.09309994926433282,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.309994926433283e-06,
+      "loss": 1.57983971,
+      "memory(GiB)": 74.41,
+      "step": 3670,
+      "train_speed(iter/s)": 1.331489
+    },
+    {
+      "acc": 0.65053821,
+      "epoch": 0.09322678843226788,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.322678843226789e-06,
+      "loss": 1.60091743,
+      "memory(GiB)": 74.41,
+      "step": 3675,
+      "train_speed(iter/s)": 1.331677
+    },
+    {
+      "acc": 0.66121783,
+      "epoch": 0.09335362760020294,
+      "grad_norm": 4.625,
+      "learning_rate": 9.335362760020295e-06,
+      "loss": 1.58782597,
+      "memory(GiB)": 74.41,
+      "step": 3680,
+      "train_speed(iter/s)": 1.331869
+    },
+    {
+      "acc": 0.66092768,
+      "epoch": 0.093480466768138,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.348046676813801e-06,
+      "loss": 1.51315746,
+      "memory(GiB)": 74.41,
+      "step": 3685,
+      "train_speed(iter/s)": 1.332056
+    },
+    {
+      "acc": 0.65990548,
+      "epoch": 0.09360730593607305,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.360730593607307e-06,
+      "loss": 1.55032921,
+      "memory(GiB)": 74.41,
+      "step": 3690,
+      "train_speed(iter/s)": 1.332234
+    },
+    {
+      "acc": 0.68456926,
+      "epoch": 0.09373414510400811,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.373414510400813e-06,
+      "loss": 1.42659454,
+      "memory(GiB)": 74.41,
+      "step": 3695,
+      "train_speed(iter/s)": 1.332413
+    },
+    {
+      "acc": 0.65050287,
+      "epoch": 0.09386098427194317,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.386098427194317e-06,
+      "loss": 1.61858654,
+      "memory(GiB)": 74.41,
+      "step": 3700,
+      "train_speed(iter/s)": 1.332597
+    },
+    {
+      "acc": 0.65403738,
+      "epoch": 0.09398782343987823,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.398782343987825e-06,
+      "loss": 1.61903915,
+      "memory(GiB)": 74.41,
+      "step": 3705,
+      "train_speed(iter/s)": 1.332773
+    },
+    {
+      "acc": 0.66778979,
+      "epoch": 0.09411466260781329,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.41146626078133e-06,
+      "loss": 1.46366386,
+      "memory(GiB)": 74.41,
+      "step": 3710,
+      "train_speed(iter/s)": 1.332937
+    },
+    {
+      "acc": 0.67651148,
+      "epoch": 0.09424150177574835,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.424150177574835e-06,
+      "loss": 1.50717621,
+      "memory(GiB)": 74.41,
+      "step": 3715,
+      "train_speed(iter/s)": 1.333113
+    },
+    {
+      "acc": 0.66236567,
+      "epoch": 0.0943683409436834,
+      "grad_norm": 4.0,
+      "learning_rate": 9.436834094368341e-06,
+      "loss": 1.50785751,
+      "memory(GiB)": 74.41,
+      "step": 3720,
+      "train_speed(iter/s)": 1.333281
+    },
+    {
+      "acc": 0.68052611,
+      "epoch": 0.09449518011161846,
+      "grad_norm": 4.125,
+      "learning_rate": 9.449518011161849e-06,
+      "loss": 1.46900444,
+      "memory(GiB)": 74.41,
+      "step": 3725,
+      "train_speed(iter/s)": 1.333449
+    },
+    {
+      "acc": 0.68314638,
+      "epoch": 0.09462201927955352,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.462201927955353e-06,
+      "loss": 1.43477764,
+      "memory(GiB)": 74.41,
+      "step": 3730,
+      "train_speed(iter/s)": 1.333631
+    },
+    {
+      "acc": 0.65517035,
+      "epoch": 0.09474885844748858,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.474885844748859e-06,
+      "loss": 1.5454443,
+      "memory(GiB)": 74.41,
+      "step": 3735,
+      "train_speed(iter/s)": 1.333807
+    },
+    {
+      "acc": 0.66310425,
+      "epoch": 0.09487569761542364,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.487569761542365e-06,
+      "loss": 1.49432602,
+      "memory(GiB)": 74.41,
+      "step": 3740,
+      "train_speed(iter/s)": 1.333975
+    },
+    {
+      "acc": 0.67852273,
+      "epoch": 0.0950025367833587,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.50025367833587e-06,
+      "loss": 1.49595079,
+      "memory(GiB)": 74.41,
+      "step": 3745,
+      "train_speed(iter/s)": 1.334139
+    },
+    {
+      "acc": 0.66234112,
+      "epoch": 0.09512937595129375,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.512937595129377e-06,
+      "loss": 1.53565054,
+      "memory(GiB)": 74.41,
+      "step": 3750,
+      "train_speed(iter/s)": 1.334318
+    },
+    {
+      "acc": 0.6702517,
+      "epoch": 0.09525621511922881,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.525621511922883e-06,
+      "loss": 1.56963005,
+      "memory(GiB)": 74.41,
+      "step": 3755,
+      "train_speed(iter/s)": 1.334489
+    },
+    {
+      "acc": 0.65332537,
+      "epoch": 0.09538305428716387,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.538305428716389e-06,
+      "loss": 1.55208721,
+      "memory(GiB)": 74.41,
+      "step": 3760,
+      "train_speed(iter/s)": 1.33466
+    },
+    {
+      "acc": 0.66909952,
+      "epoch": 0.09550989345509893,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.550989345509895e-06,
+      "loss": 1.56708193,
+      "memory(GiB)": 74.41,
+      "step": 3765,
+      "train_speed(iter/s)": 1.334831
+    },
+    {
+      "acc": 0.66686006,
+      "epoch": 0.09563673262303399,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.5636732623034e-06,
+      "loss": 1.54983759,
+      "memory(GiB)": 74.41,
+      "step": 3770,
+      "train_speed(iter/s)": 1.335007
+    },
+    {
+      "acc": 0.67115707,
+      "epoch": 0.09576357179096905,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.576357179096905e-06,
+      "loss": 1.49225578,
+      "memory(GiB)": 74.41,
+      "step": 3775,
+      "train_speed(iter/s)": 1.335172
+    },
+    {
+      "acc": 0.65652399,
+      "epoch": 0.0958904109589041,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.589041095890411e-06,
+      "loss": 1.57779875,
+      "memory(GiB)": 74.41,
+      "step": 3780,
+      "train_speed(iter/s)": 1.335321
+    },
+    {
+      "acc": 0.64945059,
+      "epoch": 0.09601725012683916,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.601725012683919e-06,
+      "loss": 1.58271952,
+      "memory(GiB)": 74.41,
+      "step": 3785,
+      "train_speed(iter/s)": 1.335489
+    },
+    {
+      "acc": 0.66475425,
+      "epoch": 0.09614408929477422,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.614408929477423e-06,
+      "loss": 1.53661976,
+      "memory(GiB)": 74.41,
+      "step": 3790,
+      "train_speed(iter/s)": 1.335659
+    },
+    {
+      "acc": 0.65427465,
+      "epoch": 0.09627092846270928,
+      "grad_norm": 2.5,
+      "learning_rate": 9.627092846270929e-06,
+      "loss": 1.5804781,
+      "memory(GiB)": 74.41,
+      "step": 3795,
+      "train_speed(iter/s)": 1.335831
+    },
+    {
+      "acc": 0.65428152,
+      "epoch": 0.09639776763064434,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.639776763064435e-06,
+      "loss": 1.55026007,
+      "memory(GiB)": 74.41,
+      "step": 3800,
+      "train_speed(iter/s)": 1.336005
+    },
+    {
+      "acc": 0.66119089,
+      "epoch": 0.0965246067985794,
+      "grad_norm": 2.5,
+      "learning_rate": 9.65246067985794e-06,
+      "loss": 1.56023436,
+      "memory(GiB)": 74.41,
+      "step": 3805,
+      "train_speed(iter/s)": 1.336179
+    },
+    {
+      "acc": 0.66636024,
+      "epoch": 0.09665144596651445,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.665144596651447e-06,
+      "loss": 1.49019394,
+      "memory(GiB)": 74.41,
+      "step": 3810,
+      "train_speed(iter/s)": 1.336316
+    },
+    {
+      "acc": 0.63028722,
+      "epoch": 0.09677828513444951,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.677828513444953e-06,
+      "loss": 1.62354813,
+      "memory(GiB)": 74.41,
+      "step": 3815,
+      "train_speed(iter/s)": 1.336496
+    },
+    {
+      "acc": 0.6556798,
+      "epoch": 0.09690512430238457,
+      "grad_norm": 3.5,
+      "learning_rate": 9.690512430238459e-06,
+      "loss": 1.61269569,
+      "memory(GiB)": 74.41,
+      "step": 3820,
+      "train_speed(iter/s)": 1.336666
+    },
+    {
+      "acc": 0.68545685,
+      "epoch": 0.09703196347031963,
+      "grad_norm": 2.75,
+      "learning_rate": 9.703196347031965e-06,
+      "loss": 1.42435951,
+      "memory(GiB)": 74.41,
+      "step": 3825,
+      "train_speed(iter/s)": 1.336816
+    },
+    {
+      "acc": 0.68508005,
+      "epoch": 0.09715880263825469,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.71588026382547e-06,
+      "loss": 1.44473553,
+      "memory(GiB)": 74.41,
+      "step": 3830,
+      "train_speed(iter/s)": 1.336956
+    },
+    {
+      "acc": 0.65260491,
+      "epoch": 0.09728564180618975,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.728564180618977e-06,
+      "loss": 1.58057976,
+      "memory(GiB)": 74.41,
+      "step": 3835,
+      "train_speed(iter/s)": 1.337122
+    },
+    {
+      "acc": 0.65975704,
+      "epoch": 0.0974124809741248,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.74124809741248e-06,
+      "loss": 1.54687347,
+      "memory(GiB)": 74.41,
+      "step": 3840,
+      "train_speed(iter/s)": 1.337287
+    },
+    {
+      "acc": 0.66391306,
+      "epoch": 0.09753932014205986,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.753932014205988e-06,
+      "loss": 1.51929398,
+      "memory(GiB)": 74.41,
+      "step": 3845,
+      "train_speed(iter/s)": 1.337452
+    },
+    {
+      "acc": 0.67653332,
+      "epoch": 0.09766615930999492,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.766615930999493e-06,
+      "loss": 1.48864813,
+      "memory(GiB)": 74.41,
+      "step": 3850,
+      "train_speed(iter/s)": 1.33761
+    },
+    {
+      "acc": 0.67974267,
+      "epoch": 0.09779299847792998,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.779299847792999e-06,
+      "loss": 1.47870502,
+      "memory(GiB)": 74.41,
+      "step": 3855,
+      "train_speed(iter/s)": 1.337769
+    },
+    {
+      "acc": 0.67558603,
+      "epoch": 0.09791983764586504,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.791983764586505e-06,
+      "loss": 1.45302172,
+      "memory(GiB)": 74.41,
+      "step": 3860,
+      "train_speed(iter/s)": 1.337925
+    },
+    {
+      "acc": 0.66798468,
+      "epoch": 0.0980466768138001,
+      "grad_norm": 3.75,
+      "learning_rate": 9.80466768138001e-06,
+      "loss": 1.53495979,
+      "memory(GiB)": 74.41,
+      "step": 3865,
+      "train_speed(iter/s)": 1.338085
+    },
+    {
+      "acc": 0.65955973,
+      "epoch": 0.09817351598173515,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.817351598173517e-06,
+      "loss": 1.56391277,
+      "memory(GiB)": 74.41,
+      "step": 3870,
+      "train_speed(iter/s)": 1.338251
+    },
+    {
+      "acc": 0.66559258,
+      "epoch": 0.09830035514967021,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.830035514967023e-06,
+      "loss": 1.57412682,
+      "memory(GiB)": 74.41,
+      "step": 3875,
+      "train_speed(iter/s)": 1.338413
+    },
+    {
+      "acc": 0.65490689,
+      "epoch": 0.09842719431760527,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.842719431760529e-06,
+      "loss": 1.61827946,
+      "memory(GiB)": 74.41,
+      "step": 3880,
+      "train_speed(iter/s)": 1.338578
+    },
+    {
+      "acc": 0.66896324,
+      "epoch": 0.09855403348554033,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.855403348554034e-06,
+      "loss": 1.5008852,
+      "memory(GiB)": 74.41,
+      "step": 3885,
+      "train_speed(iter/s)": 1.338737
+    },
+    {
+      "acc": 0.64943862,
+      "epoch": 0.09868087265347539,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.86808726534754e-06,
+      "loss": 1.57165804,
+      "memory(GiB)": 74.41,
+      "step": 3890,
+      "train_speed(iter/s)": 1.338899
+    },
+    {
+      "acc": 0.66481986,
+      "epoch": 0.09880771182141045,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.880771182141046e-06,
+      "loss": 1.61635857,
+      "memory(GiB)": 74.41,
+      "step": 3895,
+      "train_speed(iter/s)": 1.339065
+    },
+    {
+      "acc": 0.67637162,
+      "epoch": 0.0989345509893455,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.89345509893455e-06,
+      "loss": 1.50531044,
+      "memory(GiB)": 74.41,
+      "step": 3900,
+      "train_speed(iter/s)": 1.339225
+    },
+    {
+      "acc": 0.6657362,
+      "epoch": 0.09906139015728056,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.906139015728058e-06,
+      "loss": 1.54008093,
+      "memory(GiB)": 74.41,
+      "step": 3905,
+      "train_speed(iter/s)": 1.339389
+    },
+    {
+      "acc": 0.65387039,
+      "epoch": 0.09918822932521562,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.918822932521563e-06,
+      "loss": 1.55647259,
+      "memory(GiB)": 74.41,
+      "step": 3910,
+      "train_speed(iter/s)": 1.339519
+    },
+    {
+      "acc": 0.66066661,
+      "epoch": 0.09931506849315068,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.931506849315069e-06,
+      "loss": 1.52518625,
+      "memory(GiB)": 74.41,
+      "step": 3915,
+      "train_speed(iter/s)": 1.33968
+    },
+    {
+      "acc": 0.65816932,
+      "epoch": 0.09944190766108574,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.944190766108575e-06,
+      "loss": 1.55633802,
+      "memory(GiB)": 74.41,
+      "step": 3920,
+      "train_speed(iter/s)": 1.339837
+    },
+    {
+      "acc": 0.66554718,
+      "epoch": 0.0995687468290208,
+      "grad_norm": 3.25,
+      "learning_rate": 9.95687468290208e-06,
+      "loss": 1.4444231,
+      "memory(GiB)": 74.41,
+      "step": 3925,
+      "train_speed(iter/s)": 1.339994
+    },
+    {
+      "acc": 0.66726665,
+      "epoch": 0.09969558599695585,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.969558599695586e-06,
+      "loss": 1.5267519,
+      "memory(GiB)": 74.41,
+      "step": 3930,
+      "train_speed(iter/s)": 1.340129
+    },
+    {
+      "acc": 0.66697912,
+      "epoch": 0.09982242516489091,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.982242516489092e-06,
+      "loss": 1.55427876,
+      "memory(GiB)": 74.41,
+      "step": 3935,
+      "train_speed(iter/s)": 1.34029
+    },
+    {
+      "acc": 0.67474923,
+      "epoch": 0.09994926433282597,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.994926433282598e-06,
+      "loss": 1.4655118,
+      "memory(GiB)": 74.41,
+      "step": 3940,
+      "train_speed(iter/s)": 1.340446
+    },
+    {
+      "acc": 0.68629117,
+      "epoch": 0.10007610350076103,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.999999960413982e-06,
+      "loss": 1.43725309,
+      "memory(GiB)": 74.41,
+      "step": 3945,
+      "train_speed(iter/s)": 1.340595
+    },
+    {
+      "acc": 0.66494503,
+      "epoch": 0.10020294266869609,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.99999971849943e-06,
+      "loss": 1.55422611,
+      "memory(GiB)": 74.41,
+      "step": 3950,
+      "train_speed(iter/s)": 1.340755
+    },
+    {
+      "acc": 0.67058187,
+      "epoch": 0.10032978183663115,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.999999256662563e-06,
+      "loss": 1.56513824,
+      "memory(GiB)": 74.41,
+      "step": 3955,
+      "train_speed(iter/s)": 1.340914
+    },
+    {
+      "acc": 0.68514066,
+      "epoch": 0.1004566210045662,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.999998574903408e-06,
+      "loss": 1.47113867,
+      "memory(GiB)": 74.41,
+      "step": 3960,
+      "train_speed(iter/s)": 1.341074
+    },
+    {
+      "acc": 0.6672142,
+      "epoch": 0.10058346017250126,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.99999767322199e-06,
+      "loss": 1.52289553,
+      "memory(GiB)": 74.41,
+      "step": 3965,
+      "train_speed(iter/s)": 1.341226
+    },
+    {
+      "acc": 0.66511259,
+      "epoch": 0.10071029934043632,
+      "grad_norm": 2.75,
+      "learning_rate": 9.999996551618353e-06,
+      "loss": 1.46662569,
+      "memory(GiB)": 74.41,
+      "step": 3970,
+      "train_speed(iter/s)": 1.341373
+    },
+    {
+      "acc": 0.65567312,
+      "epoch": 0.10083713850837138,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.999995210092545e-06,
+      "loss": 1.52560835,
+      "memory(GiB)": 74.41,
+      "step": 3975,
+      "train_speed(iter/s)": 1.34153
+    },
+    {
+      "acc": 0.65891199,
+      "epoch": 0.10096397767630644,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.999993648644622e-06,
+      "loss": 1.49317408,
+      "memory(GiB)": 74.41,
+      "step": 3980,
+      "train_speed(iter/s)": 1.341692
+    },
+    {
+      "acc": 0.66814318,
+      "epoch": 0.1010908168442415,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.999991867274656e-06,
+      "loss": 1.53233738,
+      "memory(GiB)": 74.41,
+      "step": 3985,
+      "train_speed(iter/s)": 1.341849
+    },
+    {
+      "acc": 0.6676466,
+      "epoch": 0.10121765601217655,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.999989865982725e-06,
+      "loss": 1.45945673,
+      "memory(GiB)": 74.41,
+      "step": 3990,
+      "train_speed(iter/s)": 1.342006
+    },
+    {
+      "acc": 0.65121412,
+      "epoch": 0.10134449518011161,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.999987644768917e-06,
+      "loss": 1.48436022,
+      "memory(GiB)": 74.41,
+      "step": 3995,
+      "train_speed(iter/s)": 1.342156
+    },
+    {
+      "acc": 0.67046523,
+      "epoch": 0.10147133434804667,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.999985203633327e-06,
+      "loss": 1.52903118,
+      "memory(GiB)": 74.41,
+      "step": 4000,
+      "train_speed(iter/s)": 1.342304
+    },
+    {
+      "epoch": 0.10147133434804667,
+      "eval_acc": 0.6559277613149646,
+      "eval_loss": 1.4645648002624512,
+      "eval_runtime": 69.5234,
+      "eval_samples_per_second": 91.624,
+      "eval_steps_per_second": 22.913,
+      "step": 4000
+    },
+    {
+      "acc": 0.66568537,
+      "epoch": 0.10159817351598173,
+      "grad_norm": 3.75,
+      "learning_rate": 9.999982542576065e-06,
+      "loss": 1.4878643,
+      "memory(GiB)": 74.41,
+      "step": 4005,
+      "train_speed(iter/s)": 1.306484
+    },
+    {
+      "acc": 0.66784754,
+      "epoch": 0.10172501268391679,
+      "grad_norm": 4.125,
+      "learning_rate": 9.999979661597247e-06,
+      "loss": 1.52460175,
+      "memory(GiB)": 74.41,
+      "step": 4010,
+      "train_speed(iter/s)": 1.306654
+    },
+    {
+      "acc": 0.67531509,
+      "epoch": 0.10185185185185185,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.999976560697002e-06,
+      "loss": 1.46255007,
+      "memory(GiB)": 74.41,
+      "step": 4015,
+      "train_speed(iter/s)": 1.30683
+    },
+    {
+      "acc": 0.67784286,
+      "epoch": 0.1019786910197869,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.999973239875462e-06,
+      "loss": 1.49459171,
+      "memory(GiB)": 74.41,
+      "step": 4020,
+      "train_speed(iter/s)": 1.307006
+    },
+    {
+      "acc": 0.67188425,
+      "epoch": 0.10210553018772196,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.999969699132776e-06,
+      "loss": 1.48588963,
+      "memory(GiB)": 74.41,
+      "step": 4025,
+      "train_speed(iter/s)": 1.307183
+    },
+    {
+      "acc": 0.65230284,
+      "epoch": 0.10223236935565702,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.999965938469102e-06,
+      "loss": 1.54305906,
+      "memory(GiB)": 74.41,
+      "step": 4030,
+      "train_speed(iter/s)": 1.307357
+    },
+    {
+      "acc": 0.6602695,
+      "epoch": 0.10235920852359208,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.9999619578846e-06,
+      "loss": 1.52192421,
+      "memory(GiB)": 74.41,
+      "step": 4035,
+      "train_speed(iter/s)": 1.307531
+    },
+    {
+      "acc": 0.67116671,
+      "epoch": 0.10248604769152714,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.999957757379451e-06,
+      "loss": 1.50585899,
+      "memory(GiB)": 74.41,
+      "step": 4040,
+      "train_speed(iter/s)": 1.307705
+    },
+    {
+      "acc": 0.68313093,
+      "epoch": 0.1026128868594622,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.999953336953834e-06,
+      "loss": 1.42542734,
+      "memory(GiB)": 74.41,
+      "step": 4045,
+      "train_speed(iter/s)": 1.307823
+    },
+    {
+      "acc": 0.6582571,
+      "epoch": 0.10273972602739725,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.999948696607946e-06,
+      "loss": 1.52769489,
+      "memory(GiB)": 74.41,
+      "step": 4050,
+      "train_speed(iter/s)": 1.307996
+    },
+    {
+      "acc": 0.66601915,
+      "epoch": 0.10286656519533231,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.999943836341992e-06,
+      "loss": 1.49454317,
+      "memory(GiB)": 74.41,
+      "step": 4055,
+      "train_speed(iter/s)": 1.308172
+    },
+    {
+      "acc": 0.69215527,
+      "epoch": 0.10299340436326737,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.999938756156185e-06,
+      "loss": 1.40324678,
+      "memory(GiB)": 74.41,
+      "step": 4060,
+      "train_speed(iter/s)": 1.308337
+    },
+    {
+      "acc": 0.66666937,
+      "epoch": 0.10312024353120243,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.999933456050747e-06,
+      "loss": 1.55898819,
+      "memory(GiB)": 74.41,
+      "step": 4065,
+      "train_speed(iter/s)": 1.308504
+    },
+    {
+      "acc": 0.65882306,
+      "epoch": 0.10324708269913749,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.999927936025914e-06,
+      "loss": 1.5662796,
+      "memory(GiB)": 74.41,
+      "step": 4070,
+      "train_speed(iter/s)": 1.308674
+    },
+    {
+      "acc": 0.67338576,
+      "epoch": 0.10337392186707255,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.999922196081928e-06,
+      "loss": 1.47453909,
+      "memory(GiB)": 74.41,
+      "step": 4075,
+      "train_speed(iter/s)": 1.30885
+    },
+    {
+      "acc": 0.66933041,
+      "epoch": 0.1035007610350076,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.99991623621904e-06,
+      "loss": 1.480445,
+      "memory(GiB)": 74.41,
+      "step": 4080,
+      "train_speed(iter/s)": 1.309026
+    },
+    {
+      "acc": 0.66785889,
+      "epoch": 0.10362760020294266,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.999910056437512e-06,
+      "loss": 1.52627954,
+      "memory(GiB)": 74.41,
+      "step": 4085,
+      "train_speed(iter/s)": 1.309194
+    },
+    {
+      "acc": 0.67075872,
+      "epoch": 0.10375443937087772,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.999903656737618e-06,
+      "loss": 1.46716213,
+      "memory(GiB)": 74.41,
+      "step": 4090,
+      "train_speed(iter/s)": 1.309361
+    },
+    {
+      "acc": 0.66925669,
+      "epoch": 0.10388127853881278,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.999897037119637e-06,
+      "loss": 1.46509256,
+      "memory(GiB)": 74.41,
+      "step": 4095,
+      "train_speed(iter/s)": 1.309532
+    },
+    {
+      "acc": 0.66908813,
+      "epoch": 0.10400811770674784,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.999890197583862e-06,
+      "loss": 1.53523006,
+      "memory(GiB)": 74.41,
+      "step": 4100,
+      "train_speed(iter/s)": 1.309709
+    },
+    {
+      "acc": 0.65242405,
+      "epoch": 0.1041349568746829,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.999883138130593e-06,
+      "loss": 1.54104128,
+      "memory(GiB)": 74.41,
+      "step": 4105,
+      "train_speed(iter/s)": 1.309887
+    },
+    {
+      "acc": 0.67606897,
+      "epoch": 0.10426179604261795,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.999875858760143e-06,
+      "loss": 1.49755116,
+      "memory(GiB)": 74.41,
+      "step": 4110,
+      "train_speed(iter/s)": 1.310056
+    },
+    {
+      "acc": 0.66849222,
+      "epoch": 0.10438863521055301,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.999868359472826e-06,
+      "loss": 1.58808851,
+      "memory(GiB)": 74.41,
+      "step": 4115,
+      "train_speed(iter/s)": 1.310231
+    },
+    {
+      "acc": 0.67327023,
+      "epoch": 0.10451547437848807,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.999860640268977e-06,
+      "loss": 1.52819195,
+      "memory(GiB)": 74.41,
+      "step": 4120,
+      "train_speed(iter/s)": 1.310402
+    },
+    {
+      "acc": 0.68104138,
+      "epoch": 0.10464231354642313,
+      "grad_norm": 3.0,
+      "learning_rate": 9.999852701148935e-06,
+      "loss": 1.48551693,
+      "memory(GiB)": 74.41,
+      "step": 4125,
+      "train_speed(iter/s)": 1.310573
+    },
+    {
+      "acc": 0.68626475,
+      "epoch": 0.10476915271435819,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.999844542113049e-06,
+      "loss": 1.51182346,
+      "memory(GiB)": 74.41,
+      "step": 4130,
+      "train_speed(iter/s)": 1.310748
+    },
+    {
+      "acc": 0.65181551,
+      "epoch": 0.10489599188229325,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.999836163161675e-06,
+      "loss": 1.57443657,
+      "memory(GiB)": 74.41,
+      "step": 4135,
+      "train_speed(iter/s)": 1.310926
+    },
+    {
+      "acc": 0.69658623,
+      "epoch": 0.1050228310502283,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.999827564295187e-06,
+      "loss": 1.37806911,
+      "memory(GiB)": 74.41,
+      "step": 4140,
+      "train_speed(iter/s)": 1.311104
+    },
+    {
+      "acc": 0.67230577,
+      "epoch": 0.10514967021816336,
+      "grad_norm": 2.875,
+      "learning_rate": 9.999818745513958e-06,
+      "loss": 1.47897177,
+      "memory(GiB)": 74.41,
+      "step": 4145,
+      "train_speed(iter/s)": 1.311193
+    },
+    {
+      "acc": 0.6658782,
+      "epoch": 0.10527650938609842,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.99980970681838e-06,
+      "loss": 1.5156867,
+      "memory(GiB)": 74.41,
+      "step": 4150,
+      "train_speed(iter/s)": 1.311359
+    },
+    {
+      "acc": 0.68427377,
+      "epoch": 0.10540334855403348,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.999800448208846e-06,
+      "loss": 1.45277157,
+      "memory(GiB)": 74.41,
+      "step": 4155,
+      "train_speed(iter/s)": 1.311532
+    },
+    {
+      "acc": 0.6690567,
+      "epoch": 0.10553018772196854,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.999790969685767e-06,
+      "loss": 1.47996883,
+      "memory(GiB)": 74.41,
+      "step": 4160,
+      "train_speed(iter/s)": 1.311708
+    },
+    {
+      "acc": 0.67072258,
+      "epoch": 0.1056570268899036,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.999781271249559e-06,
+      "loss": 1.51541729,
+      "memory(GiB)": 74.41,
+      "step": 4165,
+      "train_speed(iter/s)": 1.311887
+    },
+    {
+      "acc": 0.67255821,
+      "epoch": 0.10578386605783865,
+      "grad_norm": 2.875,
+      "learning_rate": 9.999771352900647e-06,
+      "loss": 1.47417526,
+      "memory(GiB)": 74.41,
+      "step": 4170,
+      "train_speed(iter/s)": 1.312068
+    },
+    {
+      "acc": 0.68145399,
+      "epoch": 0.10591070522577371,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.999761214639469e-06,
+      "loss": 1.38901329,
+      "memory(GiB)": 74.41,
+      "step": 4175,
+      "train_speed(iter/s)": 1.31224
+    },
+    {
+      "acc": 0.6806015,
+      "epoch": 0.10603754439370877,
+      "grad_norm": 2.875,
+      "learning_rate": 9.999750856466472e-06,
+      "loss": 1.40244246,
+      "memory(GiB)": 74.41,
+      "step": 4180,
+      "train_speed(iter/s)": 1.31241
+    },
+    {
+      "acc": 0.67742186,
+      "epoch": 0.10616438356164383,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.99974027838211e-06,
+      "loss": 1.4206727,
+      "memory(GiB)": 74.41,
+      "step": 4185,
+      "train_speed(iter/s)": 1.312588
+    },
+    {
+      "acc": 0.6788826,
+      "epoch": 0.10629122272957889,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.999729480386846e-06,
+      "loss": 1.46190777,
+      "memory(GiB)": 74.41,
+      "step": 4190,
+      "train_speed(iter/s)": 1.312763
+    },
+    {
+      "acc": 0.67368855,
+      "epoch": 0.10641806189751395,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.999718462481157e-06,
+      "loss": 1.46354504,
+      "memory(GiB)": 74.41,
+      "step": 4195,
+      "train_speed(iter/s)": 1.312935
+    },
+    {
+      "acc": 0.68497181,
+      "epoch": 0.106544901065449,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.99970722466553e-06,
+      "loss": 1.4745985,
+      "memory(GiB)": 74.41,
+      "step": 4200,
+      "train_speed(iter/s)": 1.313114
+    },
+    {
+      "acc": 0.67653174,
+      "epoch": 0.10667174023338406,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.999695766940458e-06,
+      "loss": 1.44064255,
+      "memory(GiB)": 74.41,
+      "step": 4205,
+      "train_speed(iter/s)": 1.313286
+    },
+    {
+      "acc": 0.66195803,
+      "epoch": 0.10679857940131912,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.999684089306442e-06,
+      "loss": 1.50672283,
+      "memory(GiB)": 74.41,
+      "step": 4210,
+      "train_speed(iter/s)": 1.313455
+    },
+    {
+      "acc": 0.67030663,
+      "epoch": 0.10692541856925418,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.999672191763999e-06,
+      "loss": 1.47845783,
+      "memory(GiB)": 74.41,
+      "step": 4215,
+      "train_speed(iter/s)": 1.313625
+    },
+    {
+      "acc": 0.67737041,
+      "epoch": 0.10705225773718924,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.99966007431365e-06,
+      "loss": 1.3868453,
+      "memory(GiB)": 74.41,
+      "step": 4220,
+      "train_speed(iter/s)": 1.31379
+    },
+    {
+      "acc": 0.67311249,
+      "epoch": 0.1071790969051243,
+      "grad_norm": 2.5,
+      "learning_rate": 9.99964773695593e-06,
+      "loss": 1.48974056,
+      "memory(GiB)": 74.41,
+      "step": 4225,
+      "train_speed(iter/s)": 1.313956
+    },
+    {
+      "acc": 0.68769093,
+      "epoch": 0.10730593607305935,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.999635179691381e-06,
+      "loss": 1.52323427,
+      "memory(GiB)": 74.41,
+      "step": 4230,
+      "train_speed(iter/s)": 1.31412
+    },
+    {
+      "acc": 0.66809845,
+      "epoch": 0.10743277524099441,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.999622402520553e-06,
+      "loss": 1.52554207,
+      "memory(GiB)": 74.41,
+      "step": 4235,
+      "train_speed(iter/s)": 1.314302
+    },
+    {
+      "acc": 0.66274004,
+      "epoch": 0.10755961440892947,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.999609405444012e-06,
+      "loss": 1.51015797,
+      "memory(GiB)": 74.41,
+      "step": 4240,
+      "train_speed(iter/s)": 1.314481
+    },
+    {
+      "acc": 0.67617941,
+      "epoch": 0.10768645357686453,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.999596188462328e-06,
+      "loss": 1.48674488,
+      "memory(GiB)": 74.41,
+      "step": 4245,
+      "train_speed(iter/s)": 1.314658
+    },
+    {
+      "acc": 0.67757788,
+      "epoch": 0.10781329274479959,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.99958275157608e-06,
+      "loss": 1.46855984,
+      "memory(GiB)": 74.41,
+      "step": 4250,
+      "train_speed(iter/s)": 1.31483
+    },
+    {
+      "acc": 0.67494669,
+      "epoch": 0.10794013191273465,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.999569094785862e-06,
+      "loss": 1.53940477,
+      "memory(GiB)": 74.41,
+      "step": 4255,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.68242197,
+      "epoch": 0.1080669710806697,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.999555218092273e-06,
+      "loss": 1.45918045,
+      "memory(GiB)": 74.41,
+      "step": 4260,
+      "train_speed(iter/s)": 1.315185
+    },
+    {
+      "acc": 0.6817873,
+      "epoch": 0.10819381024860476,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.999541121495926e-06,
+      "loss": 1.48703394,
+      "memory(GiB)": 74.41,
+      "step": 4265,
+      "train_speed(iter/s)": 1.315366
+    },
+    {
+      "acc": 0.6637857,
+      "epoch": 0.10832064941653982,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.999526804997439e-06,
+      "loss": 1.52147179,
+      "memory(GiB)": 74.41,
+      "step": 4270,
+      "train_speed(iter/s)": 1.315546
+    },
+    {
+      "acc": 0.66408257,
+      "epoch": 0.10844748858447488,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.99951226859744e-06,
+      "loss": 1.56843023,
+      "memory(GiB)": 74.41,
+      "step": 4275,
+      "train_speed(iter/s)": 1.3157
+    },
+    {
+      "acc": 0.6769114,
+      "epoch": 0.10857432775240994,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.999497512296572e-06,
+      "loss": 1.50284309,
+      "memory(GiB)": 74.41,
+      "step": 4280,
+      "train_speed(iter/s)": 1.315871
+    },
+    {
+      "acc": 0.6767333,
+      "epoch": 0.108701166920345,
+      "grad_norm": 5.15625,
+      "learning_rate": 9.999482536095483e-06,
+      "loss": 1.49143486,
+      "memory(GiB)": 74.41,
+      "step": 4285,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.67211847,
+      "epoch": 0.10882800608828005,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.999467339994827e-06,
+      "loss": 1.47893095,
+      "memory(GiB)": 74.41,
+      "step": 4290,
+      "train_speed(iter/s)": 1.316221
+    },
+    {
+      "acc": 0.67917709,
+      "epoch": 0.10895484525621511,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.99945192399528e-06,
+      "loss": 1.45635271,
+      "memory(GiB)": 74.41,
+      "step": 4295,
+      "train_speed(iter/s)": 1.31639
+    },
+    {
+      "acc": 0.66396275,
+      "epoch": 0.10908168442415017,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.999436288097515e-06,
+      "loss": 1.48150997,
+      "memory(GiB)": 74.41,
+      "step": 4300,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.66919603,
+      "epoch": 0.10920852359208523,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.99942043230222e-06,
+      "loss": 1.50598469,
+      "memory(GiB)": 74.41,
+      "step": 4305,
+      "train_speed(iter/s)": 1.316717
+    },
+    {
+      "acc": 0.66882048,
+      "epoch": 0.10933536276002029,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.999404356610095e-06,
+      "loss": 1.41077328,
+      "memory(GiB)": 74.41,
+      "step": 4310,
+      "train_speed(iter/s)": 1.316886
+    },
+    {
+      "acc": 0.67469664,
+      "epoch": 0.10946220192795535,
+      "grad_norm": 3.25,
+      "learning_rate": 9.999388061021846e-06,
+      "loss": 1.49336529,
+      "memory(GiB)": 74.41,
+      "step": 4315,
+      "train_speed(iter/s)": 1.317056
+    },
+    {
+      "acc": 0.66853371,
+      "epoch": 0.1095890410958904,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.99937154553819e-06,
+      "loss": 1.51188068,
+      "memory(GiB)": 74.41,
+      "step": 4320,
+      "train_speed(iter/s)": 1.317226
+    },
+    {
+      "acc": 0.65006242,
+      "epoch": 0.10971588026382546,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.999354810159852e-06,
+      "loss": 1.50654049,
+      "memory(GiB)": 74.41,
+      "step": 4325,
+      "train_speed(iter/s)": 1.317387
+    },
+    {
+      "acc": 0.66229887,
+      "epoch": 0.10984271943176052,
+      "grad_norm": 3.125,
+      "learning_rate": 9.999337854887567e-06,
+      "loss": 1.49300222,
+      "memory(GiB)": 74.41,
+      "step": 4330,
+      "train_speed(iter/s)": 1.317551
+    },
+    {
+      "acc": 0.6801734,
+      "epoch": 0.10996955859969558,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.999320679722086e-06,
+      "loss": 1.49224205,
+      "memory(GiB)": 74.41,
+      "step": 4335,
+      "train_speed(iter/s)": 1.317725
+    },
+    {
+      "acc": 0.67600551,
+      "epoch": 0.11009639776763064,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.999303284664159e-06,
+      "loss": 1.4802927,
+      "memory(GiB)": 74.41,
+      "step": 4340,
+      "train_speed(iter/s)": 1.317896
+    },
+    {
+      "acc": 0.69371319,
+      "epoch": 0.1102232369355657,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.999285669714555e-06,
+      "loss": 1.43609085,
+      "memory(GiB)": 74.41,
+      "step": 4345,
+      "train_speed(iter/s)": 1.318063
+    },
+    {
+      "acc": 0.6929162,
+      "epoch": 0.11035007610350075,
+      "grad_norm": 3.0,
+      "learning_rate": 9.999267834874044e-06,
+      "loss": 1.37357235,
+      "memory(GiB)": 74.41,
+      "step": 4350,
+      "train_speed(iter/s)": 1.31824
+    },
+    {
+      "acc": 0.67220907,
+      "epoch": 0.11047691527143581,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.999249780143416e-06,
+      "loss": 1.44588776,
+      "memory(GiB)": 74.41,
+      "step": 4355,
+      "train_speed(iter/s)": 1.318405
+    },
+    {
+      "acc": 0.67073865,
+      "epoch": 0.11060375443937087,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.999231505523463e-06,
+      "loss": 1.49198589,
+      "memory(GiB)": 74.41,
+      "step": 4360,
+      "train_speed(iter/s)": 1.318571
+    },
+    {
+      "acc": 0.65534611,
+      "epoch": 0.11073059360730593,
+      "grad_norm": 3.0,
+      "learning_rate": 9.999213011014987e-06,
+      "loss": 1.54008198,
+      "memory(GiB)": 74.41,
+      "step": 4365,
+      "train_speed(iter/s)": 1.318734
+    },
+    {
+      "acc": 0.66665831,
+      "epoch": 0.11085743277524099,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.999194296618805e-06,
+      "loss": 1.50369415,
+      "memory(GiB)": 74.41,
+      "step": 4370,
+      "train_speed(iter/s)": 1.318904
+    },
+    {
+      "acc": 0.68175788,
+      "epoch": 0.11098427194317605,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.999175362335735e-06,
+      "loss": 1.47474575,
+      "memory(GiB)": 74.41,
+      "step": 4375,
+      "train_speed(iter/s)": 1.319074
+    },
+    {
+      "acc": 0.67292752,
+      "epoch": 0.1111111111111111,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.999156208166614e-06,
+      "loss": 1.48732119,
+      "memory(GiB)": 74.41,
+      "step": 4380,
+      "train_speed(iter/s)": 1.319246
+    },
+    {
+      "acc": 0.67052231,
+      "epoch": 0.11123795027904616,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.999136834112284e-06,
+      "loss": 1.47251225,
+      "memory(GiB)": 74.41,
+      "step": 4385,
+      "train_speed(iter/s)": 1.319413
+    },
+    {
+      "acc": 0.66807299,
+      "epoch": 0.11136478944698122,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.999117240173597e-06,
+      "loss": 1.51710911,
+      "memory(GiB)": 74.41,
+      "step": 4390,
+      "train_speed(iter/s)": 1.31958
+    },
+    {
+      "acc": 0.66392374,
+      "epoch": 0.11149162861491628,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.999097426351412e-06,
+      "loss": 1.55685024,
+      "memory(GiB)": 74.41,
+      "step": 4395,
+      "train_speed(iter/s)": 1.319755
+    },
+    {
+      "acc": 0.67131495,
+      "epoch": 0.11161846778285134,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.999077392646606e-06,
+      "loss": 1.47995167,
+      "memory(GiB)": 74.41,
+      "step": 4400,
+      "train_speed(iter/s)": 1.319924
+    },
+    {
+      "acc": 0.67708235,
+      "epoch": 0.1117453069507864,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.999057139060055e-06,
+      "loss": 1.4841466,
+      "memory(GiB)": 74.41,
+      "step": 4405,
+      "train_speed(iter/s)": 1.320093
+    },
+    {
+      "acc": 0.68041668,
+      "epoch": 0.11187214611872145,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.999036665592653e-06,
+      "loss": 1.45451403,
+      "memory(GiB)": 74.41,
+      "step": 4410,
+      "train_speed(iter/s)": 1.320263
+    },
+    {
+      "acc": 0.68460665,
+      "epoch": 0.11199898528665651,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.999015972245298e-06,
+      "loss": 1.4740859,
+      "memory(GiB)": 74.41,
+      "step": 4415,
+      "train_speed(iter/s)": 1.320428
+    },
+    {
+      "acc": 0.6704875,
+      "epoch": 0.11212582445459157,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.998995059018901e-06,
+      "loss": 1.48508625,
+      "memory(GiB)": 74.41,
+      "step": 4420,
+      "train_speed(iter/s)": 1.320594
+    },
+    {
+      "acc": 0.66462712,
+      "epoch": 0.11225266362252663,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.998973925914384e-06,
+      "loss": 1.45482473,
+      "memory(GiB)": 74.41,
+      "step": 4425,
+      "train_speed(iter/s)": 1.320742
+    },
+    {
+      "acc": 0.66697855,
+      "epoch": 0.11237950279046169,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.998952572932675e-06,
+      "loss": 1.51032858,
+      "memory(GiB)": 74.41,
+      "step": 4430,
+      "train_speed(iter/s)": 1.320904
+    },
+    {
+      "acc": 0.64980903,
+      "epoch": 0.11250634195839675,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.998931000074712e-06,
+      "loss": 1.58975716,
+      "memory(GiB)": 74.41,
+      "step": 4435,
+      "train_speed(iter/s)": 1.321074
+    },
+    {
+      "acc": 0.66433291,
+      "epoch": 0.1126331811263318,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.998909207341446e-06,
+      "loss": 1.47459249,
+      "memory(GiB)": 74.41,
+      "step": 4440,
+      "train_speed(iter/s)": 1.321243
+    },
+    {
+      "acc": 0.65861745,
+      "epoch": 0.11276002029426686,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.998887194733833e-06,
+      "loss": 1.56254463,
+      "memory(GiB)": 74.41,
+      "step": 4445,
+      "train_speed(iter/s)": 1.32141
+    },
+    {
+      "acc": 0.68397269,
+      "epoch": 0.11288685946220192,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.998864962252843e-06,
+      "loss": 1.42910557,
+      "memory(GiB)": 74.41,
+      "step": 4450,
+      "train_speed(iter/s)": 1.321542
+    },
+    {
+      "acc": 0.66248326,
+      "epoch": 0.11301369863013698,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.998842509899456e-06,
+      "loss": 1.54856176,
+      "memory(GiB)": 74.41,
+      "step": 4455,
+      "train_speed(iter/s)": 1.321709
+    },
+    {
+      "acc": 0.67078896,
+      "epoch": 0.11314053779807204,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.998819837674655e-06,
+      "loss": 1.48127041,
+      "memory(GiB)": 74.41,
+      "step": 4460,
+      "train_speed(iter/s)": 1.321872
+    },
+    {
+      "acc": 0.68337669,
+      "epoch": 0.1132673769660071,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.99879694557944e-06,
+      "loss": 1.44259462,
+      "memory(GiB)": 74.41,
+      "step": 4465,
+      "train_speed(iter/s)": 1.32204
+    },
+    {
+      "acc": 0.68209591,
+      "epoch": 0.11339421613394216,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.998773833614816e-06,
+      "loss": 1.48064919,
+      "memory(GiB)": 74.41,
+      "step": 4470,
+      "train_speed(iter/s)": 1.322208
+    },
+    {
+      "acc": 0.68829956,
+      "epoch": 0.11352105530187721,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.998750501781803e-06,
+      "loss": 1.38605137,
+      "memory(GiB)": 74.41,
+      "step": 4475,
+      "train_speed(iter/s)": 1.32237
+    },
+    {
+      "acc": 0.68245234,
+      "epoch": 0.11364789446981227,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.998726950081425e-06,
+      "loss": 1.49022751,
+      "memory(GiB)": 74.41,
+      "step": 4480,
+      "train_speed(iter/s)": 1.322539
+    },
+    {
+      "acc": 0.67548499,
+      "epoch": 0.11377473363774733,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.998703178514717e-06,
+      "loss": 1.5494153,
+      "memory(GiB)": 74.41,
+      "step": 4485,
+      "train_speed(iter/s)": 1.322702
+    },
+    {
+      "acc": 0.66147699,
+      "epoch": 0.11390157280568239,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.998679187082724e-06,
+      "loss": 1.46620655,
+      "memory(GiB)": 74.41,
+      "step": 4490,
+      "train_speed(iter/s)": 1.322868
+    },
+    {
+      "acc": 0.66168613,
+      "epoch": 0.11402841197361745,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.998654975786506e-06,
+      "loss": 1.45899658,
+      "memory(GiB)": 74.41,
+      "step": 4495,
+      "train_speed(iter/s)": 1.323033
+    },
+    {
+      "acc": 0.67297244,
+      "epoch": 0.1141552511415525,
+      "grad_norm": 3.25,
+      "learning_rate": 9.998630544627123e-06,
+      "loss": 1.45518017,
+      "memory(GiB)": 74.41,
+      "step": 4500,
+      "train_speed(iter/s)": 1.323206
+    },
+    {
+      "acc": 0.68147469,
+      "epoch": 0.11428209030948756,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.998605893605653e-06,
+      "loss": 1.44620113,
+      "memory(GiB)": 74.41,
+      "step": 4505,
+      "train_speed(iter/s)": 1.323375
+    },
+    {
+      "acc": 0.67424159,
+      "epoch": 0.11440892947742262,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.998581022723178e-06,
+      "loss": 1.52440109,
+      "memory(GiB)": 74.41,
+      "step": 4510,
+      "train_speed(iter/s)": 1.32354
+    },
+    {
+      "acc": 0.65664759,
+      "epoch": 0.11453576864535768,
+      "grad_norm": 3.25,
+      "learning_rate": 9.998555931980792e-06,
+      "loss": 1.53879776,
+      "memory(GiB)": 74.41,
+      "step": 4515,
+      "train_speed(iter/s)": 1.3237
+    },
+    {
+      "acc": 0.66626773,
+      "epoch": 0.11466260781329274,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.998530621379599e-06,
+      "loss": 1.51170197,
+      "memory(GiB)": 74.41,
+      "step": 4520,
+      "train_speed(iter/s)": 1.323854
+    },
+    {
+      "acc": 0.67709274,
+      "epoch": 0.1147894469812278,
+      "grad_norm": 2.40625,
+      "learning_rate": 9.998505090920713e-06,
+      "loss": 1.48293648,
+      "memory(GiB)": 74.41,
+      "step": 4525,
+      "train_speed(iter/s)": 1.324012
+    },
+    {
+      "acc": 0.68277764,
+      "epoch": 0.11491628614916286,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.998479340605257e-06,
+      "loss": 1.4591362,
+      "memory(GiB)": 74.41,
+      "step": 4530,
+      "train_speed(iter/s)": 1.324173
+    },
+    {
+      "acc": 0.67874904,
+      "epoch": 0.11504312531709791,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.99845337043436e-06,
+      "loss": 1.45485706,
+      "memory(GiB)": 74.41,
+      "step": 4535,
+      "train_speed(iter/s)": 1.324334
+    },
+    {
+      "acc": 0.6644361,
+      "epoch": 0.11516996448503297,
+      "grad_norm": 3.75,
+      "learning_rate": 9.998427180409171e-06,
+      "loss": 1.60532799,
+      "memory(GiB)": 74.41,
+      "step": 4540,
+      "train_speed(iter/s)": 1.324474
+    },
+    {
+      "acc": 0.66719456,
+      "epoch": 0.11529680365296803,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.998400770530836e-06,
+      "loss": 1.5034709,
+      "memory(GiB)": 74.41,
+      "step": 4545,
+      "train_speed(iter/s)": 1.32463
+    },
+    {
+      "acc": 0.66686759,
+      "epoch": 0.11542364282090309,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.99837414080052e-06,
+      "loss": 1.48067989,
+      "memory(GiB)": 74.41,
+      "step": 4550,
+      "train_speed(iter/s)": 1.324788
+    },
+    {
+      "acc": 0.6652092,
+      "epoch": 0.11555048198883815,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.998347291219393e-06,
+      "loss": 1.50015774,
+      "memory(GiB)": 74.41,
+      "step": 4555,
+      "train_speed(iter/s)": 1.324943
+    },
+    {
+      "acc": 0.68141556,
+      "epoch": 0.1156773211567732,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.998320221788635e-06,
+      "loss": 1.46804113,
+      "memory(GiB)": 74.41,
+      "step": 4560,
+      "train_speed(iter/s)": 1.325098
+    },
+    {
+      "acc": 0.64479847,
+      "epoch": 0.11580416032470826,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.998292932509438e-06,
+      "loss": 1.65546417,
+      "memory(GiB)": 74.41,
+      "step": 4565,
+      "train_speed(iter/s)": 1.325254
+    },
+    {
+      "acc": 0.65872478,
+      "epoch": 0.11593099949264332,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.998265423383003e-06,
+      "loss": 1.55099192,
+      "memory(GiB)": 74.41,
+      "step": 4570,
+      "train_speed(iter/s)": 1.325406
+    },
+    {
+      "acc": 0.67568188,
+      "epoch": 0.11605783866057838,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.998237694410537e-06,
+      "loss": 1.48669205,
+      "memory(GiB)": 74.41,
+      "step": 4575,
+      "train_speed(iter/s)": 1.32556
+    },
+    {
+      "acc": 0.68014956,
+      "epoch": 0.11618467782851344,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.998209745593264e-06,
+      "loss": 1.43002949,
+      "memory(GiB)": 74.41,
+      "step": 4580,
+      "train_speed(iter/s)": 1.325715
+    },
+    {
+      "acc": 0.66947289,
+      "epoch": 0.1163115169964485,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.99818157693241e-06,
+      "loss": 1.50296373,
+      "memory(GiB)": 74.41,
+      "step": 4585,
+      "train_speed(iter/s)": 1.325865
+    },
+    {
+      "acc": 0.68716621,
+      "epoch": 0.11643835616438356,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.998153188429216e-06,
+      "loss": 1.36154575,
+      "memory(GiB)": 74.41,
+      "step": 4590,
+      "train_speed(iter/s)": 1.32601
+    },
+    {
+      "acc": 0.67082081,
+      "epoch": 0.11656519533231861,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.99812458008493e-06,
+      "loss": 1.45181417,
+      "memory(GiB)": 74.41,
+      "step": 4595,
+      "train_speed(iter/s)": 1.326154
+    },
+    {
+      "acc": 0.67618895,
+      "epoch": 0.11669203450025367,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.998095751900806e-06,
+      "loss": 1.51805611,
+      "memory(GiB)": 74.41,
+      "step": 4600,
+      "train_speed(iter/s)": 1.326298
+    },
+    {
+      "acc": 0.66177635,
+      "epoch": 0.11681887366818873,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.99806670387812e-06,
+      "loss": 1.49660454,
+      "memory(GiB)": 74.41,
+      "step": 4605,
+      "train_speed(iter/s)": 1.326452
+    },
+    {
+      "acc": 0.67629166,
+      "epoch": 0.11694571283612379,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.998037436018144e-06,
+      "loss": 1.49622555,
+      "memory(GiB)": 74.41,
+      "step": 4610,
+      "train_speed(iter/s)": 1.326599
+    },
+    {
+      "acc": 0.67961283,
+      "epoch": 0.11707255200405885,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.998007948322168e-06,
+      "loss": 1.42723475,
+      "memory(GiB)": 74.41,
+      "step": 4615,
+      "train_speed(iter/s)": 1.326734
+    },
+    {
+      "acc": 0.66750212,
+      "epoch": 0.1171993911719939,
+      "grad_norm": 2.875,
+      "learning_rate": 9.997978240791487e-06,
+      "loss": 1.49176378,
+      "memory(GiB)": 74.41,
+      "step": 4620,
+      "train_speed(iter/s)": 1.326885
+    },
+    {
+      "acc": 0.67016382,
+      "epoch": 0.11732623033992896,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.99794831342741e-06,
+      "loss": 1.52620354,
+      "memory(GiB)": 74.41,
+      "step": 4625,
+      "train_speed(iter/s)": 1.327037
+    },
+    {
+      "acc": 0.67022319,
+      "epoch": 0.11745306950786402,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.99791816623125e-06,
+      "loss": 1.52248545,
+      "memory(GiB)": 74.41,
+      "step": 4630,
+      "train_speed(iter/s)": 1.327188
+    },
+    {
+      "acc": 0.66981182,
+      "epoch": 0.11757990867579908,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.997887799204335e-06,
+      "loss": 1.45147533,
+      "memory(GiB)": 74.41,
+      "step": 4635,
+      "train_speed(iter/s)": 1.327329
+    },
+    {
+      "acc": 0.65856056,
+      "epoch": 0.11770674784373414,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.997857212348e-06,
+      "loss": 1.51395998,
+      "memory(GiB)": 74.41,
+      "step": 4640,
+      "train_speed(iter/s)": 1.327476
+    },
+    {
+      "acc": 0.67336893,
+      "epoch": 0.1178335870116692,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.997826405663593e-06,
+      "loss": 1.48139896,
+      "memory(GiB)": 74.41,
+      "step": 4645,
+      "train_speed(iter/s)": 1.32762
+    },
+    {
+      "acc": 0.69231782,
+      "epoch": 0.11796042617960426,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.997795379152468e-06,
+      "loss": 1.35688334,
+      "memory(GiB)": 74.41,
+      "step": 4650,
+      "train_speed(iter/s)": 1.327767
+    },
+    {
+      "acc": 0.64404244,
+      "epoch": 0.11808726534753931,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.997764132815985e-06,
+      "loss": 1.55673237,
+      "memory(GiB)": 74.41,
+      "step": 4655,
+      "train_speed(iter/s)": 1.327916
+    },
+    {
+      "acc": 0.66965251,
+      "epoch": 0.11821410451547437,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.997732666655524e-06,
+      "loss": 1.54983416,
+      "memory(GiB)": 74.41,
+      "step": 4660,
+      "train_speed(iter/s)": 1.328059
+    },
+    {
+      "acc": 0.6565773,
+      "epoch": 0.11834094368340943,
+      "grad_norm": 3.0,
+      "learning_rate": 9.997700980672469e-06,
+      "loss": 1.52206669,
+      "memory(GiB)": 74.41,
+      "step": 4665,
+      "train_speed(iter/s)": 1.328204
+    },
+    {
+      "acc": 0.6835885,
+      "epoch": 0.11846778285134449,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.997669074868208e-06,
+      "loss": 1.44246502,
+      "memory(GiB)": 74.41,
+      "step": 4670,
+      "train_speed(iter/s)": 1.328345
+    },
+    {
+      "acc": 0.68478107,
+      "epoch": 0.11859462201927955,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.997636949244151e-06,
+      "loss": 1.48518314,
+      "memory(GiB)": 74.41,
+      "step": 4675,
+      "train_speed(iter/s)": 1.328468
+    },
+    {
+      "acc": 0.66209311,
+      "epoch": 0.1187214611872146,
+      "grad_norm": 3.125,
+      "learning_rate": 9.997604603801707e-06,
+      "loss": 1.54220543,
+      "memory(GiB)": 74.41,
+      "step": 4680,
+      "train_speed(iter/s)": 1.328611
+    },
+    {
+      "acc": 0.65118084,
+      "epoch": 0.11884830035514966,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.9975720385423e-06,
+      "loss": 1.57128601,
+      "memory(GiB)": 74.41,
+      "step": 4685,
+      "train_speed(iter/s)": 1.328753
+    },
+    {
+      "acc": 0.67937117,
+      "epoch": 0.11897513952308472,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.997539253467361e-06,
+      "loss": 1.5065485,
+      "memory(GiB)": 74.41,
+      "step": 4690,
+      "train_speed(iter/s)": 1.328897
+    },
+    {
+      "acc": 0.67064023,
+      "epoch": 0.11910197869101978,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.997506248578334e-06,
+      "loss": 1.49135513,
+      "memory(GiB)": 74.41,
+      "step": 4695,
+      "train_speed(iter/s)": 1.329041
+    },
+    {
+      "acc": 0.67917252,
+      "epoch": 0.11922881785895484,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.997473023876671e-06,
+      "loss": 1.54235888,
+      "memory(GiB)": 74.41,
+      "step": 4700,
+      "train_speed(iter/s)": 1.329186
+    },
+    {
+      "acc": 0.66986547,
+      "epoch": 0.1193556570268899,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.997439579363831e-06,
+      "loss": 1.56781349,
+      "memory(GiB)": 74.41,
+      "step": 4705,
+      "train_speed(iter/s)": 1.329319
+    },
+    {
+      "acc": 0.66297946,
+      "epoch": 0.11948249619482496,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.997405915041288e-06,
+      "loss": 1.4766654,
+      "memory(GiB)": 74.41,
+      "step": 4710,
+      "train_speed(iter/s)": 1.329466
+    },
+    {
+      "acc": 0.6790206,
+      "epoch": 0.11960933536276001,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.99737203091052e-06,
+      "loss": 1.45135422,
+      "memory(GiB)": 74.41,
+      "step": 4715,
+      "train_speed(iter/s)": 1.329607
+    },
+    {
+      "acc": 0.68656054,
+      "epoch": 0.11973617453069507,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.997337926973018e-06,
+      "loss": 1.44021225,
+      "memory(GiB)": 74.41,
+      "step": 4720,
+      "train_speed(iter/s)": 1.329754
+    },
+    {
+      "acc": 0.66749592,
+      "epoch": 0.11986301369863013,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.997303603230282e-06,
+      "loss": 1.49547596,
+      "memory(GiB)": 74.41,
+      "step": 4725,
+      "train_speed(iter/s)": 1.329877
+    },
+    {
+      "acc": 0.68116827,
+      "epoch": 0.11998985286656519,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.997269059683822e-06,
+      "loss": 1.43948898,
+      "memory(GiB)": 74.41,
+      "step": 4730,
+      "train_speed(iter/s)": 1.330022
+    },
+    {
+      "acc": 0.66667705,
+      "epoch": 0.12011669203450025,
+      "grad_norm": 3.625,
+      "learning_rate": 9.997234296335159e-06,
+      "loss": 1.50837641,
+      "memory(GiB)": 74.41,
+      "step": 4735,
+      "train_speed(iter/s)": 1.330169
+    },
+    {
+      "acc": 0.67879791,
+      "epoch": 0.1202435312024353,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.997199313185821e-06,
+      "loss": 1.46406517,
+      "memory(GiB)": 74.41,
+      "step": 4740,
+      "train_speed(iter/s)": 1.330315
+    },
+    {
+      "acc": 0.68766599,
+      "epoch": 0.12037037037037036,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.997164110237345e-06,
+      "loss": 1.36739044,
+      "memory(GiB)": 74.41,
+      "step": 4745,
+      "train_speed(iter/s)": 1.330464
+    },
+    {
+      "acc": 0.67388668,
+      "epoch": 0.12049720953830542,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.99712868749128e-06,
+      "loss": 1.47251844,
+      "memory(GiB)": 74.41,
+      "step": 4750,
+      "train_speed(iter/s)": 1.330604
+    },
+    {
+      "acc": 0.69367981,
+      "epoch": 0.12062404870624048,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.997093044949186e-06,
+      "loss": 1.46491814,
+      "memory(GiB)": 74.41,
+      "step": 4755,
+      "train_speed(iter/s)": 1.330754
+    },
+    {
+      "acc": 0.67574244,
+      "epoch": 0.12075088787417554,
+      "grad_norm": 3.125,
+      "learning_rate": 9.997057182612631e-06,
+      "loss": 1.43958941,
+      "memory(GiB)": 74.41,
+      "step": 4760,
+      "train_speed(iter/s)": 1.330897
+    },
+    {
+      "acc": 0.67430558,
+      "epoch": 0.1208777270421106,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.997021100483188e-06,
+      "loss": 1.48250885,
+      "memory(GiB)": 74.41,
+      "step": 4765,
+      "train_speed(iter/s)": 1.331047
+    },
+    {
+      "acc": 0.68293786,
+      "epoch": 0.12100456621004566,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.996984798562448e-06,
+      "loss": 1.41871901,
+      "memory(GiB)": 74.41,
+      "step": 4770,
+      "train_speed(iter/s)": 1.331191
+    },
+    {
+      "acc": 0.66476855,
+      "epoch": 0.12113140537798071,
+      "grad_norm": 3.375,
+      "learning_rate": 9.996948276852008e-06,
+      "loss": 1.48826284,
+      "memory(GiB)": 74.41,
+      "step": 4775,
+      "train_speed(iter/s)": 1.331333
+    },
+    {
+      "acc": 0.66523142,
+      "epoch": 0.12125824454591577,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.99691153535347e-06,
+      "loss": 1.52338619,
+      "memory(GiB)": 74.41,
+      "step": 4780,
+      "train_speed(iter/s)": 1.331475
+    },
+    {
+      "acc": 0.67699375,
+      "epoch": 0.12138508371385083,
+      "grad_norm": 3.0,
+      "learning_rate": 9.996874574068457e-06,
+      "loss": 1.45959587,
+      "memory(GiB)": 74.41,
+      "step": 4785,
+      "train_speed(iter/s)": 1.331612
+    },
+    {
+      "acc": 0.6678587,
+      "epoch": 0.12151192288178589,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.996837392998586e-06,
+      "loss": 1.46391773,
+      "memory(GiB)": 74.41,
+      "step": 4790,
+      "train_speed(iter/s)": 1.331748
+    },
+    {
+      "acc": 0.67532272,
+      "epoch": 0.12163876204972095,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.996799992145501e-06,
+      "loss": 1.48310652,
+      "memory(GiB)": 74.41,
+      "step": 4795,
+      "train_speed(iter/s)": 1.331884
+    },
+    {
+      "acc": 0.6779758,
+      "epoch": 0.121765601217656,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.996762371510843e-06,
+      "loss": 1.51433468,
+      "memory(GiB)": 74.41,
+      "step": 4800,
+      "train_speed(iter/s)": 1.332018
+    },
+    {
+      "acc": 0.67878542,
+      "epoch": 0.12189244038559106,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.996724531096264e-06,
+      "loss": 1.45806847,
+      "memory(GiB)": 74.41,
+      "step": 4805,
+      "train_speed(iter/s)": 1.332156
+    },
+    {
+      "acc": 0.67410221,
+      "epoch": 0.12201927955352612,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.996686470903434e-06,
+      "loss": 1.47716055,
+      "memory(GiB)": 74.41,
+      "step": 4810,
+      "train_speed(iter/s)": 1.332298
+    },
+    {
+      "acc": 0.6701025,
+      "epoch": 0.12214611872146118,
+      "grad_norm": 2.875,
+      "learning_rate": 9.996648190934025e-06,
+      "loss": 1.42997646,
+      "memory(GiB)": 74.41,
+      "step": 4815,
+      "train_speed(iter/s)": 1.332435
+    },
+    {
+      "acc": 0.68003702,
+      "epoch": 0.12227295788939624,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.996609691189718e-06,
+      "loss": 1.46451263,
+      "memory(GiB)": 74.41,
+      "step": 4820,
+      "train_speed(iter/s)": 1.332577
+    },
+    {
+      "acc": 0.68431902,
+      "epoch": 0.1223997970573313,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.996570971672209e-06,
+      "loss": 1.43506641,
+      "memory(GiB)": 74.41,
+      "step": 4825,
+      "train_speed(iter/s)": 1.332714
+    },
+    {
+      "acc": 0.66046624,
+      "epoch": 0.12252663622526636,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.996532032383202e-06,
+      "loss": 1.56521149,
+      "memory(GiB)": 74.41,
+      "step": 4830,
+      "train_speed(iter/s)": 1.332849
+    },
+    {
+      "acc": 0.67588091,
+      "epoch": 0.12265347539320141,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.996492873324406e-06,
+      "loss": 1.54269781,
+      "memory(GiB)": 74.41,
+      "step": 4835,
+      "train_speed(iter/s)": 1.332987
+    },
+    {
+      "acc": 0.67505503,
+      "epoch": 0.12278031456113647,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.996453494497546e-06,
+      "loss": 1.50036602,
+      "memory(GiB)": 74.41,
+      "step": 4840,
+      "train_speed(iter/s)": 1.333119
+    },
+    {
+      "acc": 0.67111483,
+      "epoch": 0.12290715372907153,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.996413895904355e-06,
+      "loss": 1.49330263,
+      "memory(GiB)": 74.41,
+      "step": 4845,
+      "train_speed(iter/s)": 1.333262
+    },
+    {
+      "acc": 0.67991438,
+      "epoch": 0.12303399289700659,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.996374077546573e-06,
+      "loss": 1.45041294,
+      "memory(GiB)": 74.41,
+      "step": 4850,
+      "train_speed(iter/s)": 1.333402
+    },
+    {
+      "acc": 0.6795722,
+      "epoch": 0.12316083206494165,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.996334039425952e-06,
+      "loss": 1.44384108,
+      "memory(GiB)": 74.41,
+      "step": 4855,
+      "train_speed(iter/s)": 1.333541
+    },
+    {
+      "acc": 0.6939476,
+      "epoch": 0.1232876712328767,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.996293781544255e-06,
+      "loss": 1.402248,
+      "memory(GiB)": 74.41,
+      "step": 4860,
+      "train_speed(iter/s)": 1.333678
+    },
+    {
+      "acc": 0.67524948,
+      "epoch": 0.12341451040081176,
+      "grad_norm": 2.875,
+      "learning_rate": 9.996253303903247e-06,
+      "loss": 1.43993673,
+      "memory(GiB)": 74.41,
+      "step": 4865,
+      "train_speed(iter/s)": 1.333815
+    },
+    {
+      "acc": 0.66970091,
+      "epoch": 0.12354134956874682,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.996212606504713e-06,
+      "loss": 1.41931171,
+      "memory(GiB)": 74.41,
+      "step": 4870,
+      "train_speed(iter/s)": 1.333948
+    },
+    {
+      "acc": 0.68554864,
+      "epoch": 0.12366818873668188,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.996171689350444e-06,
+      "loss": 1.47563438,
+      "memory(GiB)": 74.41,
+      "step": 4875,
+      "train_speed(iter/s)": 1.33406
+    },
+    {
+      "acc": 0.68592763,
+      "epoch": 0.12379502790461694,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.996130552442237e-06,
+      "loss": 1.40998783,
+      "memory(GiB)": 74.41,
+      "step": 4880,
+      "train_speed(iter/s)": 1.334199
+    },
+    {
+      "acc": 0.6703527,
+      "epoch": 0.123921867072552,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.996089195781902e-06,
+      "loss": 1.38099308,
+      "memory(GiB)": 74.41,
+      "step": 4885,
+      "train_speed(iter/s)": 1.334326
+    },
+    {
+      "acc": 0.66551981,
+      "epoch": 0.12404870624048706,
+      "grad_norm": 3.0,
+      "learning_rate": 9.996047619371256e-06,
+      "loss": 1.50060539,
+      "memory(GiB)": 74.41,
+      "step": 4890,
+      "train_speed(iter/s)": 1.334466
+    },
+    {
+      "acc": 0.6788096,
+      "epoch": 0.12417554540842211,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.996005823212132e-06,
+      "loss": 1.46518192,
+      "memory(GiB)": 74.41,
+      "step": 4895,
+      "train_speed(iter/s)": 1.334596
+    },
+    {
+      "acc": 0.66149321,
+      "epoch": 0.12430238457635717,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.995963807306368e-06,
+      "loss": 1.45551853,
+      "memory(GiB)": 74.41,
+      "step": 4900,
+      "train_speed(iter/s)": 1.334732
+    },
+    {
+      "acc": 0.6776166,
+      "epoch": 0.12442922374429223,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.995921571655808e-06,
+      "loss": 1.46151619,
+      "memory(GiB)": 74.41,
+      "step": 4905,
+      "train_speed(iter/s)": 1.334863
+    },
+    {
+      "acc": 0.66754417,
+      "epoch": 0.12455606291222729,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.995879116262312e-06,
+      "loss": 1.51841745,
+      "memory(GiB)": 74.41,
+      "step": 4910,
+      "train_speed(iter/s)": 1.335001
+    },
+    {
+      "acc": 0.68776064,
+      "epoch": 0.12468290208016235,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.995836441127749e-06,
+      "loss": 1.38561096,
+      "memory(GiB)": 74.41,
+      "step": 4915,
+      "train_speed(iter/s)": 1.335131
+    },
+    {
+      "acc": 0.68441229,
+      "epoch": 0.1248097412480974,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.995793546253993e-06,
+      "loss": 1.40038691,
+      "memory(GiB)": 74.41,
+      "step": 4920,
+      "train_speed(iter/s)": 1.335271
+    },
+    {
+      "acc": 0.68087473,
+      "epoch": 0.12493658041603246,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.995750431642933e-06,
+      "loss": 1.46369896,
+      "memory(GiB)": 74.41,
+      "step": 4925,
+      "train_speed(iter/s)": 1.335401
+    },
+    {
+      "acc": 0.67522473,
+      "epoch": 0.12506341958396752,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.995707097296465e-06,
+      "loss": 1.47813015,
+      "memory(GiB)": 74.41,
+      "step": 4930,
+      "train_speed(iter/s)": 1.335527
+    },
+    {
+      "acc": 0.67564821,
+      "epoch": 0.1251902587519026,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.995663543216493e-06,
+      "loss": 1.50508022,
+      "memory(GiB)": 74.41,
+      "step": 4935,
+      "train_speed(iter/s)": 1.335653
+    },
+    {
+      "acc": 0.68515635,
+      "epoch": 0.12531709791983764,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.995619769404936e-06,
+      "loss": 1.46487827,
+      "memory(GiB)": 74.41,
+      "step": 4940,
+      "train_speed(iter/s)": 1.335779
+    },
+    {
+      "acc": 0.66345377,
+      "epoch": 0.1254439370877727,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.995575775863717e-06,
+      "loss": 1.528825,
+      "memory(GiB)": 74.41,
+      "step": 4945,
+      "train_speed(iter/s)": 1.335901
+    },
+    {
+      "acc": 0.69001646,
+      "epoch": 0.12557077625570776,
+      "grad_norm": 3.5,
+      "learning_rate": 9.995531562594773e-06,
+      "loss": 1.44699345,
+      "memory(GiB)": 74.41,
+      "step": 4950,
+      "train_speed(iter/s)": 1.336014
+    },
+    {
+      "acc": 0.68774014,
+      "epoch": 0.12569761542364283,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.995487129600046e-06,
+      "loss": 1.41482325,
+      "memory(GiB)": 74.41,
+      "step": 4955,
+      "train_speed(iter/s)": 1.33614
+    },
+    {
+      "acc": 0.68150949,
+      "epoch": 0.12582445459157787,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.995442476881491e-06,
+      "loss": 1.43466883,
+      "memory(GiB)": 74.41,
+      "step": 4960,
+      "train_speed(iter/s)": 1.336264
+    },
+    {
+      "acc": 0.67829962,
+      "epoch": 0.12595129375951294,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.995397604441076e-06,
+      "loss": 1.42660933,
+      "memory(GiB)": 74.41,
+      "step": 4965,
+      "train_speed(iter/s)": 1.336387
+    },
+    {
+      "acc": 0.68255768,
+      "epoch": 0.126078132927448,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.995352512280767e-06,
+      "loss": 1.50516148,
+      "memory(GiB)": 74.41,
+      "step": 4970,
+      "train_speed(iter/s)": 1.336507
+    },
+    {
+      "acc": 0.68473272,
+      "epoch": 0.12620497209538306,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.995307200402555e-06,
+      "loss": 1.4734745,
+      "memory(GiB)": 74.41,
+      "step": 4975,
+      "train_speed(iter/s)": 1.336632
+    },
+    {
+      "acc": 0.67716751,
+      "epoch": 0.1263318112633181,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.995261668808429e-06,
+      "loss": 1.40496216,
+      "memory(GiB)": 74.41,
+      "step": 4980,
+      "train_speed(iter/s)": 1.33676
+    },
+    {
+      "acc": 0.68947949,
+      "epoch": 0.12645865043125318,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.995215917500395e-06,
+      "loss": 1.41297464,
+      "memory(GiB)": 74.41,
+      "step": 4985,
+      "train_speed(iter/s)": 1.336878
+    },
+    {
+      "acc": 0.66119204,
+      "epoch": 0.12658548959918822,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.995169946480459e-06,
+      "loss": 1.46938229,
+      "memory(GiB)": 74.41,
+      "step": 4990,
+      "train_speed(iter/s)": 1.337
+    },
+    {
+      "acc": 0.67144785,
+      "epoch": 0.1267123287671233,
+      "grad_norm": 3.5,
+      "learning_rate": 9.99512375575065e-06,
+      "loss": 1.454035,
+      "memory(GiB)": 74.41,
+      "step": 4995,
+      "train_speed(iter/s)": 1.337124
+    },
+    {
+      "acc": 0.67853374,
+      "epoch": 0.12683916793505834,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.995077345312994e-06,
+      "loss": 1.46266308,
+      "memory(GiB)": 74.41,
+      "step": 5000,
+      "train_speed(iter/s)": 1.337246
+    },
+    {
+      "epoch": 0.12683916793505834,
+      "eval_acc": 0.6608386219685627,
+      "eval_loss": 1.4359569549560547,
+      "eval_runtime": 69.8705,
+      "eval_samples_per_second": 91.169,
+      "eval_steps_per_second": 22.799,
+      "step": 5000
+    },
+    {
+      "acc": 0.66414042,
+      "epoch": 0.1269660071029934,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.995030715169535e-06,
+      "loss": 1.51429749,
+      "memory(GiB)": 74.41,
+      "step": 5005,
+      "train_speed(iter/s)": 1.30852
+    },
+    {
+      "acc": 0.66264291,
+      "epoch": 0.12709284627092846,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.994983865322327e-06,
+      "loss": 1.54873114,
+      "memory(GiB)": 74.41,
+      "step": 5010,
+      "train_speed(iter/s)": 1.308663
+    },
+    {
+      "acc": 0.67407112,
+      "epoch": 0.12721968543886353,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.994936795773424e-06,
+      "loss": 1.48143425,
+      "memory(GiB)": 74.41,
+      "step": 5015,
+      "train_speed(iter/s)": 1.308795
+    },
+    {
+      "acc": 0.67631884,
+      "epoch": 0.12734652460679857,
+      "grad_norm": 2.5,
+      "learning_rate": 9.994889506524903e-06,
+      "loss": 1.46913261,
+      "memory(GiB)": 74.41,
+      "step": 5020,
+      "train_speed(iter/s)": 1.308933
+    },
+    {
+      "acc": 0.67536669,
+      "epoch": 0.12747336377473364,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.994841997578839e-06,
+      "loss": 1.50589952,
+      "memory(GiB)": 74.41,
+      "step": 5025,
+      "train_speed(iter/s)": 1.309078
+    },
+    {
+      "acc": 0.67990618,
+      "epoch": 0.1276002029426687,
+      "grad_norm": 2.375,
+      "learning_rate": 9.994794268937325e-06,
+      "loss": 1.48718519,
+      "memory(GiB)": 74.41,
+      "step": 5030,
+      "train_speed(iter/s)": 1.309214
+    },
+    {
+      "acc": 0.66243863,
+      "epoch": 0.12772704211060376,
+      "grad_norm": 4.125,
+      "learning_rate": 9.994746320602457e-06,
+      "loss": 1.52609987,
+      "memory(GiB)": 74.41,
+      "step": 5035,
+      "train_speed(iter/s)": 1.30936
+    },
+    {
+      "acc": 0.679634,
+      "epoch": 0.1278538812785388,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.994698152576347e-06,
+      "loss": 1.50541201,
+      "memory(GiB)": 74.41,
+      "step": 5040,
+      "train_speed(iter/s)": 1.309506
+    },
+    {
+      "acc": 0.6862957,
+      "epoch": 0.12798072044647388,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.994649764861114e-06,
+      "loss": 1.43780212,
+      "memory(GiB)": 74.41,
+      "step": 5045,
+      "train_speed(iter/s)": 1.309651
+    },
+    {
+      "acc": 0.6739593,
+      "epoch": 0.12810755961440892,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.994601157458882e-06,
+      "loss": 1.4502512,
+      "memory(GiB)": 74.41,
+      "step": 5050,
+      "train_speed(iter/s)": 1.309794
+    },
+    {
+      "acc": 0.65548387,
+      "epoch": 0.128234398782344,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.994552330371792e-06,
+      "loss": 1.53416185,
+      "memory(GiB)": 74.41,
+      "step": 5055,
+      "train_speed(iter/s)": 1.309858
+    },
+    {
+      "acc": 0.66185856,
+      "epoch": 0.12836123795027904,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.994503283601993e-06,
+      "loss": 1.49054518,
+      "memory(GiB)": 74.41,
+      "step": 5060,
+      "train_speed(iter/s)": 1.309987
+    },
+    {
+      "acc": 0.66021624,
+      "epoch": 0.1284880771182141,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.99445401715164e-06,
+      "loss": 1.50916061,
+      "memory(GiB)": 74.41,
+      "step": 5065,
+      "train_speed(iter/s)": 1.31013
+    },
+    {
+      "acc": 0.67346339,
+      "epoch": 0.12861491628614916,
+      "grad_norm": 2.875,
+      "learning_rate": 9.994404531022901e-06,
+      "loss": 1.45671196,
+      "memory(GiB)": 74.41,
+      "step": 5070,
+      "train_speed(iter/s)": 1.310272
+    },
+    {
+      "acc": 0.68538084,
+      "epoch": 0.12874175545408423,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.994354825217954e-06,
+      "loss": 1.40544147,
+      "memory(GiB)": 74.41,
+      "step": 5075,
+      "train_speed(iter/s)": 1.310413
+    },
+    {
+      "acc": 0.67029953,
+      "epoch": 0.12886859462201927,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.99430489973898e-06,
+      "loss": 1.48404655,
+      "memory(GiB)": 74.41,
+      "step": 5080,
+      "train_speed(iter/s)": 1.310556
+    },
+    {
+      "acc": 0.65929613,
+      "epoch": 0.12899543378995434,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.994254754588182e-06,
+      "loss": 1.5579051,
+      "memory(GiB)": 74.41,
+      "step": 5085,
+      "train_speed(iter/s)": 1.310698
+    },
+    {
+      "acc": 0.67429667,
+      "epoch": 0.1291222729578894,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.99420438976776e-06,
+      "loss": 1.39907589,
+      "memory(GiB)": 74.41,
+      "step": 5090,
+      "train_speed(iter/s)": 1.310839
+    },
+    {
+      "acc": 0.66815977,
+      "epoch": 0.12924911212582446,
+      "grad_norm": 3.25,
+      "learning_rate": 9.994153805279932e-06,
+      "loss": 1.48352737,
+      "memory(GiB)": 74.41,
+      "step": 5095,
+      "train_speed(iter/s)": 1.310936
+    },
+    {
+      "acc": 0.6815011,
+      "epoch": 0.1293759512937595,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.994103001126923e-06,
+      "loss": 1.40687656,
+      "memory(GiB)": 74.41,
+      "step": 5100,
+      "train_speed(iter/s)": 1.311079
+    },
+    {
+      "acc": 0.67764544,
+      "epoch": 0.12950279046169458,
+      "grad_norm": 3.375,
+      "learning_rate": 9.994051977310966e-06,
+      "loss": 1.44971943,
+      "memory(GiB)": 74.41,
+      "step": 5105,
+      "train_speed(iter/s)": 1.311228
+    },
+    {
+      "acc": 0.65383282,
+      "epoch": 0.12962962962962962,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.994000733834307e-06,
+      "loss": 1.56493654,
+      "memory(GiB)": 74.41,
+      "step": 5110,
+      "train_speed(iter/s)": 1.31137
+    },
+    {
+      "acc": 0.68292179,
+      "epoch": 0.1297564687975647,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.993949270699197e-06,
+      "loss": 1.45260258,
+      "memory(GiB)": 74.41,
+      "step": 5115,
+      "train_speed(iter/s)": 1.311512
+    },
+    {
+      "acc": 0.70052814,
+      "epoch": 0.12988330796549974,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.993897587907904e-06,
+      "loss": 1.39044323,
+      "memory(GiB)": 74.41,
+      "step": 5120,
+      "train_speed(iter/s)": 1.311659
+    },
+    {
+      "acc": 0.66884108,
+      "epoch": 0.1300101471334348,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.993845685462697e-06,
+      "loss": 1.46424484,
+      "memory(GiB)": 74.41,
+      "step": 5125,
+      "train_speed(iter/s)": 1.3118
+    },
+    {
+      "acc": 0.65105453,
+      "epoch": 0.13013698630136986,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.993793563365864e-06,
+      "loss": 1.52029819,
+      "memory(GiB)": 74.41,
+      "step": 5130,
+      "train_speed(iter/s)": 1.311939
+    },
+    {
+      "acc": 0.67307301,
+      "epoch": 0.13026382546930493,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.993741221619692e-06,
+      "loss": 1.45985947,
+      "memory(GiB)": 74.41,
+      "step": 5135,
+      "train_speed(iter/s)": 1.312079
+    },
+    {
+      "acc": 0.67504768,
+      "epoch": 0.13039066463723997,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.993688660226486e-06,
+      "loss": 1.47754765,
+      "memory(GiB)": 74.41,
+      "step": 5140,
+      "train_speed(iter/s)": 1.312214
+    },
+    {
+      "acc": 0.69010491,
+      "epoch": 0.13051750380517504,
+      "grad_norm": 4.0,
+      "learning_rate": 9.993635879188557e-06,
+      "loss": 1.40250454,
+      "memory(GiB)": 74.41,
+      "step": 5145,
+      "train_speed(iter/s)": 1.312335
+    },
+    {
+      "acc": 0.6670228,
+      "epoch": 0.1306443429731101,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.993582878508229e-06,
+      "loss": 1.45732174,
+      "memory(GiB)": 74.41,
+      "step": 5150,
+      "train_speed(iter/s)": 1.312473
+    },
+    {
+      "acc": 0.6658977,
+      "epoch": 0.13077118214104516,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.993529658187829e-06,
+      "loss": 1.49448195,
+      "memory(GiB)": 74.41,
+      "step": 5155,
+      "train_speed(iter/s)": 1.312616
+    },
+    {
+      "acc": 0.6834197,
+      "epoch": 0.1308980213089802,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.9934762182297e-06,
+      "loss": 1.45485058,
+      "memory(GiB)": 74.41,
+      "step": 5160,
+      "train_speed(iter/s)": 1.312759
+    },
+    {
+      "acc": 0.68130121,
+      "epoch": 0.13102486047691528,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.993422558636194e-06,
+      "loss": 1.42725382,
+      "memory(GiB)": 74.41,
+      "step": 5165,
+      "train_speed(iter/s)": 1.312901
+    },
+    {
+      "acc": 0.68009906,
+      "epoch": 0.13115169964485032,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.99336867940967e-06,
+      "loss": 1.44670391,
+      "memory(GiB)": 74.41,
+      "step": 5170,
+      "train_speed(iter/s)": 1.313024
+    },
+    {
+      "acc": 0.68022432,
+      "epoch": 0.1312785388127854,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.993314580552497e-06,
+      "loss": 1.39735098,
+      "memory(GiB)": 74.41,
+      "step": 5175,
+      "train_speed(iter/s)": 1.313161
+    },
+    {
+      "acc": 0.67470751,
+      "epoch": 0.13140537798072044,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.993260262067054e-06,
+      "loss": 1.42325058,
+      "memory(GiB)": 74.41,
+      "step": 5180,
+      "train_speed(iter/s)": 1.313297
+    },
+    {
+      "acc": 0.66762638,
+      "epoch": 0.1315322171486555,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.993205723955734e-06,
+      "loss": 1.49978714,
+      "memory(GiB)": 74.41,
+      "step": 5185,
+      "train_speed(iter/s)": 1.313433
+    },
+    {
+      "acc": 0.68277774,
+      "epoch": 0.13165905631659056,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.993150966220933e-06,
+      "loss": 1.4042243,
+      "memory(GiB)": 74.41,
+      "step": 5190,
+      "train_speed(iter/s)": 1.313571
+    },
+    {
+      "acc": 0.66499376,
+      "epoch": 0.13178589548452563,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.993095988865057e-06,
+      "loss": 1.48225698,
+      "memory(GiB)": 74.41,
+      "step": 5195,
+      "train_speed(iter/s)": 1.31371
+    },
+    {
+      "acc": 0.66776953,
+      "epoch": 0.13191273465246067,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.99304079189053e-06,
+      "loss": 1.51196709,
+      "memory(GiB)": 74.41,
+      "step": 5200,
+      "train_speed(iter/s)": 1.313849
+    },
+    {
+      "acc": 0.67792778,
+      "epoch": 0.13203957382039574,
+      "grad_norm": 6.125,
+      "learning_rate": 9.992985375299775e-06,
+      "loss": 1.41972761,
+      "memory(GiB)": 74.41,
+      "step": 5205,
+      "train_speed(iter/s)": 1.313976
+    },
+    {
+      "acc": 0.66629086,
+      "epoch": 0.1321664129883308,
+      "grad_norm": 4.125,
+      "learning_rate": 9.992929739095232e-06,
+      "loss": 1.47064505,
+      "memory(GiB)": 74.41,
+      "step": 5210,
+      "train_speed(iter/s)": 1.314114
+    },
+    {
+      "acc": 0.66554432,
+      "epoch": 0.13229325215626586,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.992873883279345e-06,
+      "loss": 1.51637993,
+      "memory(GiB)": 74.41,
+      "step": 5215,
+      "train_speed(iter/s)": 1.314244
+    },
+    {
+      "acc": 0.66465168,
+      "epoch": 0.1324200913242009,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.992817807854575e-06,
+      "loss": 1.49373102,
+      "memory(GiB)": 74.41,
+      "step": 5220,
+      "train_speed(iter/s)": 1.314385
+    },
+    {
+      "acc": 0.68335385,
+      "epoch": 0.13254693049213598,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.992761512823386e-06,
+      "loss": 1.46242962,
+      "memory(GiB)": 74.41,
+      "step": 5225,
+      "train_speed(iter/s)": 1.314519
+    },
+    {
+      "acc": 0.67038183,
+      "epoch": 0.13267376966007102,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.992704998188255e-06,
+      "loss": 1.45943947,
+      "memory(GiB)": 74.41,
+      "step": 5230,
+      "train_speed(iter/s)": 1.314653
+    },
+    {
+      "acc": 0.65213909,
+      "epoch": 0.1328006088280061,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.992648263951668e-06,
+      "loss": 1.56592617,
+      "memory(GiB)": 74.41,
+      "step": 5235,
+      "train_speed(iter/s)": 1.314792
+    },
+    {
+      "acc": 0.67219305,
+      "epoch": 0.13292744799594114,
+      "grad_norm": 2.875,
+      "learning_rate": 9.992591310116118e-06,
+      "loss": 1.51808167,
+      "memory(GiB)": 74.41,
+      "step": 5240,
+      "train_speed(iter/s)": 1.314925
+    },
+    {
+      "acc": 0.671525,
+      "epoch": 0.1330542871638762,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.992534136684112e-06,
+      "loss": 1.44265671,
+      "memory(GiB)": 74.41,
+      "step": 5245,
+      "train_speed(iter/s)": 1.315058
+    },
+    {
+      "acc": 0.67576609,
+      "epoch": 0.13318112633181126,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.992476743658165e-06,
+      "loss": 1.44828701,
+      "memory(GiB)": 74.41,
+      "step": 5250,
+      "train_speed(iter/s)": 1.315192
+    },
+    {
+      "acc": 0.68872824,
+      "epoch": 0.13330796549974633,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.992419131040803e-06,
+      "loss": 1.39019938,
+      "memory(GiB)": 74.41,
+      "step": 5255,
+      "train_speed(iter/s)": 1.315324
+    },
+    {
+      "acc": 0.68724899,
+      "epoch": 0.13343480466768137,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.992361298834555e-06,
+      "loss": 1.37071142,
+      "memory(GiB)": 74.41,
+      "step": 5260,
+      "train_speed(iter/s)": 1.315466
+    },
+    {
+      "acc": 0.67438569,
+      "epoch": 0.13356164383561644,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.99230324704197e-06,
+      "loss": 1.51138229,
+      "memory(GiB)": 74.41,
+      "step": 5265,
+      "train_speed(iter/s)": 1.315599
+    },
+    {
+      "acc": 0.67844381,
+      "epoch": 0.1336884830035515,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.992244975665598e-06,
+      "loss": 1.44413147,
+      "memory(GiB)": 74.41,
+      "step": 5270,
+      "train_speed(iter/s)": 1.315737
+    },
+    {
+      "acc": 0.67436996,
+      "epoch": 0.13381532217148656,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.992186484708003e-06,
+      "loss": 1.4860302,
+      "memory(GiB)": 74.41,
+      "step": 5275,
+      "train_speed(iter/s)": 1.315871
+    },
+    {
+      "acc": 0.68284712,
+      "epoch": 0.1339421613394216,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.992127774171759e-06,
+      "loss": 1.40780163,
+      "memory(GiB)": 74.41,
+      "step": 5280,
+      "train_speed(iter/s)": 1.316007
+    },
+    {
+      "acc": 0.66165814,
+      "epoch": 0.13406900050735668,
+      "grad_norm": 3.25,
+      "learning_rate": 9.992068844059446e-06,
+      "loss": 1.59108582,
+      "memory(GiB)": 74.41,
+      "step": 5285,
+      "train_speed(iter/s)": 1.316144
+    },
+    {
+      "acc": 0.66397309,
+      "epoch": 0.13419583967529172,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.992009694373658e-06,
+      "loss": 1.47345829,
+      "memory(GiB)": 74.41,
+      "step": 5290,
+      "train_speed(iter/s)": 1.316272
+    },
+    {
+      "acc": 0.67620754,
+      "epoch": 0.1343226788432268,
+      "grad_norm": 3.875,
+      "learning_rate": 9.991950325116995e-06,
+      "loss": 1.47917576,
+      "memory(GiB)": 74.41,
+      "step": 5295,
+      "train_speed(iter/s)": 1.316407
+    },
+    {
+      "acc": 0.66644521,
+      "epoch": 0.13444951801116184,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.99189073629207e-06,
+      "loss": 1.44303188,
+      "memory(GiB)": 74.41,
+      "step": 5300,
+      "train_speed(iter/s)": 1.316533
+    },
+    {
+      "acc": 0.66349211,
+      "epoch": 0.1345763571790969,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.991830927901505e-06,
+      "loss": 1.56156445,
+      "memory(GiB)": 74.41,
+      "step": 5305,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.6750639,
+      "epoch": 0.13470319634703196,
+      "grad_norm": 3.125,
+      "learning_rate": 9.991770899947925e-06,
+      "loss": 1.41290503,
+      "memory(GiB)": 74.41,
+      "step": 5310,
+      "train_speed(iter/s)": 1.316792
+    },
+    {
+      "acc": 0.67474327,
+      "epoch": 0.13483003551496703,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.991710652433977e-06,
+      "loss": 1.4732769,
+      "memory(GiB)": 74.41,
+      "step": 5315,
+      "train_speed(iter/s)": 1.316923
+    },
+    {
+      "acc": 0.67381182,
+      "epoch": 0.13495687468290207,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.991650185362308e-06,
+      "loss": 1.47163076,
+      "memory(GiB)": 74.41,
+      "step": 5320,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.67389021,
+      "epoch": 0.13508371385083714,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.991589498735577e-06,
+      "loss": 1.43954639,
+      "memory(GiB)": 74.41,
+      "step": 5325,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.68659286,
+      "epoch": 0.1352105530187722,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.991528592556454e-06,
+      "loss": 1.40721159,
+      "memory(GiB)": 74.41,
+      "step": 5330,
+      "train_speed(iter/s)": 1.317321
+    },
+    {
+      "acc": 0.66312041,
+      "epoch": 0.13533739218670726,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.991467466827618e-06,
+      "loss": 1.54625053,
+      "memory(GiB)": 74.41,
+      "step": 5335,
+      "train_speed(iter/s)": 1.317452
+    },
+    {
+      "acc": 0.68024783,
+      "epoch": 0.1354642313546423,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.99140612155176e-06,
+      "loss": 1.48221207,
+      "memory(GiB)": 74.41,
+      "step": 5340,
+      "train_speed(iter/s)": 1.317582
+    },
+    {
+      "acc": 0.66318727,
+      "epoch": 0.13559107052257738,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.991344556731572e-06,
+      "loss": 1.50610676,
+      "memory(GiB)": 74.41,
+      "step": 5345,
+      "train_speed(iter/s)": 1.317715
+    },
+    {
+      "acc": 0.68796473,
+      "epoch": 0.13571790969051242,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.991282772369766e-06,
+      "loss": 1.4288991,
+      "memory(GiB)": 74.41,
+      "step": 5350,
+      "train_speed(iter/s)": 1.317849
+    },
+    {
+      "acc": 0.67604733,
+      "epoch": 0.1358447488584475,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.99122076846906e-06,
+      "loss": 1.4361022,
+      "memory(GiB)": 74.41,
+      "step": 5355,
+      "train_speed(iter/s)": 1.317979
+    },
+    {
+      "acc": 0.68241739,
+      "epoch": 0.13597158802638254,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.991158545032181e-06,
+      "loss": 1.45159664,
+      "memory(GiB)": 74.41,
+      "step": 5360,
+      "train_speed(iter/s)": 1.318111
+    },
+    {
+      "acc": 0.68970242,
+      "epoch": 0.1360984271943176,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.991096102061865e-06,
+      "loss": 1.40896711,
+      "memory(GiB)": 74.41,
+      "step": 5365,
+      "train_speed(iter/s)": 1.318249
+    },
+    {
+      "acc": 0.67420306,
+      "epoch": 0.13622526636225266,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.991033439560858e-06,
+      "loss": 1.47338696,
+      "memory(GiB)": 74.41,
+      "step": 5370,
+      "train_speed(iter/s)": 1.318381
+    },
+    {
+      "acc": 0.66488967,
+      "epoch": 0.13635210553018773,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.990970557531918e-06,
+      "loss": 1.55452347,
+      "memory(GiB)": 74.41,
+      "step": 5375,
+      "train_speed(iter/s)": 1.318513
+    },
+    {
+      "acc": 0.68728437,
+      "epoch": 0.13647894469812277,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.990907455977809e-06,
+      "loss": 1.4196394,
+      "memory(GiB)": 74.41,
+      "step": 5380,
+      "train_speed(iter/s)": 1.318643
+    },
+    {
+      "acc": 0.6631248,
+      "epoch": 0.13660578386605784,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.990844134901308e-06,
+      "loss": 1.51547356,
+      "memory(GiB)": 74.41,
+      "step": 5385,
+      "train_speed(iter/s)": 1.31878
+    },
+    {
+      "acc": 0.69585829,
+      "epoch": 0.1367326230339929,
+      "grad_norm": 3.125,
+      "learning_rate": 9.9907805943052e-06,
+      "loss": 1.3806735,
+      "memory(GiB)": 74.41,
+      "step": 5390,
+      "train_speed(iter/s)": 1.318909
+    },
+    {
+      "acc": 0.66924019,
+      "epoch": 0.13685946220192796,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.990716834192278e-06,
+      "loss": 1.40786781,
+      "memory(GiB)": 74.41,
+      "step": 5395,
+      "train_speed(iter/s)": 1.319037
+    },
+    {
+      "acc": 0.66605959,
+      "epoch": 0.136986301369863,
+      "grad_norm": 3.125,
+      "learning_rate": 9.990652854565348e-06,
+      "loss": 1.4843646,
+      "memory(GiB)": 74.41,
+      "step": 5400,
+      "train_speed(iter/s)": 1.319167
+    },
+    {
+      "acc": 0.66530266,
+      "epoch": 0.13711314053779808,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.990588655427225e-06,
+      "loss": 1.55083303,
+      "memory(GiB)": 74.41,
+      "step": 5405,
+      "train_speed(iter/s)": 1.319294
+    },
+    {
+      "acc": 0.68000975,
+      "epoch": 0.13723997970573312,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.99052423678073e-06,
+      "loss": 1.46145096,
+      "memory(GiB)": 74.41,
+      "step": 5410,
+      "train_speed(iter/s)": 1.319425
+    },
+    {
+      "acc": 0.68537655,
+      "epoch": 0.1373668188736682,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.990459598628697e-06,
+      "loss": 1.40474434,
+      "memory(GiB)": 74.41,
+      "step": 5415,
+      "train_speed(iter/s)": 1.319559
+    },
+    {
+      "acc": 0.67669477,
+      "epoch": 0.13749365804160324,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.990394740973972e-06,
+      "loss": 1.38833351,
+      "memory(GiB)": 74.41,
+      "step": 5420,
+      "train_speed(iter/s)": 1.31968
+    },
+    {
+      "acc": 0.68024025,
+      "epoch": 0.1376204972095383,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.990329663819405e-06,
+      "loss": 1.50072689,
+      "memory(GiB)": 74.41,
+      "step": 5425,
+      "train_speed(iter/s)": 1.319811
+    },
+    {
+      "acc": 0.67766657,
+      "epoch": 0.13774733637747336,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.99026436716786e-06,
+      "loss": 1.42446575,
+      "memory(GiB)": 74.41,
+      "step": 5430,
+      "train_speed(iter/s)": 1.319936
+    },
+    {
+      "acc": 0.67358165,
+      "epoch": 0.13787417554540843,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.990198851022207e-06,
+      "loss": 1.46434498,
+      "memory(GiB)": 74.41,
+      "step": 5435,
+      "train_speed(iter/s)": 1.320066
+    },
+    {
+      "acc": 0.6777771,
+      "epoch": 0.13800101471334347,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.99013311538533e-06,
+      "loss": 1.44285946,
+      "memory(GiB)": 74.41,
+      "step": 5440,
+      "train_speed(iter/s)": 1.320194
+    },
+    {
+      "acc": 0.68178253,
+      "epoch": 0.13812785388127855,
+      "grad_norm": 3.375,
+      "learning_rate": 9.99006716026012e-06,
+      "loss": 1.48958969,
+      "memory(GiB)": 74.41,
+      "step": 5445,
+      "train_speed(iter/s)": 1.320306
+    },
+    {
+      "acc": 0.67343578,
+      "epoch": 0.1382546930492136,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.990000985649475e-06,
+      "loss": 1.49231796,
+      "memory(GiB)": 74.41,
+      "step": 5450,
+      "train_speed(iter/s)": 1.320432
+    },
+    {
+      "acc": 0.66956115,
+      "epoch": 0.13838153221714866,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.989934591556308e-06,
+      "loss": 1.44769735,
+      "memory(GiB)": 74.41,
+      "step": 5455,
+      "train_speed(iter/s)": 1.320559
+    },
+    {
+      "acc": 0.66259384,
+      "epoch": 0.1385083713850837,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.98986797798354e-06,
+      "loss": 1.52780714,
+      "memory(GiB)": 74.41,
+      "step": 5460,
+      "train_speed(iter/s)": 1.320685
+    },
+    {
+      "acc": 0.67124734,
+      "epoch": 0.13863521055301878,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.989801144934102e-06,
+      "loss": 1.44861336,
+      "memory(GiB)": 74.41,
+      "step": 5465,
+      "train_speed(iter/s)": 1.320811
+    },
+    {
+      "acc": 0.6817421,
+      "epoch": 0.13876204972095382,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.98973409241093e-06,
+      "loss": 1.45624142,
+      "memory(GiB)": 74.41,
+      "step": 5470,
+      "train_speed(iter/s)": 1.320943
+    },
+    {
+      "acc": 0.66689086,
+      "epoch": 0.1388888888888889,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.989666820416974e-06,
+      "loss": 1.48923454,
+      "memory(GiB)": 74.41,
+      "step": 5475,
+      "train_speed(iter/s)": 1.321069
+    },
+    {
+      "acc": 0.67675085,
+      "epoch": 0.13901572805682394,
+      "grad_norm": 4.90625,
+      "learning_rate": 9.989599328955195e-06,
+      "loss": 1.51229401,
+      "memory(GiB)": 74.41,
+      "step": 5480,
+      "train_speed(iter/s)": 1.321194
+    },
+    {
+      "acc": 0.67773046,
+      "epoch": 0.139142567224759,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.98953161802856e-06,
+      "loss": 1.44271612,
+      "memory(GiB)": 74.41,
+      "step": 5485,
+      "train_speed(iter/s)": 1.321311
+    },
+    {
+      "acc": 0.69303894,
+      "epoch": 0.13926940639269406,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.98946368764005e-06,
+      "loss": 1.39192066,
+      "memory(GiB)": 74.41,
+      "step": 5490,
+      "train_speed(iter/s)": 1.321436
+    },
+    {
+      "acc": 0.69491491,
+      "epoch": 0.13939624556062913,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.989395537792647e-06,
+      "loss": 1.43225937,
+      "memory(GiB)": 74.41,
+      "step": 5495,
+      "train_speed(iter/s)": 1.321565
+    },
+    {
+      "acc": 0.66719732,
+      "epoch": 0.13952308472856417,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.989327168489356e-06,
+      "loss": 1.4718915,
+      "memory(GiB)": 74.41,
+      "step": 5500,
+      "train_speed(iter/s)": 1.321693
+    },
+    {
+      "acc": 0.67620239,
+      "epoch": 0.13964992389649925,
+      "grad_norm": 2.875,
+      "learning_rate": 9.989258579733179e-06,
+      "loss": 1.47505198,
+      "memory(GiB)": 74.41,
+      "step": 5505,
+      "train_speed(iter/s)": 1.321821
+    },
+    {
+      "acc": 0.668994,
+      "epoch": 0.1397767630644343,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.989189771527133e-06,
+      "loss": 1.43469181,
+      "memory(GiB)": 74.41,
+      "step": 5510,
+      "train_speed(iter/s)": 1.321953
+    },
+    {
+      "acc": 0.67099843,
+      "epoch": 0.13990360223236936,
+      "grad_norm": 2.875,
+      "learning_rate": 9.989120743874248e-06,
+      "loss": 1.48247786,
+      "memory(GiB)": 74.41,
+      "step": 5515,
+      "train_speed(iter/s)": 1.322081
+    },
+    {
+      "acc": 0.66891327,
+      "epoch": 0.1400304414003044,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.989051496777556e-06,
+      "loss": 1.53904562,
+      "memory(GiB)": 74.41,
+      "step": 5520,
+      "train_speed(iter/s)": 1.32221
+    },
+    {
+      "acc": 0.66619916,
+      "epoch": 0.14015728056823948,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.988982030240104e-06,
+      "loss": 1.50380411,
+      "memory(GiB)": 74.41,
+      "step": 5525,
+      "train_speed(iter/s)": 1.322331
+    },
+    {
+      "acc": 0.6741478,
+      "epoch": 0.14028411973617452,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.988912344264949e-06,
+      "loss": 1.5130826,
+      "memory(GiB)": 74.41,
+      "step": 5530,
+      "train_speed(iter/s)": 1.322457
+    },
+    {
+      "acc": 0.67934475,
+      "epoch": 0.1404109589041096,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.988842438855156e-06,
+      "loss": 1.40677891,
+      "memory(GiB)": 74.41,
+      "step": 5535,
+      "train_speed(iter/s)": 1.322582
+    },
+    {
+      "acc": 0.66568899,
+      "epoch": 0.14053779807204464,
+      "grad_norm": 3.625,
+      "learning_rate": 9.988772314013799e-06,
+      "loss": 1.47276525,
+      "memory(GiB)": 74.41,
+      "step": 5540,
+      "train_speed(iter/s)": 1.322709
+    },
+    {
+      "acc": 0.67035012,
+      "epoch": 0.1406646372399797,
+      "grad_norm": 3.125,
+      "learning_rate": 9.988701969743961e-06,
+      "loss": 1.45721273,
+      "memory(GiB)": 74.41,
+      "step": 5545,
+      "train_speed(iter/s)": 1.322842
+    },
+    {
+      "acc": 0.67886105,
+      "epoch": 0.14079147640791476,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.98863140604874e-06,
+      "loss": 1.47562962,
+      "memory(GiB)": 74.41,
+      "step": 5550,
+      "train_speed(iter/s)": 1.322969
+    },
+    {
+      "acc": 0.67969818,
+      "epoch": 0.14091831557584983,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.988560622931233e-06,
+      "loss": 1.44630022,
+      "memory(GiB)": 74.41,
+      "step": 5555,
+      "train_speed(iter/s)": 1.323087
+    },
+    {
+      "acc": 0.69383931,
+      "epoch": 0.14104515474378487,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.988489620394562e-06,
+      "loss": 1.42455463,
+      "memory(GiB)": 74.41,
+      "step": 5560,
+      "train_speed(iter/s)": 1.323216
+    },
+    {
+      "acc": 0.68613052,
+      "epoch": 0.14117199391171995,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.988418398441842e-06,
+      "loss": 1.41271257,
+      "memory(GiB)": 74.41,
+      "step": 5565,
+      "train_speed(iter/s)": 1.323341
+    },
+    {
+      "acc": 0.65636492,
+      "epoch": 0.141298833079655,
+      "grad_norm": 3.25,
+      "learning_rate": 9.98834695707621e-06,
+      "loss": 1.55587215,
+      "memory(GiB)": 74.41,
+      "step": 5570,
+      "train_speed(iter/s)": 1.323473
+    },
+    {
+      "acc": 0.66223726,
+      "epoch": 0.14142567224759006,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.98827529630081e-06,
+      "loss": 1.48361416,
+      "memory(GiB)": 74.41,
+      "step": 5575,
+      "train_speed(iter/s)": 1.323609
+    },
+    {
+      "acc": 0.68168488,
+      "epoch": 0.1415525114155251,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.988203416118788e-06,
+      "loss": 1.39744835,
+      "memory(GiB)": 74.41,
+      "step": 5580,
+      "train_speed(iter/s)": 1.323737
+    },
+    {
+      "acc": 0.68375525,
+      "epoch": 0.14167935058346018,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.98813131653331e-06,
+      "loss": 1.4359664,
+      "memory(GiB)": 74.41,
+      "step": 5585,
+      "train_speed(iter/s)": 1.323869
+    },
+    {
+      "acc": 0.66030507,
+      "epoch": 0.14180618975139522,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.988058997547548e-06,
+      "loss": 1.46749029,
+      "memory(GiB)": 74.41,
+      "step": 5590,
+      "train_speed(iter/s)": 1.323994
+    },
+    {
+      "acc": 0.66670837,
+      "epoch": 0.1419330289193303,
+      "grad_norm": 4.75,
+      "learning_rate": 9.987986459164678e-06,
+      "loss": 1.55378685,
+      "memory(GiB)": 74.41,
+      "step": 5595,
+      "train_speed(iter/s)": 1.324127
+    },
+    {
+      "acc": 0.67393637,
+      "epoch": 0.14205986808726534,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.987913701387897e-06,
+      "loss": 1.4591711,
+      "memory(GiB)": 74.41,
+      "step": 5600,
+      "train_speed(iter/s)": 1.324256
+    },
+    {
+      "acc": 0.66163654,
+      "epoch": 0.1421867072552004,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.9878407242204e-06,
+      "loss": 1.50209656,
+      "memory(GiB)": 74.41,
+      "step": 5605,
+      "train_speed(iter/s)": 1.324384
+    },
+    {
+      "acc": 0.67107725,
+      "epoch": 0.14231354642313546,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.9877675276654e-06,
+      "loss": 1.43066006,
+      "memory(GiB)": 74.41,
+      "step": 5610,
+      "train_speed(iter/s)": 1.324511
+    },
+    {
+      "acc": 0.68052077,
+      "epoch": 0.14244038559107053,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.987694111726114e-06,
+      "loss": 1.42242727,
+      "memory(GiB)": 74.41,
+      "step": 5615,
+      "train_speed(iter/s)": 1.324638
+    },
+    {
+      "acc": 0.68275967,
+      "epoch": 0.14256722475900557,
+      "grad_norm": 2.75,
+      "learning_rate": 9.987620476405774e-06,
+      "loss": 1.42960892,
+      "memory(GiB)": 74.41,
+      "step": 5620,
+      "train_speed(iter/s)": 1.324768
+    },
+    {
+      "acc": 0.68630466,
+      "epoch": 0.14269406392694065,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.987546621707616e-06,
+      "loss": 1.43038988,
+      "memory(GiB)": 74.41,
+      "step": 5625,
+      "train_speed(iter/s)": 1.324889
+    },
+    {
+      "acc": 0.66758313,
+      "epoch": 0.1428209030948757,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.98747254763489e-06,
+      "loss": 1.48348064,
+      "memory(GiB)": 74.41,
+      "step": 5630,
+      "train_speed(iter/s)": 1.325015
+    },
+    {
+      "acc": 0.66321049,
+      "epoch": 0.14294774226281076,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.987398254190855e-06,
+      "loss": 1.51671391,
+      "memory(GiB)": 74.41,
+      "step": 5635,
+      "train_speed(iter/s)": 1.325141
+    },
+    {
+      "acc": 0.67576532,
+      "epoch": 0.1430745814307458,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.987323741378777e-06,
+      "loss": 1.49405441,
+      "memory(GiB)": 74.41,
+      "step": 5640,
+      "train_speed(iter/s)": 1.325265
+    },
+    {
+      "acc": 0.69233437,
+      "epoch": 0.14320142059868088,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.987249009201934e-06,
+      "loss": 1.40569649,
+      "memory(GiB)": 74.41,
+      "step": 5645,
+      "train_speed(iter/s)": 1.325392
+    },
+    {
+      "acc": 0.68413944,
+      "epoch": 0.14332825976661592,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.987174057663613e-06,
+      "loss": 1.42640572,
+      "memory(GiB)": 74.41,
+      "step": 5650,
+      "train_speed(iter/s)": 1.32552
+    },
+    {
+      "acc": 0.67004395,
+      "epoch": 0.143455098934551,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.987098886767111e-06,
+      "loss": 1.46639032,
+      "memory(GiB)": 74.41,
+      "step": 5655,
+      "train_speed(iter/s)": 1.325648
+    },
+    {
+      "acc": 0.66027412,
+      "epoch": 0.14358193810248604,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.987023496515734e-06,
+      "loss": 1.54882936,
+      "memory(GiB)": 74.41,
+      "step": 5660,
+      "train_speed(iter/s)": 1.325763
+    },
+    {
+      "acc": 0.67224212,
+      "epoch": 0.1437087772704211,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.9869478869128e-06,
+      "loss": 1.47377834,
+      "memory(GiB)": 74.41,
+      "step": 5665,
+      "train_speed(iter/s)": 1.325893
+    },
+    {
+      "acc": 0.66580467,
+      "epoch": 0.14383561643835616,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.98687205796163e-06,
+      "loss": 1.51950064,
+      "memory(GiB)": 74.41,
+      "step": 5670,
+      "train_speed(iter/s)": 1.325985
+    },
+    {
+      "acc": 0.67456326,
+      "epoch": 0.14396245560629123,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.986796009665562e-06,
+      "loss": 1.40329494,
+      "memory(GiB)": 74.41,
+      "step": 5675,
+      "train_speed(iter/s)": 1.32611
+    },
+    {
+      "acc": 0.67849097,
+      "epoch": 0.14408929477422627,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.986719742027944e-06,
+      "loss": 1.39257736,
+      "memory(GiB)": 74.41,
+      "step": 5680,
+      "train_speed(iter/s)": 1.326235
+    },
+    {
+      "acc": 0.67778139,
+      "epoch": 0.14421613394216135,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.986643255052125e-06,
+      "loss": 1.43970556,
+      "memory(GiB)": 74.41,
+      "step": 5685,
+      "train_speed(iter/s)": 1.326364
+    },
+    {
+      "acc": 0.67100382,
+      "epoch": 0.1443429731100964,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.986566548741473e-06,
+      "loss": 1.44389887,
+      "memory(GiB)": 74.41,
+      "step": 5690,
+      "train_speed(iter/s)": 1.326491
+    },
+    {
+      "acc": 0.68498144,
+      "epoch": 0.14446981227803146,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.98648962309936e-06,
+      "loss": 1.44494476,
+      "memory(GiB)": 74.41,
+      "step": 5695,
+      "train_speed(iter/s)": 1.326614
+    },
+    {
+      "acc": 0.68308759,
+      "epoch": 0.1445966514459665,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.986412478129171e-06,
+      "loss": 1.3816658,
+      "memory(GiB)": 74.41,
+      "step": 5700,
+      "train_speed(iter/s)": 1.326735
+    },
+    {
+      "acc": 0.67953053,
+      "epoch": 0.14472349061390158,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.9863351138343e-06,
+      "loss": 1.42906961,
+      "memory(GiB)": 74.41,
+      "step": 5705,
+      "train_speed(iter/s)": 1.326861
+    },
+    {
+      "acc": 0.68642354,
+      "epoch": 0.14485032978183662,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.986257530218146e-06,
+      "loss": 1.40588169,
+      "memory(GiB)": 74.41,
+      "step": 5710,
+      "train_speed(iter/s)": 1.326989
+    },
+    {
+      "acc": 0.67285957,
+      "epoch": 0.1449771689497717,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.986179727284124e-06,
+      "loss": 1.4339426,
+      "memory(GiB)": 74.41,
+      "step": 5715,
+      "train_speed(iter/s)": 1.327115
+    },
+    {
+      "acc": 0.65860806,
+      "epoch": 0.14510400811770674,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.986101705035656e-06,
+      "loss": 1.50875168,
+      "memory(GiB)": 74.41,
+      "step": 5720,
+      "train_speed(iter/s)": 1.327233
+    },
+    {
+      "acc": 0.66784,
+      "epoch": 0.1452308472856418,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.986023463476175e-06,
+      "loss": 1.52321329,
+      "memory(GiB)": 74.41,
+      "step": 5725,
+      "train_speed(iter/s)": 1.327357
+    },
+    {
+      "acc": 0.66872253,
+      "epoch": 0.14535768645357686,
+      "grad_norm": 3.125,
+      "learning_rate": 9.985945002609119e-06,
+      "loss": 1.49432068,
+      "memory(GiB)": 74.41,
+      "step": 5730,
+      "train_speed(iter/s)": 1.327482
+    },
+    {
+      "acc": 0.67970715,
+      "epoch": 0.14548452562151193,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.985866322437942e-06,
+      "loss": 1.52032089,
+      "memory(GiB)": 74.41,
+      "step": 5735,
+      "train_speed(iter/s)": 1.32761
+    },
+    {
+      "acc": 0.68226662,
+      "epoch": 0.14561136478944697,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.985787422966105e-06,
+      "loss": 1.4466094,
+      "memory(GiB)": 74.41,
+      "step": 5740,
+      "train_speed(iter/s)": 1.327732
+    },
+    {
+      "acc": 0.67653313,
+      "epoch": 0.14573820395738205,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.985708304197075e-06,
+      "loss": 1.54679623,
+      "memory(GiB)": 74.41,
+      "step": 5745,
+      "train_speed(iter/s)": 1.327853
+    },
+    {
+      "acc": 0.67613993,
+      "epoch": 0.1458650431253171,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.985628966134336e-06,
+      "loss": 1.46932364,
+      "memory(GiB)": 74.41,
+      "step": 5750,
+      "train_speed(iter/s)": 1.32798
+    },
+    {
+      "acc": 0.67247124,
+      "epoch": 0.14599188229325216,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.985549408781377e-06,
+      "loss": 1.44778271,
+      "memory(GiB)": 74.41,
+      "step": 5755,
+      "train_speed(iter/s)": 1.328087
+    },
+    {
+      "acc": 0.66912937,
+      "epoch": 0.1461187214611872,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.985469632141693e-06,
+      "loss": 1.50241432,
+      "memory(GiB)": 74.41,
+      "step": 5760,
+      "train_speed(iter/s)": 1.328209
+    },
+    {
+      "acc": 0.66960969,
+      "epoch": 0.14624556062912228,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.985389636218797e-06,
+      "loss": 1.4625102,
+      "memory(GiB)": 74.41,
+      "step": 5765,
+      "train_speed(iter/s)": 1.328323
+    },
+    {
+      "acc": 0.67095127,
+      "epoch": 0.14637239979705732,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.985309421016207e-06,
+      "loss": 1.4584774,
+      "memory(GiB)": 74.41,
+      "step": 5770,
+      "train_speed(iter/s)": 1.328444
+    },
+    {
+      "acc": 0.6832509,
+      "epoch": 0.1464992389649924,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.985228986537451e-06,
+      "loss": 1.43589792,
+      "memory(GiB)": 74.41,
+      "step": 5775,
+      "train_speed(iter/s)": 1.328568
+    },
+    {
+      "acc": 0.67679062,
+      "epoch": 0.14662607813292744,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.985148332786068e-06,
+      "loss": 1.41989422,
+      "memory(GiB)": 74.41,
+      "step": 5780,
+      "train_speed(iter/s)": 1.328692
+    },
+    {
+      "acc": 0.67695398,
+      "epoch": 0.1467529173008625,
+      "grad_norm": 3.375,
+      "learning_rate": 9.985067459765603e-06,
+      "loss": 1.48500957,
+      "memory(GiB)": 74.41,
+      "step": 5785,
+      "train_speed(iter/s)": 1.328816
+    },
+    {
+      "acc": 0.67666731,
+      "epoch": 0.14687975646879756,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.984986367479615e-06,
+      "loss": 1.4901082,
+      "memory(GiB)": 74.41,
+      "step": 5790,
+      "train_speed(iter/s)": 1.328931
+    },
+    {
+      "acc": 0.68309522,
+      "epoch": 0.14700659563673263,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.984905055931668e-06,
+      "loss": 1.43481121,
+      "memory(GiB)": 74.41,
+      "step": 5795,
+      "train_speed(iter/s)": 1.329049
+    },
+    {
+      "acc": 0.65112691,
+      "epoch": 0.14713343480466767,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.984823525125342e-06,
+      "loss": 1.52671738,
+      "memory(GiB)": 74.41,
+      "step": 5800,
+      "train_speed(iter/s)": 1.329171
+    },
+    {
+      "acc": 0.66602678,
+      "epoch": 0.14726027397260275,
+      "grad_norm": 3.125,
+      "learning_rate": 9.984741775064222e-06,
+      "loss": 1.49231052,
+      "memory(GiB)": 74.41,
+      "step": 5805,
+      "train_speed(iter/s)": 1.329288
+    },
+    {
+      "acc": 0.6663784,
+      "epoch": 0.1473871131405378,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.984659805751904e-06,
+      "loss": 1.5365591,
+      "memory(GiB)": 74.41,
+      "step": 5810,
+      "train_speed(iter/s)": 1.329401
+    },
+    {
+      "acc": 0.68376894,
+      "epoch": 0.14751395230847286,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.984577617191993e-06,
+      "loss": 1.43436604,
+      "memory(GiB)": 74.41,
+      "step": 5815,
+      "train_speed(iter/s)": 1.329513
+    },
+    {
+      "acc": 0.66587582,
+      "epoch": 0.1476407914764079,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.984495209388102e-06,
+      "loss": 1.51071091,
+      "memory(GiB)": 74.41,
+      "step": 5820,
+      "train_speed(iter/s)": 1.32963
+    },
+    {
+      "acc": 0.66028914,
+      "epoch": 0.14776763064434298,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.984412582343859e-06,
+      "loss": 1.52795696,
+      "memory(GiB)": 74.41,
+      "step": 5825,
+      "train_speed(iter/s)": 1.329747
+    },
+    {
+      "acc": 0.6772841,
+      "epoch": 0.14789446981227802,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.984329736062896e-06,
+      "loss": 1.4106348,
+      "memory(GiB)": 74.41,
+      "step": 5830,
+      "train_speed(iter/s)": 1.329863
+    },
+    {
+      "acc": 0.66106005,
+      "epoch": 0.1480213089802131,
+      "grad_norm": 2.875,
+      "learning_rate": 9.984246670548858e-06,
+      "loss": 1.48321609,
+      "memory(GiB)": 74.41,
+      "step": 5835,
+      "train_speed(iter/s)": 1.329977
+    },
+    {
+      "acc": 0.67662067,
+      "epoch": 0.14814814814814814,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.984163385805398e-06,
+      "loss": 1.51368103,
+      "memory(GiB)": 74.41,
+      "step": 5840,
+      "train_speed(iter/s)": 1.330088
+    },
+    {
+      "acc": 0.67393694,
+      "epoch": 0.1482749873160832,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.984079881836182e-06,
+      "loss": 1.48277817,
+      "memory(GiB)": 74.41,
+      "step": 5845,
+      "train_speed(iter/s)": 1.330207
+    },
+    {
+      "acc": 0.67487364,
+      "epoch": 0.14840182648401826,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.983996158644877e-06,
+      "loss": 1.43262262,
+      "memory(GiB)": 74.41,
+      "step": 5850,
+      "train_speed(iter/s)": 1.330317
+    },
+    {
+      "acc": 0.67476578,
+      "epoch": 0.14852866565195333,
+      "grad_norm": 2.5,
+      "learning_rate": 9.983912216235172e-06,
+      "loss": 1.38087339,
+      "memory(GiB)": 74.41,
+      "step": 5855,
+      "train_speed(iter/s)": 1.330432
+    },
+    {
+      "acc": 0.68206239,
+      "epoch": 0.14865550481988837,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.983828054610754e-06,
+      "loss": 1.47352724,
+      "memory(GiB)": 74.41,
+      "step": 5860,
+      "train_speed(iter/s)": 1.330545
+    },
+    {
+      "acc": 0.68086476,
+      "epoch": 0.14878234398782345,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.983743673775328e-06,
+      "loss": 1.49459362,
+      "memory(GiB)": 74.41,
+      "step": 5865,
+      "train_speed(iter/s)": 1.330644
+    },
+    {
+      "acc": 0.67520857,
+      "epoch": 0.1489091831557585,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.983659073732604e-06,
+      "loss": 1.45500126,
+      "memory(GiB)": 74.41,
+      "step": 5870,
+      "train_speed(iter/s)": 1.330764
+    },
+    {
+      "acc": 0.67538538,
+      "epoch": 0.14903602232369356,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.983574254486303e-06,
+      "loss": 1.46491451,
+      "memory(GiB)": 74.41,
+      "step": 5875,
+      "train_speed(iter/s)": 1.330877
+    },
+    {
+      "acc": 0.68383317,
+      "epoch": 0.1491628614916286,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.983489216040158e-06,
+      "loss": 1.45136061,
+      "memory(GiB)": 74.41,
+      "step": 5880,
+      "train_speed(iter/s)": 1.330989
+    },
+    {
+      "acc": 0.68471317,
+      "epoch": 0.14928970065956368,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.983403958397907e-06,
+      "loss": 1.44710627,
+      "memory(GiB)": 74.41,
+      "step": 5885,
+      "train_speed(iter/s)": 1.331105
+    },
+    {
+      "acc": 0.67280602,
+      "epoch": 0.14941653982749872,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.9833184815633e-06,
+      "loss": 1.62267265,
+      "memory(GiB)": 74.41,
+      "step": 5890,
+      "train_speed(iter/s)": 1.331217
+    },
+    {
+      "acc": 0.66782923,
+      "epoch": 0.1495433789954338,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.983232785540097e-06,
+      "loss": 1.55785322,
+      "memory(GiB)": 74.41,
+      "step": 5895,
+      "train_speed(iter/s)": 1.331313
+    },
+    {
+      "acc": 0.66646562,
+      "epoch": 0.14967021816336884,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.983146870332068e-06,
+      "loss": 1.47144756,
+      "memory(GiB)": 74.41,
+      "step": 5900,
+      "train_speed(iter/s)": 1.331427
+    },
+    {
+      "acc": 0.6738615,
+      "epoch": 0.1497970573313039,
+      "grad_norm": 5.75,
+      "learning_rate": 9.98306073594299e-06,
+      "loss": 1.50128069,
+      "memory(GiB)": 74.41,
+      "step": 5905,
+      "train_speed(iter/s)": 1.331539
+    },
+    {
+      "acc": 0.69000626,
+      "epoch": 0.14992389649923896,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.982974382376656e-06,
+      "loss": 1.42517776,
+      "memory(GiB)": 74.41,
+      "step": 5910,
+      "train_speed(iter/s)": 1.331647
+    },
+    {
+      "acc": 0.68917923,
+      "epoch": 0.15005073566717403,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.98288780963686e-06,
+      "loss": 1.4264616,
+      "memory(GiB)": 74.41,
+      "step": 5915,
+      "train_speed(iter/s)": 1.33175
+    },
+    {
+      "acc": 0.67096195,
+      "epoch": 0.15017757483510907,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.98280101772741e-06,
+      "loss": 1.47339535,
+      "memory(GiB)": 74.41,
+      "step": 5920,
+      "train_speed(iter/s)": 1.33186
+    },
+    {
+      "acc": 0.67293692,
+      "epoch": 0.15030441400304415,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.982714006652126e-06,
+      "loss": 1.51313038,
+      "memory(GiB)": 74.41,
+      "step": 5925,
+      "train_speed(iter/s)": 1.331966
+    },
+    {
+      "acc": 0.6714468,
+      "epoch": 0.1504312531709792,
+      "grad_norm": 2.75,
+      "learning_rate": 9.982626776414834e-06,
+      "loss": 1.4445982,
+      "memory(GiB)": 74.41,
+      "step": 5930,
+      "train_speed(iter/s)": 1.332076
+    },
+    {
+      "acc": 0.66601801,
+      "epoch": 0.15055809233891426,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.98253932701937e-06,
+      "loss": 1.46733093,
+      "memory(GiB)": 74.41,
+      "step": 5935,
+      "train_speed(iter/s)": 1.33217
+    },
+    {
+      "acc": 0.67825823,
+      "epoch": 0.1506849315068493,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.98245165846958e-06,
+      "loss": 1.52465038,
+      "memory(GiB)": 74.41,
+      "step": 5940,
+      "train_speed(iter/s)": 1.332276
+    },
+    {
+      "acc": 0.68103414,
+      "epoch": 0.15081177067478438,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.982363770769323e-06,
+      "loss": 1.44981155,
+      "memory(GiB)": 74.41,
+      "step": 5945,
+      "train_speed(iter/s)": 1.332386
+    },
+    {
+      "acc": 0.67710657,
+      "epoch": 0.15093860984271942,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.98227566392246e-06,
+      "loss": 1.45123138,
+      "memory(GiB)": 74.41,
+      "step": 5950,
+      "train_speed(iter/s)": 1.332476
+    },
+    {
+      "acc": 0.67353849,
+      "epoch": 0.1510654490106545,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.982187337932871e-06,
+      "loss": 1.42747765,
+      "memory(GiB)": 74.41,
+      "step": 5955,
+      "train_speed(iter/s)": 1.332582
+    },
+    {
+      "acc": 0.66164818,
+      "epoch": 0.15119228817858954,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.98209879280444e-06,
+      "loss": 1.51977062,
+      "memory(GiB)": 74.41,
+      "step": 5960,
+      "train_speed(iter/s)": 1.332689
+    },
+    {
+      "acc": 0.66978612,
+      "epoch": 0.1513191273465246,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.982010028541057e-06,
+      "loss": 1.50127926,
+      "memory(GiB)": 74.41,
+      "step": 5965,
+      "train_speed(iter/s)": 1.332797
+    },
+    {
+      "acc": 0.67614326,
+      "epoch": 0.15144596651445966,
+      "grad_norm": 3.375,
+      "learning_rate": 9.981921045146633e-06,
+      "loss": 1.41253462,
+      "memory(GiB)": 74.41,
+      "step": 5970,
+      "train_speed(iter/s)": 1.332903
+    },
+    {
+      "acc": 0.68572135,
+      "epoch": 0.15157280568239473,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.981831842625079e-06,
+      "loss": 1.40189838,
+      "memory(GiB)": 74.41,
+      "step": 5975,
+      "train_speed(iter/s)": 1.333009
+    },
+    {
+      "acc": 0.67715535,
+      "epoch": 0.15169964485032977,
+      "grad_norm": 3.125,
+      "learning_rate": 9.981742420980316e-06,
+      "loss": 1.48626108,
+      "memory(GiB)": 74.41,
+      "step": 5980,
+      "train_speed(iter/s)": 1.333121
+    },
+    {
+      "acc": 0.67456465,
+      "epoch": 0.15182648401826485,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.981652780216281e-06,
+      "loss": 1.4688508,
+      "memory(GiB)": 74.41,
+      "step": 5985,
+      "train_speed(iter/s)": 1.333236
+    },
+    {
+      "acc": 0.69072752,
+      "epoch": 0.1519533231861999,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.981562920336915e-06,
+      "loss": 1.3819972,
+      "memory(GiB)": 74.41,
+      "step": 5990,
+      "train_speed(iter/s)": 1.333344
+    },
+    {
+      "acc": 0.66027627,
+      "epoch": 0.15208016235413496,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.98147284134617e-06,
+      "loss": 1.5185338,
+      "memory(GiB)": 74.41,
+      "step": 5995,
+      "train_speed(iter/s)": 1.333452
+    },
+    {
+      "acc": 0.67649632,
+      "epoch": 0.15220700152207,
+      "grad_norm": 3.5,
+      "learning_rate": 9.981382543248011e-06,
+      "loss": 1.49696827,
+      "memory(GiB)": 74.41,
+      "step": 6000,
+      "train_speed(iter/s)": 1.333558
+    },
+    {
+      "epoch": 0.15220700152207,
+      "eval_acc": 0.6637609975420637,
+      "eval_loss": 1.4196269512176514,
+      "eval_runtime": 69.5447,
+      "eval_samples_per_second": 91.596,
+      "eval_steps_per_second": 22.906,
+      "step": 6000
+    },
+    {
+      "acc": 0.67350378,
+      "epoch": 0.15233384069000508,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.981292026046406e-06,
+      "loss": 1.45170393,
+      "memory(GiB)": 74.41,
+      "step": 6005,
+      "train_speed(iter/s)": 1.309724
+    },
+    {
+      "acc": 0.67141185,
+      "epoch": 0.15246067985794012,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.981201289745337e-06,
+      "loss": 1.47604589,
+      "memory(GiB)": 74.41,
+      "step": 6010,
+      "train_speed(iter/s)": 1.309846
+    },
+    {
+      "acc": 0.67012568,
+      "epoch": 0.1525875190258752,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.981110334348796e-06,
+      "loss": 1.49991121,
+      "memory(GiB)": 74.41,
+      "step": 6015,
+      "train_speed(iter/s)": 1.30997
+    },
+    {
+      "acc": 0.68577023,
+      "epoch": 0.15271435819381024,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.981019159860782e-06,
+      "loss": 1.40321951,
+      "memory(GiB)": 74.41,
+      "step": 6020,
+      "train_speed(iter/s)": 1.310102
+    },
+    {
+      "acc": 0.67344093,
+      "epoch": 0.1528411973617453,
+      "grad_norm": 4.375,
+      "learning_rate": 9.98092776628531e-06,
+      "loss": 1.44506454,
+      "memory(GiB)": 74.41,
+      "step": 6025,
+      "train_speed(iter/s)": 1.310227
+    },
+    {
+      "acc": 0.6822082,
+      "epoch": 0.15296803652968036,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.980836153626396e-06,
+      "loss": 1.4086689,
+      "memory(GiB)": 74.41,
+      "step": 6030,
+      "train_speed(iter/s)": 1.31035
+    },
+    {
+      "acc": 0.68966064,
+      "epoch": 0.15309487569761543,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.980744321888068e-06,
+      "loss": 1.45931091,
+      "memory(GiB)": 74.41,
+      "step": 6035,
+      "train_speed(iter/s)": 1.310473
+    },
+    {
+      "acc": 0.66883264,
+      "epoch": 0.15322171486555047,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.98065227107437e-06,
+      "loss": 1.56373882,
+      "memory(GiB)": 74.41,
+      "step": 6040,
+      "train_speed(iter/s)": 1.310599
+    },
+    {
+      "acc": 0.67347336,
+      "epoch": 0.15334855403348555,
+      "grad_norm": 2.75,
+      "learning_rate": 9.980560001189346e-06,
+      "loss": 1.47441416,
+      "memory(GiB)": 74.41,
+      "step": 6045,
+      "train_speed(iter/s)": 1.310722
+    },
+    {
+      "acc": 0.66093922,
+      "epoch": 0.1534753932014206,
+      "grad_norm": 3.0,
+      "learning_rate": 9.980467512237058e-06,
+      "loss": 1.52750111,
+      "memory(GiB)": 74.41,
+      "step": 6050,
+      "train_speed(iter/s)": 1.310851
+    },
+    {
+      "acc": 0.66299181,
+      "epoch": 0.15360223236935566,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.98037480422157e-06,
+      "loss": 1.46424932,
+      "memory(GiB)": 84.03,
+      "step": 6055,
+      "train_speed(iter/s)": 1.31095
+    },
+    {
+      "acc": 0.68041239,
+      "epoch": 0.1537290715372907,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.980281877146964e-06,
+      "loss": 1.47238178,
+      "memory(GiB)": 84.03,
+      "step": 6060,
+      "train_speed(iter/s)": 1.311079
+    },
+    {
+      "acc": 0.67586026,
+      "epoch": 0.15385591070522578,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.980188731017327e-06,
+      "loss": 1.4881237,
+      "memory(GiB)": 84.03,
+      "step": 6065,
+      "train_speed(iter/s)": 1.311205
+    },
+    {
+      "acc": 0.69314947,
+      "epoch": 0.15398274987316082,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.980095365836753e-06,
+      "loss": 1.4290554,
+      "memory(GiB)": 84.03,
+      "step": 6070,
+      "train_speed(iter/s)": 1.311324
+    },
+    {
+      "acc": 0.67606382,
+      "epoch": 0.1541095890410959,
+      "grad_norm": 3.625,
+      "learning_rate": 9.980001781609353e-06,
+      "loss": 1.47579374,
+      "memory(GiB)": 84.03,
+      "step": 6075,
+      "train_speed(iter/s)": 1.311429
+    },
+    {
+      "acc": 0.6716146,
+      "epoch": 0.15423642820903094,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.979907978339236e-06,
+      "loss": 1.48150253,
+      "memory(GiB)": 84.03,
+      "step": 6080,
+      "train_speed(iter/s)": 1.31156
+    },
+    {
+      "acc": 0.68094735,
+      "epoch": 0.154363267376966,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.979813956030535e-06,
+      "loss": 1.41302595,
+      "memory(GiB)": 84.03,
+      "step": 6085,
+      "train_speed(iter/s)": 1.311687
+    },
+    {
+      "acc": 0.66669474,
+      "epoch": 0.15449010654490106,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.979719714687384e-06,
+      "loss": 1.41963129,
+      "memory(GiB)": 84.03,
+      "step": 6090,
+      "train_speed(iter/s)": 1.31181
+    },
+    {
+      "acc": 0.6713995,
+      "epoch": 0.15461694571283613,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.979625254313924e-06,
+      "loss": 1.55340834,
+      "memory(GiB)": 84.03,
+      "step": 6095,
+      "train_speed(iter/s)": 1.311936
+    },
+    {
+      "acc": 0.66735377,
+      "epoch": 0.15474378488077117,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.979530574914316e-06,
+      "loss": 1.48315926,
+      "memory(GiB)": 84.03,
+      "step": 6100,
+      "train_speed(iter/s)": 1.312062
+    },
+    {
+      "acc": 0.67481165,
+      "epoch": 0.15487062404870625,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.97943567649272e-06,
+      "loss": 1.41142712,
+      "memory(GiB)": 84.03,
+      "step": 6105,
+      "train_speed(iter/s)": 1.312142
+    },
+    {
+      "acc": 0.67369375,
+      "epoch": 0.1549974632166413,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.979340559053311e-06,
+      "loss": 1.41358795,
+      "memory(GiB)": 84.03,
+      "step": 6110,
+      "train_speed(iter/s)": 1.312259
+    },
+    {
+      "acc": 0.67145967,
+      "epoch": 0.15512430238457636,
+      "grad_norm": 2.5,
+      "learning_rate": 9.979245222600273e-06,
+      "loss": 1.46554585,
+      "memory(GiB)": 84.03,
+      "step": 6115,
+      "train_speed(iter/s)": 1.312377
+    },
+    {
+      "acc": 0.66695099,
+      "epoch": 0.1552511415525114,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.979149667137801e-06,
+      "loss": 1.51609306,
+      "memory(GiB)": 84.03,
+      "step": 6120,
+      "train_speed(iter/s)": 1.3125
+    },
+    {
+      "acc": 0.66231642,
+      "epoch": 0.15537798072044648,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.979053892670094e-06,
+      "loss": 1.48381729,
+      "memory(GiB)": 84.03,
+      "step": 6125,
+      "train_speed(iter/s)": 1.312623
+    },
+    {
+      "acc": 0.66478152,
+      "epoch": 0.15550481988838152,
+      "grad_norm": 2.875,
+      "learning_rate": 9.978957899201369e-06,
+      "loss": 1.52829514,
+      "memory(GiB)": 84.03,
+      "step": 6130,
+      "train_speed(iter/s)": 1.312745
+    },
+    {
+      "acc": 0.68607569,
+      "epoch": 0.1556316590563166,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.978861686735845e-06,
+      "loss": 1.45974789,
+      "memory(GiB)": 84.03,
+      "step": 6135,
+      "train_speed(iter/s)": 1.312864
+    },
+    {
+      "acc": 0.67329264,
+      "epoch": 0.15575849822425164,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.978765255277756e-06,
+      "loss": 1.49201088,
+      "memory(GiB)": 84.03,
+      "step": 6140,
+      "train_speed(iter/s)": 1.312984
+    },
+    {
+      "acc": 0.68481865,
+      "epoch": 0.1558853373921867,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.97866860483134e-06,
+      "loss": 1.46981373,
+      "memory(GiB)": 84.03,
+      "step": 6145,
+      "train_speed(iter/s)": 1.313105
+    },
+    {
+      "acc": 0.68808346,
+      "epoch": 0.15601217656012176,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.978571735400853e-06,
+      "loss": 1.41622105,
+      "memory(GiB)": 84.03,
+      "step": 6150,
+      "train_speed(iter/s)": 1.313223
+    },
+    {
+      "acc": 0.66818619,
+      "epoch": 0.15613901572805683,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.978474646990552e-06,
+      "loss": 1.43968687,
+      "memory(GiB)": 84.03,
+      "step": 6155,
+      "train_speed(iter/s)": 1.313328
+    },
+    {
+      "acc": 0.68007145,
+      "epoch": 0.15626585489599187,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.97837733960471e-06,
+      "loss": 1.47557316,
+      "memory(GiB)": 84.03,
+      "step": 6160,
+      "train_speed(iter/s)": 1.313428
+    },
+    {
+      "acc": 0.69619884,
+      "epoch": 0.15639269406392695,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.978279813247605e-06,
+      "loss": 1.40981655,
+      "memory(GiB)": 84.03,
+      "step": 6165,
+      "train_speed(iter/s)": 1.313554
+    },
+    {
+      "acc": 0.68758507,
+      "epoch": 0.156519533231862,
+      "grad_norm": 2.75,
+      "learning_rate": 9.978182067923528e-06,
+      "loss": 1.41813755,
+      "memory(GiB)": 84.03,
+      "step": 6170,
+      "train_speed(iter/s)": 1.313675
+    },
+    {
+      "acc": 0.66634378,
+      "epoch": 0.15664637239979706,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.978084103636778e-06,
+      "loss": 1.56330185,
+      "memory(GiB)": 84.03,
+      "step": 6175,
+      "train_speed(iter/s)": 1.313794
+    },
+    {
+      "acc": 0.68371439,
+      "epoch": 0.1567732115677321,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.977985920391661e-06,
+      "loss": 1.43494091,
+      "memory(GiB)": 84.03,
+      "step": 6180,
+      "train_speed(iter/s)": 1.313915
+    },
+    {
+      "acc": 0.67680445,
+      "epoch": 0.15690005073566718,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.977887518192501e-06,
+      "loss": 1.49738455,
+      "memory(GiB)": 84.03,
+      "step": 6185,
+      "train_speed(iter/s)": 1.314031
+    },
+    {
+      "acc": 0.66000605,
+      "epoch": 0.15702688990360222,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.977788897043622e-06,
+      "loss": 1.54282265,
+      "memory(GiB)": 84.03,
+      "step": 6190,
+      "train_speed(iter/s)": 1.314152
+    },
+    {
+      "acc": 0.6755558,
+      "epoch": 0.1571537290715373,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.977690056949363e-06,
+      "loss": 1.45911703,
+      "memory(GiB)": 84.03,
+      "step": 6195,
+      "train_speed(iter/s)": 1.314271
+    },
+    {
+      "acc": 0.66880121,
+      "epoch": 0.15728056823947234,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.977590997914072e-06,
+      "loss": 1.48055134,
+      "memory(GiB)": 84.03,
+      "step": 6200,
+      "train_speed(iter/s)": 1.31439
+    },
+    {
+      "acc": 0.67375402,
+      "epoch": 0.1574074074074074,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.977491719942106e-06,
+      "loss": 1.43741331,
+      "memory(GiB)": 84.03,
+      "step": 6205,
+      "train_speed(iter/s)": 1.314504
+    },
+    {
+      "acc": 0.66632061,
+      "epoch": 0.15753424657534246,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.97739222303783e-06,
+      "loss": 1.47162094,
+      "memory(GiB)": 84.03,
+      "step": 6210,
+      "train_speed(iter/s)": 1.314624
+    },
+    {
+      "acc": 0.68299274,
+      "epoch": 0.15766108574327753,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.977292507205623e-06,
+      "loss": 1.45073261,
+      "memory(GiB)": 84.03,
+      "step": 6215,
+      "train_speed(iter/s)": 1.314749
+    },
+    {
+      "acc": 0.65131822,
+      "epoch": 0.15778792491121257,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.977192572449868e-06,
+      "loss": 1.52430439,
+      "memory(GiB)": 84.03,
+      "step": 6220,
+      "train_speed(iter/s)": 1.314875
+    },
+    {
+      "acc": 0.67760496,
+      "epoch": 0.15791476407914765,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.977092418774962e-06,
+      "loss": 1.46370068,
+      "memory(GiB)": 84.03,
+      "step": 6225,
+      "train_speed(iter/s)": 1.314949
+    },
+    {
+      "acc": 0.66538744,
+      "epoch": 0.1580416032470827,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.976992046185313e-06,
+      "loss": 1.52403641,
+      "memory(GiB)": 84.03,
+      "step": 6230,
+      "train_speed(iter/s)": 1.315063
+    },
+    {
+      "acc": 0.66769114,
+      "epoch": 0.15816844241501776,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.97689145468533e-06,
+      "loss": 1.47242098,
+      "memory(GiB)": 84.03,
+      "step": 6235,
+      "train_speed(iter/s)": 1.315189
+    },
+    {
+      "acc": 0.69257011,
+      "epoch": 0.1582952815829528,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.976790644279442e-06,
+      "loss": 1.39216967,
+      "memory(GiB)": 84.03,
+      "step": 6240,
+      "train_speed(iter/s)": 1.315314
+    },
+    {
+      "acc": 0.68026557,
+      "epoch": 0.15842212075088788,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.976689614972082e-06,
+      "loss": 1.40147457,
+      "memory(GiB)": 84.03,
+      "step": 6245,
+      "train_speed(iter/s)": 1.315431
+    },
+    {
+      "acc": 0.67826395,
+      "epoch": 0.15854895991882292,
+      "grad_norm": 4.25,
+      "learning_rate": 9.976588366767693e-06,
+      "loss": 1.50207605,
+      "memory(GiB)": 84.03,
+      "step": 6250,
+      "train_speed(iter/s)": 1.315555
+    },
+    {
+      "acc": 0.67186813,
+      "epoch": 0.158675799086758,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.976486899670729e-06,
+      "loss": 1.44094934,
+      "memory(GiB)": 84.03,
+      "step": 6255,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.6764843,
+      "epoch": 0.15880263825469304,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.976385213685652e-06,
+      "loss": 1.4605484,
+      "memory(GiB)": 84.03,
+      "step": 6260,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.67415466,
+      "epoch": 0.1589294774226281,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.976283308816937e-06,
+      "loss": 1.4808898,
+      "memory(GiB)": 84.03,
+      "step": 6265,
+      "train_speed(iter/s)": 1.315916
+    },
+    {
+      "acc": 0.68341036,
+      "epoch": 0.15905631659056316,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.976181185069063e-06,
+      "loss": 1.40281134,
+      "memory(GiB)": 84.03,
+      "step": 6270,
+      "train_speed(iter/s)": 1.316038
+    },
+    {
+      "acc": 0.68350744,
+      "epoch": 0.15918315575849823,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.976078842446522e-06,
+      "loss": 1.42629099,
+      "memory(GiB)": 84.03,
+      "step": 6275,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.66405363,
+      "epoch": 0.15930999492643327,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.97597628095382e-06,
+      "loss": 1.58796339,
+      "memory(GiB)": 84.03,
+      "step": 6280,
+      "train_speed(iter/s)": 1.316265
+    },
+    {
+      "acc": 0.66049261,
+      "epoch": 0.15943683409436835,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.975873500595464e-06,
+      "loss": 1.55238256,
+      "memory(GiB)": 84.03,
+      "step": 6285,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.67769146,
+      "epoch": 0.1595636732623034,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.975770501375974e-06,
+      "loss": 1.50274725,
+      "memory(GiB)": 84.03,
+      "step": 6290,
+      "train_speed(iter/s)": 1.316497
+    },
+    {
+      "acc": 0.6766901,
+      "epoch": 0.15969051243023846,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.975667283299884e-06,
+      "loss": 1.44078388,
+      "memory(GiB)": 84.03,
+      "step": 6295,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "acc": 0.67321858,
+      "epoch": 0.1598173515981735,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.975563846371732e-06,
+      "loss": 1.46467361,
+      "memory(GiB)": 84.03,
+      "step": 6300,
+      "train_speed(iter/s)": 1.316737
+    },
+    {
+      "acc": 0.67651296,
+      "epoch": 0.15994419076610858,
+      "grad_norm": 2.875,
+      "learning_rate": 9.975460190596068e-06,
+      "loss": 1.43891869,
+      "memory(GiB)": 84.03,
+      "step": 6305,
+      "train_speed(iter/s)": 1.316853
+    },
+    {
+      "acc": 0.67306108,
+      "epoch": 0.16007102993404362,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.975356315977451e-06,
+      "loss": 1.48238955,
+      "memory(GiB)": 84.03,
+      "step": 6310,
+      "train_speed(iter/s)": 1.316969
+    },
+    {
+      "acc": 0.68066874,
+      "epoch": 0.1601978691019787,
+      "grad_norm": 3.625,
+      "learning_rate": 9.975252222520449e-06,
+      "loss": 1.43060093,
+      "memory(GiB)": 84.03,
+      "step": 6315,
+      "train_speed(iter/s)": 1.317081
+    },
+    {
+      "acc": 0.67389855,
+      "epoch": 0.16032470826991374,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.97514791022964e-06,
+      "loss": 1.47459612,
+      "memory(GiB)": 84.03,
+      "step": 6320,
+      "train_speed(iter/s)": 1.317201
+    },
+    {
+      "acc": 0.68304076,
+      "epoch": 0.1604515474378488,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.975043379109617e-06,
+      "loss": 1.44840536,
+      "memory(GiB)": 84.03,
+      "step": 6325,
+      "train_speed(iter/s)": 1.317319
+    },
+    {
+      "acc": 0.68322802,
+      "epoch": 0.16057838660578386,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.974938629164973e-06,
+      "loss": 1.42595997,
+      "memory(GiB)": 84.03,
+      "step": 6330,
+      "train_speed(iter/s)": 1.317438
+    },
+    {
+      "acc": 0.67848287,
+      "epoch": 0.16070522577371893,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.974833660400315e-06,
+      "loss": 1.49422693,
+      "memory(GiB)": 84.03,
+      "step": 6335,
+      "train_speed(iter/s)": 1.317556
+    },
+    {
+      "acc": 0.67346263,
+      "epoch": 0.16083206494165397,
+      "grad_norm": 3.5,
+      "learning_rate": 9.974728472820264e-06,
+      "loss": 1.42535658,
+      "memory(GiB)": 84.03,
+      "step": 6340,
+      "train_speed(iter/s)": 1.317677
+    },
+    {
+      "acc": 0.6731668,
+      "epoch": 0.16095890410958905,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.97462306642944e-06,
+      "loss": 1.47330332,
+      "memory(GiB)": 84.03,
+      "step": 6345,
+      "train_speed(iter/s)": 1.317791
+    },
+    {
+      "acc": 0.69153981,
+      "epoch": 0.1610857432775241,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.974517441232487e-06,
+      "loss": 1.38750954,
+      "memory(GiB)": 84.03,
+      "step": 6350,
+      "train_speed(iter/s)": 1.317912
+    },
+    {
+      "acc": 0.6777123,
+      "epoch": 0.16121258244545916,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.974411597234046e-06,
+      "loss": 1.47005043,
+      "memory(GiB)": 84.03,
+      "step": 6355,
+      "train_speed(iter/s)": 1.318025
+    },
+    {
+      "acc": 0.67098198,
+      "epoch": 0.1613394216133942,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.974305534438774e-06,
+      "loss": 1.44151535,
+      "memory(GiB)": 84.03,
+      "step": 6360,
+      "train_speed(iter/s)": 1.31815
+    },
+    {
+      "acc": 0.68478169,
+      "epoch": 0.16146626078132928,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.974199252851338e-06,
+      "loss": 1.45341263,
+      "memory(GiB)": 84.03,
+      "step": 6365,
+      "train_speed(iter/s)": 1.318275
+    },
+    {
+      "acc": 0.68154378,
+      "epoch": 0.16159309994926432,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.974092752476408e-06,
+      "loss": 1.42956314,
+      "memory(GiB)": 84.03,
+      "step": 6370,
+      "train_speed(iter/s)": 1.318402
+    },
+    {
+      "acc": 0.68693876,
+      "epoch": 0.1617199391171994,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.973986033318673e-06,
+      "loss": 1.42827063,
+      "memory(GiB)": 84.03,
+      "step": 6375,
+      "train_speed(iter/s)": 1.318527
+    },
+    {
+      "acc": 0.6944397,
+      "epoch": 0.16184677828513444,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.973879095382824e-06,
+      "loss": 1.37208462,
+      "memory(GiB)": 84.03,
+      "step": 6380,
+      "train_speed(iter/s)": 1.318645
+    },
+    {
+      "acc": 0.68893852,
+      "epoch": 0.1619736174530695,
+      "grad_norm": 3.625,
+      "learning_rate": 9.973771938673564e-06,
+      "loss": 1.47604694,
+      "memory(GiB)": 84.03,
+      "step": 6385,
+      "train_speed(iter/s)": 1.318767
+    },
+    {
+      "acc": 0.67243013,
+      "epoch": 0.16210045662100456,
+      "grad_norm": 3.375,
+      "learning_rate": 9.973664563195609e-06,
+      "loss": 1.51858091,
+      "memory(GiB)": 84.03,
+      "step": 6390,
+      "train_speed(iter/s)": 1.318885
+    },
+    {
+      "acc": 0.67083964,
+      "epoch": 0.16222729578893963,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.973556968953682e-06,
+      "loss": 1.48236942,
+      "memory(GiB)": 84.03,
+      "step": 6395,
+      "train_speed(iter/s)": 1.319007
+    },
+    {
+      "acc": 0.67153382,
+      "epoch": 0.16235413495687467,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.973449155952512e-06,
+      "loss": 1.45307884,
+      "memory(GiB)": 84.03,
+      "step": 6400,
+      "train_speed(iter/s)": 1.319123
+    },
+    {
+      "acc": 0.6848743,
+      "epoch": 0.16248097412480975,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.973341124196847e-06,
+      "loss": 1.41210938,
+      "memory(GiB)": 84.03,
+      "step": 6405,
+      "train_speed(iter/s)": 1.319247
+    },
+    {
+      "acc": 0.69141006,
+      "epoch": 0.1626078132927448,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.973232873691431e-06,
+      "loss": 1.37265701,
+      "memory(GiB)": 84.03,
+      "step": 6410,
+      "train_speed(iter/s)": 1.319363
+    },
+    {
+      "acc": 0.6733892,
+      "epoch": 0.16273465246067986,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.973124404441031e-06,
+      "loss": 1.45809002,
+      "memory(GiB)": 84.03,
+      "step": 6415,
+      "train_speed(iter/s)": 1.319463
+    },
+    {
+      "acc": 0.66711111,
+      "epoch": 0.1628614916286149,
+      "grad_norm": 3.125,
+      "learning_rate": 9.973015716450416e-06,
+      "loss": 1.53318901,
+      "memory(GiB)": 84.03,
+      "step": 6420,
+      "train_speed(iter/s)": 1.319575
+    },
+    {
+      "acc": 0.68696651,
+      "epoch": 0.16298833079654998,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.972906809724367e-06,
+      "loss": 1.41335182,
+      "memory(GiB)": 84.03,
+      "step": 6425,
+      "train_speed(iter/s)": 1.319691
+    },
+    {
+      "acc": 0.68468218,
+      "epoch": 0.16311516996448502,
+      "grad_norm": 4.125,
+      "learning_rate": 9.972797684267674e-06,
+      "loss": 1.39561062,
+      "memory(GiB)": 84.03,
+      "step": 6430,
+      "train_speed(iter/s)": 1.319804
+    },
+    {
+      "acc": 0.68814998,
+      "epoch": 0.1632420091324201,
+      "grad_norm": 2.875,
+      "learning_rate": 9.972688340085137e-06,
+      "loss": 1.38954792,
+      "memory(GiB)": 84.03,
+      "step": 6435,
+      "train_speed(iter/s)": 1.319913
+    },
+    {
+      "acc": 0.68015051,
+      "epoch": 0.16336884830035514,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.972578777181565e-06,
+      "loss": 1.42977343,
+      "memory(GiB)": 84.03,
+      "step": 6440,
+      "train_speed(iter/s)": 1.320023
+    },
+    {
+      "acc": 0.67601366,
+      "epoch": 0.1634956874682902,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.972468995561778e-06,
+      "loss": 1.41601868,
+      "memory(GiB)": 84.03,
+      "step": 6445,
+      "train_speed(iter/s)": 1.320137
+    },
+    {
+      "acc": 0.6802474,
+      "epoch": 0.16362252663622526,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.972358995230604e-06,
+      "loss": 1.39051218,
+      "memory(GiB)": 84.03,
+      "step": 6450,
+      "train_speed(iter/s)": 1.320252
+    },
+    {
+      "acc": 0.68285036,
+      "epoch": 0.16374936580416033,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.97224877619288e-06,
+      "loss": 1.46564932,
+      "memory(GiB)": 84.03,
+      "step": 6455,
+      "train_speed(iter/s)": 1.320366
+    },
+    {
+      "acc": 0.66727772,
+      "epoch": 0.16387620497209537,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.972138338453457e-06,
+      "loss": 1.50595322,
+      "memory(GiB)": 84.03,
+      "step": 6460,
+      "train_speed(iter/s)": 1.320482
+    },
+    {
+      "acc": 0.68236628,
+      "epoch": 0.16400304414003045,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.972027682017191e-06,
+      "loss": 1.44456167,
+      "memory(GiB)": 84.03,
+      "step": 6465,
+      "train_speed(iter/s)": 1.320593
+    },
+    {
+      "acc": 0.66293626,
+      "epoch": 0.1641298833079655,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.971916806888948e-06,
+      "loss": 1.54165726,
+      "memory(GiB)": 84.03,
+      "step": 6470,
+      "train_speed(iter/s)": 1.320703
+    },
+    {
+      "acc": 0.67273626,
+      "epoch": 0.16425672247590056,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.971805713073606e-06,
+      "loss": 1.50497427,
+      "memory(GiB)": 84.03,
+      "step": 6475,
+      "train_speed(iter/s)": 1.320815
+    },
+    {
+      "acc": 0.67484398,
+      "epoch": 0.1643835616438356,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.971694400576053e-06,
+      "loss": 1.41637478,
+      "memory(GiB)": 84.03,
+      "step": 6480,
+      "train_speed(iter/s)": 1.320914
+    },
+    {
+      "acc": 0.68074188,
+      "epoch": 0.16451040081177068,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.971582869401182e-06,
+      "loss": 1.46963329,
+      "memory(GiB)": 84.03,
+      "step": 6485,
+      "train_speed(iter/s)": 1.32102
+    },
+    {
+      "acc": 0.68326287,
+      "epoch": 0.16463723997970572,
+      "grad_norm": 2.625,
+      "learning_rate": 9.9714711195539e-06,
+      "loss": 1.43971148,
+      "memory(GiB)": 84.03,
+      "step": 6490,
+      "train_speed(iter/s)": 1.321131
+    },
+    {
+      "acc": 0.67750559,
+      "epoch": 0.1647640791476408,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.97135915103912e-06,
+      "loss": 1.42492027,
+      "memory(GiB)": 84.03,
+      "step": 6495,
+      "train_speed(iter/s)": 1.321248
+    },
+    {
+      "acc": 0.68341432,
+      "epoch": 0.16489091831557584,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.971246963861772e-06,
+      "loss": 1.44431391,
+      "memory(GiB)": 84.03,
+      "step": 6500,
+      "train_speed(iter/s)": 1.321363
+    },
+    {
+      "acc": 0.68359885,
+      "epoch": 0.1650177574835109,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.971134558026786e-06,
+      "loss": 1.43128643,
+      "memory(GiB)": 84.03,
+      "step": 6505,
+      "train_speed(iter/s)": 1.32148
+    },
+    {
+      "acc": 0.68299708,
+      "epoch": 0.16514459665144596,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.971021933539108e-06,
+      "loss": 1.42038116,
+      "memory(GiB)": 84.03,
+      "step": 6510,
+      "train_speed(iter/s)": 1.3216
+    },
+    {
+      "acc": 0.68828382,
+      "epoch": 0.16527143581938103,
+      "grad_norm": 3.625,
+      "learning_rate": 9.97090909040369e-06,
+      "loss": 1.40049725,
+      "memory(GiB)": 84.03,
+      "step": 6515,
+      "train_speed(iter/s)": 1.321724
+    },
+    {
+      "acc": 0.67432528,
+      "epoch": 0.16539827498731607,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.970796028625499e-06,
+      "loss": 1.48123407,
+      "memory(GiB)": 84.03,
+      "step": 6520,
+      "train_speed(iter/s)": 1.32184
+    },
+    {
+      "acc": 0.68123083,
+      "epoch": 0.16552511415525115,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.970682748209505e-06,
+      "loss": 1.48922253,
+      "memory(GiB)": 84.03,
+      "step": 6525,
+      "train_speed(iter/s)": 1.321958
+    },
+    {
+      "acc": 0.68106556,
+      "epoch": 0.1656519533231862,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.97056924916069e-06,
+      "loss": 1.4452198,
+      "memory(GiB)": 84.03,
+      "step": 6530,
+      "train_speed(iter/s)": 1.322077
+    },
+    {
+      "acc": 0.68931541,
+      "epoch": 0.16577879249112126,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.970455531484049e-06,
+      "loss": 1.40451517,
+      "memory(GiB)": 84.03,
+      "step": 6535,
+      "train_speed(iter/s)": 1.322198
+    },
+    {
+      "acc": 0.68670344,
+      "epoch": 0.1659056316590563,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.97034159518458e-06,
+      "loss": 1.43749552,
+      "memory(GiB)": 84.03,
+      "step": 6540,
+      "train_speed(iter/s)": 1.322316
+    },
+    {
+      "acc": 0.67501898,
+      "epoch": 0.16603247082699138,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.9702274402673e-06,
+      "loss": 1.45324984,
+      "memory(GiB)": 84.03,
+      "step": 6545,
+      "train_speed(iter/s)": 1.322434
+    },
+    {
+      "acc": 0.68596907,
+      "epoch": 0.16615930999492642,
+      "grad_norm": 3.0,
+      "learning_rate": 9.970113066737223e-06,
+      "loss": 1.44301023,
+      "memory(GiB)": 84.03,
+      "step": 6550,
+      "train_speed(iter/s)": 1.322548
+    },
+    {
+      "acc": 0.6758235,
+      "epoch": 0.1662861491628615,
+      "grad_norm": 3.375,
+      "learning_rate": 9.969998474599386e-06,
+      "loss": 1.43410072,
+      "memory(GiB)": 84.03,
+      "step": 6555,
+      "train_speed(iter/s)": 1.322663
+    },
+    {
+      "acc": 0.68839293,
+      "epoch": 0.16641298833079654,
+      "grad_norm": 3.875,
+      "learning_rate": 9.969883663858826e-06,
+      "loss": 1.39234533,
+      "memory(GiB)": 84.03,
+      "step": 6560,
+      "train_speed(iter/s)": 1.321005
+    },
+    {
+      "acc": 0.69817762,
+      "epoch": 0.1665398274987316,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.969768634520593e-06,
+      "loss": 1.40835552,
+      "memory(GiB)": 84.03,
+      "step": 6565,
+      "train_speed(iter/s)": 1.321119
+    },
+    {
+      "acc": 0.70060635,
+      "epoch": 0.16666666666666666,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.969653386589749e-06,
+      "loss": 1.3562521,
+      "memory(GiB)": 84.03,
+      "step": 6570,
+      "train_speed(iter/s)": 1.321237
+    },
+    {
+      "acc": 0.68665152,
+      "epoch": 0.16679350583460173,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.96953792007136e-06,
+      "loss": 1.4430006,
+      "memory(GiB)": 84.03,
+      "step": 6575,
+      "train_speed(iter/s)": 1.321355
+    },
+    {
+      "acc": 0.66642728,
+      "epoch": 0.16692034500253677,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.969422234970506e-06,
+      "loss": 1.49913692,
+      "memory(GiB)": 84.03,
+      "step": 6580,
+      "train_speed(iter/s)": 1.32147
+    },
+    {
+      "acc": 0.69424238,
+      "epoch": 0.16704718417047185,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.969306331292273e-06,
+      "loss": 1.38590527,
+      "memory(GiB)": 84.03,
+      "step": 6585,
+      "train_speed(iter/s)": 1.321572
+    },
+    {
+      "acc": 0.67323561,
+      "epoch": 0.1671740233384069,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.969190209041764e-06,
+      "loss": 1.52743301,
+      "memory(GiB)": 84.03,
+      "step": 6590,
+      "train_speed(iter/s)": 1.321683
+    },
+    {
+      "acc": 0.66785965,
+      "epoch": 0.16730086250634196,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.969073868224082e-06,
+      "loss": 1.47988796,
+      "memory(GiB)": 84.03,
+      "step": 6595,
+      "train_speed(iter/s)": 1.321795
+    },
+    {
+      "acc": 0.67088342,
+      "epoch": 0.167427701674277,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.968957308844346e-06,
+      "loss": 1.56117172,
+      "memory(GiB)": 84.03,
+      "step": 6600,
+      "train_speed(iter/s)": 1.321908
+    },
+    {
+      "acc": 0.67405233,
+      "epoch": 0.16755454084221208,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.968840530907684e-06,
+      "loss": 1.49264374,
+      "memory(GiB)": 84.03,
+      "step": 6605,
+      "train_speed(iter/s)": 1.322018
+    },
+    {
+      "acc": 0.67089033,
+      "epoch": 0.16768138001014712,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.96872353441923e-06,
+      "loss": 1.49277792,
+      "memory(GiB)": 84.03,
+      "step": 6610,
+      "train_speed(iter/s)": 1.322134
+    },
+    {
+      "acc": 0.67435732,
+      "epoch": 0.1678082191780822,
+      "grad_norm": 3.625,
+      "learning_rate": 9.968606319384131e-06,
+      "loss": 1.50213432,
+      "memory(GiB)": 84.03,
+      "step": 6615,
+      "train_speed(iter/s)": 1.322247
+    },
+    {
+      "acc": 0.67240167,
+      "epoch": 0.16793505834601724,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.968488885807544e-06,
+      "loss": 1.52199707,
+      "memory(GiB)": 84.03,
+      "step": 6620,
+      "train_speed(iter/s)": 1.322346
+    },
+    {
+      "acc": 0.67695351,
+      "epoch": 0.1680618975139523,
+      "grad_norm": 2.625,
+      "learning_rate": 9.968371233694633e-06,
+      "loss": 1.42570362,
+      "memory(GiB)": 84.03,
+      "step": 6625,
+      "train_speed(iter/s)": 1.322456
+    },
+    {
+      "acc": 0.67939644,
+      "epoch": 0.16818873668188736,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.968253363050573e-06,
+      "loss": 1.45981617,
+      "memory(GiB)": 84.03,
+      "step": 6630,
+      "train_speed(iter/s)": 1.322562
+    },
+    {
+      "acc": 0.68097906,
+      "epoch": 0.16831557584982243,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.968135273880547e-06,
+      "loss": 1.39907064,
+      "memory(GiB)": 84.03,
+      "step": 6635,
+      "train_speed(iter/s)": 1.322668
+    },
+    {
+      "acc": 0.67201824,
+      "epoch": 0.16844241501775747,
+      "grad_norm": 4.125,
+      "learning_rate": 9.968016966189753e-06,
+      "loss": 1.45031796,
+      "memory(GiB)": 84.03,
+      "step": 6640,
+      "train_speed(iter/s)": 1.322777
+    },
+    {
+      "acc": 0.68637385,
+      "epoch": 0.16856925418569255,
+      "grad_norm": 3.625,
+      "learning_rate": 9.96789843998339e-06,
+      "loss": 1.35989351,
+      "memory(GiB)": 84.03,
+      "step": 6645,
+      "train_speed(iter/s)": 1.322886
+    },
+    {
+      "acc": 0.6789042,
+      "epoch": 0.1686960933536276,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.967779695266675e-06,
+      "loss": 1.44609766,
+      "memory(GiB)": 84.03,
+      "step": 6650,
+      "train_speed(iter/s)": 1.322989
+    },
+    {
+      "acc": 0.67634153,
+      "epoch": 0.16882293252156266,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.967660732044828e-06,
+      "loss": 1.45321503,
+      "memory(GiB)": 84.03,
+      "step": 6655,
+      "train_speed(iter/s)": 1.323105
+    },
+    {
+      "acc": 0.67773724,
+      "epoch": 0.1689497716894977,
+      "grad_norm": 2.75,
+      "learning_rate": 9.967541550323085e-06,
+      "loss": 1.46208839,
+      "memory(GiB)": 84.03,
+      "step": 6660,
+      "train_speed(iter/s)": 1.323217
+    },
+    {
+      "acc": 0.68695474,
+      "epoch": 0.16907661085743278,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.967422150106685e-06,
+      "loss": 1.45379906,
+      "memory(GiB)": 84.03,
+      "step": 6665,
+      "train_speed(iter/s)": 1.323333
+    },
+    {
+      "acc": 0.67199945,
+      "epoch": 0.16920345002536782,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.96730253140088e-06,
+      "loss": 1.44753876,
+      "memory(GiB)": 84.03,
+      "step": 6670,
+      "train_speed(iter/s)": 1.323449
+    },
+    {
+      "acc": 0.67275047,
+      "epoch": 0.1693302891933029,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.967182694210933e-06,
+      "loss": 1.46594572,
+      "memory(GiB)": 84.03,
+      "step": 6675,
+      "train_speed(iter/s)": 1.323564
+    },
+    {
+      "acc": 0.68547063,
+      "epoch": 0.16945712836123794,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.967062638542116e-06,
+      "loss": 1.43129683,
+      "memory(GiB)": 84.03,
+      "step": 6680,
+      "train_speed(iter/s)": 1.323677
+    },
+    {
+      "acc": 0.68039246,
+      "epoch": 0.169583967529173,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.966942364399706e-06,
+      "loss": 1.45226822,
+      "memory(GiB)": 84.03,
+      "step": 6685,
+      "train_speed(iter/s)": 1.32379
+    },
+    {
+      "acc": 0.66649275,
+      "epoch": 0.16971080669710806,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.966821871788995e-06,
+      "loss": 1.4777132,
+      "memory(GiB)": 84.03,
+      "step": 6690,
+      "train_speed(iter/s)": 1.323906
+    },
+    {
+      "acc": 0.6805603,
+      "epoch": 0.16983764586504313,
+      "grad_norm": 3.25,
+      "learning_rate": 9.966701160715283e-06,
+      "loss": 1.45332289,
+      "memory(GiB)": 84.03,
+      "step": 6695,
+      "train_speed(iter/s)": 1.324023
+    },
+    {
+      "acc": 0.68742733,
+      "epoch": 0.16996448503297817,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.96658023118388e-06,
+      "loss": 1.44522667,
+      "memory(GiB)": 84.03,
+      "step": 6700,
+      "train_speed(iter/s)": 1.324133
+    },
+    {
+      "acc": 0.68340836,
+      "epoch": 0.17009132420091325,
+      "grad_norm": 3.375,
+      "learning_rate": 9.966459083200102e-06,
+      "loss": 1.45029564,
+      "memory(GiB)": 84.03,
+      "step": 6705,
+      "train_speed(iter/s)": 1.324238
+    },
+    {
+      "acc": 0.66717319,
+      "epoch": 0.1702181633688483,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.966337716769283e-06,
+      "loss": 1.51814022,
+      "memory(GiB)": 84.03,
+      "step": 6710,
+      "train_speed(iter/s)": 1.324347
+    },
+    {
+      "acc": 0.66444573,
+      "epoch": 0.17034500253678336,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.966216131896755e-06,
+      "loss": 1.50667706,
+      "memory(GiB)": 84.03,
+      "step": 6715,
+      "train_speed(iter/s)": 1.32446
+    },
+    {
+      "acc": 0.66625814,
+      "epoch": 0.1704718417047184,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.966094328587871e-06,
+      "loss": 1.46497936,
+      "memory(GiB)": 84.03,
+      "step": 6720,
+      "train_speed(iter/s)": 1.324569
+    },
+    {
+      "acc": 0.67484603,
+      "epoch": 0.17059868087265348,
+      "grad_norm": 2.875,
+      "learning_rate": 9.965972306847986e-06,
+      "loss": 1.5178091,
+      "memory(GiB)": 84.03,
+      "step": 6725,
+      "train_speed(iter/s)": 1.324675
+    },
+    {
+      "acc": 0.69521112,
+      "epoch": 0.17072552004058852,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.965850066682468e-06,
+      "loss": 1.35668163,
+      "memory(GiB)": 84.03,
+      "step": 6730,
+      "train_speed(iter/s)": 1.32478
+    },
+    {
+      "acc": 0.67168932,
+      "epoch": 0.1708523592085236,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.965727608096692e-06,
+      "loss": 1.44138927,
+      "memory(GiB)": 84.03,
+      "step": 6735,
+      "train_speed(iter/s)": 1.324884
+    },
+    {
+      "acc": 0.68202744,
+      "epoch": 0.17097919837645864,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.965604931096045e-06,
+      "loss": 1.48506231,
+      "memory(GiB)": 84.03,
+      "step": 6740,
+      "train_speed(iter/s)": 1.324992
+    },
+    {
+      "acc": 0.67131748,
+      "epoch": 0.1711060375443937,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.965482035685925e-06,
+      "loss": 1.50202665,
+      "memory(GiB)": 84.03,
+      "step": 6745,
+      "train_speed(iter/s)": 1.3251
+    },
+    {
+      "acc": 0.69023943,
+      "epoch": 0.17123287671232876,
+      "grad_norm": 3.0,
+      "learning_rate": 9.965358921871735e-06,
+      "loss": 1.41043186,
+      "memory(GiB)": 84.03,
+      "step": 6750,
+      "train_speed(iter/s)": 1.325212
+    },
+    {
+      "acc": 0.69200039,
+      "epoch": 0.17135971588026383,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.965235589658891e-06,
+      "loss": 1.41213284,
+      "memory(GiB)": 84.03,
+      "step": 6755,
+      "train_speed(iter/s)": 1.325321
+    },
+    {
+      "acc": 0.69493322,
+      "epoch": 0.17148655504819887,
+      "grad_norm": 2.21875,
+      "learning_rate": 9.965112039052817e-06,
+      "loss": 1.44045172,
+      "memory(GiB)": 84.03,
+      "step": 6760,
+      "train_speed(iter/s)": 1.325433
+    },
+    {
+      "acc": 0.68552008,
+      "epoch": 0.17161339421613395,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.964988270058948e-06,
+      "loss": 1.4898325,
+      "memory(GiB)": 84.03,
+      "step": 6765,
+      "train_speed(iter/s)": 1.325542
+    },
+    {
+      "acc": 0.67255669,
+      "epoch": 0.171740233384069,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.96486428268273e-06,
+      "loss": 1.43044319,
+      "memory(GiB)": 84.03,
+      "step": 6770,
+      "train_speed(iter/s)": 1.325648
+    },
+    {
+      "acc": 0.67979031,
+      "epoch": 0.17186707255200406,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.964740076929612e-06,
+      "loss": 1.43200703,
+      "memory(GiB)": 84.03,
+      "step": 6775,
+      "train_speed(iter/s)": 1.325757
+    },
+    {
+      "acc": 0.66918106,
+      "epoch": 0.1719939117199391,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.964615652805059e-06,
+      "loss": 1.51224289,
+      "memory(GiB)": 84.03,
+      "step": 6780,
+      "train_speed(iter/s)": 1.325867
+    },
+    {
+      "acc": 0.68714666,
+      "epoch": 0.17212075088787418,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.964491010314545e-06,
+      "loss": 1.42595119,
+      "memory(GiB)": 84.03,
+      "step": 6785,
+      "train_speed(iter/s)": 1.325973
+    },
+    {
+      "acc": 0.68133354,
+      "epoch": 0.17224759005580922,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.964366149463552e-06,
+      "loss": 1.45480986,
+      "memory(GiB)": 84.03,
+      "step": 6790,
+      "train_speed(iter/s)": 1.326077
+    },
+    {
+      "acc": 0.69556971,
+      "epoch": 0.1723744292237443,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.96424107025757e-06,
+      "loss": 1.36624832,
+      "memory(GiB)": 84.03,
+      "step": 6795,
+      "train_speed(iter/s)": 1.326179
+    },
+    {
+      "acc": 0.68241591,
+      "epoch": 0.17250126839167934,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.964115772702104e-06,
+      "loss": 1.38947525,
+      "memory(GiB)": 84.03,
+      "step": 6800,
+      "train_speed(iter/s)": 1.326286
+    },
+    {
+      "acc": 0.67640886,
+      "epoch": 0.1726281075596144,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.963990256802662e-06,
+      "loss": 1.49233313,
+      "memory(GiB)": 84.03,
+      "step": 6805,
+      "train_speed(iter/s)": 1.326388
+    },
+    {
+      "acc": 0.69005947,
+      "epoch": 0.17275494672754946,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.963864522564765e-06,
+      "loss": 1.37665796,
+      "memory(GiB)": 84.03,
+      "step": 6810,
+      "train_speed(iter/s)": 1.326491
+    },
+    {
+      "acc": 0.68142166,
+      "epoch": 0.17288178589548453,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.963738569993945e-06,
+      "loss": 1.48349905,
+      "memory(GiB)": 84.03,
+      "step": 6815,
+      "train_speed(iter/s)": 1.326592
+    },
+    {
+      "acc": 0.68787861,
+      "epoch": 0.17300862506341957,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.963612399095743e-06,
+      "loss": 1.45882072,
+      "memory(GiB)": 84.03,
+      "step": 6820,
+      "train_speed(iter/s)": 1.326698
+    },
+    {
+      "acc": 0.66207733,
+      "epoch": 0.17313546423135465,
+      "grad_norm": 3.375,
+      "learning_rate": 9.963486009875705e-06,
+      "loss": 1.51938553,
+      "memory(GiB)": 84.03,
+      "step": 6825,
+      "train_speed(iter/s)": 1.326808
+    },
+    {
+      "acc": 0.68319798,
+      "epoch": 0.1732623033992897,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.963359402339393e-06,
+      "loss": 1.45501394,
+      "memory(GiB)": 84.03,
+      "step": 6830,
+      "train_speed(iter/s)": 1.326909
+    },
+    {
+      "acc": 0.67128172,
+      "epoch": 0.17338914256722476,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.963232576492373e-06,
+      "loss": 1.52493582,
+      "memory(GiB)": 84.03,
+      "step": 6835,
+      "train_speed(iter/s)": 1.327015
+    },
+    {
+      "acc": 0.6743495,
+      "epoch": 0.1735159817351598,
+      "grad_norm": 2.875,
+      "learning_rate": 9.963105532340226e-06,
+      "loss": 1.52561321,
+      "memory(GiB)": 84.03,
+      "step": 6840,
+      "train_speed(iter/s)": 1.327106
+    },
+    {
+      "acc": 0.67272425,
+      "epoch": 0.17364282090309488,
+      "grad_norm": 3.75,
+      "learning_rate": 9.962978269888538e-06,
+      "loss": 1.48959007,
+      "memory(GiB)": 84.03,
+      "step": 6845,
+      "train_speed(iter/s)": 1.325761
+    },
+    {
+      "acc": 0.66643348,
+      "epoch": 0.17376966007102992,
+      "grad_norm": 2.625,
+      "learning_rate": 9.96285078914291e-06,
+      "loss": 1.41744976,
+      "memory(GiB)": 84.03,
+      "step": 6850,
+      "train_speed(iter/s)": 1.325861
+    },
+    {
+      "acc": 0.66866293,
+      "epoch": 0.173896499238965,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.962723090108944e-06,
+      "loss": 1.45672398,
+      "memory(GiB)": 84.03,
+      "step": 6855,
+      "train_speed(iter/s)": 1.325968
+    },
+    {
+      "acc": 0.6734477,
+      "epoch": 0.17402333840690004,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.962595172792261e-06,
+      "loss": 1.47723694,
+      "memory(GiB)": 84.03,
+      "step": 6860,
+      "train_speed(iter/s)": 1.326075
+    },
+    {
+      "acc": 0.68085685,
+      "epoch": 0.1741501775748351,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.962467037198487e-06,
+      "loss": 1.37029514,
+      "memory(GiB)": 84.03,
+      "step": 6865,
+      "train_speed(iter/s)": 1.326178
+    },
+    {
+      "acc": 0.67521067,
+      "epoch": 0.17427701674277016,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.962338683333254e-06,
+      "loss": 1.46367369,
+      "memory(GiB)": 84.03,
+      "step": 6870,
+      "train_speed(iter/s)": 1.326283
+    },
+    {
+      "acc": 0.67081261,
+      "epoch": 0.17440385591070523,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.962210111202212e-06,
+      "loss": 1.53500528,
+      "memory(GiB)": 84.03,
+      "step": 6875,
+      "train_speed(iter/s)": 1.326392
+    },
+    {
+      "acc": 0.67676725,
+      "epoch": 0.17453069507864027,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.962081320811015e-06,
+      "loss": 1.45632095,
+      "memory(GiB)": 84.03,
+      "step": 6880,
+      "train_speed(iter/s)": 1.32649
+    },
+    {
+      "acc": 0.68870268,
+      "epoch": 0.17465753424657535,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.961952312165327e-06,
+      "loss": 1.43509655,
+      "memory(GiB)": 84.03,
+      "step": 6885,
+      "train_speed(iter/s)": 1.326577
+    },
+    {
+      "acc": 0.68707471,
+      "epoch": 0.1747843734145104,
+      "grad_norm": 2.625,
+      "learning_rate": 9.961823085270823e-06,
+      "loss": 1.41879101,
+      "memory(GiB)": 84.03,
+      "step": 6890,
+      "train_speed(iter/s)": 1.326675
+    },
+    {
+      "acc": 0.68200164,
+      "epoch": 0.17491121258244546,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.961693640133187e-06,
+      "loss": 1.35287151,
+      "memory(GiB)": 84.03,
+      "step": 6895,
+      "train_speed(iter/s)": 1.326782
+    },
+    {
+      "acc": 0.67957006,
+      "epoch": 0.1750380517503805,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.961563976758112e-06,
+      "loss": 1.44992065,
+      "memory(GiB)": 84.03,
+      "step": 6900,
+      "train_speed(iter/s)": 1.326886
+    },
+    {
+      "acc": 0.68149853,
+      "epoch": 0.17516489091831558,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.961434095151301e-06,
+      "loss": 1.46768274,
+      "memory(GiB)": 84.03,
+      "step": 6905,
+      "train_speed(iter/s)": 1.32699
+    },
+    {
+      "acc": 0.67518568,
+      "epoch": 0.17529173008625062,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.961303995318467e-06,
+      "loss": 1.40401545,
+      "memory(GiB)": 84.03,
+      "step": 6910,
+      "train_speed(iter/s)": 1.327093
+    },
+    {
+      "acc": 0.68169885,
+      "epoch": 0.1754185692541857,
+      "grad_norm": 8.5,
+      "learning_rate": 9.961173677265334e-06,
+      "loss": 1.42359543,
+      "memory(GiB)": 84.03,
+      "step": 6915,
+      "train_speed(iter/s)": 1.327199
+    },
+    {
+      "acc": 0.67442713,
+      "epoch": 0.17554540842212074,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.961043140997632e-06,
+      "loss": 1.42162514,
+      "memory(GiB)": 84.03,
+      "step": 6920,
+      "train_speed(iter/s)": 1.327297
+    },
+    {
+      "acc": 0.66386724,
+      "epoch": 0.1756722475900558,
+      "grad_norm": 3.25,
+      "learning_rate": 9.960912386521104e-06,
+      "loss": 1.47871132,
+      "memory(GiB)": 84.03,
+      "step": 6925,
+      "train_speed(iter/s)": 1.327386
+    },
+    {
+      "acc": 0.67381535,
+      "epoch": 0.17579908675799086,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.9607814138415e-06,
+      "loss": 1.45070648,
+      "memory(GiB)": 84.03,
+      "step": 6930,
+      "train_speed(iter/s)": 1.32749
+    },
+    {
+      "acc": 0.67737508,
+      "epoch": 0.17592592592592593,
+      "grad_norm": 4.25,
+      "learning_rate": 9.96065022296458e-06,
+      "loss": 1.46409359,
+      "memory(GiB)": 84.03,
+      "step": 6935,
+      "train_speed(iter/s)": 1.327591
+    },
+    {
+      "acc": 0.68725786,
+      "epoch": 0.17605276509386097,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.960518813896117e-06,
+      "loss": 1.39846497,
+      "memory(GiB)": 84.03,
+      "step": 6940,
+      "train_speed(iter/s)": 1.327681
+    },
+    {
+      "acc": 0.67600069,
+      "epoch": 0.17617960426179605,
+      "grad_norm": 3.25,
+      "learning_rate": 9.960387186641887e-06,
+      "loss": 1.45503941,
+      "memory(GiB)": 84.03,
+      "step": 6945,
+      "train_speed(iter/s)": 1.327786
+    },
+    {
+      "acc": 0.67949853,
+      "epoch": 0.1763064434297311,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.960255341207686e-06,
+      "loss": 1.40665035,
+      "memory(GiB)": 84.03,
+      "step": 6950,
+      "train_speed(iter/s)": 1.32789
+    },
+    {
+      "acc": 0.69468079,
+      "epoch": 0.17643328259766616,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.960123277599305e-06,
+      "loss": 1.39744101,
+      "memory(GiB)": 84.03,
+      "step": 6955,
+      "train_speed(iter/s)": 1.327995
+    },
+    {
+      "acc": 0.66049366,
+      "epoch": 0.1765601217656012,
+      "grad_norm": 3.5,
+      "learning_rate": 9.959990995822559e-06,
+      "loss": 1.47330599,
+      "memory(GiB)": 84.03,
+      "step": 6960,
+      "train_speed(iter/s)": 1.3281
+    },
+    {
+      "acc": 0.68365307,
+      "epoch": 0.17668696093353628,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.959858495883263e-06,
+      "loss": 1.41128139,
+      "memory(GiB)": 84.03,
+      "step": 6965,
+      "train_speed(iter/s)": 1.328202
+    },
+    {
+      "acc": 0.68864241,
+      "epoch": 0.17681380010147132,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.959725777787249e-06,
+      "loss": 1.38568306,
+      "memory(GiB)": 84.03,
+      "step": 6970,
+      "train_speed(iter/s)": 1.328303
+    },
+    {
+      "acc": 0.68245506,
+      "epoch": 0.1769406392694064,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.959592841540349e-06,
+      "loss": 1.45053539,
+      "memory(GiB)": 84.03,
+      "step": 6975,
+      "train_speed(iter/s)": 1.326832
+    },
+    {
+      "acc": 0.66536908,
+      "epoch": 0.17706747843734144,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.959459687148414e-06,
+      "loss": 1.44794655,
+      "memory(GiB)": 84.03,
+      "step": 6980,
+      "train_speed(iter/s)": 1.326938
+    },
+    {
+      "acc": 0.66374092,
+      "epoch": 0.1771943176052765,
+      "grad_norm": 3.5,
+      "learning_rate": 9.959326314617299e-06,
+      "loss": 1.48032999,
+      "memory(GiB)": 84.03,
+      "step": 6985,
+      "train_speed(iter/s)": 1.325461
+    },
+    {
+      "acc": 0.67092271,
+      "epoch": 0.17732115677321156,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.95919272395287e-06,
+      "loss": 1.46768456,
+      "memory(GiB)": 84.03,
+      "step": 6990,
+      "train_speed(iter/s)": 1.325564
+    },
+    {
+      "acc": 0.66382222,
+      "epoch": 0.17744799594114663,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.959058915161006e-06,
+      "loss": 1.55699778,
+      "memory(GiB)": 84.03,
+      "step": 6995,
+      "train_speed(iter/s)": 1.325666
+    },
+    {
+      "acc": 0.66125469,
+      "epoch": 0.17757483510908167,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.95892488824759e-06,
+      "loss": 1.51882839,
+      "memory(GiB)": 84.03,
+      "step": 7000,
+      "train_speed(iter/s)": 1.325758
+    },
+    {
+      "epoch": 0.17757483510908167,
+      "eval_acc": 0.6656709623123396,
+      "eval_loss": 1.408115267753601,
+      "eval_runtime": 70.3583,
+      "eval_samples_per_second": 90.537,
+      "eval_steps_per_second": 22.641,
+      "step": 7000
+    },
+    {
+      "acc": 0.69738526,
+      "epoch": 0.17770167427701675,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.958790643218515e-06,
+      "loss": 1.34354811,
+      "memory(GiB)": 84.03,
+      "step": 7005,
+      "train_speed(iter/s)": 1.305351
+    },
+    {
+      "acc": 0.67537589,
+      "epoch": 0.1778285134449518,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.95865618007969e-06,
+      "loss": 1.41198549,
+      "memory(GiB)": 84.03,
+      "step": 7010,
+      "train_speed(iter/s)": 1.305444
+    },
+    {
+      "acc": 0.67915568,
+      "epoch": 0.17795535261288686,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.958521498837029e-06,
+      "loss": 1.45032997,
+      "memory(GiB)": 84.03,
+      "step": 7015,
+      "train_speed(iter/s)": 1.305552
+    },
+    {
+      "acc": 0.6774868,
+      "epoch": 0.1780821917808219,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.95838659949645e-06,
+      "loss": 1.48344288,
+      "memory(GiB)": 84.03,
+      "step": 7020,
+      "train_speed(iter/s)": 1.305654
+    },
+    {
+      "acc": 0.69721746,
+      "epoch": 0.17820903094875698,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.958251482063894e-06,
+      "loss": 1.39726954,
+      "memory(GiB)": 84.03,
+      "step": 7025,
+      "train_speed(iter/s)": 1.305759
+    },
+    {
+      "acc": 0.67271423,
+      "epoch": 0.17833587011669202,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.9581161465453e-06,
+      "loss": 1.4687006,
+      "memory(GiB)": 84.03,
+      "step": 7030,
+      "train_speed(iter/s)": 1.305868
+    },
+    {
+      "acc": 0.67304783,
+      "epoch": 0.1784627092846271,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.957980592946621e-06,
+      "loss": 1.46409664,
+      "memory(GiB)": 84.03,
+      "step": 7035,
+      "train_speed(iter/s)": 1.305974
+    },
+    {
+      "acc": 0.68899403,
+      "epoch": 0.17858954845256214,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.957844821273822e-06,
+      "loss": 1.39345961,
+      "memory(GiB)": 84.03,
+      "step": 7040,
+      "train_speed(iter/s)": 1.306077
+    },
+    {
+      "acc": 0.69238262,
+      "epoch": 0.1787163876204972,
+      "grad_norm": 2.75,
+      "learning_rate": 9.95770883153287e-06,
+      "loss": 1.34865952,
+      "memory(GiB)": 84.03,
+      "step": 7045,
+      "train_speed(iter/s)": 1.306187
+    },
+    {
+      "acc": 0.68097448,
+      "epoch": 0.17884322678843226,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.957572623729749e-06,
+      "loss": 1.5037859,
+      "memory(GiB)": 84.03,
+      "step": 7050,
+      "train_speed(iter/s)": 1.306293
+    },
+    {
+      "acc": 0.66523604,
+      "epoch": 0.17897006595636733,
+      "grad_norm": 3.625,
+      "learning_rate": 9.957436197870451e-06,
+      "loss": 1.46537209,
+      "memory(GiB)": 84.03,
+      "step": 7055,
+      "train_speed(iter/s)": 1.3064
+    },
+    {
+      "acc": 0.67959223,
+      "epoch": 0.17909690512430237,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.957299553960975e-06,
+      "loss": 1.49611158,
+      "memory(GiB)": 84.03,
+      "step": 7060,
+      "train_speed(iter/s)": 1.30651
+    },
+    {
+      "acc": 0.68119822,
+      "epoch": 0.17922374429223745,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.957162692007334e-06,
+      "loss": 1.39288635,
+      "memory(GiB)": 84.03,
+      "step": 7065,
+      "train_speed(iter/s)": 1.306611
+    },
+    {
+      "acc": 0.68606024,
+      "epoch": 0.1793505834601725,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.957025612015543e-06,
+      "loss": 1.44091291,
+      "memory(GiB)": 84.03,
+      "step": 7070,
+      "train_speed(iter/s)": 1.306715
+    },
+    {
+      "acc": 0.68718615,
+      "epoch": 0.17947742262810756,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.956888313991636e-06,
+      "loss": 1.38391647,
+      "memory(GiB)": 84.03,
+      "step": 7075,
+      "train_speed(iter/s)": 1.306826
+    },
+    {
+      "acc": 0.6738739,
+      "epoch": 0.1796042617960426,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.956750797941648e-06,
+      "loss": 1.46603212,
+      "memory(GiB)": 84.03,
+      "step": 7080,
+      "train_speed(iter/s)": 1.306928
+    },
+    {
+      "acc": 0.67291355,
+      "epoch": 0.17973110096397768,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.95661306387163e-06,
+      "loss": 1.50964909,
+      "memory(GiB)": 84.03,
+      "step": 7085,
+      "train_speed(iter/s)": 1.307036
+    },
+    {
+      "acc": 0.6754921,
+      "epoch": 0.17985794013191272,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.95647511178764e-06,
+      "loss": 1.37670689,
+      "memory(GiB)": 84.03,
+      "step": 7090,
+      "train_speed(iter/s)": 1.307135
+    },
+    {
+      "acc": 0.68139009,
+      "epoch": 0.1799847792998478,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.956336941695747e-06,
+      "loss": 1.47903299,
+      "memory(GiB)": 84.03,
+      "step": 7095,
+      "train_speed(iter/s)": 1.307239
+    },
+    {
+      "acc": 0.68117304,
+      "epoch": 0.18011161846778284,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.956198553602026e-06,
+      "loss": 1.3909481,
+      "memory(GiB)": 84.03,
+      "step": 7100,
+      "train_speed(iter/s)": 1.307346
+    },
+    {
+      "acc": 0.64900866,
+      "epoch": 0.1802384576357179,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.956059947512563e-06,
+      "loss": 1.53884592,
+      "memory(GiB)": 84.03,
+      "step": 7105,
+      "train_speed(iter/s)": 1.307426
+    },
+    {
+      "acc": 0.67373705,
+      "epoch": 0.18036529680365296,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.95592112343346e-06,
+      "loss": 1.45439167,
+      "memory(GiB)": 84.03,
+      "step": 7110,
+      "train_speed(iter/s)": 1.307531
+    },
+    {
+      "acc": 0.68399754,
+      "epoch": 0.18049213597158803,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.955782081370818e-06,
+      "loss": 1.40017004,
+      "memory(GiB)": 84.03,
+      "step": 7115,
+      "train_speed(iter/s)": 1.307635
+    },
+    {
+      "acc": 0.69125991,
+      "epoch": 0.18061897513952307,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.955642821330752e-06,
+      "loss": 1.43639622,
+      "memory(GiB)": 84.03,
+      "step": 7120,
+      "train_speed(iter/s)": 1.307737
+    },
+    {
+      "acc": 0.69536862,
+      "epoch": 0.18074581430745815,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.95550334331939e-06,
+      "loss": 1.35474072,
+      "memory(GiB)": 84.03,
+      "step": 7125,
+      "train_speed(iter/s)": 1.307835
+    },
+    {
+      "acc": 0.68481469,
+      "epoch": 0.1808726534753932,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.955363647342868e-06,
+      "loss": 1.38030787,
+      "memory(GiB)": 84.03,
+      "step": 7130,
+      "train_speed(iter/s)": 1.30789
+    },
+    {
+      "acc": 0.67103715,
+      "epoch": 0.18099949264332826,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.955223733407327e-06,
+      "loss": 1.48803749,
+      "memory(GiB)": 84.03,
+      "step": 7135,
+      "train_speed(iter/s)": 1.307998
+    },
+    {
+      "acc": 0.68106079,
+      "epoch": 0.1811263318112633,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.955083601518924e-06,
+      "loss": 1.46575174,
+      "memory(GiB)": 84.03,
+      "step": 7140,
+      "train_speed(iter/s)": 1.308103
+    },
+    {
+      "acc": 0.65999794,
+      "epoch": 0.18125317097919838,
+      "grad_norm": 3.0,
+      "learning_rate": 9.95494325168382e-06,
+      "loss": 1.52575588,
+      "memory(GiB)": 84.03,
+      "step": 7145,
+      "train_speed(iter/s)": 1.308208
+    },
+    {
+      "acc": 0.67221708,
+      "epoch": 0.18138001014713342,
+      "grad_norm": 2.875,
+      "learning_rate": 9.954802683908192e-06,
+      "loss": 1.50568676,
+      "memory(GiB)": 84.03,
+      "step": 7150,
+      "train_speed(iter/s)": 1.308314
+    },
+    {
+      "acc": 0.68123727,
+      "epoch": 0.1815068493150685,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.954661898198216e-06,
+      "loss": 1.44009342,
+      "memory(GiB)": 84.03,
+      "step": 7155,
+      "train_speed(iter/s)": 1.308424
+    },
+    {
+      "acc": 0.66840439,
+      "epoch": 0.18163368848300354,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.954520894560092e-06,
+      "loss": 1.47081337,
+      "memory(GiB)": 84.03,
+      "step": 7160,
+      "train_speed(iter/s)": 1.30853
+    },
+    {
+      "acc": 0.67477932,
+      "epoch": 0.1817605276509386,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.954379673000018e-06,
+      "loss": 1.47401209,
+      "memory(GiB)": 84.03,
+      "step": 7165,
+      "train_speed(iter/s)": 1.308631
+    },
+    {
+      "acc": 0.67438231,
+      "epoch": 0.18188736681887366,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.954238233524208e-06,
+      "loss": 1.48946419,
+      "memory(GiB)": 84.03,
+      "step": 7170,
+      "train_speed(iter/s)": 1.308735
+    },
+    {
+      "acc": 0.67583365,
+      "epoch": 0.18201420598680873,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.954096576138879e-06,
+      "loss": 1.48761139,
+      "memory(GiB)": 84.03,
+      "step": 7175,
+      "train_speed(iter/s)": 1.308839
+    },
+    {
+      "acc": 0.67588234,
+      "epoch": 0.18214104515474377,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.953954700850264e-06,
+      "loss": 1.48669815,
+      "memory(GiB)": 84.03,
+      "step": 7180,
+      "train_speed(iter/s)": 1.30895
+    },
+    {
+      "acc": 0.67458658,
+      "epoch": 0.18226788432267885,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.953812607664607e-06,
+      "loss": 1.49057693,
+      "memory(GiB)": 84.03,
+      "step": 7185,
+      "train_speed(iter/s)": 1.309055
+    },
+    {
+      "acc": 0.68518915,
+      "epoch": 0.1823947234906139,
+      "grad_norm": 3.25,
+      "learning_rate": 9.95367029658815e-06,
+      "loss": 1.44901924,
+      "memory(GiB)": 84.03,
+      "step": 7190,
+      "train_speed(iter/s)": 1.309166
+    },
+    {
+      "acc": 0.67003198,
+      "epoch": 0.18252156265854896,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.953527767627159e-06,
+      "loss": 1.405159,
+      "memory(GiB)": 84.03,
+      "step": 7195,
+      "train_speed(iter/s)": 1.309271
+    },
+    {
+      "acc": 0.6983532,
+      "epoch": 0.182648401826484,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.9533850207879e-06,
+      "loss": 1.35821362,
+      "memory(GiB)": 84.03,
+      "step": 7200,
+      "train_speed(iter/s)": 1.309373
+    },
+    {
+      "acc": 0.68865299,
+      "epoch": 0.18277524099441908,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.953242056076652e-06,
+      "loss": 1.4442131,
+      "memory(GiB)": 84.03,
+      "step": 7205,
+      "train_speed(iter/s)": 1.309477
+    },
+    {
+      "acc": 0.66602869,
+      "epoch": 0.18290208016235412,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.953098873499705e-06,
+      "loss": 1.49825249,
+      "memory(GiB)": 84.03,
+      "step": 7210,
+      "train_speed(iter/s)": 1.309587
+    },
+    {
+      "acc": 0.68344612,
+      "epoch": 0.1830289193302892,
+      "grad_norm": 2.75,
+      "learning_rate": 9.952955473063356e-06,
+      "loss": 1.39401865,
+      "memory(GiB)": 84.03,
+      "step": 7215,
+      "train_speed(iter/s)": 1.309692
+    },
+    {
+      "acc": 0.66496706,
+      "epoch": 0.18315575849822424,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.952811854773911e-06,
+      "loss": 1.50802059,
+      "memory(GiB)": 84.03,
+      "step": 7220,
+      "train_speed(iter/s)": 1.309796
+    },
+    {
+      "acc": 0.66755295,
+      "epoch": 0.1832825976661593,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.952668018637687e-06,
+      "loss": 1.48519135,
+      "memory(GiB)": 84.03,
+      "step": 7225,
+      "train_speed(iter/s)": 1.309901
+    },
+    {
+      "acc": 0.67847137,
+      "epoch": 0.18340943683409436,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.952523964661014e-06,
+      "loss": 1.43496857,
+      "memory(GiB)": 84.03,
+      "step": 7230,
+      "train_speed(iter/s)": 1.310005
+    },
+    {
+      "acc": 0.67980013,
+      "epoch": 0.18353627600202943,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.952379692850222e-06,
+      "loss": 1.45012016,
+      "memory(GiB)": 84.03,
+      "step": 7235,
+      "train_speed(iter/s)": 1.310113
+    },
+    {
+      "acc": 0.68692102,
+      "epoch": 0.18366311516996447,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.952235203211663e-06,
+      "loss": 1.39391155,
+      "memory(GiB)": 84.03,
+      "step": 7240,
+      "train_speed(iter/s)": 1.310218
+    },
+    {
+      "acc": 0.68799305,
+      "epoch": 0.18378995433789955,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.952090495751689e-06,
+      "loss": 1.41769028,
+      "memory(GiB)": 84.03,
+      "step": 7245,
+      "train_speed(iter/s)": 1.310323
+    },
+    {
+      "acc": 0.67789345,
+      "epoch": 0.1839167935058346,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.951945570476666e-06,
+      "loss": 1.44823208,
+      "memory(GiB)": 84.03,
+      "step": 7250,
+      "train_speed(iter/s)": 1.310428
+    },
+    {
+      "acc": 0.66524448,
+      "epoch": 0.18404363267376966,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.951800427392968e-06,
+      "loss": 1.51081676,
+      "memory(GiB)": 84.03,
+      "step": 7255,
+      "train_speed(iter/s)": 1.310529
+    },
+    {
+      "acc": 0.67477579,
+      "epoch": 0.1841704718417047,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.951655066506977e-06,
+      "loss": 1.50624962,
+      "memory(GiB)": 84.03,
+      "step": 7260,
+      "train_speed(iter/s)": 1.310633
+    },
+    {
+      "acc": 0.67412033,
+      "epoch": 0.18429731100963978,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.951509487825091e-06,
+      "loss": 1.43544178,
+      "memory(GiB)": 84.03,
+      "step": 7265,
+      "train_speed(iter/s)": 1.310734
+    },
+    {
+      "acc": 0.66520596,
+      "epoch": 0.18442415017757482,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.95136369135371e-06,
+      "loss": 1.50105076,
+      "memory(GiB)": 84.03,
+      "step": 7270,
+      "train_speed(iter/s)": 1.310828
+    },
+    {
+      "acc": 0.67623215,
+      "epoch": 0.1845509893455099,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.951217677099248e-06,
+      "loss": 1.47483644,
+      "memory(GiB)": 84.03,
+      "step": 7275,
+      "train_speed(iter/s)": 1.310933
+    },
+    {
+      "acc": 0.67984581,
+      "epoch": 0.18467782851344494,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.951071445068125e-06,
+      "loss": 1.43034477,
+      "memory(GiB)": 84.03,
+      "step": 7280,
+      "train_speed(iter/s)": 1.311035
+    },
+    {
+      "acc": 0.67290678,
+      "epoch": 0.18480466768138,
+      "grad_norm": 3.25,
+      "learning_rate": 9.950924995266778e-06,
+      "loss": 1.45238743,
+      "memory(GiB)": 84.03,
+      "step": 7285,
+      "train_speed(iter/s)": 1.311139
+    },
+    {
+      "acc": 0.68125076,
+      "epoch": 0.18493150684931506,
+      "grad_norm": 3.25,
+      "learning_rate": 9.950778327701643e-06,
+      "loss": 1.43590965,
+      "memory(GiB)": 84.03,
+      "step": 7290,
+      "train_speed(iter/s)": 1.31124
+    },
+    {
+      "acc": 0.69386015,
+      "epoch": 0.18505834601725013,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.950631442379175e-06,
+      "loss": 1.42613373,
+      "memory(GiB)": 84.03,
+      "step": 7295,
+      "train_speed(iter/s)": 1.311346
+    },
+    {
+      "acc": 0.67993126,
+      "epoch": 0.18518518518518517,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.950484339305832e-06,
+      "loss": 1.42139616,
+      "memory(GiB)": 84.03,
+      "step": 7300,
+      "train_speed(iter/s)": 1.311451
+    },
+    {
+      "acc": 0.68248596,
+      "epoch": 0.18531202435312025,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.950337018488086e-06,
+      "loss": 1.46876621,
+      "memory(GiB)": 84.03,
+      "step": 7305,
+      "train_speed(iter/s)": 1.31155
+    },
+    {
+      "acc": 0.6674315,
+      "epoch": 0.1854388635210553,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.950189479932417e-06,
+      "loss": 1.46871643,
+      "memory(GiB)": 84.03,
+      "step": 7310,
+      "train_speed(iter/s)": 1.311649
+    },
+    {
+      "acc": 0.68134809,
+      "epoch": 0.18556570268899036,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.950041723645312e-06,
+      "loss": 1.44708014,
+      "memory(GiB)": 84.03,
+      "step": 7315,
+      "train_speed(iter/s)": 1.311749
+    },
+    {
+      "acc": 0.67354293,
+      "epoch": 0.1856925418569254,
+      "grad_norm": 2.75,
+      "learning_rate": 9.949893749633273e-06,
+      "loss": 1.48168688,
+      "memory(GiB)": 84.03,
+      "step": 7320,
+      "train_speed(iter/s)": 1.311848
+    },
+    {
+      "acc": 0.67603221,
+      "epoch": 0.18581938102486048,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.949745557902806e-06,
+      "loss": 1.50283823,
+      "memory(GiB)": 84.03,
+      "step": 7325,
+      "train_speed(iter/s)": 1.311948
+    },
+    {
+      "acc": 0.68161259,
+      "epoch": 0.18594622019279552,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.949597148460433e-06,
+      "loss": 1.47765341,
+      "memory(GiB)": 84.03,
+      "step": 7330,
+      "train_speed(iter/s)": 1.312045
+    },
+    {
+      "acc": 0.67087021,
+      "epoch": 0.1860730593607306,
+      "grad_norm": 3.125,
+      "learning_rate": 9.949448521312676e-06,
+      "loss": 1.55303516,
+      "memory(GiB)": 84.03,
+      "step": 7335,
+      "train_speed(iter/s)": 1.312143
+    },
+    {
+      "acc": 0.67450504,
+      "epoch": 0.18619989852866564,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.949299676466077e-06,
+      "loss": 1.476161,
+      "memory(GiB)": 84.03,
+      "step": 7340,
+      "train_speed(iter/s)": 1.312242
+    },
+    {
+      "acc": 0.67457752,
+      "epoch": 0.1863267376966007,
+      "grad_norm": 3.125,
+      "learning_rate": 9.94915061392718e-06,
+      "loss": 1.41340923,
+      "memory(GiB)": 84.03,
+      "step": 7345,
+      "train_speed(iter/s)": 1.312344
+    },
+    {
+      "acc": 0.68947668,
+      "epoch": 0.18645357686453576,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.949001333702543e-06,
+      "loss": 1.38865452,
+      "memory(GiB)": 84.03,
+      "step": 7350,
+      "train_speed(iter/s)": 1.312443
+    },
+    {
+      "acc": 0.66936011,
+      "epoch": 0.18658041603247083,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.948851835798732e-06,
+      "loss": 1.55544338,
+      "memory(GiB)": 84.03,
+      "step": 7355,
+      "train_speed(iter/s)": 1.312545
+    },
+    {
+      "acc": 0.66420641,
+      "epoch": 0.18670725520040587,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.948702120222323e-06,
+      "loss": 1.49988871,
+      "memory(GiB)": 84.03,
+      "step": 7360,
+      "train_speed(iter/s)": 1.31264
+    },
+    {
+      "acc": 0.68937979,
+      "epoch": 0.18683409436834095,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.9485521869799e-06,
+      "loss": 1.40261984,
+      "memory(GiB)": 84.03,
+      "step": 7365,
+      "train_speed(iter/s)": 1.312734
+    },
+    {
+      "acc": 0.69927645,
+      "epoch": 0.186960933536276,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.948402036078057e-06,
+      "loss": 1.35144348,
+      "memory(GiB)": 84.03,
+      "step": 7370,
+      "train_speed(iter/s)": 1.312834
+    },
+    {
+      "acc": 0.66904383,
+      "epoch": 0.18708777270421106,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.948251667523401e-06,
+      "loss": 1.49804792,
+      "memory(GiB)": 84.03,
+      "step": 7375,
+      "train_speed(iter/s)": 1.312932
+    },
+    {
+      "acc": 0.66316681,
+      "epoch": 0.1872146118721461,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.948101081322544e-06,
+      "loss": 1.39315157,
+      "memory(GiB)": 84.03,
+      "step": 7380,
+      "train_speed(iter/s)": 1.313027
+    },
+    {
+      "acc": 0.68414564,
+      "epoch": 0.18734145104008118,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.947950277482109e-06,
+      "loss": 1.41299896,
+      "memory(GiB)": 84.03,
+      "step": 7385,
+      "train_speed(iter/s)": 1.31313
+    },
+    {
+      "acc": 0.67990084,
+      "epoch": 0.18746829020801623,
+      "grad_norm": 3.125,
+      "learning_rate": 9.94779925600873e-06,
+      "loss": 1.44724674,
+      "memory(GiB)": 84.03,
+      "step": 7390,
+      "train_speed(iter/s)": 1.313232
+    },
+    {
+      "acc": 0.69267759,
+      "epoch": 0.1875951293759513,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.947648016909048e-06,
+      "loss": 1.41744938,
+      "memory(GiB)": 84.03,
+      "step": 7395,
+      "train_speed(iter/s)": 1.313334
+    },
+    {
+      "acc": 0.68370476,
+      "epoch": 0.18772196854388634,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.947496560189717e-06,
+      "loss": 1.43576851,
+      "memory(GiB)": 84.03,
+      "step": 7400,
+      "train_speed(iter/s)": 1.313434
+    },
+    {
+      "acc": 0.68065386,
+      "epoch": 0.18784880771182141,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.9473448858574e-06,
+      "loss": 1.45861034,
+      "memory(GiB)": 84.03,
+      "step": 7405,
+      "train_speed(iter/s)": 1.31353
+    },
+    {
+      "acc": 0.65323114,
+      "epoch": 0.18797564687975646,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.947192993918765e-06,
+      "loss": 1.53791122,
+      "memory(GiB)": 84.03,
+      "step": 7410,
+      "train_speed(iter/s)": 1.313629
+    },
+    {
+      "acc": 0.67184048,
+      "epoch": 0.18810248604769153,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.947040884380496e-06,
+      "loss": 1.47103405,
+      "memory(GiB)": 84.03,
+      "step": 7415,
+      "train_speed(iter/s)": 1.313725
+    },
+    {
+      "acc": 0.68964567,
+      "epoch": 0.18822932521562658,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.946888557249281e-06,
+      "loss": 1.43060274,
+      "memory(GiB)": 84.03,
+      "step": 7420,
+      "train_speed(iter/s)": 1.313825
+    },
+    {
+      "acc": 0.68989468,
+      "epoch": 0.18835616438356165,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.946736012531821e-06,
+      "loss": 1.41697702,
+      "memory(GiB)": 84.03,
+      "step": 7425,
+      "train_speed(iter/s)": 1.313926
+    },
+    {
+      "acc": 0.68144455,
+      "epoch": 0.1884830035514967,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.946583250234826e-06,
+      "loss": 1.43501472,
+      "memory(GiB)": 84.03,
+      "step": 7430,
+      "train_speed(iter/s)": 1.314028
+    },
+    {
+      "acc": 0.68761392,
+      "epoch": 0.18860984271943176,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.946430270365015e-06,
+      "loss": 1.38412247,
+      "memory(GiB)": 84.03,
+      "step": 7435,
+      "train_speed(iter/s)": 1.314128
+    },
+    {
+      "acc": 0.68826871,
+      "epoch": 0.1887366818873668,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.946277072929115e-06,
+      "loss": 1.43229542,
+      "memory(GiB)": 84.03,
+      "step": 7440,
+      "train_speed(iter/s)": 1.314224
+    },
+    {
+      "acc": 0.68151021,
+      "epoch": 0.18886352105530188,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.946123657933867e-06,
+      "loss": 1.45132332,
+      "memory(GiB)": 84.03,
+      "step": 7445,
+      "train_speed(iter/s)": 1.314319
+    },
+    {
+      "acc": 0.67372036,
+      "epoch": 0.18899036022323693,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.945970025386018e-06,
+      "loss": 1.39355183,
+      "memory(GiB)": 84.03,
+      "step": 7450,
+      "train_speed(iter/s)": 1.314415
+    },
+    {
+      "acc": 0.67898121,
+      "epoch": 0.189117199391172,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.945816175292326e-06,
+      "loss": 1.40652752,
+      "memory(GiB)": 84.03,
+      "step": 7455,
+      "train_speed(iter/s)": 1.314508
+    },
+    {
+      "acc": 0.67498064,
+      "epoch": 0.18924403855910704,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.945662107659554e-06,
+      "loss": 1.49501972,
+      "memory(GiB)": 84.03,
+      "step": 7460,
+      "train_speed(iter/s)": 1.314609
+    },
+    {
+      "acc": 0.6889605,
+      "epoch": 0.18937087772704211,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.945507822494485e-06,
+      "loss": 1.42675552,
+      "memory(GiB)": 84.03,
+      "step": 7465,
+      "train_speed(iter/s)": 1.314707
+    },
+    {
+      "acc": 0.6774672,
+      "epoch": 0.18949771689497716,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.9453533198039e-06,
+      "loss": 1.44061108,
+      "memory(GiB)": 84.03,
+      "step": 7470,
+      "train_speed(iter/s)": 1.314801
+    },
+    {
+      "acc": 0.66861944,
+      "epoch": 0.18962455606291223,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.945198599594598e-06,
+      "loss": 1.44858541,
+      "memory(GiB)": 84.03,
+      "step": 7475,
+      "train_speed(iter/s)": 1.3149
+    },
+    {
+      "acc": 0.66350565,
+      "epoch": 0.18975139523084728,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.945043661873381e-06,
+      "loss": 1.50099716,
+      "memory(GiB)": 84.03,
+      "step": 7480,
+      "train_speed(iter/s)": 1.314997
+    },
+    {
+      "acc": 0.68524041,
+      "epoch": 0.18987823439878235,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.944888506647066e-06,
+      "loss": 1.4075717,
+      "memory(GiB)": 84.03,
+      "step": 7485,
+      "train_speed(iter/s)": 1.315088
+    },
+    {
+      "acc": 0.6847971,
+      "epoch": 0.1900050735667174,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.944733133922479e-06,
+      "loss": 1.43658047,
+      "memory(GiB)": 84.03,
+      "step": 7490,
+      "train_speed(iter/s)": 1.31518
+    },
+    {
+      "acc": 0.67941046,
+      "epoch": 0.19013191273465246,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.944577543706451e-06,
+      "loss": 1.43483753,
+      "memory(GiB)": 84.03,
+      "step": 7495,
+      "train_speed(iter/s)": 1.315278
+    },
+    {
+      "acc": 0.67788687,
+      "epoch": 0.1902587519025875,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.944421736005825e-06,
+      "loss": 1.43910522,
+      "memory(GiB)": 84.03,
+      "step": 7500,
+      "train_speed(iter/s)": 1.315374
+    },
+    {
+      "acc": 0.66725903,
+      "epoch": 0.19038559107052258,
+      "grad_norm": 2.875,
+      "learning_rate": 9.944265710827459e-06,
+      "loss": 1.51340065,
+      "memory(GiB)": 84.03,
+      "step": 7505,
+      "train_speed(iter/s)": 1.315468
+    },
+    {
+      "acc": 0.68192854,
+      "epoch": 0.19051243023845763,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.944109468178208e-06,
+      "loss": 1.41813679,
+      "memory(GiB)": 84.03,
+      "step": 7510,
+      "train_speed(iter/s)": 1.315562
+    },
+    {
+      "acc": 0.67877798,
+      "epoch": 0.1906392694063927,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.943953008064953e-06,
+      "loss": 1.44127426,
+      "memory(GiB)": 84.03,
+      "step": 7515,
+      "train_speed(iter/s)": 1.31566
+    },
+    {
+      "acc": 0.68017912,
+      "epoch": 0.19076610857432774,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.94379633049457e-06,
+      "loss": 1.45065117,
+      "memory(GiB)": 84.03,
+      "step": 7520,
+      "train_speed(iter/s)": 1.315759
+    },
+    {
+      "acc": 0.6824482,
+      "epoch": 0.19089294774226281,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.943639435473952e-06,
+      "loss": 1.45466499,
+      "memory(GiB)": 84.03,
+      "step": 7525,
+      "train_speed(iter/s)": 1.315858
+    },
+    {
+      "acc": 0.69053373,
+      "epoch": 0.19101978691019786,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.94348232301e-06,
+      "loss": 1.40129337,
+      "memory(GiB)": 84.03,
+      "step": 7530,
+      "train_speed(iter/s)": 1.315953
+    },
+    {
+      "acc": 0.67054586,
+      "epoch": 0.19114662607813293,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.943324993109624e-06,
+      "loss": 1.46407557,
+      "memory(GiB)": 84.03,
+      "step": 7535,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.68723621,
+      "epoch": 0.19127346524606798,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.943167445779745e-06,
+      "loss": 1.38581934,
+      "memory(GiB)": 84.03,
+      "step": 7540,
+      "train_speed(iter/s)": 1.31614
+    },
+    {
+      "acc": 0.69762678,
+      "epoch": 0.19140030441400305,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.94300968102729e-06,
+      "loss": 1.38730621,
+      "memory(GiB)": 84.03,
+      "step": 7545,
+      "train_speed(iter/s)": 1.316238
+    },
+    {
+      "acc": 0.68571568,
+      "epoch": 0.1915271435819381,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.942851698859204e-06,
+      "loss": 1.42588005,
+      "memory(GiB)": 84.03,
+      "step": 7550,
+      "train_speed(iter/s)": 1.316331
+    },
+    {
+      "acc": 0.69416919,
+      "epoch": 0.19165398274987316,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.94269349928243e-06,
+      "loss": 1.37223206,
+      "memory(GiB)": 84.03,
+      "step": 7555,
+      "train_speed(iter/s)": 1.316428
+    },
+    {
+      "acc": 0.69289274,
+      "epoch": 0.1917808219178082,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.942535082303927e-06,
+      "loss": 1.37428894,
+      "memory(GiB)": 84.03,
+      "step": 7560,
+      "train_speed(iter/s)": 1.316513
+    },
+    {
+      "acc": 0.67496805,
+      "epoch": 0.19190766108574328,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.942376447930666e-06,
+      "loss": 1.45934057,
+      "memory(GiB)": 84.03,
+      "step": 7565,
+      "train_speed(iter/s)": 1.316612
+    },
+    {
+      "acc": 0.69286795,
+      "epoch": 0.19203450025367833,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.942217596169623e-06,
+      "loss": 1.34902401,
+      "memory(GiB)": 84.03,
+      "step": 7570,
+      "train_speed(iter/s)": 1.3167
+    },
+    {
+      "acc": 0.66057596,
+      "epoch": 0.1921613394216134,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.942058527027785e-06,
+      "loss": 1.48827353,
+      "memory(GiB)": 84.03,
+      "step": 7575,
+      "train_speed(iter/s)": 1.316784
+    },
+    {
+      "acc": 0.66577826,
+      "epoch": 0.19228817858954844,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.941899240512147e-06,
+      "loss": 1.45141344,
+      "memory(GiB)": 84.03,
+      "step": 7580,
+      "train_speed(iter/s)": 1.316874
+    },
+    {
+      "acc": 0.67463646,
+      "epoch": 0.19241501775748351,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.941739736629716e-06,
+      "loss": 1.4525424,
+      "memory(GiB)": 84.03,
+      "step": 7585,
+      "train_speed(iter/s)": 1.316965
+    },
+    {
+      "acc": 0.68377275,
+      "epoch": 0.19254185692541856,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.941580015387509e-06,
+      "loss": 1.38758593,
+      "memory(GiB)": 84.03,
+      "step": 7590,
+      "train_speed(iter/s)": 1.317051
+    },
+    {
+      "acc": 0.65429235,
+      "epoch": 0.19266869609335363,
+      "grad_norm": 4.9375,
+      "learning_rate": 9.94142007679255e-06,
+      "loss": 1.5320941,
+      "memory(GiB)": 84.03,
+      "step": 7595,
+      "train_speed(iter/s)": 1.317149
+    },
+    {
+      "acc": 0.6965529,
+      "epoch": 0.19279553526128868,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.941259920851874e-06,
+      "loss": 1.37761059,
+      "memory(GiB)": 84.03,
+      "step": 7600,
+      "train_speed(iter/s)": 1.317241
+    },
+    {
+      "acc": 0.68115063,
+      "epoch": 0.19292237442922375,
+      "grad_norm": 4.25,
+      "learning_rate": 9.941099547572527e-06,
+      "loss": 1.44935398,
+      "memory(GiB)": 84.03,
+      "step": 7605,
+      "train_speed(iter/s)": 1.317341
+    },
+    {
+      "acc": 0.68697271,
+      "epoch": 0.1930492135971588,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.94093895696156e-06,
+      "loss": 1.42814198,
+      "memory(GiB)": 84.03,
+      "step": 7610,
+      "train_speed(iter/s)": 1.317432
+    },
+    {
+      "acc": 0.67490063,
+      "epoch": 0.19317605276509386,
+      "grad_norm": 3.375,
+      "learning_rate": 9.940778149026038e-06,
+      "loss": 1.46020784,
+      "memory(GiB)": 84.03,
+      "step": 7615,
+      "train_speed(iter/s)": 1.317526
+    },
+    {
+      "acc": 0.67825308,
+      "epoch": 0.1933028919330289,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.940617123773036e-06,
+      "loss": 1.46240873,
+      "memory(GiB)": 84.03,
+      "step": 7620,
+      "train_speed(iter/s)": 1.31762
+    },
+    {
+      "acc": 0.69209251,
+      "epoch": 0.19342973110096398,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.940455881209632e-06,
+      "loss": 1.35018225,
+      "memory(GiB)": 84.03,
+      "step": 7625,
+      "train_speed(iter/s)": 1.317713
+    },
+    {
+      "acc": 0.6828917,
+      "epoch": 0.19355657026889903,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.940294421342922e-06,
+      "loss": 1.35995159,
+      "memory(GiB)": 84.03,
+      "step": 7630,
+      "train_speed(iter/s)": 1.317805
+    },
+    {
+      "acc": 0.66362257,
+      "epoch": 0.1936834094368341,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.940132744180007e-06,
+      "loss": 1.49226894,
+      "memory(GiB)": 84.03,
+      "step": 7635,
+      "train_speed(iter/s)": 1.3179
+    },
+    {
+      "acc": 0.68995533,
+      "epoch": 0.19381024860476914,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.939970849727995e-06,
+      "loss": 1.3889183,
+      "memory(GiB)": 84.03,
+      "step": 7640,
+      "train_speed(iter/s)": 1.31799
+    },
+    {
+      "acc": 0.66431642,
+      "epoch": 0.19393708777270421,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.939808737994013e-06,
+      "loss": 1.51099482,
+      "memory(GiB)": 84.03,
+      "step": 7645,
+      "train_speed(iter/s)": 1.318085
+    },
+    {
+      "acc": 0.6835844,
+      "epoch": 0.19406392694063926,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.939646408985186e-06,
+      "loss": 1.39035568,
+      "memory(GiB)": 84.03,
+      "step": 7650,
+      "train_speed(iter/s)": 1.318179
+    },
+    {
+      "acc": 0.68823462,
+      "epoch": 0.19419076610857433,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.939483862708658e-06,
+      "loss": 1.40803146,
+      "memory(GiB)": 84.03,
+      "step": 7655,
+      "train_speed(iter/s)": 1.318274
+    },
+    {
+      "acc": 0.67263041,
+      "epoch": 0.19431760527650938,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.939321099171575e-06,
+      "loss": 1.41558456,
+      "memory(GiB)": 84.03,
+      "step": 7660,
+      "train_speed(iter/s)": 1.318369
+    },
+    {
+      "acc": 0.67389736,
+      "epoch": 0.19444444444444445,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.939158118381097e-06,
+      "loss": 1.48367462,
+      "memory(GiB)": 84.03,
+      "step": 7665,
+      "train_speed(iter/s)": 1.318463
+    },
+    {
+      "acc": 0.67939835,
+      "epoch": 0.1945712836123795,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.938994920344395e-06,
+      "loss": 1.4643074,
+      "memory(GiB)": 84.03,
+      "step": 7670,
+      "train_speed(iter/s)": 1.318556
+    },
+    {
+      "acc": 0.68373051,
+      "epoch": 0.19469812278031456,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.938831505068645e-06,
+      "loss": 1.38951416,
+      "memory(GiB)": 84.03,
+      "step": 7675,
+      "train_speed(iter/s)": 1.318653
+    },
+    {
+      "acc": 0.68185391,
+      "epoch": 0.1948249619482496,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.938667872561035e-06,
+      "loss": 1.37099571,
+      "memory(GiB)": 84.03,
+      "step": 7680,
+      "train_speed(iter/s)": 1.318736
+    },
+    {
+      "acc": 0.69897089,
+      "epoch": 0.19495180111618468,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.938504022828762e-06,
+      "loss": 1.35489922,
+      "memory(GiB)": 84.03,
+      "step": 7685,
+      "train_speed(iter/s)": 1.318836
+    },
+    {
+      "acc": 0.67795639,
+      "epoch": 0.19507864028411973,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.938339955879033e-06,
+      "loss": 1.42139912,
+      "memory(GiB)": 84.03,
+      "step": 7690,
+      "train_speed(iter/s)": 1.318921
+    },
+    {
+      "acc": 0.66791234,
+      "epoch": 0.1952054794520548,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.938175671719064e-06,
+      "loss": 1.46826105,
+      "memory(GiB)": 84.03,
+      "step": 7695,
+      "train_speed(iter/s)": 1.319016
+    },
+    {
+      "acc": 0.66470747,
+      "epoch": 0.19533231861998984,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.938011170356083e-06,
+      "loss": 1.43862038,
+      "memory(GiB)": 84.03,
+      "step": 7700,
+      "train_speed(iter/s)": 1.319106
+    },
+    {
+      "acc": 0.67323523,
+      "epoch": 0.19545915778792491,
+      "grad_norm": 4.5,
+      "learning_rate": 9.937846451797324e-06,
+      "loss": 1.43579435,
+      "memory(GiB)": 84.03,
+      "step": 7705,
+      "train_speed(iter/s)": 1.319199
+    },
+    {
+      "acc": 0.68678741,
+      "epoch": 0.19558599695585996,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.93768151605003e-06,
+      "loss": 1.39313469,
+      "memory(GiB)": 84.03,
+      "step": 7710,
+      "train_speed(iter/s)": 1.319287
+    },
+    {
+      "acc": 0.68875055,
+      "epoch": 0.19571283612379503,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.93751636312146e-06,
+      "loss": 1.39899912,
+      "memory(GiB)": 84.03,
+      "step": 7715,
+      "train_speed(iter/s)": 1.319379
+    },
+    {
+      "acc": 0.66380777,
+      "epoch": 0.19583967529173008,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.937350993018875e-06,
+      "loss": 1.52167645,
+      "memory(GiB)": 84.03,
+      "step": 7720,
+      "train_speed(iter/s)": 1.31947
+    },
+    {
+      "acc": 0.67729416,
+      "epoch": 0.19596651445966515,
+      "grad_norm": 2.75,
+      "learning_rate": 9.93718540574955e-06,
+      "loss": 1.4212079,
+      "memory(GiB)": 84.03,
+      "step": 7725,
+      "train_speed(iter/s)": 1.319566
+    },
+    {
+      "acc": 0.66409655,
+      "epoch": 0.1960933536276002,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.937019601320768e-06,
+      "loss": 1.49378157,
+      "memory(GiB)": 84.03,
+      "step": 7730,
+      "train_speed(iter/s)": 1.319657
+    },
+    {
+      "acc": 0.67241449,
+      "epoch": 0.19622019279553526,
+      "grad_norm": 3.125,
+      "learning_rate": 9.936853579739823e-06,
+      "loss": 1.44417133,
+      "memory(GiB)": 84.03,
+      "step": 7735,
+      "train_speed(iter/s)": 1.31975
+    },
+    {
+      "acc": 0.66704335,
+      "epoch": 0.1963470319634703,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.936687341014015e-06,
+      "loss": 1.50907316,
+      "memory(GiB)": 84.03,
+      "step": 7740,
+      "train_speed(iter/s)": 1.319842
+    },
+    {
+      "acc": 0.68493929,
+      "epoch": 0.19647387113140538,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.936520885150655e-06,
+      "loss": 1.37331657,
+      "memory(GiB)": 84.03,
+      "step": 7745,
+      "train_speed(iter/s)": 1.319938
+    },
+    {
+      "acc": 0.675179,
+      "epoch": 0.19660071029934043,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.936354212157068e-06,
+      "loss": 1.49210329,
+      "memory(GiB)": 84.03,
+      "step": 7750,
+      "train_speed(iter/s)": 1.320029
+    },
+    {
+      "acc": 0.67343616,
+      "epoch": 0.1967275494672755,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.936187322040584e-06,
+      "loss": 1.4736433,
+      "memory(GiB)": 84.03,
+      "step": 7755,
+      "train_speed(iter/s)": 1.32012
+    },
+    {
+      "acc": 0.66771679,
+      "epoch": 0.19685438863521054,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.936020214808544e-06,
+      "loss": 1.49195976,
+      "memory(GiB)": 84.03,
+      "step": 7760,
+      "train_speed(iter/s)": 1.320211
+    },
+    {
+      "acc": 0.68670335,
+      "epoch": 0.19698122780314561,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.935852890468297e-06,
+      "loss": 1.43899727,
+      "memory(GiB)": 84.03,
+      "step": 7765,
+      "train_speed(iter/s)": 1.320303
+    },
+    {
+      "acc": 0.67139673,
+      "epoch": 0.19710806697108066,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.935685349027201e-06,
+      "loss": 1.44236298,
+      "memory(GiB)": 84.03,
+      "step": 7770,
+      "train_speed(iter/s)": 1.320393
+    },
+    {
+      "acc": 0.67762966,
+      "epoch": 0.19723490613901573,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.935517590492627e-06,
+      "loss": 1.43566055,
+      "memory(GiB)": 84.03,
+      "step": 7775,
+      "train_speed(iter/s)": 1.320486
+    },
+    {
+      "acc": 0.69273314,
+      "epoch": 0.19736174530695078,
+      "grad_norm": 2.875,
+      "learning_rate": 9.935349614871957e-06,
+      "loss": 1.39784718,
+      "memory(GiB)": 84.03,
+      "step": 7780,
+      "train_speed(iter/s)": 1.320579
+    },
+    {
+      "acc": 0.6888485,
+      "epoch": 0.19748858447488585,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.935181422172574e-06,
+      "loss": 1.45105133,
+      "memory(GiB)": 84.03,
+      "step": 7785,
+      "train_speed(iter/s)": 1.320672
+    },
+    {
+      "acc": 0.66913505,
+      "epoch": 0.1976154236428209,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.935013012401878e-06,
+      "loss": 1.54231205,
+      "memory(GiB)": 84.03,
+      "step": 7790,
+      "train_speed(iter/s)": 1.320762
+    },
+    {
+      "acc": 0.68240309,
+      "epoch": 0.19774226281075596,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.934844385567275e-06,
+      "loss": 1.38855476,
+      "memory(GiB)": 84.03,
+      "step": 7795,
+      "train_speed(iter/s)": 1.320853
+    },
+    {
+      "acc": 0.68014274,
+      "epoch": 0.197869101978691,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.934675541676186e-06,
+      "loss": 1.41418219,
+      "memory(GiB)": 84.03,
+      "step": 7800,
+      "train_speed(iter/s)": 1.320945
+    },
+    {
+      "acc": 0.69413924,
+      "epoch": 0.19799594114662608,
+      "grad_norm": 3.625,
+      "learning_rate": 9.934506480736034e-06,
+      "loss": 1.38484278,
+      "memory(GiB)": 84.03,
+      "step": 7805,
+      "train_speed(iter/s)": 1.321037
+    },
+    {
+      "acc": 0.6909873,
+      "epoch": 0.19812278031456113,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.934337202754257e-06,
+      "loss": 1.41167126,
+      "memory(GiB)": 84.03,
+      "step": 7810,
+      "train_speed(iter/s)": 1.32113
+    },
+    {
+      "acc": 0.68179002,
+      "epoch": 0.1982496194824962,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.934167707738298e-06,
+      "loss": 1.4743474,
+      "memory(GiB)": 84.03,
+      "step": 7815,
+      "train_speed(iter/s)": 1.321222
+    },
+    {
+      "acc": 0.67986736,
+      "epoch": 0.19837645865043124,
+      "grad_norm": 3.75,
+      "learning_rate": 9.933997995695615e-06,
+      "loss": 1.42975845,
+      "memory(GiB)": 84.03,
+      "step": 7820,
+      "train_speed(iter/s)": 1.321303
+    },
+    {
+      "acc": 0.67408085,
+      "epoch": 0.19850329781836631,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.93382806663367e-06,
+      "loss": 1.49499416,
+      "memory(GiB)": 84.03,
+      "step": 7825,
+      "train_speed(iter/s)": 1.321393
+    },
+    {
+      "acc": 0.68233223,
+      "epoch": 0.19863013698630136,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.933657920559939e-06,
+      "loss": 1.41208658,
+      "memory(GiB)": 84.03,
+      "step": 7830,
+      "train_speed(iter/s)": 1.321487
+    },
+    {
+      "acc": 0.67893314,
+      "epoch": 0.19875697615423643,
+      "grad_norm": 3.0,
+      "learning_rate": 9.933487557481905e-06,
+      "loss": 1.45575218,
+      "memory(GiB)": 84.03,
+      "step": 7835,
+      "train_speed(iter/s)": 1.321575
+    },
+    {
+      "acc": 0.68127775,
+      "epoch": 0.19888381532217148,
+      "grad_norm": 2.875,
+      "learning_rate": 9.933316977407063e-06,
+      "loss": 1.50590324,
+      "memory(GiB)": 84.03,
+      "step": 7840,
+      "train_speed(iter/s)": 1.321664
+    },
+    {
+      "acc": 0.69573269,
+      "epoch": 0.19901065449010655,
+      "grad_norm": 2.75,
+      "learning_rate": 9.933146180342914e-06,
+      "loss": 1.33563299,
+      "memory(GiB)": 84.03,
+      "step": 7845,
+      "train_speed(iter/s)": 1.321753
+    },
+    {
+      "acc": 0.68755722,
+      "epoch": 0.1991374936580416,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.932975166296972e-06,
+      "loss": 1.43208256,
+      "memory(GiB)": 84.03,
+      "step": 7850,
+      "train_speed(iter/s)": 1.321842
+    },
+    {
+      "acc": 0.67921944,
+      "epoch": 0.19926433282597666,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.932803935276757e-06,
+      "loss": 1.38357077,
+      "memory(GiB)": 84.03,
+      "step": 7855,
+      "train_speed(iter/s)": 1.321932
+    },
+    {
+      "acc": 0.6736824,
+      "epoch": 0.1993911719939117,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.932632487289802e-06,
+      "loss": 1.49717922,
+      "memory(GiB)": 84.03,
+      "step": 7860,
+      "train_speed(iter/s)": 1.322024
+    },
+    {
+      "acc": 0.67750263,
+      "epoch": 0.19951801116184678,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.932460822343649e-06,
+      "loss": 1.42597103,
+      "memory(GiB)": 84.03,
+      "step": 7865,
+      "train_speed(iter/s)": 1.322105
+    },
+    {
+      "acc": 0.66489992,
+      "epoch": 0.19964485032978183,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.932288940445845e-06,
+      "loss": 1.57015553,
+      "memory(GiB)": 84.03,
+      "step": 7870,
+      "train_speed(iter/s)": 1.322193
+    },
+    {
+      "acc": 0.68323755,
+      "epoch": 0.1997716894977169,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.932116841603954e-06,
+      "loss": 1.39531469,
+      "memory(GiB)": 84.03,
+      "step": 7875,
+      "train_speed(iter/s)": 1.32228
+    },
+    {
+      "acc": 0.6840003,
+      "epoch": 0.19989852866565194,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.931944525825542e-06,
+      "loss": 1.42870617,
+      "memory(GiB)": 84.03,
+      "step": 7880,
+      "train_speed(iter/s)": 1.322367
+    },
+    {
+      "acc": 0.67854748,
+      "epoch": 0.20002536783358701,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.931771993118191e-06,
+      "loss": 1.36690464,
+      "memory(GiB)": 84.03,
+      "step": 7885,
+      "train_speed(iter/s)": 1.322451
+    },
+    {
+      "acc": 0.67764301,
+      "epoch": 0.20015220700152206,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.931599243489489e-06,
+      "loss": 1.46186943,
+      "memory(GiB)": 84.03,
+      "step": 7890,
+      "train_speed(iter/s)": 1.322543
+    },
+    {
+      "acc": 0.6804203,
+      "epoch": 0.20027904616945713,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.931426276947037e-06,
+      "loss": 1.42847013,
+      "memory(GiB)": 84.03,
+      "step": 7895,
+      "train_speed(iter/s)": 1.322633
+    },
+    {
+      "acc": 0.67254262,
+      "epoch": 0.20040588533739218,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.931253093498437e-06,
+      "loss": 1.42366667,
+      "memory(GiB)": 84.03,
+      "step": 7900,
+      "train_speed(iter/s)": 1.322717
+    },
+    {
+      "acc": 0.67056856,
+      "epoch": 0.20053272450532725,
+      "grad_norm": 2.3125,
+      "learning_rate": 9.93107969315131e-06,
+      "loss": 1.42955036,
+      "memory(GiB)": 84.03,
+      "step": 7905,
+      "train_speed(iter/s)": 1.322806
+    },
+    {
+      "acc": 0.69048443,
+      "epoch": 0.2006595636732623,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.930906075913281e-06,
+      "loss": 1.44253578,
+      "memory(GiB)": 84.03,
+      "step": 7910,
+      "train_speed(iter/s)": 1.322898
+    },
+    {
+      "acc": 0.6948247,
+      "epoch": 0.20078640284119736,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.93073224179199e-06,
+      "loss": 1.39068289,
+      "memory(GiB)": 84.03,
+      "step": 7915,
+      "train_speed(iter/s)": 1.322985
+    },
+    {
+      "acc": 0.66753607,
+      "epoch": 0.2009132420091324,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.93055819079508e-06,
+      "loss": 1.49200363,
+      "memory(GiB)": 84.03,
+      "step": 7920,
+      "train_speed(iter/s)": 1.323072
+    },
+    {
+      "acc": 0.681954,
+      "epoch": 0.20104008117706748,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.930383922930207e-06,
+      "loss": 1.50620823,
+      "memory(GiB)": 84.03,
+      "step": 7925,
+      "train_speed(iter/s)": 1.323166
+    },
+    {
+      "acc": 0.67144437,
+      "epoch": 0.20116692034500253,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.930209438205038e-06,
+      "loss": 1.45755367,
+      "memory(GiB)": 84.03,
+      "step": 7930,
+      "train_speed(iter/s)": 1.323255
+    },
+    {
+      "acc": 0.68091507,
+      "epoch": 0.2012937595129376,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.930034736627245e-06,
+      "loss": 1.43965921,
+      "memory(GiB)": 84.03,
+      "step": 7935,
+      "train_speed(iter/s)": 1.323346
+    },
+    {
+      "acc": 0.68182707,
+      "epoch": 0.20142059868087264,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.929859818204514e-06,
+      "loss": 1.41062145,
+      "memory(GiB)": 84.03,
+      "step": 7940,
+      "train_speed(iter/s)": 1.323439
+    },
+    {
+      "acc": 0.69579096,
+      "epoch": 0.20154743784880771,
+      "grad_norm": 2.75,
+      "learning_rate": 9.929684682944538e-06,
+      "loss": 1.3060873,
+      "memory(GiB)": 84.03,
+      "step": 7945,
+      "train_speed(iter/s)": 1.323525
+    },
+    {
+      "acc": 0.67020531,
+      "epoch": 0.20167427701674276,
+      "grad_norm": 3.0,
+      "learning_rate": 9.929509330855018e-06,
+      "loss": 1.47111626,
+      "memory(GiB)": 84.03,
+      "step": 7950,
+      "train_speed(iter/s)": 1.323613
+    },
+    {
+      "acc": 0.68020039,
+      "epoch": 0.20180111618467783,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.929333761943672e-06,
+      "loss": 1.49631386,
+      "memory(GiB)": 84.03,
+      "step": 7955,
+      "train_speed(iter/s)": 1.323704
+    },
+    {
+      "acc": 0.66592965,
+      "epoch": 0.20192795535261288,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.929157976218218e-06,
+      "loss": 1.52492876,
+      "memory(GiB)": 84.03,
+      "step": 7960,
+      "train_speed(iter/s)": 1.323796
+    },
+    {
+      "acc": 0.6769516,
+      "epoch": 0.20205479452054795,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.928981973686388e-06,
+      "loss": 1.48214436,
+      "memory(GiB)": 84.03,
+      "step": 7965,
+      "train_speed(iter/s)": 1.323885
+    },
+    {
+      "acc": 0.68802834,
+      "epoch": 0.202181633688483,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.928805754355926e-06,
+      "loss": 1.3488678,
+      "memory(GiB)": 84.03,
+      "step": 7970,
+      "train_speed(iter/s)": 1.32397
+    },
+    {
+      "acc": 0.6820622,
+      "epoch": 0.20230847285641806,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.92862931823458e-06,
+      "loss": 1.40432386,
+      "memory(GiB)": 84.03,
+      "step": 7975,
+      "train_speed(iter/s)": 1.324059
+    },
+    {
+      "acc": 0.6842659,
+      "epoch": 0.2024353120243531,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.928452665330113e-06,
+      "loss": 1.47018433,
+      "memory(GiB)": 84.03,
+      "step": 7980,
+      "train_speed(iter/s)": 1.324142
+    },
+    {
+      "acc": 0.67363539,
+      "epoch": 0.20256215119228818,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.928275795650293e-06,
+      "loss": 1.4857913,
+      "memory(GiB)": 84.03,
+      "step": 7985,
+      "train_speed(iter/s)": 1.324234
+    },
+    {
+      "acc": 0.69394097,
+      "epoch": 0.20268899036022323,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.928098709202901e-06,
+      "loss": 1.39835224,
+      "memory(GiB)": 84.03,
+      "step": 7990,
+      "train_speed(iter/s)": 1.324323
+    },
+    {
+      "acc": 0.67374868,
+      "epoch": 0.2028158295281583,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.927921405995727e-06,
+      "loss": 1.46991777,
+      "memory(GiB)": 84.03,
+      "step": 7995,
+      "train_speed(iter/s)": 1.324409
+    },
+    {
+      "acc": 0.69145174,
+      "epoch": 0.20294266869609334,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.927743886036566e-06,
+      "loss": 1.46408701,
+      "memory(GiB)": 84.03,
+      "step": 8000,
+      "train_speed(iter/s)": 1.324477
+    },
+    {
+      "epoch": 0.20294266869609334,
+      "eval_acc": 0.6673002587411273,
+      "eval_loss": 1.3998042345046997,
+      "eval_runtime": 70.056,
+      "eval_samples_per_second": 90.927,
+      "eval_steps_per_second": 22.739,
+      "step": 8000
+    },
+    {
+      "acc": 0.69222608,
+      "epoch": 0.20306950786402841,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.927566149333228e-06,
+      "loss": 1.3733078,
+      "memory(GiB)": 84.03,
+      "step": 8005,
+      "train_speed(iter/s)": 1.306694
+    },
+    {
+      "acc": 0.67182026,
+      "epoch": 0.20319634703196346,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.92738819589353e-06,
+      "loss": 1.42372398,
+      "memory(GiB)": 84.03,
+      "step": 8010,
+      "train_speed(iter/s)": 1.306789
+    },
+    {
+      "acc": 0.68068123,
+      "epoch": 0.20332318619989853,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.927210025725301e-06,
+      "loss": 1.42104197,
+      "memory(GiB)": 84.03,
+      "step": 8015,
+      "train_speed(iter/s)": 1.306889
+    },
+    {
+      "acc": 0.67508979,
+      "epoch": 0.20345002536783358,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.927031638836377e-06,
+      "loss": 1.47574005,
+      "memory(GiB)": 84.03,
+      "step": 8020,
+      "train_speed(iter/s)": 1.306987
+    },
+    {
+      "acc": 0.6761662,
+      "epoch": 0.20357686453576865,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.926853035234603e-06,
+      "loss": 1.45118294,
+      "memory(GiB)": 84.03,
+      "step": 8025,
+      "train_speed(iter/s)": 1.307072
+    },
+    {
+      "acc": 0.68874674,
+      "epoch": 0.2037037037037037,
+      "grad_norm": 3.5,
+      "learning_rate": 9.926674214927836e-06,
+      "loss": 1.49361095,
+      "memory(GiB)": 84.03,
+      "step": 8030,
+      "train_speed(iter/s)": 1.30717
+    },
+    {
+      "acc": 0.68041549,
+      "epoch": 0.20383054287163876,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.926495177923941e-06,
+      "loss": 1.39434586,
+      "memory(GiB)": 84.03,
+      "step": 8035,
+      "train_speed(iter/s)": 1.307226
+    },
+    {
+      "acc": 0.67497392,
+      "epoch": 0.2039573820395738,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.926315924230794e-06,
+      "loss": 1.43837137,
+      "memory(GiB)": 84.03,
+      "step": 8040,
+      "train_speed(iter/s)": 1.307309
+    },
+    {
+      "acc": 0.67716241,
+      "epoch": 0.20408422120750888,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.926136453856277e-06,
+      "loss": 1.45549746,
+      "memory(GiB)": 84.03,
+      "step": 8045,
+      "train_speed(iter/s)": 1.307401
+    },
+    {
+      "acc": 0.68908148,
+      "epoch": 0.20421106037544393,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.925956766808286e-06,
+      "loss": 1.43444633,
+      "memory(GiB)": 84.03,
+      "step": 8050,
+      "train_speed(iter/s)": 1.307495
+    },
+    {
+      "acc": 0.67160096,
+      "epoch": 0.204337899543379,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.925776863094723e-06,
+      "loss": 1.45596294,
+      "memory(GiB)": 84.03,
+      "step": 8055,
+      "train_speed(iter/s)": 1.307591
+    },
+    {
+      "acc": 0.67125115,
+      "epoch": 0.20446473871131404,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.9255967427235e-06,
+      "loss": 1.48053017,
+      "memory(GiB)": 84.03,
+      "step": 8060,
+      "train_speed(iter/s)": 1.307689
+    },
+    {
+      "acc": 0.6861659,
+      "epoch": 0.20459157787924911,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.925416405702544e-06,
+      "loss": 1.4075983,
+      "memory(GiB)": 84.03,
+      "step": 8065,
+      "train_speed(iter/s)": 1.307788
+    },
+    {
+      "acc": 0.68027048,
+      "epoch": 0.20471841704718416,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.925235852039783e-06,
+      "loss": 1.46062984,
+      "memory(GiB)": 84.03,
+      "step": 8070,
+      "train_speed(iter/s)": 1.307889
+    },
+    {
+      "acc": 0.67161069,
+      "epoch": 0.20484525621511923,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.92505508174316e-06,
+      "loss": 1.49725113,
+      "memory(GiB)": 84.03,
+      "step": 8075,
+      "train_speed(iter/s)": 1.307988
+    },
+    {
+      "acc": 0.68513231,
+      "epoch": 0.20497209538305428,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.924874094820625e-06,
+      "loss": 1.41439342,
+      "memory(GiB)": 84.03,
+      "step": 8080,
+      "train_speed(iter/s)": 1.308083
+    },
+    {
+      "acc": 0.67578197,
+      "epoch": 0.20509893455098935,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.924692891280139e-06,
+      "loss": 1.45497446,
+      "memory(GiB)": 84.03,
+      "step": 8085,
+      "train_speed(iter/s)": 1.30818
+    },
+    {
+      "acc": 0.68612537,
+      "epoch": 0.2052257737189244,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.924511471129673e-06,
+      "loss": 1.38021412,
+      "memory(GiB)": 84.03,
+      "step": 8090,
+      "train_speed(iter/s)": 1.308275
+    },
+    {
+      "acc": 0.67523303,
+      "epoch": 0.20535261288685946,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.924329834377206e-06,
+      "loss": 1.41934319,
+      "memory(GiB)": 84.03,
+      "step": 8095,
+      "train_speed(iter/s)": 1.308371
+    },
+    {
+      "acc": 0.67806692,
+      "epoch": 0.2054794520547945,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.924147981030728e-06,
+      "loss": 1.42056322,
+      "memory(GiB)": 84.03,
+      "step": 8100,
+      "train_speed(iter/s)": 1.308471
+    },
+    {
+      "acc": 0.69006243,
+      "epoch": 0.20560629122272958,
+      "grad_norm": 2.875,
+      "learning_rate": 9.923965911098235e-06,
+      "loss": 1.3882163,
+      "memory(GiB)": 84.03,
+      "step": 8105,
+      "train_speed(iter/s)": 1.308563
+    },
+    {
+      "acc": 0.65992627,
+      "epoch": 0.20573313039066463,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.92378362458774e-06,
+      "loss": 1.5303586,
+      "memory(GiB)": 84.03,
+      "step": 8110,
+      "train_speed(iter/s)": 1.308661
+    },
+    {
+      "acc": 0.6805141,
+      "epoch": 0.2058599695585997,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.923601121507256e-06,
+      "loss": 1.43631296,
+      "memory(GiB)": 84.03,
+      "step": 8115,
+      "train_speed(iter/s)": 1.308764
+    },
+    {
+      "acc": 0.66393437,
+      "epoch": 0.20598680872653474,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.923418401864812e-06,
+      "loss": 1.46221209,
+      "memory(GiB)": 84.03,
+      "step": 8120,
+      "train_speed(iter/s)": 1.308841
+    },
+    {
+      "acc": 0.67954936,
+      "epoch": 0.20611364789446981,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.923235465668447e-06,
+      "loss": 1.39333305,
+      "memory(GiB)": 84.03,
+      "step": 8125,
+      "train_speed(iter/s)": 1.308944
+    },
+    {
+      "acc": 0.68614674,
+      "epoch": 0.20624048706240486,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.923052312926204e-06,
+      "loss": 1.41946783,
+      "memory(GiB)": 84.03,
+      "step": 8130,
+      "train_speed(iter/s)": 1.309044
+    },
+    {
+      "acc": 0.67585592,
+      "epoch": 0.20636732623033993,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.922868943646142e-06,
+      "loss": 1.40654697,
+      "memory(GiB)": 84.03,
+      "step": 8135,
+      "train_speed(iter/s)": 1.309144
+    },
+    {
+      "acc": 0.67382069,
+      "epoch": 0.20649416539827498,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.922685357836324e-06,
+      "loss": 1.42129393,
+      "memory(GiB)": 84.03,
+      "step": 8140,
+      "train_speed(iter/s)": 1.309239
+    },
+    {
+      "acc": 0.67692757,
+      "epoch": 0.20662100456621005,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.922501555504827e-06,
+      "loss": 1.45043364,
+      "memory(GiB)": 84.03,
+      "step": 8145,
+      "train_speed(iter/s)": 1.309337
+    },
+    {
+      "acc": 0.67441196,
+      "epoch": 0.2067478437341451,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.922317536659733e-06,
+      "loss": 1.40523777,
+      "memory(GiB)": 84.03,
+      "step": 8150,
+      "train_speed(iter/s)": 1.309434
+    },
+    {
+      "acc": 0.68538322,
+      "epoch": 0.20687468290208016,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.922133301309136e-06,
+      "loss": 1.42200851,
+      "memory(GiB)": 84.03,
+      "step": 8155,
+      "train_speed(iter/s)": 1.309533
+    },
+    {
+      "acc": 0.66709776,
+      "epoch": 0.2070015220700152,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.921948849461142e-06,
+      "loss": 1.45843668,
+      "memory(GiB)": 84.03,
+      "step": 8160,
+      "train_speed(iter/s)": 1.309629
+    },
+    {
+      "acc": 0.69068999,
+      "epoch": 0.20712836123795028,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.921764181123864e-06,
+      "loss": 1.38236685,
+      "memory(GiB)": 84.03,
+      "step": 8165,
+      "train_speed(iter/s)": 1.309729
+    },
+    {
+      "acc": 0.69216032,
+      "epoch": 0.20725520040588533,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.921579296305421e-06,
+      "loss": 1.41059704,
+      "memory(GiB)": 84.03,
+      "step": 8170,
+      "train_speed(iter/s)": 1.309814
+    },
+    {
+      "acc": 0.67081766,
+      "epoch": 0.2073820395738204,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.921394195013949e-06,
+      "loss": 1.4119772,
+      "memory(GiB)": 84.03,
+      "step": 8175,
+      "train_speed(iter/s)": 1.309913
+    },
+    {
+      "acc": 0.6841558,
+      "epoch": 0.20750887874175544,
+      "grad_norm": 2.875,
+      "learning_rate": 9.921208877257586e-06,
+      "loss": 1.47068624,
+      "memory(GiB)": 84.03,
+      "step": 8180,
+      "train_speed(iter/s)": 1.310009
+    },
+    {
+      "acc": 0.67593002,
+      "epoch": 0.20763571790969051,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.921023343044486e-06,
+      "loss": 1.44050417,
+      "memory(GiB)": 84.03,
+      "step": 8185,
+      "train_speed(iter/s)": 1.310093
+    },
+    {
+      "acc": 0.68991909,
+      "epoch": 0.20776255707762556,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.92083759238281e-06,
+      "loss": 1.36594152,
+      "memory(GiB)": 84.03,
+      "step": 8190,
+      "train_speed(iter/s)": 1.310192
+    },
+    {
+      "acc": 0.68835778,
+      "epoch": 0.20788939624556063,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.920651625280725e-06,
+      "loss": 1.47512665,
+      "memory(GiB)": 84.03,
+      "step": 8195,
+      "train_speed(iter/s)": 1.31029
+    },
+    {
+      "acc": 0.67550449,
+      "epoch": 0.20801623541349568,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.920465441746412e-06,
+      "loss": 1.45360546,
+      "memory(GiB)": 84.03,
+      "step": 8200,
+      "train_speed(iter/s)": 1.310392
+    },
+    {
+      "acc": 0.67797861,
+      "epoch": 0.20814307458143075,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.920279041788062e-06,
+      "loss": 1.46601267,
+      "memory(GiB)": 84.03,
+      "step": 8205,
+      "train_speed(iter/s)": 1.310492
+    },
+    {
+      "acc": 0.68519058,
+      "epoch": 0.2082699137493658,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.920092425413871e-06,
+      "loss": 1.43697548,
+      "memory(GiB)": 84.03,
+      "step": 8210,
+      "train_speed(iter/s)": 1.310589
+    },
+    {
+      "acc": 0.68783326,
+      "epoch": 0.20839675291730086,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.919905592632048e-06,
+      "loss": 1.35149193,
+      "memory(GiB)": 84.03,
+      "step": 8215,
+      "train_speed(iter/s)": 1.310685
+    },
+    {
+      "acc": 0.66899996,
+      "epoch": 0.2085235920852359,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.919718543450813e-06,
+      "loss": 1.41606646,
+      "memory(GiB)": 84.03,
+      "step": 8220,
+      "train_speed(iter/s)": 1.310785
+    },
+    {
+      "acc": 0.67716517,
+      "epoch": 0.20865043125317098,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.919531277878391e-06,
+      "loss": 1.45671673,
+      "memory(GiB)": 84.03,
+      "step": 8225,
+      "train_speed(iter/s)": 1.31088
+    },
+    {
+      "acc": 0.67673931,
+      "epoch": 0.20877727042110603,
+      "grad_norm": 4.78125,
+      "learning_rate": 9.91934379592302e-06,
+      "loss": 1.35650225,
+      "memory(GiB)": 84.03,
+      "step": 8230,
+      "train_speed(iter/s)": 1.310973
+    },
+    {
+      "acc": 0.69266634,
+      "epoch": 0.2089041095890411,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.919156097592944e-06,
+      "loss": 1.33897533,
+      "memory(GiB)": 84.03,
+      "step": 8235,
+      "train_speed(iter/s)": 1.31107
+    },
+    {
+      "acc": 0.68816013,
+      "epoch": 0.20903094875697614,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.91896818289642e-06,
+      "loss": 1.37552547,
+      "memory(GiB)": 84.03,
+      "step": 8240,
+      "train_speed(iter/s)": 1.311169
+    },
+    {
+      "acc": 0.68619137,
+      "epoch": 0.20915778792491121,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.918780051841716e-06,
+      "loss": 1.42564545,
+      "memory(GiB)": 84.03,
+      "step": 8245,
+      "train_speed(iter/s)": 1.311268
+    },
+    {
+      "acc": 0.67573323,
+      "epoch": 0.20928462709284626,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.918591704437103e-06,
+      "loss": 1.42791843,
+      "memory(GiB)": 84.03,
+      "step": 8250,
+      "train_speed(iter/s)": 1.31136
+    },
+    {
+      "acc": 0.67933979,
+      "epoch": 0.20941146626078133,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.918403140690866e-06,
+      "loss": 1.42071314,
+      "memory(GiB)": 84.03,
+      "step": 8255,
+      "train_speed(iter/s)": 1.311459
+    },
+    {
+      "acc": 0.67666588,
+      "epoch": 0.20953830542871638,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.918214360611302e-06,
+      "loss": 1.44158745,
+      "memory(GiB)": 84.03,
+      "step": 8260,
+      "train_speed(iter/s)": 1.311556
+    },
+    {
+      "acc": 0.68067484,
+      "epoch": 0.20966514459665145,
+      "grad_norm": 3.75,
+      "learning_rate": 9.918025364206712e-06,
+      "loss": 1.42340298,
+      "memory(GiB)": 84.03,
+      "step": 8265,
+      "train_speed(iter/s)": 1.311649
+    },
+    {
+      "acc": 0.67943306,
+      "epoch": 0.2097919837645865,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.917836151485407e-06,
+      "loss": 1.41947203,
+      "memory(GiB)": 84.03,
+      "step": 8270,
+      "train_speed(iter/s)": 1.311742
+    },
+    {
+      "acc": 0.68797197,
+      "epoch": 0.20991882293252156,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.917646722455713e-06,
+      "loss": 1.44841022,
+      "memory(GiB)": 84.03,
+      "step": 8275,
+      "train_speed(iter/s)": 1.311841
+    },
+    {
+      "acc": 0.67583652,
+      "epoch": 0.2100456621004566,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.91745707712596e-06,
+      "loss": 1.40882158,
+      "memory(GiB)": 84.03,
+      "step": 8280,
+      "train_speed(iter/s)": 1.311935
+    },
+    {
+      "acc": 0.68146586,
+      "epoch": 0.21017250126839168,
+      "grad_norm": 3.375,
+      "learning_rate": 9.91726721550449e-06,
+      "loss": 1.43994646,
+      "memory(GiB)": 84.03,
+      "step": 8285,
+      "train_speed(iter/s)": 1.312034
+    },
+    {
+      "acc": 0.67499781,
+      "epoch": 0.21029934043632673,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.917077137599653e-06,
+      "loss": 1.37999783,
+      "memory(GiB)": 84.03,
+      "step": 8290,
+      "train_speed(iter/s)": 1.312128
+    },
+    {
+      "acc": 0.66738396,
+      "epoch": 0.2104261796042618,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.916886843419811e-06,
+      "loss": 1.48493242,
+      "memory(GiB)": 84.03,
+      "step": 8295,
+      "train_speed(iter/s)": 1.312224
+    },
+    {
+      "acc": 0.69147782,
+      "epoch": 0.21055301877219684,
+      "grad_norm": 3.25,
+      "learning_rate": 9.916696332973334e-06,
+      "loss": 1.43515244,
+      "memory(GiB)": 84.03,
+      "step": 8300,
+      "train_speed(iter/s)": 1.312313
+    },
+    {
+      "acc": 0.68236837,
+      "epoch": 0.21067985794013191,
+      "grad_norm": 3.125,
+      "learning_rate": 9.9165056062686e-06,
+      "loss": 1.41096516,
+      "memory(GiB)": 84.03,
+      "step": 8305,
+      "train_speed(iter/s)": 1.312409
+    },
+    {
+      "acc": 0.67217216,
+      "epoch": 0.21080669710806696,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.916314663314e-06,
+      "loss": 1.45464211,
+      "memory(GiB)": 84.03,
+      "step": 8310,
+      "train_speed(iter/s)": 1.312505
+    },
+    {
+      "acc": 0.68234081,
+      "epoch": 0.21093353627600203,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.91612350411793e-06,
+      "loss": 1.43128338,
+      "memory(GiB)": 84.03,
+      "step": 8315,
+      "train_speed(iter/s)": 1.312601
+    },
+    {
+      "acc": 0.6969943,
+      "epoch": 0.21106037544393708,
+      "grad_norm": 3.375,
+      "learning_rate": 9.9159321286888e-06,
+      "loss": 1.38041449,
+      "memory(GiB)": 84.03,
+      "step": 8320,
+      "train_speed(iter/s)": 1.312695
+    },
+    {
+      "acc": 0.67716184,
+      "epoch": 0.21118721461187215,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.915740537035026e-06,
+      "loss": 1.46225014,
+      "memory(GiB)": 84.03,
+      "step": 8325,
+      "train_speed(iter/s)": 1.312791
+    },
+    {
+      "acc": 0.67961955,
+      "epoch": 0.2113140537798072,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.915548729165036e-06,
+      "loss": 1.46477842,
+      "memory(GiB)": 84.03,
+      "step": 8330,
+      "train_speed(iter/s)": 1.312891
+    },
+    {
+      "acc": 0.70381374,
+      "epoch": 0.21144089294774226,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.915356705087269e-06,
+      "loss": 1.36538229,
+      "memory(GiB)": 84.03,
+      "step": 8335,
+      "train_speed(iter/s)": 1.312985
+    },
+    {
+      "acc": 0.67501602,
+      "epoch": 0.2115677321156773,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.915164464810166e-06,
+      "loss": 1.4686779,
+      "memory(GiB)": 84.03,
+      "step": 8340,
+      "train_speed(iter/s)": 1.313079
+    },
+    {
+      "acc": 0.69054585,
+      "epoch": 0.21169457128361238,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.914972008342186e-06,
+      "loss": 1.40777817,
+      "memory(GiB)": 84.03,
+      "step": 8345,
+      "train_speed(iter/s)": 1.313171
+    },
+    {
+      "acc": 0.66407981,
+      "epoch": 0.21182141045154743,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.914779335691793e-06,
+      "loss": 1.49298878,
+      "memory(GiB)": 84.03,
+      "step": 8350,
+      "train_speed(iter/s)": 1.313266
+    },
+    {
+      "acc": 0.67076483,
+      "epoch": 0.2119482496194825,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.914586446867463e-06,
+      "loss": 1.48290958,
+      "memory(GiB)": 84.03,
+      "step": 8355,
+      "train_speed(iter/s)": 1.313362
+    },
+    {
+      "acc": 0.68522444,
+      "epoch": 0.21207508878741754,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.914393341877678e-06,
+      "loss": 1.44680519,
+      "memory(GiB)": 84.03,
+      "step": 8360,
+      "train_speed(iter/s)": 1.31346
+    },
+    {
+      "acc": 0.6780961,
+      "epoch": 0.21220192795535261,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.914200020730932e-06,
+      "loss": 1.45645924,
+      "memory(GiB)": 84.03,
+      "step": 8365,
+      "train_speed(iter/s)": 1.313557
+    },
+    {
+      "acc": 0.68333206,
+      "epoch": 0.21232876712328766,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.914006483435732e-06,
+      "loss": 1.40687027,
+      "memory(GiB)": 84.03,
+      "step": 8370,
+      "train_speed(iter/s)": 1.313656
+    },
+    {
+      "acc": 0.66640701,
+      "epoch": 0.21245560629122273,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.913812730000585e-06,
+      "loss": 1.47037239,
+      "memory(GiB)": 84.03,
+      "step": 8375,
+      "train_speed(iter/s)": 1.313742
+    },
+    {
+      "acc": 0.69216528,
+      "epoch": 0.21258244545915778,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.913618760434015e-06,
+      "loss": 1.43702946,
+      "memory(GiB)": 84.03,
+      "step": 8380,
+      "train_speed(iter/s)": 1.313842
+    },
+    {
+      "acc": 0.68422346,
+      "epoch": 0.21270928462709285,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.913424574744555e-06,
+      "loss": 1.36071262,
+      "memory(GiB)": 84.03,
+      "step": 8385,
+      "train_speed(iter/s)": 1.313939
+    },
+    {
+      "acc": 0.6912859,
+      "epoch": 0.2128361237950279,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.913230172940744e-06,
+      "loss": 1.41428566,
+      "memory(GiB)": 84.03,
+      "step": 8390,
+      "train_speed(iter/s)": 1.314031
+    },
+    {
+      "acc": 0.66158571,
+      "epoch": 0.21296296296296297,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.913035555031136e-06,
+      "loss": 1.47546616,
+      "memory(GiB)": 84.03,
+      "step": 8395,
+      "train_speed(iter/s)": 1.31413
+    },
+    {
+      "acc": 0.68236475,
+      "epoch": 0.213089802130898,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.912840721024288e-06,
+      "loss": 1.38791523,
+      "memory(GiB)": 84.03,
+      "step": 8400,
+      "train_speed(iter/s)": 1.314226
+    },
+    {
+      "acc": 0.70418639,
+      "epoch": 0.21321664129883308,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.91264567092877e-06,
+      "loss": 1.31762142,
+      "memory(GiB)": 84.03,
+      "step": 8405,
+      "train_speed(iter/s)": 1.314325
+    },
+    {
+      "acc": 0.67564454,
+      "epoch": 0.21334348046676813,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.912450404753164e-06,
+      "loss": 1.50291653,
+      "memory(GiB)": 84.03,
+      "step": 8410,
+      "train_speed(iter/s)": 1.314427
+    },
+    {
+      "acc": 0.69132857,
+      "epoch": 0.2134703196347032,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.912254922506057e-06,
+      "loss": 1.42909889,
+      "memory(GiB)": 84.03,
+      "step": 8415,
+      "train_speed(iter/s)": 1.314525
+    },
+    {
+      "acc": 0.68211889,
+      "epoch": 0.21359715880263824,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.912059224196044e-06,
+      "loss": 1.44782219,
+      "memory(GiB)": 84.03,
+      "step": 8420,
+      "train_speed(iter/s)": 1.314624
+    },
+    {
+      "acc": 0.68064756,
+      "epoch": 0.21372399797057332,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.911863309831738e-06,
+      "loss": 1.38500652,
+      "memory(GiB)": 84.03,
+      "step": 8425,
+      "train_speed(iter/s)": 1.314718
+    },
+    {
+      "acc": 0.66556873,
+      "epoch": 0.21385083713850836,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.911667179421753e-06,
+      "loss": 1.53193493,
+      "memory(GiB)": 84.03,
+      "step": 8430,
+      "train_speed(iter/s)": 1.314815
+    },
+    {
+      "acc": 0.68741741,
+      "epoch": 0.21397767630644343,
+      "grad_norm": 2.625,
+      "learning_rate": 9.911470832974717e-06,
+      "loss": 1.37571259,
+      "memory(GiB)": 84.03,
+      "step": 8435,
+      "train_speed(iter/s)": 1.314907
+    },
+    {
+      "acc": 0.67405648,
+      "epoch": 0.21410451547437848,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.911274270499265e-06,
+      "loss": 1.39071074,
+      "memory(GiB)": 84.03,
+      "step": 8440,
+      "train_speed(iter/s)": 1.315002
+    },
+    {
+      "acc": 0.69064293,
+      "epoch": 0.21423135464231355,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.911077492004044e-06,
+      "loss": 1.4144557,
+      "memory(GiB)": 84.03,
+      "step": 8445,
+      "train_speed(iter/s)": 1.315097
+    },
+    {
+      "acc": 0.67696414,
+      "epoch": 0.2143581938102486,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.910880497497707e-06,
+      "loss": 1.40022688,
+      "memory(GiB)": 84.03,
+      "step": 8450,
+      "train_speed(iter/s)": 1.315193
+    },
+    {
+      "acc": 0.68263502,
+      "epoch": 0.21448503297818367,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.910683286988922e-06,
+      "loss": 1.39699297,
+      "memory(GiB)": 84.03,
+      "step": 8455,
+      "train_speed(iter/s)": 1.315285
+    },
+    {
+      "acc": 0.67229004,
+      "epoch": 0.2146118721461187,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.910485860486361e-06,
+      "loss": 1.45075626,
+      "memory(GiB)": 84.03,
+      "step": 8460,
+      "train_speed(iter/s)": 1.315378
+    },
+    {
+      "acc": 0.67726669,
+      "epoch": 0.21473871131405378,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.910288217998707e-06,
+      "loss": 1.4754261,
+      "memory(GiB)": 84.03,
+      "step": 8465,
+      "train_speed(iter/s)": 1.315471
+    },
+    {
+      "acc": 0.67623482,
+      "epoch": 0.21486555048198883,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.910090359534654e-06,
+      "loss": 1.44915962,
+      "memory(GiB)": 84.03,
+      "step": 8470,
+      "train_speed(iter/s)": 1.31556
+    },
+    {
+      "acc": 0.68616548,
+      "epoch": 0.2149923896499239,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.909892285102907e-06,
+      "loss": 1.43427238,
+      "memory(GiB)": 84.03,
+      "step": 8475,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.68497357,
+      "epoch": 0.21511922881785894,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.909693994712174e-06,
+      "loss": 1.4095046,
+      "memory(GiB)": 84.03,
+      "step": 8480,
+      "train_speed(iter/s)": 1.31574
+    },
+    {
+      "acc": 0.69230442,
+      "epoch": 0.21524606798579402,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.909495488371181e-06,
+      "loss": 1.34298325,
+      "memory(GiB)": 84.03,
+      "step": 8485,
+      "train_speed(iter/s)": 1.315826
+    },
+    {
+      "acc": 0.68056459,
+      "epoch": 0.21537290715372906,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.909296766088657e-06,
+      "loss": 1.39448032,
+      "memory(GiB)": 84.03,
+      "step": 8490,
+      "train_speed(iter/s)": 1.315919
+    },
+    {
+      "acc": 0.68902712,
+      "epoch": 0.21549974632166413,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.909097827873341e-06,
+      "loss": 1.39837608,
+      "memory(GiB)": 84.03,
+      "step": 8495,
+      "train_speed(iter/s)": 1.31601
+    },
+    {
+      "acc": 0.68817744,
+      "epoch": 0.21562658548959918,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.908898673733986e-06,
+      "loss": 1.44989948,
+      "memory(GiB)": 84.03,
+      "step": 8500,
+      "train_speed(iter/s)": 1.316101
+    },
+    {
+      "acc": 0.68973808,
+      "epoch": 0.21575342465753425,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.90869930367935e-06,
+      "loss": 1.4046917,
+      "memory(GiB)": 84.03,
+      "step": 8505,
+      "train_speed(iter/s)": 1.31619
+    },
+    {
+      "acc": 0.66532106,
+      "epoch": 0.2158802638254693,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.908499717718203e-06,
+      "loss": 1.49695854,
+      "memory(GiB)": 84.03,
+      "step": 8510,
+      "train_speed(iter/s)": 1.316283
+    },
+    {
+      "acc": 0.68529425,
+      "epoch": 0.21600710299340437,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.908299915859325e-06,
+      "loss": 1.44650364,
+      "memory(GiB)": 84.03,
+      "step": 8515,
+      "train_speed(iter/s)": 1.316372
+    },
+    {
+      "acc": 0.68925481,
+      "epoch": 0.2161339421613394,
+      "grad_norm": 3.0,
+      "learning_rate": 9.908099898111502e-06,
+      "loss": 1.39845247,
+      "memory(GiB)": 84.03,
+      "step": 8520,
+      "train_speed(iter/s)": 1.316462
+    },
+    {
+      "acc": 0.67790456,
+      "epoch": 0.21626078132927448,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.907899664483533e-06,
+      "loss": 1.40204325,
+      "memory(GiB)": 84.03,
+      "step": 8525,
+      "train_speed(iter/s)": 1.316547
+    },
+    {
+      "acc": 0.66339641,
+      "epoch": 0.21638762049720953,
+      "grad_norm": 2.75,
+      "learning_rate": 9.907699214984223e-06,
+      "loss": 1.45523682,
+      "memory(GiB)": 84.03,
+      "step": 8530,
+      "train_speed(iter/s)": 1.31664
+    },
+    {
+      "acc": 0.66954446,
+      "epoch": 0.2165144596651446,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.90749854962239e-06,
+      "loss": 1.44180641,
+      "memory(GiB)": 84.03,
+      "step": 8535,
+      "train_speed(iter/s)": 1.316729
+    },
+    {
+      "acc": 0.68241644,
+      "epoch": 0.21664129883307964,
+      "grad_norm": 3.625,
+      "learning_rate": 9.907297668406863e-06,
+      "loss": 1.38817768,
+      "memory(GiB)": 84.03,
+      "step": 8540,
+      "train_speed(iter/s)": 1.316821
+    },
+    {
+      "acc": 0.67608285,
+      "epoch": 0.21676813800101472,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.907096571346474e-06,
+      "loss": 1.44112911,
+      "memory(GiB)": 84.03,
+      "step": 8545,
+      "train_speed(iter/s)": 1.316912
+    },
+    {
+      "acc": 0.68149195,
+      "epoch": 0.21689497716894976,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.906895258450067e-06,
+      "loss": 1.38389587,
+      "memory(GiB)": 84.03,
+      "step": 8550,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.68785386,
+      "epoch": 0.21702181633688483,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.9066937297265e-06,
+      "loss": 1.38715105,
+      "memory(GiB)": 84.03,
+      "step": 8555,
+      "train_speed(iter/s)": 1.317089
+    },
+    {
+      "acc": 0.67188835,
+      "epoch": 0.21714865550481988,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.906491985184637e-06,
+      "loss": 1.38184528,
+      "memory(GiB)": 84.03,
+      "step": 8560,
+      "train_speed(iter/s)": 1.317179
+    },
+    {
+      "acc": 0.66390362,
+      "epoch": 0.21727549467275495,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.906290024833349e-06,
+      "loss": 1.4729722,
+      "memory(GiB)": 84.03,
+      "step": 8565,
+      "train_speed(iter/s)": 1.317262
+    },
+    {
+      "acc": 0.67786255,
+      "epoch": 0.21740233384069,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.906087848681523e-06,
+      "loss": 1.4338397,
+      "memory(GiB)": 84.03,
+      "step": 8570,
+      "train_speed(iter/s)": 1.317352
+    },
+    {
+      "acc": 0.68311296,
+      "epoch": 0.21752917300862507,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.905885456738046e-06,
+      "loss": 1.43986492,
+      "memory(GiB)": 84.03,
+      "step": 8575,
+      "train_speed(iter/s)": 1.31744
+    },
+    {
+      "acc": 0.68679357,
+      "epoch": 0.2176560121765601,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.905682849011826e-06,
+      "loss": 1.43497267,
+      "memory(GiB)": 84.03,
+      "step": 8580,
+      "train_speed(iter/s)": 1.317532
+    },
+    {
+      "acc": 0.67329416,
+      "epoch": 0.21778285134449518,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.905480025511772e-06,
+      "loss": 1.46913557,
+      "memory(GiB)": 84.03,
+      "step": 8585,
+      "train_speed(iter/s)": 1.317622
+    },
+    {
+      "acc": 0.67301025,
+      "epoch": 0.21790969051243023,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.905276986246804e-06,
+      "loss": 1.44978495,
+      "memory(GiB)": 84.03,
+      "step": 8590,
+      "train_speed(iter/s)": 1.317714
+    },
+    {
+      "acc": 0.67723265,
+      "epoch": 0.2180365296803653,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.905073731225854e-06,
+      "loss": 1.36265745,
+      "memory(GiB)": 84.03,
+      "step": 8595,
+      "train_speed(iter/s)": 1.317805
+    },
+    {
+      "acc": 0.66148529,
+      "epoch": 0.21816336884830034,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.904870260457861e-06,
+      "loss": 1.439118,
+      "memory(GiB)": 84.03,
+      "step": 8600,
+      "train_speed(iter/s)": 1.317894
+    },
+    {
+      "acc": 0.6683105,
+      "epoch": 0.21829020801623542,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.904666573951777e-06,
+      "loss": 1.44968214,
+      "memory(GiB)": 84.03,
+      "step": 8605,
+      "train_speed(iter/s)": 1.317983
+    },
+    {
+      "acc": 0.67425146,
+      "epoch": 0.21841704718417046,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.904462671716559e-06,
+      "loss": 1.48530483,
+      "memory(GiB)": 84.03,
+      "step": 8610,
+      "train_speed(iter/s)": 1.318062
+    },
+    {
+      "acc": 0.6669229,
+      "epoch": 0.21854388635210553,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.904258553761175e-06,
+      "loss": 1.50680161,
+      "memory(GiB)": 84.03,
+      "step": 8615,
+      "train_speed(iter/s)": 1.318154
+    },
+    {
+      "acc": 0.67928562,
+      "epoch": 0.21867072552004058,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.904054220094603e-06,
+      "loss": 1.46858616,
+      "memory(GiB)": 84.03,
+      "step": 8620,
+      "train_speed(iter/s)": 1.318243
+    },
+    {
+      "acc": 0.67172546,
+      "epoch": 0.21879756468797565,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.903849670725833e-06,
+      "loss": 1.45664463,
+      "memory(GiB)": 84.03,
+      "step": 8625,
+      "train_speed(iter/s)": 1.318329
+    },
+    {
+      "acc": 0.67768226,
+      "epoch": 0.2189244038559107,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.903644905663861e-06,
+      "loss": 1.41793842,
+      "memory(GiB)": 84.03,
+      "step": 8630,
+      "train_speed(iter/s)": 1.318421
+    },
+    {
+      "acc": 0.67503667,
+      "epoch": 0.21905124302384577,
+      "grad_norm": 4.125,
+      "learning_rate": 9.90343992491769e-06,
+      "loss": 1.42101068,
+      "memory(GiB)": 84.03,
+      "step": 8635,
+      "train_speed(iter/s)": 1.31849
+    },
+    {
+      "acc": 0.68329511,
+      "epoch": 0.2191780821917808,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.903234728496341e-06,
+      "loss": 1.41762257,
+      "memory(GiB)": 84.03,
+      "step": 8640,
+      "train_speed(iter/s)": 1.318572
+    },
+    {
+      "acc": 0.66814876,
+      "epoch": 0.21930492135971588,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.903029316408838e-06,
+      "loss": 1.5038476,
+      "memory(GiB)": 84.03,
+      "step": 8645,
+      "train_speed(iter/s)": 1.318664
+    },
+    {
+      "acc": 0.68791857,
+      "epoch": 0.21943176052765093,
+      "grad_norm": 2.875,
+      "learning_rate": 9.902823688664214e-06,
+      "loss": 1.38463554,
+      "memory(GiB)": 84.03,
+      "step": 8650,
+      "train_speed(iter/s)": 1.318755
+    },
+    {
+      "acc": 0.68180146,
+      "epoch": 0.219558599695586,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.902617845271514e-06,
+      "loss": 1.39640236,
+      "memory(GiB)": 84.03,
+      "step": 8655,
+      "train_speed(iter/s)": 1.318846
+    },
+    {
+      "acc": 0.69795041,
+      "epoch": 0.21968543886352104,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.902411786239794e-06,
+      "loss": 1.36029997,
+      "memory(GiB)": 84.03,
+      "step": 8660,
+      "train_speed(iter/s)": 1.318937
+    },
+    {
+      "acc": 0.67986498,
+      "epoch": 0.21981227803145612,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.902205511578114e-06,
+      "loss": 1.42857609,
+      "memory(GiB)": 84.03,
+      "step": 8665,
+      "train_speed(iter/s)": 1.319027
+    },
+    {
+      "acc": 0.67464981,
+      "epoch": 0.21993911719939116,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.90199902129555e-06,
+      "loss": 1.46685448,
+      "memory(GiB)": 84.03,
+      "step": 8670,
+      "train_speed(iter/s)": 1.319118
+    },
+    {
+      "acc": 0.66588278,
+      "epoch": 0.22006595636732623,
+      "grad_norm": 3.25,
+      "learning_rate": 9.901792315401184e-06,
+      "loss": 1.48431463,
+      "memory(GiB)": 84.03,
+      "step": 8675,
+      "train_speed(iter/s)": 1.319208
+    },
+    {
+      "acc": 0.68496885,
+      "epoch": 0.22019279553526128,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.901585393904104e-06,
+      "loss": 1.34194126,
+      "memory(GiB)": 84.03,
+      "step": 8680,
+      "train_speed(iter/s)": 1.319296
+    },
+    {
+      "acc": 0.67065711,
+      "epoch": 0.22031963470319635,
+      "grad_norm": 2.875,
+      "learning_rate": 9.901378256813418e-06,
+      "loss": 1.46803188,
+      "memory(GiB)": 84.03,
+      "step": 8685,
+      "train_speed(iter/s)": 1.319387
+    },
+    {
+      "acc": 0.68841085,
+      "epoch": 0.2204464738711314,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.901170904138232e-06,
+      "loss": 1.381707,
+      "memory(GiB)": 84.03,
+      "step": 8690,
+      "train_speed(iter/s)": 1.319478
+    },
+    {
+      "acc": 0.66446342,
+      "epoch": 0.22057331303906647,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.900963335887667e-06,
+      "loss": 1.50616579,
+      "memory(GiB)": 84.03,
+      "step": 8695,
+      "train_speed(iter/s)": 1.319567
+    },
+    {
+      "acc": 0.68452721,
+      "epoch": 0.2207001522070015,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.900755552070852e-06,
+      "loss": 1.44200897,
+      "memory(GiB)": 84.03,
+      "step": 8700,
+      "train_speed(iter/s)": 1.319655
+    },
+    {
+      "acc": 0.69591198,
+      "epoch": 0.22082699137493658,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.900547552696931e-06,
+      "loss": 1.37768583,
+      "memory(GiB)": 84.03,
+      "step": 8705,
+      "train_speed(iter/s)": 1.319743
+    },
+    {
+      "acc": 0.68510504,
+      "epoch": 0.22095383054287163,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.900339337775046e-06,
+      "loss": 1.4653574,
+      "memory(GiB)": 84.03,
+      "step": 8710,
+      "train_speed(iter/s)": 1.319834
+    },
+    {
+      "acc": 0.68105645,
+      "epoch": 0.2210806697108067,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.90013090731436e-06,
+      "loss": 1.46134796,
+      "memory(GiB)": 84.03,
+      "step": 8715,
+      "train_speed(iter/s)": 1.319925
+    },
+    {
+      "acc": 0.6712338,
+      "epoch": 0.22120750887874174,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.89992226132404e-06,
+      "loss": 1.49151955,
+      "memory(GiB)": 84.03,
+      "step": 8720,
+      "train_speed(iter/s)": 1.320015
+    },
+    {
+      "acc": 0.68230128,
+      "epoch": 0.22133434804667682,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.899713399813261e-06,
+      "loss": 1.40642033,
+      "memory(GiB)": 84.03,
+      "step": 8725,
+      "train_speed(iter/s)": 1.320105
+    },
+    {
+      "acc": 0.68452754,
+      "epoch": 0.22146118721461186,
+      "grad_norm": 3.25,
+      "learning_rate": 9.899504322791212e-06,
+      "loss": 1.42414856,
+      "memory(GiB)": 84.03,
+      "step": 8730,
+      "train_speed(iter/s)": 1.320195
+    },
+    {
+      "acc": 0.67539282,
+      "epoch": 0.22158802638254693,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.899295030267086e-06,
+      "loss": 1.48622303,
+      "memory(GiB)": 84.03,
+      "step": 8735,
+      "train_speed(iter/s)": 1.320284
+    },
+    {
+      "acc": 0.67659674,
+      "epoch": 0.22171486555048198,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.899085522250094e-06,
+      "loss": 1.42314196,
+      "memory(GiB)": 84.03,
+      "step": 8740,
+      "train_speed(iter/s)": 1.320372
+    },
+    {
+      "acc": 0.68971472,
+      "epoch": 0.22184170471841705,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.898875798749446e-06,
+      "loss": 1.43578148,
+      "memory(GiB)": 84.03,
+      "step": 8745,
+      "train_speed(iter/s)": 1.320462
+    },
+    {
+      "acc": 0.67636681,
+      "epoch": 0.2219685438863521,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.898665859774367e-06,
+      "loss": 1.45242615,
+      "memory(GiB)": 84.03,
+      "step": 8750,
+      "train_speed(iter/s)": 1.320551
+    },
+    {
+      "acc": 0.6816328,
+      "epoch": 0.22209538305428717,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.898455705334095e-06,
+      "loss": 1.45989227,
+      "memory(GiB)": 84.03,
+      "step": 8755,
+      "train_speed(iter/s)": 1.320642
+    },
+    {
+      "acc": 0.67376366,
+      "epoch": 0.2222222222222222,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.89824533543787e-06,
+      "loss": 1.44470892,
+      "memory(GiB)": 84.03,
+      "step": 8760,
+      "train_speed(iter/s)": 1.320729
+    },
+    {
+      "acc": 0.66320901,
+      "epoch": 0.22234906139015728,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.898034750094946e-06,
+      "loss": 1.48706617,
+      "memory(GiB)": 84.03,
+      "step": 8765,
+      "train_speed(iter/s)": 1.320816
+    },
+    {
+      "acc": 0.66899896,
+      "epoch": 0.22247590055809233,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.897823949314586e-06,
+      "loss": 1.49470863,
+      "memory(GiB)": 84.03,
+      "step": 8770,
+      "train_speed(iter/s)": 1.320905
+    },
+    {
+      "acc": 0.67915878,
+      "epoch": 0.2226027397260274,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.897612933106061e-06,
+      "loss": 1.49629364,
+      "memory(GiB)": 84.03,
+      "step": 8775,
+      "train_speed(iter/s)": 1.320997
+    },
+    {
+      "acc": 0.68282342,
+      "epoch": 0.22272957889396244,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.897401701478654e-06,
+      "loss": 1.40296783,
+      "memory(GiB)": 84.03,
+      "step": 8780,
+      "train_speed(iter/s)": 1.321085
+    },
+    {
+      "acc": 0.68313909,
+      "epoch": 0.22285641806189752,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.897190254441653e-06,
+      "loss": 1.43856869,
+      "memory(GiB)": 84.03,
+      "step": 8785,
+      "train_speed(iter/s)": 1.321168
+    },
+    {
+      "acc": 0.66763806,
+      "epoch": 0.22298325722983256,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.896978592004363e-06,
+      "loss": 1.48587408,
+      "memory(GiB)": 84.03,
+      "step": 8790,
+      "train_speed(iter/s)": 1.321255
+    },
+    {
+      "acc": 0.66962199,
+      "epoch": 0.22311009639776763,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.896766714176089e-06,
+      "loss": 1.4704011,
+      "memory(GiB)": 84.03,
+      "step": 8795,
+      "train_speed(iter/s)": 1.321341
+    },
+    {
+      "acc": 0.67680612,
+      "epoch": 0.22323693556570268,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.896554620966152e-06,
+      "loss": 1.44826593,
+      "memory(GiB)": 84.03,
+      "step": 8800,
+      "train_speed(iter/s)": 1.321431
+    },
+    {
+      "acc": 0.68062844,
+      "epoch": 0.22336377473363775,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.896342312383883e-06,
+      "loss": 1.41705246,
+      "memory(GiB)": 84.03,
+      "step": 8805,
+      "train_speed(iter/s)": 1.321518
+    },
+    {
+      "acc": 0.68600359,
+      "epoch": 0.2234906139015728,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.896129788438617e-06,
+      "loss": 1.4292182,
+      "memory(GiB)": 84.03,
+      "step": 8810,
+      "train_speed(iter/s)": 1.321601
+    },
+    {
+      "acc": 0.68425746,
+      "epoch": 0.22361745306950787,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.895917049139704e-06,
+      "loss": 1.40172024,
+      "memory(GiB)": 84.03,
+      "step": 8815,
+      "train_speed(iter/s)": 1.321689
+    },
+    {
+      "acc": 0.67077942,
+      "epoch": 0.2237442922374429,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.895704094496502e-06,
+      "loss": 1.46337318,
+      "memory(GiB)": 84.03,
+      "step": 8820,
+      "train_speed(iter/s)": 1.321774
+    },
+    {
+      "acc": 0.69682398,
+      "epoch": 0.22387113140537798,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.895490924518372e-06,
+      "loss": 1.36605778,
+      "memory(GiB)": 84.03,
+      "step": 8825,
+      "train_speed(iter/s)": 1.321856
+    },
+    {
+      "acc": 0.68831277,
+      "epoch": 0.22399797057331303,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.895277539214698e-06,
+      "loss": 1.34647522,
+      "memory(GiB)": 84.03,
+      "step": 8830,
+      "train_speed(iter/s)": 1.321937
+    },
+    {
+      "acc": 0.67484426,
+      "epoch": 0.2241248097412481,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.895063938594859e-06,
+      "loss": 1.48877563,
+      "memory(GiB)": 84.03,
+      "step": 8835,
+      "train_speed(iter/s)": 1.322022
+    },
+    {
+      "acc": 0.6675478,
+      "epoch": 0.22425164890918314,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.894850122668256e-06,
+      "loss": 1.45653992,
+      "memory(GiB)": 84.03,
+      "step": 8840,
+      "train_speed(iter/s)": 1.322106
+    },
+    {
+      "acc": 0.68448992,
+      "epoch": 0.22437848807711822,
+      "grad_norm": 2.5,
+      "learning_rate": 9.89463609144429e-06,
+      "loss": 1.38898506,
+      "memory(GiB)": 84.03,
+      "step": 8845,
+      "train_speed(iter/s)": 1.322189
+    },
+    {
+      "acc": 0.68658662,
+      "epoch": 0.22450532724505326,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.894421844932375e-06,
+      "loss": 1.43553429,
+      "memory(GiB)": 84.03,
+      "step": 8850,
+      "train_speed(iter/s)": 1.322269
+    },
+    {
+      "acc": 0.69827971,
+      "epoch": 0.22463216641298833,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.894207383141937e-06,
+      "loss": 1.35634336,
+      "memory(GiB)": 84.03,
+      "step": 8855,
+      "train_speed(iter/s)": 1.322356
+    },
+    {
+      "acc": 0.67727256,
+      "epoch": 0.22475900558092338,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.893992706082405e-06,
+      "loss": 1.39159756,
+      "memory(GiB)": 84.03,
+      "step": 8860,
+      "train_speed(iter/s)": 1.322444
+    },
+    {
+      "acc": 0.67784328,
+      "epoch": 0.22488584474885845,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.893777813763223e-06,
+      "loss": 1.4437088,
+      "memory(GiB)": 84.03,
+      "step": 8865,
+      "train_speed(iter/s)": 1.322529
+    },
+    {
+      "acc": 0.67887707,
+      "epoch": 0.2250126839167935,
+      "grad_norm": 4.125,
+      "learning_rate": 9.893562706193847e-06,
+      "loss": 1.51149006,
+      "memory(GiB)": 84.03,
+      "step": 8870,
+      "train_speed(iter/s)": 1.322618
+    },
+    {
+      "acc": 0.69085188,
+      "epoch": 0.22513952308472857,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.893347383383732e-06,
+      "loss": 1.40996532,
+      "memory(GiB)": 84.03,
+      "step": 8875,
+      "train_speed(iter/s)": 1.322709
+    },
+    {
+      "acc": 0.69383187,
+      "epoch": 0.2252663622526636,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.893131845342352e-06,
+      "loss": 1.43056831,
+      "memory(GiB)": 84.03,
+      "step": 8880,
+      "train_speed(iter/s)": 1.322798
+    },
+    {
+      "acc": 0.68943863,
+      "epoch": 0.22539320142059868,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.892916092079188e-06,
+      "loss": 1.44087696,
+      "memory(GiB)": 84.03,
+      "step": 8885,
+      "train_speed(iter/s)": 1.322884
+    },
+    {
+      "acc": 0.68122873,
+      "epoch": 0.22552004058853373,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.89270012360373e-06,
+      "loss": 1.43814697,
+      "memory(GiB)": 84.03,
+      "step": 8890,
+      "train_speed(iter/s)": 1.322971
+    },
+    {
+      "acc": 0.68947091,
+      "epoch": 0.2256468797564688,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.892483939925476e-06,
+      "loss": 1.408144,
+      "memory(GiB)": 84.03,
+      "step": 8895,
+      "train_speed(iter/s)": 1.323056
+    },
+    {
+      "acc": 0.67011194,
+      "epoch": 0.22577371892440384,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.892267541053933e-06,
+      "loss": 1.43852158,
+      "memory(GiB)": 84.03,
+      "step": 8900,
+      "train_speed(iter/s)": 1.323141
+    },
+    {
+      "acc": 0.68212276,
+      "epoch": 0.22590055809233892,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.892050926998624e-06,
+      "loss": 1.39410191,
+      "memory(GiB)": 84.03,
+      "step": 8905,
+      "train_speed(iter/s)": 1.32322
+    },
+    {
+      "acc": 0.67628069,
+      "epoch": 0.22602739726027396,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.891834097769071e-06,
+      "loss": 1.47494879,
+      "memory(GiB)": 84.03,
+      "step": 8910,
+      "train_speed(iter/s)": 1.323305
+    },
+    {
+      "acc": 0.66344261,
+      "epoch": 0.22615423642820903,
+      "grad_norm": 2.625,
+      "learning_rate": 9.891617053374816e-06,
+      "loss": 1.5094676,
+      "memory(GiB)": 84.03,
+      "step": 8915,
+      "train_speed(iter/s)": 1.323387
+    },
+    {
+      "acc": 0.67373161,
+      "epoch": 0.22628107559614408,
+      "grad_norm": 3.5,
+      "learning_rate": 9.891399793825403e-06,
+      "loss": 1.45829449,
+      "memory(GiB)": 84.03,
+      "step": 8920,
+      "train_speed(iter/s)": 1.323472
+    },
+    {
+      "acc": 0.6836298,
+      "epoch": 0.22640791476407915,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.891182319130387e-06,
+      "loss": 1.5021307,
+      "memory(GiB)": 84.03,
+      "step": 8925,
+      "train_speed(iter/s)": 1.323561
+    },
+    {
+      "acc": 0.68618116,
+      "epoch": 0.2265347539320142,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.890964629299336e-06,
+      "loss": 1.47441196,
+      "memory(GiB)": 84.03,
+      "step": 8930,
+      "train_speed(iter/s)": 1.323644
+    },
+    {
+      "acc": 0.67266092,
+      "epoch": 0.22666159309994927,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.890746724341825e-06,
+      "loss": 1.50048027,
+      "memory(GiB)": 84.03,
+      "step": 8935,
+      "train_speed(iter/s)": 1.32372
+    },
+    {
+      "acc": 0.68296785,
+      "epoch": 0.2267884322678843,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.890528604267436e-06,
+      "loss": 1.35963755,
+      "memory(GiB)": 84.03,
+      "step": 8940,
+      "train_speed(iter/s)": 1.323801
+    },
+    {
+      "acc": 0.68923221,
+      "epoch": 0.22691527143581938,
+      "grad_norm": 3.625,
+      "learning_rate": 9.890310269085765e-06,
+      "loss": 1.40730801,
+      "memory(GiB)": 84.03,
+      "step": 8945,
+      "train_speed(iter/s)": 1.323884
+    },
+    {
+      "acc": 0.65722032,
+      "epoch": 0.22704211060375443,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.890091718806414e-06,
+      "loss": 1.48802834,
+      "memory(GiB)": 84.03,
+      "step": 8950,
+      "train_speed(iter/s)": 1.323967
+    },
+    {
+      "acc": 0.68133812,
+      "epoch": 0.2271689497716895,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.889872953438996e-06,
+      "loss": 1.42898092,
+      "memory(GiB)": 84.03,
+      "step": 8955,
+      "train_speed(iter/s)": 1.324048
+    },
+    {
+      "acc": 0.67995338,
+      "epoch": 0.22729578893962454,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.889653972993136e-06,
+      "loss": 1.45546093,
+      "memory(GiB)": 84.03,
+      "step": 8960,
+      "train_speed(iter/s)": 1.324134
+    },
+    {
+      "acc": 0.67855239,
+      "epoch": 0.22742262810755962,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.889434777478464e-06,
+      "loss": 1.43797169,
+      "memory(GiB)": 84.03,
+      "step": 8965,
+      "train_speed(iter/s)": 1.324218
+    },
+    {
+      "acc": 0.67748775,
+      "epoch": 0.22754946727549466,
+      "grad_norm": 2.875,
+      "learning_rate": 9.88921536690462e-06,
+      "loss": 1.47336349,
+      "memory(GiB)": 84.03,
+      "step": 8970,
+      "train_speed(iter/s)": 1.324303
+    },
+    {
+      "acc": 0.66654854,
+      "epoch": 0.22767630644342973,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.888995741281252e-06,
+      "loss": 1.50892601,
+      "memory(GiB)": 84.03,
+      "step": 8975,
+      "train_speed(iter/s)": 1.324385
+    },
+    {
+      "acc": 0.68562498,
+      "epoch": 0.22780314561136478,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.888775900618028e-06,
+      "loss": 1.37359867,
+      "memory(GiB)": 84.03,
+      "step": 8980,
+      "train_speed(iter/s)": 1.324472
+    },
+    {
+      "acc": 0.67732725,
+      "epoch": 0.22792998477929985,
+      "grad_norm": 3.5,
+      "learning_rate": 9.88855584492461e-06,
+      "loss": 1.46777935,
+      "memory(GiB)": 84.03,
+      "step": 8985,
+      "train_speed(iter/s)": 1.32456
+    },
+    {
+      "acc": 0.67360964,
+      "epoch": 0.2280568239472349,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.888335574210681e-06,
+      "loss": 1.46550159,
+      "memory(GiB)": 84.03,
+      "step": 8990,
+      "train_speed(iter/s)": 1.324643
+    },
+    {
+      "acc": 0.68332992,
+      "epoch": 0.22818366311516997,
+      "grad_norm": 3.25,
+      "learning_rate": 9.888115088485931e-06,
+      "loss": 1.3413435,
+      "memory(GiB)": 84.03,
+      "step": 8995,
+      "train_speed(iter/s)": 1.324727
+    },
+    {
+      "acc": 0.67556086,
+      "epoch": 0.228310502283105,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.887894387760053e-06,
+      "loss": 1.43862495,
+      "memory(GiB)": 84.03,
+      "step": 9000,
+      "train_speed(iter/s)": 1.324811
+    },
+    {
+      "epoch": 0.228310502283105,
+      "eval_acc": 0.6684805693557785,
+      "eval_loss": 1.3934357166290283,
+      "eval_runtime": 69.4532,
+      "eval_samples_per_second": 91.716,
+      "eval_steps_per_second": 22.936,
+      "step": 9000
+    },
+    {
+      "acc": 0.67909002,
+      "epoch": 0.22843734145104008,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.887673472042757e-06,
+      "loss": 1.47005329,
+      "memory(GiB)": 84.03,
+      "step": 9005,
+      "train_speed(iter/s)": 1.309099
+    },
+    {
+      "acc": 0.67954073,
+      "epoch": 0.22856418061897513,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.88745234134376e-06,
+      "loss": 1.44287643,
+      "memory(GiB)": 84.03,
+      "step": 9010,
+      "train_speed(iter/s)": 1.309174
+    },
+    {
+      "acc": 0.6730361,
+      "epoch": 0.2286910197869102,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.887230995672789e-06,
+      "loss": 1.45247173,
+      "memory(GiB)": 84.03,
+      "step": 9015,
+      "train_speed(iter/s)": 1.309257
+    },
+    {
+      "acc": 0.68357801,
+      "epoch": 0.22881785895484524,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.887009435039578e-06,
+      "loss": 1.3739748,
+      "memory(GiB)": 84.03,
+      "step": 9020,
+      "train_speed(iter/s)": 1.309342
+    },
+    {
+      "acc": 0.67232924,
+      "epoch": 0.22894469812278032,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.886787659453873e-06,
+      "loss": 1.40815725,
+      "memory(GiB)": 84.03,
+      "step": 9025,
+      "train_speed(iter/s)": 1.309425
+    },
+    {
+      "acc": 0.67591691,
+      "epoch": 0.22907153729071536,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.886565668925429e-06,
+      "loss": 1.42533016,
+      "memory(GiB)": 84.03,
+      "step": 9030,
+      "train_speed(iter/s)": 1.309508
+    },
+    {
+      "acc": 0.68339405,
+      "epoch": 0.22919837645865043,
+      "grad_norm": 5.59375,
+      "learning_rate": 9.88634346346401e-06,
+      "loss": 1.42587967,
+      "memory(GiB)": 84.03,
+      "step": 9035,
+      "train_speed(iter/s)": 1.309594
+    },
+    {
+      "acc": 0.68229442,
+      "epoch": 0.22932521562658548,
+      "grad_norm": 2.75,
+      "learning_rate": 9.88612104307939e-06,
+      "loss": 1.40520267,
+      "memory(GiB)": 84.03,
+      "step": 9040,
+      "train_speed(iter/s)": 1.309679
+    },
+    {
+      "acc": 0.69107671,
+      "epoch": 0.22945205479452055,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.885898407781352e-06,
+      "loss": 1.40074282,
+      "memory(GiB)": 84.03,
+      "step": 9045,
+      "train_speed(iter/s)": 1.309765
+    },
+    {
+      "acc": 0.68377647,
+      "epoch": 0.2295788939624556,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.885675557579686e-06,
+      "loss": 1.50677099,
+      "memory(GiB)": 84.03,
+      "step": 9050,
+      "train_speed(iter/s)": 1.309855
+    },
+    {
+      "acc": 0.67914181,
+      "epoch": 0.22970573313039067,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.885452492484198e-06,
+      "loss": 1.36587315,
+      "memory(GiB)": 84.03,
+      "step": 9055,
+      "train_speed(iter/s)": 1.309943
+    },
+    {
+      "acc": 0.67501936,
+      "epoch": 0.2298325722983257,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.885229212504697e-06,
+      "loss": 1.43792801,
+      "memory(GiB)": 84.03,
+      "step": 9060,
+      "train_speed(iter/s)": 1.310033
+    },
+    {
+      "acc": 0.67893209,
+      "epoch": 0.22995941146626078,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.885005717651002e-06,
+      "loss": 1.40172138,
+      "memory(GiB)": 84.03,
+      "step": 9065,
+      "train_speed(iter/s)": 1.31012
+    },
+    {
+      "acc": 0.68629003,
+      "epoch": 0.23008625063419583,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.88478200793295e-06,
+      "loss": 1.39182119,
+      "memory(GiB)": 84.03,
+      "step": 9070,
+      "train_speed(iter/s)": 1.310199
+    },
+    {
+      "acc": 0.68306789,
+      "epoch": 0.2302130898021309,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.884558083360372e-06,
+      "loss": 1.39317741,
+      "memory(GiB)": 84.03,
+      "step": 9075,
+      "train_speed(iter/s)": 1.310287
+    },
+    {
+      "acc": 0.69361525,
+      "epoch": 0.23033992897006594,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.884333943943123e-06,
+      "loss": 1.33644924,
+      "memory(GiB)": 84.03,
+      "step": 9080,
+      "train_speed(iter/s)": 1.310373
+    },
+    {
+      "acc": 0.68455915,
+      "epoch": 0.23046676813800102,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.884109589691062e-06,
+      "loss": 1.44066582,
+      "memory(GiB)": 84.03,
+      "step": 9085,
+      "train_speed(iter/s)": 1.310463
+    },
+    {
+      "acc": 0.67895169,
+      "epoch": 0.23059360730593606,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.883885020614052e-06,
+      "loss": 1.47626953,
+      "memory(GiB)": 84.03,
+      "step": 9090,
+      "train_speed(iter/s)": 1.310554
+    },
+    {
+      "acc": 0.69318953,
+      "epoch": 0.23072044647387113,
+      "grad_norm": 2.625,
+      "learning_rate": 9.883660236721977e-06,
+      "loss": 1.33342667,
+      "memory(GiB)": 84.03,
+      "step": 9095,
+      "train_speed(iter/s)": 1.310642
+    },
+    {
+      "acc": 0.67201533,
+      "epoch": 0.23084728564180618,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.883435238024718e-06,
+      "loss": 1.4595829,
+      "memory(GiB)": 84.03,
+      "step": 9100,
+      "train_speed(iter/s)": 1.310733
+    },
+    {
+      "acc": 0.69347935,
+      "epoch": 0.23097412480974125,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.883210024532176e-06,
+      "loss": 1.35239315,
+      "memory(GiB)": 84.03,
+      "step": 9105,
+      "train_speed(iter/s)": 1.310819
+    },
+    {
+      "acc": 0.691854,
+      "epoch": 0.2311009639776763,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.882984596254255e-06,
+      "loss": 1.43713074,
+      "memory(GiB)": 84.03,
+      "step": 9110,
+      "train_speed(iter/s)": 1.310909
+    },
+    {
+      "acc": 0.68151407,
+      "epoch": 0.23122780314561137,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.88275895320087e-06,
+      "loss": 1.45632114,
+      "memory(GiB)": 84.03,
+      "step": 9115,
+      "train_speed(iter/s)": 1.310989
+    },
+    {
+      "acc": 0.68365388,
+      "epoch": 0.2313546423135464,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.882533095381947e-06,
+      "loss": 1.45072708,
+      "memory(GiB)": 84.03,
+      "step": 9120,
+      "train_speed(iter/s)": 1.311072
+    },
+    {
+      "acc": 0.67326717,
+      "epoch": 0.23148148148148148,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.882307022807419e-06,
+      "loss": 1.41874971,
+      "memory(GiB)": 84.03,
+      "step": 9125,
+      "train_speed(iter/s)": 1.311158
+    },
+    {
+      "acc": 0.67532883,
+      "epoch": 0.23160832064941653,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.88208073548723e-06,
+      "loss": 1.52434359,
+      "memory(GiB)": 84.03,
+      "step": 9130,
+      "train_speed(iter/s)": 1.311246
+    },
+    {
+      "acc": 0.67948365,
+      "epoch": 0.2317351598173516,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.881854233431333e-06,
+      "loss": 1.43893795,
+      "memory(GiB)": 84.03,
+      "step": 9135,
+      "train_speed(iter/s)": 1.311333
+    },
+    {
+      "acc": 0.67838717,
+      "epoch": 0.23186199898528664,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.881627516649692e-06,
+      "loss": 1.42291784,
+      "memory(GiB)": 84.03,
+      "step": 9140,
+      "train_speed(iter/s)": 1.311419
+    },
+    {
+      "acc": 0.66027532,
+      "epoch": 0.23198883815322172,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.881400585152278e-06,
+      "loss": 1.51608286,
+      "memory(GiB)": 84.03,
+      "step": 9145,
+      "train_speed(iter/s)": 1.311506
+    },
+    {
+      "acc": 0.68982182,
+      "epoch": 0.23211567732115676,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.881173438949072e-06,
+      "loss": 1.44998283,
+      "memory(GiB)": 84.03,
+      "step": 9150,
+      "train_speed(iter/s)": 1.311595
+    },
+    {
+      "acc": 0.68820124,
+      "epoch": 0.23224251648909183,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.880946078050064e-06,
+      "loss": 1.3856987,
+      "memory(GiB)": 84.03,
+      "step": 9155,
+      "train_speed(iter/s)": 1.311683
+    },
+    {
+      "acc": 0.66608295,
+      "epoch": 0.23236935565702688,
+      "grad_norm": 2.875,
+      "learning_rate": 9.880718502465258e-06,
+      "loss": 1.49154358,
+      "memory(GiB)": 84.03,
+      "step": 9160,
+      "train_speed(iter/s)": 1.311764
+    },
+    {
+      "acc": 0.66892633,
+      "epoch": 0.23249619482496195,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.88049071220466e-06,
+      "loss": 1.47393303,
+      "memory(GiB)": 84.03,
+      "step": 9165,
+      "train_speed(iter/s)": 1.311852
+    },
+    {
+      "acc": 0.67538261,
+      "epoch": 0.232623033992897,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.88026270727829e-06,
+      "loss": 1.45453272,
+      "memory(GiB)": 84.03,
+      "step": 9170,
+      "train_speed(iter/s)": 1.311942
+    },
+    {
+      "acc": 0.67522297,
+      "epoch": 0.23274987316083207,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.880034487696179e-06,
+      "loss": 1.43073235,
+      "memory(GiB)": 84.03,
+      "step": 9175,
+      "train_speed(iter/s)": 1.312027
+    },
+    {
+      "acc": 0.67537527,
+      "epoch": 0.2328767123287671,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.879806053468361e-06,
+      "loss": 1.45121765,
+      "memory(GiB)": 84.03,
+      "step": 9180,
+      "train_speed(iter/s)": 1.312113
+    },
+    {
+      "acc": 0.6803575,
+      "epoch": 0.23300355149670218,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.879577404604889e-06,
+      "loss": 1.44439211,
+      "memory(GiB)": 84.03,
+      "step": 9185,
+      "train_speed(iter/s)": 1.312198
+    },
+    {
+      "acc": 0.66594472,
+      "epoch": 0.23313039066463723,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.879348541115816e-06,
+      "loss": 1.45645752,
+      "memory(GiB)": 84.03,
+      "step": 9190,
+      "train_speed(iter/s)": 1.312284
+    },
+    {
+      "acc": 0.67499676,
+      "epoch": 0.2332572298325723,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.879119463011208e-06,
+      "loss": 1.43738775,
+      "memory(GiB)": 84.03,
+      "step": 9195,
+      "train_speed(iter/s)": 1.312371
+    },
+    {
+      "acc": 0.69053597,
+      "epoch": 0.23338406900050734,
+      "grad_norm": 3.25,
+      "learning_rate": 9.878890170301143e-06,
+      "loss": 1.39407244,
+      "memory(GiB)": 84.03,
+      "step": 9200,
+      "train_speed(iter/s)": 1.312455
+    },
+    {
+      "acc": 0.69450688,
+      "epoch": 0.23351090816844242,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.878660662995706e-06,
+      "loss": 1.40217667,
+      "memory(GiB)": 84.03,
+      "step": 9205,
+      "train_speed(iter/s)": 1.312544
+    },
+    {
+      "acc": 0.67968383,
+      "epoch": 0.23363774733637746,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.878430941104991e-06,
+      "loss": 1.43128271,
+      "memory(GiB)": 84.03,
+      "step": 9210,
+      "train_speed(iter/s)": 1.312597
+    },
+    {
+      "acc": 0.67851548,
+      "epoch": 0.23376458650431253,
+      "grad_norm": 3.375,
+      "learning_rate": 9.878201004639104e-06,
+      "loss": 1.50149565,
+      "memory(GiB)": 84.03,
+      "step": 9215,
+      "train_speed(iter/s)": 1.312683
+    },
+    {
+      "acc": 0.69577813,
+      "epoch": 0.23389142567224758,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.877970853608156e-06,
+      "loss": 1.39178734,
+      "memory(GiB)": 84.03,
+      "step": 9220,
+      "train_speed(iter/s)": 1.312772
+    },
+    {
+      "acc": 0.67788,
+      "epoch": 0.23401826484018265,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.87774048802227e-06,
+      "loss": 1.4340229,
+      "memory(GiB)": 84.03,
+      "step": 9225,
+      "train_speed(iter/s)": 1.312848
+    },
+    {
+      "acc": 0.67690315,
+      "epoch": 0.2341451040081177,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.877509907891583e-06,
+      "loss": 1.34960032,
+      "memory(GiB)": 84.03,
+      "step": 9230,
+      "train_speed(iter/s)": 1.312938
+    },
+    {
+      "acc": 0.69593868,
+      "epoch": 0.23427194317605277,
+      "grad_norm": 3.0,
+      "learning_rate": 9.877279113226232e-06,
+      "loss": 1.3260047,
+      "memory(GiB)": 84.03,
+      "step": 9235,
+      "train_speed(iter/s)": 1.313021
+    },
+    {
+      "acc": 0.68842597,
+      "epoch": 0.2343987823439878,
+      "grad_norm": 2.875,
+      "learning_rate": 9.87704810403637e-06,
+      "loss": 1.38822956,
+      "memory(GiB)": 84.03,
+      "step": 9240,
+      "train_speed(iter/s)": 1.313106
+    },
+    {
+      "acc": 0.67779236,
+      "epoch": 0.23452562151192288,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.876816880332157e-06,
+      "loss": 1.45777912,
+      "memory(GiB)": 84.03,
+      "step": 9245,
+      "train_speed(iter/s)": 1.313189
+    },
+    {
+      "acc": 0.684867,
+      "epoch": 0.23465246067985793,
+      "grad_norm": 2.328125,
+      "learning_rate": 9.876585442123765e-06,
+      "loss": 1.42217007,
+      "memory(GiB)": 84.03,
+      "step": 9250,
+      "train_speed(iter/s)": 1.313274
+    },
+    {
+      "acc": 0.68420658,
+      "epoch": 0.234779299847793,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.876353789421373e-06,
+      "loss": 1.44369268,
+      "memory(GiB)": 84.03,
+      "step": 9255,
+      "train_speed(iter/s)": 1.313358
+    },
+    {
+      "acc": 0.67105403,
+      "epoch": 0.23490613901572804,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.876121922235171e-06,
+      "loss": 1.50645809,
+      "memory(GiB)": 84.03,
+      "step": 9260,
+      "train_speed(iter/s)": 1.313445
+    },
+    {
+      "acc": 0.68001318,
+      "epoch": 0.23503297818366312,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.875889840575356e-06,
+      "loss": 1.49196453,
+      "memory(GiB)": 84.03,
+      "step": 9265,
+      "train_speed(iter/s)": 1.313532
+    },
+    {
+      "acc": 0.67205734,
+      "epoch": 0.23515981735159816,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.875657544452135e-06,
+      "loss": 1.43081207,
+      "memory(GiB)": 84.03,
+      "step": 9270,
+      "train_speed(iter/s)": 1.313614
+    },
+    {
+      "acc": 0.66637235,
+      "epoch": 0.23528665651953323,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.875425033875728e-06,
+      "loss": 1.42399845,
+      "memory(GiB)": 84.03,
+      "step": 9275,
+      "train_speed(iter/s)": 1.313701
+    },
+    {
+      "acc": 0.69640112,
+      "epoch": 0.23541349568746828,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.875192308856363e-06,
+      "loss": 1.43968048,
+      "memory(GiB)": 84.03,
+      "step": 9280,
+      "train_speed(iter/s)": 1.313787
+    },
+    {
+      "acc": 0.69227295,
+      "epoch": 0.23554033485540335,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.87495936940427e-06,
+      "loss": 1.38581562,
+      "memory(GiB)": 84.03,
+      "step": 9285,
+      "train_speed(iter/s)": 1.313875
+    },
+    {
+      "acc": 0.67052898,
+      "epoch": 0.2356671740233384,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.874726215529702e-06,
+      "loss": 1.47326202,
+      "memory(GiB)": 84.03,
+      "step": 9290,
+      "train_speed(iter/s)": 1.313962
+    },
+    {
+      "acc": 0.68813133,
+      "epoch": 0.23579401319127347,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.87449284724291e-06,
+      "loss": 1.38447733,
+      "memory(GiB)": 84.03,
+      "step": 9295,
+      "train_speed(iter/s)": 1.314049
+    },
+    {
+      "acc": 0.68543549,
+      "epoch": 0.2359208523592085,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.874259264554159e-06,
+      "loss": 1.45233421,
+      "memory(GiB)": 84.03,
+      "step": 9300,
+      "train_speed(iter/s)": 1.314135
+    },
+    {
+      "acc": 0.69366817,
+      "epoch": 0.23604769152714358,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.874025467473722e-06,
+      "loss": 1.36652107,
+      "memory(GiB)": 84.03,
+      "step": 9305,
+      "train_speed(iter/s)": 1.31422
+    },
+    {
+      "acc": 0.66591487,
+      "epoch": 0.23617453069507863,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.873791456011887e-06,
+      "loss": 1.46629848,
+      "memory(GiB)": 84.03,
+      "step": 9310,
+      "train_speed(iter/s)": 1.314304
+    },
+    {
+      "acc": 0.6826148,
+      "epoch": 0.2363013698630137,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.873557230178942e-06,
+      "loss": 1.42551079,
+      "memory(GiB)": 84.03,
+      "step": 9315,
+      "train_speed(iter/s)": 1.314391
+    },
+    {
+      "acc": 0.67972093,
+      "epoch": 0.23642820903094874,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.873322789985191e-06,
+      "loss": 1.39456501,
+      "memory(GiB)": 84.03,
+      "step": 9320,
+      "train_speed(iter/s)": 1.314476
+    },
+    {
+      "acc": 0.67539167,
+      "epoch": 0.23655504819888382,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.873088135440949e-06,
+      "loss": 1.4477684,
+      "memory(GiB)": 84.03,
+      "step": 9325,
+      "train_speed(iter/s)": 1.314553
+    },
+    {
+      "acc": 0.65522208,
+      "epoch": 0.23668188736681886,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.87285326655653e-06,
+      "loss": 1.50778961,
+      "memory(GiB)": 84.03,
+      "step": 9330,
+      "train_speed(iter/s)": 1.314627
+    },
+    {
+      "acc": 0.67891307,
+      "epoch": 0.23680872653475393,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.87261818334227e-06,
+      "loss": 1.41056776,
+      "memory(GiB)": 84.03,
+      "step": 9335,
+      "train_speed(iter/s)": 1.314713
+    },
+    {
+      "acc": 0.65603476,
+      "epoch": 0.23693556570268898,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.87238288580851e-06,
+      "loss": 1.50008039,
+      "memory(GiB)": 84.03,
+      "step": 9340,
+      "train_speed(iter/s)": 1.314797
+    },
+    {
+      "acc": 0.67995701,
+      "epoch": 0.23706240487062405,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.872147373965594e-06,
+      "loss": 1.43530998,
+      "memory(GiB)": 84.03,
+      "step": 9345,
+      "train_speed(iter/s)": 1.314883
+    },
+    {
+      "acc": 0.66755981,
+      "epoch": 0.2371892440385591,
+      "grad_norm": 4.375,
+      "learning_rate": 9.871911647823884e-06,
+      "loss": 1.53725643,
+      "memory(GiB)": 84.03,
+      "step": 9350,
+      "train_speed(iter/s)": 1.314971
+    },
+    {
+      "acc": 0.68585963,
+      "epoch": 0.23731608320649417,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.871675707393749e-06,
+      "loss": 1.41693516,
+      "memory(GiB)": 84.03,
+      "step": 9355,
+      "train_speed(iter/s)": 1.315056
+    },
+    {
+      "acc": 0.68776536,
+      "epoch": 0.2374429223744292,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.871439552685566e-06,
+      "loss": 1.43133411,
+      "memory(GiB)": 84.03,
+      "step": 9360,
+      "train_speed(iter/s)": 1.315139
+    },
+    {
+      "acc": 0.67792888,
+      "epoch": 0.23756976154236428,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.871203183709723e-06,
+      "loss": 1.44563503,
+      "memory(GiB)": 84.03,
+      "step": 9365,
+      "train_speed(iter/s)": 1.315224
+    },
+    {
+      "acc": 0.6920702,
+      "epoch": 0.23769660071029933,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.870966600476614e-06,
+      "loss": 1.40379915,
+      "memory(GiB)": 84.03,
+      "step": 9370,
+      "train_speed(iter/s)": 1.315307
+    },
+    {
+      "acc": 0.69509621,
+      "epoch": 0.2378234398782344,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.870729802996647e-06,
+      "loss": 1.40617199,
+      "memory(GiB)": 84.03,
+      "step": 9375,
+      "train_speed(iter/s)": 1.315393
+    },
+    {
+      "acc": 0.67882495,
+      "epoch": 0.23795027904616944,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.870492791280239e-06,
+      "loss": 1.39430447,
+      "memory(GiB)": 84.03,
+      "step": 9380,
+      "train_speed(iter/s)": 1.315477
+    },
+    {
+      "acc": 0.67627792,
+      "epoch": 0.23807711821410452,
+      "grad_norm": 4.875,
+      "learning_rate": 9.87025556533781e-06,
+      "loss": 1.39250784,
+      "memory(GiB)": 84.03,
+      "step": 9385,
+      "train_speed(iter/s)": 1.315559
+    },
+    {
+      "acc": 0.70598407,
+      "epoch": 0.23820395738203956,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.870018125179799e-06,
+      "loss": 1.33902178,
+      "memory(GiB)": 84.03,
+      "step": 9390,
+      "train_speed(iter/s)": 1.315642
+    },
+    {
+      "acc": 0.69155102,
+      "epoch": 0.23833079654997463,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.869780470816647e-06,
+      "loss": 1.33387766,
+      "memory(GiB)": 84.03,
+      "step": 9395,
+      "train_speed(iter/s)": 1.315723
+    },
+    {
+      "acc": 0.69773474,
+      "epoch": 0.23845763571790968,
+      "grad_norm": 2.5,
+      "learning_rate": 9.869542602258809e-06,
+      "loss": 1.35466537,
+      "memory(GiB)": 84.03,
+      "step": 9400,
+      "train_speed(iter/s)": 1.315803
+    },
+    {
+      "acc": 0.67398801,
+      "epoch": 0.23858447488584475,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.869304519516745e-06,
+      "loss": 1.52385607,
+      "memory(GiB)": 84.03,
+      "step": 9405,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.68254881,
+      "epoch": 0.2387113140537798,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.869066222600928e-06,
+      "loss": 1.45661621,
+      "memory(GiB)": 84.03,
+      "step": 9410,
+      "train_speed(iter/s)": 1.315964
+    },
+    {
+      "acc": 0.68282728,
+      "epoch": 0.23883815322171487,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.86882771152184e-06,
+      "loss": 1.40920486,
+      "memory(GiB)": 84.03,
+      "step": 9415,
+      "train_speed(iter/s)": 1.316041
+    },
+    {
+      "acc": 0.6685986,
+      "epoch": 0.2389649923896499,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.868588986289973e-06,
+      "loss": 1.56194153,
+      "memory(GiB)": 84.03,
+      "step": 9420,
+      "train_speed(iter/s)": 1.316122
+    },
+    {
+      "acc": 0.66190119,
+      "epoch": 0.23909183155758498,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.868350046915825e-06,
+      "loss": 1.51770515,
+      "memory(GiB)": 84.03,
+      "step": 9425,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.67646661,
+      "epoch": 0.23921867072552003,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.868110893409906e-06,
+      "loss": 1.49256039,
+      "memory(GiB)": 84.03,
+      "step": 9430,
+      "train_speed(iter/s)": 1.31629
+    },
+    {
+      "acc": 0.70233703,
+      "epoch": 0.2393455098934551,
+      "grad_norm": 3.875,
+      "learning_rate": 9.867871525782735e-06,
+      "loss": 1.36929321,
+      "memory(GiB)": 84.03,
+      "step": 9435,
+      "train_speed(iter/s)": 1.316376
+    },
+    {
+      "acc": 0.68064117,
+      "epoch": 0.23947234906139014,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.86763194404484e-06,
+      "loss": 1.43110933,
+      "memory(GiB)": 84.03,
+      "step": 9440,
+      "train_speed(iter/s)": 1.316462
+    },
+    {
+      "acc": 0.66729283,
+      "epoch": 0.23959918822932522,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.867392148206762e-06,
+      "loss": 1.46242313,
+      "memory(GiB)": 84.03,
+      "step": 9445,
+      "train_speed(iter/s)": 1.316542
+    },
+    {
+      "acc": 0.67945309,
+      "epoch": 0.23972602739726026,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.867152138279043e-06,
+      "loss": 1.43903503,
+      "memory(GiB)": 84.03,
+      "step": 9450,
+      "train_speed(iter/s)": 1.316624
+    },
+    {
+      "acc": 0.68202424,
+      "epoch": 0.23985286656519533,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.866911914272246e-06,
+      "loss": 1.38851652,
+      "memory(GiB)": 84.03,
+      "step": 9455,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.68467937,
+      "epoch": 0.23997970573313038,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.866671476196931e-06,
+      "loss": 1.42180691,
+      "memory(GiB)": 84.03,
+      "step": 9460,
+      "train_speed(iter/s)": 1.31679
+    },
+    {
+      "acc": 0.68919768,
+      "epoch": 0.24010654490106545,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.866430824063678e-06,
+      "loss": 1.3800148,
+      "memory(GiB)": 84.03,
+      "step": 9465,
+      "train_speed(iter/s)": 1.31687
+    },
+    {
+      "acc": 0.69198942,
+      "epoch": 0.2402333840690005,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.86618995788307e-06,
+      "loss": 1.37564373,
+      "memory(GiB)": 93.76,
+      "step": 9470,
+      "train_speed(iter/s)": 1.316937
+    },
+    {
+      "acc": 0.66267934,
+      "epoch": 0.24036022323693557,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.865948877665702e-06,
+      "loss": 1.45800877,
+      "memory(GiB)": 93.76,
+      "step": 9475,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.68175745,
+      "epoch": 0.2404870624048706,
+      "grad_norm": 3.75,
+      "learning_rate": 9.865707583422178e-06,
+      "loss": 1.44921055,
+      "memory(GiB)": 93.76,
+      "step": 9480,
+      "train_speed(iter/s)": 1.317101
+    },
+    {
+      "acc": 0.67957277,
+      "epoch": 0.24061390157280568,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.865466075163108e-06,
+      "loss": 1.43240433,
+      "memory(GiB)": 93.76,
+      "step": 9485,
+      "train_speed(iter/s)": 1.317183
+    },
+    {
+      "acc": 0.69667454,
+      "epoch": 0.24074074074074073,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.86522435289912e-06,
+      "loss": 1.38396654,
+      "memory(GiB)": 93.76,
+      "step": 9490,
+      "train_speed(iter/s)": 1.317264
+    },
+    {
+      "acc": 0.66818547,
+      "epoch": 0.2408675799086758,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.864982416640843e-06,
+      "loss": 1.51250324,
+      "memory(GiB)": 93.76,
+      "step": 9495,
+      "train_speed(iter/s)": 1.317347
+    },
+    {
+      "acc": 0.6829308,
+      "epoch": 0.24099441907661084,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.864740266398918e-06,
+      "loss": 1.43318043,
+      "memory(GiB)": 93.76,
+      "step": 9500,
+      "train_speed(iter/s)": 1.31743
+    },
+    {
+      "acc": 0.68914089,
+      "epoch": 0.24112125824454592,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.864497902183996e-06,
+      "loss": 1.38145618,
+      "memory(GiB)": 93.76,
+      "step": 9505,
+      "train_speed(iter/s)": 1.317511
+    },
+    {
+      "acc": 0.69036136,
+      "epoch": 0.24124809741248096,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.864255324006738e-06,
+      "loss": 1.3941494,
+      "memory(GiB)": 93.76,
+      "step": 9510,
+      "train_speed(iter/s)": 1.317593
+    },
+    {
+      "acc": 0.66944723,
+      "epoch": 0.24137493658041603,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.864012531877814e-06,
+      "loss": 1.50283699,
+      "memory(GiB)": 93.76,
+      "step": 9515,
+      "train_speed(iter/s)": 1.317677
+    },
+    {
+      "acc": 0.66870332,
+      "epoch": 0.24150177574835108,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.863769525807903e-06,
+      "loss": 1.48253479,
+      "memory(GiB)": 93.76,
+      "step": 9520,
+      "train_speed(iter/s)": 1.31776
+    },
+    {
+      "acc": 0.70016055,
+      "epoch": 0.24162861491628615,
+      "grad_norm": 3.25,
+      "learning_rate": 9.863526305807694e-06,
+      "loss": 1.34306049,
+      "memory(GiB)": 93.76,
+      "step": 9525,
+      "train_speed(iter/s)": 1.317845
+    },
+    {
+      "acc": 0.68171797,
+      "epoch": 0.2417554540842212,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.863282871887882e-06,
+      "loss": 1.49825525,
+      "memory(GiB)": 93.76,
+      "step": 9530,
+      "train_speed(iter/s)": 1.317931
+    },
+    {
+      "acc": 0.68217869,
+      "epoch": 0.24188229325215627,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.863039224059177e-06,
+      "loss": 1.43053675,
+      "memory(GiB)": 93.76,
+      "step": 9535,
+      "train_speed(iter/s)": 1.31801
+    },
+    {
+      "acc": 0.68154278,
+      "epoch": 0.2420091324200913,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.862795362332293e-06,
+      "loss": 1.3784214,
+      "memory(GiB)": 93.76,
+      "step": 9540,
+      "train_speed(iter/s)": 1.318089
+    },
+    {
+      "acc": 0.69521079,
+      "epoch": 0.24213597158802638,
+      "grad_norm": 3.0,
+      "learning_rate": 9.862551286717961e-06,
+      "loss": 1.3801116,
+      "memory(GiB)": 93.76,
+      "step": 9545,
+      "train_speed(iter/s)": 1.318172
+    },
+    {
+      "acc": 0.68453879,
+      "epoch": 0.24226281075596143,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.862306997226914e-06,
+      "loss": 1.39400158,
+      "memory(GiB)": 93.76,
+      "step": 9550,
+      "train_speed(iter/s)": 1.31825
+    },
+    {
+      "acc": 0.67447324,
+      "epoch": 0.2423896499238965,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.862062493869895e-06,
+      "loss": 1.45840311,
+      "memory(GiB)": 93.76,
+      "step": 9555,
+      "train_speed(iter/s)": 1.318329
+    },
+    {
+      "acc": 0.68349257,
+      "epoch": 0.24251648909183154,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.861817776657661e-06,
+      "loss": 1.37065248,
+      "memory(GiB)": 93.76,
+      "step": 9560,
+      "train_speed(iter/s)": 1.318406
+    },
+    {
+      "acc": 0.67111783,
+      "epoch": 0.24264332825976662,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.861572845600973e-06,
+      "loss": 1.45604973,
+      "memory(GiB)": 93.76,
+      "step": 9565,
+      "train_speed(iter/s)": 1.318487
+    },
+    {
+      "acc": 0.67911301,
+      "epoch": 0.24277016742770166,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.861327700710608e-06,
+      "loss": 1.38650141,
+      "memory(GiB)": 93.76,
+      "step": 9570,
+      "train_speed(iter/s)": 1.318567
+    },
+    {
+      "acc": 0.67480974,
+      "epoch": 0.24289700659563673,
+      "grad_norm": 3.0,
+      "learning_rate": 9.861082341997345e-06,
+      "loss": 1.47019482,
+      "memory(GiB)": 93.76,
+      "step": 9575,
+      "train_speed(iter/s)": 1.31865
+    },
+    {
+      "acc": 0.68483829,
+      "epoch": 0.24302384576357178,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.860836769471977e-06,
+      "loss": 1.40375004,
+      "memory(GiB)": 93.76,
+      "step": 9580,
+      "train_speed(iter/s)": 1.318732
+    },
+    {
+      "acc": 0.67409525,
+      "epoch": 0.24315068493150685,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.860590983145307e-06,
+      "loss": 1.48943214,
+      "memory(GiB)": 93.76,
+      "step": 9585,
+      "train_speed(iter/s)": 1.318814
+    },
+    {
+      "acc": 0.67990246,
+      "epoch": 0.2432775240994419,
+      "grad_norm": 3.625,
+      "learning_rate": 9.860344983028146e-06,
+      "loss": 1.44887028,
+      "memory(GiB)": 93.76,
+      "step": 9590,
+      "train_speed(iter/s)": 1.318896
+    },
+    {
+      "acc": 0.67374649,
+      "epoch": 0.24340436326737697,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.86009876913131e-06,
+      "loss": 1.46806545,
+      "memory(GiB)": 93.76,
+      "step": 9595,
+      "train_speed(iter/s)": 1.318977
+    },
+    {
+      "acc": 0.67905297,
+      "epoch": 0.243531202435312,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.859852341465633e-06,
+      "loss": 1.4843565,
+      "memory(GiB)": 93.76,
+      "step": 9600,
+      "train_speed(iter/s)": 1.319061
+    },
+    {
+      "acc": 0.6821878,
+      "epoch": 0.24365804160324708,
+      "grad_norm": 3.0,
+      "learning_rate": 9.859605700041951e-06,
+      "loss": 1.46332464,
+      "memory(GiB)": 93.76,
+      "step": 9605,
+      "train_speed(iter/s)": 1.319136
+    },
+    {
+      "acc": 0.7017065,
+      "epoch": 0.24378488077118213,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.859358844871113e-06,
+      "loss": 1.34398022,
+      "memory(GiB)": 93.76,
+      "step": 9610,
+      "train_speed(iter/s)": 1.319217
+    },
+    {
+      "acc": 0.68478508,
+      "epoch": 0.2439117199391172,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.859111775963981e-06,
+      "loss": 1.41572456,
+      "memory(GiB)": 93.76,
+      "step": 9615,
+      "train_speed(iter/s)": 1.3193
+    },
+    {
+      "acc": 0.68865566,
+      "epoch": 0.24403855910705224,
+      "grad_norm": 3.0,
+      "learning_rate": 9.858864493331417e-06,
+      "loss": 1.41181431,
+      "memory(GiB)": 93.76,
+      "step": 9620,
+      "train_speed(iter/s)": 1.31938
+    },
+    {
+      "acc": 0.66261959,
+      "epoch": 0.24416539827498732,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.858616996984297e-06,
+      "loss": 1.50776348,
+      "memory(GiB)": 93.76,
+      "step": 9625,
+      "train_speed(iter/s)": 1.319463
+    },
+    {
+      "acc": 0.66796713,
+      "epoch": 0.24429223744292236,
+      "grad_norm": 3.0,
+      "learning_rate": 9.858369286933513e-06,
+      "loss": 1.46185951,
+      "memory(GiB)": 93.76,
+      "step": 9630,
+      "train_speed(iter/s)": 1.319542
+    },
+    {
+      "acc": 0.68332753,
+      "epoch": 0.24441907661085743,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.858121363189954e-06,
+      "loss": 1.43373928,
+      "memory(GiB)": 93.76,
+      "step": 9635,
+      "train_speed(iter/s)": 1.319621
+    },
+    {
+      "acc": 0.67353287,
+      "epoch": 0.24454591577879248,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.85787322576453e-06,
+      "loss": 1.51023569,
+      "memory(GiB)": 93.76,
+      "step": 9640,
+      "train_speed(iter/s)": 1.319704
+    },
+    {
+      "acc": 0.67769651,
+      "epoch": 0.24467275494672755,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.85762487466815e-06,
+      "loss": 1.43726168,
+      "memory(GiB)": 93.76,
+      "step": 9645,
+      "train_speed(iter/s)": 1.319784
+    },
+    {
+      "acc": 0.69284477,
+      "epoch": 0.2447995941146626,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.857376309911741e-06,
+      "loss": 1.32456913,
+      "memory(GiB)": 93.76,
+      "step": 9650,
+      "train_speed(iter/s)": 1.319865
+    },
+    {
+      "acc": 0.67259922,
+      "epoch": 0.24492643328259767,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.857127531506237e-06,
+      "loss": 1.48577366,
+      "memory(GiB)": 93.76,
+      "step": 9655,
+      "train_speed(iter/s)": 1.319945
+    },
+    {
+      "acc": 0.68021173,
+      "epoch": 0.2450532724505327,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.856878539462577e-06,
+      "loss": 1.42892094,
+      "memory(GiB)": 93.76,
+      "step": 9660,
+      "train_speed(iter/s)": 1.320024
+    },
+    {
+      "acc": 0.69523549,
+      "epoch": 0.24518011161846778,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.856629333791716e-06,
+      "loss": 1.36193981,
+      "memory(GiB)": 93.76,
+      "step": 9665,
+      "train_speed(iter/s)": 1.320103
+    },
+    {
+      "acc": 0.67922378,
+      "epoch": 0.24530695078640283,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.856379914504612e-06,
+      "loss": 1.43835545,
+      "memory(GiB)": 93.76,
+      "step": 9670,
+      "train_speed(iter/s)": 1.320179
+    },
+    {
+      "acc": 0.68078823,
+      "epoch": 0.2454337899543379,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.856130281612237e-06,
+      "loss": 1.45848961,
+      "memory(GiB)": 93.76,
+      "step": 9675,
+      "train_speed(iter/s)": 1.320258
+    },
+    {
+      "acc": 0.67439818,
+      "epoch": 0.24556062912227294,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.855880435125572e-06,
+      "loss": 1.39037094,
+      "memory(GiB)": 93.76,
+      "step": 9680,
+      "train_speed(iter/s)": 1.320337
+    },
+    {
+      "acc": 0.67947378,
+      "epoch": 0.24568746829020802,
+      "grad_norm": 3.25,
+      "learning_rate": 9.855630375055604e-06,
+      "loss": 1.43067064,
+      "memory(GiB)": 93.76,
+      "step": 9685,
+      "train_speed(iter/s)": 1.320414
+    },
+    {
+      "acc": 0.68154192,
+      "epoch": 0.24581430745814306,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.855380101413336e-06,
+      "loss": 1.43296423,
+      "memory(GiB)": 93.76,
+      "step": 9690,
+      "train_speed(iter/s)": 1.320488
+    },
+    {
+      "acc": 0.68075433,
+      "epoch": 0.24594114662607813,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.855129614209771e-06,
+      "loss": 1.41093388,
+      "memory(GiB)": 93.76,
+      "step": 9695,
+      "train_speed(iter/s)": 1.320564
+    },
+    {
+      "acc": 0.68623571,
+      "epoch": 0.24606798579401318,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.85487891345593e-06,
+      "loss": 1.46732693,
+      "memory(GiB)": 93.76,
+      "step": 9700,
+      "train_speed(iter/s)": 1.320642
+    },
+    {
+      "acc": 0.66578569,
+      "epoch": 0.24619482496194825,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.85462799916284e-06,
+      "loss": 1.48664598,
+      "memory(GiB)": 93.76,
+      "step": 9705,
+      "train_speed(iter/s)": 1.32072
+    },
+    {
+      "acc": 0.67233696,
+      "epoch": 0.2463216641298833,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.854376871341535e-06,
+      "loss": 1.47396374,
+      "memory(GiB)": 93.76,
+      "step": 9710,
+      "train_speed(iter/s)": 1.320793
+    },
+    {
+      "acc": 0.66588821,
+      "epoch": 0.24644850329781837,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.854125530003063e-06,
+      "loss": 1.48761272,
+      "memory(GiB)": 93.76,
+      "step": 9715,
+      "train_speed(iter/s)": 1.320873
+    },
+    {
+      "acc": 0.68799429,
+      "epoch": 0.2465753424657534,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.853873975158476e-06,
+      "loss": 1.38804474,
+      "memory(GiB)": 93.76,
+      "step": 9720,
+      "train_speed(iter/s)": 1.320949
+    },
+    {
+      "acc": 0.67504864,
+      "epoch": 0.24670218163368848,
+      "grad_norm": 3.0,
+      "learning_rate": 9.853622206818842e-06,
+      "loss": 1.48541336,
+      "memory(GiB)": 93.76,
+      "step": 9725,
+      "train_speed(iter/s)": 1.321023
+    },
+    {
+      "acc": 0.68161778,
+      "epoch": 0.24682902080162353,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.853370224995233e-06,
+      "loss": 1.42324009,
+      "memory(GiB)": 93.76,
+      "step": 9730,
+      "train_speed(iter/s)": 1.321099
+    },
+    {
+      "acc": 0.69487791,
+      "epoch": 0.2469558599695586,
+      "grad_norm": 3.0,
+      "learning_rate": 9.853118029698733e-06,
+      "loss": 1.35707359,
+      "memory(GiB)": 93.76,
+      "step": 9735,
+      "train_speed(iter/s)": 1.321157
+    },
+    {
+      "acc": 0.68766155,
+      "epoch": 0.24708269913749364,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.852865620940436e-06,
+      "loss": 1.45484657,
+      "memory(GiB)": 93.76,
+      "step": 9740,
+      "train_speed(iter/s)": 1.321229
+    },
+    {
+      "acc": 0.68050303,
+      "epoch": 0.24720953830542872,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.85261299873144e-06,
+      "loss": 1.43673897,
+      "memory(GiB)": 93.76,
+      "step": 9745,
+      "train_speed(iter/s)": 1.321308
+    },
+    {
+      "acc": 0.69828467,
+      "epoch": 0.24733637747336376,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.85236016308286e-06,
+      "loss": 1.39915609,
+      "memory(GiB)": 93.76,
+      "step": 9750,
+      "train_speed(iter/s)": 1.321387
+    },
+    {
+      "acc": 0.68661218,
+      "epoch": 0.24746321664129883,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.852107114005816e-06,
+      "loss": 1.44194984,
+      "memory(GiB)": 93.76,
+      "step": 9755,
+      "train_speed(iter/s)": 1.32146
+    },
+    {
+      "acc": 0.70407424,
+      "epoch": 0.24759005580923388,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.851853851511437e-06,
+      "loss": 1.37923088,
+      "memory(GiB)": 93.76,
+      "step": 9760,
+      "train_speed(iter/s)": 1.321538
+    },
+    {
+      "acc": 0.6867692,
+      "epoch": 0.24771689497716895,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.851600375610864e-06,
+      "loss": 1.40179787,
+      "memory(GiB)": 93.76,
+      "step": 9765,
+      "train_speed(iter/s)": 1.321618
+    },
+    {
+      "acc": 0.68123245,
+      "epoch": 0.247843734145104,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.851346686315246e-06,
+      "loss": 1.44075823,
+      "memory(GiB)": 93.76,
+      "step": 9770,
+      "train_speed(iter/s)": 1.321692
+    },
+    {
+      "acc": 0.67268467,
+      "epoch": 0.24797057331303907,
+      "grad_norm": 2.625,
+      "learning_rate": 9.851092783635742e-06,
+      "loss": 1.49451008,
+      "memory(GiB)": 93.76,
+      "step": 9775,
+      "train_speed(iter/s)": 1.321769
+    },
+    {
+      "acc": 0.68065453,
+      "epoch": 0.2480974124809741,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.850838667583518e-06,
+      "loss": 1.40958385,
+      "memory(GiB)": 93.76,
+      "step": 9780,
+      "train_speed(iter/s)": 1.321845
+    },
+    {
+      "acc": 0.68943186,
+      "epoch": 0.24822425164890918,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.850584338169752e-06,
+      "loss": 1.3800148,
+      "memory(GiB)": 93.76,
+      "step": 9785,
+      "train_speed(iter/s)": 1.321922
+    },
+    {
+      "acc": 0.67500005,
+      "epoch": 0.24835109081684423,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.85032979540563e-06,
+      "loss": 1.44119396,
+      "memory(GiB)": 93.76,
+      "step": 9790,
+      "train_speed(iter/s)": 1.322001
+    },
+    {
+      "acc": 0.67247376,
+      "epoch": 0.2484779299847793,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.85007503930235e-06,
+      "loss": 1.45250359,
+      "memory(GiB)": 93.76,
+      "step": 9795,
+      "train_speed(iter/s)": 1.322079
+    },
+    {
+      "acc": 0.68435516,
+      "epoch": 0.24860476915271434,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.849820069871114e-06,
+      "loss": 1.4075634,
+      "memory(GiB)": 93.76,
+      "step": 9800,
+      "train_speed(iter/s)": 1.322157
+    },
+    {
+      "acc": 0.68064008,
+      "epoch": 0.24873160832064942,
+      "grad_norm": 3.75,
+      "learning_rate": 9.849564887123138e-06,
+      "loss": 1.44379349,
+      "memory(GiB)": 93.76,
+      "step": 9805,
+      "train_speed(iter/s)": 1.322234
+    },
+    {
+      "acc": 0.6675241,
+      "epoch": 0.24885844748858446,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.849309491069647e-06,
+      "loss": 1.44936991,
+      "memory(GiB)": 93.76,
+      "step": 9810,
+      "train_speed(iter/s)": 1.322311
+    },
+    {
+      "acc": 0.68091555,
+      "epoch": 0.24898528665651953,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.849053881721876e-06,
+      "loss": 1.38546772,
+      "memory(GiB)": 93.76,
+      "step": 9815,
+      "train_speed(iter/s)": 1.322389
+    },
+    {
+      "acc": 0.68581128,
+      "epoch": 0.24911212582445458,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.848798059091064e-06,
+      "loss": 1.45667515,
+      "memory(GiB)": 93.76,
+      "step": 9820,
+      "train_speed(iter/s)": 1.322467
+    },
+    {
+      "acc": 0.68609962,
+      "epoch": 0.24923896499238965,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.848542023188466e-06,
+      "loss": 1.44928503,
+      "memory(GiB)": 93.76,
+      "step": 9825,
+      "train_speed(iter/s)": 1.322539
+    },
+    {
+      "acc": 0.70357208,
+      "epoch": 0.2493658041603247,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.848285774025342e-06,
+      "loss": 1.32036247,
+      "memory(GiB)": 93.76,
+      "step": 9830,
+      "train_speed(iter/s)": 1.322616
+    },
+    {
+      "acc": 0.67644033,
+      "epoch": 0.24949264332825977,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.848029311612963e-06,
+      "loss": 1.4268486,
+      "memory(GiB)": 93.76,
+      "step": 9835,
+      "train_speed(iter/s)": 1.322693
+    },
+    {
+      "acc": 0.68404713,
+      "epoch": 0.2496194824961948,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.84777263596261e-06,
+      "loss": 1.42398129,
+      "memory(GiB)": 93.76,
+      "step": 9840,
+      "train_speed(iter/s)": 1.32277
+    },
+    {
+      "acc": 0.674084,
+      "epoch": 0.24974632166412988,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.847515747085573e-06,
+      "loss": 1.46803112,
+      "memory(GiB)": 93.76,
+      "step": 9845,
+      "train_speed(iter/s)": 1.322847
+    },
+    {
+      "acc": 0.68006659,
+      "epoch": 0.24987316083206493,
+      "grad_norm": 2.75,
+      "learning_rate": 9.847258644993151e-06,
+      "loss": 1.46226816,
+      "memory(GiB)": 93.76,
+      "step": 9850,
+      "train_speed(iter/s)": 1.32292
+    },
+    {
+      "acc": 0.66462469,
+      "epoch": 0.25,
+      "grad_norm": 2.75,
+      "learning_rate": 9.847001329696653e-06,
+      "loss": 1.47678661,
+      "memory(GiB)": 93.76,
+      "step": 9855,
+      "train_speed(iter/s)": 1.322996
+    },
+    {
+      "acc": 0.67995396,
+      "epoch": 0.25012683916793504,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.846743801207395e-06,
+      "loss": 1.41272507,
+      "memory(GiB)": 93.76,
+      "step": 9860,
+      "train_speed(iter/s)": 1.323072
+    },
+    {
+      "acc": 0.67065735,
+      "epoch": 0.25025367833587014,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.846486059536706e-06,
+      "loss": 1.46339684,
+      "memory(GiB)": 103.55,
+      "step": 9865,
+      "train_speed(iter/s)": 1.323129
+    },
+    {
+      "acc": 0.67683983,
+      "epoch": 0.2503805175038052,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.846228104695922e-06,
+      "loss": 1.43174019,
+      "memory(GiB)": 103.55,
+      "step": 9870,
+      "train_speed(iter/s)": 1.323207
+    },
+    {
+      "acc": 0.66574955,
+      "epoch": 0.25050735667174023,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.84596993669639e-06,
+      "loss": 1.54087257,
+      "memory(GiB)": 103.55,
+      "step": 9875,
+      "train_speed(iter/s)": 1.323286
+    },
+    {
+      "acc": 0.68101425,
+      "epoch": 0.2506341958396753,
+      "grad_norm": 6.90625,
+      "learning_rate": 9.845711555549464e-06,
+      "loss": 1.4107029,
+      "memory(GiB)": 103.55,
+      "step": 9880,
+      "train_speed(iter/s)": 1.323362
+    },
+    {
+      "acc": 0.68041811,
+      "epoch": 0.2507610350076104,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.845452961266509e-06,
+      "loss": 1.40414886,
+      "memory(GiB)": 103.55,
+      "step": 9885,
+      "train_speed(iter/s)": 1.323437
+    },
+    {
+      "acc": 0.68644519,
+      "epoch": 0.2508878741755454,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.845194153858899e-06,
+      "loss": 1.4184042,
+      "memory(GiB)": 103.55,
+      "step": 9890,
+      "train_speed(iter/s)": 1.323514
+    },
+    {
+      "acc": 0.68026648,
+      "epoch": 0.25101471334348047,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.844935133338018e-06,
+      "loss": 1.37730131,
+      "memory(GiB)": 103.55,
+      "step": 9895,
+      "train_speed(iter/s)": 1.323588
+    },
+    {
+      "acc": 0.65919342,
+      "epoch": 0.2511415525114155,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.84467589971526e-06,
+      "loss": 1.52961578,
+      "memory(GiB)": 103.55,
+      "step": 9900,
+      "train_speed(iter/s)": 1.323661
+    },
+    {
+      "acc": 0.67672415,
+      "epoch": 0.2512683916793506,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.844416453002027e-06,
+      "loss": 1.45850115,
+      "memory(GiB)": 103.55,
+      "step": 9905,
+      "train_speed(iter/s)": 1.323724
+    },
+    {
+      "acc": 0.68315177,
+      "epoch": 0.25139523084728566,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.844156793209725e-06,
+      "loss": 1.4085453,
+      "memory(GiB)": 103.55,
+      "step": 9910,
+      "train_speed(iter/s)": 1.323801
+    },
+    {
+      "acc": 0.68720455,
+      "epoch": 0.2515220700152207,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.843896920349783e-06,
+      "loss": 1.41604767,
+      "memory(GiB)": 103.55,
+      "step": 9915,
+      "train_speed(iter/s)": 1.323878
+    },
+    {
+      "acc": 0.68393903,
+      "epoch": 0.25164890918315574,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.843636834433627e-06,
+      "loss": 1.44866142,
+      "memory(GiB)": 103.55,
+      "step": 9920,
+      "train_speed(iter/s)": 1.323957
+    },
+    {
+      "acc": 0.68078394,
+      "epoch": 0.25177574835109084,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.843376535472698e-06,
+      "loss": 1.41035252,
+      "memory(GiB)": 103.55,
+      "step": 9925,
+      "train_speed(iter/s)": 1.324034
+    },
+    {
+      "acc": 0.69674215,
+      "epoch": 0.2519025875190259,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.843116023478445e-06,
+      "loss": 1.34639101,
+      "memory(GiB)": 103.55,
+      "step": 9930,
+      "train_speed(iter/s)": 1.324108
+    },
+    {
+      "acc": 0.67656407,
+      "epoch": 0.25202942668696093,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.842855298462327e-06,
+      "loss": 1.45146332,
+      "memory(GiB)": 103.55,
+      "step": 9935,
+      "train_speed(iter/s)": 1.324184
+    },
+    {
+      "acc": 0.67593193,
+      "epoch": 0.252156265854896,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.84259436043581e-06,
+      "loss": 1.4394496,
+      "memory(GiB)": 103.55,
+      "step": 9940,
+      "train_speed(iter/s)": 1.324261
+    },
+    {
+      "acc": 0.68415928,
+      "epoch": 0.2522831050228311,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.842333209410372e-06,
+      "loss": 1.41075211,
+      "memory(GiB)": 103.55,
+      "step": 9945,
+      "train_speed(iter/s)": 1.324335
+    },
+    {
+      "acc": 0.68679538,
+      "epoch": 0.2524099441907661,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.842071845397502e-06,
+      "loss": 1.33731613,
+      "memory(GiB)": 103.55,
+      "step": 9950,
+      "train_speed(iter/s)": 1.324412
+    },
+    {
+      "acc": 0.67182646,
+      "epoch": 0.25253678335870117,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.841810268408692e-06,
+      "loss": 1.46188736,
+      "memory(GiB)": 103.55,
+      "step": 9955,
+      "train_speed(iter/s)": 1.324489
+    },
+    {
+      "acc": 0.6844018,
+      "epoch": 0.2526636225266362,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.841548478455451e-06,
+      "loss": 1.47429409,
+      "memory(GiB)": 103.55,
+      "step": 9960,
+      "train_speed(iter/s)": 1.324563
+    },
+    {
+      "acc": 0.68647189,
+      "epoch": 0.2527904616945713,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.841286475549291e-06,
+      "loss": 1.46064367,
+      "memory(GiB)": 103.55,
+      "step": 9965,
+      "train_speed(iter/s)": 1.324637
+    },
+    {
+      "acc": 0.68212109,
+      "epoch": 0.25291730086250636,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.841024259701737e-06,
+      "loss": 1.40103273,
+      "memory(GiB)": 103.55,
+      "step": 9970,
+      "train_speed(iter/s)": 1.324711
+    },
+    {
+      "acc": 0.67887249,
+      "epoch": 0.2530441400304414,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.840761830924323e-06,
+      "loss": 1.43572884,
+      "memory(GiB)": 103.55,
+      "step": 9975,
+      "train_speed(iter/s)": 1.324786
+    },
+    {
+      "acc": 0.66993284,
+      "epoch": 0.25317097919837644,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.84049918922859e-06,
+      "loss": 1.4634531,
+      "memory(GiB)": 103.55,
+      "step": 9980,
+      "train_speed(iter/s)": 1.324864
+    },
+    {
+      "acc": 0.67143688,
+      "epoch": 0.25329781836631154,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.840236334626091e-06,
+      "loss": 1.42774639,
+      "memory(GiB)": 103.55,
+      "step": 9985,
+      "train_speed(iter/s)": 1.324937
+    },
+    {
+      "acc": 0.68185239,
+      "epoch": 0.2534246575342466,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.83997326712839e-06,
+      "loss": 1.40808277,
+      "memory(GiB)": 103.55,
+      "step": 9990,
+      "train_speed(iter/s)": 1.325011
+    },
+    {
+      "acc": 0.69770985,
+      "epoch": 0.25355149670218163,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.839709986747054e-06,
+      "loss": 1.36357841,
+      "memory(GiB)": 103.55,
+      "step": 9995,
+      "train_speed(iter/s)": 1.325085
+    },
+    {
+      "acc": 0.68477201,
+      "epoch": 0.2536783358701167,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.839446493493667e-06,
+      "loss": 1.37264128,
+      "memory(GiB)": 103.55,
+      "step": 10000,
+      "train_speed(iter/s)": 1.325159
+    },
+    {
+      "epoch": 0.2536783358701167,
+      "eval_acc": 0.6693012736578978,
+      "eval_loss": 1.3885244131088257,
+      "eval_runtime": 69.1313,
+      "eval_samples_per_second": 92.143,
+      "eval_steps_per_second": 23.043,
+      "step": 10000
+    },
+    {
+      "acc": 0.68757687,
+      "epoch": 0.2538051750380518,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.839182787379815e-06,
+      "loss": 1.36487532,
+      "memory(GiB)": 103.55,
+      "step": 10005,
+      "train_speed(iter/s)": 1.311022
+    },
+    {
+      "acc": 0.68414311,
+      "epoch": 0.2539320142059868,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.8389188684171e-06,
+      "loss": 1.38915787,
+      "memory(GiB)": 103.55,
+      "step": 10010,
+      "train_speed(iter/s)": 1.311103
+    },
+    {
+      "acc": 0.68516779,
+      "epoch": 0.25405885337392187,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.838654736617128e-06,
+      "loss": 1.33549871,
+      "memory(GiB)": 103.55,
+      "step": 10015,
+      "train_speed(iter/s)": 1.311184
+    },
+    {
+      "acc": 0.68704333,
+      "epoch": 0.2541856925418569,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.838390391991517e-06,
+      "loss": 1.40608444,
+      "memory(GiB)": 103.55,
+      "step": 10020,
+      "train_speed(iter/s)": 1.311265
+    },
+    {
+      "acc": 0.67796764,
+      "epoch": 0.254312531709792,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.838125834551895e-06,
+      "loss": 1.45913467,
+      "memory(GiB)": 103.55,
+      "step": 10025,
+      "train_speed(iter/s)": 1.311339
+    },
+    {
+      "acc": 0.68167882,
+      "epoch": 0.25443937087772706,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.837861064309899e-06,
+      "loss": 1.4413662,
+      "memory(GiB)": 103.55,
+      "step": 10030,
+      "train_speed(iter/s)": 1.31142
+    },
+    {
+      "acc": 0.68415589,
+      "epoch": 0.2545662100456621,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.837596081277173e-06,
+      "loss": 1.35038576,
+      "memory(GiB)": 103.55,
+      "step": 10035,
+      "train_speed(iter/s)": 1.311489
+    },
+    {
+      "acc": 0.69183168,
+      "epoch": 0.25469304921359714,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.837330885465373e-06,
+      "loss": 1.42365227,
+      "memory(GiB)": 103.55,
+      "step": 10040,
+      "train_speed(iter/s)": 1.31157
+    },
+    {
+      "acc": 0.68065081,
+      "epoch": 0.25481988838153224,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.837065476886163e-06,
+      "loss": 1.44919844,
+      "memory(GiB)": 103.55,
+      "step": 10045,
+      "train_speed(iter/s)": 1.311651
+    },
+    {
+      "acc": 0.67587323,
+      "epoch": 0.2549467275494673,
+      "grad_norm": 3.75,
+      "learning_rate": 9.83679985555122e-06,
+      "loss": 1.43420696,
+      "memory(GiB)": 103.55,
+      "step": 10050,
+      "train_speed(iter/s)": 1.311731
+    },
+    {
+      "acc": 0.67729011,
+      "epoch": 0.25507356671740233,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.836534021472222e-06,
+      "loss": 1.43121109,
+      "memory(GiB)": 103.55,
+      "step": 10055,
+      "train_speed(iter/s)": 1.311811
+    },
+    {
+      "acc": 0.68388872,
+      "epoch": 0.2552004058853374,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.836267974660866e-06,
+      "loss": 1.44481163,
+      "memory(GiB)": 103.55,
+      "step": 10060,
+      "train_speed(iter/s)": 1.31189
+    },
+    {
+      "acc": 0.67942362,
+      "epoch": 0.2553272450532725,
+      "grad_norm": 3.125,
+      "learning_rate": 9.836001715128851e-06,
+      "loss": 1.45349636,
+      "memory(GiB)": 103.55,
+      "step": 10065,
+      "train_speed(iter/s)": 1.311969
+    },
+    {
+      "acc": 0.68752117,
+      "epoch": 0.2554540842212075,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.835735242887889e-06,
+      "loss": 1.41230135,
+      "memory(GiB)": 103.55,
+      "step": 10070,
+      "train_speed(iter/s)": 1.312053
+    },
+    {
+      "acc": 0.68746653,
+      "epoch": 0.25558092338914257,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.835468557949701e-06,
+      "loss": 1.40399246,
+      "memory(GiB)": 103.55,
+      "step": 10075,
+      "train_speed(iter/s)": 1.312133
+    },
+    {
+      "acc": 0.67599564,
+      "epoch": 0.2557077625570776,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.83520166032602e-06,
+      "loss": 1.43576517,
+      "memory(GiB)": 103.55,
+      "step": 10080,
+      "train_speed(iter/s)": 1.312191
+    },
+    {
+      "acc": 0.67216816,
+      "epoch": 0.2558346017250127,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.834934550028579e-06,
+      "loss": 1.41582022,
+      "memory(GiB)": 103.55,
+      "step": 10085,
+      "train_speed(iter/s)": 1.312269
+    },
+    {
+      "acc": 0.67887468,
+      "epoch": 0.25596144089294776,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.83466722706913e-06,
+      "loss": 1.44489422,
+      "memory(GiB)": 103.55,
+      "step": 10090,
+      "train_speed(iter/s)": 1.312349
+    },
+    {
+      "acc": 0.68086691,
+      "epoch": 0.2560882800608828,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.834399691459433e-06,
+      "loss": 1.44331322,
+      "memory(GiB)": 103.55,
+      "step": 10095,
+      "train_speed(iter/s)": 1.31243
+    },
+    {
+      "acc": 0.67431726,
+      "epoch": 0.25621511922881784,
+      "grad_norm": 3.125,
+      "learning_rate": 9.83413194321125e-06,
+      "loss": 1.41821041,
+      "memory(GiB)": 103.55,
+      "step": 10100,
+      "train_speed(iter/s)": 1.312509
+    },
+    {
+      "acc": 0.69339409,
+      "epoch": 0.25634195839675294,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.833863982336365e-06,
+      "loss": 1.37829542,
+      "memory(GiB)": 103.55,
+      "step": 10105,
+      "train_speed(iter/s)": 1.312585
+    },
+    {
+      "acc": 0.69121032,
+      "epoch": 0.256468797564688,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.83359580884656e-06,
+      "loss": 1.42013922,
+      "memory(GiB)": 103.55,
+      "step": 10110,
+      "train_speed(iter/s)": 1.312667
+    },
+    {
+      "acc": 0.69023695,
+      "epoch": 0.25659563673262303,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.83332742275363e-06,
+      "loss": 1.41327734,
+      "memory(GiB)": 103.55,
+      "step": 10115,
+      "train_speed(iter/s)": 1.31271
+    },
+    {
+      "acc": 0.66245141,
+      "epoch": 0.2567224759005581,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.833058824069382e-06,
+      "loss": 1.47697105,
+      "memory(GiB)": 103.55,
+      "step": 10120,
+      "train_speed(iter/s)": 1.312792
+    },
+    {
+      "acc": 0.68747487,
+      "epoch": 0.2568493150684932,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.832790012805626e-06,
+      "loss": 1.38594255,
+      "memory(GiB)": 103.55,
+      "step": 10125,
+      "train_speed(iter/s)": 1.31287
+    },
+    {
+      "acc": 0.68339472,
+      "epoch": 0.2569761542364282,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.832520988974191e-06,
+      "loss": 1.36582165,
+      "memory(GiB)": 103.55,
+      "step": 10130,
+      "train_speed(iter/s)": 1.31295
+    },
+    {
+      "acc": 0.6913013,
+      "epoch": 0.25710299340436327,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.832251752586907e-06,
+      "loss": 1.36766233,
+      "memory(GiB)": 103.55,
+      "step": 10135,
+      "train_speed(iter/s)": 1.313025
+    },
+    {
+      "acc": 0.68531375,
+      "epoch": 0.2572298325722983,
+      "grad_norm": 2.625,
+      "learning_rate": 9.831982303655617e-06,
+      "loss": 1.35796518,
+      "memory(GiB)": 103.55,
+      "step": 10140,
+      "train_speed(iter/s)": 1.313105
+    },
+    {
+      "acc": 0.68350234,
+      "epoch": 0.2573566717402334,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.83171264219217e-06,
+      "loss": 1.40313511,
+      "memory(GiB)": 103.55,
+      "step": 10145,
+      "train_speed(iter/s)": 1.313186
+    },
+    {
+      "acc": 0.70655327,
+      "epoch": 0.25748351090816846,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.831442768208429e-06,
+      "loss": 1.26845789,
+      "memory(GiB)": 103.55,
+      "step": 10150,
+      "train_speed(iter/s)": 1.313256
+    },
+    {
+      "acc": 0.70020533,
+      "epoch": 0.2576103500761035,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.831172681716265e-06,
+      "loss": 1.38891201,
+      "memory(GiB)": 103.55,
+      "step": 10155,
+      "train_speed(iter/s)": 1.313333
+    },
+    {
+      "acc": 0.69886508,
+      "epoch": 0.25773718924403854,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.830902382727556e-06,
+      "loss": 1.38369246,
+      "memory(GiB)": 103.55,
+      "step": 10160,
+      "train_speed(iter/s)": 1.313411
+    },
+    {
+      "acc": 0.69333248,
+      "epoch": 0.25786402841197364,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.830631871254193e-06,
+      "loss": 1.36332369,
+      "memory(GiB)": 103.55,
+      "step": 10165,
+      "train_speed(iter/s)": 1.31349
+    },
+    {
+      "acc": 0.6886106,
+      "epoch": 0.2579908675799087,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.830361147308074e-06,
+      "loss": 1.39977722,
+      "memory(GiB)": 103.55,
+      "step": 10170,
+      "train_speed(iter/s)": 1.313569
+    },
+    {
+      "acc": 0.67446971,
+      "epoch": 0.25811770674784373,
+      "grad_norm": 2.625,
+      "learning_rate": 9.830090210901104e-06,
+      "loss": 1.43945427,
+      "memory(GiB)": 103.55,
+      "step": 10175,
+      "train_speed(iter/s)": 1.31365
+    },
+    {
+      "acc": 0.67462883,
+      "epoch": 0.2582445459157788,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.829819062045203e-06,
+      "loss": 1.43966274,
+      "memory(GiB)": 103.55,
+      "step": 10180,
+      "train_speed(iter/s)": 1.313729
+    },
+    {
+      "acc": 0.69995565,
+      "epoch": 0.2583713850837139,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.829547700752295e-06,
+      "loss": 1.32462997,
+      "memory(GiB)": 103.55,
+      "step": 10185,
+      "train_speed(iter/s)": 1.31381
+    },
+    {
+      "acc": 0.66583385,
+      "epoch": 0.2584982242516489,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.829276127034315e-06,
+      "loss": 1.48206329,
+      "memory(GiB)": 103.55,
+      "step": 10190,
+      "train_speed(iter/s)": 1.313887
+    },
+    {
+      "acc": 0.66647844,
+      "epoch": 0.25862506341958397,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.829004340903214e-06,
+      "loss": 1.44538364,
+      "memory(GiB)": 103.55,
+      "step": 10195,
+      "train_speed(iter/s)": 1.313965
+    },
+    {
+      "acc": 0.67575207,
+      "epoch": 0.258751902587519,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.82873234237094e-06,
+      "loss": 1.48054981,
+      "memory(GiB)": 103.55,
+      "step": 10200,
+      "train_speed(iter/s)": 1.314044
+    },
+    {
+      "acc": 0.66673379,
+      "epoch": 0.2588787417554541,
+      "grad_norm": 4.125,
+      "learning_rate": 9.828460131449457e-06,
+      "loss": 1.49738445,
+      "memory(GiB)": 103.55,
+      "step": 10205,
+      "train_speed(iter/s)": 1.314122
+    },
+    {
+      "acc": 0.66584005,
+      "epoch": 0.25900558092338916,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.828187708150743e-06,
+      "loss": 1.4471199,
+      "memory(GiB)": 103.55,
+      "step": 10210,
+      "train_speed(iter/s)": 1.3142
+    },
+    {
+      "acc": 0.69438505,
+      "epoch": 0.2591324200913242,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.827915072486776e-06,
+      "loss": 1.37052517,
+      "memory(GiB)": 103.55,
+      "step": 10215,
+      "train_speed(iter/s)": 1.314279
+    },
+    {
+      "acc": 0.68566608,
+      "epoch": 0.25925925925925924,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.827642224469547e-06,
+      "loss": 1.40551538,
+      "memory(GiB)": 103.55,
+      "step": 10220,
+      "train_speed(iter/s)": 1.31434
+    },
+    {
+      "acc": 0.67547159,
+      "epoch": 0.25938609842719434,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.827369164111062e-06,
+      "loss": 1.48625317,
+      "memory(GiB)": 103.55,
+      "step": 10225,
+      "train_speed(iter/s)": 1.31442
+    },
+    {
+      "acc": 0.67358875,
+      "epoch": 0.2595129375951294,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.827095891423328e-06,
+      "loss": 1.49376764,
+      "memory(GiB)": 103.55,
+      "step": 10230,
+      "train_speed(iter/s)": 1.314499
+    },
+    {
+      "acc": 0.70116043,
+      "epoch": 0.25963977676306443,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.826822406418366e-06,
+      "loss": 1.31537075,
+      "memory(GiB)": 103.55,
+      "step": 10235,
+      "train_speed(iter/s)": 1.314575
+    },
+    {
+      "acc": 0.69151459,
+      "epoch": 0.2597666159309995,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.826548709108202e-06,
+      "loss": 1.33353615,
+      "memory(GiB)": 103.55,
+      "step": 10240,
+      "train_speed(iter/s)": 1.314653
+    },
+    {
+      "acc": 0.67947164,
+      "epoch": 0.2598934550989346,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.826274799504878e-06,
+      "loss": 1.3836772,
+      "memory(GiB)": 103.55,
+      "step": 10245,
+      "train_speed(iter/s)": 1.31473
+    },
+    {
+      "acc": 0.68525,
+      "epoch": 0.2600202942668696,
+      "grad_norm": 2.75,
+      "learning_rate": 9.82600067762044e-06,
+      "loss": 1.42802114,
+      "memory(GiB)": 103.55,
+      "step": 10250,
+      "train_speed(iter/s)": 1.314809
+    },
+    {
+      "acc": 0.68564482,
+      "epoch": 0.26014713343480467,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.825726343466947e-06,
+      "loss": 1.41030855,
+      "memory(GiB)": 103.55,
+      "step": 10255,
+      "train_speed(iter/s)": 1.314883
+    },
+    {
+      "acc": 0.68253098,
+      "epoch": 0.2602739726027397,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.825451797056462e-06,
+      "loss": 1.4113308,
+      "memory(GiB)": 103.55,
+      "step": 10260,
+      "train_speed(iter/s)": 1.31496
+    },
+    {
+      "acc": 0.67882051,
+      "epoch": 0.2604008117706748,
+      "grad_norm": 3.375,
+      "learning_rate": 9.825177038401064e-06,
+      "loss": 1.41874294,
+      "memory(GiB)": 103.55,
+      "step": 10265,
+      "train_speed(iter/s)": 1.315037
+    },
+    {
+      "acc": 0.67190042,
+      "epoch": 0.26052765093860986,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.824902067512838e-06,
+      "loss": 1.45532761,
+      "memory(GiB)": 103.55,
+      "step": 10270,
+      "train_speed(iter/s)": 1.315115
+    },
+    {
+      "acc": 0.68848152,
+      "epoch": 0.2606544901065449,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.824626884403877e-06,
+      "loss": 1.42056484,
+      "memory(GiB)": 103.55,
+      "step": 10275,
+      "train_speed(iter/s)": 1.315194
+    },
+    {
+      "acc": 0.67909842,
+      "epoch": 0.26078132927447994,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.824351489086283e-06,
+      "loss": 1.40245285,
+      "memory(GiB)": 103.55,
+      "step": 10280,
+      "train_speed(iter/s)": 1.315264
+    },
+    {
+      "acc": 0.69471407,
+      "epoch": 0.26090816844241504,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.824075881572176e-06,
+      "loss": 1.41281462,
+      "memory(GiB)": 103.55,
+      "step": 10285,
+      "train_speed(iter/s)": 1.315342
+    },
+    {
+      "acc": 0.69321446,
+      "epoch": 0.2610350076103501,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.823800061873669e-06,
+      "loss": 1.42319336,
+      "memory(GiB)": 103.55,
+      "step": 10290,
+      "train_speed(iter/s)": 1.315418
+    },
+    {
+      "acc": 0.68218646,
+      "epoch": 0.26116184677828513,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.8235240300029e-06,
+      "loss": 1.45947895,
+      "memory(GiB)": 103.55,
+      "step": 10295,
+      "train_speed(iter/s)": 1.315496
+    },
+    {
+      "acc": 0.67824612,
+      "epoch": 0.2612886859462202,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.82324778597201e-06,
+      "loss": 1.41673851,
+      "memory(GiB)": 103.55,
+      "step": 10300,
+      "train_speed(iter/s)": 1.315573
+    },
+    {
+      "acc": 0.67704663,
+      "epoch": 0.2614155251141553,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.822971329793147e-06,
+      "loss": 1.4401618,
+      "memory(GiB)": 103.55,
+      "step": 10305,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.697715,
+      "epoch": 0.2615423642820903,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.822694661478471e-06,
+      "loss": 1.33578329,
+      "memory(GiB)": 103.55,
+      "step": 10310,
+      "train_speed(iter/s)": 1.315724
+    },
+    {
+      "acc": 0.6750021,
+      "epoch": 0.26166920345002537,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.822417781040154e-06,
+      "loss": 1.45309258,
+      "memory(GiB)": 103.55,
+      "step": 10315,
+      "train_speed(iter/s)": 1.315804
+    },
+    {
+      "acc": 0.67167993,
+      "epoch": 0.2617960426179604,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.822140688490372e-06,
+      "loss": 1.43638783,
+      "memory(GiB)": 103.55,
+      "step": 10320,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.68546028,
+      "epoch": 0.2619228817858955,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.821863383841312e-06,
+      "loss": 1.4073637,
+      "memory(GiB)": 103.55,
+      "step": 10325,
+      "train_speed(iter/s)": 1.315956
+    },
+    {
+      "acc": 0.69717789,
+      "epoch": 0.26204972095383056,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.821585867105173e-06,
+      "loss": 1.41012011,
+      "memory(GiB)": 103.55,
+      "step": 10330,
+      "train_speed(iter/s)": 1.316033
+    },
+    {
+      "acc": 0.67126184,
+      "epoch": 0.2621765601217656,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.821308138294162e-06,
+      "loss": 1.45890503,
+      "memory(GiB)": 103.55,
+      "step": 10335,
+      "train_speed(iter/s)": 1.316109
+    },
+    {
+      "acc": 0.68111258,
+      "epoch": 0.26230339928970065,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.821030197420492e-06,
+      "loss": 1.47048702,
+      "memory(GiB)": 103.55,
+      "step": 10340,
+      "train_speed(iter/s)": 1.316186
+    },
+    {
+      "acc": 0.68122787,
+      "epoch": 0.26243023845763574,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.820752044496389e-06,
+      "loss": 1.40793858,
+      "memory(GiB)": 103.55,
+      "step": 10345,
+      "train_speed(iter/s)": 1.316258
+    },
+    {
+      "acc": 0.67648878,
+      "epoch": 0.2625570776255708,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.82047367953409e-06,
+      "loss": 1.36804552,
+      "memory(GiB)": 103.55,
+      "step": 10350,
+      "train_speed(iter/s)": 1.31633
+    },
+    {
+      "acc": 0.69531832,
+      "epoch": 0.26268391679350583,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.820195102545835e-06,
+      "loss": 1.37648659,
+      "memory(GiB)": 103.55,
+      "step": 10355,
+      "train_speed(iter/s)": 1.316403
+    },
+    {
+      "acc": 0.65743847,
+      "epoch": 0.2628107559614409,
+      "grad_norm": 2.625,
+      "learning_rate": 9.81991631354388e-06,
+      "loss": 1.42561026,
+      "memory(GiB)": 103.55,
+      "step": 10360,
+      "train_speed(iter/s)": 1.316482
+    },
+    {
+      "acc": 0.6708952,
+      "epoch": 0.262937595129376,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.819637312540485e-06,
+      "loss": 1.49082355,
+      "memory(GiB)": 103.55,
+      "step": 10365,
+      "train_speed(iter/s)": 1.316558
+    },
+    {
+      "acc": 0.67878151,
+      "epoch": 0.263064434297311,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.819358099547923e-06,
+      "loss": 1.38612251,
+      "memory(GiB)": 103.55,
+      "step": 10370,
+      "train_speed(iter/s)": 1.316628
+    },
+    {
+      "acc": 0.66217608,
+      "epoch": 0.26319127346524607,
+      "grad_norm": 2.875,
+      "learning_rate": 9.819078674578474e-06,
+      "loss": 1.51770277,
+      "memory(GiB)": 103.55,
+      "step": 10375,
+      "train_speed(iter/s)": 1.316704
+    },
+    {
+      "acc": 0.66839733,
+      "epoch": 0.2633181126331811,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.818799037644432e-06,
+      "loss": 1.50183849,
+      "memory(GiB)": 103.55,
+      "step": 10380,
+      "train_speed(iter/s)": 1.316783
+    },
+    {
+      "acc": 0.6818121,
+      "epoch": 0.2634449518011162,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.818519188758092e-06,
+      "loss": 1.42590618,
+      "memory(GiB)": 103.55,
+      "step": 10385,
+      "train_speed(iter/s)": 1.316858
+    },
+    {
+      "acc": 0.67782626,
+      "epoch": 0.26357179096905126,
+      "grad_norm": 2.875,
+      "learning_rate": 9.818239127931765e-06,
+      "loss": 1.42514534,
+      "memory(GiB)": 103.55,
+      "step": 10390,
+      "train_speed(iter/s)": 1.316934
+    },
+    {
+      "acc": 0.67301917,
+      "epoch": 0.2636986301369863,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.81795885517777e-06,
+      "loss": 1.5003479,
+      "memory(GiB)": 103.55,
+      "step": 10395,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.68467102,
+      "epoch": 0.26382546930492135,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.817678370508434e-06,
+      "loss": 1.45326252,
+      "memory(GiB)": 103.55,
+      "step": 10400,
+      "train_speed(iter/s)": 1.317086
+    },
+    {
+      "acc": 0.6717658,
+      "epoch": 0.26395230847285645,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.817397673936093e-06,
+      "loss": 1.46253147,
+      "memory(GiB)": 103.55,
+      "step": 10405,
+      "train_speed(iter/s)": 1.317163
+    },
+    {
+      "acc": 0.66947246,
+      "epoch": 0.2640791476407915,
+      "grad_norm": 3.375,
+      "learning_rate": 9.817116765473095e-06,
+      "loss": 1.49295425,
+      "memory(GiB)": 103.55,
+      "step": 10410,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.68336287,
+      "epoch": 0.26420598680872653,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.816835645131795e-06,
+      "loss": 1.37493515,
+      "memory(GiB)": 103.55,
+      "step": 10415,
+      "train_speed(iter/s)": 1.317317
+    },
+    {
+      "acc": 0.69205961,
+      "epoch": 0.2643328259766616,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.816554312924555e-06,
+      "loss": 1.39892883,
+      "memory(GiB)": 103.55,
+      "step": 10420,
+      "train_speed(iter/s)": 1.317394
+    },
+    {
+      "acc": 0.69153109,
+      "epoch": 0.2644596651445967,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.816272768863756e-06,
+      "loss": 1.40556793,
+      "memory(GiB)": 103.55,
+      "step": 10425,
+      "train_speed(iter/s)": 1.317469
+    },
+    {
+      "acc": 0.68840899,
+      "epoch": 0.2645865043125317,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.815991012961773e-06,
+      "loss": 1.39226561,
+      "memory(GiB)": 103.55,
+      "step": 10430,
+      "train_speed(iter/s)": 1.31754
+    },
+    {
+      "acc": 0.68502665,
+      "epoch": 0.26471334348046677,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.815709045231008e-06,
+      "loss": 1.47720089,
+      "memory(GiB)": 103.55,
+      "step": 10435,
+      "train_speed(iter/s)": 1.317618
+    },
+    {
+      "acc": 0.67319384,
+      "epoch": 0.2648401826484018,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.815426865683858e-06,
+      "loss": 1.49918756,
+      "memory(GiB)": 103.55,
+      "step": 10440,
+      "train_speed(iter/s)": 1.317694
+    },
+    {
+      "acc": 0.69266262,
+      "epoch": 0.2649670218163369,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.815144474332732e-06,
+      "loss": 1.36381454,
+      "memory(GiB)": 103.55,
+      "step": 10445,
+      "train_speed(iter/s)": 1.317768
+    },
+    {
+      "acc": 0.69266424,
+      "epoch": 0.26509386098427196,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.814861871190056e-06,
+      "loss": 1.38476982,
+      "memory(GiB)": 103.55,
+      "step": 10450,
+      "train_speed(iter/s)": 1.317846
+    },
+    {
+      "acc": 0.68796511,
+      "epoch": 0.265220700152207,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.814579056268256e-06,
+      "loss": 1.35146389,
+      "memory(GiB)": 103.55,
+      "step": 10455,
+      "train_speed(iter/s)": 1.317922
+    },
+    {
+      "acc": 0.68783612,
+      "epoch": 0.26534753932014205,
+      "grad_norm": 3.0,
+      "learning_rate": 9.814296029579776e-06,
+      "loss": 1.41145144,
+      "memory(GiB)": 103.55,
+      "step": 10460,
+      "train_speed(iter/s)": 1.317997
+    },
+    {
+      "acc": 0.66472998,
+      "epoch": 0.26547437848807715,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.814012791137063e-06,
+      "loss": 1.50025558,
+      "memory(GiB)": 103.55,
+      "step": 10465,
+      "train_speed(iter/s)": 1.318073
+    },
+    {
+      "acc": 0.67352219,
+      "epoch": 0.2656012176560122,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.81372934095257e-06,
+      "loss": 1.45615005,
+      "memory(GiB)": 103.55,
+      "step": 10470,
+      "train_speed(iter/s)": 1.318139
+    },
+    {
+      "acc": 0.6632844,
+      "epoch": 0.26572805682394723,
+      "grad_norm": 5.6875,
+      "learning_rate": 9.813445679038773e-06,
+      "loss": 1.47928085,
+      "memory(GiB)": 103.55,
+      "step": 10475,
+      "train_speed(iter/s)": 1.318216
+    },
+    {
+      "acc": 0.67069707,
+      "epoch": 0.2658548959918823,
+      "grad_norm": 5.40625,
+      "learning_rate": 9.813161805408145e-06,
+      "loss": 1.47584534,
+      "memory(GiB)": 103.55,
+      "step": 10480,
+      "train_speed(iter/s)": 1.318289
+    },
+    {
+      "acc": 0.687081,
+      "epoch": 0.2659817351598174,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.812877720073169e-06,
+      "loss": 1.44086246,
+      "memory(GiB)": 103.55,
+      "step": 10485,
+      "train_speed(iter/s)": 1.318363
+    },
+    {
+      "acc": 0.67575006,
+      "epoch": 0.2661085743277524,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.812593423046344e-06,
+      "loss": 1.46342678,
+      "memory(GiB)": 103.55,
+      "step": 10490,
+      "train_speed(iter/s)": 1.31844
+    },
+    {
+      "acc": 0.66907291,
+      "epoch": 0.26623541349568747,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.812308914340174e-06,
+      "loss": 1.52630472,
+      "memory(GiB)": 103.55,
+      "step": 10495,
+      "train_speed(iter/s)": 1.318514
+    },
+    {
+      "acc": 0.68357472,
+      "epoch": 0.2663622526636225,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.812024193967171e-06,
+      "loss": 1.47609901,
+      "memory(GiB)": 103.55,
+      "step": 10500,
+      "train_speed(iter/s)": 1.318589
+    },
+    {
+      "acc": 0.67935219,
+      "epoch": 0.2664890918315576,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.811739261939861e-06,
+      "loss": 1.46302776,
+      "memory(GiB)": 103.55,
+      "step": 10505,
+      "train_speed(iter/s)": 1.318662
+    },
+    {
+      "acc": 0.68028259,
+      "epoch": 0.26661593099949266,
+      "grad_norm": 2.5,
+      "learning_rate": 9.811454118270775e-06,
+      "loss": 1.45839033,
+      "memory(GiB)": 103.55,
+      "step": 10510,
+      "train_speed(iter/s)": 1.318734
+    },
+    {
+      "acc": 0.66906366,
+      "epoch": 0.2667427701674277,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.811168762972457e-06,
+      "loss": 1.43841295,
+      "memory(GiB)": 103.55,
+      "step": 10515,
+      "train_speed(iter/s)": 1.318808
+    },
+    {
+      "acc": 0.67588387,
+      "epoch": 0.26686960933536275,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.810883196057454e-06,
+      "loss": 1.456917,
+      "memory(GiB)": 103.55,
+      "step": 10520,
+      "train_speed(iter/s)": 1.318879
+    },
+    {
+      "acc": 0.6836082,
+      "epoch": 0.26699644850329785,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.81059741753833e-06,
+      "loss": 1.39072695,
+      "memory(GiB)": 103.55,
+      "step": 10525,
+      "train_speed(iter/s)": 1.318954
+    },
+    {
+      "acc": 0.68261046,
+      "epoch": 0.2671232876712329,
+      "grad_norm": 3.0,
+      "learning_rate": 9.810311427427653e-06,
+      "loss": 1.4329565,
+      "memory(GiB)": 103.55,
+      "step": 10530,
+      "train_speed(iter/s)": 1.319028
+    },
+    {
+      "acc": 0.69176664,
+      "epoch": 0.26725012683916793,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.810025225738005e-06,
+      "loss": 1.42812347,
+      "memory(GiB)": 103.55,
+      "step": 10535,
+      "train_speed(iter/s)": 1.3191
+    },
+    {
+      "acc": 0.64608707,
+      "epoch": 0.267376966007103,
+      "grad_norm": 3.875,
+      "learning_rate": 9.809738812481971e-06,
+      "loss": 1.5295352,
+      "memory(GiB)": 103.55,
+      "step": 10540,
+      "train_speed(iter/s)": 1.319171
+    },
+    {
+      "acc": 0.67941213,
+      "epoch": 0.2675038051750381,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.809452187672149e-06,
+      "loss": 1.42042627,
+      "memory(GiB)": 103.55,
+      "step": 10545,
+      "train_speed(iter/s)": 1.319242
+    },
+    {
+      "acc": 0.67114878,
+      "epoch": 0.2676306443429731,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.809165351321149e-06,
+      "loss": 1.45718727,
+      "memory(GiB)": 103.55,
+      "step": 10550,
+      "train_speed(iter/s)": 1.319315
+    },
+    {
+      "acc": 0.69472952,
+      "epoch": 0.26775748351090817,
+      "grad_norm": 3.625,
+      "learning_rate": 9.808878303441585e-06,
+      "loss": 1.38773308,
+      "memory(GiB)": 103.55,
+      "step": 10555,
+      "train_speed(iter/s)": 1.319388
+    },
+    {
+      "acc": 0.68277321,
+      "epoch": 0.2678843226788432,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.808591044046083e-06,
+      "loss": 1.43276272,
+      "memory(GiB)": 103.55,
+      "step": 10560,
+      "train_speed(iter/s)": 1.319464
+    },
+    {
+      "acc": 0.67614956,
+      "epoch": 0.2680111618467783,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.808303573147277e-06,
+      "loss": 1.43820133,
+      "memory(GiB)": 103.55,
+      "step": 10565,
+      "train_speed(iter/s)": 1.319521
+    },
+    {
+      "acc": 0.69080157,
+      "epoch": 0.26813800101471336,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.808015890757812e-06,
+      "loss": 1.4251543,
+      "memory(GiB)": 103.55,
+      "step": 10570,
+      "train_speed(iter/s)": 1.319593
+    },
+    {
+      "acc": 0.68801842,
+      "epoch": 0.2682648401826484,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.807727996890343e-06,
+      "loss": 1.39425297,
+      "memory(GiB)": 103.55,
+      "step": 10575,
+      "train_speed(iter/s)": 1.319665
+    },
+    {
+      "acc": 0.68143206,
+      "epoch": 0.26839167935058345,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.807439891557533e-06,
+      "loss": 1.42957363,
+      "memory(GiB)": 103.55,
+      "step": 10580,
+      "train_speed(iter/s)": 1.319739
+    },
+    {
+      "acc": 0.6762557,
+      "epoch": 0.26851851851851855,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.80715157477205e-06,
+      "loss": 1.47412567,
+      "memory(GiB)": 103.55,
+      "step": 10585,
+      "train_speed(iter/s)": 1.319809
+    },
+    {
+      "acc": 0.67838993,
+      "epoch": 0.2686453576864536,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.806863046546581e-06,
+      "loss": 1.47360268,
+      "memory(GiB)": 103.55,
+      "step": 10590,
+      "train_speed(iter/s)": 1.319883
+    },
+    {
+      "acc": 0.67921047,
+      "epoch": 0.26877219685438863,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.806574306893814e-06,
+      "loss": 1.4365839,
+      "memory(GiB)": 103.55,
+      "step": 10595,
+      "train_speed(iter/s)": 1.319955
+    },
+    {
+      "acc": 0.68314381,
+      "epoch": 0.2688990360223237,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.806285355826447e-06,
+      "loss": 1.40715923,
+      "memory(GiB)": 103.55,
+      "step": 10600,
+      "train_speed(iter/s)": 1.320029
+    },
+    {
+      "acc": 0.69741211,
+      "epoch": 0.2690258751902588,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.805996193357194e-06,
+      "loss": 1.35298176,
+      "memory(GiB)": 103.55,
+      "step": 10605,
+      "train_speed(iter/s)": 1.320106
+    },
+    {
+      "acc": 0.665938,
+      "epoch": 0.2691527143581938,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.80570681949877e-06,
+      "loss": 1.45799236,
+      "memory(GiB)": 103.55,
+      "step": 10610,
+      "train_speed(iter/s)": 1.320179
+    },
+    {
+      "acc": 0.67439685,
+      "epoch": 0.26927955352612887,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.805417234263905e-06,
+      "loss": 1.43799706,
+      "memory(GiB)": 103.55,
+      "step": 10615,
+      "train_speed(iter/s)": 1.320251
+    },
+    {
+      "acc": 0.67354231,
+      "epoch": 0.2694063926940639,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.805127437665333e-06,
+      "loss": 1.51822777,
+      "memory(GiB)": 103.55,
+      "step": 10620,
+      "train_speed(iter/s)": 1.320324
+    },
+    {
+      "acc": 0.67443228,
+      "epoch": 0.269533231861999,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.804837429715805e-06,
+      "loss": 1.46038713,
+      "memory(GiB)": 103.55,
+      "step": 10625,
+      "train_speed(iter/s)": 1.320397
+    },
+    {
+      "acc": 0.69694033,
+      "epoch": 0.26966007102993406,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.804547210428074e-06,
+      "loss": 1.4049901,
+      "memory(GiB)": 103.55,
+      "step": 10630,
+      "train_speed(iter/s)": 1.320469
+    },
+    {
+      "acc": 0.67513342,
+      "epoch": 0.2697869101978691,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.804256779814906e-06,
+      "loss": 1.45858765,
+      "memory(GiB)": 103.55,
+      "step": 10635,
+      "train_speed(iter/s)": 1.320542
+    },
+    {
+      "acc": 0.68266277,
+      "epoch": 0.26991374936580415,
+      "grad_norm": 3.0,
+      "learning_rate": 9.803966137889076e-06,
+      "loss": 1.38912315,
+      "memory(GiB)": 103.55,
+      "step": 10640,
+      "train_speed(iter/s)": 1.320608
+    },
+    {
+      "acc": 0.66295538,
+      "epoch": 0.27004058853373925,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.803675284663368e-06,
+      "loss": 1.51194782,
+      "memory(GiB)": 103.55,
+      "step": 10645,
+      "train_speed(iter/s)": 1.32068
+    },
+    {
+      "acc": 0.68201218,
+      "epoch": 0.2701674277016743,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.803384220150571e-06,
+      "loss": 1.45984364,
+      "memory(GiB)": 103.55,
+      "step": 10650,
+      "train_speed(iter/s)": 1.320755
+    },
+    {
+      "acc": 0.67419419,
+      "epoch": 0.27029426686960933,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.803092944363493e-06,
+      "loss": 1.48058081,
+      "memory(GiB)": 103.55,
+      "step": 10655,
+      "train_speed(iter/s)": 1.320829
+    },
+    {
+      "acc": 0.69287257,
+      "epoch": 0.2704211060375444,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.802801457314943e-06,
+      "loss": 1.39590349,
+      "memory(GiB)": 103.55,
+      "step": 10660,
+      "train_speed(iter/s)": 1.320903
+    },
+    {
+      "acc": 0.68714161,
+      "epoch": 0.2705479452054795,
+      "grad_norm": 4.25,
+      "learning_rate": 9.802509759017741e-06,
+      "loss": 1.39016762,
+      "memory(GiB)": 103.55,
+      "step": 10665,
+      "train_speed(iter/s)": 1.320975
+    },
+    {
+      "acc": 0.66668439,
+      "epoch": 0.2706747843734145,
+      "grad_norm": 2.25,
+      "learning_rate": 9.802217849484719e-06,
+      "loss": 1.51326809,
+      "memory(GiB)": 103.55,
+      "step": 10670,
+      "train_speed(iter/s)": 1.321049
+    },
+    {
+      "acc": 0.68189898,
+      "epoch": 0.27080162354134957,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.801925728728715e-06,
+      "loss": 1.4209486,
+      "memory(GiB)": 103.55,
+      "step": 10675,
+      "train_speed(iter/s)": 1.321124
+    },
+    {
+      "acc": 0.68306289,
+      "epoch": 0.2709284627092846,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.801633396762577e-06,
+      "loss": 1.43766432,
+      "memory(GiB)": 103.55,
+      "step": 10680,
+      "train_speed(iter/s)": 1.321199
+    },
+    {
+      "acc": 0.68920431,
+      "epoch": 0.2710553018772197,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.801340853599167e-06,
+      "loss": 1.46258907,
+      "memory(GiB)": 103.55,
+      "step": 10685,
+      "train_speed(iter/s)": 1.321275
+    },
+    {
+      "acc": 0.6985549,
+      "epoch": 0.27118214104515476,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.801048099251348e-06,
+      "loss": 1.38358984,
+      "memory(GiB)": 103.55,
+      "step": 10690,
+      "train_speed(iter/s)": 1.321349
+    },
+    {
+      "acc": 0.68179975,
+      "epoch": 0.2713089802130898,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.800755133731999e-06,
+      "loss": 1.41958923,
+      "memory(GiB)": 103.55,
+      "step": 10695,
+      "train_speed(iter/s)": 1.321421
+    },
+    {
+      "acc": 0.68180556,
+      "epoch": 0.27143581938102485,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.800461957054006e-06,
+      "loss": 1.42467804,
+      "memory(GiB)": 103.55,
+      "step": 10700,
+      "train_speed(iter/s)": 1.321491
+    },
+    {
+      "acc": 0.68688617,
+      "epoch": 0.27156265854895995,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.800168569230261e-06,
+      "loss": 1.39949665,
+      "memory(GiB)": 103.55,
+      "step": 10705,
+      "train_speed(iter/s)": 1.321561
+    },
+    {
+      "acc": 0.68919754,
+      "epoch": 0.271689497716895,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.799874970273674e-06,
+      "loss": 1.37865839,
+      "memory(GiB)": 103.55,
+      "step": 10710,
+      "train_speed(iter/s)": 1.321632
+    },
+    {
+      "acc": 0.68276138,
+      "epoch": 0.27181633688483003,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.799581160197156e-06,
+      "loss": 1.42672291,
+      "memory(GiB)": 103.55,
+      "step": 10715,
+      "train_speed(iter/s)": 1.321704
+    },
+    {
+      "acc": 0.68935022,
+      "epoch": 0.2719431760527651,
+      "grad_norm": 2.5,
+      "learning_rate": 9.799287139013628e-06,
+      "loss": 1.37812777,
+      "memory(GiB)": 103.55,
+      "step": 10720,
+      "train_speed(iter/s)": 1.321774
+    },
+    {
+      "acc": 0.6884213,
+      "epoch": 0.2720700152207002,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.798992906736028e-06,
+      "loss": 1.3584671,
+      "memory(GiB)": 103.55,
+      "step": 10725,
+      "train_speed(iter/s)": 1.321845
+    },
+    {
+      "acc": 0.67818794,
+      "epoch": 0.2721968543886352,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.79869846337729e-06,
+      "loss": 1.44809561,
+      "memory(GiB)": 103.55,
+      "step": 10730,
+      "train_speed(iter/s)": 1.321913
+    },
+    {
+      "acc": 0.67672515,
+      "epoch": 0.27232369355657027,
+      "grad_norm": 3.5,
+      "learning_rate": 9.79840380895037e-06,
+      "loss": 1.43774204,
+      "memory(GiB)": 103.55,
+      "step": 10735,
+      "train_speed(iter/s)": 1.321972
+    },
+    {
+      "acc": 0.69053798,
+      "epoch": 0.2724505327245053,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.798108943468228e-06,
+      "loss": 1.33569412,
+      "memory(GiB)": 103.55,
+      "step": 10740,
+      "train_speed(iter/s)": 1.322043
+    },
+    {
+      "acc": 0.68036294,
+      "epoch": 0.2725773718924404,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.797813866943832e-06,
+      "loss": 1.38552723,
+      "memory(GiB)": 103.55,
+      "step": 10745,
+      "train_speed(iter/s)": 1.322111
+    },
+    {
+      "acc": 0.69211817,
+      "epoch": 0.27270421106037546,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.797518579390162e-06,
+      "loss": 1.31044626,
+      "memory(GiB)": 103.55,
+      "step": 10750,
+      "train_speed(iter/s)": 1.322181
+    },
+    {
+      "acc": 0.68979878,
+      "epoch": 0.2728310502283105,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.797223080820204e-06,
+      "loss": 1.40587425,
+      "memory(GiB)": 103.55,
+      "step": 10755,
+      "train_speed(iter/s)": 1.322248
+    },
+    {
+      "acc": 0.68244705,
+      "epoch": 0.27295788939624555,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.796927371246958e-06,
+      "loss": 1.40622587,
+      "memory(GiB)": 103.55,
+      "step": 10760,
+      "train_speed(iter/s)": 1.322318
+    },
+    {
+      "acc": 0.68367457,
+      "epoch": 0.27308472856418065,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.796631450683431e-06,
+      "loss": 1.41333399,
+      "memory(GiB)": 103.55,
+      "step": 10765,
+      "train_speed(iter/s)": 1.322387
+    },
+    {
+      "acc": 0.66530929,
+      "epoch": 0.2732115677321157,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.796335319142637e-06,
+      "loss": 1.41991405,
+      "memory(GiB)": 103.55,
+      "step": 10770,
+      "train_speed(iter/s)": 1.322457
+    },
+    {
+      "acc": 0.68509564,
+      "epoch": 0.27333840690005073,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.796038976637599e-06,
+      "loss": 1.41270676,
+      "memory(GiB)": 103.55,
+      "step": 10775,
+      "train_speed(iter/s)": 1.322523
+    },
+    {
+      "acc": 0.68615313,
+      "epoch": 0.2734652460679858,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.795742423181355e-06,
+      "loss": 1.42458286,
+      "memory(GiB)": 103.55,
+      "step": 10780,
+      "train_speed(iter/s)": 1.32259
+    },
+    {
+      "acc": 0.66389751,
+      "epoch": 0.2735920852359209,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.795445658786948e-06,
+      "loss": 1.53224258,
+      "memory(GiB)": 103.55,
+      "step": 10785,
+      "train_speed(iter/s)": 1.322659
+    },
+    {
+      "acc": 0.68079281,
+      "epoch": 0.2737189244038559,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.795148683467431e-06,
+      "loss": 1.43396139,
+      "memory(GiB)": 103.55,
+      "step": 10790,
+      "train_speed(iter/s)": 1.32273
+    },
+    {
+      "acc": 0.66622353,
+      "epoch": 0.27384576357179097,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.794851497235866e-06,
+      "loss": 1.45826206,
+      "memory(GiB)": 103.55,
+      "step": 10795,
+      "train_speed(iter/s)": 1.322797
+    },
+    {
+      "acc": 0.67817283,
+      "epoch": 0.273972602739726,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.794554100105325e-06,
+      "loss": 1.41341648,
+      "memory(GiB)": 103.55,
+      "step": 10800,
+      "train_speed(iter/s)": 1.322866
+    },
+    {
+      "acc": 0.69173326,
+      "epoch": 0.2740994419076611,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.794256492088888e-06,
+      "loss": 1.37746859,
+      "memory(GiB)": 103.55,
+      "step": 10805,
+      "train_speed(iter/s)": 1.322937
+    },
+    {
+      "acc": 0.68067408,
+      "epoch": 0.27422628107559616,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.793958673199647e-06,
+      "loss": 1.46386223,
+      "memory(GiB)": 103.55,
+      "step": 10810,
+      "train_speed(iter/s)": 1.323007
+    },
+    {
+      "acc": 0.68892317,
+      "epoch": 0.2743531202435312,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.793660643450697e-06,
+      "loss": 1.40558996,
+      "memory(GiB)": 103.55,
+      "step": 10815,
+      "train_speed(iter/s)": 1.323076
+    },
+    {
+      "acc": 0.67427673,
+      "epoch": 0.27447995941146625,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.793362402855152e-06,
+      "loss": 1.41780729,
+      "memory(GiB)": 103.55,
+      "step": 10820,
+      "train_speed(iter/s)": 1.32314
+    },
+    {
+      "acc": 0.67530608,
+      "epoch": 0.27460679857940135,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.79306395142613e-06,
+      "loss": 1.41832771,
+      "memory(GiB)": 103.55,
+      "step": 10825,
+      "train_speed(iter/s)": 1.323208
+    },
+    {
+      "acc": 0.68983593,
+      "epoch": 0.2747336377473364,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.792765289176751e-06,
+      "loss": 1.36875696,
+      "memory(GiB)": 103.55,
+      "step": 10830,
+      "train_speed(iter/s)": 1.323275
+    },
+    {
+      "acc": 0.67874999,
+      "epoch": 0.27486047691527143,
+      "grad_norm": 2.875,
+      "learning_rate": 9.79246641612016e-06,
+      "loss": 1.41995983,
+      "memory(GiB)": 103.55,
+      "step": 10835,
+      "train_speed(iter/s)": 1.323341
+    },
+    {
+      "acc": 0.67920384,
+      "epoch": 0.2749873160832065,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.792167332269498e-06,
+      "loss": 1.40397644,
+      "memory(GiB)": 103.55,
+      "step": 10840,
+      "train_speed(iter/s)": 1.323405
+    },
+    {
+      "acc": 0.69072719,
+      "epoch": 0.2751141552511416,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.791868037637922e-06,
+      "loss": 1.40719662,
+      "memory(GiB)": 103.55,
+      "step": 10845,
+      "train_speed(iter/s)": 1.323467
+    },
+    {
+      "acc": 0.67968473,
+      "epoch": 0.2752409944190766,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.791568532238594e-06,
+      "loss": 1.44329071,
+      "memory(GiB)": 103.55,
+      "step": 10850,
+      "train_speed(iter/s)": 1.32353
+    },
+    {
+      "acc": 0.68221407,
+      "epoch": 0.27536783358701167,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.79126881608469e-06,
+      "loss": 1.43378954,
+      "memory(GiB)": 103.55,
+      "step": 10855,
+      "train_speed(iter/s)": 1.323595
+    },
+    {
+      "acc": 0.68560085,
+      "epoch": 0.2754946727549467,
+      "grad_norm": 3.25,
+      "learning_rate": 9.790968889189392e-06,
+      "loss": 1.41876717,
+      "memory(GiB)": 103.55,
+      "step": 10860,
+      "train_speed(iter/s)": 1.323656
+    },
+    {
+      "acc": 0.6854229,
+      "epoch": 0.2756215119228818,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.790668751565893e-06,
+      "loss": 1.43450222,
+      "memory(GiB)": 103.55,
+      "step": 10865,
+      "train_speed(iter/s)": 1.323723
+    },
+    {
+      "acc": 0.68286095,
+      "epoch": 0.27574835109081686,
+      "grad_norm": 2.625,
+      "learning_rate": 9.790368403227391e-06,
+      "loss": 1.45778484,
+      "memory(GiB)": 103.55,
+      "step": 10870,
+      "train_speed(iter/s)": 1.323786
+    },
+    {
+      "acc": 0.67689152,
+      "epoch": 0.2758751902587519,
+      "grad_norm": 2.5,
+      "learning_rate": 9.7900678441871e-06,
+      "loss": 1.37830219,
+      "memory(GiB)": 103.55,
+      "step": 10875,
+      "train_speed(iter/s)": 1.323848
+    },
+    {
+      "acc": 0.67691593,
+      "epoch": 0.27600202942668695,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.78976707445824e-06,
+      "loss": 1.38832808,
+      "memory(GiB)": 103.55,
+      "step": 10880,
+      "train_speed(iter/s)": 1.323908
+    },
+    {
+      "acc": 0.68562036,
+      "epoch": 0.27612886859462205,
+      "grad_norm": 3.25,
+      "learning_rate": 9.78946609405404e-06,
+      "loss": 1.42718086,
+      "memory(GiB)": 103.55,
+      "step": 10885,
+      "train_speed(iter/s)": 1.323961
+    },
+    {
+      "acc": 0.69569707,
+      "epoch": 0.2762557077625571,
+      "grad_norm": 4.75,
+      "learning_rate": 9.789164902987738e-06,
+      "loss": 1.39215183,
+      "memory(GiB)": 103.55,
+      "step": 10890,
+      "train_speed(iter/s)": 1.324028
+    },
+    {
+      "acc": 0.69965954,
+      "epoch": 0.27638254693049213,
+      "grad_norm": 2.75,
+      "learning_rate": 9.78886350127258e-06,
+      "loss": 1.37610922,
+      "memory(GiB)": 103.58,
+      "step": 10895,
+      "train_speed(iter/s)": 1.324089
+    },
+    {
+      "acc": 0.67550907,
+      "epoch": 0.2765093860984272,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.788561888921825e-06,
+      "loss": 1.45530519,
+      "memory(GiB)": 103.58,
+      "step": 10900,
+      "train_speed(iter/s)": 1.324158
+    },
+    {
+      "acc": 0.68894062,
+      "epoch": 0.2766362252663623,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.788260065948738e-06,
+      "loss": 1.44272728,
+      "memory(GiB)": 103.58,
+      "step": 10905,
+      "train_speed(iter/s)": 1.324227
+    },
+    {
+      "acc": 0.66868515,
+      "epoch": 0.2767630644342973,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.787958032366596e-06,
+      "loss": 1.47879639,
+      "memory(GiB)": 103.58,
+      "step": 10910,
+      "train_speed(iter/s)": 1.32429
+    },
+    {
+      "acc": 0.68829417,
+      "epoch": 0.27688990360223237,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.787655788188684e-06,
+      "loss": 1.3655426,
+      "memory(GiB)": 103.58,
+      "step": 10915,
+      "train_speed(iter/s)": 1.324355
+    },
+    {
+      "acc": 0.69507675,
+      "epoch": 0.2770167427701674,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.787353333428293e-06,
+      "loss": 1.33330259,
+      "memory(GiB)": 103.58,
+      "step": 10920,
+      "train_speed(iter/s)": 1.324422
+    },
+    {
+      "acc": 0.67910442,
+      "epoch": 0.2771435819381025,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.78705066809873e-06,
+      "loss": 1.44372282,
+      "memory(GiB)": 103.58,
+      "step": 10925,
+      "train_speed(iter/s)": 1.324489
+    },
+    {
+      "acc": 0.66817064,
+      "epoch": 0.27727042110603756,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.786747792213304e-06,
+      "loss": 1.5027009,
+      "memory(GiB)": 103.58,
+      "step": 10930,
+      "train_speed(iter/s)": 1.324556
+    },
+    {
+      "acc": 0.69571829,
+      "epoch": 0.2773972602739726,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.78644470578534e-06,
+      "loss": 1.39268618,
+      "memory(GiB)": 103.58,
+      "step": 10935,
+      "train_speed(iter/s)": 1.324623
+    },
+    {
+      "acc": 0.66904836,
+      "epoch": 0.27752409944190765,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.78614140882817e-06,
+      "loss": 1.50770864,
+      "memory(GiB)": 103.58,
+      "step": 10940,
+      "train_speed(iter/s)": 1.324687
+    },
+    {
+      "acc": 0.69654884,
+      "epoch": 0.27765093860984275,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.78583790135513e-06,
+      "loss": 1.35862761,
+      "memory(GiB)": 103.58,
+      "step": 10945,
+      "train_speed(iter/s)": 1.32475
+    },
+    {
+      "acc": 0.68685131,
+      "epoch": 0.2777777777777778,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.785534183379571e-06,
+      "loss": 1.41011457,
+      "memory(GiB)": 103.58,
+      "step": 10950,
+      "train_speed(iter/s)": 1.324817
+    },
+    {
+      "acc": 0.67952662,
+      "epoch": 0.27790461694571283,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.785230254914855e-06,
+      "loss": 1.45626087,
+      "memory(GiB)": 103.58,
+      "step": 10955,
+      "train_speed(iter/s)": 1.324878
+    },
+    {
+      "acc": 0.67016811,
+      "epoch": 0.2780314561136479,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.784926115974346e-06,
+      "loss": 1.43742313,
+      "memory(GiB)": 103.58,
+      "step": 10960,
+      "train_speed(iter/s)": 1.324944
+    },
+    {
+      "acc": 0.69019384,
+      "epoch": 0.278158295281583,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.784621766571424e-06,
+      "loss": 1.42032375,
+      "memory(GiB)": 103.58,
+      "step": 10965,
+      "train_speed(iter/s)": 1.325008
+    },
+    {
+      "acc": 0.67785468,
+      "epoch": 0.278285134449518,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.784317206719475e-06,
+      "loss": 1.52022858,
+      "memory(GiB)": 103.58,
+      "step": 10970,
+      "train_speed(iter/s)": 1.325075
+    },
+    {
+      "acc": 0.67361813,
+      "epoch": 0.27841197361745307,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.784012436431896e-06,
+      "loss": 1.45668879,
+      "memory(GiB)": 103.58,
+      "step": 10975,
+      "train_speed(iter/s)": 1.32514
+    },
+    {
+      "acc": 0.68371215,
+      "epoch": 0.2785388127853881,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.78370745572209e-06,
+      "loss": 1.44674873,
+      "memory(GiB)": 103.58,
+      "step": 10980,
+      "train_speed(iter/s)": 1.325207
+    },
+    {
+      "acc": 0.67880764,
+      "epoch": 0.2786656519533232,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.783402264603471e-06,
+      "loss": 1.37799788,
+      "memory(GiB)": 103.58,
+      "step": 10985,
+      "train_speed(iter/s)": 1.325272
+    },
+    {
+      "acc": 0.67181635,
+      "epoch": 0.27879249112125826,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.783096863089465e-06,
+      "loss": 1.42931643,
+      "memory(GiB)": 103.58,
+      "step": 10990,
+      "train_speed(iter/s)": 1.325336
+    },
+    {
+      "acc": 0.70034018,
+      "epoch": 0.2789193302891933,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.782791251193505e-06,
+      "loss": 1.28506184,
+      "memory(GiB)": 103.58,
+      "step": 10995,
+      "train_speed(iter/s)": 1.325403
+    },
+    {
+      "acc": 0.68100429,
+      "epoch": 0.27904616945712835,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.782485428929032e-06,
+      "loss": 1.39873724,
+      "memory(GiB)": 103.58,
+      "step": 11000,
+      "train_speed(iter/s)": 1.32547
+    },
+    {
+      "epoch": 0.27904616945712835,
+      "eval_acc": 0.6700877297397762,
+      "eval_loss": 1.3843111991882324,
+      "eval_runtime": 70.4077,
+      "eval_samples_per_second": 90.473,
+      "eval_steps_per_second": 22.625,
+      "step": 11000
+    },
+    {
+      "acc": 0.69485216,
+      "epoch": 0.27917300862506345,
+      "grad_norm": 3.125,
+      "learning_rate": 9.782179396309496e-06,
+      "loss": 1.42331867,
+      "memory(GiB)": 103.58,
+      "step": 11005,
+      "train_speed(iter/s)": 1.312386
+    },
+    {
+      "acc": 0.66633325,
+      "epoch": 0.2792998477929985,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.78187315334836e-06,
+      "loss": 1.44804249,
+      "memory(GiB)": 103.58,
+      "step": 11010,
+      "train_speed(iter/s)": 1.312454
+    },
+    {
+      "acc": 0.68011818,
+      "epoch": 0.27942668696093353,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.781566700059094e-06,
+      "loss": 1.40870762,
+      "memory(GiB)": 103.58,
+      "step": 11015,
+      "train_speed(iter/s)": 1.312524
+    },
+    {
+      "acc": 0.67622747,
+      "epoch": 0.2795535261288686,
+      "grad_norm": 4.125,
+      "learning_rate": 9.781260036455176e-06,
+      "loss": 1.47030544,
+      "memory(GiB)": 103.58,
+      "step": 11020,
+      "train_speed(iter/s)": 1.312592
+    },
+    {
+      "acc": 0.69299622,
+      "epoch": 0.2796803652968037,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.780953162550093e-06,
+      "loss": 1.38226824,
+      "memory(GiB)": 103.58,
+      "step": 11025,
+      "train_speed(iter/s)": 1.312634
+    },
+    {
+      "acc": 0.67534447,
+      "epoch": 0.2798072044647387,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.780646078357346e-06,
+      "loss": 1.43549786,
+      "memory(GiB)": 103.58,
+      "step": 11030,
+      "train_speed(iter/s)": 1.312705
+    },
+    {
+      "acc": 0.68595991,
+      "epoch": 0.27993404363267377,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.78033878389044e-06,
+      "loss": 1.4638999,
+      "memory(GiB)": 103.58,
+      "step": 11035,
+      "train_speed(iter/s)": 1.312773
+    },
+    {
+      "acc": 0.68646374,
+      "epoch": 0.2800608828006088,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.780031279162892e-06,
+      "loss": 1.45880823,
+      "memory(GiB)": 103.58,
+      "step": 11040,
+      "train_speed(iter/s)": 1.312838
+    },
+    {
+      "acc": 0.69259882,
+      "epoch": 0.2801877219685439,
+      "grad_norm": 4.25,
+      "learning_rate": 9.779723564188228e-06,
+      "loss": 1.46240864,
+      "memory(GiB)": 103.58,
+      "step": 11045,
+      "train_speed(iter/s)": 1.312909
+    },
+    {
+      "acc": 0.67648673,
+      "epoch": 0.28031456113647896,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.77941563897998e-06,
+      "loss": 1.4562746,
+      "memory(GiB)": 103.58,
+      "step": 11050,
+      "train_speed(iter/s)": 1.312979
+    },
+    {
+      "acc": 0.68641949,
+      "epoch": 0.280441400304414,
+      "grad_norm": 2.875,
+      "learning_rate": 9.779107503551695e-06,
+      "loss": 1.44600925,
+      "memory(GiB)": 103.58,
+      "step": 11055,
+      "train_speed(iter/s)": 1.313047
+    },
+    {
+      "acc": 0.69191017,
+      "epoch": 0.28056823947234905,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.778799157916926e-06,
+      "loss": 1.39904728,
+      "memory(GiB)": 103.58,
+      "step": 11060,
+      "train_speed(iter/s)": 1.313115
+    },
+    {
+      "acc": 0.67351828,
+      "epoch": 0.28069507864028415,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.77849060208923e-06,
+      "loss": 1.43726311,
+      "memory(GiB)": 103.58,
+      "step": 11065,
+      "train_speed(iter/s)": 1.313169
+    },
+    {
+      "acc": 0.66636643,
+      "epoch": 0.2808219178082192,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.778181836082185e-06,
+      "loss": 1.4479847,
+      "memory(GiB)": 103.58,
+      "step": 11070,
+      "train_speed(iter/s)": 1.313237
+    },
+    {
+      "acc": 0.68871474,
+      "epoch": 0.28094875697615423,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.777872859909373e-06,
+      "loss": 1.4163847,
+      "memory(GiB)": 103.58,
+      "step": 11075,
+      "train_speed(iter/s)": 1.313306
+    },
+    {
+      "acc": 0.70154552,
+      "epoch": 0.2810755961440893,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.777563673584376e-06,
+      "loss": 1.33125267,
+      "memory(GiB)": 103.58,
+      "step": 11080,
+      "train_speed(iter/s)": 1.313375
+    },
+    {
+      "acc": 0.68206825,
+      "epoch": 0.2812024353120244,
+      "grad_norm": 3.125,
+      "learning_rate": 9.777254277120801e-06,
+      "loss": 1.36812725,
+      "memory(GiB)": 103.58,
+      "step": 11085,
+      "train_speed(iter/s)": 1.313442
+    },
+    {
+      "acc": 0.69708114,
+      "epoch": 0.2813292744799594,
+      "grad_norm": 2.75,
+      "learning_rate": 9.776944670532253e-06,
+      "loss": 1.38789063,
+      "memory(GiB)": 103.58,
+      "step": 11090,
+      "train_speed(iter/s)": 1.313502
+    },
+    {
+      "acc": 0.68893161,
+      "epoch": 0.28145611364789447,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.776634853832352e-06,
+      "loss": 1.38095322,
+      "memory(GiB)": 103.58,
+      "step": 11095,
+      "train_speed(iter/s)": 1.31357
+    },
+    {
+      "acc": 0.68040795,
+      "epoch": 0.2815829528158295,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.776324827034724e-06,
+      "loss": 1.4212574,
+      "memory(GiB)": 103.58,
+      "step": 11100,
+      "train_speed(iter/s)": 1.313637
+    },
+    {
+      "acc": 0.67813711,
+      "epoch": 0.2817097919837646,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.776014590153005e-06,
+      "loss": 1.4283165,
+      "memory(GiB)": 103.58,
+      "step": 11105,
+      "train_speed(iter/s)": 1.313705
+    },
+    {
+      "acc": 0.69041996,
+      "epoch": 0.28183663115169966,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.77570414320084e-06,
+      "loss": 1.36910191,
+      "memory(GiB)": 103.58,
+      "step": 11110,
+      "train_speed(iter/s)": 1.313774
+    },
+    {
+      "acc": 0.68832541,
+      "epoch": 0.2819634703196347,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.775393486191884e-06,
+      "loss": 1.38099108,
+      "memory(GiB)": 103.58,
+      "step": 11115,
+      "train_speed(iter/s)": 1.313843
+    },
+    {
+      "acc": 0.67196722,
+      "epoch": 0.28209030948756975,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.775082619139805e-06,
+      "loss": 1.44677963,
+      "memory(GiB)": 103.58,
+      "step": 11120,
+      "train_speed(iter/s)": 1.313914
+    },
+    {
+      "acc": 0.69682565,
+      "epoch": 0.28221714865550485,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.77477154205827e-06,
+      "loss": 1.37261791,
+      "memory(GiB)": 103.58,
+      "step": 11125,
+      "train_speed(iter/s)": 1.313985
+    },
+    {
+      "acc": 0.67407312,
+      "epoch": 0.2823439878234399,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.774460254960968e-06,
+      "loss": 1.4499239,
+      "memory(GiB)": 103.58,
+      "step": 11130,
+      "train_speed(iter/s)": 1.314054
+    },
+    {
+      "acc": 0.69480772,
+      "epoch": 0.28247082699137493,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.774148757861584e-06,
+      "loss": 1.37106171,
+      "memory(GiB)": 103.58,
+      "step": 11135,
+      "train_speed(iter/s)": 1.314121
+    },
+    {
+      "acc": 0.68130307,
+      "epoch": 0.28259766615931,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.773837050773824e-06,
+      "loss": 1.45896606,
+      "memory(GiB)": 103.58,
+      "step": 11140,
+      "train_speed(iter/s)": 1.314192
+    },
+    {
+      "acc": 0.66603403,
+      "epoch": 0.2827245053272451,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.773525133711399e-06,
+      "loss": 1.47616711,
+      "memory(GiB)": 103.58,
+      "step": 11145,
+      "train_speed(iter/s)": 1.314261
+    },
+    {
+      "acc": 0.67787676,
+      "epoch": 0.2828513444951801,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.773213006688024e-06,
+      "loss": 1.4614686,
+      "memory(GiB)": 103.58,
+      "step": 11150,
+      "train_speed(iter/s)": 1.314329
+    },
+    {
+      "acc": 0.66683297,
+      "epoch": 0.28297818366311517,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.77290066971743e-06,
+      "loss": 1.52399387,
+      "memory(GiB)": 103.58,
+      "step": 11155,
+      "train_speed(iter/s)": 1.314398
+    },
+    {
+      "acc": 0.68438239,
+      "epoch": 0.2831050228310502,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.772588122813358e-06,
+      "loss": 1.43468571,
+      "memory(GiB)": 103.58,
+      "step": 11160,
+      "train_speed(iter/s)": 1.314466
+    },
+    {
+      "acc": 0.69987621,
+      "epoch": 0.2832318619989853,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.772275365989548e-06,
+      "loss": 1.44010439,
+      "memory(GiB)": 103.58,
+      "step": 11165,
+      "train_speed(iter/s)": 1.314538
+    },
+    {
+      "acc": 0.69377027,
+      "epoch": 0.28335870116692036,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.771962399259764e-06,
+      "loss": 1.39739685,
+      "memory(GiB)": 103.58,
+      "step": 11170,
+      "train_speed(iter/s)": 1.314606
+    },
+    {
+      "acc": 0.69223456,
+      "epoch": 0.2834855403348554,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.771649222637767e-06,
+      "loss": 1.33924713,
+      "memory(GiB)": 103.58,
+      "step": 11175,
+      "train_speed(iter/s)": 1.314673
+    },
+    {
+      "acc": 0.66084533,
+      "epoch": 0.28361237950279045,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.771335836137332e-06,
+      "loss": 1.45109997,
+      "memory(GiB)": 103.58,
+      "step": 11180,
+      "train_speed(iter/s)": 1.314744
+    },
+    {
+      "acc": 0.69679217,
+      "epoch": 0.28373921867072555,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.771022239772248e-06,
+      "loss": 1.33105927,
+      "memory(GiB)": 103.58,
+      "step": 11185,
+      "train_speed(iter/s)": 1.314813
+    },
+    {
+      "acc": 0.67491651,
+      "epoch": 0.2838660578386606,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.770708433556302e-06,
+      "loss": 1.47011423,
+      "memory(GiB)": 103.58,
+      "step": 11190,
+      "train_speed(iter/s)": 1.314882
+    },
+    {
+      "acc": 0.67663546,
+      "epoch": 0.28399289700659563,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.7703944175033e-06,
+      "loss": 1.43640852,
+      "memory(GiB)": 103.58,
+      "step": 11195,
+      "train_speed(iter/s)": 1.314949
+    },
+    {
+      "acc": 0.69196243,
+      "epoch": 0.2841197361745307,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.770080191627054e-06,
+      "loss": 1.36660604,
+      "memory(GiB)": 103.58,
+      "step": 11200,
+      "train_speed(iter/s)": 1.315015
+    },
+    {
+      "acc": 0.69008713,
+      "epoch": 0.2842465753424658,
+      "grad_norm": 2.5,
+      "learning_rate": 9.769765755941383e-06,
+      "loss": 1.40352268,
+      "memory(GiB)": 103.58,
+      "step": 11205,
+      "train_speed(iter/s)": 1.315082
+    },
+    {
+      "acc": 0.67903218,
+      "epoch": 0.2843734145104008,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.76945111046012e-06,
+      "loss": 1.40967941,
+      "memory(GiB)": 103.58,
+      "step": 11210,
+      "train_speed(iter/s)": 1.315153
+    },
+    {
+      "acc": 0.67041464,
+      "epoch": 0.28450025367833587,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.769136255197103e-06,
+      "loss": 1.51183834,
+      "memory(GiB)": 103.58,
+      "step": 11215,
+      "train_speed(iter/s)": 1.315221
+    },
+    {
+      "acc": 0.67470751,
+      "epoch": 0.2846270928462709,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.768821190166179e-06,
+      "loss": 1.42143688,
+      "memory(GiB)": 103.58,
+      "step": 11220,
+      "train_speed(iter/s)": 1.315289
+    },
+    {
+      "acc": 0.6848434,
+      "epoch": 0.284753932014206,
+      "grad_norm": 3.0,
+      "learning_rate": 9.76850591538121e-06,
+      "loss": 1.33720436,
+      "memory(GiB)": 103.58,
+      "step": 11225,
+      "train_speed(iter/s)": 1.315359
+    },
+    {
+      "acc": 0.67531796,
+      "epoch": 0.28488077118214106,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.76819043085606e-06,
+      "loss": 1.4223341,
+      "memory(GiB)": 103.58,
+      "step": 11230,
+      "train_speed(iter/s)": 1.315417
+    },
+    {
+      "acc": 0.69479079,
+      "epoch": 0.2850076103500761,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.767874736604605e-06,
+      "loss": 1.35555801,
+      "memory(GiB)": 103.58,
+      "step": 11235,
+      "train_speed(iter/s)": 1.315485
+    },
+    {
+      "acc": 0.69132667,
+      "epoch": 0.28513444951801115,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.767558832640734e-06,
+      "loss": 1.35112829,
+      "memory(GiB)": 103.58,
+      "step": 11240,
+      "train_speed(iter/s)": 1.315556
+    },
+    {
+      "acc": 0.68309474,
+      "epoch": 0.28526128868594625,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.76724271897834e-06,
+      "loss": 1.36818142,
+      "memory(GiB)": 103.58,
+      "step": 11245,
+      "train_speed(iter/s)": 1.315623
+    },
+    {
+      "acc": 0.66896462,
+      "epoch": 0.2853881278538813,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.766926395631326e-06,
+      "loss": 1.46236048,
+      "memory(GiB)": 103.58,
+      "step": 11250,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.70289893,
+      "epoch": 0.28551496702181633,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.766609862613607e-06,
+      "loss": 1.3334404,
+      "memory(GiB)": 103.58,
+      "step": 11255,
+      "train_speed(iter/s)": 1.315757
+    },
+    {
+      "acc": 0.68945961,
+      "epoch": 0.2856418061897514,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.766293119939104e-06,
+      "loss": 1.44700079,
+      "memory(GiB)": 103.58,
+      "step": 11260,
+      "train_speed(iter/s)": 1.315825
+    },
+    {
+      "acc": 0.68823376,
+      "epoch": 0.2857686453576865,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.76597616762175e-06,
+      "loss": 1.40591297,
+      "memory(GiB)": 103.58,
+      "step": 11265,
+      "train_speed(iter/s)": 1.315892
+    },
+    {
+      "acc": 0.67278461,
+      "epoch": 0.2858954845256215,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.765659005675488e-06,
+      "loss": 1.4014946,
+      "memory(GiB)": 103.58,
+      "step": 11270,
+      "train_speed(iter/s)": 1.315957
+    },
+    {
+      "acc": 0.66843481,
+      "epoch": 0.28602232369355657,
+      "grad_norm": 3.0,
+      "learning_rate": 9.765341634114263e-06,
+      "loss": 1.45914364,
+      "memory(GiB)": 103.58,
+      "step": 11275,
+      "train_speed(iter/s)": 1.316024
+    },
+    {
+      "acc": 0.68000078,
+      "epoch": 0.2861491628614916,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.765024052952037e-06,
+      "loss": 1.43038168,
+      "memory(GiB)": 103.58,
+      "step": 11280,
+      "train_speed(iter/s)": 1.316092
+    },
+    {
+      "acc": 0.67878337,
+      "epoch": 0.2862760020294267,
+      "grad_norm": 2.75,
+      "learning_rate": 9.76470626220278e-06,
+      "loss": 1.45821934,
+      "memory(GiB)": 103.58,
+      "step": 11285,
+      "train_speed(iter/s)": 1.316159
+    },
+    {
+      "acc": 0.67809639,
+      "epoch": 0.28640284119736176,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.76438826188047e-06,
+      "loss": 1.42670441,
+      "memory(GiB)": 103.58,
+      "step": 11290,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.66834517,
+      "epoch": 0.2865296803652968,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.76407005199909e-06,
+      "loss": 1.45032015,
+      "memory(GiB)": 103.58,
+      "step": 11295,
+      "train_speed(iter/s)": 1.316285
+    },
+    {
+      "acc": 0.67724237,
+      "epoch": 0.28665651953323185,
+      "grad_norm": 2.3125,
+      "learning_rate": 9.76375163257264e-06,
+      "loss": 1.41110344,
+      "memory(GiB)": 103.58,
+      "step": 11300,
+      "train_speed(iter/s)": 1.316351
+    },
+    {
+      "acc": 0.68664875,
+      "epoch": 0.28678335870116695,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.763433003615124e-06,
+      "loss": 1.40386925,
+      "memory(GiB)": 103.58,
+      "step": 11305,
+      "train_speed(iter/s)": 1.316412
+    },
+    {
+      "acc": 0.67890549,
+      "epoch": 0.286910197869102,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.763114165140559e-06,
+      "loss": 1.50228872,
+      "memory(GiB)": 103.58,
+      "step": 11310,
+      "train_speed(iter/s)": 1.316474
+    },
+    {
+      "acc": 0.67139869,
+      "epoch": 0.28703703703703703,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.762795117162967e-06,
+      "loss": 1.46661158,
+      "memory(GiB)": 103.58,
+      "step": 11315,
+      "train_speed(iter/s)": 1.316541
+    },
+    {
+      "acc": 0.69790697,
+      "epoch": 0.2871638762049721,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.76247585969638e-06,
+      "loss": 1.3748455,
+      "memory(GiB)": 103.58,
+      "step": 11320,
+      "train_speed(iter/s)": 1.316607
+    },
+    {
+      "acc": 0.68984137,
+      "epoch": 0.2872907153729072,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.762156392754842e-06,
+      "loss": 1.37347469,
+      "memory(GiB)": 103.58,
+      "step": 11325,
+      "train_speed(iter/s)": 1.316673
+    },
+    {
+      "acc": 0.65758433,
+      "epoch": 0.2874175545408422,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.761836716352405e-06,
+      "loss": 1.44809074,
+      "memory(GiB)": 103.58,
+      "step": 11330,
+      "train_speed(iter/s)": 1.31674
+    },
+    {
+      "acc": 0.67094965,
+      "epoch": 0.28754439370877727,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.761516830503128e-06,
+      "loss": 1.47430096,
+      "memory(GiB)": 103.58,
+      "step": 11335,
+      "train_speed(iter/s)": 1.316807
+    },
+    {
+      "acc": 0.68410683,
+      "epoch": 0.2876712328767123,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.761196735221083e-06,
+      "loss": 1.43832703,
+      "memory(GiB)": 103.58,
+      "step": 11340,
+      "train_speed(iter/s)": 1.316876
+    },
+    {
+      "acc": 0.67906718,
+      "epoch": 0.2877980720446474,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.76087643052035e-06,
+      "loss": 1.42234421,
+      "memory(GiB)": 103.58,
+      "step": 11345,
+      "train_speed(iter/s)": 1.316941
+    },
+    {
+      "acc": 0.67745728,
+      "epoch": 0.28792491121258246,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.760555916415015e-06,
+      "loss": 1.43473883,
+      "memory(GiB)": 103.58,
+      "step": 11350,
+      "train_speed(iter/s)": 1.317008
+    },
+    {
+      "acc": 0.68709946,
+      "epoch": 0.2880517503805175,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.760235192919175e-06,
+      "loss": 1.37202568,
+      "memory(GiB)": 103.58,
+      "step": 11355,
+      "train_speed(iter/s)": 1.317075
+    },
+    {
+      "acc": 0.67553,
+      "epoch": 0.28817858954845255,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.75991426004694e-06,
+      "loss": 1.44872131,
+      "memory(GiB)": 103.58,
+      "step": 11360,
+      "train_speed(iter/s)": 1.317139
+    },
+    {
+      "acc": 0.67701855,
+      "epoch": 0.28830542871638765,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.759593117812423e-06,
+      "loss": 1.41427803,
+      "memory(GiB)": 103.58,
+      "step": 11365,
+      "train_speed(iter/s)": 1.317204
+    },
+    {
+      "acc": 0.69779811,
+      "epoch": 0.2884322678843227,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.75927176622975e-06,
+      "loss": 1.36166973,
+      "memory(GiB)": 103.58,
+      "step": 11370,
+      "train_speed(iter/s)": 1.31727
+    },
+    {
+      "acc": 0.6724884,
+      "epoch": 0.28855910705225774,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.758950205313057e-06,
+      "loss": 1.46871624,
+      "memory(GiB)": 103.58,
+      "step": 11375,
+      "train_speed(iter/s)": 1.317335
+    },
+    {
+      "acc": 0.6950388,
+      "epoch": 0.2886859462201928,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.758628435076488e-06,
+      "loss": 1.35349207,
+      "memory(GiB)": 103.58,
+      "step": 11380,
+      "train_speed(iter/s)": 1.317403
+    },
+    {
+      "acc": 0.67503481,
+      "epoch": 0.2888127853881279,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.758306455534193e-06,
+      "loss": 1.43632717,
+      "memory(GiB)": 103.58,
+      "step": 11385,
+      "train_speed(iter/s)": 1.31747
+    },
+    {
+      "acc": 0.67203341,
+      "epoch": 0.2889396245560629,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.757984266700336e-06,
+      "loss": 1.45215092,
+      "memory(GiB)": 103.58,
+      "step": 11390,
+      "train_speed(iter/s)": 1.317538
+    },
+    {
+      "acc": 0.68174438,
+      "epoch": 0.28906646372399797,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.75766186858909e-06,
+      "loss": 1.39275513,
+      "memory(GiB)": 103.58,
+      "step": 11395,
+      "train_speed(iter/s)": 1.317605
+    },
+    {
+      "acc": 0.68110962,
+      "epoch": 0.289193302891933,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.757339261214631e-06,
+      "loss": 1.48581457,
+      "memory(GiB)": 103.58,
+      "step": 11400,
+      "train_speed(iter/s)": 1.317674
+    },
+    {
+      "acc": 0.67480536,
+      "epoch": 0.2893201420598681,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.757016444591152e-06,
+      "loss": 1.42646332,
+      "memory(GiB)": 103.58,
+      "step": 11405,
+      "train_speed(iter/s)": 1.31774
+    },
+    {
+      "acc": 0.6754354,
+      "epoch": 0.28944698122780316,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.756693418732852e-06,
+      "loss": 1.43132915,
+      "memory(GiB)": 103.58,
+      "step": 11410,
+      "train_speed(iter/s)": 1.31781
+    },
+    {
+      "acc": 0.68476362,
+      "epoch": 0.2895738203957382,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.756370183653938e-06,
+      "loss": 1.48429108,
+      "memory(GiB)": 103.58,
+      "step": 11415,
+      "train_speed(iter/s)": 1.317877
+    },
+    {
+      "acc": 0.68445668,
+      "epoch": 0.28970065956367325,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.756046739368628e-06,
+      "loss": 1.38570461,
+      "memory(GiB)": 103.58,
+      "step": 11420,
+      "train_speed(iter/s)": 1.317945
+    },
+    {
+      "acc": 0.67493734,
+      "epoch": 0.28982749873160835,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.755723085891147e-06,
+      "loss": 1.46835585,
+      "memory(GiB)": 103.58,
+      "step": 11425,
+      "train_speed(iter/s)": 1.318007
+    },
+    {
+      "acc": 0.69106426,
+      "epoch": 0.2899543378995434,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.755399223235734e-06,
+      "loss": 1.35736074,
+      "memory(GiB)": 103.58,
+      "step": 11430,
+      "train_speed(iter/s)": 1.318076
+    },
+    {
+      "acc": 0.68166943,
+      "epoch": 0.29008117706747844,
+      "grad_norm": 3.75,
+      "learning_rate": 9.75507515141663e-06,
+      "loss": 1.40743866,
+      "memory(GiB)": 103.58,
+      "step": 11435,
+      "train_speed(iter/s)": 1.318145
+    },
+    {
+      "acc": 0.67547092,
+      "epoch": 0.2902080162354135,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.75475087044809e-06,
+      "loss": 1.41307354,
+      "memory(GiB)": 103.58,
+      "step": 11440,
+      "train_speed(iter/s)": 1.318209
+    },
+    {
+      "acc": 0.6772779,
+      "epoch": 0.2903348554033486,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.754426380344382e-06,
+      "loss": 1.45347519,
+      "memory(GiB)": 103.58,
+      "step": 11445,
+      "train_speed(iter/s)": 1.318275
+    },
+    {
+      "acc": 0.67196808,
+      "epoch": 0.2904616945712836,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.754101681119772e-06,
+      "loss": 1.47723179,
+      "memory(GiB)": 103.58,
+      "step": 11450,
+      "train_speed(iter/s)": 1.318343
+    },
+    {
+      "acc": 0.66674995,
+      "epoch": 0.29058853373921867,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.753776772788545e-06,
+      "loss": 1.46673269,
+      "memory(GiB)": 103.58,
+      "step": 11455,
+      "train_speed(iter/s)": 1.318411
+    },
+    {
+      "acc": 0.68515825,
+      "epoch": 0.2907153729071537,
+      "grad_norm": 2.625,
+      "learning_rate": 9.753451655364992e-06,
+      "loss": 1.4252636,
+      "memory(GiB)": 103.58,
+      "step": 11460,
+      "train_speed(iter/s)": 1.318477
+    },
+    {
+      "acc": 0.68068709,
+      "epoch": 0.2908422120750888,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.75312632886341e-06,
+      "loss": 1.48398485,
+      "memory(GiB)": 103.58,
+      "step": 11465,
+      "train_speed(iter/s)": 1.318545
+    },
+    {
+      "acc": 0.70529089,
+      "epoch": 0.29096905124302386,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.752800793298113e-06,
+      "loss": 1.34524755,
+      "memory(GiB)": 103.58,
+      "step": 11470,
+      "train_speed(iter/s)": 1.318612
+    },
+    {
+      "acc": 0.69060612,
+      "epoch": 0.2910958904109589,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.752475048683419e-06,
+      "loss": 1.41270704,
+      "memory(GiB)": 103.58,
+      "step": 11475,
+      "train_speed(iter/s)": 1.318678
+    },
+    {
+      "acc": 0.67877617,
+      "epoch": 0.29122272957889395,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.752149095033651e-06,
+      "loss": 1.38366337,
+      "memory(GiB)": 103.58,
+      "step": 11480,
+      "train_speed(iter/s)": 1.318743
+    },
+    {
+      "acc": 0.69210849,
+      "epoch": 0.29134956874682905,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.75182293236315e-06,
+      "loss": 1.37219048,
+      "memory(GiB)": 103.58,
+      "step": 11485,
+      "train_speed(iter/s)": 1.31881
+    },
+    {
+      "acc": 0.67602258,
+      "epoch": 0.2914764079147641,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.751496560686262e-06,
+      "loss": 1.33447685,
+      "memory(GiB)": 103.58,
+      "step": 11490,
+      "train_speed(iter/s)": 1.318879
+    },
+    {
+      "acc": 0.68230553,
+      "epoch": 0.29160324708269914,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.751169980017341e-06,
+      "loss": 1.39866352,
+      "memory(GiB)": 103.58,
+      "step": 11495,
+      "train_speed(iter/s)": 1.318946
+    },
+    {
+      "acc": 0.69088902,
+      "epoch": 0.2917300862506342,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.750843190370752e-06,
+      "loss": 1.37974968,
+      "memory(GiB)": 103.58,
+      "step": 11500,
+      "train_speed(iter/s)": 1.319012
+    },
+    {
+      "acc": 0.70392179,
+      "epoch": 0.2918569254185693,
+      "grad_norm": 2.265625,
+      "learning_rate": 9.750516191760868e-06,
+      "loss": 1.36129131,
+      "memory(GiB)": 103.58,
+      "step": 11505,
+      "train_speed(iter/s)": 1.319075
+    },
+    {
+      "acc": 0.68842468,
+      "epoch": 0.2919837645865043,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.750188984202073e-06,
+      "loss": 1.39052715,
+      "memory(GiB)": 103.58,
+      "step": 11510,
+      "train_speed(iter/s)": 1.319142
+    },
+    {
+      "acc": 0.67272272,
+      "epoch": 0.29211060375443937,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.749861567708759e-06,
+      "loss": 1.48831005,
+      "memory(GiB)": 103.58,
+      "step": 11515,
+      "train_speed(iter/s)": 1.319208
+    },
+    {
+      "acc": 0.67092652,
+      "epoch": 0.2922374429223744,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.749533942295323e-06,
+      "loss": 1.46219349,
+      "memory(GiB)": 103.58,
+      "step": 11520,
+      "train_speed(iter/s)": 1.319274
+    },
+    {
+      "acc": 0.68692174,
+      "epoch": 0.2923642820903095,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.749206107976183e-06,
+      "loss": 1.37869797,
+      "memory(GiB)": 103.58,
+      "step": 11525,
+      "train_speed(iter/s)": 1.319343
+    },
+    {
+      "acc": 0.67634921,
+      "epoch": 0.29249112125824456,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.748878064765753e-06,
+      "loss": 1.45270443,
+      "memory(GiB)": 103.58,
+      "step": 11530,
+      "train_speed(iter/s)": 1.319409
+    },
+    {
+      "acc": 0.67955055,
+      "epoch": 0.2926179604261796,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.748549812678466e-06,
+      "loss": 1.46452589,
+      "memory(GiB)": 103.58,
+      "step": 11535,
+      "train_speed(iter/s)": 1.319477
+    },
+    {
+      "acc": 0.68255491,
+      "epoch": 0.29274479959411465,
+      "grad_norm": 2.75,
+      "learning_rate": 9.748221351728754e-06,
+      "loss": 1.45473995,
+      "memory(GiB)": 103.58,
+      "step": 11540,
+      "train_speed(iter/s)": 1.319543
+    },
+    {
+      "acc": 0.68816175,
+      "epoch": 0.29287163876204975,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.747892681931067e-06,
+      "loss": 1.40150719,
+      "memory(GiB)": 103.58,
+      "step": 11545,
+      "train_speed(iter/s)": 1.319609
+    },
+    {
+      "acc": 0.67849607,
+      "epoch": 0.2929984779299848,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.747563803299865e-06,
+      "loss": 1.41210775,
+      "memory(GiB)": 103.58,
+      "step": 11550,
+      "train_speed(iter/s)": 1.319675
+    },
+    {
+      "acc": 0.68381233,
+      "epoch": 0.29312531709791984,
+      "grad_norm": 3.125,
+      "learning_rate": 9.74723471584961e-06,
+      "loss": 1.39703369,
+      "memory(GiB)": 103.58,
+      "step": 11555,
+      "train_speed(iter/s)": 1.319741
+    },
+    {
+      "acc": 0.67869835,
+      "epoch": 0.2932521562658549,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.746905419594777e-06,
+      "loss": 1.38541088,
+      "memory(GiB)": 103.58,
+      "step": 11560,
+      "train_speed(iter/s)": 1.319808
+    },
+    {
+      "acc": 0.67093654,
+      "epoch": 0.29337899543379,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.746575914549851e-06,
+      "loss": 1.43809948,
+      "memory(GiB)": 103.58,
+      "step": 11565,
+      "train_speed(iter/s)": 1.319873
+    },
+    {
+      "acc": 0.66853647,
+      "epoch": 0.293505834601725,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.746246200729323e-06,
+      "loss": 1.50691767,
+      "memory(GiB)": 103.58,
+      "step": 11570,
+      "train_speed(iter/s)": 1.319938
+    },
+    {
+      "acc": 0.68634448,
+      "epoch": 0.29363267376966007,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.745916278147696e-06,
+      "loss": 1.39381733,
+      "memory(GiB)": 103.58,
+      "step": 11575,
+      "train_speed(iter/s)": 1.320003
+    },
+    {
+      "acc": 0.6829319,
+      "epoch": 0.2937595129375951,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.745586146819484e-06,
+      "loss": 1.43174114,
+      "memory(GiB)": 103.58,
+      "step": 11580,
+      "train_speed(iter/s)": 1.320066
+    },
+    {
+      "acc": 0.67830381,
+      "epoch": 0.2938863521055302,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.745255806759205e-06,
+      "loss": 1.50164661,
+      "memory(GiB)": 103.58,
+      "step": 11585,
+      "train_speed(iter/s)": 1.320131
+    },
+    {
+      "acc": 0.68274479,
+      "epoch": 0.29401319127346526,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.74492525798139e-06,
+      "loss": 1.44249096,
+      "memory(GiB)": 103.58,
+      "step": 11590,
+      "train_speed(iter/s)": 1.320194
+    },
+    {
+      "acc": 0.66587954,
+      "epoch": 0.2941400304414003,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.744594500500578e-06,
+      "loss": 1.44870453,
+      "memory(GiB)": 103.58,
+      "step": 11595,
+      "train_speed(iter/s)": 1.320259
+    },
+    {
+      "acc": 0.67501016,
+      "epoch": 0.29426686960933535,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.744263534331315e-06,
+      "loss": 1.39511385,
+      "memory(GiB)": 103.58,
+      "step": 11600,
+      "train_speed(iter/s)": 1.320322
+    },
+    {
+      "acc": 0.6875762,
+      "epoch": 0.29439370877727045,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.743932359488161e-06,
+      "loss": 1.40538692,
+      "memory(GiB)": 103.58,
+      "step": 11605,
+      "train_speed(iter/s)": 1.320387
+    },
+    {
+      "acc": 0.6734045,
+      "epoch": 0.2945205479452055,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.743600975985681e-06,
+      "loss": 1.44619923,
+      "memory(GiB)": 103.58,
+      "step": 11610,
+      "train_speed(iter/s)": 1.320452
+    },
+    {
+      "acc": 0.68181648,
+      "epoch": 0.29464738711314054,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.743269383838452e-06,
+      "loss": 1.44993095,
+      "memory(GiB)": 103.58,
+      "step": 11615,
+      "train_speed(iter/s)": 1.320518
+    },
+    {
+      "acc": 0.68582606,
+      "epoch": 0.2947742262810756,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.74293758306106e-06,
+      "loss": 1.39530315,
+      "memory(GiB)": 103.58,
+      "step": 11620,
+      "train_speed(iter/s)": 1.320585
+    },
+    {
+      "acc": 0.6631372,
+      "epoch": 0.2949010654490107,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.742605573668096e-06,
+      "loss": 1.44046421,
+      "memory(GiB)": 103.58,
+      "step": 11625,
+      "train_speed(iter/s)": 1.320649
+    },
+    {
+      "acc": 0.67966771,
+      "epoch": 0.2950279046169457,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.742273355674164e-06,
+      "loss": 1.47607536,
+      "memory(GiB)": 103.58,
+      "step": 11630,
+      "train_speed(iter/s)": 1.320715
+    },
+    {
+      "acc": 0.69115944,
+      "epoch": 0.29515474378488077,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.741940929093879e-06,
+      "loss": 1.41931696,
+      "memory(GiB)": 103.58,
+      "step": 11635,
+      "train_speed(iter/s)": 1.320778
+    },
+    {
+      "acc": 0.69391041,
+      "epoch": 0.2952815829528158,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.741608293941858e-06,
+      "loss": 1.35682278,
+      "memory(GiB)": 103.58,
+      "step": 11640,
+      "train_speed(iter/s)": 1.320834
+    },
+    {
+      "acc": 0.69222879,
+      "epoch": 0.2954084221207509,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.741275450232736e-06,
+      "loss": 1.43391075,
+      "memory(GiB)": 103.58,
+      "step": 11645,
+      "train_speed(iter/s)": 1.320898
+    },
+    {
+      "acc": 0.69206638,
+      "epoch": 0.29553526128868596,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.740942397981151e-06,
+      "loss": 1.32817192,
+      "memory(GiB)": 103.58,
+      "step": 11650,
+      "train_speed(iter/s)": 1.320963
+    },
+    {
+      "acc": 0.68402047,
+      "epoch": 0.295662100456621,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.740609137201752e-06,
+      "loss": 1.422824,
+      "memory(GiB)": 103.58,
+      "step": 11655,
+      "train_speed(iter/s)": 1.321028
+    },
+    {
+      "acc": 0.68832073,
+      "epoch": 0.29578893962455605,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.7402756679092e-06,
+      "loss": 1.41569633,
+      "memory(GiB)": 103.58,
+      "step": 11660,
+      "train_speed(iter/s)": 1.321092
+    },
+    {
+      "acc": 0.68028259,
+      "epoch": 0.29591577879249115,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.739941990118157e-06,
+      "loss": 1.37503319,
+      "memory(GiB)": 103.58,
+      "step": 11665,
+      "train_speed(iter/s)": 1.321155
+    },
+    {
+      "acc": 0.68956795,
+      "epoch": 0.2960426179604262,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.739608103843306e-06,
+      "loss": 1.40482378,
+      "memory(GiB)": 103.58,
+      "step": 11670,
+      "train_speed(iter/s)": 1.321212
+    },
+    {
+      "acc": 0.67655005,
+      "epoch": 0.29616945712836124,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.739274009099328e-06,
+      "loss": 1.43868809,
+      "memory(GiB)": 103.58,
+      "step": 11675,
+      "train_speed(iter/s)": 1.321279
+    },
+    {
+      "acc": 0.69778233,
+      "epoch": 0.2962962962962963,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.738939705900922e-06,
+      "loss": 1.40186539,
+      "memory(GiB)": 103.58,
+      "step": 11680,
+      "train_speed(iter/s)": 1.321345
+    },
+    {
+      "acc": 0.69065018,
+      "epoch": 0.2964231354642314,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.738605194262787e-06,
+      "loss": 1.37151909,
+      "memory(GiB)": 103.58,
+      "step": 11685,
+      "train_speed(iter/s)": 1.321411
+    },
+    {
+      "acc": 0.68008084,
+      "epoch": 0.2965499746321664,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.738270474199641e-06,
+      "loss": 1.41408691,
+      "memory(GiB)": 103.58,
+      "step": 11690,
+      "train_speed(iter/s)": 1.321478
+    },
+    {
+      "acc": 0.67175527,
+      "epoch": 0.29667681380010147,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.737935545726205e-06,
+      "loss": 1.46417007,
+      "memory(GiB)": 103.58,
+      "step": 11695,
+      "train_speed(iter/s)": 1.321544
+    },
+    {
+      "acc": 0.68535199,
+      "epoch": 0.2968036529680365,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.737600408857208e-06,
+      "loss": 1.40932941,
+      "memory(GiB)": 103.58,
+      "step": 11700,
+      "train_speed(iter/s)": 1.321608
+    },
+    {
+      "acc": 0.67640748,
+      "epoch": 0.2969304921359716,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.737265063607395e-06,
+      "loss": 1.43054953,
+      "memory(GiB)": 103.58,
+      "step": 11705,
+      "train_speed(iter/s)": 1.321672
+    },
+    {
+      "acc": 0.68319907,
+      "epoch": 0.29705733130390666,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.736929509991515e-06,
+      "loss": 1.39165335,
+      "memory(GiB)": 103.58,
+      "step": 11710,
+      "train_speed(iter/s)": 1.321734
+    },
+    {
+      "acc": 0.68700895,
+      "epoch": 0.2971841704718417,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.736593748024325e-06,
+      "loss": 1.43684025,
+      "memory(GiB)": 103.58,
+      "step": 11715,
+      "train_speed(iter/s)": 1.3218
+    },
+    {
+      "acc": 0.68820462,
+      "epoch": 0.29731100963977675,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.736257777720595e-06,
+      "loss": 1.35562391,
+      "memory(GiB)": 103.58,
+      "step": 11720,
+      "train_speed(iter/s)": 1.321865
+    },
+    {
+      "acc": 0.69315586,
+      "epoch": 0.29743784880771185,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.735921599095101e-06,
+      "loss": 1.37566662,
+      "memory(GiB)": 103.58,
+      "step": 11725,
+      "train_speed(iter/s)": 1.321931
+    },
+    {
+      "acc": 0.69340353,
+      "epoch": 0.2975646879756469,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.735585212162633e-06,
+      "loss": 1.37137737,
+      "memory(GiB)": 103.58,
+      "step": 11730,
+      "train_speed(iter/s)": 1.321996
+    },
+    {
+      "acc": 0.6777431,
+      "epoch": 0.29769152714358194,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.735248616937983e-06,
+      "loss": 1.44319744,
+      "memory(GiB)": 103.58,
+      "step": 11735,
+      "train_speed(iter/s)": 1.322062
+    },
+    {
+      "acc": 0.68265767,
+      "epoch": 0.297818366311517,
+      "grad_norm": 3.0,
+      "learning_rate": 9.734911813435957e-06,
+      "loss": 1.38505421,
+      "memory(GiB)": 103.58,
+      "step": 11740,
+      "train_speed(iter/s)": 1.322125
+    },
+    {
+      "acc": 0.6830678,
+      "epoch": 0.2979452054794521,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.73457480167137e-06,
+      "loss": 1.53464088,
+      "memory(GiB)": 103.58,
+      "step": 11745,
+      "train_speed(iter/s)": 1.322191
+    },
+    {
+      "acc": 0.69080925,
+      "epoch": 0.2980720446473871,
+      "grad_norm": 3.375,
+      "learning_rate": 9.734237581659045e-06,
+      "loss": 1.41795015,
+      "memory(GiB)": 103.58,
+      "step": 11750,
+      "train_speed(iter/s)": 1.322255
+    },
+    {
+      "acc": 0.6831665,
+      "epoch": 0.29819888381532217,
+      "grad_norm": 3.25,
+      "learning_rate": 9.733900153413813e-06,
+      "loss": 1.38505001,
+      "memory(GiB)": 103.58,
+      "step": 11755,
+      "train_speed(iter/s)": 1.322319
+    },
+    {
+      "acc": 0.68449059,
+      "epoch": 0.2983257229832572,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.733562516950519e-06,
+      "loss": 1.41081181,
+      "memory(GiB)": 103.58,
+      "step": 11760,
+      "train_speed(iter/s)": 1.322383
+    },
+    {
+      "acc": 0.67271113,
+      "epoch": 0.2984525621511923,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.73322467228401e-06,
+      "loss": 1.42109184,
+      "memory(GiB)": 103.58,
+      "step": 11765,
+      "train_speed(iter/s)": 1.322446
+    },
+    {
+      "acc": 0.68628702,
+      "epoch": 0.29857940131912736,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.73288661942915e-06,
+      "loss": 1.38794918,
+      "memory(GiB)": 103.58,
+      "step": 11770,
+      "train_speed(iter/s)": 1.32251
+    },
+    {
+      "acc": 0.69289627,
+      "epoch": 0.2987062404870624,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.732548358400802e-06,
+      "loss": 1.38462276,
+      "memory(GiB)": 103.58,
+      "step": 11775,
+      "train_speed(iter/s)": 1.322574
+    },
+    {
+      "acc": 0.6804986,
+      "epoch": 0.29883307965499745,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.73220988921385e-06,
+      "loss": 1.37049637,
+      "memory(GiB)": 103.58,
+      "step": 11780,
+      "train_speed(iter/s)": 1.322637
+    },
+    {
+      "acc": 0.6774086,
+      "epoch": 0.29895991882293255,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.73187121188318e-06,
+      "loss": 1.3751874,
+      "memory(GiB)": 103.58,
+      "step": 11785,
+      "train_speed(iter/s)": 1.322701
+    },
+    {
+      "acc": 0.69294891,
+      "epoch": 0.2990867579908676,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.731532326423686e-06,
+      "loss": 1.38502283,
+      "memory(GiB)": 103.58,
+      "step": 11790,
+      "train_speed(iter/s)": 1.322766
+    },
+    {
+      "acc": 0.70020199,
+      "epoch": 0.29921359715880264,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.731193232850277e-06,
+      "loss": 1.3471776,
+      "memory(GiB)": 103.58,
+      "step": 11795,
+      "train_speed(iter/s)": 1.322828
+    },
+    {
+      "acc": 0.67677112,
+      "epoch": 0.2993404363267377,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.730853931177866e-06,
+      "loss": 1.44407959,
+      "memory(GiB)": 103.58,
+      "step": 11800,
+      "train_speed(iter/s)": 1.322893
+    },
+    {
+      "acc": 0.68813715,
+      "epoch": 0.2994672754946728,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.730514421421378e-06,
+      "loss": 1.4002327,
+      "memory(GiB)": 103.58,
+      "step": 11805,
+      "train_speed(iter/s)": 1.322958
+    },
+    {
+      "acc": 0.67809305,
+      "epoch": 0.2995941146626078,
+      "grad_norm": 2.625,
+      "learning_rate": 9.730174703595745e-06,
+      "loss": 1.42838917,
+      "memory(GiB)": 103.58,
+      "step": 11810,
+      "train_speed(iter/s)": 1.323019
+    },
+    {
+      "acc": 0.68998647,
+      "epoch": 0.29972095383054287,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.72983477771591e-06,
+      "loss": 1.36091452,
+      "memory(GiB)": 103.58,
+      "step": 11815,
+      "train_speed(iter/s)": 1.323081
+    },
+    {
+      "acc": 0.67219334,
+      "epoch": 0.2998477929984779,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.729494643796823e-06,
+      "loss": 1.46501198,
+      "memory(GiB)": 103.58,
+      "step": 11820,
+      "train_speed(iter/s)": 1.323145
+    },
+    {
+      "acc": 0.68217058,
+      "epoch": 0.299974632166413,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.729154301853448e-06,
+      "loss": 1.41578255,
+      "memory(GiB)": 103.58,
+      "step": 11825,
+      "train_speed(iter/s)": 1.323209
+    },
+    {
+      "acc": 0.67931037,
+      "epoch": 0.30010147133434806,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.72881375190075e-06,
+      "loss": 1.44828548,
+      "memory(GiB)": 103.58,
+      "step": 11830,
+      "train_speed(iter/s)": 1.323271
+    },
+    {
+      "acc": 0.69607091,
+      "epoch": 0.3002283105022831,
+      "grad_norm": 2.625,
+      "learning_rate": 9.728472993953712e-06,
+      "loss": 1.37709389,
+      "memory(GiB)": 103.58,
+      "step": 11835,
+      "train_speed(iter/s)": 1.323331
+    },
+    {
+      "acc": 0.67902532,
+      "epoch": 0.30035514967021815,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.728132028027323e-06,
+      "loss": 1.34889069,
+      "memory(GiB)": 103.58,
+      "step": 11840,
+      "train_speed(iter/s)": 1.323394
+    },
+    {
+      "acc": 0.69760818,
+      "epoch": 0.30048198883815325,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.727790854136573e-06,
+      "loss": 1.39651222,
+      "memory(GiB)": 103.58,
+      "step": 11845,
+      "train_speed(iter/s)": 1.323458
+    },
+    {
+      "acc": 0.69172049,
+      "epoch": 0.3006088280060883,
+      "grad_norm": 2.5,
+      "learning_rate": 9.727449472296476e-06,
+      "loss": 1.35363789,
+      "memory(GiB)": 103.58,
+      "step": 11850,
+      "train_speed(iter/s)": 1.323517
+    },
+    {
+      "acc": 0.68143711,
+      "epoch": 0.30073566717402334,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.727107882522045e-06,
+      "loss": 1.46351233,
+      "memory(GiB)": 103.58,
+      "step": 11855,
+      "train_speed(iter/s)": 1.32358
+    },
+    {
+      "acc": 0.68945017,
+      "epoch": 0.3008625063419584,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.726766084828303e-06,
+      "loss": 1.39248104,
+      "memory(GiB)": 103.58,
+      "step": 11860,
+      "train_speed(iter/s)": 1.323644
+    },
+    {
+      "acc": 0.67263136,
+      "epoch": 0.3009893455098935,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.726424079230286e-06,
+      "loss": 1.395821,
+      "memory(GiB)": 103.58,
+      "step": 11865,
+      "train_speed(iter/s)": 1.323704
+    },
+    {
+      "acc": 0.67693133,
+      "epoch": 0.3011161846778285,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.726081865743036e-06,
+      "loss": 1.41477642,
+      "memory(GiB)": 103.58,
+      "step": 11870,
+      "train_speed(iter/s)": 1.323766
+    },
+    {
+      "acc": 0.70289555,
+      "epoch": 0.30124302384576357,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.725739444381603e-06,
+      "loss": 1.37224121,
+      "memory(GiB)": 103.58,
+      "step": 11875,
+      "train_speed(iter/s)": 1.323827
+    },
+    {
+      "acc": 0.68243408,
+      "epoch": 0.3013698630136986,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.725396815161053e-06,
+      "loss": 1.43773422,
+      "memory(GiB)": 103.58,
+      "step": 11880,
+      "train_speed(iter/s)": 1.323887
+    },
+    {
+      "acc": 0.67814617,
+      "epoch": 0.3014967021816337,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.725053978096453e-06,
+      "loss": 1.44822998,
+      "memory(GiB)": 103.58,
+      "step": 11885,
+      "train_speed(iter/s)": 1.323951
+    },
+    {
+      "acc": 0.67784615,
+      "epoch": 0.30162354134956876,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.724710933202884e-06,
+      "loss": 1.47383518,
+      "memory(GiB)": 103.58,
+      "step": 11890,
+      "train_speed(iter/s)": 1.324012
+    },
+    {
+      "acc": 0.68405037,
+      "epoch": 0.3017503805175038,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.724367680495432e-06,
+      "loss": 1.42436295,
+      "memory(GiB)": 103.58,
+      "step": 11895,
+      "train_speed(iter/s)": 1.324071
+    },
+    {
+      "acc": 0.69699221,
+      "epoch": 0.30187721968543885,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.724024219989198e-06,
+      "loss": 1.35782728,
+      "memory(GiB)": 103.58,
+      "step": 11900,
+      "train_speed(iter/s)": 1.32413
+    },
+    {
+      "acc": 0.67668495,
+      "epoch": 0.30200405885337395,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.723680551699286e-06,
+      "loss": 1.41583519,
+      "memory(GiB)": 103.58,
+      "step": 11905,
+      "train_speed(iter/s)": 1.324189
+    },
+    {
+      "acc": 0.67396021,
+      "epoch": 0.302130898021309,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.723336675640815e-06,
+      "loss": 1.5003581,
+      "memory(GiB)": 103.58,
+      "step": 11910,
+      "train_speed(iter/s)": 1.324251
+    },
+    {
+      "acc": 0.6819499,
+      "epoch": 0.30225773718924404,
+      "grad_norm": 3.0,
+      "learning_rate": 9.722992591828908e-06,
+      "loss": 1.42074299,
+      "memory(GiB)": 103.58,
+      "step": 11915,
+      "train_speed(iter/s)": 1.324309
+    },
+    {
+      "acc": 0.70056458,
+      "epoch": 0.3023845763571791,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.722648300278701e-06,
+      "loss": 1.31127625,
+      "memory(GiB)": 103.58,
+      "step": 11920,
+      "train_speed(iter/s)": 1.324366
+    },
+    {
+      "acc": 0.69738083,
+      "epoch": 0.3025114155251142,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.722303801005338e-06,
+      "loss": 1.40940895,
+      "memory(GiB)": 103.58,
+      "step": 11925,
+      "train_speed(iter/s)": 1.324429
+    },
+    {
+      "acc": 0.68786569,
+      "epoch": 0.3026382546930492,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.721959094023968e-06,
+      "loss": 1.45398102,
+      "memory(GiB)": 103.58,
+      "step": 11930,
+      "train_speed(iter/s)": 1.324482
+    },
+    {
+      "acc": 0.70048642,
+      "epoch": 0.30276509386098427,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.721614179349754e-06,
+      "loss": 1.32469692,
+      "memory(GiB)": 103.58,
+      "step": 11935,
+      "train_speed(iter/s)": 1.324544
+    },
+    {
+      "acc": 0.69434166,
+      "epoch": 0.3028919330289193,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.72126905699787e-06,
+      "loss": 1.43518724,
+      "memory(GiB)": 103.58,
+      "step": 11940,
+      "train_speed(iter/s)": 1.324605
+    },
+    {
+      "acc": 0.66956115,
+      "epoch": 0.3030187721968544,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.720923726983493e-06,
+      "loss": 1.39562244,
+      "memory(GiB)": 103.58,
+      "step": 11945,
+      "train_speed(iter/s)": 1.324666
+    },
+    {
+      "acc": 0.67340236,
+      "epoch": 0.30314561136478946,
+      "grad_norm": 2.875,
+      "learning_rate": 9.720578189321814e-06,
+      "loss": 1.42238884,
+      "memory(GiB)": 103.58,
+      "step": 11950,
+      "train_speed(iter/s)": 1.324728
+    },
+    {
+      "acc": 0.68628025,
+      "epoch": 0.3032724505327245,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.72023244402803e-06,
+      "loss": 1.41785746,
+      "memory(GiB)": 103.58,
+      "step": 11955,
+      "train_speed(iter/s)": 1.324788
+    },
+    {
+      "acc": 0.67694178,
+      "epoch": 0.30339928970065955,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.719886491117348e-06,
+      "loss": 1.40558271,
+      "memory(GiB)": 103.58,
+      "step": 11960,
+      "train_speed(iter/s)": 1.324848
+    },
+    {
+      "acc": 0.68418751,
+      "epoch": 0.30352612886859465,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.719540330604986e-06,
+      "loss": 1.36943493,
+      "memory(GiB)": 103.58,
+      "step": 11965,
+      "train_speed(iter/s)": 1.324908
+    },
+    {
+      "acc": 0.68996482,
+      "epoch": 0.3036529680365297,
+      "grad_norm": 3.25,
+      "learning_rate": 9.71919396250617e-06,
+      "loss": 1.42027502,
+      "memory(GiB)": 103.58,
+      "step": 11970,
+      "train_speed(iter/s)": 1.324971
+    },
+    {
+      "acc": 0.66376486,
+      "epoch": 0.30377980720446474,
+      "grad_norm": 3.0,
+      "learning_rate": 9.718847386836131e-06,
+      "loss": 1.48014994,
+      "memory(GiB)": 103.58,
+      "step": 11975,
+      "train_speed(iter/s)": 1.325031
+    },
+    {
+      "acc": 0.67924938,
+      "epoch": 0.3039066463723998,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.718500603610119e-06,
+      "loss": 1.43547764,
+      "memory(GiB)": 103.58,
+      "step": 11980,
+      "train_speed(iter/s)": 1.325093
+    },
+    {
+      "acc": 0.67148046,
+      "epoch": 0.3040334855403349,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.718153612843382e-06,
+      "loss": 1.44479132,
+      "memory(GiB)": 103.58,
+      "step": 11985,
+      "train_speed(iter/s)": 1.325153
+    },
+    {
+      "acc": 0.67101898,
+      "epoch": 0.3041603247082699,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.717806414551186e-06,
+      "loss": 1.45858879,
+      "memory(GiB)": 103.58,
+      "step": 11990,
+      "train_speed(iter/s)": 1.325195
+    },
+    {
+      "acc": 0.68815699,
+      "epoch": 0.30428716387620497,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.717459008748798e-06,
+      "loss": 1.42060509,
+      "memory(GiB)": 103.58,
+      "step": 11995,
+      "train_speed(iter/s)": 1.325254
+    },
+    {
+      "acc": 0.67989783,
+      "epoch": 0.30441400304414,
+      "grad_norm": 3.5,
+      "learning_rate": 9.717111395451501e-06,
+      "loss": 1.39170332,
+      "memory(GiB)": 103.58,
+      "step": 12000,
+      "train_speed(iter/s)": 1.325316
+    },
+    {
+      "epoch": 0.30441400304414,
+      "eval_acc": 0.6707860593037169,
+      "eval_loss": 1.3804293870925903,
+      "eval_runtime": 69.0511,
+      "eval_samples_per_second": 92.251,
+      "eval_steps_per_second": 23.07,
+      "step": 12000
+    },
+    {
+      "acc": 0.66932788,
+      "epoch": 0.3045408422120751,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.716763574674586e-06,
+      "loss": 1.48469954,
+      "memory(GiB)": 103.58,
+      "step": 12005,
+      "train_speed(iter/s)": 1.313535
+    },
+    {
+      "acc": 0.69438539,
+      "epoch": 0.30466768138001016,
+      "grad_norm": 2.875,
+      "learning_rate": 9.71641554643335e-06,
+      "loss": 1.32045574,
+      "memory(GiB)": 103.58,
+      "step": 12010,
+      "train_speed(iter/s)": 1.313594
+    },
+    {
+      "acc": 0.68928251,
+      "epoch": 0.3047945205479452,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.7160673107431e-06,
+      "loss": 1.36017532,
+      "memory(GiB)": 103.58,
+      "step": 12015,
+      "train_speed(iter/s)": 1.313657
+    },
+    {
+      "acc": 0.65869589,
+      "epoch": 0.30492135971588025,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.715718867619155e-06,
+      "loss": 1.49884386,
+      "memory(GiB)": 103.58,
+      "step": 12020,
+      "train_speed(iter/s)": 1.313709
+    },
+    {
+      "acc": 0.68113089,
+      "epoch": 0.30504819888381535,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.715370217076838e-06,
+      "loss": 1.38338089,
+      "memory(GiB)": 103.58,
+      "step": 12025,
+      "train_speed(iter/s)": 1.313766
+    },
+    {
+      "acc": 0.6719636,
+      "epoch": 0.3051750380517504,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.715021359131489e-06,
+      "loss": 1.44073257,
+      "memory(GiB)": 103.58,
+      "step": 12030,
+      "train_speed(iter/s)": 1.313827
+    },
+    {
+      "acc": 0.67615352,
+      "epoch": 0.30530187721968544,
+      "grad_norm": 4.375,
+      "learning_rate": 9.714672293798449e-06,
+      "loss": 1.45939264,
+      "memory(GiB)": 103.58,
+      "step": 12035,
+      "train_speed(iter/s)": 1.31389
+    },
+    {
+      "acc": 0.66876945,
+      "epoch": 0.3054287163876205,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.71432302109307e-06,
+      "loss": 1.46166649,
+      "memory(GiB)": 103.58,
+      "step": 12040,
+      "train_speed(iter/s)": 1.313952
+    },
+    {
+      "acc": 0.68192368,
+      "epoch": 0.3055555555555556,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.713973541030716e-06,
+      "loss": 1.39736118,
+      "memory(GiB)": 103.58,
+      "step": 12045,
+      "train_speed(iter/s)": 1.314007
+    },
+    {
+      "acc": 0.68601546,
+      "epoch": 0.3056823947234906,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.713623853626763e-06,
+      "loss": 1.34832439,
+      "memory(GiB)": 103.58,
+      "step": 12050,
+      "train_speed(iter/s)": 1.314069
+    },
+    {
+      "acc": 0.68446517,
+      "epoch": 0.30580923389142567,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.713273958896586e-06,
+      "loss": 1.41273937,
+      "memory(GiB)": 103.58,
+      "step": 12055,
+      "train_speed(iter/s)": 1.314129
+    },
+    {
+      "acc": 0.68728304,
+      "epoch": 0.3059360730593607,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.712923856855578e-06,
+      "loss": 1.36727772,
+      "memory(GiB)": 103.58,
+      "step": 12060,
+      "train_speed(iter/s)": 1.314192
+    },
+    {
+      "acc": 0.67681484,
+      "epoch": 0.3060629122272958,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.712573547519134e-06,
+      "loss": 1.43593521,
+      "memory(GiB)": 103.58,
+      "step": 12065,
+      "train_speed(iter/s)": 1.314254
+    },
+    {
+      "acc": 0.66494584,
+      "epoch": 0.30618975139523086,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.712223030902668e-06,
+      "loss": 1.49808121,
+      "memory(GiB)": 103.58,
+      "step": 12070,
+      "train_speed(iter/s)": 1.314314
+    },
+    {
+      "acc": 0.68936119,
+      "epoch": 0.3063165905631659,
+      "grad_norm": 3.375,
+      "learning_rate": 9.711872307021594e-06,
+      "loss": 1.39755945,
+      "memory(GiB)": 103.58,
+      "step": 12075,
+      "train_speed(iter/s)": 1.314375
+    },
+    {
+      "acc": 0.68133945,
+      "epoch": 0.30644342973110095,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.711521375891339e-06,
+      "loss": 1.45846176,
+      "memory(GiB)": 103.58,
+      "step": 12080,
+      "train_speed(iter/s)": 1.314436
+    },
+    {
+      "acc": 0.68538923,
+      "epoch": 0.30657026889903605,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.71117023752734e-06,
+      "loss": 1.43656292,
+      "memory(GiB)": 103.58,
+      "step": 12085,
+      "train_speed(iter/s)": 1.314493
+    },
+    {
+      "acc": 0.67518415,
+      "epoch": 0.3066971080669711,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.710818891945037e-06,
+      "loss": 1.38579855,
+      "memory(GiB)": 103.58,
+      "step": 12090,
+      "train_speed(iter/s)": 1.314542
+    },
+    {
+      "acc": 0.68935146,
+      "epoch": 0.30682394723490614,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.71046733915989e-06,
+      "loss": 1.3779726,
+      "memory(GiB)": 103.58,
+      "step": 12095,
+      "train_speed(iter/s)": 1.314592
+    },
+    {
+      "acc": 0.6832057,
+      "epoch": 0.3069507864028412,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.710115579187356e-06,
+      "loss": 1.42827978,
+      "memory(GiB)": 103.58,
+      "step": 12100,
+      "train_speed(iter/s)": 1.314652
+    },
+    {
+      "acc": 0.6747457,
+      "epoch": 0.3070776255707763,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.709763612042911e-06,
+      "loss": 1.44994602,
+      "memory(GiB)": 103.58,
+      "step": 12105,
+      "train_speed(iter/s)": 1.314712
+    },
+    {
+      "acc": 0.68288522,
+      "epoch": 0.3072044647387113,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.709411437742035e-06,
+      "loss": 1.3345561,
+      "memory(GiB)": 103.58,
+      "step": 12110,
+      "train_speed(iter/s)": 1.314769
+    },
+    {
+      "acc": 0.68991551,
+      "epoch": 0.30733130390664637,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.709059056300218e-06,
+      "loss": 1.36811934,
+      "memory(GiB)": 103.58,
+      "step": 12115,
+      "train_speed(iter/s)": 1.314827
+    },
+    {
+      "acc": 0.68177204,
+      "epoch": 0.3074581430745814,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.708706467732958e-06,
+      "loss": 1.39175968,
+      "memory(GiB)": 103.58,
+      "step": 12120,
+      "train_speed(iter/s)": 1.314882
+    },
+    {
+      "acc": 0.68296967,
+      "epoch": 0.3075849822425165,
+      "grad_norm": 3.125,
+      "learning_rate": 9.708353672055766e-06,
+      "loss": 1.46997204,
+      "memory(GiB)": 103.58,
+      "step": 12125,
+      "train_speed(iter/s)": 1.314943
+    },
+    {
+      "acc": 0.67745476,
+      "epoch": 0.30771182141045156,
+      "grad_norm": 3.375,
+      "learning_rate": 9.708000669284158e-06,
+      "loss": 1.46234732,
+      "memory(GiB)": 103.58,
+      "step": 12130,
+      "train_speed(iter/s)": 1.315002
+    },
+    {
+      "acc": 0.69118218,
+      "epoch": 0.3078386605783866,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.707647459433661e-06,
+      "loss": 1.42247791,
+      "memory(GiB)": 103.58,
+      "step": 12135,
+      "train_speed(iter/s)": 1.315062
+    },
+    {
+      "acc": 0.68420935,
+      "epoch": 0.30796549974632165,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.707294042519808e-06,
+      "loss": 1.39925022,
+      "memory(GiB)": 103.58,
+      "step": 12140,
+      "train_speed(iter/s)": 1.31512
+    },
+    {
+      "acc": 0.68936882,
+      "epoch": 0.30809233891425675,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.70694041855815e-06,
+      "loss": 1.49211407,
+      "memory(GiB)": 103.58,
+      "step": 12145,
+      "train_speed(iter/s)": 1.315179
+    },
+    {
+      "acc": 0.6793901,
+      "epoch": 0.3082191780821918,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.706586587564236e-06,
+      "loss": 1.47697735,
+      "memory(GiB)": 103.58,
+      "step": 12150,
+      "train_speed(iter/s)": 1.315233
+    },
+    {
+      "acc": 0.67701693,
+      "epoch": 0.30834601725012684,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.70623254955363e-06,
+      "loss": 1.41851053,
+      "memory(GiB)": 103.58,
+      "step": 12155,
+      "train_speed(iter/s)": 1.315288
+    },
+    {
+      "acc": 0.6891768,
+      "epoch": 0.3084728564180619,
+      "grad_norm": 4.5,
+      "learning_rate": 9.705878304541905e-06,
+      "loss": 1.3411973,
+      "memory(GiB)": 103.58,
+      "step": 12160,
+      "train_speed(iter/s)": 1.315349
+    },
+    {
+      "acc": 0.68265829,
+      "epoch": 0.308599695585997,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.705523852544643e-06,
+      "loss": 1.3939044,
+      "memory(GiB)": 103.58,
+      "step": 12165,
+      "train_speed(iter/s)": 1.315407
+    },
+    {
+      "acc": 0.67093191,
+      "epoch": 0.308726534753932,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.705169193577434e-06,
+      "loss": 1.50192852,
+      "memory(GiB)": 103.58,
+      "step": 12170,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.69697943,
+      "epoch": 0.30885337392186707,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.704814327655874e-06,
+      "loss": 1.39934902,
+      "memory(GiB)": 103.58,
+      "step": 12175,
+      "train_speed(iter/s)": 1.315527
+    },
+    {
+      "acc": 0.69821167,
+      "epoch": 0.3089802130898021,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.704459254795575e-06,
+      "loss": 1.38301144,
+      "memory(GiB)": 103.58,
+      "step": 12180,
+      "train_speed(iter/s)": 1.315584
+    },
+    {
+      "acc": 0.66800179,
+      "epoch": 0.3091070522577372,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.704103975012155e-06,
+      "loss": 1.44489899,
+      "memory(GiB)": 103.58,
+      "step": 12185,
+      "train_speed(iter/s)": 1.315645
+    },
+    {
+      "acc": 0.68642216,
+      "epoch": 0.30923389142567226,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.703748488321238e-06,
+      "loss": 1.36050358,
+      "memory(GiB)": 103.58,
+      "step": 12190,
+      "train_speed(iter/s)": 1.315703
+    },
+    {
+      "acc": 0.6974359,
+      "epoch": 0.3093607305936073,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.703392794738464e-06,
+      "loss": 1.35445251,
+      "memory(GiB)": 103.58,
+      "step": 12195,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.68185673,
+      "epoch": 0.30948756976154235,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.703036894279476e-06,
+      "loss": 1.43791742,
+      "memory(GiB)": 103.58,
+      "step": 12200,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.70159516,
+      "epoch": 0.30961440892947745,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.702680786959925e-06,
+      "loss": 1.33150187,
+      "memory(GiB)": 103.58,
+      "step": 12205,
+      "train_speed(iter/s)": 1.315848
+    },
+    {
+      "acc": 0.6909482,
+      "epoch": 0.3097412480974125,
+      "grad_norm": 2.875,
+      "learning_rate": 9.702324472795481e-06,
+      "loss": 1.40938091,
+      "memory(GiB)": 103.58,
+      "step": 12210,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.67703338,
+      "epoch": 0.30986808726534754,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.70196795180181e-06,
+      "loss": 1.42846956,
+      "memory(GiB)": 103.58,
+      "step": 12215,
+      "train_speed(iter/s)": 1.315963
+    },
+    {
+      "acc": 0.6837697,
+      "epoch": 0.3099949264332826,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.701611223994596e-06,
+      "loss": 1.43263321,
+      "memory(GiB)": 103.58,
+      "step": 12220,
+      "train_speed(iter/s)": 1.31602
+    },
+    {
+      "acc": 0.68712788,
+      "epoch": 0.3101217656012177,
+      "grad_norm": 2.75,
+      "learning_rate": 9.701254289389529e-06,
+      "loss": 1.37322483,
+      "memory(GiB)": 103.58,
+      "step": 12225,
+      "train_speed(iter/s)": 1.316079
+    },
+    {
+      "acc": 0.68548689,
+      "epoch": 0.3102486047691527,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.700897148002308e-06,
+      "loss": 1.43592968,
+      "memory(GiB)": 103.58,
+      "step": 12230,
+      "train_speed(iter/s)": 1.316138
+    },
+    {
+      "acc": 0.68106394,
+      "epoch": 0.31037544393708777,
+      "grad_norm": 3.0,
+      "learning_rate": 9.700539799848645e-06,
+      "loss": 1.45099602,
+      "memory(GiB)": 103.58,
+      "step": 12235,
+      "train_speed(iter/s)": 1.3162
+    },
+    {
+      "acc": 0.68113508,
+      "epoch": 0.3105022831050228,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.700182244944252e-06,
+      "loss": 1.40697861,
+      "memory(GiB)": 103.58,
+      "step": 12240,
+      "train_speed(iter/s)": 1.31626
+    },
+    {
+      "acc": 0.67560616,
+      "epoch": 0.3106291222729579,
+      "grad_norm": 3.0,
+      "learning_rate": 9.69982448330486e-06,
+      "loss": 1.45037918,
+      "memory(GiB)": 103.58,
+      "step": 12245,
+      "train_speed(iter/s)": 1.316321
+    },
+    {
+      "acc": 0.67244358,
+      "epoch": 0.31075596144089296,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.699466514946206e-06,
+      "loss": 1.5311182,
+      "memory(GiB)": 103.58,
+      "step": 12250,
+      "train_speed(iter/s)": 1.316379
+    },
+    {
+      "acc": 0.7001852,
+      "epoch": 0.310882800608828,
+      "grad_norm": 3.625,
+      "learning_rate": 9.699108339884032e-06,
+      "loss": 1.34744301,
+      "memory(GiB)": 103.58,
+      "step": 12255,
+      "train_speed(iter/s)": 1.316437
+    },
+    {
+      "acc": 0.66857457,
+      "epoch": 0.31100963977676305,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.698749958134093e-06,
+      "loss": 1.49382019,
+      "memory(GiB)": 103.58,
+      "step": 12260,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.67169843,
+      "epoch": 0.31113647894469815,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.698391369712152e-06,
+      "loss": 1.49703274,
+      "memory(GiB)": 103.58,
+      "step": 12265,
+      "train_speed(iter/s)": 1.316558
+    },
+    {
+      "acc": 0.67646294,
+      "epoch": 0.3112633181126332,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.698032574633982e-06,
+      "loss": 1.41688557,
+      "memory(GiB)": 103.58,
+      "step": 12270,
+      "train_speed(iter/s)": 1.316618
+    },
+    {
+      "acc": 0.67809834,
+      "epoch": 0.31139015728056824,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.697673572915364e-06,
+      "loss": 1.40882835,
+      "memory(GiB)": 103.58,
+      "step": 12275,
+      "train_speed(iter/s)": 1.316677
+    },
+    {
+      "acc": 0.67666287,
+      "epoch": 0.3115169964485033,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.697314364572087e-06,
+      "loss": 1.43289146,
+      "memory(GiB)": 103.58,
+      "step": 12280,
+      "train_speed(iter/s)": 1.316737
+    },
+    {
+      "acc": 0.68821354,
+      "epoch": 0.3116438356164384,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.696954949619955e-06,
+      "loss": 1.45686512,
+      "memory(GiB)": 103.58,
+      "step": 12285,
+      "train_speed(iter/s)": 1.316797
+    },
+    {
+      "acc": 0.68688822,
+      "epoch": 0.3117706747843734,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.696595328074774e-06,
+      "loss": 1.39289742,
+      "memory(GiB)": 103.58,
+      "step": 12290,
+      "train_speed(iter/s)": 1.316859
+    },
+    {
+      "acc": 0.69069328,
+      "epoch": 0.31189751395230847,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.69623549995236e-06,
+      "loss": 1.42972755,
+      "memory(GiB)": 103.58,
+      "step": 12295,
+      "train_speed(iter/s)": 1.316921
+    },
+    {
+      "acc": 0.67230015,
+      "epoch": 0.3120243531202435,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.695875465268543e-06,
+      "loss": 1.47167912,
+      "memory(GiB)": 103.58,
+      "step": 12300,
+      "train_speed(iter/s)": 1.316981
+    },
+    {
+      "acc": 0.68530879,
+      "epoch": 0.3121511922881786,
+      "grad_norm": 4.0,
+      "learning_rate": 9.695515224039156e-06,
+      "loss": 1.44340658,
+      "memory(GiB)": 103.58,
+      "step": 12305,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.68359423,
+      "epoch": 0.31227803145611366,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.695154776280047e-06,
+      "loss": 1.38178415,
+      "memory(GiB)": 103.58,
+      "step": 12310,
+      "train_speed(iter/s)": 1.317108
+    },
+    {
+      "acc": 0.68181252,
+      "epoch": 0.3124048706240487,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.694794122007067e-06,
+      "loss": 1.45760765,
+      "memory(GiB)": 103.58,
+      "step": 12315,
+      "train_speed(iter/s)": 1.317167
+    },
+    {
+      "acc": 0.68719339,
+      "epoch": 0.31253170979198375,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.694433261236083e-06,
+      "loss": 1.40486488,
+      "memory(GiB)": 103.58,
+      "step": 12320,
+      "train_speed(iter/s)": 1.31723
+    },
+    {
+      "acc": 0.68268175,
+      "epoch": 0.31265854895991885,
+      "grad_norm": 3.0,
+      "learning_rate": 9.694072193982962e-06,
+      "loss": 1.46527042,
+      "memory(GiB)": 103.58,
+      "step": 12325,
+      "train_speed(iter/s)": 1.317292
+    },
+    {
+      "acc": 0.68213224,
+      "epoch": 0.3127853881278539,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.69371092026359e-06,
+      "loss": 1.43906384,
+      "memory(GiB)": 103.58,
+      "step": 12330,
+      "train_speed(iter/s)": 1.317353
+    },
+    {
+      "acc": 0.70958004,
+      "epoch": 0.31291222729578894,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.693349440093855e-06,
+      "loss": 1.36095591,
+      "memory(GiB)": 103.58,
+      "step": 12335,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.6949635,
+      "epoch": 0.313039066463724,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.69298775348966e-06,
+      "loss": 1.31893253,
+      "memory(GiB)": 103.58,
+      "step": 12340,
+      "train_speed(iter/s)": 1.317465
+    },
+    {
+      "acc": 0.68084049,
+      "epoch": 0.3131659056316591,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.69262586046691e-06,
+      "loss": 1.42202053,
+      "memory(GiB)": 103.58,
+      "step": 12345,
+      "train_speed(iter/s)": 1.317526
+    },
+    {
+      "acc": 0.70036592,
+      "epoch": 0.3132927447995941,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.692263761041521e-06,
+      "loss": 1.4159339,
+      "memory(GiB)": 103.58,
+      "step": 12350,
+      "train_speed(iter/s)": 1.31759
+    },
+    {
+      "acc": 0.68163605,
+      "epoch": 0.31341958396752917,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.691901455229425e-06,
+      "loss": 1.39514809,
+      "memory(GiB)": 103.58,
+      "step": 12355,
+      "train_speed(iter/s)": 1.31765
+    },
+    {
+      "acc": 0.68205833,
+      "epoch": 0.3135464231354642,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.691538943046552e-06,
+      "loss": 1.43387251,
+      "memory(GiB)": 103.58,
+      "step": 12360,
+      "train_speed(iter/s)": 1.317715
+    },
+    {
+      "acc": 0.68377624,
+      "epoch": 0.3136732623033993,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.691176224508853e-06,
+      "loss": 1.44979496,
+      "memory(GiB)": 103.58,
+      "step": 12365,
+      "train_speed(iter/s)": 1.317775
+    },
+    {
+      "acc": 0.68447361,
+      "epoch": 0.31380010147133436,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.690813299632278e-06,
+      "loss": 1.37071371,
+      "memory(GiB)": 103.58,
+      "step": 12370,
+      "train_speed(iter/s)": 1.317839
+    },
+    {
+      "acc": 0.68357158,
+      "epoch": 0.3139269406392694,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.690450168432793e-06,
+      "loss": 1.43228607,
+      "memory(GiB)": 103.58,
+      "step": 12375,
+      "train_speed(iter/s)": 1.317901
+    },
+    {
+      "acc": 0.67361755,
+      "epoch": 0.31405377980720445,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.690086830926366e-06,
+      "loss": 1.42127094,
+      "memory(GiB)": 103.58,
+      "step": 12380,
+      "train_speed(iter/s)": 1.317966
+    },
+    {
+      "acc": 0.66168289,
+      "epoch": 0.31418061897513955,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.689723287128981e-06,
+      "loss": 1.45612268,
+      "memory(GiB)": 103.58,
+      "step": 12385,
+      "train_speed(iter/s)": 1.318028
+    },
+    {
+      "acc": 0.68345747,
+      "epoch": 0.3143074581430746,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.689359537056628e-06,
+      "loss": 1.40988922,
+      "memory(GiB)": 103.58,
+      "step": 12390,
+      "train_speed(iter/s)": 1.318092
+    },
+    {
+      "acc": 0.68116822,
+      "epoch": 0.31443429731100964,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.688995580725304e-06,
+      "loss": 1.37664557,
+      "memory(GiB)": 103.58,
+      "step": 12395,
+      "train_speed(iter/s)": 1.318156
+    },
+    {
+      "acc": 0.69140215,
+      "epoch": 0.3145611364789447,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.688631418151022e-06,
+      "loss": 1.34995079,
+      "memory(GiB)": 103.58,
+      "step": 12400,
+      "train_speed(iter/s)": 1.318215
+    },
+    {
+      "acc": 0.69188066,
+      "epoch": 0.3146879756468798,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.688267049349796e-06,
+      "loss": 1.38404675,
+      "memory(GiB)": 103.58,
+      "step": 12405,
+      "train_speed(iter/s)": 1.318277
+    },
+    {
+      "acc": 0.67555652,
+      "epoch": 0.3148148148148148,
+      "grad_norm": 3.0,
+      "learning_rate": 9.687902474337654e-06,
+      "loss": 1.50037575,
+      "memory(GiB)": 103.58,
+      "step": 12410,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "acc": 0.68460913,
+      "epoch": 0.31494165398274987,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.687537693130631e-06,
+      "loss": 1.38251419,
+      "memory(GiB)": 103.58,
+      "step": 12415,
+      "train_speed(iter/s)": 1.3184
+    },
+    {
+      "acc": 0.67380099,
+      "epoch": 0.3150684931506849,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.687172705744773e-06,
+      "loss": 1.45886412,
+      "memory(GiB)": 103.58,
+      "step": 12420,
+      "train_speed(iter/s)": 1.31846
+    },
+    {
+      "acc": 0.66226826,
+      "epoch": 0.31519533231862,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.686807512196132e-06,
+      "loss": 1.48117924,
+      "memory(GiB)": 103.58,
+      "step": 12425,
+      "train_speed(iter/s)": 1.31852
+    },
+    {
+      "acc": 0.68475647,
+      "epoch": 0.31532217148655506,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.68644211250077e-06,
+      "loss": 1.39193592,
+      "memory(GiB)": 103.58,
+      "step": 12430,
+      "train_speed(iter/s)": 1.318582
+    },
+    {
+      "acc": 0.66392331,
+      "epoch": 0.3154490106544901,
+      "grad_norm": 3.5,
+      "learning_rate": 9.686076506674761e-06,
+      "loss": 1.4580946,
+      "memory(GiB)": 103.58,
+      "step": 12435,
+      "train_speed(iter/s)": 1.318643
+    },
+    {
+      "acc": 0.66910906,
+      "epoch": 0.31557584982242515,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.685710694734187e-06,
+      "loss": 1.43788509,
+      "memory(GiB)": 103.58,
+      "step": 12440,
+      "train_speed(iter/s)": 1.318702
+    },
+    {
+      "acc": 0.68206358,
+      "epoch": 0.31570268899036025,
+      "grad_norm": 3.25,
+      "learning_rate": 9.685344676695135e-06,
+      "loss": 1.40816727,
+      "memory(GiB)": 103.58,
+      "step": 12445,
+      "train_speed(iter/s)": 1.318765
+    },
+    {
+      "acc": 0.67978139,
+      "epoch": 0.3158295281582953,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.684978452573706e-06,
+      "loss": 1.44237442,
+      "memory(GiB)": 103.58,
+      "step": 12450,
+      "train_speed(iter/s)": 1.318826
+    },
+    {
+      "acc": 0.69154034,
+      "epoch": 0.31595636732623034,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.684612022386008e-06,
+      "loss": 1.34683933,
+      "memory(GiB)": 103.58,
+      "step": 12455,
+      "train_speed(iter/s)": 1.318872
+    },
+    {
+      "acc": 0.67872143,
+      "epoch": 0.3160832064941654,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.68424538614816e-06,
+      "loss": 1.4418479,
+      "memory(GiB)": 103.58,
+      "step": 12460,
+      "train_speed(iter/s)": 1.318934
+    },
+    {
+      "acc": 0.68056192,
+      "epoch": 0.3162100456621005,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.683878543876281e-06,
+      "loss": 1.43249683,
+      "memory(GiB)": 103.58,
+      "step": 12465,
+      "train_speed(iter/s)": 1.318996
+    },
+    {
+      "acc": 0.68974457,
+      "epoch": 0.3163368848300355,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.683511495586516e-06,
+      "loss": 1.3805892,
+      "memory(GiB)": 103.58,
+      "step": 12470,
+      "train_speed(iter/s)": 1.319057
+    },
+    {
+      "acc": 0.68872771,
+      "epoch": 0.31646372399797057,
+      "grad_norm": 3.0,
+      "learning_rate": 9.683144241295003e-06,
+      "loss": 1.35925093,
+      "memory(GiB)": 103.58,
+      "step": 12475,
+      "train_speed(iter/s)": 1.319106
+    },
+    {
+      "acc": 0.68259754,
+      "epoch": 0.3165905631659056,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.682776781017899e-06,
+      "loss": 1.4079752,
+      "memory(GiB)": 103.58,
+      "step": 12480,
+      "train_speed(iter/s)": 1.319166
+    },
+    {
+      "acc": 0.67111511,
+      "epoch": 0.3167174023338407,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.682409114771364e-06,
+      "loss": 1.44368992,
+      "memory(GiB)": 103.58,
+      "step": 12485,
+      "train_speed(iter/s)": 1.319218
+    },
+    {
+      "acc": 0.66696715,
+      "epoch": 0.31684424150177576,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.682041242571571e-06,
+      "loss": 1.41707659,
+      "memory(GiB)": 103.58,
+      "step": 12490,
+      "train_speed(iter/s)": 1.319279
+    },
+    {
+      "acc": 0.6854641,
+      "epoch": 0.3169710806697108,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.681673164434701e-06,
+      "loss": 1.43747349,
+      "memory(GiB)": 103.58,
+      "step": 12495,
+      "train_speed(iter/s)": 1.31934
+    },
+    {
+      "acc": 0.69668741,
+      "epoch": 0.31709791983764585,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.681304880376942e-06,
+      "loss": 1.38131084,
+      "memory(GiB)": 103.58,
+      "step": 12500,
+      "train_speed(iter/s)": 1.3194
+    },
+    {
+      "acc": 0.67805634,
+      "epoch": 0.31722475900558095,
+      "grad_norm": 3.0,
+      "learning_rate": 9.680936390414495e-06,
+      "loss": 1.51082039,
+      "memory(GiB)": 103.58,
+      "step": 12505,
+      "train_speed(iter/s)": 1.319461
+    },
+    {
+      "acc": 0.68025656,
+      "epoch": 0.317351598173516,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.680567694563566e-06,
+      "loss": 1.42461996,
+      "memory(GiB)": 103.58,
+      "step": 12510,
+      "train_speed(iter/s)": 1.31952
+    },
+    {
+      "acc": 0.68717461,
+      "epoch": 0.31747843734145104,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.680198792840371e-06,
+      "loss": 1.42069502,
+      "memory(GiB)": 103.58,
+      "step": 12515,
+      "train_speed(iter/s)": 1.319577
+    },
+    {
+      "acc": 0.67235317,
+      "epoch": 0.3176052765093861,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.67982968526114e-06,
+      "loss": 1.41346207,
+      "memory(GiB)": 103.58,
+      "step": 12520,
+      "train_speed(iter/s)": 1.319636
+    },
+    {
+      "acc": 0.694665,
+      "epoch": 0.3177321156773212,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.679460371842104e-06,
+      "loss": 1.35996389,
+      "memory(GiB)": 103.58,
+      "step": 12525,
+      "train_speed(iter/s)": 1.319692
+    },
+    {
+      "acc": 0.67447939,
+      "epoch": 0.3178589548452562,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.679090852599508e-06,
+      "loss": 1.431633,
+      "memory(GiB)": 103.58,
+      "step": 12530,
+      "train_speed(iter/s)": 1.319749
+    },
+    {
+      "acc": 0.67421565,
+      "epoch": 0.31798579401319127,
+      "grad_norm": 3.125,
+      "learning_rate": 9.678721127549608e-06,
+      "loss": 1.44905224,
+      "memory(GiB)": 103.58,
+      "step": 12535,
+      "train_speed(iter/s)": 1.319805
+    },
+    {
+      "acc": 0.66589355,
+      "epoch": 0.3181126331811263,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.678351196708662e-06,
+      "loss": 1.53565359,
+      "memory(GiB)": 103.58,
+      "step": 12540,
+      "train_speed(iter/s)": 1.319866
+    },
+    {
+      "acc": 0.68080897,
+      "epoch": 0.3182394723490614,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.677981060092943e-06,
+      "loss": 1.46287365,
+      "memory(GiB)": 103.58,
+      "step": 12545,
+      "train_speed(iter/s)": 1.319922
+    },
+    {
+      "acc": 0.66408386,
+      "epoch": 0.31836631151699646,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.677610717718732e-06,
+      "loss": 1.48941193,
+      "memory(GiB)": 103.58,
+      "step": 12550,
+      "train_speed(iter/s)": 1.319978
+    },
+    {
+      "acc": 0.65710783,
+      "epoch": 0.3184931506849315,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.677240169602317e-06,
+      "loss": 1.50266981,
+      "memory(GiB)": 103.58,
+      "step": 12555,
+      "train_speed(iter/s)": 1.320036
+    },
+    {
+      "acc": 0.68037677,
+      "epoch": 0.31861998985286655,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.676869415759999e-06,
+      "loss": 1.46036482,
+      "memory(GiB)": 103.58,
+      "step": 12560,
+      "train_speed(iter/s)": 1.320086
+    },
+    {
+      "acc": 0.66503887,
+      "epoch": 0.31874682902080165,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.67649845620808e-06,
+      "loss": 1.50657091,
+      "memory(GiB)": 103.58,
+      "step": 12565,
+      "train_speed(iter/s)": 1.320146
+    },
+    {
+      "acc": 0.67930326,
+      "epoch": 0.3188736681887367,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.676127290962883e-06,
+      "loss": 1.46083088,
+      "memory(GiB)": 103.58,
+      "step": 12570,
+      "train_speed(iter/s)": 1.320201
+    },
+    {
+      "acc": 0.67400565,
+      "epoch": 0.31900050735667174,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.675755920040728e-06,
+      "loss": 1.43120995,
+      "memory(GiB)": 103.58,
+      "step": 12575,
+      "train_speed(iter/s)": 1.320258
+    },
+    {
+      "acc": 0.69316816,
+      "epoch": 0.3191273465246068,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.675384343457954e-06,
+      "loss": 1.40083284,
+      "memory(GiB)": 103.58,
+      "step": 12580,
+      "train_speed(iter/s)": 1.320313
+    },
+    {
+      "acc": 0.67797585,
+      "epoch": 0.3192541856925419,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.675012561230901e-06,
+      "loss": 1.46818476,
+      "memory(GiB)": 103.58,
+      "step": 12585,
+      "train_speed(iter/s)": 1.32037
+    },
+    {
+      "acc": 0.68675976,
+      "epoch": 0.3193810248604769,
+      "grad_norm": 3.25,
+      "learning_rate": 9.674640573375924e-06,
+      "loss": 1.44634256,
+      "memory(GiB)": 103.58,
+      "step": 12590,
+      "train_speed(iter/s)": 1.320429
+    },
+    {
+      "acc": 0.68516245,
+      "epoch": 0.31950786402841197,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.674268379909383e-06,
+      "loss": 1.46056299,
+      "memory(GiB)": 103.58,
+      "step": 12595,
+      "train_speed(iter/s)": 1.320485
+    },
+    {
+      "acc": 0.68364964,
+      "epoch": 0.319634703196347,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.67389598084765e-06,
+      "loss": 1.41789036,
+      "memory(GiB)": 103.58,
+      "step": 12600,
+      "train_speed(iter/s)": 1.320536
+    },
+    {
+      "acc": 0.68702655,
+      "epoch": 0.3197615423642821,
+      "grad_norm": 4.0,
+      "learning_rate": 9.673523376207103e-06,
+      "loss": 1.37538042,
+      "memory(GiB)": 103.58,
+      "step": 12605,
+      "train_speed(iter/s)": 1.320595
+    },
+    {
+      "acc": 0.67947764,
+      "epoch": 0.31988838153221716,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.673150566004135e-06,
+      "loss": 1.42423649,
+      "memory(GiB)": 103.58,
+      "step": 12610,
+      "train_speed(iter/s)": 1.320654
+    },
+    {
+      "acc": 0.67261276,
+      "epoch": 0.3200152207001522,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.672777550255137e-06,
+      "loss": 1.50257978,
+      "memory(GiB)": 103.58,
+      "step": 12615,
+      "train_speed(iter/s)": 1.320712
+    },
+    {
+      "acc": 0.68355618,
+      "epoch": 0.32014205986808725,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.672404328976523e-06,
+      "loss": 1.44009285,
+      "memory(GiB)": 103.58,
+      "step": 12620,
+      "train_speed(iter/s)": 1.320771
+    },
+    {
+      "acc": 0.68613863,
+      "epoch": 0.32026889903602235,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.672030902184706e-06,
+      "loss": 1.40223389,
+      "memory(GiB)": 103.58,
+      "step": 12625,
+      "train_speed(iter/s)": 1.320829
+    },
+    {
+      "acc": 0.6822793,
+      "epoch": 0.3203957382039574,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.671657269896108e-06,
+      "loss": 1.4967226,
+      "memory(GiB)": 103.58,
+      "step": 12630,
+      "train_speed(iter/s)": 1.320888
+    },
+    {
+      "acc": 0.68814936,
+      "epoch": 0.32052257737189244,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.671283432127169e-06,
+      "loss": 1.40568972,
+      "memory(GiB)": 103.58,
+      "step": 12635,
+      "train_speed(iter/s)": 1.320947
+    },
+    {
+      "acc": 0.68649483,
+      "epoch": 0.3206494165398275,
+      "grad_norm": 3.0,
+      "learning_rate": 9.670909388894328e-06,
+      "loss": 1.36818047,
+      "memory(GiB)": 103.58,
+      "step": 12640,
+      "train_speed(iter/s)": 1.321007
+    },
+    {
+      "acc": 0.68237152,
+      "epoch": 0.3207762557077626,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.670535140214037e-06,
+      "loss": 1.44337816,
+      "memory(GiB)": 103.58,
+      "step": 12645,
+      "train_speed(iter/s)": 1.321064
+    },
+    {
+      "acc": 0.6805295,
+      "epoch": 0.3209030948756976,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.670160686102759e-06,
+      "loss": 1.42307673,
+      "memory(GiB)": 103.58,
+      "step": 12650,
+      "train_speed(iter/s)": 1.321122
+    },
+    {
+      "acc": 0.6716691,
+      "epoch": 0.32102993404363267,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.669786026576962e-06,
+      "loss": 1.45979233,
+      "memory(GiB)": 103.58,
+      "step": 12655,
+      "train_speed(iter/s)": 1.32118
+    },
+    {
+      "acc": 0.69132528,
+      "epoch": 0.3211567732115677,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.669411161653127e-06,
+      "loss": 1.39090538,
+      "memory(GiB)": 103.58,
+      "step": 12660,
+      "train_speed(iter/s)": 1.321238
+    },
+    {
+      "acc": 0.67459431,
+      "epoch": 0.3212836123795028,
+      "grad_norm": 2.875,
+      "learning_rate": 9.669036091347742e-06,
+      "loss": 1.38879852,
+      "memory(GiB)": 103.58,
+      "step": 12665,
+      "train_speed(iter/s)": 1.321299
+    },
+    {
+      "acc": 0.68647799,
+      "epoch": 0.32141045154743786,
+      "grad_norm": 3.5,
+      "learning_rate": 9.668660815677304e-06,
+      "loss": 1.37750473,
+      "memory(GiB)": 103.58,
+      "step": 12670,
+      "train_speed(iter/s)": 1.321354
+    },
+    {
+      "acc": 0.70642977,
+      "epoch": 0.3215372907153729,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.668285334658319e-06,
+      "loss": 1.38140697,
+      "memory(GiB)": 103.58,
+      "step": 12675,
+      "train_speed(iter/s)": 1.321412
+    },
+    {
+      "acc": 0.67493982,
+      "epoch": 0.32166412988330795,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.667909648307302e-06,
+      "loss": 1.46997261,
+      "memory(GiB)": 103.58,
+      "step": 12680,
+      "train_speed(iter/s)": 1.32147
+    },
+    {
+      "acc": 0.69371538,
+      "epoch": 0.32179096905124305,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.66753375664078e-06,
+      "loss": 1.35765514,
+      "memory(GiB)": 103.58,
+      "step": 12685,
+      "train_speed(iter/s)": 1.321529
+    },
+    {
+      "acc": 0.68206778,
+      "epoch": 0.3219178082191781,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.667157659675284e-06,
+      "loss": 1.434622,
+      "memory(GiB)": 103.58,
+      "step": 12690,
+      "train_speed(iter/s)": 1.321588
+    },
+    {
+      "acc": 0.68361106,
+      "epoch": 0.32204464738711314,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.666781357427355e-06,
+      "loss": 1.43791313,
+      "memory(GiB)": 103.58,
+      "step": 12695,
+      "train_speed(iter/s)": 1.321646
+    },
+    {
+      "acc": 0.68113008,
+      "epoch": 0.3221714865550482,
+      "grad_norm": 3.125,
+      "learning_rate": 9.666404849913546e-06,
+      "loss": 1.40974178,
+      "memory(GiB)": 103.58,
+      "step": 12700,
+      "train_speed(iter/s)": 1.321706
+    },
+    {
+      "acc": 0.69028387,
+      "epoch": 0.3222983257229833,
+      "grad_norm": 3.75,
+      "learning_rate": 9.66602813715042e-06,
+      "loss": 1.38329353,
+      "memory(GiB)": 103.58,
+      "step": 12705,
+      "train_speed(iter/s)": 1.321763
+    },
+    {
+      "acc": 0.69276919,
+      "epoch": 0.3224251648909183,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.665651219154543e-06,
+      "loss": 1.45745697,
+      "memory(GiB)": 103.58,
+      "step": 12710,
+      "train_speed(iter/s)": 1.321819
+    },
+    {
+      "acc": 0.68857999,
+      "epoch": 0.32255200405885337,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.665274095942495e-06,
+      "loss": 1.46388474,
+      "memory(GiB)": 103.58,
+      "step": 12715,
+      "train_speed(iter/s)": 1.321879
+    },
+    {
+      "acc": 0.68470168,
+      "epoch": 0.3226788432267884,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.664896767530862e-06,
+      "loss": 1.45458059,
+      "memory(GiB)": 103.58,
+      "step": 12720,
+      "train_speed(iter/s)": 1.321937
+    },
+    {
+      "acc": 0.6789228,
+      "epoch": 0.3228056823947235,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.664519233936242e-06,
+      "loss": 1.45069418,
+      "memory(GiB)": 103.58,
+      "step": 12725,
+      "train_speed(iter/s)": 1.321994
+    },
+    {
+      "acc": 0.68244848,
+      "epoch": 0.32293252156265856,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.664141495175242e-06,
+      "loss": 1.34855022,
+      "memory(GiB)": 103.58,
+      "step": 12730,
+      "train_speed(iter/s)": 1.32204
+    },
+    {
+      "acc": 0.68822508,
+      "epoch": 0.3230593607305936,
+      "grad_norm": 2.75,
+      "learning_rate": 9.663763551264476e-06,
+      "loss": 1.4269887,
+      "memory(GiB)": 103.58,
+      "step": 12735,
+      "train_speed(iter/s)": 1.322098
+    },
+    {
+      "acc": 0.68722415,
+      "epoch": 0.32318619989852865,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.663385402220565e-06,
+      "loss": 1.38351002,
+      "memory(GiB)": 103.58,
+      "step": 12740,
+      "train_speed(iter/s)": 1.322157
+    },
+    {
+      "acc": 0.69180264,
+      "epoch": 0.32331303906646375,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.663007048060144e-06,
+      "loss": 1.47873468,
+      "memory(GiB)": 103.58,
+      "step": 12745,
+      "train_speed(iter/s)": 1.322216
+    },
+    {
+      "acc": 0.6984127,
+      "epoch": 0.3234398782343988,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.662628488799854e-06,
+      "loss": 1.35473156,
+      "memory(GiB)": 103.58,
+      "step": 12750,
+      "train_speed(iter/s)": 1.322273
+    },
+    {
+      "acc": 0.67651,
+      "epoch": 0.32356671740233384,
+      "grad_norm": 3.375,
+      "learning_rate": 9.662249724456346e-06,
+      "loss": 1.41068306,
+      "memory(GiB)": 103.58,
+      "step": 12755,
+      "train_speed(iter/s)": 1.322333
+    },
+    {
+      "acc": 0.68535185,
+      "epoch": 0.3236935565702689,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.661870755046278e-06,
+      "loss": 1.41712599,
+      "memory(GiB)": 103.58,
+      "step": 12760,
+      "train_speed(iter/s)": 1.32239
+    },
+    {
+      "acc": 0.67639422,
+      "epoch": 0.323820395738204,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.661491580586322e-06,
+      "loss": 1.36070709,
+      "memory(GiB)": 103.58,
+      "step": 12765,
+      "train_speed(iter/s)": 1.322447
+    },
+    {
+      "acc": 0.67846127,
+      "epoch": 0.323947234906139,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.661112201093156e-06,
+      "loss": 1.41868277,
+      "memory(GiB)": 103.58,
+      "step": 12770,
+      "train_speed(iter/s)": 1.322504
+    },
+    {
+      "acc": 0.67977972,
+      "epoch": 0.32407407407407407,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.660732616583463e-06,
+      "loss": 1.46903229,
+      "memory(GiB)": 103.58,
+      "step": 12775,
+      "train_speed(iter/s)": 1.322552
+    },
+    {
+      "acc": 0.69649773,
+      "epoch": 0.3242009132420091,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.660352827073941e-06,
+      "loss": 1.38854761,
+      "memory(GiB)": 103.58,
+      "step": 12780,
+      "train_speed(iter/s)": 1.322608
+    },
+    {
+      "acc": 0.66713762,
+      "epoch": 0.3243277524099442,
+      "grad_norm": 2.875,
+      "learning_rate": 9.659972832581295e-06,
+      "loss": 1.45436993,
+      "memory(GiB)": 103.58,
+      "step": 12785,
+      "train_speed(iter/s)": 1.322665
+    },
+    {
+      "acc": 0.67523537,
+      "epoch": 0.32445459157787926,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.659592633122238e-06,
+      "loss": 1.40017643,
+      "memory(GiB)": 103.58,
+      "step": 12790,
+      "train_speed(iter/s)": 1.322722
+    },
+    {
+      "acc": 0.68135743,
+      "epoch": 0.3245814307458143,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.659212228713495e-06,
+      "loss": 1.43737555,
+      "memory(GiB)": 103.58,
+      "step": 12795,
+      "train_speed(iter/s)": 1.322779
+    },
+    {
+      "acc": 0.68427258,
+      "epoch": 0.32470826991374935,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.658831619371793e-06,
+      "loss": 1.42623987,
+      "memory(GiB)": 103.58,
+      "step": 12800,
+      "train_speed(iter/s)": 1.322836
+    },
+    {
+      "acc": 0.67852907,
+      "epoch": 0.32483510908168445,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.658450805113879e-06,
+      "loss": 1.45946016,
+      "memory(GiB)": 103.58,
+      "step": 12805,
+      "train_speed(iter/s)": 1.322893
+    },
+    {
+      "acc": 0.68953691,
+      "epoch": 0.3249619482496195,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.6580697859565e-06,
+      "loss": 1.37313528,
+      "memory(GiB)": 103.58,
+      "step": 12810,
+      "train_speed(iter/s)": 1.322946
+    },
+    {
+      "acc": 0.70454712,
+      "epoch": 0.32508878741755454,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.657688561916414e-06,
+      "loss": 1.2950182,
+      "memory(GiB)": 103.58,
+      "step": 12815,
+      "train_speed(iter/s)": 1.323003
+    },
+    {
+      "acc": 0.6915659,
+      "epoch": 0.3252156265854896,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.65730713301039e-06,
+      "loss": 1.37618084,
+      "memory(GiB)": 103.58,
+      "step": 12820,
+      "train_speed(iter/s)": 1.32306
+    },
+    {
+      "acc": 0.6796658,
+      "epoch": 0.3253424657534247,
+      "grad_norm": 3.25,
+      "learning_rate": 9.656925499255206e-06,
+      "loss": 1.40961809,
+      "memory(GiB)": 103.58,
+      "step": 12825,
+      "train_speed(iter/s)": 1.323118
+    },
+    {
+      "acc": 0.68632312,
+      "epoch": 0.3254693049213597,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.656543660667646e-06,
+      "loss": 1.37854528,
+      "memory(GiB)": 103.58,
+      "step": 12830,
+      "train_speed(iter/s)": 1.323175
+    },
+    {
+      "acc": 0.69571562,
+      "epoch": 0.32559614408929477,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.656161617264507e-06,
+      "loss": 1.33589277,
+      "memory(GiB)": 103.58,
+      "step": 12835,
+      "train_speed(iter/s)": 1.323233
+    },
+    {
+      "acc": 0.67088623,
+      "epoch": 0.3257229832572298,
+      "grad_norm": 3.625,
+      "learning_rate": 9.65577936906259e-06,
+      "loss": 1.41394577,
+      "memory(GiB)": 103.58,
+      "step": 12840,
+      "train_speed(iter/s)": 1.323283
+    },
+    {
+      "acc": 0.66471071,
+      "epoch": 0.3258498224251649,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.65539691607871e-06,
+      "loss": 1.43362675,
+      "memory(GiB)": 103.58,
+      "step": 12845,
+      "train_speed(iter/s)": 1.32334
+    },
+    {
+      "acc": 0.67338471,
+      "epoch": 0.32597666159309996,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.65501425832969e-06,
+      "loss": 1.47292471,
+      "memory(GiB)": 103.58,
+      "step": 12850,
+      "train_speed(iter/s)": 1.323398
+    },
+    {
+      "acc": 0.68055711,
+      "epoch": 0.326103500761035,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.65463139583236e-06,
+      "loss": 1.41743841,
+      "memory(GiB)": 103.58,
+      "step": 12855,
+      "train_speed(iter/s)": 1.323454
+    },
+    {
+      "acc": 0.67844915,
+      "epoch": 0.32623033992897005,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.65424832860356e-06,
+      "loss": 1.42313881,
+      "memory(GiB)": 103.58,
+      "step": 12860,
+      "train_speed(iter/s)": 1.323511
+    },
+    {
+      "acc": 0.68286619,
+      "epoch": 0.32635717909690515,
+      "grad_norm": 3.0,
+      "learning_rate": 9.653865056660136e-06,
+      "loss": 1.44259806,
+      "memory(GiB)": 103.58,
+      "step": 12865,
+      "train_speed(iter/s)": 1.323567
+    },
+    {
+      "acc": 0.68135424,
+      "epoch": 0.3264840182648402,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.653481580018951e-06,
+      "loss": 1.45052843,
+      "memory(GiB)": 103.58,
+      "step": 12870,
+      "train_speed(iter/s)": 1.323622
+    },
+    {
+      "acc": 0.67376475,
+      "epoch": 0.32661085743277524,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.653097898696869e-06,
+      "loss": 1.47943745,
+      "memory(GiB)": 103.58,
+      "step": 12875,
+      "train_speed(iter/s)": 1.32368
+    },
+    {
+      "acc": 0.67855015,
+      "epoch": 0.3267376966007103,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.652714012710766e-06,
+      "loss": 1.45299358,
+      "memory(GiB)": 103.58,
+      "step": 12880,
+      "train_speed(iter/s)": 1.323734
+    },
+    {
+      "acc": 0.67754855,
+      "epoch": 0.3268645357686454,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.65232992207753e-06,
+      "loss": 1.48792305,
+      "memory(GiB)": 103.58,
+      "step": 12885,
+      "train_speed(iter/s)": 1.32379
+    },
+    {
+      "acc": 0.67892122,
+      "epoch": 0.3269913749365804,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.651945626814052e-06,
+      "loss": 1.46509323,
+      "memory(GiB)": 103.58,
+      "step": 12890,
+      "train_speed(iter/s)": 1.323848
+    },
+    {
+      "acc": 0.68559828,
+      "epoch": 0.32711821410451547,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.651561126937236e-06,
+      "loss": 1.40213814,
+      "memory(GiB)": 103.58,
+      "step": 12895,
+      "train_speed(iter/s)": 1.3239
+    },
+    {
+      "acc": 0.67900949,
+      "epoch": 0.3272450532724505,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.651176422463994e-06,
+      "loss": 1.48282833,
+      "memory(GiB)": 103.58,
+      "step": 12900,
+      "train_speed(iter/s)": 1.323956
+    },
+    {
+      "acc": 0.68942699,
+      "epoch": 0.3273718924403856,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.650791513411246e-06,
+      "loss": 1.35836258,
+      "memory(GiB)": 103.58,
+      "step": 12905,
+      "train_speed(iter/s)": 1.324008
+    },
+    {
+      "acc": 0.67262025,
+      "epoch": 0.32749873160832066,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.650406399795924e-06,
+      "loss": 1.45789232,
+      "memory(GiB)": 103.58,
+      "step": 12910,
+      "train_speed(iter/s)": 1.324063
+    },
+    {
+      "acc": 0.67767591,
+      "epoch": 0.3276255707762557,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.650021081634965e-06,
+      "loss": 1.44825954,
+      "memory(GiB)": 103.58,
+      "step": 12915,
+      "train_speed(iter/s)": 1.324118
+    },
+    {
+      "acc": 0.67489882,
+      "epoch": 0.32775240994419075,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.649635558945318e-06,
+      "loss": 1.43747597,
+      "memory(GiB)": 103.58,
+      "step": 12920,
+      "train_speed(iter/s)": 1.324175
+    },
+    {
+      "acc": 0.67507882,
+      "epoch": 0.32787924911212585,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.649249831743941e-06,
+      "loss": 1.45647831,
+      "memory(GiB)": 103.58,
+      "step": 12925,
+      "train_speed(iter/s)": 1.324231
+    },
+    {
+      "acc": 0.68155775,
+      "epoch": 0.3280060882800609,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.6488639000478e-06,
+      "loss": 1.45696487,
+      "memory(GiB)": 103.58,
+      "step": 12930,
+      "train_speed(iter/s)": 1.324286
+    },
+    {
+      "acc": 0.69436822,
+      "epoch": 0.32813292744799594,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.648477763873868e-06,
+      "loss": 1.41298981,
+      "memory(GiB)": 103.58,
+      "step": 12935,
+      "train_speed(iter/s)": 1.324344
+    },
+    {
+      "acc": 0.69172125,
+      "epoch": 0.328259766615931,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.648091423239128e-06,
+      "loss": 1.36302004,
+      "memory(GiB)": 103.58,
+      "step": 12940,
+      "train_speed(iter/s)": 1.3244
+    },
+    {
+      "acc": 0.68816204,
+      "epoch": 0.3283866057838661,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.647704878160576e-06,
+      "loss": 1.368929,
+      "memory(GiB)": 103.58,
+      "step": 12945,
+      "train_speed(iter/s)": 1.324452
+    },
+    {
+      "acc": 0.66898556,
+      "epoch": 0.3285134449518011,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.647318128655213e-06,
+      "loss": 1.47056589,
+      "memory(GiB)": 103.58,
+      "step": 12950,
+      "train_speed(iter/s)": 1.324507
+    },
+    {
+      "acc": 0.69919701,
+      "epoch": 0.32864028411973617,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.64693117474005e-06,
+      "loss": 1.35862427,
+      "memory(GiB)": 103.58,
+      "step": 12955,
+      "train_speed(iter/s)": 1.324559
+    },
+    {
+      "acc": 0.68703537,
+      "epoch": 0.3287671232876712,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.646544016432109e-06,
+      "loss": 1.40418558,
+      "memory(GiB)": 103.58,
+      "step": 12960,
+      "train_speed(iter/s)": 1.324609
+    },
+    {
+      "acc": 0.69234858,
+      "epoch": 0.3288939624556063,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.646156653748415e-06,
+      "loss": 1.38478441,
+      "memory(GiB)": 103.58,
+      "step": 12965,
+      "train_speed(iter/s)": 1.324665
+    },
+    {
+      "acc": 0.68936119,
+      "epoch": 0.32902080162354136,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.645769086706008e-06,
+      "loss": 1.46246271,
+      "memory(GiB)": 103.58,
+      "step": 12970,
+      "train_speed(iter/s)": 1.324719
+    },
+    {
+      "acc": 0.68722153,
+      "epoch": 0.3291476407914764,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.645381315321934e-06,
+      "loss": 1.42210321,
+      "memory(GiB)": 103.58,
+      "step": 12975,
+      "train_speed(iter/s)": 1.324774
+    },
+    {
+      "acc": 0.67871947,
+      "epoch": 0.32927447995941145,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.64499333961325e-06,
+      "loss": 1.4610198,
+      "memory(GiB)": 103.58,
+      "step": 12980,
+      "train_speed(iter/s)": 1.324829
+    },
+    {
+      "acc": 0.69772272,
+      "epoch": 0.32940131912734655,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.64460515959702e-06,
+      "loss": 1.31917028,
+      "memory(GiB)": 103.58,
+      "step": 12985,
+      "train_speed(iter/s)": 1.324885
+    },
+    {
+      "acc": 0.67469606,
+      "epoch": 0.3295281582952816,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.64421677529032e-06,
+      "loss": 1.43970976,
+      "memory(GiB)": 103.58,
+      "step": 12990,
+      "train_speed(iter/s)": 1.32494
+    },
+    {
+      "acc": 0.69671397,
+      "epoch": 0.32965499746321664,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.64382818671023e-06,
+      "loss": 1.34917059,
+      "memory(GiB)": 103.58,
+      "step": 12995,
+      "train_speed(iter/s)": 1.324993
+    },
+    {
+      "acc": 0.69387283,
+      "epoch": 0.3297818366311517,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.643439393873844e-06,
+      "loss": 1.36491146,
+      "memory(GiB)": 103.58,
+      "step": 13000,
+      "train_speed(iter/s)": 1.325042
+    },
+    {
+      "epoch": 0.3297818366311517,
+      "eval_acc": 0.6712738876115416,
+      "eval_loss": 1.3780618906021118,
+      "eval_runtime": 69.1147,
+      "eval_samples_per_second": 92.166,
+      "eval_steps_per_second": 23.049,
+      "step": 13000
+    },
+    {
+      "acc": 0.6962584,
+      "epoch": 0.3299086757990868,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.643050396798262e-06,
+      "loss": 1.43750772,
+      "memory(GiB)": 103.58,
+      "step": 13005,
+      "train_speed(iter/s)": 1.314158
+    },
+    {
+      "acc": 0.69663563,
+      "epoch": 0.3300355149670218,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.642661195500593e-06,
+      "loss": 1.31615257,
+      "memory(GiB)": 103.58,
+      "step": 13010,
+      "train_speed(iter/s)": 1.314217
+    },
+    {
+      "acc": 0.67992096,
+      "epoch": 0.33016235413495687,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.642271789997956e-06,
+      "loss": 1.41814108,
+      "memory(GiB)": 103.58,
+      "step": 13015,
+      "train_speed(iter/s)": 1.314273
+    },
+    {
+      "acc": 0.68312998,
+      "epoch": 0.3302891933028919,
+      "grad_norm": 2.625,
+      "learning_rate": 9.64188218030748e-06,
+      "loss": 1.33970976,
+      "memory(GiB)": 103.58,
+      "step": 13020,
+      "train_speed(iter/s)": 1.31433
+    },
+    {
+      "acc": 0.69069505,
+      "epoch": 0.330416032470827,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.641492366446301e-06,
+      "loss": 1.40943813,
+      "memory(GiB)": 103.58,
+      "step": 13025,
+      "train_speed(iter/s)": 1.314386
+    },
+    {
+      "acc": 0.6834549,
+      "epoch": 0.33054287163876206,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.641102348431565e-06,
+      "loss": 1.39937782,
+      "memory(GiB)": 103.58,
+      "step": 13030,
+      "train_speed(iter/s)": 1.314443
+    },
+    {
+      "acc": 0.67296295,
+      "epoch": 0.3306697108066971,
+      "grad_norm": 3.0,
+      "learning_rate": 9.640712126280429e-06,
+      "loss": 1.42477846,
+      "memory(GiB)": 103.58,
+      "step": 13035,
+      "train_speed(iter/s)": 1.314502
+    },
+    {
+      "acc": 0.68280172,
+      "epoch": 0.33079654997463215,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.640321700010053e-06,
+      "loss": 1.39759407,
+      "memory(GiB)": 103.58,
+      "step": 13040,
+      "train_speed(iter/s)": 1.314548
+    },
+    {
+      "acc": 0.68275647,
+      "epoch": 0.33092338914256725,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.63993106963761e-06,
+      "loss": 1.42882996,
+      "memory(GiB)": 103.58,
+      "step": 13045,
+      "train_speed(iter/s)": 1.314607
+    },
+    {
+      "acc": 0.67983732,
+      "epoch": 0.3310502283105023,
+      "grad_norm": 2.75,
+      "learning_rate": 9.639540235180283e-06,
+      "loss": 1.41104069,
+      "memory(GiB)": 103.58,
+      "step": 13050,
+      "train_speed(iter/s)": 1.314664
+    },
+    {
+      "acc": 0.69362268,
+      "epoch": 0.33117706747843734,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.639149196655263e-06,
+      "loss": 1.37143917,
+      "memory(GiB)": 103.58,
+      "step": 13055,
+      "train_speed(iter/s)": 1.314724
+    },
+    {
+      "acc": 0.67486324,
+      "epoch": 0.3313039066463724,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.638757954079749e-06,
+      "loss": 1.41074781,
+      "memory(GiB)": 103.58,
+      "step": 13060,
+      "train_speed(iter/s)": 1.31478
+    },
+    {
+      "acc": 0.69070964,
+      "epoch": 0.3314307458143075,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.638366507470948e-06,
+      "loss": 1.3738925,
+      "memory(GiB)": 103.58,
+      "step": 13065,
+      "train_speed(iter/s)": 1.314838
+    },
+    {
+      "acc": 0.69327121,
+      "epoch": 0.3315575849822425,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.637974856846082e-06,
+      "loss": 1.40082788,
+      "memory(GiB)": 103.58,
+      "step": 13070,
+      "train_speed(iter/s)": 1.314894
+    },
+    {
+      "acc": 0.69429874,
+      "epoch": 0.33168442415017757,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.637583002222373e-06,
+      "loss": 1.33612309,
+      "memory(GiB)": 103.58,
+      "step": 13075,
+      "train_speed(iter/s)": 1.314944
+    },
+    {
+      "acc": 0.68622046,
+      "epoch": 0.3318112633181126,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.637190943617059e-06,
+      "loss": 1.406462,
+      "memory(GiB)": 103.58,
+      "step": 13080,
+      "train_speed(iter/s)": 1.315002
+    },
+    {
+      "acc": 0.67864308,
+      "epoch": 0.3319381024860477,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.636798681047383e-06,
+      "loss": 1.38253479,
+      "memory(GiB)": 103.58,
+      "step": 13085,
+      "train_speed(iter/s)": 1.31506
+    },
+    {
+      "acc": 0.70230999,
+      "epoch": 0.33206494165398276,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.6364062145306e-06,
+      "loss": 1.38139601,
+      "memory(GiB)": 103.58,
+      "step": 13090,
+      "train_speed(iter/s)": 1.315119
+    },
+    {
+      "acc": 0.68688927,
+      "epoch": 0.3321917808219178,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.636013544083971e-06,
+      "loss": 1.41858959,
+      "memory(GiB)": 103.58,
+      "step": 13095,
+      "train_speed(iter/s)": 1.315166
+    },
+    {
+      "acc": 0.68239698,
+      "epoch": 0.33231861998985285,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.635620669724768e-06,
+      "loss": 1.4148035,
+      "memory(GiB)": 103.58,
+      "step": 13100,
+      "train_speed(iter/s)": 1.315207
+    },
+    {
+      "acc": 0.6861763,
+      "epoch": 0.33244545915778795,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.635227591470272e-06,
+      "loss": 1.44053459,
+      "memory(GiB)": 103.58,
+      "step": 13105,
+      "train_speed(iter/s)": 1.315243
+    },
+    {
+      "acc": 0.69065304,
+      "epoch": 0.332572298325723,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.63483430933777e-06,
+      "loss": 1.38115826,
+      "memory(GiB)": 103.58,
+      "step": 13110,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.68365946,
+      "epoch": 0.33269913749365804,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.634440823344565e-06,
+      "loss": 1.44275064,
+      "memory(GiB)": 103.58,
+      "step": 13115,
+      "train_speed(iter/s)": 1.315362
+    },
+    {
+      "acc": 0.6991631,
+      "epoch": 0.3328259766615931,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.634047133507959e-06,
+      "loss": 1.35068741,
+      "memory(GiB)": 103.58,
+      "step": 13120,
+      "train_speed(iter/s)": 1.315423
+    },
+    {
+      "acc": 0.69678741,
+      "epoch": 0.3329528158295282,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.63365323984527e-06,
+      "loss": 1.34963827,
+      "memory(GiB)": 103.58,
+      "step": 13125,
+      "train_speed(iter/s)": 1.315482
+    },
+    {
+      "acc": 0.69904661,
+      "epoch": 0.3330796549974632,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.633259142373825e-06,
+      "loss": 1.35703897,
+      "memory(GiB)": 103.58,
+      "step": 13130,
+      "train_speed(iter/s)": 1.315535
+    },
+    {
+      "acc": 0.66996965,
+      "epoch": 0.33320649416539827,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.632864841110957e-06,
+      "loss": 1.42675533,
+      "memory(GiB)": 103.58,
+      "step": 13135,
+      "train_speed(iter/s)": 1.315589
+    },
+    {
+      "acc": 0.67911353,
+      "epoch": 0.3333333333333333,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.632470336074009e-06,
+      "loss": 1.47730465,
+      "memory(GiB)": 103.58,
+      "step": 13140,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.69801416,
+      "epoch": 0.3334601725012684,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.632075627280333e-06,
+      "loss": 1.34708786,
+      "memory(GiB)": 103.58,
+      "step": 13145,
+      "train_speed(iter/s)": 1.315712
+    },
+    {
+      "acc": 0.6778882,
+      "epoch": 0.33358701166920346,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.631680714747292e-06,
+      "loss": 1.42490454,
+      "memory(GiB)": 103.58,
+      "step": 13150,
+      "train_speed(iter/s)": 1.315769
+    },
+    {
+      "acc": 0.69052877,
+      "epoch": 0.3337138508371385,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.63128559849225e-06,
+      "loss": 1.36631603,
+      "memory(GiB)": 103.58,
+      "step": 13155,
+      "train_speed(iter/s)": 1.315828
+    },
+    {
+      "acc": 0.68220472,
+      "epoch": 0.33384069000507355,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.630890278532594e-06,
+      "loss": 1.37567749,
+      "memory(GiB)": 103.58,
+      "step": 13160,
+      "train_speed(iter/s)": 1.315888
+    },
+    {
+      "acc": 0.65825305,
+      "epoch": 0.33396752917300865,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.630494754885706e-06,
+      "loss": 1.49503803,
+      "memory(GiB)": 103.58,
+      "step": 13165,
+      "train_speed(iter/s)": 1.315943
+    },
+    {
+      "acc": 0.68789735,
+      "epoch": 0.3340943683409437,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.630099027568986e-06,
+      "loss": 1.39400034,
+      "memory(GiB)": 103.58,
+      "step": 13170,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.69492521,
+      "epoch": 0.33422120750887874,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.629703096599839e-06,
+      "loss": 1.42697592,
+      "memory(GiB)": 103.58,
+      "step": 13175,
+      "train_speed(iter/s)": 1.316059
+    },
+    {
+      "acc": 0.68379154,
+      "epoch": 0.3343480466768138,
+      "grad_norm": 2.875,
+      "learning_rate": 9.629306961995678e-06,
+      "loss": 1.41401405,
+      "memory(GiB)": 103.58,
+      "step": 13180,
+      "train_speed(iter/s)": 1.316114
+    },
+    {
+      "acc": 0.66198592,
+      "epoch": 0.3344748858447489,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.62891062377393e-06,
+      "loss": 1.4624609,
+      "memory(GiB)": 103.58,
+      "step": 13185,
+      "train_speed(iter/s)": 1.316174
+    },
+    {
+      "acc": 0.68970957,
+      "epoch": 0.3346017250126839,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.628514081952026e-06,
+      "loss": 1.40876198,
+      "memory(GiB)": 103.58,
+      "step": 13190,
+      "train_speed(iter/s)": 1.316228
+    },
+    {
+      "acc": 0.65745583,
+      "epoch": 0.33472856418061897,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.628117336547408e-06,
+      "loss": 1.51519108,
+      "memory(GiB)": 103.58,
+      "step": 13195,
+      "train_speed(iter/s)": 1.316286
+    },
+    {
+      "acc": 0.66945295,
+      "epoch": 0.334855403348554,
+      "grad_norm": 2.5,
+      "learning_rate": 9.627720387577525e-06,
+      "loss": 1.42803402,
+      "memory(GiB)": 103.58,
+      "step": 13200,
+      "train_speed(iter/s)": 1.316344
+    },
+    {
+      "acc": 0.67875032,
+      "epoch": 0.3349822425164891,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.62732323505984e-06,
+      "loss": 1.46455946,
+      "memory(GiB)": 103.58,
+      "step": 13205,
+      "train_speed(iter/s)": 1.316405
+    },
+    {
+      "acc": 0.68704605,
+      "epoch": 0.33510908168442416,
+      "grad_norm": 2.75,
+      "learning_rate": 9.62692587901182e-06,
+      "loss": 1.35209475,
+      "memory(GiB)": 103.58,
+      "step": 13210,
+      "train_speed(iter/s)": 1.316463
+    },
+    {
+      "acc": 0.68050489,
+      "epoch": 0.3352359208523592,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.62652831945094e-06,
+      "loss": 1.43059063,
+      "memory(GiB)": 103.58,
+      "step": 13215,
+      "train_speed(iter/s)": 1.31652
+    },
+    {
+      "acc": 0.69515796,
+      "epoch": 0.33536276002029425,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.626130556394689e-06,
+      "loss": 1.44006681,
+      "memory(GiB)": 103.58,
+      "step": 13220,
+      "train_speed(iter/s)": 1.316579
+    },
+    {
+      "acc": 0.69013224,
+      "epoch": 0.33548959918822935,
+      "grad_norm": 2.875,
+      "learning_rate": 9.625732589860562e-06,
+      "loss": 1.46015949,
+      "memory(GiB)": 103.58,
+      "step": 13225,
+      "train_speed(iter/s)": 1.316636
+    },
+    {
+      "acc": 0.66726341,
+      "epoch": 0.3356164383561644,
+      "grad_norm": 3.25,
+      "learning_rate": 9.625334419866064e-06,
+      "loss": 1.4765069,
+      "memory(GiB)": 103.58,
+      "step": 13230,
+      "train_speed(iter/s)": 1.316696
+    },
+    {
+      "acc": 0.68442669,
+      "epoch": 0.33574327752409944,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.624936046428708e-06,
+      "loss": 1.39833374,
+      "memory(GiB)": 103.58,
+      "step": 13235,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.68845768,
+      "epoch": 0.3358701166920345,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.624537469566015e-06,
+      "loss": 1.3946352,
+      "memory(GiB)": 103.58,
+      "step": 13240,
+      "train_speed(iter/s)": 1.31681
+    },
+    {
+      "acc": 0.6911212,
+      "epoch": 0.3359969558599696,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.624138689295516e-06,
+      "loss": 1.34838333,
+      "memory(GiB)": 103.58,
+      "step": 13245,
+      "train_speed(iter/s)": 1.316868
+    },
+    {
+      "acc": 0.67953267,
+      "epoch": 0.3361237950279046,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.623739705634753e-06,
+      "loss": 1.37580442,
+      "memory(GiB)": 103.58,
+      "step": 13250,
+      "train_speed(iter/s)": 1.316926
+    },
+    {
+      "acc": 0.689713,
+      "epoch": 0.33625063419583967,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.623340518601274e-06,
+      "loss": 1.41959238,
+      "memory(GiB)": 103.58,
+      "step": 13255,
+      "train_speed(iter/s)": 1.316983
+    },
+    {
+      "acc": 0.67598653,
+      "epoch": 0.3363774733637747,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.622941128212639e-06,
+      "loss": 1.51135683,
+      "memory(GiB)": 103.58,
+      "step": 13260,
+      "train_speed(iter/s)": 1.317042
+    },
+    {
+      "acc": 0.68476248,
+      "epoch": 0.3365043125317098,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.622541534486411e-06,
+      "loss": 1.42286329,
+      "memory(GiB)": 103.58,
+      "step": 13265,
+      "train_speed(iter/s)": 1.317102
+    },
+    {
+      "acc": 0.68209257,
+      "epoch": 0.33663115169964486,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.62214173744017e-06,
+      "loss": 1.41897774,
+      "memory(GiB)": 103.58,
+      "step": 13270,
+      "train_speed(iter/s)": 1.317159
+    },
+    {
+      "acc": 0.67317281,
+      "epoch": 0.3367579908675799,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.6217417370915e-06,
+      "loss": 1.41852942,
+      "memory(GiB)": 103.58,
+      "step": 13275,
+      "train_speed(iter/s)": 1.317217
+    },
+    {
+      "acc": 0.6896903,
+      "epoch": 0.33688483003551495,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.62134153345799e-06,
+      "loss": 1.37739372,
+      "memory(GiB)": 103.58,
+      "step": 13280,
+      "train_speed(iter/s)": 1.317276
+    },
+    {
+      "acc": 0.68407383,
+      "epoch": 0.33701166920345005,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.620941126557248e-06,
+      "loss": 1.35517302,
+      "memory(GiB)": 103.58,
+      "step": 13285,
+      "train_speed(iter/s)": 1.317334
+    },
+    {
+      "acc": 0.69830847,
+      "epoch": 0.3371385083713851,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.620540516406885e-06,
+      "loss": 1.32179375,
+      "memory(GiB)": 103.58,
+      "step": 13290,
+      "train_speed(iter/s)": 1.317391
+    },
+    {
+      "acc": 0.67472014,
+      "epoch": 0.33726534753932014,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.620139703024522e-06,
+      "loss": 1.40538464,
+      "memory(GiB)": 103.58,
+      "step": 13295,
+      "train_speed(iter/s)": 1.317449
+    },
+    {
+      "acc": 0.66405258,
+      "epoch": 0.3373921867072552,
+      "grad_norm": 3.75,
+      "learning_rate": 9.619738686427785e-06,
+      "loss": 1.46373682,
+      "memory(GiB)": 103.58,
+      "step": 13300,
+      "train_speed(iter/s)": 1.317508
+    },
+    {
+      "acc": 0.68042121,
+      "epoch": 0.3375190258751903,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.619337466634317e-06,
+      "loss": 1.39914217,
+      "memory(GiB)": 103.58,
+      "step": 13305,
+      "train_speed(iter/s)": 1.316941
+    },
+    {
+      "acc": 0.69284315,
+      "epoch": 0.3376458650431253,
+      "grad_norm": 3.0,
+      "learning_rate": 9.618936043661762e-06,
+      "loss": 1.35667076,
+      "memory(GiB)": 103.58,
+      "step": 13310,
+      "train_speed(iter/s)": 1.316999
+    },
+    {
+      "acc": 0.67313585,
+      "epoch": 0.33777270421106037,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.618534417527779e-06,
+      "loss": 1.43184681,
+      "memory(GiB)": 103.58,
+      "step": 13315,
+      "train_speed(iter/s)": 1.317057
+    },
+    {
+      "acc": 0.68800535,
+      "epoch": 0.3378995433789954,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.61813258825003e-06,
+      "loss": 1.36976624,
+      "memory(GiB)": 103.58,
+      "step": 13320,
+      "train_speed(iter/s)": 1.317116
+    },
+    {
+      "acc": 0.68217988,
+      "epoch": 0.3380263825469305,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.617730555846191e-06,
+      "loss": 1.40200634,
+      "memory(GiB)": 103.58,
+      "step": 13325,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.67759676,
+      "epoch": 0.33815322171486556,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.617328320333947e-06,
+      "loss": 1.39845724,
+      "memory(GiB)": 103.58,
+      "step": 13330,
+      "train_speed(iter/s)": 1.317229
+    },
+    {
+      "acc": 0.70635657,
+      "epoch": 0.3382800608828006,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.616925881730989e-06,
+      "loss": 1.30388069,
+      "memory(GiB)": 103.58,
+      "step": 13335,
+      "train_speed(iter/s)": 1.317279
+    },
+    {
+      "acc": 0.68948345,
+      "epoch": 0.33840690005073565,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.616523240055017e-06,
+      "loss": 1.39845448,
+      "memory(GiB)": 103.58,
+      "step": 13340,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.6686379,
+      "epoch": 0.33853373921867075,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.616120395323743e-06,
+      "loss": 1.46902103,
+      "memory(GiB)": 103.58,
+      "step": 13345,
+      "train_speed(iter/s)": 1.317389
+    },
+    {
+      "acc": 0.68865066,
+      "epoch": 0.3386605783866058,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.615717347554882e-06,
+      "loss": 1.37765942,
+      "memory(GiB)": 103.58,
+      "step": 13350,
+      "train_speed(iter/s)": 1.317446
+    },
+    {
+      "acc": 0.67456083,
+      "epoch": 0.33878741755454084,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.615314096766166e-06,
+      "loss": 1.40090322,
+      "memory(GiB)": 103.58,
+      "step": 13355,
+      "train_speed(iter/s)": 1.317502
+    },
+    {
+      "acc": 0.67672787,
+      "epoch": 0.3389142567224759,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.61491064297533e-06,
+      "loss": 1.45599995,
+      "memory(GiB)": 103.58,
+      "step": 13360,
+      "train_speed(iter/s)": 1.317556
+    },
+    {
+      "acc": 0.67828064,
+      "epoch": 0.339041095890411,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.614506986200119e-06,
+      "loss": 1.45947456,
+      "memory(GiB)": 103.58,
+      "step": 13365,
+      "train_speed(iter/s)": 1.31761
+    },
+    {
+      "acc": 0.67984991,
+      "epoch": 0.339167935058346,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.61410312645829e-06,
+      "loss": 1.41996346,
+      "memory(GiB)": 103.58,
+      "step": 13370,
+      "train_speed(iter/s)": 1.317667
+    },
+    {
+      "acc": 0.67828236,
+      "epoch": 0.33929477422628107,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.613699063767603e-06,
+      "loss": 1.43997908,
+      "memory(GiB)": 103.58,
+      "step": 13375,
+      "train_speed(iter/s)": 1.317716
+    },
+    {
+      "acc": 0.68080425,
+      "epoch": 0.3394216133942161,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.613294798145833e-06,
+      "loss": 1.45058403,
+      "memory(GiB)": 103.58,
+      "step": 13380,
+      "train_speed(iter/s)": 1.317772
+    },
+    {
+      "acc": 0.69026074,
+      "epoch": 0.3395484525621512,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.612890329610762e-06,
+      "loss": 1.36553478,
+      "memory(GiB)": 103.58,
+      "step": 13385,
+      "train_speed(iter/s)": 1.317829
+    },
+    {
+      "acc": 0.69182687,
+      "epoch": 0.33967529173008626,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.612485658180178e-06,
+      "loss": 1.39205704,
+      "memory(GiB)": 103.58,
+      "step": 13390,
+      "train_speed(iter/s)": 1.317885
+    },
+    {
+      "acc": 0.69051208,
+      "epoch": 0.3398021308980213,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.612080783871882e-06,
+      "loss": 1.38217688,
+      "memory(GiB)": 103.58,
+      "step": 13395,
+      "train_speed(iter/s)": 1.317941
+    },
+    {
+      "acc": 0.68424349,
+      "epoch": 0.33992897006595635,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.611675706703682e-06,
+      "loss": 1.37451248,
+      "memory(GiB)": 103.58,
+      "step": 13400,
+      "train_speed(iter/s)": 1.317996
+    },
+    {
+      "acc": 0.67382174,
+      "epoch": 0.34005580923389145,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.611270426693395e-06,
+      "loss": 1.45724735,
+      "memory(GiB)": 103.58,
+      "step": 13405,
+      "train_speed(iter/s)": 1.318054
+    },
+    {
+      "acc": 0.66585941,
+      "epoch": 0.3401826484018265,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.610864943858847e-06,
+      "loss": 1.439011,
+      "memory(GiB)": 103.58,
+      "step": 13410,
+      "train_speed(iter/s)": 1.31811
+    },
+    {
+      "acc": 0.69472647,
+      "epoch": 0.34030948756976154,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.61045925821787e-06,
+      "loss": 1.37923403,
+      "memory(GiB)": 103.58,
+      "step": 13415,
+      "train_speed(iter/s)": 1.318168
+    },
+    {
+      "acc": 0.6783988,
+      "epoch": 0.3404363267376966,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.610053369788314e-06,
+      "loss": 1.41731262,
+      "memory(GiB)": 103.58,
+      "step": 13420,
+      "train_speed(iter/s)": 1.318224
+    },
+    {
+      "acc": 0.68010378,
+      "epoch": 0.3405631659056317,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.609647278588027e-06,
+      "loss": 1.50451965,
+      "memory(GiB)": 103.58,
+      "step": 13425,
+      "train_speed(iter/s)": 1.31828
+    },
+    {
+      "acc": 0.68525648,
+      "epoch": 0.3406900050735667,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.609240984634871e-06,
+      "loss": 1.36103745,
+      "memory(GiB)": 103.58,
+      "step": 13430,
+      "train_speed(iter/s)": 1.318336
+    },
+    {
+      "acc": 0.69297915,
+      "epoch": 0.34081684424150177,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.608834487946719e-06,
+      "loss": 1.41444349,
+      "memory(GiB)": 103.58,
+      "step": 13435,
+      "train_speed(iter/s)": 1.318388
+    },
+    {
+      "acc": 0.66920424,
+      "epoch": 0.3409436834094368,
+      "grad_norm": 3.5,
+      "learning_rate": 9.60842778854145e-06,
+      "loss": 1.41299973,
+      "memory(GiB)": 103.58,
+      "step": 13440,
+      "train_speed(iter/s)": 1.318444
+    },
+    {
+      "acc": 0.691574,
+      "epoch": 0.3410705225773719,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.60802088643695e-06,
+      "loss": 1.40427084,
+      "memory(GiB)": 103.58,
+      "step": 13445,
+      "train_speed(iter/s)": 1.318498
+    },
+    {
+      "acc": 0.67503147,
+      "epoch": 0.34119736174530696,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.60761378165112e-06,
+      "loss": 1.40574465,
+      "memory(GiB)": 103.58,
+      "step": 13450,
+      "train_speed(iter/s)": 1.318548
+    },
+    {
+      "acc": 0.69699574,
+      "epoch": 0.341324200913242,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.607206474201863e-06,
+      "loss": 1.37335491,
+      "memory(GiB)": 103.58,
+      "step": 13455,
+      "train_speed(iter/s)": 1.318601
+    },
+    {
+      "acc": 0.68399992,
+      "epoch": 0.34145104008117705,
+      "grad_norm": 3.375,
+      "learning_rate": 9.606798964107096e-06,
+      "loss": 1.47422371,
+      "memory(GiB)": 103.58,
+      "step": 13460,
+      "train_speed(iter/s)": 1.318653
+    },
+    {
+      "acc": 0.69506664,
+      "epoch": 0.34157787924911215,
+      "grad_norm": 3.375,
+      "learning_rate": 9.60639125138474e-06,
+      "loss": 1.33859196,
+      "memory(GiB)": 103.58,
+      "step": 13465,
+      "train_speed(iter/s)": 1.318709
+    },
+    {
+      "acc": 0.6926055,
+      "epoch": 0.3417047184170472,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.605983336052735e-06,
+      "loss": 1.40810833,
+      "memory(GiB)": 103.58,
+      "step": 13470,
+      "train_speed(iter/s)": 1.318765
+    },
+    {
+      "acc": 0.67082257,
+      "epoch": 0.34183155758498224,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.605575218129017e-06,
+      "loss": 1.43456879,
+      "memory(GiB)": 103.58,
+      "step": 13475,
+      "train_speed(iter/s)": 1.318822
+    },
+    {
+      "acc": 0.66083651,
+      "epoch": 0.3419583967529173,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.605166897631539e-06,
+      "loss": 1.49457912,
+      "memory(GiB)": 103.58,
+      "step": 13480,
+      "train_speed(iter/s)": 1.318876
+    },
+    {
+      "acc": 0.69121895,
+      "epoch": 0.3420852359208524,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.604758374578259e-06,
+      "loss": 1.40658112,
+      "memory(GiB)": 103.58,
+      "step": 13485,
+      "train_speed(iter/s)": 1.318928
+    },
+    {
+      "acc": 0.67676897,
+      "epoch": 0.3422120750887874,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.604349648987148e-06,
+      "loss": 1.40288877,
+      "memory(GiB)": 103.58,
+      "step": 13490,
+      "train_speed(iter/s)": 1.318984
+    },
+    {
+      "acc": 0.68645468,
+      "epoch": 0.34233891425672247,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.603940720876181e-06,
+      "loss": 1.43618832,
+      "memory(GiB)": 103.58,
+      "step": 13495,
+      "train_speed(iter/s)": 1.319038
+    },
+    {
+      "acc": 0.68012362,
+      "epoch": 0.3424657534246575,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.603531590263348e-06,
+      "loss": 1.41157246,
+      "memory(GiB)": 103.58,
+      "step": 13500,
+      "train_speed(iter/s)": 1.319091
+    },
+    {
+      "acc": 0.68009119,
+      "epoch": 0.3425925925925926,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.603122257166641e-06,
+      "loss": 1.41806421,
+      "memory(GiB)": 103.58,
+      "step": 13505,
+      "train_speed(iter/s)": 1.319145
+    },
+    {
+      "acc": 0.69644651,
+      "epoch": 0.34271943176052766,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.602712721604066e-06,
+      "loss": 1.35852518,
+      "memory(GiB)": 103.58,
+      "step": 13510,
+      "train_speed(iter/s)": 1.319199
+    },
+    {
+      "acc": 0.67394266,
+      "epoch": 0.3428462709284627,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.602302983593637e-06,
+      "loss": 1.4491293,
+      "memory(GiB)": 103.58,
+      "step": 13515,
+      "train_speed(iter/s)": 1.319256
+    },
+    {
+      "acc": 0.68187346,
+      "epoch": 0.34297311009639775,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.601893043153372e-06,
+      "loss": 1.41034393,
+      "memory(GiB)": 103.58,
+      "step": 13520,
+      "train_speed(iter/s)": 1.319307
+    },
+    {
+      "acc": 0.67551451,
+      "epoch": 0.34309994926433285,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.601482900301308e-06,
+      "loss": 1.41042767,
+      "memory(GiB)": 103.58,
+      "step": 13525,
+      "train_speed(iter/s)": 1.319361
+    },
+    {
+      "acc": 0.68390923,
+      "epoch": 0.3432267884322679,
+      "grad_norm": 3.125,
+      "learning_rate": 9.60107255505548e-06,
+      "loss": 1.41836681,
+      "memory(GiB)": 103.58,
+      "step": 13530,
+      "train_speed(iter/s)": 1.319415
+    },
+    {
+      "acc": 0.66911874,
+      "epoch": 0.34335362760020294,
+      "grad_norm": 2.75,
+      "learning_rate": 9.60066200743394e-06,
+      "loss": 1.52876787,
+      "memory(GiB)": 103.58,
+      "step": 13535,
+      "train_speed(iter/s)": 1.319471
+    },
+    {
+      "acc": 0.67658987,
+      "epoch": 0.343480466768138,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.600251257454744e-06,
+      "loss": 1.40301857,
+      "memory(GiB)": 103.58,
+      "step": 13540,
+      "train_speed(iter/s)": 1.319527
+    },
+    {
+      "acc": 0.66734118,
+      "epoch": 0.3436073059360731,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.599840305135959e-06,
+      "loss": 1.44921646,
+      "memory(GiB)": 103.58,
+      "step": 13545,
+      "train_speed(iter/s)": 1.319582
+    },
+    {
+      "acc": 0.6881741,
+      "epoch": 0.3437341451040081,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.59942915049566e-06,
+      "loss": 1.34421358,
+      "memory(GiB)": 103.58,
+      "step": 13550,
+      "train_speed(iter/s)": 1.319639
+    },
+    {
+      "acc": 0.66577749,
+      "epoch": 0.34386098427194317,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.599017793551933e-06,
+      "loss": 1.41541615,
+      "memory(GiB)": 103.58,
+      "step": 13555,
+      "train_speed(iter/s)": 1.319693
+    },
+    {
+      "acc": 0.67021723,
+      "epoch": 0.3439878234398782,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.598606234322869e-06,
+      "loss": 1.52072258,
+      "memory(GiB)": 103.58,
+      "step": 13560,
+      "train_speed(iter/s)": 1.319749
+    },
+    {
+      "acc": 0.68978558,
+      "epoch": 0.3441146626078133,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.598194472826574e-06,
+      "loss": 1.38800373,
+      "memory(GiB)": 103.58,
+      "step": 13565,
+      "train_speed(iter/s)": 1.319805
+    },
+    {
+      "acc": 0.67349024,
+      "epoch": 0.34424150177574836,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.597782509081154e-06,
+      "loss": 1.43993521,
+      "memory(GiB)": 103.58,
+      "step": 13570,
+      "train_speed(iter/s)": 1.319862
+    },
+    {
+      "acc": 0.6849824,
+      "epoch": 0.3443683409436834,
+      "grad_norm": 2.75,
+      "learning_rate": 9.597370343104733e-06,
+      "loss": 1.37893963,
+      "memory(GiB)": 103.58,
+      "step": 13575,
+      "train_speed(iter/s)": 1.319918
+    },
+    {
+      "acc": 0.69873772,
+      "epoch": 0.34449518011161845,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.596957974915438e-06,
+      "loss": 1.32922249,
+      "memory(GiB)": 103.58,
+      "step": 13580,
+      "train_speed(iter/s)": 1.31997
+    },
+    {
+      "acc": 0.68102503,
+      "epoch": 0.34462201927955355,
+      "grad_norm": 2.375,
+      "learning_rate": 9.596545404531408e-06,
+      "loss": 1.42819548,
+      "memory(GiB)": 103.58,
+      "step": 13585,
+      "train_speed(iter/s)": 1.319386
+    },
+    {
+      "acc": 0.66988111,
+      "epoch": 0.3447488584474886,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.596132631970788e-06,
+      "loss": 1.39812975,
+      "memory(GiB)": 103.58,
+      "step": 13590,
+      "train_speed(iter/s)": 1.319443
+    },
+    {
+      "acc": 0.70390253,
+      "epoch": 0.34487569761542364,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.595719657251735e-06,
+      "loss": 1.34582825,
+      "memory(GiB)": 103.58,
+      "step": 13595,
+      "train_speed(iter/s)": 1.319499
+    },
+    {
+      "acc": 0.69277067,
+      "epoch": 0.3450025367833587,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.595306480392413e-06,
+      "loss": 1.37304401,
+      "memory(GiB)": 103.58,
+      "step": 13600,
+      "train_speed(iter/s)": 1.319555
+    },
+    {
+      "acc": 0.68630791,
+      "epoch": 0.3451293759512938,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.594893101410995e-06,
+      "loss": 1.44016895,
+      "memory(GiB)": 103.58,
+      "step": 13605,
+      "train_speed(iter/s)": 1.31961
+    },
+    {
+      "acc": 0.67565227,
+      "epoch": 0.3452562151192288,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.594479520325665e-06,
+      "loss": 1.4181798,
+      "memory(GiB)": 103.58,
+      "step": 13610,
+      "train_speed(iter/s)": 1.319663
+    },
+    {
+      "acc": 0.67314825,
+      "epoch": 0.34538305428716387,
+      "grad_norm": 3.25,
+      "learning_rate": 9.594065737154611e-06,
+      "loss": 1.39146156,
+      "memory(GiB)": 103.58,
+      "step": 13615,
+      "train_speed(iter/s)": 1.319719
+    },
+    {
+      "acc": 0.68481727,
+      "epoch": 0.3455098934550989,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.593651751916037e-06,
+      "loss": 1.35130825,
+      "memory(GiB)": 103.58,
+      "step": 13620,
+      "train_speed(iter/s)": 1.319773
+    },
+    {
+      "acc": 0.69647202,
+      "epoch": 0.345636732623034,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.593237564628149e-06,
+      "loss": 1.29013882,
+      "memory(GiB)": 103.58,
+      "step": 13625,
+      "train_speed(iter/s)": 1.319829
+    },
+    {
+      "acc": 0.69431567,
+      "epoch": 0.34576357179096906,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.592823175309164e-06,
+      "loss": 1.30613222,
+      "memory(GiB)": 103.58,
+      "step": 13630,
+      "train_speed(iter/s)": 1.319885
+    },
+    {
+      "acc": 0.68694282,
+      "epoch": 0.3458904109589041,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.592408583977311e-06,
+      "loss": 1.36852598,
+      "memory(GiB)": 103.58,
+      "step": 13635,
+      "train_speed(iter/s)": 1.319937
+    },
+    {
+      "acc": 0.66812806,
+      "epoch": 0.34601725012683915,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.591993790650826e-06,
+      "loss": 1.3933979,
+      "memory(GiB)": 103.58,
+      "step": 13640,
+      "train_speed(iter/s)": 1.319993
+    },
+    {
+      "acc": 0.68821702,
+      "epoch": 0.34614408929477425,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.591578795347952e-06,
+      "loss": 1.44424286,
+      "memory(GiB)": 103.58,
+      "step": 13645,
+      "train_speed(iter/s)": 1.320049
+    },
+    {
+      "acc": 0.68353209,
+      "epoch": 0.3462709284627093,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.591163598086943e-06,
+      "loss": 1.40578279,
+      "memory(GiB)": 103.58,
+      "step": 13650,
+      "train_speed(iter/s)": 1.320103
+    },
+    {
+      "acc": 0.68113394,
+      "epoch": 0.34639776763064434,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.59074819888606e-06,
+      "loss": 1.42256165,
+      "memory(GiB)": 103.58,
+      "step": 13655,
+      "train_speed(iter/s)": 1.320161
+    },
+    {
+      "acc": 0.69126463,
+      "epoch": 0.3465246067985794,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.590332597763575e-06,
+      "loss": 1.36765823,
+      "memory(GiB)": 103.58,
+      "step": 13660,
+      "train_speed(iter/s)": 1.320216
+    },
+    {
+      "acc": 0.68721385,
+      "epoch": 0.3466514459665145,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.589916794737768e-06,
+      "loss": 1.44843502,
+      "memory(GiB)": 103.58,
+      "step": 13665,
+      "train_speed(iter/s)": 1.32027
+    },
+    {
+      "acc": 0.68237214,
+      "epoch": 0.3467782851344495,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.589500789826927e-06,
+      "loss": 1.48482504,
+      "memory(GiB)": 103.58,
+      "step": 13670,
+      "train_speed(iter/s)": 1.320323
+    },
+    {
+      "acc": 0.6690196,
+      "epoch": 0.34690512430238457,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.589084583049353e-06,
+      "loss": 1.51101761,
+      "memory(GiB)": 103.58,
+      "step": 13675,
+      "train_speed(iter/s)": 1.319745
+    },
+    {
+      "acc": 0.68867478,
+      "epoch": 0.3470319634703196,
+      "grad_norm": 3.5,
+      "learning_rate": 9.588668174423348e-06,
+      "loss": 1.4016263,
+      "memory(GiB)": 103.58,
+      "step": 13680,
+      "train_speed(iter/s)": 1.3198
+    },
+    {
+      "acc": 0.67519684,
+      "epoch": 0.3471588026382547,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.588251563967232e-06,
+      "loss": 1.46799507,
+      "memory(GiB)": 103.58,
+      "step": 13685,
+      "train_speed(iter/s)": 1.319856
+    },
+    {
+      "acc": 0.67556763,
+      "epoch": 0.34728564180618976,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.587834751699326e-06,
+      "loss": 1.47412672,
+      "memory(GiB)": 103.58,
+      "step": 13690,
+      "train_speed(iter/s)": 1.319912
+    },
+    {
+      "acc": 0.67940679,
+      "epoch": 0.3474124809741248,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.587417737637963e-06,
+      "loss": 1.46347179,
+      "memory(GiB)": 103.58,
+      "step": 13695,
+      "train_speed(iter/s)": 1.319967
+    },
+    {
+      "acc": 0.68636999,
+      "epoch": 0.34753932014205985,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.587000521801488e-06,
+      "loss": 1.38201141,
+      "memory(GiB)": 103.58,
+      "step": 13700,
+      "train_speed(iter/s)": 1.320018
+    },
+    {
+      "acc": 0.68947644,
+      "epoch": 0.34766615930999495,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.58658310420825e-06,
+      "loss": 1.40582075,
+      "memory(GiB)": 103.58,
+      "step": 13705,
+      "train_speed(iter/s)": 1.320074
+    },
+    {
+      "acc": 0.68384867,
+      "epoch": 0.34779299847793,
+      "grad_norm": 3.125,
+      "learning_rate": 9.58616548487661e-06,
+      "loss": 1.39474173,
+      "memory(GiB)": 103.58,
+      "step": 13710,
+      "train_speed(iter/s)": 1.320129
+    },
+    {
+      "acc": 0.68299093,
+      "epoch": 0.34791983764586504,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.585747663824936e-06,
+      "loss": 1.46125832,
+      "memory(GiB)": 103.58,
+      "step": 13715,
+      "train_speed(iter/s)": 1.319516
+    },
+    {
+      "acc": 0.69340563,
+      "epoch": 0.3480466768138001,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.585329641071606e-06,
+      "loss": 1.38938198,
+      "memory(GiB)": 103.58,
+      "step": 13720,
+      "train_speed(iter/s)": 1.319571
+    },
+    {
+      "acc": 0.68086896,
+      "epoch": 0.3481735159817352,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.584911416635007e-06,
+      "loss": 1.42356033,
+      "memory(GiB)": 103.58,
+      "step": 13725,
+      "train_speed(iter/s)": 1.319626
+    },
+    {
+      "acc": 0.6819706,
+      "epoch": 0.3483003551496702,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.584492990533533e-06,
+      "loss": 1.42305565,
+      "memory(GiB)": 103.58,
+      "step": 13730,
+      "train_speed(iter/s)": 1.31968
+    },
+    {
+      "acc": 0.69345393,
+      "epoch": 0.34842719431760527,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.58407436278559e-06,
+      "loss": 1.32317142,
+      "memory(GiB)": 103.58,
+      "step": 13735,
+      "train_speed(iter/s)": 1.319734
+    },
+    {
+      "acc": 0.67556047,
+      "epoch": 0.3485540334855403,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.583655533409588e-06,
+      "loss": 1.41217165,
+      "memory(GiB)": 103.58,
+      "step": 13740,
+      "train_speed(iter/s)": 1.31979
+    },
+    {
+      "acc": 0.68535585,
+      "epoch": 0.3486808726534754,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.583236502423952e-06,
+      "loss": 1.40956354,
+      "memory(GiB)": 103.58,
+      "step": 13745,
+      "train_speed(iter/s)": 1.319843
+    },
+    {
+      "acc": 0.68324986,
+      "epoch": 0.34880771182141046,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.582817269847112e-06,
+      "loss": 1.46595497,
+      "memory(GiB)": 103.58,
+      "step": 13750,
+      "train_speed(iter/s)": 1.319891
+    },
+    {
+      "acc": 0.68636975,
+      "epoch": 0.3489345509893455,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.582397835697509e-06,
+      "loss": 1.37293625,
+      "memory(GiB)": 103.58,
+      "step": 13755,
+      "train_speed(iter/s)": 1.319944
+    },
+    {
+      "acc": 0.68872313,
+      "epoch": 0.34906139015728055,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.581978199993587e-06,
+      "loss": 1.39980545,
+      "memory(GiB)": 103.58,
+      "step": 13760,
+      "train_speed(iter/s)": 1.319997
+    },
+    {
+      "acc": 0.67855158,
+      "epoch": 0.34918822932521565,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.58155836275381e-06,
+      "loss": 1.4051796,
+      "memory(GiB)": 103.58,
+      "step": 13765,
+      "train_speed(iter/s)": 1.32005
+    },
+    {
+      "acc": 0.68162861,
+      "epoch": 0.3493150684931507,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.581138323996639e-06,
+      "loss": 1.47915726,
+      "memory(GiB)": 103.58,
+      "step": 13770,
+      "train_speed(iter/s)": 1.320105
+    },
+    {
+      "acc": 0.69672828,
+      "epoch": 0.34944190766108574,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.580718083740553e-06,
+      "loss": 1.45576611,
+      "memory(GiB)": 103.58,
+      "step": 13775,
+      "train_speed(iter/s)": 1.320158
+    },
+    {
+      "acc": 0.71125374,
+      "epoch": 0.3495687468290208,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.580297642004032e-06,
+      "loss": 1.3214695,
+      "memory(GiB)": 103.58,
+      "step": 13780,
+      "train_speed(iter/s)": 1.320208
+    },
+    {
+      "acc": 0.68341675,
+      "epoch": 0.3496955859969559,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.579876998805573e-06,
+      "loss": 1.46356297,
+      "memory(GiB)": 103.58,
+      "step": 13785,
+      "train_speed(iter/s)": 1.320261
+    },
+    {
+      "acc": 0.69001036,
+      "epoch": 0.3498224251648909,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.579456154163676e-06,
+      "loss": 1.42348833,
+      "memory(GiB)": 103.58,
+      "step": 13790,
+      "train_speed(iter/s)": 1.320316
+    },
+    {
+      "acc": 0.67900634,
+      "epoch": 0.34994926433282597,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.57903510809685e-06,
+      "loss": 1.4040081,
+      "memory(GiB)": 103.58,
+      "step": 13795,
+      "train_speed(iter/s)": 1.320369
+    },
+    {
+      "acc": 0.6802763,
+      "epoch": 0.350076103500761,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.578613860623617e-06,
+      "loss": 1.39285889,
+      "memory(GiB)": 103.58,
+      "step": 13800,
+      "train_speed(iter/s)": 1.320422
+    },
+    {
+      "acc": 0.69428253,
+      "epoch": 0.3502029426686961,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.578192411762503e-06,
+      "loss": 1.31060543,
+      "memory(GiB)": 103.58,
+      "step": 13805,
+      "train_speed(iter/s)": 1.320475
+    },
+    {
+      "acc": 0.67509999,
+      "epoch": 0.35032978183663116,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.577770761532049e-06,
+      "loss": 1.46534557,
+      "memory(GiB)": 103.58,
+      "step": 13810,
+      "train_speed(iter/s)": 1.320528
+    },
+    {
+      "acc": 0.68640432,
+      "epoch": 0.3504566210045662,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.577348909950797e-06,
+      "loss": 1.43123322,
+      "memory(GiB)": 103.58,
+      "step": 13815,
+      "train_speed(iter/s)": 1.320575
+    },
+    {
+      "acc": 0.69384742,
+      "epoch": 0.35058346017250125,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.576926857037303e-06,
+      "loss": 1.26134939,
+      "memory(GiB)": 103.58,
+      "step": 13820,
+      "train_speed(iter/s)": 1.320627
+    },
+    {
+      "acc": 0.69414129,
+      "epoch": 0.35071029934043635,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.576504602810133e-06,
+      "loss": 1.40944386,
+      "memory(GiB)": 103.58,
+      "step": 13825,
+      "train_speed(iter/s)": 1.320683
+    },
+    {
+      "acc": 0.68363857,
+      "epoch": 0.3508371385083714,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.576082147287858e-06,
+      "loss": 1.37879181,
+      "memory(GiB)": 103.58,
+      "step": 13830,
+      "train_speed(iter/s)": 1.320735
+    },
+    {
+      "acc": 0.68442197,
+      "epoch": 0.35096397767630644,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.575659490489058e-06,
+      "loss": 1.36711226,
+      "memory(GiB)": 103.58,
+      "step": 13835,
+      "train_speed(iter/s)": 1.320786
+    },
+    {
+      "acc": 0.68934884,
+      "epoch": 0.3510908168442415,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.575236632432325e-06,
+      "loss": 1.42008591,
+      "memory(GiB)": 103.58,
+      "step": 13840,
+      "train_speed(iter/s)": 1.320841
+    },
+    {
+      "acc": 0.69181027,
+      "epoch": 0.3512176560121766,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.574813573136259e-06,
+      "loss": 1.41748123,
+      "memory(GiB)": 103.58,
+      "step": 13845,
+      "train_speed(iter/s)": 1.320893
+    },
+    {
+      "acc": 0.68924398,
+      "epoch": 0.3513444951801116,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.574390312619466e-06,
+      "loss": 1.39449148,
+      "memory(GiB)": 103.58,
+      "step": 13850,
+      "train_speed(iter/s)": 1.320947
+    },
+    {
+      "acc": 0.68714743,
+      "epoch": 0.35147133434804667,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.573966850900565e-06,
+      "loss": 1.41097527,
+      "memory(GiB)": 103.58,
+      "step": 13855,
+      "train_speed(iter/s)": 1.321
+    },
+    {
+      "acc": 0.67669406,
+      "epoch": 0.3515981735159817,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.57354318799818e-06,
+      "loss": 1.38393135,
+      "memory(GiB)": 103.58,
+      "step": 13860,
+      "train_speed(iter/s)": 1.321052
+    },
+    {
+      "acc": 0.68707061,
+      "epoch": 0.3517250126839168,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.573119323930946e-06,
+      "loss": 1.3937418,
+      "memory(GiB)": 103.58,
+      "step": 13865,
+      "train_speed(iter/s)": 1.321105
+    },
+    {
+      "acc": 0.68726187,
+      "epoch": 0.35185185185185186,
+      "grad_norm": 2.875,
+      "learning_rate": 9.572695258717507e-06,
+      "loss": 1.42207813,
+      "memory(GiB)": 103.58,
+      "step": 13870,
+      "train_speed(iter/s)": 1.321157
+    },
+    {
+      "acc": 0.68021641,
+      "epoch": 0.3519786910197869,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.572270992376513e-06,
+      "loss": 1.42305508,
+      "memory(GiB)": 103.58,
+      "step": 13875,
+      "train_speed(iter/s)": 1.321209
+    },
+    {
+      "acc": 0.68625412,
+      "epoch": 0.35210553018772195,
+      "grad_norm": 2.625,
+      "learning_rate": 9.571846524926629e-06,
+      "loss": 1.40971375,
+      "memory(GiB)": 103.58,
+      "step": 13880,
+      "train_speed(iter/s)": 1.321261
+    },
+    {
+      "acc": 0.6952889,
+      "epoch": 0.35223236935565705,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.571421856386522e-06,
+      "loss": 1.4410759,
+      "memory(GiB)": 103.58,
+      "step": 13885,
+      "train_speed(iter/s)": 1.321315
+    },
+    {
+      "acc": 0.68302555,
+      "epoch": 0.3523592085235921,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.570996986774872e-06,
+      "loss": 1.40820198,
+      "memory(GiB)": 103.58,
+      "step": 13890,
+      "train_speed(iter/s)": 1.321365
+    },
+    {
+      "acc": 0.68568053,
+      "epoch": 0.35248604769152714,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.570571916110366e-06,
+      "loss": 1.44875822,
+      "memory(GiB)": 103.58,
+      "step": 13895,
+      "train_speed(iter/s)": 1.32141
+    },
+    {
+      "acc": 0.68833375,
+      "epoch": 0.3526128868594622,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.570146644411705e-06,
+      "loss": 1.39808359,
+      "memory(GiB)": 103.58,
+      "step": 13900,
+      "train_speed(iter/s)": 1.321463
+    },
+    {
+      "acc": 0.68126683,
+      "epoch": 0.3527397260273973,
+      "grad_norm": 3.125,
+      "learning_rate": 9.569721171697587e-06,
+      "loss": 1.42431946,
+      "memory(GiB)": 103.58,
+      "step": 13905,
+      "train_speed(iter/s)": 1.321517
+    },
+    {
+      "acc": 0.68029327,
+      "epoch": 0.3528665651953323,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.569295497986727e-06,
+      "loss": 1.47627926,
+      "memory(GiB)": 103.58,
+      "step": 13910,
+      "train_speed(iter/s)": 1.321573
+    },
+    {
+      "acc": 0.70311427,
+      "epoch": 0.35299340436326737,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.568869623297855e-06,
+      "loss": 1.37537088,
+      "memory(GiB)": 103.58,
+      "step": 13915,
+      "train_speed(iter/s)": 1.321627
+    },
+    {
+      "acc": 0.67339764,
+      "epoch": 0.3531202435312024,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.568443547649697e-06,
+      "loss": 1.52320805,
+      "memory(GiB)": 103.58,
+      "step": 13920,
+      "train_speed(iter/s)": 1.32168
+    },
+    {
+      "acc": 0.687327,
+      "epoch": 0.3532470826991375,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.568017271060994e-06,
+      "loss": 1.43092575,
+      "memory(GiB)": 103.58,
+      "step": 13925,
+      "train_speed(iter/s)": 1.321732
+    },
+    {
+      "acc": 0.69329386,
+      "epoch": 0.35337392186707256,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.567590793550498e-06,
+      "loss": 1.37756796,
+      "memory(GiB)": 103.58,
+      "step": 13930,
+      "train_speed(iter/s)": 1.321785
+    },
+    {
+      "acc": 0.68744631,
+      "epoch": 0.3535007610350076,
+      "grad_norm": 3.375,
+      "learning_rate": 9.567164115136965e-06,
+      "loss": 1.41956081,
+      "memory(GiB)": 103.58,
+      "step": 13935,
+      "train_speed(iter/s)": 1.321837
+    },
+    {
+      "acc": 0.69063158,
+      "epoch": 0.35362760020294265,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.566737235839166e-06,
+      "loss": 1.40798569,
+      "memory(GiB)": 103.58,
+      "step": 13940,
+      "train_speed(iter/s)": 1.32189
+    },
+    {
+      "acc": 0.67877951,
+      "epoch": 0.35375443937087775,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.566310155675871e-06,
+      "loss": 1.42974081,
+      "memory(GiB)": 103.58,
+      "step": 13945,
+      "train_speed(iter/s)": 1.321941
+    },
+    {
+      "acc": 0.67493143,
+      "epoch": 0.3538812785388128,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.56588287466587e-06,
+      "loss": 1.38417225,
+      "memory(GiB)": 103.58,
+      "step": 13950,
+      "train_speed(iter/s)": 1.321993
+    },
+    {
+      "acc": 0.69627261,
+      "epoch": 0.35400811770674784,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.565455392827954e-06,
+      "loss": 1.37189331,
+      "memory(GiB)": 103.58,
+      "step": 13955,
+      "train_speed(iter/s)": 1.322045
+    },
+    {
+      "acc": 0.67932558,
+      "epoch": 0.3541349568746829,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.565027710180927e-06,
+      "loss": 1.42271938,
+      "memory(GiB)": 103.58,
+      "step": 13960,
+      "train_speed(iter/s)": 1.322097
+    },
+    {
+      "acc": 0.69756861,
+      "epoch": 0.354261796042618,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.5645998267436e-06,
+      "loss": 1.30514832,
+      "memory(GiB)": 103.58,
+      "step": 13965,
+      "train_speed(iter/s)": 1.32215
+    },
+    {
+      "acc": 0.67665815,
+      "epoch": 0.354388635210553,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.564171742534794e-06,
+      "loss": 1.41825962,
+      "memory(GiB)": 103.58,
+      "step": 13970,
+      "train_speed(iter/s)": 1.322202
+    },
+    {
+      "acc": 0.68148413,
+      "epoch": 0.35451547437848807,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.563743457573336e-06,
+      "loss": 1.45504265,
+      "memory(GiB)": 103.58,
+      "step": 13975,
+      "train_speed(iter/s)": 1.322253
+    },
+    {
+      "acc": 0.67521553,
+      "epoch": 0.3546423135464231,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.563314971878065e-06,
+      "loss": 1.44160233,
+      "memory(GiB)": 103.58,
+      "step": 13980,
+      "train_speed(iter/s)": 1.322305
+    },
+    {
+      "acc": 0.68193455,
+      "epoch": 0.3547691527143582,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.562886285467828e-06,
+      "loss": 1.39183836,
+      "memory(GiB)": 103.58,
+      "step": 13985,
+      "train_speed(iter/s)": 1.322358
+    },
+    {
+      "acc": 0.67769361,
+      "epoch": 0.35489599188229326,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.56245739836148e-06,
+      "loss": 1.43717518,
+      "memory(GiB)": 103.58,
+      "step": 13990,
+      "train_speed(iter/s)": 1.322413
+    },
+    {
+      "acc": 0.67904215,
+      "epoch": 0.3550228310502283,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.562028310577887e-06,
+      "loss": 1.40191498,
+      "memory(GiB)": 103.58,
+      "step": 13995,
+      "train_speed(iter/s)": 1.322466
+    },
+    {
+      "acc": 0.67679815,
+      "epoch": 0.35514967021816335,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.56159902213592e-06,
+      "loss": 1.43341417,
+      "memory(GiB)": 103.58,
+      "step": 14000,
+      "train_speed(iter/s)": 1.32252
+    },
+    {
+      "epoch": 0.35514967021816335,
+      "eval_acc": 0.6716982314135536,
+      "eval_loss": 1.3755980730056763,
+      "eval_runtime": 69.3092,
+      "eval_samples_per_second": 91.907,
+      "eval_steps_per_second": 22.984,
+      "step": 14000
+    },
+    {
+      "acc": 0.67910738,
+      "epoch": 0.35527650938609845,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.561169533054462e-06,
+      "loss": 1.41957989,
+      "memory(GiB)": 103.58,
+      "step": 14005,
+      "train_speed(iter/s)": 1.312417
+    },
+    {
+      "acc": 0.699822,
+      "epoch": 0.3554033485540335,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.560739843352404e-06,
+      "loss": 1.42510681,
+      "memory(GiB)": 103.58,
+      "step": 14010,
+      "train_speed(iter/s)": 1.31245
+    },
+    {
+      "acc": 0.67851777,
+      "epoch": 0.35553018772196854,
+      "grad_norm": 3.375,
+      "learning_rate": 9.560309953048645e-06,
+      "loss": 1.46958361,
+      "memory(GiB)": 103.58,
+      "step": 14015,
+      "train_speed(iter/s)": 1.312501
+    },
+    {
+      "acc": 0.67383585,
+      "epoch": 0.3556570268899036,
+      "grad_norm": 2.75,
+      "learning_rate": 9.559879862162095e-06,
+      "loss": 1.45790062,
+      "memory(GiB)": 103.58,
+      "step": 14020,
+      "train_speed(iter/s)": 1.312556
+    },
+    {
+      "acc": 0.67893825,
+      "epoch": 0.3557838660578387,
+      "grad_norm": 3.375,
+      "learning_rate": 9.55944957071167e-06,
+      "loss": 1.46276827,
+      "memory(GiB)": 103.58,
+      "step": 14025,
+      "train_speed(iter/s)": 1.31261
+    },
+    {
+      "acc": 0.67912035,
+      "epoch": 0.3559107052257737,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.559019078716295e-06,
+      "loss": 1.44471083,
+      "memory(GiB)": 103.58,
+      "step": 14030,
+      "train_speed(iter/s)": 1.312664
+    },
+    {
+      "acc": 0.68332849,
+      "epoch": 0.35603754439370877,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.558588386194907e-06,
+      "loss": 1.474545,
+      "memory(GiB)": 103.58,
+      "step": 14035,
+      "train_speed(iter/s)": 1.312717
+    },
+    {
+      "acc": 0.6834548,
+      "epoch": 0.3561643835616438,
+      "grad_norm": 5.3125,
+      "learning_rate": 9.55815749316645e-06,
+      "loss": 1.36885586,
+      "memory(GiB)": 103.58,
+      "step": 14040,
+      "train_speed(iter/s)": 1.312771
+    },
+    {
+      "acc": 0.6699266,
+      "epoch": 0.3562912227295789,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.557726399649875e-06,
+      "loss": 1.43712025,
+      "memory(GiB)": 103.58,
+      "step": 14045,
+      "train_speed(iter/s)": 1.312822
+    },
+    {
+      "acc": 0.68360538,
+      "epoch": 0.35641806189751396,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.557295105664144e-06,
+      "loss": 1.41117706,
+      "memory(GiB)": 103.58,
+      "step": 14050,
+      "train_speed(iter/s)": 1.312876
+    },
+    {
+      "acc": 0.67352042,
+      "epoch": 0.356544901065449,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.556863611228228e-06,
+      "loss": 1.45141449,
+      "memory(GiB)": 103.58,
+      "step": 14055,
+      "train_speed(iter/s)": 1.312926
+    },
+    {
+      "acc": 0.69126415,
+      "epoch": 0.35667174023338405,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.556431916361105e-06,
+      "loss": 1.39947014,
+      "memory(GiB)": 103.58,
+      "step": 14060,
+      "train_speed(iter/s)": 1.312979
+    },
+    {
+      "acc": 0.67772241,
+      "epoch": 0.35679857940131915,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.556000021081764e-06,
+      "loss": 1.40714207,
+      "memory(GiB)": 103.58,
+      "step": 14065,
+      "train_speed(iter/s)": 1.313034
+    },
+    {
+      "acc": 0.69741726,
+      "epoch": 0.3569254185692542,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.5555679254092e-06,
+      "loss": 1.32001257,
+      "memory(GiB)": 103.58,
+      "step": 14070,
+      "train_speed(iter/s)": 1.313089
+    },
+    {
+      "acc": 0.67709112,
+      "epoch": 0.35705225773718924,
+      "grad_norm": 3.375,
+      "learning_rate": 9.55513562936242e-06,
+      "loss": 1.50639572,
+      "memory(GiB)": 103.58,
+      "step": 14075,
+      "train_speed(iter/s)": 1.313141
+    },
+    {
+      "acc": 0.66623683,
+      "epoch": 0.3571790969051243,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.554703132960437e-06,
+      "loss": 1.52240257,
+      "memory(GiB)": 103.58,
+      "step": 14080,
+      "train_speed(iter/s)": 1.313193
+    },
+    {
+      "acc": 0.6981143,
+      "epoch": 0.3573059360730594,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.554270436222277e-06,
+      "loss": 1.32877922,
+      "memory(GiB)": 103.58,
+      "step": 14085,
+      "train_speed(iter/s)": 1.313242
+    },
+    {
+      "acc": 0.67796359,
+      "epoch": 0.3574327752409944,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.553837539166969e-06,
+      "loss": 1.39791288,
+      "memory(GiB)": 103.58,
+      "step": 14090,
+      "train_speed(iter/s)": 1.313297
+    },
+    {
+      "acc": 0.67942142,
+      "epoch": 0.35755961440892947,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.553404441813554e-06,
+      "loss": 1.46771717,
+      "memory(GiB)": 103.58,
+      "step": 14095,
+      "train_speed(iter/s)": 1.313349
+    },
+    {
+      "acc": 0.68752241,
+      "epoch": 0.3576864535768645,
+      "grad_norm": 4.25,
+      "learning_rate": 9.552971144181083e-06,
+      "loss": 1.36150255,
+      "memory(GiB)": 103.58,
+      "step": 14100,
+      "train_speed(iter/s)": 1.313401
+    },
+    {
+      "acc": 0.67906666,
+      "epoch": 0.3578132927447996,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.552537646288612e-06,
+      "loss": 1.42692347,
+      "memory(GiB)": 103.58,
+      "step": 14105,
+      "train_speed(iter/s)": 1.313454
+    },
+    {
+      "acc": 0.70551233,
+      "epoch": 0.35794013191273466,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.552103948155211e-06,
+      "loss": 1.35764847,
+      "memory(GiB)": 103.58,
+      "step": 14110,
+      "train_speed(iter/s)": 1.313509
+    },
+    {
+      "acc": 0.69494205,
+      "epoch": 0.3580669710806697,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.551670049799954e-06,
+      "loss": 1.39718113,
+      "memory(GiB)": 103.58,
+      "step": 14115,
+      "train_speed(iter/s)": 1.313564
+    },
+    {
+      "acc": 0.67600589,
+      "epoch": 0.35819381024860475,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.551235951241927e-06,
+      "loss": 1.39975386,
+      "memory(GiB)": 103.58,
+      "step": 14120,
+      "train_speed(iter/s)": 1.313617
+    },
+    {
+      "acc": 0.67997694,
+      "epoch": 0.35832064941653985,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.550801652500223e-06,
+      "loss": 1.51227207,
+      "memory(GiB)": 103.58,
+      "step": 14125,
+      "train_speed(iter/s)": 1.313669
+    },
+    {
+      "acc": 0.68436089,
+      "epoch": 0.3584474885844749,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.550367153593944e-06,
+      "loss": 1.42600861,
+      "memory(GiB)": 103.58,
+      "step": 14130,
+      "train_speed(iter/s)": 1.313726
+    },
+    {
+      "acc": 0.68076382,
+      "epoch": 0.35857432775240994,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.549932454542202e-06,
+      "loss": 1.42336483,
+      "memory(GiB)": 103.58,
+      "step": 14135,
+      "train_speed(iter/s)": 1.313779
+    },
+    {
+      "acc": 0.69637756,
+      "epoch": 0.358701166920345,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.549497555364115e-06,
+      "loss": 1.34899302,
+      "memory(GiB)": 103.58,
+      "step": 14140,
+      "train_speed(iter/s)": 1.313833
+    },
+    {
+      "acc": 0.68571134,
+      "epoch": 0.3588280060882801,
+      "grad_norm": 3.625,
+      "learning_rate": 9.549062456078816e-06,
+      "loss": 1.39712162,
+      "memory(GiB)": 103.58,
+      "step": 14145,
+      "train_speed(iter/s)": 1.313886
+    },
+    {
+      "acc": 0.68506021,
+      "epoch": 0.3589548452562151,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.54862715670544e-06,
+      "loss": 1.39785128,
+      "memory(GiB)": 103.58,
+      "step": 14150,
+      "train_speed(iter/s)": 1.313937
+    },
+    {
+      "acc": 0.68565736,
+      "epoch": 0.35908168442415017,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.548191657263132e-06,
+      "loss": 1.39474936,
+      "memory(GiB)": 103.58,
+      "step": 14155,
+      "train_speed(iter/s)": 1.313992
+    },
+    {
+      "acc": 0.69479923,
+      "epoch": 0.3592085235920852,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.547755957771049e-06,
+      "loss": 1.39789639,
+      "memory(GiB)": 103.58,
+      "step": 14160,
+      "train_speed(iter/s)": 1.314044
+    },
+    {
+      "acc": 0.68900414,
+      "epoch": 0.3593353627600203,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.547320058248356e-06,
+      "loss": 1.37763424,
+      "memory(GiB)": 103.58,
+      "step": 14165,
+      "train_speed(iter/s)": 1.314098
+    },
+    {
+      "acc": 0.68689523,
+      "epoch": 0.35946220192795536,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.546883958714223e-06,
+      "loss": 1.45291042,
+      "memory(GiB)": 103.58,
+      "step": 14170,
+      "train_speed(iter/s)": 1.314154
+    },
+    {
+      "acc": 0.65436797,
+      "epoch": 0.3595890410958904,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.546447659187834e-06,
+      "loss": 1.49967003,
+      "memory(GiB)": 103.58,
+      "step": 14175,
+      "train_speed(iter/s)": 1.314206
+    },
+    {
+      "acc": 0.69412317,
+      "epoch": 0.35971588026382545,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.546011159688377e-06,
+      "loss": 1.37823324,
+      "memory(GiB)": 103.58,
+      "step": 14180,
+      "train_speed(iter/s)": 1.314261
+    },
+    {
+      "acc": 0.68350563,
+      "epoch": 0.35984271943176055,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.545574460235055e-06,
+      "loss": 1.41572857,
+      "memory(GiB)": 103.58,
+      "step": 14185,
+      "train_speed(iter/s)": 1.31431
+    },
+    {
+      "acc": 0.68803415,
+      "epoch": 0.3599695585996956,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.545137560847071e-06,
+      "loss": 1.36574154,
+      "memory(GiB)": 103.58,
+      "step": 14190,
+      "train_speed(iter/s)": 1.314363
+    },
+    {
+      "acc": 0.68054376,
+      "epoch": 0.36009639776763064,
+      "grad_norm": 3.125,
+      "learning_rate": 9.544700461543647e-06,
+      "loss": 1.40438137,
+      "memory(GiB)": 103.58,
+      "step": 14195,
+      "train_speed(iter/s)": 1.314412
+    },
+    {
+      "acc": 0.69706779,
+      "epoch": 0.3602232369355657,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.544263162344005e-06,
+      "loss": 1.38108826,
+      "memory(GiB)": 103.58,
+      "step": 14200,
+      "train_speed(iter/s)": 1.314463
+    },
+    {
+      "acc": 0.67383132,
+      "epoch": 0.3603500761035008,
+      "grad_norm": 3.125,
+      "learning_rate": 9.54382566326738e-06,
+      "loss": 1.4580286,
+      "memory(GiB)": 103.58,
+      "step": 14205,
+      "train_speed(iter/s)": 1.314515
+    },
+    {
+      "acc": 0.67701778,
+      "epoch": 0.3604769152714358,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.543387964333018e-06,
+      "loss": 1.40252171,
+      "memory(GiB)": 103.58,
+      "step": 14210,
+      "train_speed(iter/s)": 1.314565
+    },
+    {
+      "acc": 0.66741743,
+      "epoch": 0.36060375443937087,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.542950065560165e-06,
+      "loss": 1.42765675,
+      "memory(GiB)": 103.58,
+      "step": 14215,
+      "train_speed(iter/s)": 1.314616
+    },
+    {
+      "acc": 0.66758652,
+      "epoch": 0.3607305936073059,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.542511966968087e-06,
+      "loss": 1.50375004,
+      "memory(GiB)": 103.58,
+      "step": 14220,
+      "train_speed(iter/s)": 1.314669
+    },
+    {
+      "acc": 0.68875275,
+      "epoch": 0.360857432775241,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.542073668576052e-06,
+      "loss": 1.32807474,
+      "memory(GiB)": 103.58,
+      "step": 14225,
+      "train_speed(iter/s)": 1.314722
+    },
+    {
+      "acc": 0.69319506,
+      "epoch": 0.36098427194317606,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.541635170403338e-06,
+      "loss": 1.35744247,
+      "memory(GiB)": 103.58,
+      "step": 14230,
+      "train_speed(iter/s)": 1.314772
+    },
+    {
+      "acc": 0.67639151,
+      "epoch": 0.3611111111111111,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.541196472469234e-06,
+      "loss": 1.39521999,
+      "memory(GiB)": 103.58,
+      "step": 14235,
+      "train_speed(iter/s)": 1.31482
+    },
+    {
+      "acc": 0.68212552,
+      "epoch": 0.36123795027904615,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.540757574793032e-06,
+      "loss": 1.42297382,
+      "memory(GiB)": 103.58,
+      "step": 14240,
+      "train_speed(iter/s)": 1.314874
+    },
+    {
+      "acc": 0.67975016,
+      "epoch": 0.36136478944698125,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.540318477394039e-06,
+      "loss": 1.42566948,
+      "memory(GiB)": 103.58,
+      "step": 14245,
+      "train_speed(iter/s)": 1.314928
+    },
+    {
+      "acc": 0.69557562,
+      "epoch": 0.3614916286149163,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.539879180291568e-06,
+      "loss": 1.3586031,
+      "memory(GiB)": 103.58,
+      "step": 14250,
+      "train_speed(iter/s)": 1.314981
+    },
+    {
+      "acc": 0.68469658,
+      "epoch": 0.36161846778285134,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.539439683504943e-06,
+      "loss": 1.40200329,
+      "memory(GiB)": 103.58,
+      "step": 14255,
+      "train_speed(iter/s)": 1.315032
+    },
+    {
+      "acc": 0.67163844,
+      "epoch": 0.3617453069507864,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.538999987053492e-06,
+      "loss": 1.50527382,
+      "memory(GiB)": 103.58,
+      "step": 14260,
+      "train_speed(iter/s)": 1.315086
+    },
+    {
+      "acc": 0.6774703,
+      "epoch": 0.3618721461187215,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.538560090956557e-06,
+      "loss": 1.44611235,
+      "memory(GiB)": 103.58,
+      "step": 14265,
+      "train_speed(iter/s)": 1.315135
+    },
+    {
+      "acc": 0.69191141,
+      "epoch": 0.3619989852866565,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.538119995233485e-06,
+      "loss": 1.40282869,
+      "memory(GiB)": 103.58,
+      "step": 14270,
+      "train_speed(iter/s)": 1.315185
+    },
+    {
+      "acc": 0.69070253,
+      "epoch": 0.36212582445459157,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.537679699903637e-06,
+      "loss": 1.4094677,
+      "memory(GiB)": 103.58,
+      "step": 14275,
+      "train_speed(iter/s)": 1.315237
+    },
+    {
+      "acc": 0.70221682,
+      "epoch": 0.3622526636225266,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.537239204986375e-06,
+      "loss": 1.29365101,
+      "memory(GiB)": 103.58,
+      "step": 14280,
+      "train_speed(iter/s)": 1.315288
+    },
+    {
+      "acc": 0.669947,
+      "epoch": 0.3623795027904617,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.536798510501075e-06,
+      "loss": 1.43445234,
+      "memory(GiB)": 103.58,
+      "step": 14285,
+      "train_speed(iter/s)": 1.315341
+    },
+    {
+      "acc": 0.68518896,
+      "epoch": 0.36250634195839676,
+      "grad_norm": 6.3125,
+      "learning_rate": 9.536357616467123e-06,
+      "loss": 1.39248924,
+      "memory(GiB)": 103.58,
+      "step": 14290,
+      "train_speed(iter/s)": 1.315396
+    },
+    {
+      "acc": 0.68897376,
+      "epoch": 0.3626331811263318,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.535916522903908e-06,
+      "loss": 1.36954384,
+      "memory(GiB)": 103.58,
+      "step": 14295,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.66773982,
+      "epoch": 0.36276002029426685,
+      "grad_norm": 3.75,
+      "learning_rate": 9.535475229830832e-06,
+      "loss": 1.49354296,
+      "memory(GiB)": 103.58,
+      "step": 14300,
+      "train_speed(iter/s)": 1.315498
+    },
+    {
+      "acc": 0.67857366,
+      "epoch": 0.36288685946220195,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.535033737267308e-06,
+      "loss": 1.36144838,
+      "memory(GiB)": 103.58,
+      "step": 14305,
+      "train_speed(iter/s)": 1.315549
+    },
+    {
+      "acc": 0.68389778,
+      "epoch": 0.363013698630137,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.534592045232752e-06,
+      "loss": 1.33101482,
+      "memory(GiB)": 103.58,
+      "step": 14310,
+      "train_speed(iter/s)": 1.3156
+    },
+    {
+      "acc": 0.67882662,
+      "epoch": 0.36314053779807204,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.534150153746591e-06,
+      "loss": 1.49811735,
+      "memory(GiB)": 103.58,
+      "step": 14315,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.67661486,
+      "epoch": 0.3632673769660071,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.533708062828264e-06,
+      "loss": 1.3984416,
+      "memory(GiB)": 103.58,
+      "step": 14320,
+      "train_speed(iter/s)": 1.315705
+    },
+    {
+      "acc": 0.67896023,
+      "epoch": 0.3633942161339422,
+      "grad_norm": 3.0,
+      "learning_rate": 9.533265772497216e-06,
+      "loss": 1.4684413,
+      "memory(GiB)": 103.58,
+      "step": 14325,
+      "train_speed(iter/s)": 1.315756
+    },
+    {
+      "acc": 0.69408998,
+      "epoch": 0.3635210553018772,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.532823282772899e-06,
+      "loss": 1.37793388,
+      "memory(GiB)": 103.58,
+      "step": 14330,
+      "train_speed(iter/s)": 1.315807
+    },
+    {
+      "acc": 0.67539234,
+      "epoch": 0.36364789446981227,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.532380593674775e-06,
+      "loss": 1.40496025,
+      "memory(GiB)": 103.58,
+      "step": 14335,
+      "train_speed(iter/s)": 1.315856
+    },
+    {
+      "acc": 0.67932649,
+      "epoch": 0.3637747336377473,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.531937705222319e-06,
+      "loss": 1.45443115,
+      "memory(GiB)": 103.58,
+      "step": 14340,
+      "train_speed(iter/s)": 1.315908
+    },
+    {
+      "acc": 0.6738049,
+      "epoch": 0.3639015728056824,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.531494617435006e-06,
+      "loss": 1.4017417,
+      "memory(GiB)": 103.58,
+      "step": 14345,
+      "train_speed(iter/s)": 1.315961
+    },
+    {
+      "acc": 0.6842577,
+      "epoch": 0.36402841197361746,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.531051330332331e-06,
+      "loss": 1.37059746,
+      "memory(GiB)": 103.58,
+      "step": 14350,
+      "train_speed(iter/s)": 1.316015
+    },
+    {
+      "acc": 0.67064648,
+      "epoch": 0.3641552511415525,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.530607843933788e-06,
+      "loss": 1.45690041,
+      "memory(GiB)": 103.58,
+      "step": 14355,
+      "train_speed(iter/s)": 1.316065
+    },
+    {
+      "acc": 0.68407192,
+      "epoch": 0.36428209030948755,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.530164158258883e-06,
+      "loss": 1.34965649,
+      "memory(GiB)": 103.58,
+      "step": 14360,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.69032669,
+      "epoch": 0.36440892947742265,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.529720273327135e-06,
+      "loss": 1.38649235,
+      "memory(GiB)": 103.58,
+      "step": 14365,
+      "train_speed(iter/s)": 1.31617
+    },
+    {
+      "acc": 0.67872,
+      "epoch": 0.3645357686453577,
+      "grad_norm": 3.625,
+      "learning_rate": 9.529276189158063e-06,
+      "loss": 1.45774622,
+      "memory(GiB)": 103.58,
+      "step": 14370,
+      "train_speed(iter/s)": 1.316222
+    },
+    {
+      "acc": 0.6869071,
+      "epoch": 0.36466260781329274,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.528831905771205e-06,
+      "loss": 1.41424255,
+      "memory(GiB)": 103.58,
+      "step": 14375,
+      "train_speed(iter/s)": 1.316273
+    },
+    {
+      "acc": 0.67963185,
+      "epoch": 0.3647894469812278,
+      "grad_norm": 3.75,
+      "learning_rate": 9.528387423186098e-06,
+      "loss": 1.44471655,
+      "memory(GiB)": 103.58,
+      "step": 14380,
+      "train_speed(iter/s)": 1.316325
+    },
+    {
+      "acc": 0.67926221,
+      "epoch": 0.3649162861491629,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.527942741422297e-06,
+      "loss": 1.39108219,
+      "memory(GiB)": 103.58,
+      "step": 14385,
+      "train_speed(iter/s)": 1.316378
+    },
+    {
+      "acc": 0.67461376,
+      "epoch": 0.3650431253170979,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.527497860499355e-06,
+      "loss": 1.50502186,
+      "memory(GiB)": 103.58,
+      "step": 14390,
+      "train_speed(iter/s)": 1.31643
+    },
+    {
+      "acc": 0.69668379,
+      "epoch": 0.36516996448503297,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.527052780436845e-06,
+      "loss": 1.30743055,
+      "memory(GiB)": 103.58,
+      "step": 14395,
+      "train_speed(iter/s)": 1.316482
+    },
+    {
+      "acc": 0.67809682,
+      "epoch": 0.365296803652968,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.52660750125434e-06,
+      "loss": 1.39875383,
+      "memory(GiB)": 103.58,
+      "step": 14400,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.68440609,
+      "epoch": 0.3654236428209031,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.52616202297143e-06,
+      "loss": 1.42829132,
+      "memory(GiB)": 103.58,
+      "step": 14405,
+      "train_speed(iter/s)": 1.316585
+    },
+    {
+      "acc": 0.67633495,
+      "epoch": 0.36555048198883816,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.525716345607706e-06,
+      "loss": 1.41851406,
+      "memory(GiB)": 103.58,
+      "step": 14410,
+      "train_speed(iter/s)": 1.316634
+    },
+    {
+      "acc": 0.6607687,
+      "epoch": 0.3656773211567732,
+      "grad_norm": 2.75,
+      "learning_rate": 9.52527046918277e-06,
+      "loss": 1.48206959,
+      "memory(GiB)": 103.58,
+      "step": 14415,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.66489429,
+      "epoch": 0.36580416032470825,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.524824393716235e-06,
+      "loss": 1.4866993,
+      "memory(GiB)": 103.58,
+      "step": 14420,
+      "train_speed(iter/s)": 1.316734
+    },
+    {
+      "acc": 0.67854934,
+      "epoch": 0.36593099949264335,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.524378119227722e-06,
+      "loss": 1.40110769,
+      "memory(GiB)": 103.58,
+      "step": 14425,
+      "train_speed(iter/s)": 1.316787
+    },
+    {
+      "acc": 0.68699675,
+      "epoch": 0.3660578386605784,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.523931645736858e-06,
+      "loss": 1.4237154,
+      "memory(GiB)": 103.58,
+      "step": 14430,
+      "train_speed(iter/s)": 1.316841
+    },
+    {
+      "acc": 0.68139668,
+      "epoch": 0.36618467782851344,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.523484973263283e-06,
+      "loss": 1.43569031,
+      "memory(GiB)": 103.58,
+      "step": 14435,
+      "train_speed(iter/s)": 1.316891
+    },
+    {
+      "acc": 0.69837427,
+      "epoch": 0.3663115169964485,
+      "grad_norm": 3.5,
+      "learning_rate": 9.523038101826644e-06,
+      "loss": 1.36664162,
+      "memory(GiB)": 103.58,
+      "step": 14440,
+      "train_speed(iter/s)": 1.316944
+    },
+    {
+      "acc": 0.66168041,
+      "epoch": 0.3664383561643836,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.522591031446596e-06,
+      "loss": 1.49072189,
+      "memory(GiB)": 103.58,
+      "step": 14445,
+      "train_speed(iter/s)": 1.316995
+    },
+    {
+      "acc": 0.68697176,
+      "epoch": 0.3665651953323186,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.522143762142801e-06,
+      "loss": 1.36484184,
+      "memory(GiB)": 103.58,
+      "step": 14450,
+      "train_speed(iter/s)": 1.317047
+    },
+    {
+      "acc": 0.69969635,
+      "epoch": 0.36669203450025367,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.521696293934934e-06,
+      "loss": 1.36798325,
+      "memory(GiB)": 103.58,
+      "step": 14455,
+      "train_speed(iter/s)": 1.317098
+    },
+    {
+      "acc": 0.67956872,
+      "epoch": 0.3668188736681887,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.521248626842676e-06,
+      "loss": 1.43998079,
+      "memory(GiB)": 103.58,
+      "step": 14460,
+      "train_speed(iter/s)": 1.317151
+    },
+    {
+      "acc": 0.67291956,
+      "epoch": 0.3669457128361238,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.520800760885716e-06,
+      "loss": 1.45842562,
+      "memory(GiB)": 103.58,
+      "step": 14465,
+      "train_speed(iter/s)": 1.317204
+    },
+    {
+      "acc": 0.68054037,
+      "epoch": 0.36707255200405886,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.520352696083756e-06,
+      "loss": 1.37788324,
+      "memory(GiB)": 103.58,
+      "step": 14470,
+      "train_speed(iter/s)": 1.317256
+    },
+    {
+      "acc": 0.69234786,
+      "epoch": 0.3671993911719939,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.519904432456504e-06,
+      "loss": 1.32905378,
+      "memory(GiB)": 103.58,
+      "step": 14475,
+      "train_speed(iter/s)": 1.317309
+    },
+    {
+      "acc": 0.70114908,
+      "epoch": 0.36732623033992895,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.519455970023672e-06,
+      "loss": 1.32123384,
+      "memory(GiB)": 103.58,
+      "step": 14480,
+      "train_speed(iter/s)": 1.317362
+    },
+    {
+      "acc": 0.67649951,
+      "epoch": 0.36745306950786405,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.519007308804991e-06,
+      "loss": 1.4241971,
+      "memory(GiB)": 103.58,
+      "step": 14485,
+      "train_speed(iter/s)": 1.317407
+    },
+    {
+      "acc": 0.67552471,
+      "epoch": 0.3675799086757991,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.518558448820193e-06,
+      "loss": 1.36441097,
+      "memory(GiB)": 103.58,
+      "step": 14490,
+      "train_speed(iter/s)": 1.317459
+    },
+    {
+      "acc": 0.68302374,
+      "epoch": 0.36770674784373414,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.518109390089017e-06,
+      "loss": 1.40718622,
+      "memory(GiB)": 103.58,
+      "step": 14495,
+      "train_speed(iter/s)": 1.317512
+    },
+    {
+      "acc": 0.67325349,
+      "epoch": 0.3678335870116692,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.517660132631222e-06,
+      "loss": 1.46716728,
+      "memory(GiB)": 103.58,
+      "step": 14500,
+      "train_speed(iter/s)": 1.317565
+    },
+    {
+      "acc": 0.70104146,
+      "epoch": 0.3679604261796043,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.517210676466561e-06,
+      "loss": 1.40424652,
+      "memory(GiB)": 103.58,
+      "step": 14505,
+      "train_speed(iter/s)": 1.317619
+    },
+    {
+      "acc": 0.68281059,
+      "epoch": 0.3680872653475393,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.516761021614809e-06,
+      "loss": 1.39461575,
+      "memory(GiB)": 103.58,
+      "step": 14510,
+      "train_speed(iter/s)": 1.317667
+    },
+    {
+      "acc": 0.68660188,
+      "epoch": 0.36821410451547437,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.51631116809574e-06,
+      "loss": 1.40411663,
+      "memory(GiB)": 103.58,
+      "step": 14515,
+      "train_speed(iter/s)": 1.317714
+    },
+    {
+      "acc": 0.67458158,
+      "epoch": 0.3683409436834094,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.515861115929144e-06,
+      "loss": 1.48781471,
+      "memory(GiB)": 103.58,
+      "step": 14520,
+      "train_speed(iter/s)": 1.317768
+    },
+    {
+      "acc": 0.68273201,
+      "epoch": 0.3684677828513445,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.515410865134812e-06,
+      "loss": 1.40447121,
+      "memory(GiB)": 103.58,
+      "step": 14525,
+      "train_speed(iter/s)": 1.317819
+    },
+    {
+      "acc": 0.69620275,
+      "epoch": 0.36859462201927956,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.514960415732551e-06,
+      "loss": 1.33710833,
+      "memory(GiB)": 103.58,
+      "step": 14530,
+      "train_speed(iter/s)": 1.31787
+    },
+    {
+      "acc": 0.67783785,
+      "epoch": 0.3687214611872146,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.514509767742172e-06,
+      "loss": 1.4639389,
+      "memory(GiB)": 103.58,
+      "step": 14535,
+      "train_speed(iter/s)": 1.317922
+    },
+    {
+      "acc": 0.68120699,
+      "epoch": 0.36884830035514965,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.5140589211835e-06,
+      "loss": 1.40461531,
+      "memory(GiB)": 103.58,
+      "step": 14540,
+      "train_speed(iter/s)": 1.317973
+    },
+    {
+      "acc": 0.68953519,
+      "epoch": 0.36897513952308475,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.513607876076363e-06,
+      "loss": 1.46138601,
+      "memory(GiB)": 103.58,
+      "step": 14545,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.69569535,
+      "epoch": 0.3691019786910198,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.513156632440598e-06,
+      "loss": 1.37434988,
+      "memory(GiB)": 103.58,
+      "step": 14550,
+      "train_speed(iter/s)": 1.318075
+    },
+    {
+      "acc": 0.6810132,
+      "epoch": 0.36922881785895484,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.512705190296055e-06,
+      "loss": 1.42156801,
+      "memory(GiB)": 103.58,
+      "step": 14555,
+      "train_speed(iter/s)": 1.318126
+    },
+    {
+      "acc": 0.69375319,
+      "epoch": 0.3693556570268899,
+      "grad_norm": 2.875,
+      "learning_rate": 9.512253549662588e-06,
+      "loss": 1.36109629,
+      "memory(GiB)": 103.58,
+      "step": 14560,
+      "train_speed(iter/s)": 1.318177
+    },
+    {
+      "acc": 0.68727083,
+      "epoch": 0.369482496194825,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.511801710560066e-06,
+      "loss": 1.48085842,
+      "memory(GiB)": 103.58,
+      "step": 14565,
+      "train_speed(iter/s)": 1.318229
+    },
+    {
+      "acc": 0.69372988,
+      "epoch": 0.36960933536276,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.511349673008364e-06,
+      "loss": 1.36045017,
+      "memory(GiB)": 103.58,
+      "step": 14570,
+      "train_speed(iter/s)": 1.31828
+    },
+    {
+      "acc": 0.68540998,
+      "epoch": 0.36973617453069507,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.510897437027358e-06,
+      "loss": 1.4130372,
+      "memory(GiB)": 103.58,
+      "step": 14575,
+      "train_speed(iter/s)": 1.318329
+    },
+    {
+      "acc": 0.69208136,
+      "epoch": 0.3698630136986301,
+      "grad_norm": 3.375,
+      "learning_rate": 9.510445002636943e-06,
+      "loss": 1.40549622,
+      "memory(GiB)": 103.58,
+      "step": 14580,
+      "train_speed(iter/s)": 1.318382
+    },
+    {
+      "acc": 0.6974113,
+      "epoch": 0.3699898528665652,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.50999236985702e-06,
+      "loss": 1.38590145,
+      "memory(GiB)": 103.58,
+      "step": 14585,
+      "train_speed(iter/s)": 1.318432
+    },
+    {
+      "acc": 0.68241663,
+      "epoch": 0.37011669203450026,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.509539538707497e-06,
+      "loss": 1.46578598,
+      "memory(GiB)": 103.58,
+      "step": 14590,
+      "train_speed(iter/s)": 1.318483
+    },
+    {
+      "acc": 0.67789135,
+      "epoch": 0.3702435312024353,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.50908650920829e-06,
+      "loss": 1.40185471,
+      "memory(GiB)": 103.58,
+      "step": 14595,
+      "train_speed(iter/s)": 1.318533
+    },
+    {
+      "acc": 0.67564278,
+      "epoch": 0.37037037037037035,
+      "grad_norm": 2.625,
+      "learning_rate": 9.50863328137933e-06,
+      "loss": 1.39575033,
+      "memory(GiB)": 103.58,
+      "step": 14600,
+      "train_speed(iter/s)": 1.318581
+    },
+    {
+      "acc": 0.68614626,
+      "epoch": 0.37049720953830545,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.508179855240545e-06,
+      "loss": 1.37208557,
+      "memory(GiB)": 103.58,
+      "step": 14605,
+      "train_speed(iter/s)": 1.31863
+    },
+    {
+      "acc": 0.68858919,
+      "epoch": 0.3706240487062405,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.507726230811884e-06,
+      "loss": 1.46599417,
+      "memory(GiB)": 103.58,
+      "step": 14610,
+      "train_speed(iter/s)": 1.31868
+    },
+    {
+      "acc": 0.67957172,
+      "epoch": 0.37075088787417554,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.507272408113298e-06,
+      "loss": 1.38050957,
+      "memory(GiB)": 103.58,
+      "step": 14615,
+      "train_speed(iter/s)": 1.318729
+    },
+    {
+      "acc": 0.66567879,
+      "epoch": 0.3708777270421106,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.506818387164748e-06,
+      "loss": 1.46035538,
+      "memory(GiB)": 103.58,
+      "step": 14620,
+      "train_speed(iter/s)": 1.318777
+    },
+    {
+      "acc": 0.67802186,
+      "epoch": 0.3710045662100457,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.506364167986204e-06,
+      "loss": 1.46798468,
+      "memory(GiB)": 103.58,
+      "step": 14625,
+      "train_speed(iter/s)": 1.318819
+    },
+    {
+      "acc": 0.69207821,
+      "epoch": 0.3711314053779807,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.505909750597644e-06,
+      "loss": 1.42180119,
+      "memory(GiB)": 103.58,
+      "step": 14630,
+      "train_speed(iter/s)": 1.318871
+    },
+    {
+      "acc": 0.67685919,
+      "epoch": 0.37125824454591577,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.505455135019055e-06,
+      "loss": 1.38446674,
+      "memory(GiB)": 103.58,
+      "step": 14635,
+      "train_speed(iter/s)": 1.318922
+    },
+    {
+      "acc": 0.67986164,
+      "epoch": 0.3713850837138508,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.505000321270435e-06,
+      "loss": 1.39056396,
+      "memory(GiB)": 103.58,
+      "step": 14640,
+      "train_speed(iter/s)": 1.318975
+    },
+    {
+      "acc": 0.69385757,
+      "epoch": 0.3715119228817859,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.504545309371786e-06,
+      "loss": 1.40860748,
+      "memory(GiB)": 103.58,
+      "step": 14645,
+      "train_speed(iter/s)": 1.319025
+    },
+    {
+      "acc": 0.66716881,
+      "epoch": 0.37163876204972096,
+      "grad_norm": 2.75,
+      "learning_rate": 9.504090099343125e-06,
+      "loss": 1.46407299,
+      "memory(GiB)": 103.58,
+      "step": 14650,
+      "train_speed(iter/s)": 1.319075
+    },
+    {
+      "acc": 0.71185184,
+      "epoch": 0.371765601217656,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.50363469120447e-06,
+      "loss": 1.32774229,
+      "memory(GiB)": 103.58,
+      "step": 14655,
+      "train_speed(iter/s)": 1.319124
+    },
+    {
+      "acc": 0.66650686,
+      "epoch": 0.37189244038559105,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.503179084975855e-06,
+      "loss": 1.46908684,
+      "memory(GiB)": 103.58,
+      "step": 14660,
+      "train_speed(iter/s)": 1.319174
+    },
+    {
+      "acc": 0.67588992,
+      "epoch": 0.37201927955352615,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.502723280677319e-06,
+      "loss": 1.43501167,
+      "memory(GiB)": 103.58,
+      "step": 14665,
+      "train_speed(iter/s)": 1.319215
+    },
+    {
+      "acc": 0.68793278,
+      "epoch": 0.3721461187214612,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.50226727832891e-06,
+      "loss": 1.44878778,
+      "memory(GiB)": 103.58,
+      "step": 14670,
+      "train_speed(iter/s)": 1.319266
+    },
+    {
+      "acc": 0.66972136,
+      "epoch": 0.37227295788939624,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.501811077950685e-06,
+      "loss": 1.4474328,
+      "memory(GiB)": 103.58,
+      "step": 14675,
+      "train_speed(iter/s)": 1.319309
+    },
+    {
+      "acc": 0.68765297,
+      "epoch": 0.3723997970573313,
+      "grad_norm": 3.25,
+      "learning_rate": 9.501354679562708e-06,
+      "loss": 1.39094677,
+      "memory(GiB)": 103.58,
+      "step": 14680,
+      "train_speed(iter/s)": 1.31936
+    },
+    {
+      "acc": 0.68173876,
+      "epoch": 0.3725266362252664,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.500898083185058e-06,
+      "loss": 1.42425728,
+      "memory(GiB)": 103.58,
+      "step": 14685,
+      "train_speed(iter/s)": 1.319411
+    },
+    {
+      "acc": 0.68513689,
+      "epoch": 0.3726534753932014,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.500441288837812e-06,
+      "loss": 1.37179203,
+      "memory(GiB)": 103.58,
+      "step": 14690,
+      "train_speed(iter/s)": 1.319463
+    },
+    {
+      "acc": 0.6792532,
+      "epoch": 0.37278031456113647,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.499984296541066e-06,
+      "loss": 1.46889019,
+      "memory(GiB)": 103.58,
+      "step": 14695,
+      "train_speed(iter/s)": 1.319512
+    },
+    {
+      "acc": 0.66009822,
+      "epoch": 0.3729071537290715,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.49952710631492e-06,
+      "loss": 1.53338833,
+      "memory(GiB)": 103.58,
+      "step": 14700,
+      "train_speed(iter/s)": 1.319562
+    },
+    {
+      "acc": 0.68716521,
+      "epoch": 0.3730339928970066,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.499069718179484e-06,
+      "loss": 1.38632116,
+      "memory(GiB)": 103.58,
+      "step": 14705,
+      "train_speed(iter/s)": 1.31961
+    },
+    {
+      "acc": 0.69105506,
+      "epoch": 0.37316083206494166,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.498612132154874e-06,
+      "loss": 1.34160213,
+      "memory(GiB)": 103.58,
+      "step": 14710,
+      "train_speed(iter/s)": 1.31966
+    },
+    {
+      "acc": 0.67832465,
+      "epoch": 0.3732876712328767,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.498154348261217e-06,
+      "loss": 1.40472031,
+      "memory(GiB)": 103.58,
+      "step": 14715,
+      "train_speed(iter/s)": 1.319706
+    },
+    {
+      "acc": 0.69242959,
+      "epoch": 0.37341451040081175,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.497696366518649e-06,
+      "loss": 1.39179382,
+      "memory(GiB)": 103.58,
+      "step": 14720,
+      "train_speed(iter/s)": 1.319757
+    },
+    {
+      "acc": 0.68216019,
+      "epoch": 0.37354134956874685,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.497238186947315e-06,
+      "loss": 1.46256676,
+      "memory(GiB)": 103.58,
+      "step": 14725,
+      "train_speed(iter/s)": 1.319806
+    },
+    {
+      "acc": 0.67211256,
+      "epoch": 0.3736681887366819,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.496779809567367e-06,
+      "loss": 1.48941975,
+      "memory(GiB)": 103.58,
+      "step": 14730,
+      "train_speed(iter/s)": 1.319856
+    },
+    {
+      "acc": 0.68818874,
+      "epoch": 0.37379502790461694,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.496321234398967e-06,
+      "loss": 1.34828539,
+      "memory(GiB)": 103.58,
+      "step": 14735,
+      "train_speed(iter/s)": 1.319906
+    },
+    {
+      "acc": 0.67722292,
+      "epoch": 0.373921867072552,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.495862461462282e-06,
+      "loss": 1.41558657,
+      "memory(GiB)": 103.58,
+      "step": 14740,
+      "train_speed(iter/s)": 1.319953
+    },
+    {
+      "acc": 0.67988338,
+      "epoch": 0.3740487062404871,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.495403490777495e-06,
+      "loss": 1.42959566,
+      "memory(GiB)": 103.58,
+      "step": 14745,
+      "train_speed(iter/s)": 1.320002
+    },
+    {
+      "acc": 0.70135956,
+      "epoch": 0.3741755454084221,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.49494432236479e-06,
+      "loss": 1.34868317,
+      "memory(GiB)": 103.58,
+      "step": 14750,
+      "train_speed(iter/s)": 1.320051
+    },
+    {
+      "acc": 0.68288736,
+      "epoch": 0.37430238457635717,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.494484956244368e-06,
+      "loss": 1.41513491,
+      "memory(GiB)": 103.58,
+      "step": 14755,
+      "train_speed(iter/s)": 1.320101
+    },
+    {
+      "acc": 0.68128958,
+      "epoch": 0.3744292237442922,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.49402539243643e-06,
+      "loss": 1.4547617,
+      "memory(GiB)": 103.58,
+      "step": 14760,
+      "train_speed(iter/s)": 1.32015
+    },
+    {
+      "acc": 0.68640156,
+      "epoch": 0.3745560629122273,
+      "grad_norm": 2.75,
+      "learning_rate": 9.49356563096119e-06,
+      "loss": 1.41374302,
+      "memory(GiB)": 103.58,
+      "step": 14765,
+      "train_speed(iter/s)": 1.3202
+    },
+    {
+      "acc": 0.68415403,
+      "epoch": 0.37468290208016236,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.49310567183887e-06,
+      "loss": 1.39714241,
+      "memory(GiB)": 103.58,
+      "step": 14770,
+      "train_speed(iter/s)": 1.320252
+    },
+    {
+      "acc": 0.68352623,
+      "epoch": 0.3748097412480974,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.492645515089706e-06,
+      "loss": 1.41681213,
+      "memory(GiB)": 103.58,
+      "step": 14775,
+      "train_speed(iter/s)": 1.320304
+    },
+    {
+      "acc": 0.67981853,
+      "epoch": 0.37493658041603245,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.492185160733934e-06,
+      "loss": 1.39052258,
+      "memory(GiB)": 103.58,
+      "step": 14780,
+      "train_speed(iter/s)": 1.320353
+    },
+    {
+      "acc": 0.67967281,
+      "epoch": 0.37506341958396755,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.491724608791798e-06,
+      "loss": 1.42391205,
+      "memory(GiB)": 103.58,
+      "step": 14785,
+      "train_speed(iter/s)": 1.320402
+    },
+    {
+      "acc": 0.67461367,
+      "epoch": 0.3751902587519026,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.491263859283563e-06,
+      "loss": 1.37816248,
+      "memory(GiB)": 103.58,
+      "step": 14790,
+      "train_speed(iter/s)": 1.320452
+    },
+    {
+      "acc": 0.69312029,
+      "epoch": 0.37531709791983764,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.490802912229491e-06,
+      "loss": 1.39321699,
+      "memory(GiB)": 103.58,
+      "step": 14795,
+      "train_speed(iter/s)": 1.320502
+    },
+    {
+      "acc": 0.67217054,
+      "epoch": 0.3754439370877727,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.490341767649858e-06,
+      "loss": 1.47721386,
+      "memory(GiB)": 103.58,
+      "step": 14800,
+      "train_speed(iter/s)": 1.320551
+    },
+    {
+      "acc": 0.69383621,
+      "epoch": 0.3755707762557078,
+      "grad_norm": 3.25,
+      "learning_rate": 9.489880425564944e-06,
+      "loss": 1.35775652,
+      "memory(GiB)": 103.58,
+      "step": 14805,
+      "train_speed(iter/s)": 1.320599
+    },
+    {
+      "acc": 0.68002691,
+      "epoch": 0.37569761542364283,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.489418885995043e-06,
+      "loss": 1.40177822,
+      "memory(GiB)": 103.58,
+      "step": 14810,
+      "train_speed(iter/s)": 1.320649
+    },
+    {
+      "acc": 0.66318917,
+      "epoch": 0.3758244545915779,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.488957148960457e-06,
+      "loss": 1.50088387,
+      "memory(GiB)": 103.58,
+      "step": 14815,
+      "train_speed(iter/s)": 1.320696
+    },
+    {
+      "acc": 0.67276483,
+      "epoch": 0.3759512937595129,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.488495214481494e-06,
+      "loss": 1.39578037,
+      "memory(GiB)": 103.58,
+      "step": 14820,
+      "train_speed(iter/s)": 1.320747
+    },
+    {
+      "acc": 0.69477639,
+      "epoch": 0.376078132927448,
+      "grad_norm": 2.875,
+      "learning_rate": 9.48803308257847e-06,
+      "loss": 1.40247574,
+      "memory(GiB)": 103.58,
+      "step": 14825,
+      "train_speed(iter/s)": 1.320796
+    },
+    {
+      "acc": 0.69584365,
+      "epoch": 0.37620497209538306,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.487570753271716e-06,
+      "loss": 1.3717309,
+      "memory(GiB)": 103.58,
+      "step": 14830,
+      "train_speed(iter/s)": 1.320846
+    },
+    {
+      "acc": 0.66839314,
+      "epoch": 0.3763318112633181,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.487108226581564e-06,
+      "loss": 1.43645487,
+      "memory(GiB)": 103.58,
+      "step": 14835,
+      "train_speed(iter/s)": 1.320893
+    },
+    {
+      "acc": 0.68764534,
+      "epoch": 0.37645865043125315,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.486645502528355e-06,
+      "loss": 1.39920139,
+      "memory(GiB)": 103.58,
+      "step": 14840,
+      "train_speed(iter/s)": 1.320942
+    },
+    {
+      "acc": 0.69013252,
+      "epoch": 0.37658548959918825,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.486182581132449e-06,
+      "loss": 1.36902409,
+      "memory(GiB)": 103.58,
+      "step": 14845,
+      "train_speed(iter/s)": 1.320989
+    },
+    {
+      "acc": 0.68428187,
+      "epoch": 0.3767123287671233,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.485719462414202e-06,
+      "loss": 1.35288029,
+      "memory(GiB)": 103.58,
+      "step": 14850,
+      "train_speed(iter/s)": 1.321036
+    },
+    {
+      "acc": 0.69350243,
+      "epoch": 0.37683916793505834,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.485256146393987e-06,
+      "loss": 1.47380638,
+      "memory(GiB)": 103.58,
+      "step": 14855,
+      "train_speed(iter/s)": 1.321086
+    },
+    {
+      "acc": 0.68620739,
+      "epoch": 0.3769660071029934,
+      "grad_norm": 3.375,
+      "learning_rate": 9.484792633092182e-06,
+      "loss": 1.41181116,
+      "memory(GiB)": 103.58,
+      "step": 14860,
+      "train_speed(iter/s)": 1.321135
+    },
+    {
+      "acc": 0.69182749,
+      "epoch": 0.3770928462709285,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.484328922529172e-06,
+      "loss": 1.33957014,
+      "memory(GiB)": 103.58,
+      "step": 14865,
+      "train_speed(iter/s)": 1.321184
+    },
+    {
+      "acc": 0.69114008,
+      "epoch": 0.37721968543886353,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.483865014725356e-06,
+      "loss": 1.34595709,
+      "memory(GiB)": 103.58,
+      "step": 14870,
+      "train_speed(iter/s)": 1.32123
+    },
+    {
+      "acc": 0.68705282,
+      "epoch": 0.3773465246067986,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.483400909701139e-06,
+      "loss": 1.42262669,
+      "memory(GiB)": 103.58,
+      "step": 14875,
+      "train_speed(iter/s)": 1.321277
+    },
+    {
+      "acc": 0.67561083,
+      "epoch": 0.3774733637747336,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.482936607476931e-06,
+      "loss": 1.44354248,
+      "memory(GiB)": 103.58,
+      "step": 14880,
+      "train_speed(iter/s)": 1.321323
+    },
+    {
+      "acc": 0.67523928,
+      "epoch": 0.3776002029426687,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.482472108073157e-06,
+      "loss": 1.46803513,
+      "memory(GiB)": 103.58,
+      "step": 14885,
+      "train_speed(iter/s)": 1.32137
+    },
+    {
+      "acc": 0.67517099,
+      "epoch": 0.37772704211060376,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.482007411510245e-06,
+      "loss": 1.40851641,
+      "memory(GiB)": 103.58,
+      "step": 14890,
+      "train_speed(iter/s)": 1.321416
+    },
+    {
+      "acc": 0.69090562,
+      "epoch": 0.3778538812785388,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.48154251780864e-06,
+      "loss": 1.33897991,
+      "memory(GiB)": 103.58,
+      "step": 14895,
+      "train_speed(iter/s)": 1.321462
+    },
+    {
+      "acc": 0.68476396,
+      "epoch": 0.37798072044647385,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.481077426988782e-06,
+      "loss": 1.42288656,
+      "memory(GiB)": 103.58,
+      "step": 14900,
+      "train_speed(iter/s)": 1.32151
+    },
+    {
+      "acc": 0.67105446,
+      "epoch": 0.37810755961440895,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.480612139071134e-06,
+      "loss": 1.42766705,
+      "memory(GiB)": 103.58,
+      "step": 14905,
+      "train_speed(iter/s)": 1.321556
+    },
+    {
+      "acc": 0.67084923,
+      "epoch": 0.378234398782344,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.48014665407616e-06,
+      "loss": 1.44598637,
+      "memory(GiB)": 103.58,
+      "step": 14910,
+      "train_speed(iter/s)": 1.321602
+    },
+    {
+      "acc": 0.70536032,
+      "epoch": 0.37836123795027904,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.479680972024334e-06,
+      "loss": 1.38828335,
+      "memory(GiB)": 103.58,
+      "step": 14915,
+      "train_speed(iter/s)": 1.32165
+    },
+    {
+      "acc": 0.69640713,
+      "epoch": 0.3784880771182141,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.47921509293614e-06,
+      "loss": 1.36076651,
+      "memory(GiB)": 103.58,
+      "step": 14920,
+      "train_speed(iter/s)": 1.321697
+    },
+    {
+      "acc": 0.69137368,
+      "epoch": 0.3786149162861492,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.478749016832066e-06,
+      "loss": 1.4602026,
+      "memory(GiB)": 103.58,
+      "step": 14925,
+      "train_speed(iter/s)": 1.321744
+    },
+    {
+      "acc": 0.68708434,
+      "epoch": 0.37874175545408423,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.478282743732613e-06,
+      "loss": 1.36052189,
+      "memory(GiB)": 103.58,
+      "step": 14930,
+      "train_speed(iter/s)": 1.321791
+    },
+    {
+      "acc": 0.68509283,
+      "epoch": 0.3788685946220193,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.477816273658293e-06,
+      "loss": 1.42767181,
+      "memory(GiB)": 103.58,
+      "step": 14935,
+      "train_speed(iter/s)": 1.321841
+    },
+    {
+      "acc": 0.67984033,
+      "epoch": 0.3789954337899543,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.47734960662962e-06,
+      "loss": 1.44425383,
+      "memory(GiB)": 103.58,
+      "step": 14940,
+      "train_speed(iter/s)": 1.321887
+    },
+    {
+      "acc": 0.68639202,
+      "epoch": 0.3791222729578894,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.476882742667122e-06,
+      "loss": 1.40860519,
+      "memory(GiB)": 103.58,
+      "step": 14945,
+      "train_speed(iter/s)": 1.321936
+    },
+    {
+      "acc": 0.67897859,
+      "epoch": 0.37924911212582446,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.476415681791333e-06,
+      "loss": 1.35750742,
+      "memory(GiB)": 103.58,
+      "step": 14950,
+      "train_speed(iter/s)": 1.321985
+    },
+    {
+      "acc": 0.67991924,
+      "epoch": 0.3793759512937595,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.475948424022798e-06,
+      "loss": 1.38469086,
+      "memory(GiB)": 103.58,
+      "step": 14955,
+      "train_speed(iter/s)": 1.322034
+    },
+    {
+      "acc": 0.69297533,
+      "epoch": 0.37950279046169455,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.475480969382065e-06,
+      "loss": 1.35577717,
+      "memory(GiB)": 103.58,
+      "step": 14960,
+      "train_speed(iter/s)": 1.322082
+    },
+    {
+      "acc": 0.68193936,
+      "epoch": 0.37962962962962965,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.475013317889699e-06,
+      "loss": 1.38740139,
+      "memory(GiB)": 103.58,
+      "step": 14965,
+      "train_speed(iter/s)": 1.32213
+    },
+    {
+      "acc": 0.68903255,
+      "epoch": 0.3797564687975647,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.474545469566267e-06,
+      "loss": 1.37292318,
+      "memory(GiB)": 103.58,
+      "step": 14970,
+      "train_speed(iter/s)": 1.322179
+    },
+    {
+      "acc": 0.67497611,
+      "epoch": 0.37988330796549974,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.474077424432348e-06,
+      "loss": 1.3974082,
+      "memory(GiB)": 103.58,
+      "step": 14975,
+      "train_speed(iter/s)": 1.322226
+    },
+    {
+      "acc": 0.67931871,
+      "epoch": 0.3800101471334348,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.47360918250853e-06,
+      "loss": 1.4694067,
+      "memory(GiB)": 103.58,
+      "step": 14980,
+      "train_speed(iter/s)": 1.322275
+    },
+    {
+      "acc": 0.68299975,
+      "epoch": 0.3801369863013699,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.473140743815405e-06,
+      "loss": 1.4876543,
+      "memory(GiB)": 103.58,
+      "step": 14985,
+      "train_speed(iter/s)": 1.322327
+    },
+    {
+      "acc": 0.70347233,
+      "epoch": 0.38026382546930493,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.47267210837358e-06,
+      "loss": 1.35990562,
+      "memory(GiB)": 103.58,
+      "step": 14990,
+      "train_speed(iter/s)": 1.322374
+    },
+    {
+      "acc": 0.68559203,
+      "epoch": 0.38039066463724,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.472203276203667e-06,
+      "loss": 1.40363331,
+      "memory(GiB)": 103.58,
+      "step": 14995,
+      "train_speed(iter/s)": 1.322423
+    },
+    {
+      "acc": 0.68683825,
+      "epoch": 0.380517503805175,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.471734247326284e-06,
+      "loss": 1.42466831,
+      "memory(GiB)": 103.58,
+      "step": 15000,
+      "train_speed(iter/s)": 1.322473
+    },
+    {
+      "epoch": 0.380517503805175,
+      "eval_acc": 0.6720411312771871,
+      "eval_loss": 1.3731284141540527,
+      "eval_runtime": 69.8422,
+      "eval_samples_per_second": 91.206,
+      "eval_steps_per_second": 22.809,
+      "step": 15000
+    },
+    {
+      "acc": 0.68716784,
+      "epoch": 0.3806443429731101,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.471265021762067e-06,
+      "loss": 1.3941761,
+      "memory(GiB)": 103.58,
+      "step": 15005,
+      "train_speed(iter/s)": 1.312977
+    },
+    {
+      "acc": 0.71046457,
+      "epoch": 0.38077118214104516,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.47079559953165e-06,
+      "loss": 1.28611526,
+      "memory(GiB)": 103.58,
+      "step": 15010,
+      "train_speed(iter/s)": 1.313025
+    },
+    {
+      "acc": 0.6740325,
+      "epoch": 0.3808980213089802,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.470325980655683e-06,
+      "loss": 1.43694792,
+      "memory(GiB)": 103.58,
+      "step": 15015,
+      "train_speed(iter/s)": 1.313075
+    },
+    {
+      "acc": 0.68174877,
+      "epoch": 0.38102486047691525,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.46985616515482e-06,
+      "loss": 1.46784534,
+      "memory(GiB)": 103.58,
+      "step": 15020,
+      "train_speed(iter/s)": 1.313126
+    },
+    {
+      "acc": 0.68136859,
+      "epoch": 0.38115169964485035,
+      "grad_norm": 3.25,
+      "learning_rate": 9.469386153049727e-06,
+      "loss": 1.40315914,
+      "memory(GiB)": 103.58,
+      "step": 15025,
+      "train_speed(iter/s)": 1.31317
+    },
+    {
+      "acc": 0.68894238,
+      "epoch": 0.3812785388127854,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.468915944361076e-06,
+      "loss": 1.33626556,
+      "memory(GiB)": 103.58,
+      "step": 15030,
+      "train_speed(iter/s)": 1.313219
+    },
+    {
+      "acc": 0.69009533,
+      "epoch": 0.38140537798072044,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.468445539109551e-06,
+      "loss": 1.4185688,
+      "memory(GiB)": 103.58,
+      "step": 15035,
+      "train_speed(iter/s)": 1.313265
+    },
+    {
+      "acc": 0.67382975,
+      "epoch": 0.3815322171486555,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.46797493731584e-06,
+      "loss": 1.47820091,
+      "memory(GiB)": 103.58,
+      "step": 15040,
+      "train_speed(iter/s)": 1.313314
+    },
+    {
+      "acc": 0.6910224,
+      "epoch": 0.3816590563165906,
+      "grad_norm": 2.75,
+      "learning_rate": 9.467504139000642e-06,
+      "loss": 1.40142078,
+      "memory(GiB)": 103.58,
+      "step": 15045,
+      "train_speed(iter/s)": 1.31336
+    },
+    {
+      "acc": 0.68588328,
+      "epoch": 0.38178589548452563,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.467033144184667e-06,
+      "loss": 1.44824114,
+      "memory(GiB)": 103.58,
+      "step": 15050,
+      "train_speed(iter/s)": 1.313406
+    },
+    {
+      "acc": 0.68804159,
+      "epoch": 0.3819127346524607,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.466561952888632e-06,
+      "loss": 1.42002497,
+      "memory(GiB)": 103.58,
+      "step": 15055,
+      "train_speed(iter/s)": 1.313453
+    },
+    {
+      "acc": 0.6828948,
+      "epoch": 0.3820395738203957,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.466090565133259e-06,
+      "loss": 1.39300747,
+      "memory(GiB)": 103.58,
+      "step": 15060,
+      "train_speed(iter/s)": 1.3135
+    },
+    {
+      "acc": 0.68614759,
+      "epoch": 0.3821664129883308,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.465618980939284e-06,
+      "loss": 1.40370035,
+      "memory(GiB)": 103.58,
+      "step": 15065,
+      "train_speed(iter/s)": 1.31355
+    },
+    {
+      "acc": 0.69667387,
+      "epoch": 0.38229325215626586,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.465147200327446e-06,
+      "loss": 1.43825312,
+      "memory(GiB)": 103.58,
+      "step": 15070,
+      "train_speed(iter/s)": 1.313596
+    },
+    {
+      "acc": 0.6857316,
+      "epoch": 0.3824200913242009,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.464675223318503e-06,
+      "loss": 1.33452625,
+      "memory(GiB)": 103.58,
+      "step": 15075,
+      "train_speed(iter/s)": 1.313647
+    },
+    {
+      "acc": 0.69874868,
+      "epoch": 0.38254693049213595,
+      "grad_norm": 3.5,
+      "learning_rate": 9.464203049933207e-06,
+      "loss": 1.38652792,
+      "memory(GiB)": 103.58,
+      "step": 15080,
+      "train_speed(iter/s)": 1.313695
+    },
+    {
+      "acc": 0.71257696,
+      "epoch": 0.38267376966007105,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.463730680192332e-06,
+      "loss": 1.35586996,
+      "memory(GiB)": 103.58,
+      "step": 15085,
+      "train_speed(iter/s)": 1.313744
+    },
+    {
+      "acc": 0.68463821,
+      "epoch": 0.3828006088280061,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.46325811411665e-06,
+      "loss": 1.45762634,
+      "memory(GiB)": 103.58,
+      "step": 15090,
+      "train_speed(iter/s)": 1.313791
+    },
+    {
+      "acc": 0.68533092,
+      "epoch": 0.38292744799594114,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.462785351726951e-06,
+      "loss": 1.39121485,
+      "memory(GiB)": 103.58,
+      "step": 15095,
+      "train_speed(iter/s)": 1.313839
+    },
+    {
+      "acc": 0.68197536,
+      "epoch": 0.3830542871638762,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.462312393044027e-06,
+      "loss": 1.38551579,
+      "memory(GiB)": 103.58,
+      "step": 15100,
+      "train_speed(iter/s)": 1.313886
+    },
+    {
+      "acc": 0.70100479,
+      "epoch": 0.3831811263318113,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.46183923808868e-06,
+      "loss": 1.35820675,
+      "memory(GiB)": 103.58,
+      "step": 15105,
+      "train_speed(iter/s)": 1.313932
+    },
+    {
+      "acc": 0.69006271,
+      "epoch": 0.38330796549974633,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.461365886881724e-06,
+      "loss": 1.41478138,
+      "memory(GiB)": 103.58,
+      "step": 15110,
+      "train_speed(iter/s)": 1.313977
+    },
+    {
+      "acc": 0.68807101,
+      "epoch": 0.3834348046676814,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.460892339443977e-06,
+      "loss": 1.45500393,
+      "memory(GiB)": 103.58,
+      "step": 15115,
+      "train_speed(iter/s)": 1.314025
+    },
+    {
+      "acc": 0.68336864,
+      "epoch": 0.3835616438356164,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.460418595796268e-06,
+      "loss": 1.39254999,
+      "memory(GiB)": 103.58,
+      "step": 15120,
+      "train_speed(iter/s)": 1.314076
+    },
+    {
+      "acc": 0.6728301,
+      "epoch": 0.3836884830035515,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.459944655959437e-06,
+      "loss": 1.45025024,
+      "memory(GiB)": 103.58,
+      "step": 15125,
+      "train_speed(iter/s)": 1.314125
+    },
+    {
+      "acc": 0.69763913,
+      "epoch": 0.38381532217148656,
+      "grad_norm": 3.0,
+      "learning_rate": 9.459470519954325e-06,
+      "loss": 1.32655697,
+      "memory(GiB)": 103.58,
+      "step": 15130,
+      "train_speed(iter/s)": 1.314176
+    },
+    {
+      "acc": 0.68865037,
+      "epoch": 0.3839421613394216,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.458996187801791e-06,
+      "loss": 1.37124281,
+      "memory(GiB)": 103.58,
+      "step": 15135,
+      "train_speed(iter/s)": 1.314226
+    },
+    {
+      "acc": 0.67448802,
+      "epoch": 0.38406900050735665,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.458521659522697e-06,
+      "loss": 1.45752096,
+      "memory(GiB)": 103.58,
+      "step": 15140,
+      "train_speed(iter/s)": 1.314275
+    },
+    {
+      "acc": 0.66887112,
+      "epoch": 0.38419583967529175,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.458046935137913e-06,
+      "loss": 1.46494865,
+      "memory(GiB)": 103.58,
+      "step": 15145,
+      "train_speed(iter/s)": 1.314326
+    },
+    {
+      "acc": 0.69257259,
+      "epoch": 0.3843226788432268,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.457572014668323e-06,
+      "loss": 1.41070023,
+      "memory(GiB)": 103.58,
+      "step": 15150,
+      "train_speed(iter/s)": 1.314376
+    },
+    {
+      "acc": 0.68784103,
+      "epoch": 0.38444951801116184,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.457096898134813e-06,
+      "loss": 1.39834795,
+      "memory(GiB)": 103.58,
+      "step": 15155,
+      "train_speed(iter/s)": 1.314427
+    },
+    {
+      "acc": 0.69765453,
+      "epoch": 0.3845763571790969,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.45662158555828e-06,
+      "loss": 1.41355667,
+      "memory(GiB)": 103.58,
+      "step": 15160,
+      "train_speed(iter/s)": 1.314475
+    },
+    {
+      "acc": 0.68090816,
+      "epoch": 0.384703196347032,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.456146076959636e-06,
+      "loss": 1.3973526,
+      "memory(GiB)": 103.58,
+      "step": 15165,
+      "train_speed(iter/s)": 1.314527
+    },
+    {
+      "acc": 0.68337212,
+      "epoch": 0.38483003551496703,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.455670372359791e-06,
+      "loss": 1.45276623,
+      "memory(GiB)": 103.58,
+      "step": 15170,
+      "train_speed(iter/s)": 1.314577
+    },
+    {
+      "acc": 0.67716599,
+      "epoch": 0.3849568746829021,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.45519447177967e-06,
+      "loss": 1.40309095,
+      "memory(GiB)": 103.58,
+      "step": 15175,
+      "train_speed(iter/s)": 1.314627
+    },
+    {
+      "acc": 0.67481151,
+      "epoch": 0.3850837138508371,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.454718375240204e-06,
+      "loss": 1.47329369,
+      "memory(GiB)": 103.58,
+      "step": 15180,
+      "train_speed(iter/s)": 1.314659
+    },
+    {
+      "acc": 0.69370222,
+      "epoch": 0.3852105530187722,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.454242082762336e-06,
+      "loss": 1.37173719,
+      "memory(GiB)": 103.58,
+      "step": 15185,
+      "train_speed(iter/s)": 1.314709
+    },
+    {
+      "acc": 0.68917646,
+      "epoch": 0.38533739218670726,
+      "grad_norm": 3.75,
+      "learning_rate": 9.453765594367014e-06,
+      "loss": 1.41967316,
+      "memory(GiB)": 103.58,
+      "step": 15190,
+      "train_speed(iter/s)": 1.31476
+    },
+    {
+      "acc": 0.68396378,
+      "epoch": 0.3854642313546423,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.453288910075196e-06,
+      "loss": 1.38410406,
+      "memory(GiB)": 103.58,
+      "step": 15195,
+      "train_speed(iter/s)": 1.314812
+    },
+    {
+      "acc": 0.68221655,
+      "epoch": 0.38559107052257735,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.452812029907849e-06,
+      "loss": 1.44570026,
+      "memory(GiB)": 103.58,
+      "step": 15200,
+      "train_speed(iter/s)": 1.314861
+    },
+    {
+      "acc": 0.6841454,
+      "epoch": 0.38571790969051245,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.452334953885951e-06,
+      "loss": 1.41351032,
+      "memory(GiB)": 103.58,
+      "step": 15205,
+      "train_speed(iter/s)": 1.31491
+    },
+    {
+      "acc": 0.6835197,
+      "epoch": 0.3858447488584475,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.451857682030481e-06,
+      "loss": 1.44593639,
+      "memory(GiB)": 103.58,
+      "step": 15210,
+      "train_speed(iter/s)": 1.314959
+    },
+    {
+      "acc": 0.69478931,
+      "epoch": 0.38597158802638254,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.451380214362436e-06,
+      "loss": 1.42971182,
+      "memory(GiB)": 103.58,
+      "step": 15215,
+      "train_speed(iter/s)": 1.315008
+    },
+    {
+      "acc": 0.69395695,
+      "epoch": 0.3860984271943176,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.450902550902814e-06,
+      "loss": 1.39380655,
+      "memory(GiB)": 103.58,
+      "step": 15220,
+      "train_speed(iter/s)": 1.315057
+    },
+    {
+      "acc": 0.66965294,
+      "epoch": 0.3862252663622527,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.450424691672626e-06,
+      "loss": 1.49140844,
+      "memory(GiB)": 103.58,
+      "step": 15225,
+      "train_speed(iter/s)": 1.315105
+    },
+    {
+      "acc": 0.68685708,
+      "epoch": 0.38635210553018773,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.449946636692891e-06,
+      "loss": 1.3886364,
+      "memory(GiB)": 103.58,
+      "step": 15230,
+      "train_speed(iter/s)": 1.31515
+    },
+    {
+      "acc": 0.69151707,
+      "epoch": 0.3864789446981228,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.449468385984634e-06,
+      "loss": 1.36700134,
+      "memory(GiB)": 103.58,
+      "step": 15235,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.68954964,
+      "epoch": 0.3866057838660578,
+      "grad_norm": 3.5,
+      "learning_rate": 9.448989939568892e-06,
+      "loss": 1.38747158,
+      "memory(GiB)": 103.58,
+      "step": 15240,
+      "train_speed(iter/s)": 1.315254
+    },
+    {
+      "acc": 0.66753635,
+      "epoch": 0.3867326230339929,
+      "grad_norm": 2.875,
+      "learning_rate": 9.448511297466708e-06,
+      "loss": 1.46824856,
+      "memory(GiB)": 103.58,
+      "step": 15245,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.68463774,
+      "epoch": 0.38685946220192796,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.448032459699139e-06,
+      "loss": 1.37498093,
+      "memory(GiB)": 103.58,
+      "step": 15250,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.66724281,
+      "epoch": 0.386986301369863,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.447553426287244e-06,
+      "loss": 1.48571234,
+      "memory(GiB)": 103.58,
+      "step": 15255,
+      "train_speed(iter/s)": 1.315396
+    },
+    {
+      "acc": 0.68066835,
+      "epoch": 0.38711314053779805,
+      "grad_norm": 3.0,
+      "learning_rate": 9.44707419725209e-06,
+      "loss": 1.39218311,
+      "memory(GiB)": 103.58,
+      "step": 15260,
+      "train_speed(iter/s)": 1.315445
+    },
+    {
+      "acc": 0.67489891,
+      "epoch": 0.38723997970573315,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.446594772614759e-06,
+      "loss": 1.47339668,
+      "memory(GiB)": 103.58,
+      "step": 15265,
+      "train_speed(iter/s)": 1.315495
+    },
+    {
+      "acc": 0.68209581,
+      "epoch": 0.3873668188736682,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.446115152396335e-06,
+      "loss": 1.38658218,
+      "memory(GiB)": 103.58,
+      "step": 15270,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.70076213,
+      "epoch": 0.38749365804160324,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.445635336617919e-06,
+      "loss": 1.33027639,
+      "memory(GiB)": 103.58,
+      "step": 15275,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.67877612,
+      "epoch": 0.3876204972095383,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.445155325300612e-06,
+      "loss": 1.44295139,
+      "memory(GiB)": 103.58,
+      "step": 15280,
+      "train_speed(iter/s)": 1.315634
+    },
+    {
+      "acc": 0.68854494,
+      "epoch": 0.3877473363774734,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.444675118465528e-06,
+      "loss": 1.43597441,
+      "memory(GiB)": 103.58,
+      "step": 15285,
+      "train_speed(iter/s)": 1.315682
+    },
+    {
+      "acc": 0.68338265,
+      "epoch": 0.38787417554540843,
+      "grad_norm": 2.875,
+      "learning_rate": 9.444194716133785e-06,
+      "loss": 1.45050879,
+      "memory(GiB)": 103.58,
+      "step": 15290,
+      "train_speed(iter/s)": 1.315728
+    },
+    {
+      "acc": 0.68869324,
+      "epoch": 0.3880010147133435,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.44371411832652e-06,
+      "loss": 1.3594388,
+      "memory(GiB)": 103.58,
+      "step": 15295,
+      "train_speed(iter/s)": 1.31578
+    },
+    {
+      "acc": 0.67388129,
+      "epoch": 0.3881278538812785,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.443233325064867e-06,
+      "loss": 1.47222576,
+      "memory(GiB)": 103.58,
+      "step": 15300,
+      "train_speed(iter/s)": 1.315829
+    },
+    {
+      "acc": 0.67820864,
+      "epoch": 0.3882546930492136,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.442752336369976e-06,
+      "loss": 1.41254454,
+      "memory(GiB)": 103.58,
+      "step": 15305,
+      "train_speed(iter/s)": 1.315877
+    },
+    {
+      "acc": 0.70574565,
+      "epoch": 0.38838153221714866,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.442271152263e-06,
+      "loss": 1.33811674,
+      "memory(GiB)": 103.58,
+      "step": 15310,
+      "train_speed(iter/s)": 1.315922
+    },
+    {
+      "acc": 0.68234181,
+      "epoch": 0.3885083713850837,
+      "grad_norm": 2.75,
+      "learning_rate": 9.441789772765107e-06,
+      "loss": 1.44191761,
+      "memory(GiB)": 103.58,
+      "step": 15315,
+      "train_speed(iter/s)": 1.31597
+    },
+    {
+      "acc": 0.6769949,
+      "epoch": 0.38863521055301875,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.441308197897467e-06,
+      "loss": 1.43654804,
+      "memory(GiB)": 103.58,
+      "step": 15320,
+      "train_speed(iter/s)": 1.316018
+    },
+    {
+      "acc": 0.69306479,
+      "epoch": 0.38876204972095385,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.440826427681264e-06,
+      "loss": 1.45662937,
+      "memory(GiB)": 103.58,
+      "step": 15325,
+      "train_speed(iter/s)": 1.316062
+    },
+    {
+      "acc": 0.68501244,
+      "epoch": 0.3888888888888889,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.44034446213769e-06,
+      "loss": 1.39728966,
+      "memory(GiB)": 103.58,
+      "step": 15330,
+      "train_speed(iter/s)": 1.316112
+    },
+    {
+      "acc": 0.68518801,
+      "epoch": 0.38901572805682394,
+      "grad_norm": 3.25,
+      "learning_rate": 9.439862301287939e-06,
+      "loss": 1.40405025,
+      "memory(GiB)": 103.58,
+      "step": 15335,
+      "train_speed(iter/s)": 1.316162
+    },
+    {
+      "acc": 0.67993417,
+      "epoch": 0.389142567224759,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.439379945153223e-06,
+      "loss": 1.46301222,
+      "memory(GiB)": 103.58,
+      "step": 15340,
+      "train_speed(iter/s)": 1.316211
+    },
+    {
+      "acc": 0.69318018,
+      "epoch": 0.3892694063926941,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.438897393754755e-06,
+      "loss": 1.35441837,
+      "memory(GiB)": 103.58,
+      "step": 15345,
+      "train_speed(iter/s)": 1.31626
+    },
+    {
+      "acc": 0.68813291,
+      "epoch": 0.38939624556062913,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.438414647113762e-06,
+      "loss": 1.38235931,
+      "memory(GiB)": 103.58,
+      "step": 15350,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.68198652,
+      "epoch": 0.3895230847285642,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.437931705251478e-06,
+      "loss": 1.41232872,
+      "memory(GiB)": 103.58,
+      "step": 15355,
+      "train_speed(iter/s)": 1.316358
+    },
+    {
+      "acc": 0.68667674,
+      "epoch": 0.3896499238964992,
+      "grad_norm": 2.75,
+      "learning_rate": 9.437448568189142e-06,
+      "loss": 1.33511467,
+      "memory(GiB)": 103.58,
+      "step": 15360,
+      "train_speed(iter/s)": 1.316406
+    },
+    {
+      "acc": 0.68526721,
+      "epoch": 0.3897767630644343,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.436965235948008e-06,
+      "loss": 1.37163229,
+      "memory(GiB)": 103.58,
+      "step": 15365,
+      "train_speed(iter/s)": 1.316456
+    },
+    {
+      "acc": 0.68233089,
+      "epoch": 0.38990360223236936,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.436481708549332e-06,
+      "loss": 1.42416124,
+      "memory(GiB)": 103.58,
+      "step": 15370,
+      "train_speed(iter/s)": 1.316504
+    },
+    {
+      "acc": 0.67412686,
+      "epoch": 0.3900304414003044,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.435997986014382e-06,
+      "loss": 1.47187214,
+      "memory(GiB)": 103.58,
+      "step": 15375,
+      "train_speed(iter/s)": 1.316555
+    },
+    {
+      "acc": 0.66916566,
+      "epoch": 0.39015728056823945,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.435514068364437e-06,
+      "loss": 1.51977348,
+      "memory(GiB)": 103.58,
+      "step": 15380,
+      "train_speed(iter/s)": 1.316603
+    },
+    {
+      "acc": 0.68331275,
+      "epoch": 0.39028411973617455,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.43502995562078e-06,
+      "loss": 1.37688847,
+      "memory(GiB)": 103.58,
+      "step": 15385,
+      "train_speed(iter/s)": 1.31665
+    },
+    {
+      "acc": 0.68388519,
+      "epoch": 0.3904109589041096,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.434545647804703e-06,
+      "loss": 1.44332132,
+      "memory(GiB)": 103.58,
+      "step": 15390,
+      "train_speed(iter/s)": 1.3167
+    },
+    {
+      "acc": 0.69626107,
+      "epoch": 0.39053779807204464,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.434061144937512e-06,
+      "loss": 1.37110529,
+      "memory(GiB)": 103.58,
+      "step": 15395,
+      "train_speed(iter/s)": 1.316749
+    },
+    {
+      "acc": 0.69703159,
+      "epoch": 0.3906646372399797,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.433576447040513e-06,
+      "loss": 1.34085579,
+      "memory(GiB)": 103.58,
+      "step": 15400,
+      "train_speed(iter/s)": 1.316797
+    },
+    {
+      "acc": 0.69190102,
+      "epoch": 0.3907914764079148,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.433091554135029e-06,
+      "loss": 1.44712925,
+      "memory(GiB)": 103.58,
+      "step": 15405,
+      "train_speed(iter/s)": 1.316847
+    },
+    {
+      "acc": 0.67945051,
+      "epoch": 0.39091831557584983,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.432606466242384e-06,
+      "loss": 1.4680337,
+      "memory(GiB)": 103.58,
+      "step": 15410,
+      "train_speed(iter/s)": 1.316897
+    },
+    {
+      "acc": 0.68016739,
+      "epoch": 0.3910451547437849,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.43212118338392e-06,
+      "loss": 1.37939777,
+      "memory(GiB)": 103.58,
+      "step": 15415,
+      "train_speed(iter/s)": 1.316945
+    },
+    {
+      "acc": 0.68172431,
+      "epoch": 0.3911719939117199,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.431635705580975e-06,
+      "loss": 1.4117815,
+      "memory(GiB)": 103.58,
+      "step": 15420,
+      "train_speed(iter/s)": 1.316994
+    },
+    {
+      "acc": 0.68004341,
+      "epoch": 0.391298833079655,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.431150032854907e-06,
+      "loss": 1.40928135,
+      "memory(GiB)": 103.58,
+      "step": 15425,
+      "train_speed(iter/s)": 1.317045
+    },
+    {
+      "acc": 0.68739157,
+      "epoch": 0.39142567224759006,
+      "grad_norm": 3.5,
+      "learning_rate": 9.430664165227077e-06,
+      "loss": 1.35144949,
+      "memory(GiB)": 103.58,
+      "step": 15430,
+      "train_speed(iter/s)": 1.317092
+    },
+    {
+      "acc": 0.68345957,
+      "epoch": 0.3915525114155251,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.430178102718857e-06,
+      "loss": 1.44034672,
+      "memory(GiB)": 103.58,
+      "step": 15435,
+      "train_speed(iter/s)": 1.31714
+    },
+    {
+      "acc": 0.68994164,
+      "epoch": 0.39167935058346015,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.429691845351623e-06,
+      "loss": 1.35984774,
+      "memory(GiB)": 103.58,
+      "step": 15440,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.68942499,
+      "epoch": 0.39180618975139525,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.429205393146763e-06,
+      "loss": 1.33078003,
+      "memory(GiB)": 103.58,
+      "step": 15445,
+      "train_speed(iter/s)": 1.317238
+    },
+    {
+      "acc": 0.6779582,
+      "epoch": 0.3919330289193303,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.428718746125678e-06,
+      "loss": 1.45603037,
+      "memory(GiB)": 103.58,
+      "step": 15450,
+      "train_speed(iter/s)": 1.317286
+    },
+    {
+      "acc": 0.68715343,
+      "epoch": 0.39205986808726534,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.428231904309768e-06,
+      "loss": 1.46157122,
+      "memory(GiB)": 103.58,
+      "step": 15455,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.6719635,
+      "epoch": 0.3921867072552004,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.427744867720448e-06,
+      "loss": 1.45850601,
+      "memory(GiB)": 103.58,
+      "step": 15460,
+      "train_speed(iter/s)": 1.317382
+    },
+    {
+      "acc": 0.67855101,
+      "epoch": 0.3923135464231355,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.42725763637914e-06,
+      "loss": 1.41056833,
+      "memory(GiB)": 103.58,
+      "step": 15465,
+      "train_speed(iter/s)": 1.31743
+    },
+    {
+      "acc": 0.69903545,
+      "epoch": 0.39244038559107053,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.426770210307277e-06,
+      "loss": 1.35919094,
+      "memory(GiB)": 103.58,
+      "step": 15470,
+      "train_speed(iter/s)": 1.31747
+    },
+    {
+      "acc": 0.68017874,
+      "epoch": 0.3925672247590056,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.426282589526294e-06,
+      "loss": 1.42327042,
+      "memory(GiB)": 103.58,
+      "step": 15475,
+      "train_speed(iter/s)": 1.31752
+    },
+    {
+      "acc": 0.69410396,
+      "epoch": 0.3926940639269406,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.425794774057641e-06,
+      "loss": 1.38150568,
+      "memory(GiB)": 103.58,
+      "step": 15480,
+      "train_speed(iter/s)": 1.317567
+    },
+    {
+      "acc": 0.68584094,
+      "epoch": 0.3928209030948757,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.425306763922775e-06,
+      "loss": 1.43161306,
+      "memory(GiB)": 103.58,
+      "step": 15485,
+      "train_speed(iter/s)": 1.317616
+    },
+    {
+      "acc": 0.66425962,
+      "epoch": 0.39294774226281076,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.42481855914316e-06,
+      "loss": 1.45954323,
+      "memory(GiB)": 103.58,
+      "step": 15490,
+      "train_speed(iter/s)": 1.317665
+    },
+    {
+      "acc": 0.66625252,
+      "epoch": 0.3930745814307458,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.424330159740269e-06,
+      "loss": 1.43139305,
+      "memory(GiB)": 103.58,
+      "step": 15495,
+      "train_speed(iter/s)": 1.317714
+    },
+    {
+      "acc": 0.693472,
+      "epoch": 0.39320142059868085,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.423841565735582e-06,
+      "loss": 1.30642471,
+      "memory(GiB)": 103.58,
+      "step": 15500,
+      "train_speed(iter/s)": 1.317762
+    },
+    {
+      "acc": 0.68720818,
+      "epoch": 0.39332825976661595,
+      "grad_norm": 3.125,
+      "learning_rate": 9.423352777150597e-06,
+      "loss": 1.45912666,
+      "memory(GiB)": 103.58,
+      "step": 15505,
+      "train_speed(iter/s)": 1.317809
+    },
+    {
+      "acc": 0.69633055,
+      "epoch": 0.393455098934551,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.422863794006804e-06,
+      "loss": 1.3834362,
+      "memory(GiB)": 103.58,
+      "step": 15510,
+      "train_speed(iter/s)": 1.31786
+    },
+    {
+      "acc": 0.67598319,
+      "epoch": 0.39358193810248604,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.422374616325716e-06,
+      "loss": 1.42999868,
+      "memory(GiB)": 103.58,
+      "step": 15515,
+      "train_speed(iter/s)": 1.317909
+    },
+    {
+      "acc": 0.68136864,
+      "epoch": 0.3937087772704211,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.421885244128847e-06,
+      "loss": 1.42059078,
+      "memory(GiB)": 103.58,
+      "step": 15520,
+      "train_speed(iter/s)": 1.317956
+    },
+    {
+      "acc": 0.68917828,
+      "epoch": 0.3938356164383562,
+      "grad_norm": 3.125,
+      "learning_rate": 9.421395677437724e-06,
+      "loss": 1.38203297,
+      "memory(GiB)": 103.58,
+      "step": 15525,
+      "train_speed(iter/s)": 1.318005
+    },
+    {
+      "acc": 0.69678164,
+      "epoch": 0.39396245560629123,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.42090591627388e-06,
+      "loss": 1.37644367,
+      "memory(GiB)": 103.58,
+      "step": 15530,
+      "train_speed(iter/s)": 1.318051
+    },
+    {
+      "acc": 0.68036404,
+      "epoch": 0.3940892947742263,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.420415960658853e-06,
+      "loss": 1.38143368,
+      "memory(GiB)": 103.58,
+      "step": 15535,
+      "train_speed(iter/s)": 1.318098
+    },
+    {
+      "acc": 0.68398495,
+      "epoch": 0.3942161339421613,
+      "grad_norm": 3.0,
+      "learning_rate": 9.419925810614196e-06,
+      "loss": 1.41257076,
+      "memory(GiB)": 103.58,
+      "step": 15540,
+      "train_speed(iter/s)": 1.318144
+    },
+    {
+      "acc": 0.67691617,
+      "epoch": 0.3943429731100964,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.419435466161471e-06,
+      "loss": 1.43206997,
+      "memory(GiB)": 103.58,
+      "step": 15545,
+      "train_speed(iter/s)": 1.318193
+    },
+    {
+      "acc": 0.68003502,
+      "epoch": 0.39446981227803146,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.418944927322242e-06,
+      "loss": 1.47350674,
+      "memory(GiB)": 103.58,
+      "step": 15550,
+      "train_speed(iter/s)": 1.31824
+    },
+    {
+      "acc": 0.68953629,
+      "epoch": 0.3945966514459665,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.418454194118085e-06,
+      "loss": 1.37382002,
+      "memory(GiB)": 103.58,
+      "step": 15555,
+      "train_speed(iter/s)": 1.318287
+    },
+    {
+      "acc": 0.68792353,
+      "epoch": 0.39472349061390155,
+      "grad_norm": 3.375,
+      "learning_rate": 9.417963266570587e-06,
+      "loss": 1.44540358,
+      "memory(GiB)": 103.58,
+      "step": 15560,
+      "train_speed(iter/s)": 1.318334
+    },
+    {
+      "acc": 0.68184037,
+      "epoch": 0.39485032978183665,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.417472144701338e-06,
+      "loss": 1.39875488,
+      "memory(GiB)": 103.58,
+      "step": 15565,
+      "train_speed(iter/s)": 1.31838
+    },
+    {
+      "acc": 0.69271345,
+      "epoch": 0.3949771689497717,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.416980828531944e-06,
+      "loss": 1.35033264,
+      "memory(GiB)": 103.58,
+      "step": 15570,
+      "train_speed(iter/s)": 1.318429
+    },
+    {
+      "acc": 0.67280459,
+      "epoch": 0.39510400811770674,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.41648931808401e-06,
+      "loss": 1.43696871,
+      "memory(GiB)": 103.58,
+      "step": 15575,
+      "train_speed(iter/s)": 1.318478
+    },
+    {
+      "acc": 0.68361826,
+      "epoch": 0.3952308472856418,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.415997613379159e-06,
+      "loss": 1.39861622,
+      "memory(GiB)": 103.58,
+      "step": 15580,
+      "train_speed(iter/s)": 1.318525
+    },
+    {
+      "acc": 0.68731041,
+      "epoch": 0.3953576864535769,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.415505714439016e-06,
+      "loss": 1.38587914,
+      "memory(GiB)": 103.58,
+      "step": 15585,
+      "train_speed(iter/s)": 1.318573
+    },
+    {
+      "acc": 0.67763934,
+      "epoch": 0.39548452562151193,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.415013621285219e-06,
+      "loss": 1.4257535,
+      "memory(GiB)": 103.58,
+      "step": 15590,
+      "train_speed(iter/s)": 1.318614
+    },
+    {
+      "acc": 0.69000387,
+      "epoch": 0.395611364789447,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.41452133393941e-06,
+      "loss": 1.34788256,
+      "memory(GiB)": 103.58,
+      "step": 15595,
+      "train_speed(iter/s)": 1.318661
+    },
+    {
+      "acc": 0.69612651,
+      "epoch": 0.395738203957382,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.414028852423245e-06,
+      "loss": 1.36710529,
+      "memory(GiB)": 103.58,
+      "step": 15600,
+      "train_speed(iter/s)": 1.318708
+    },
+    {
+      "acc": 0.69501467,
+      "epoch": 0.3958650431253171,
+      "grad_norm": 4.125,
+      "learning_rate": 9.413536176758384e-06,
+      "loss": 1.34862947,
+      "memory(GiB)": 103.58,
+      "step": 15605,
+      "train_speed(iter/s)": 1.318756
+    },
+    {
+      "acc": 0.67476044,
+      "epoch": 0.39599188229325216,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.413043306966496e-06,
+      "loss": 1.42158146,
+      "memory(GiB)": 103.58,
+      "step": 15610,
+      "train_speed(iter/s)": 1.318803
+    },
+    {
+      "acc": 0.69905748,
+      "epoch": 0.3961187214611872,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.41255024306926e-06,
+      "loss": 1.35832195,
+      "memory(GiB)": 103.58,
+      "step": 15615,
+      "train_speed(iter/s)": 1.318846
+    },
+    {
+      "acc": 0.69702082,
+      "epoch": 0.39624556062912225,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.412056985088364e-06,
+      "loss": 1.41551094,
+      "memory(GiB)": 103.58,
+      "step": 15620,
+      "train_speed(iter/s)": 1.31889
+    },
+    {
+      "acc": 0.69056621,
+      "epoch": 0.39637239979705735,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.411563533045505e-06,
+      "loss": 1.41793327,
+      "memory(GiB)": 103.58,
+      "step": 15625,
+      "train_speed(iter/s)": 1.318936
+    },
+    {
+      "acc": 0.67623014,
+      "epoch": 0.3964992389649924,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.411069886962383e-06,
+      "loss": 1.46466007,
+      "memory(GiB)": 103.58,
+      "step": 15630,
+      "train_speed(iter/s)": 1.318982
+    },
+    {
+      "acc": 0.67195196,
+      "epoch": 0.39662607813292744,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.410576046860716e-06,
+      "loss": 1.44581642,
+      "memory(GiB)": 103.58,
+      "step": 15635,
+      "train_speed(iter/s)": 1.319029
+    },
+    {
+      "acc": 0.69618196,
+      "epoch": 0.3967529173008625,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.41008201276222e-06,
+      "loss": 1.35123653,
+      "memory(GiB)": 103.58,
+      "step": 15640,
+      "train_speed(iter/s)": 1.319076
+    },
+    {
+      "acc": 0.69348297,
+      "epoch": 0.3968797564687976,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.409587784688629e-06,
+      "loss": 1.42882156,
+      "memory(GiB)": 103.58,
+      "step": 15645,
+      "train_speed(iter/s)": 1.319121
+    },
+    {
+      "acc": 0.66302891,
+      "epoch": 0.39700659563673263,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.409093362661678e-06,
+      "loss": 1.49177017,
+      "memory(GiB)": 103.58,
+      "step": 15650,
+      "train_speed(iter/s)": 1.319167
+    },
+    {
+      "acc": 0.69445386,
+      "epoch": 0.3971334348046677,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.408598746703119e-06,
+      "loss": 1.37163296,
+      "memory(GiB)": 103.58,
+      "step": 15655,
+      "train_speed(iter/s)": 1.319213
+    },
+    {
+      "acc": 0.68512034,
+      "epoch": 0.3972602739726027,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.408103936834703e-06,
+      "loss": 1.43169365,
+      "memory(GiB)": 103.58,
+      "step": 15660,
+      "train_speed(iter/s)": 1.319259
+    },
+    {
+      "acc": 0.6824369,
+      "epoch": 0.3973871131405378,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.407608933078194e-06,
+      "loss": 1.4201004,
+      "memory(GiB)": 103.58,
+      "step": 15665,
+      "train_speed(iter/s)": 1.319305
+    },
+    {
+      "acc": 0.69020414,
+      "epoch": 0.39751395230847286,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.407113735455366e-06,
+      "loss": 1.39301691,
+      "memory(GiB)": 103.58,
+      "step": 15670,
+      "train_speed(iter/s)": 1.319349
+    },
+    {
+      "acc": 0.69362173,
+      "epoch": 0.3976407914764079,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.406618343988e-06,
+      "loss": 1.36174603,
+      "memory(GiB)": 103.58,
+      "step": 15675,
+      "train_speed(iter/s)": 1.319398
+    },
+    {
+      "acc": 0.68364668,
+      "epoch": 0.39776763064434295,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.406122758697885e-06,
+      "loss": 1.41324577,
+      "memory(GiB)": 103.58,
+      "step": 15680,
+      "train_speed(iter/s)": 1.319442
+    },
+    {
+      "acc": 0.68625598,
+      "epoch": 0.39789446981227805,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.405626979606819e-06,
+      "loss": 1.45233555,
+      "memory(GiB)": 103.58,
+      "step": 15685,
+      "train_speed(iter/s)": 1.319488
+    },
+    {
+      "acc": 0.68589525,
+      "epoch": 0.3980213089802131,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.405131006736608e-06,
+      "loss": 1.39389753,
+      "memory(GiB)": 103.58,
+      "step": 15690,
+      "train_speed(iter/s)": 1.319534
+    },
+    {
+      "acc": 0.69815302,
+      "epoch": 0.39814814814814814,
+      "grad_norm": 3.125,
+      "learning_rate": 9.404634840109069e-06,
+      "loss": 1.40300913,
+      "memory(GiB)": 103.58,
+      "step": 15695,
+      "train_speed(iter/s)": 1.319576
+    },
+    {
+      "acc": 0.67714453,
+      "epoch": 0.3982749873160832,
+      "grad_norm": 3.0,
+      "learning_rate": 9.404138479746022e-06,
+      "loss": 1.40200138,
+      "memory(GiB)": 103.58,
+      "step": 15700,
+      "train_speed(iter/s)": 1.319623
+    },
+    {
+      "acc": 0.68509498,
+      "epoch": 0.3984018264840183,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.403641925669304e-06,
+      "loss": 1.42601557,
+      "memory(GiB)": 103.58,
+      "step": 15705,
+      "train_speed(iter/s)": 1.319669
+    },
+    {
+      "acc": 0.68394842,
+      "epoch": 0.39852866565195333,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.403145177900752e-06,
+      "loss": 1.44966784,
+      "memory(GiB)": 103.58,
+      "step": 15710,
+      "train_speed(iter/s)": 1.319715
+    },
+    {
+      "acc": 0.68380466,
+      "epoch": 0.3986555048198884,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.402648236462217e-06,
+      "loss": 1.39413071,
+      "memory(GiB)": 103.58,
+      "step": 15715,
+      "train_speed(iter/s)": 1.319764
+    },
+    {
+      "acc": 0.68868494,
+      "epoch": 0.3987823439878234,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.402151101375557e-06,
+      "loss": 1.40544691,
+      "memory(GiB)": 103.58,
+      "step": 15720,
+      "train_speed(iter/s)": 1.319809
+    },
+    {
+      "acc": 0.67087555,
+      "epoch": 0.3989091831557585,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.401653772662638e-06,
+      "loss": 1.41930094,
+      "memory(GiB)": 103.58,
+      "step": 15725,
+      "train_speed(iter/s)": 1.319854
+    },
+    {
+      "acc": 0.67923255,
+      "epoch": 0.39903602232369356,
+      "grad_norm": 2.75,
+      "learning_rate": 9.401156250345331e-06,
+      "loss": 1.40847273,
+      "memory(GiB)": 103.58,
+      "step": 15730,
+      "train_speed(iter/s)": 1.319897
+    },
+    {
+      "acc": 0.68837032,
+      "epoch": 0.3991628614916286,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.400658534445524e-06,
+      "loss": 1.3780776,
+      "memory(GiB)": 103.58,
+      "step": 15735,
+      "train_speed(iter/s)": 1.319943
+    },
+    {
+      "acc": 0.6714623,
+      "epoch": 0.39928970065956365,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.40016062498511e-06,
+      "loss": 1.41208344,
+      "memory(GiB)": 103.58,
+      "step": 15740,
+      "train_speed(iter/s)": 1.31999
+    },
+    {
+      "acc": 0.68542476,
+      "epoch": 0.39941653982749875,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.399662521985982e-06,
+      "loss": 1.39328594,
+      "memory(GiB)": 103.58,
+      "step": 15745,
+      "train_speed(iter/s)": 1.320036
+    },
+    {
+      "acc": 0.68817387,
+      "epoch": 0.3995433789954338,
+      "grad_norm": 4.90625,
+      "learning_rate": 9.399164225470055e-06,
+      "loss": 1.4264926,
+      "memory(GiB)": 103.58,
+      "step": 15750,
+      "train_speed(iter/s)": 1.320081
+    },
+    {
+      "acc": 0.6863905,
+      "epoch": 0.39967021816336884,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.398665735459245e-06,
+      "loss": 1.42704,
+      "memory(GiB)": 103.58,
+      "step": 15755,
+      "train_speed(iter/s)": 1.320127
+    },
+    {
+      "acc": 0.66547184,
+      "epoch": 0.3997970573313039,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.398167051975475e-06,
+      "loss": 1.44853611,
+      "memory(GiB)": 103.58,
+      "step": 15760,
+      "train_speed(iter/s)": 1.320173
+    },
+    {
+      "acc": 0.68304739,
+      "epoch": 0.399923896499239,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.397668175040684e-06,
+      "loss": 1.38527184,
+      "memory(GiB)": 103.58,
+      "step": 15765,
+      "train_speed(iter/s)": 1.320214
+    },
+    {
+      "acc": 0.66888714,
+      "epoch": 0.40005073566717403,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.397169104676813e-06,
+      "loss": 1.44721889,
+      "memory(GiB)": 103.58,
+      "step": 15770,
+      "train_speed(iter/s)": 1.32026
+    },
+    {
+      "acc": 0.67461939,
+      "epoch": 0.4001775748351091,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.39666984090581e-06,
+      "loss": 1.45337944,
+      "memory(GiB)": 103.58,
+      "step": 15775,
+      "train_speed(iter/s)": 1.320304
+    },
+    {
+      "acc": 0.69226217,
+      "epoch": 0.4003044140030441,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.396170383749642e-06,
+      "loss": 1.40937338,
+      "memory(GiB)": 103.58,
+      "step": 15780,
+      "train_speed(iter/s)": 1.320345
+    },
+    {
+      "acc": 0.68398657,
+      "epoch": 0.4004312531709792,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.39567073323027e-06,
+      "loss": 1.39180737,
+      "memory(GiB)": 103.58,
+      "step": 15785,
+      "train_speed(iter/s)": 1.320392
+    },
+    {
+      "acc": 0.67507982,
+      "epoch": 0.40055809233891426,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.395170889369674e-06,
+      "loss": 1.49576616,
+      "memory(GiB)": 103.58,
+      "step": 15790,
+      "train_speed(iter/s)": 1.320439
+    },
+    {
+      "acc": 0.6876822,
+      "epoch": 0.4006849315068493,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.39467085218984e-06,
+      "loss": 1.4208065,
+      "memory(GiB)": 103.58,
+      "step": 15795,
+      "train_speed(iter/s)": 1.320474
+    },
+    {
+      "acc": 0.67547994,
+      "epoch": 0.40081177067478435,
+      "grad_norm": 2.75,
+      "learning_rate": 9.39417062171276e-06,
+      "loss": 1.41501589,
+      "memory(GiB)": 103.58,
+      "step": 15800,
+      "train_speed(iter/s)": 1.320521
+    },
+    {
+      "acc": 0.69046564,
+      "epoch": 0.40093860984271945,
+      "grad_norm": 3.125,
+      "learning_rate": 9.393670197960439e-06,
+      "loss": 1.40267563,
+      "memory(GiB)": 103.58,
+      "step": 15805,
+      "train_speed(iter/s)": 1.320565
+    },
+    {
+      "acc": 0.6802948,
+      "epoch": 0.4010654490106545,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.393169580954884e-06,
+      "loss": 1.37511282,
+      "memory(GiB)": 103.58,
+      "step": 15810,
+      "train_speed(iter/s)": 1.320611
+    },
+    {
+      "acc": 0.67560015,
+      "epoch": 0.40119228817858954,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.392668770718118e-06,
+      "loss": 1.46958294,
+      "memory(GiB)": 103.58,
+      "step": 15815,
+      "train_speed(iter/s)": 1.320655
+    },
+    {
+      "acc": 0.68857012,
+      "epoch": 0.4013191273465246,
+      "grad_norm": 3.125,
+      "learning_rate": 9.392167767272169e-06,
+      "loss": 1.42028713,
+      "memory(GiB)": 103.58,
+      "step": 15820,
+      "train_speed(iter/s)": 1.320702
+    },
+    {
+      "acc": 0.68161278,
+      "epoch": 0.4014459665144597,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.39166657063907e-06,
+      "loss": 1.43429394,
+      "memory(GiB)": 103.58,
+      "step": 15825,
+      "train_speed(iter/s)": 1.320747
+    },
+    {
+      "acc": 0.67454977,
+      "epoch": 0.40157280568239473,
+      "grad_norm": 2.625,
+      "learning_rate": 9.391165180840869e-06,
+      "loss": 1.40030422,
+      "memory(GiB)": 103.58,
+      "step": 15830,
+      "train_speed(iter/s)": 1.320792
+    },
+    {
+      "acc": 0.68023529,
+      "epoch": 0.4016996448503298,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.390663597899619e-06,
+      "loss": 1.44665537,
+      "memory(GiB)": 103.58,
+      "step": 15835,
+      "train_speed(iter/s)": 1.32084
+    },
+    {
+      "acc": 0.69543276,
+      "epoch": 0.4018264840182648,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.39016182183738e-06,
+      "loss": 1.33154001,
+      "memory(GiB)": 103.58,
+      "step": 15840,
+      "train_speed(iter/s)": 1.320887
+    },
+    {
+      "acc": 0.68428311,
+      "epoch": 0.4019533231861999,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.389659852676223e-06,
+      "loss": 1.39361925,
+      "memory(GiB)": 103.58,
+      "step": 15845,
+      "train_speed(iter/s)": 1.320934
+    },
+    {
+      "acc": 0.70813537,
+      "epoch": 0.40208016235413496,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.389157690438228e-06,
+      "loss": 1.30785675,
+      "memory(GiB)": 103.58,
+      "step": 15850,
+      "train_speed(iter/s)": 1.320978
+    },
+    {
+      "acc": 0.68908587,
+      "epoch": 0.40220700152207,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.38865533514548e-06,
+      "loss": 1.33483505,
+      "memory(GiB)": 103.58,
+      "step": 15855,
+      "train_speed(iter/s)": 1.321024
+    },
+    {
+      "acc": 0.69579315,
+      "epoch": 0.40233384069000505,
+      "grad_norm": 2.75,
+      "learning_rate": 9.388152786820078e-06,
+      "loss": 1.3284339,
+      "memory(GiB)": 103.58,
+      "step": 15860,
+      "train_speed(iter/s)": 1.321069
+    },
+    {
+      "acc": 0.66872425,
+      "epoch": 0.40246067985794015,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.387650045484124e-06,
+      "loss": 1.52663784,
+      "memory(GiB)": 103.58,
+      "step": 15865,
+      "train_speed(iter/s)": 1.321114
+    },
+    {
+      "acc": 0.68989382,
+      "epoch": 0.4025875190258752,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.387147111159734e-06,
+      "loss": 1.39600201,
+      "memory(GiB)": 103.58,
+      "step": 15870,
+      "train_speed(iter/s)": 1.32116
+    },
+    {
+      "acc": 0.68866725,
+      "epoch": 0.40271435819381024,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.386643983869025e-06,
+      "loss": 1.39983521,
+      "memory(GiB)": 103.58,
+      "step": 15875,
+      "train_speed(iter/s)": 1.321205
+    },
+    {
+      "acc": 0.68270731,
+      "epoch": 0.4028411973617453,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.386140663634127e-06,
+      "loss": 1.48297405,
+      "memory(GiB)": 103.58,
+      "step": 15880,
+      "train_speed(iter/s)": 1.321251
+    },
+    {
+      "acc": 0.6829185,
+      "epoch": 0.4029680365296804,
+      "grad_norm": 3.125,
+      "learning_rate": 9.385637150477182e-06,
+      "loss": 1.4304059,
+      "memory(GiB)": 103.58,
+      "step": 15885,
+      "train_speed(iter/s)": 1.321297
+    },
+    {
+      "acc": 0.67287178,
+      "epoch": 0.40309487569761543,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.385133444420333e-06,
+      "loss": 1.47430401,
+      "memory(GiB)": 103.58,
+      "step": 15890,
+      "train_speed(iter/s)": 1.321344
+    },
+    {
+      "acc": 0.70040703,
+      "epoch": 0.4032217148655505,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.384629545485738e-06,
+      "loss": 1.29858131,
+      "memory(GiB)": 103.58,
+      "step": 15895,
+      "train_speed(iter/s)": 1.321387
+    },
+    {
+      "acc": 0.67515674,
+      "epoch": 0.4033485540334855,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.38412545369556e-06,
+      "loss": 1.4200738,
+      "memory(GiB)": 103.58,
+      "step": 15900,
+      "train_speed(iter/s)": 1.321432
+    },
+    {
+      "acc": 0.6792439,
+      "epoch": 0.4034753932014206,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.383621169071971e-06,
+      "loss": 1.39748278,
+      "memory(GiB)": 103.58,
+      "step": 15905,
+      "train_speed(iter/s)": 1.321478
+    },
+    {
+      "acc": 0.69643416,
+      "epoch": 0.40360223236935566,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.38311669163715e-06,
+      "loss": 1.43205862,
+      "memory(GiB)": 103.58,
+      "step": 15910,
+      "train_speed(iter/s)": 1.321519
+    },
+    {
+      "acc": 0.68984847,
+      "epoch": 0.4037290715372907,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.38261202141329e-06,
+      "loss": 1.3882103,
+      "memory(GiB)": 103.58,
+      "step": 15915,
+      "train_speed(iter/s)": 1.321566
+    },
+    {
+      "acc": 0.68299799,
+      "epoch": 0.40385591070522575,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.382107158422585e-06,
+      "loss": 1.40434275,
+      "memory(GiB)": 103.58,
+      "step": 15920,
+      "train_speed(iter/s)": 1.32161
+    },
+    {
+      "acc": 0.68195314,
+      "epoch": 0.40398274987316085,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.381602102687241e-06,
+      "loss": 1.39227219,
+      "memory(GiB)": 103.58,
+      "step": 15925,
+      "train_speed(iter/s)": 1.321656
+    },
+    {
+      "acc": 0.69243422,
+      "epoch": 0.4041095890410959,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.381096854229476e-06,
+      "loss": 1.40912189,
+      "memory(GiB)": 103.58,
+      "step": 15930,
+      "train_speed(iter/s)": 1.321699
+    },
+    {
+      "acc": 0.69451246,
+      "epoch": 0.40423642820903094,
+      "grad_norm": 2.75,
+      "learning_rate": 9.38059141307151e-06,
+      "loss": 1.39503803,
+      "memory(GiB)": 103.58,
+      "step": 15935,
+      "train_speed(iter/s)": 1.321746
+    },
+    {
+      "acc": 0.68526111,
+      "epoch": 0.404363267376966,
+      "grad_norm": 2.875,
+      "learning_rate": 9.380085779235577e-06,
+      "loss": 1.38013725,
+      "memory(GiB)": 103.58,
+      "step": 15940,
+      "train_speed(iter/s)": 1.321787
+    },
+    {
+      "acc": 0.67836485,
+      "epoch": 0.4044901065449011,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.379579952743916e-06,
+      "loss": 1.43037367,
+      "memory(GiB)": 103.58,
+      "step": 15945,
+      "train_speed(iter/s)": 1.321832
+    },
+    {
+      "acc": 0.68463674,
+      "epoch": 0.40461694571283613,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.379073933618774e-06,
+      "loss": 1.39288597,
+      "memory(GiB)": 103.58,
+      "step": 15950,
+      "train_speed(iter/s)": 1.321878
+    },
+    {
+      "acc": 0.68056326,
+      "epoch": 0.4047437848807712,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.37856772188241e-06,
+      "loss": 1.4408514,
+      "memory(GiB)": 103.58,
+      "step": 15955,
+      "train_speed(iter/s)": 1.321924
+    },
+    {
+      "acc": 0.681075,
+      "epoch": 0.4048706240487062,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.378061317557088e-06,
+      "loss": 1.38370762,
+      "memory(GiB)": 103.58,
+      "step": 15960,
+      "train_speed(iter/s)": 1.321968
+    },
+    {
+      "acc": 0.70034552,
+      "epoch": 0.4049974632166413,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.377554720665083e-06,
+      "loss": 1.38945694,
+      "memory(GiB)": 103.58,
+      "step": 15965,
+      "train_speed(iter/s)": 1.322013
+    },
+    {
+      "acc": 0.68510098,
+      "epoch": 0.40512430238457636,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.377047931228677e-06,
+      "loss": 1.44678659,
+      "memory(GiB)": 103.58,
+      "step": 15970,
+      "train_speed(iter/s)": 1.322059
+    },
+    {
+      "acc": 0.69490232,
+      "epoch": 0.4052511415525114,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.376540949270161e-06,
+      "loss": 1.35986786,
+      "memory(GiB)": 103.58,
+      "step": 15975,
+      "train_speed(iter/s)": 1.322106
+    },
+    {
+      "acc": 0.68732967,
+      "epoch": 0.40537798072044645,
+      "grad_norm": 2.875,
+      "learning_rate": 9.376033774811833e-06,
+      "loss": 1.37381802,
+      "memory(GiB)": 103.58,
+      "step": 15980,
+      "train_speed(iter/s)": 1.322151
+    },
+    {
+      "acc": 0.70025687,
+      "epoch": 0.40550481988838155,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.375526407876003e-06,
+      "loss": 1.36879559,
+      "memory(GiB)": 103.58,
+      "step": 15985,
+      "train_speed(iter/s)": 1.322196
+    },
+    {
+      "acc": 0.67174029,
+      "epoch": 0.4056316590563166,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.375018848484987e-06,
+      "loss": 1.46149817,
+      "memory(GiB)": 103.58,
+      "step": 15990,
+      "train_speed(iter/s)": 1.322237
+    },
+    {
+      "acc": 0.69078121,
+      "epoch": 0.40575849822425164,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.374511096661108e-06,
+      "loss": 1.38914232,
+      "memory(GiB)": 103.58,
+      "step": 15995,
+      "train_speed(iter/s)": 1.322279
+    },
+    {
+      "acc": 0.68064003,
+      "epoch": 0.4058853373921867,
+      "grad_norm": 2.75,
+      "learning_rate": 9.374003152426701e-06,
+      "loss": 1.41427498,
+      "memory(GiB)": 103.58,
+      "step": 16000,
+      "train_speed(iter/s)": 1.322324
+    },
+    {
+      "epoch": 0.4058853373921867,
+      "eval_acc": 0.6724291385528456,
+      "eval_loss": 1.3723005056381226,
+      "eval_runtime": 69.408,
+      "eval_samples_per_second": 91.776,
+      "eval_steps_per_second": 22.951,
+      "step": 16000
+    },
+    {
+      "acc": 0.68010807,
+      "epoch": 0.4060121765601218,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.373495015804106e-06,
+      "loss": 1.40253944,
+      "memory(GiB)": 103.58,
+      "step": 16005,
+      "train_speed(iter/s)": 1.31348
+    },
+    {
+      "acc": 0.6786253,
+      "epoch": 0.40613901572805683,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.372986686815674e-06,
+      "loss": 1.43841343,
+      "memory(GiB)": 103.58,
+      "step": 16010,
+      "train_speed(iter/s)": 1.313526
+    },
+    {
+      "acc": 0.68307133,
+      "epoch": 0.4062658548959919,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.372478165483763e-06,
+      "loss": 1.37289772,
+      "memory(GiB)": 103.58,
+      "step": 16015,
+      "train_speed(iter/s)": 1.313573
+    },
+    {
+      "acc": 0.69100351,
+      "epoch": 0.4063926940639269,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.371969451830743e-06,
+      "loss": 1.42628632,
+      "memory(GiB)": 103.58,
+      "step": 16020,
+      "train_speed(iter/s)": 1.313617
+    },
+    {
+      "acc": 0.69275227,
+      "epoch": 0.406519533231862,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.371460545878986e-06,
+      "loss": 1.36847601,
+      "memory(GiB)": 103.58,
+      "step": 16025,
+      "train_speed(iter/s)": 1.313662
+    },
+    {
+      "acc": 0.68905678,
+      "epoch": 0.40664637239979706,
+      "grad_norm": 3.25,
+      "learning_rate": 9.370951447650875e-06,
+      "loss": 1.36208801,
+      "memory(GiB)": 103.58,
+      "step": 16030,
+      "train_speed(iter/s)": 1.313708
+    },
+    {
+      "acc": 0.66216836,
+      "epoch": 0.4067732115677321,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.370442157168806e-06,
+      "loss": 1.45925999,
+      "memory(GiB)": 103.58,
+      "step": 16035,
+      "train_speed(iter/s)": 1.313752
+    },
+    {
+      "acc": 0.67558002,
+      "epoch": 0.40690005073566715,
+      "grad_norm": 2.75,
+      "learning_rate": 9.369932674455177e-06,
+      "loss": 1.4234273,
+      "memory(GiB)": 103.58,
+      "step": 16040,
+      "train_speed(iter/s)": 1.313796
+    },
+    {
+      "acc": 0.69230809,
+      "epoch": 0.40702688990360225,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.3694229995324e-06,
+      "loss": 1.32735958,
+      "memory(GiB)": 103.58,
+      "step": 16045,
+      "train_speed(iter/s)": 1.313836
+    },
+    {
+      "acc": 0.69581614,
+      "epoch": 0.4071537290715373,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.368913132422891e-06,
+      "loss": 1.39417448,
+      "memory(GiB)": 103.58,
+      "step": 16050,
+      "train_speed(iter/s)": 1.31388
+    },
+    {
+      "acc": 0.68045201,
+      "epoch": 0.40728056823947234,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.368403073149079e-06,
+      "loss": 1.4233614,
+      "memory(GiB)": 103.58,
+      "step": 16055,
+      "train_speed(iter/s)": 1.313923
+    },
+    {
+      "acc": 0.67889395,
+      "epoch": 0.4074074074074074,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.367892821733393e-06,
+      "loss": 1.39257965,
+      "memory(GiB)": 103.58,
+      "step": 16060,
+      "train_speed(iter/s)": 1.313966
+    },
+    {
+      "acc": 0.69097681,
+      "epoch": 0.4075342465753425,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.367382378198282e-06,
+      "loss": 1.41413479,
+      "memory(GiB)": 103.58,
+      "step": 16065,
+      "train_speed(iter/s)": 1.314008
+    },
+    {
+      "acc": 0.68416748,
+      "epoch": 0.40766108574327753,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.366871742566193e-06,
+      "loss": 1.35897522,
+      "memory(GiB)": 103.58,
+      "step": 16070,
+      "train_speed(iter/s)": 1.314052
+    },
+    {
+      "acc": 0.6717926,
+      "epoch": 0.4077879249112126,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.36636091485959e-06,
+      "loss": 1.4251647,
+      "memory(GiB)": 103.58,
+      "step": 16075,
+      "train_speed(iter/s)": 1.314096
+    },
+    {
+      "acc": 0.68926029,
+      "epoch": 0.4079147640791476,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.365849895100939e-06,
+      "loss": 1.37492886,
+      "memory(GiB)": 103.58,
+      "step": 16080,
+      "train_speed(iter/s)": 1.314139
+    },
+    {
+      "acc": 0.69174476,
+      "epoch": 0.4080416032470827,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.365338683312714e-06,
+      "loss": 1.36665983,
+      "memory(GiB)": 103.58,
+      "step": 16085,
+      "train_speed(iter/s)": 1.314183
+    },
+    {
+      "acc": 0.68694429,
+      "epoch": 0.40816844241501776,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.364827279517408e-06,
+      "loss": 1.39739895,
+      "memory(GiB)": 103.58,
+      "step": 16090,
+      "train_speed(iter/s)": 1.314209
+    },
+    {
+      "acc": 0.67146888,
+      "epoch": 0.4082952815829528,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.36431568373751e-06,
+      "loss": 1.47544975,
+      "memory(GiB)": 103.58,
+      "step": 16095,
+      "train_speed(iter/s)": 1.314254
+    },
+    {
+      "acc": 0.6955833,
+      "epoch": 0.40842212075088785,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.363803895995522e-06,
+      "loss": 1.34990749,
+      "memory(GiB)": 103.58,
+      "step": 16100,
+      "train_speed(iter/s)": 1.314302
+    },
+    {
+      "acc": 0.68621716,
+      "epoch": 0.40854895991882295,
+      "grad_norm": 3.125,
+      "learning_rate": 9.363291916313955e-06,
+      "loss": 1.38727503,
+      "memory(GiB)": 103.58,
+      "step": 16105,
+      "train_speed(iter/s)": 1.314346
+    },
+    {
+      "acc": 0.68401928,
+      "epoch": 0.408675799086758,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.362779744715332e-06,
+      "loss": 1.40846682,
+      "memory(GiB)": 103.58,
+      "step": 16110,
+      "train_speed(iter/s)": 1.31439
+    },
+    {
+      "acc": 0.68287506,
+      "epoch": 0.40880263825469304,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.362267381222174e-06,
+      "loss": 1.41881104,
+      "memory(GiB)": 103.58,
+      "step": 16115,
+      "train_speed(iter/s)": 1.314439
+    },
+    {
+      "acc": 0.68800197,
+      "epoch": 0.4089294774226281,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.361754825857022e-06,
+      "loss": 1.39381008,
+      "memory(GiB)": 103.58,
+      "step": 16120,
+      "train_speed(iter/s)": 1.314481
+    },
+    {
+      "acc": 0.67645164,
+      "epoch": 0.4090563165905632,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.36124207864242e-06,
+      "loss": 1.42156134,
+      "memory(GiB)": 103.58,
+      "step": 16125,
+      "train_speed(iter/s)": 1.314525
+    },
+    {
+      "acc": 0.67101226,
+      "epoch": 0.40918315575849823,
+      "grad_norm": 3.25,
+      "learning_rate": 9.360729139600917e-06,
+      "loss": 1.4396719,
+      "memory(GiB)": 103.58,
+      "step": 16130,
+      "train_speed(iter/s)": 1.314561
+    },
+    {
+      "acc": 0.68714857,
+      "epoch": 0.4093099949264333,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.36021600875508e-06,
+      "loss": 1.3993001,
+      "memory(GiB)": 103.58,
+      "step": 16135,
+      "train_speed(iter/s)": 1.314605
+    },
+    {
+      "acc": 0.67382145,
+      "epoch": 0.4094368340943683,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.359702686127474e-06,
+      "loss": 1.43283205,
+      "memory(GiB)": 103.58,
+      "step": 16140,
+      "train_speed(iter/s)": 1.314649
+    },
+    {
+      "acc": 0.69365835,
+      "epoch": 0.4095636732623034,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.359189171740679e-06,
+      "loss": 1.43431396,
+      "memory(GiB)": 103.58,
+      "step": 16145,
+      "train_speed(iter/s)": 1.314695
+    },
+    {
+      "acc": 0.6771275,
+      "epoch": 0.40969051243023846,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.358675465617283e-06,
+      "loss": 1.43183794,
+      "memory(GiB)": 103.58,
+      "step": 16150,
+      "train_speed(iter/s)": 1.314741
+    },
+    {
+      "acc": 0.68121262,
+      "epoch": 0.4098173515981735,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.35816156777988e-06,
+      "loss": 1.43432274,
+      "memory(GiB)": 103.58,
+      "step": 16155,
+      "train_speed(iter/s)": 1.314787
+    },
+    {
+      "acc": 0.6685276,
+      "epoch": 0.40994419076610855,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.357647478251072e-06,
+      "loss": 1.41565151,
+      "memory(GiB)": 103.58,
+      "step": 16160,
+      "train_speed(iter/s)": 1.314832
+    },
+    {
+      "acc": 0.67774701,
+      "epoch": 0.41007102993404365,
+      "grad_norm": 3.5,
+      "learning_rate": 9.357133197053475e-06,
+      "loss": 1.46987648,
+      "memory(GiB)": 103.58,
+      "step": 16165,
+      "train_speed(iter/s)": 1.314878
+    },
+    {
+      "acc": 0.69285831,
+      "epoch": 0.4101978691019787,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.356618724209704e-06,
+      "loss": 1.40265446,
+      "memory(GiB)": 103.58,
+      "step": 16170,
+      "train_speed(iter/s)": 1.314923
+    },
+    {
+      "acc": 0.68689523,
+      "epoch": 0.41032470826991374,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.356104059742392e-06,
+      "loss": 1.36385345,
+      "memory(GiB)": 103.58,
+      "step": 16175,
+      "train_speed(iter/s)": 1.314966
+    },
+    {
+      "acc": 0.68107786,
+      "epoch": 0.4104515474378488,
+      "grad_norm": 3.0,
+      "learning_rate": 9.355589203674175e-06,
+      "loss": 1.43335018,
+      "memory(GiB)": 103.58,
+      "step": 16180,
+      "train_speed(iter/s)": 1.31501
+    },
+    {
+      "acc": 0.69561253,
+      "epoch": 0.4105783866057839,
+      "grad_norm": 3.75,
+      "learning_rate": 9.355074156027699e-06,
+      "loss": 1.36062574,
+      "memory(GiB)": 103.58,
+      "step": 16185,
+      "train_speed(iter/s)": 1.315056
+    },
+    {
+      "acc": 0.67108755,
+      "epoch": 0.41070522577371893,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.354558916825616e-06,
+      "loss": 1.45705519,
+      "memory(GiB)": 103.58,
+      "step": 16190,
+      "train_speed(iter/s)": 1.315102
+    },
+    {
+      "acc": 0.69391298,
+      "epoch": 0.410832064941654,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.354043486090592e-06,
+      "loss": 1.36750402,
+      "memory(GiB)": 103.58,
+      "step": 16195,
+      "train_speed(iter/s)": 1.315147
+    },
+    {
+      "acc": 0.67206907,
+      "epoch": 0.410958904109589,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.353527863845296e-06,
+      "loss": 1.43701077,
+      "memory(GiB)": 103.58,
+      "step": 16200,
+      "train_speed(iter/s)": 1.315194
+    },
+    {
+      "acc": 0.69664235,
+      "epoch": 0.4110857432775241,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.353012050112405e-06,
+      "loss": 1.3748786,
+      "memory(GiB)": 103.58,
+      "step": 16205,
+      "train_speed(iter/s)": 1.315238
+    },
+    {
+      "acc": 0.6794044,
+      "epoch": 0.41121258244545916,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.352496044914611e-06,
+      "loss": 1.40685005,
+      "memory(GiB)": 103.58,
+      "step": 16210,
+      "train_speed(iter/s)": 1.315284
+    },
+    {
+      "acc": 0.69127374,
+      "epoch": 0.4113394216133942,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.351979848274608e-06,
+      "loss": 1.38146896,
+      "memory(GiB)": 103.58,
+      "step": 16215,
+      "train_speed(iter/s)": 1.31533
+    },
+    {
+      "acc": 0.70420742,
+      "epoch": 0.41146626078132925,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.351463460215102e-06,
+      "loss": 1.27739115,
+      "memory(GiB)": 103.58,
+      "step": 16220,
+      "train_speed(iter/s)": 1.315375
+    },
+    {
+      "acc": 0.68960776,
+      "epoch": 0.41159309994926435,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.350946880758804e-06,
+      "loss": 1.35385551,
+      "memory(GiB)": 103.58,
+      "step": 16225,
+      "train_speed(iter/s)": 1.31542
+    },
+    {
+      "acc": 0.68594413,
+      "epoch": 0.4117199391171994,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.350430109928437e-06,
+      "loss": 1.39981289,
+      "memory(GiB)": 103.58,
+      "step": 16230,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.68926458,
+      "epoch": 0.41184677828513444,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.349913147746731e-06,
+      "loss": 1.45673552,
+      "memory(GiB)": 103.58,
+      "step": 16235,
+      "train_speed(iter/s)": 1.31551
+    },
+    {
+      "acc": 0.69099293,
+      "epoch": 0.4119736174530695,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.349395994236423e-06,
+      "loss": 1.34699945,
+      "memory(GiB)": 103.58,
+      "step": 16240,
+      "train_speed(iter/s)": 1.315554
+    },
+    {
+      "acc": 0.68673491,
+      "epoch": 0.4121004566210046,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.348878649420262e-06,
+      "loss": 1.46892357,
+      "memory(GiB)": 103.58,
+      "step": 16245,
+      "train_speed(iter/s)": 1.3156
+    },
+    {
+      "acc": 0.68101873,
+      "epoch": 0.41222729578893963,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.348361113321e-06,
+      "loss": 1.41525726,
+      "memory(GiB)": 103.58,
+      "step": 16250,
+      "train_speed(iter/s)": 1.315635
+    },
+    {
+      "acc": 0.69015775,
+      "epoch": 0.4123541349568747,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.347843385961403e-06,
+      "loss": 1.41498451,
+      "memory(GiB)": 103.58,
+      "step": 16255,
+      "train_speed(iter/s)": 1.31568
+    },
+    {
+      "acc": 0.67210741,
+      "epoch": 0.4124809741248097,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.347325467364242e-06,
+      "loss": 1.42884102,
+      "memory(GiB)": 103.58,
+      "step": 16260,
+      "train_speed(iter/s)": 1.315724
+    },
+    {
+      "acc": 0.69840312,
+      "epoch": 0.4126078132927448,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.346807357552296e-06,
+      "loss": 1.39283152,
+      "memory(GiB)": 103.58,
+      "step": 16265,
+      "train_speed(iter/s)": 1.315769
+    },
+    {
+      "acc": 0.6751399,
+      "epoch": 0.41273465246067986,
+      "grad_norm": 2.75,
+      "learning_rate": 9.346289056548357e-06,
+      "loss": 1.4408721,
+      "memory(GiB)": 103.58,
+      "step": 16270,
+      "train_speed(iter/s)": 1.315813
+    },
+    {
+      "acc": 0.69775653,
+      "epoch": 0.4128614916286149,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.345770564375221e-06,
+      "loss": 1.34579163,
+      "memory(GiB)": 103.58,
+      "step": 16275,
+      "train_speed(iter/s)": 1.315856
+    },
+    {
+      "acc": 0.67210503,
+      "epoch": 0.41298833079654995,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.345251881055692e-06,
+      "loss": 1.41237154,
+      "memory(GiB)": 103.58,
+      "step": 16280,
+      "train_speed(iter/s)": 1.315898
+    },
+    {
+      "acc": 0.68549032,
+      "epoch": 0.41311516996448505,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.344733006612585e-06,
+      "loss": 1.44196072,
+      "memory(GiB)": 103.58,
+      "step": 16285,
+      "train_speed(iter/s)": 1.315943
+    },
+    {
+      "acc": 0.69175682,
+      "epoch": 0.4132420091324201,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.344213941068724e-06,
+      "loss": 1.40356407,
+      "memory(GiB)": 103.58,
+      "step": 16290,
+      "train_speed(iter/s)": 1.315986
+    },
+    {
+      "acc": 0.67801571,
+      "epoch": 0.41336884830035514,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.343694684446937e-06,
+      "loss": 1.42512569,
+      "memory(GiB)": 103.58,
+      "step": 16295,
+      "train_speed(iter/s)": 1.316029
+    },
+    {
+      "acc": 0.6839396,
+      "epoch": 0.4134956874682902,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.343175236770065e-06,
+      "loss": 1.36129599,
+      "memory(GiB)": 103.58,
+      "step": 16300,
+      "train_speed(iter/s)": 1.316073
+    },
+    {
+      "acc": 0.68969526,
+      "epoch": 0.4136225266362253,
+      "grad_norm": 2.5,
+      "learning_rate": 9.342655598060955e-06,
+      "loss": 1.3755764,
+      "memory(GiB)": 103.58,
+      "step": 16305,
+      "train_speed(iter/s)": 1.316117
+    },
+    {
+      "acc": 0.68643723,
+      "epoch": 0.41374936580416033,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.342135768342464e-06,
+      "loss": 1.37599688,
+      "memory(GiB)": 103.58,
+      "step": 16310,
+      "train_speed(iter/s)": 1.31616
+    },
+    {
+      "acc": 0.69553652,
+      "epoch": 0.4138762049720954,
+      "grad_norm": 3.0,
+      "learning_rate": 9.341615747637454e-06,
+      "loss": 1.37399063,
+      "memory(GiB)": 103.58,
+      "step": 16315,
+      "train_speed(iter/s)": 1.316206
+    },
+    {
+      "acc": 0.67395215,
+      "epoch": 0.4140030441400304,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.3410955359688e-06,
+      "loss": 1.45377693,
+      "memory(GiB)": 103.58,
+      "step": 16320,
+      "train_speed(iter/s)": 1.316251
+    },
+    {
+      "acc": 0.69582334,
+      "epoch": 0.4141298833079655,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.340575133359385e-06,
+      "loss": 1.3815629,
+      "memory(GiB)": 103.58,
+      "step": 16325,
+      "train_speed(iter/s)": 1.316294
+    },
+    {
+      "acc": 0.68444777,
+      "epoch": 0.41425672247590056,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.340054539832095e-06,
+      "loss": 1.36861324,
+      "memory(GiB)": 103.58,
+      "step": 16330,
+      "train_speed(iter/s)": 1.316338
+    },
+    {
+      "acc": 0.67666593,
+      "epoch": 0.4143835616438356,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.339533755409828e-06,
+      "loss": 1.45684299,
+      "memory(GiB)": 103.58,
+      "step": 16335,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.67457047,
+      "epoch": 0.41451040081177065,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.339012780115492e-06,
+      "loss": 1.40252533,
+      "memory(GiB)": 103.58,
+      "step": 16340,
+      "train_speed(iter/s)": 1.316426
+    },
+    {
+      "acc": 0.68504744,
+      "epoch": 0.41463723997970575,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.338491613972002e-06,
+      "loss": 1.37405052,
+      "memory(GiB)": 103.58,
+      "step": 16345,
+      "train_speed(iter/s)": 1.316468
+    },
+    {
+      "acc": 0.70737686,
+      "epoch": 0.4147640791476408,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.337970257002282e-06,
+      "loss": 1.38829346,
+      "memory(GiB)": 103.58,
+      "step": 16350,
+      "train_speed(iter/s)": 1.316508
+    },
+    {
+      "acc": 0.68241973,
+      "epoch": 0.41489091831557584,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.337448709229261e-06,
+      "loss": 1.49754219,
+      "memory(GiB)": 103.58,
+      "step": 16355,
+      "train_speed(iter/s)": 1.316552
+    },
+    {
+      "acc": 0.69522219,
+      "epoch": 0.4150177574835109,
+      "grad_norm": 3.25,
+      "learning_rate": 9.336926970675883e-06,
+      "loss": 1.41889544,
+      "memory(GiB)": 103.58,
+      "step": 16360,
+      "train_speed(iter/s)": 1.316597
+    },
+    {
+      "acc": 0.66861277,
+      "epoch": 0.415144596651446,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.33640504136509e-06,
+      "loss": 1.47276363,
+      "memory(GiB)": 103.58,
+      "step": 16365,
+      "train_speed(iter/s)": 1.316641
+    },
+    {
+      "acc": 0.68386869,
+      "epoch": 0.41527143581938103,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.335882921319845e-06,
+      "loss": 1.3991004,
+      "memory(GiB)": 103.58,
+      "step": 16370,
+      "train_speed(iter/s)": 1.316686
+    },
+    {
+      "acc": 0.70090952,
+      "epoch": 0.4153982749873161,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.335360610563111e-06,
+      "loss": 1.42220001,
+      "memory(GiB)": 103.58,
+      "step": 16375,
+      "train_speed(iter/s)": 1.31673
+    },
+    {
+      "acc": 0.69175315,
+      "epoch": 0.4155251141552511,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.33483810911786e-06,
+      "loss": 1.402244,
+      "memory(GiB)": 103.58,
+      "step": 16380,
+      "train_speed(iter/s)": 1.316773
+    },
+    {
+      "acc": 0.66551161,
+      "epoch": 0.4156519533231862,
+      "grad_norm": 3.0,
+      "learning_rate": 9.334315417007079e-06,
+      "loss": 1.48846722,
+      "memory(GiB)": 103.58,
+      "step": 16385,
+      "train_speed(iter/s)": 1.316814
+    },
+    {
+      "acc": 0.67438288,
+      "epoch": 0.41577879249112126,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.333792534253751e-06,
+      "loss": 1.42908678,
+      "memory(GiB)": 103.58,
+      "step": 16390,
+      "train_speed(iter/s)": 1.316858
+    },
+    {
+      "acc": 0.67468452,
+      "epoch": 0.4159056316590563,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.333269460880879e-06,
+      "loss": 1.41952,
+      "memory(GiB)": 103.58,
+      "step": 16395,
+      "train_speed(iter/s)": 1.316903
+    },
+    {
+      "acc": 0.69329109,
+      "epoch": 0.41603247082699135,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.33274619691147e-06,
+      "loss": 1.34263983,
+      "memory(GiB)": 103.58,
+      "step": 16400,
+      "train_speed(iter/s)": 1.316946
+    },
+    {
+      "acc": 0.68277121,
+      "epoch": 0.41615930999492645,
+      "grad_norm": 3.25,
+      "learning_rate": 9.332222742368537e-06,
+      "loss": 1.37590799,
+      "memory(GiB)": 103.58,
+      "step": 16405,
+      "train_speed(iter/s)": 1.31699
+    },
+    {
+      "acc": 0.67837944,
+      "epoch": 0.4162861491628615,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.331699097275108e-06,
+      "loss": 1.42917328,
+      "memory(GiB)": 103.58,
+      "step": 16410,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.68529882,
+      "epoch": 0.41641298833079654,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.331175261654213e-06,
+      "loss": 1.40575867,
+      "memory(GiB)": 103.58,
+      "step": 16415,
+      "train_speed(iter/s)": 1.317078
+    },
+    {
+      "acc": 0.67889605,
+      "epoch": 0.4165398274987316,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.330651235528891e-06,
+      "loss": 1.42634983,
+      "memory(GiB)": 103.58,
+      "step": 16420,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.67380581,
+      "epoch": 0.4166666666666667,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.330127018922195e-06,
+      "loss": 1.45028553,
+      "memory(GiB)": 103.58,
+      "step": 16425,
+      "train_speed(iter/s)": 1.317164
+    },
+    {
+      "acc": 0.70309181,
+      "epoch": 0.41679350583460173,
+      "grad_norm": 2.875,
+      "learning_rate": 9.329602611857179e-06,
+      "loss": 1.30985804,
+      "memory(GiB)": 103.58,
+      "step": 16430,
+      "train_speed(iter/s)": 1.317208
+    },
+    {
+      "acc": 0.68163586,
+      "epoch": 0.4169203450025368,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.329078014356909e-06,
+      "loss": 1.37470398,
+      "memory(GiB)": 103.58,
+      "step": 16435,
+      "train_speed(iter/s)": 1.317252
+    },
+    {
+      "acc": 0.69833999,
+      "epoch": 0.4170471841704718,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.32855322644446e-06,
+      "loss": 1.31648026,
+      "memory(GiB)": 103.58,
+      "step": 16440,
+      "train_speed(iter/s)": 1.317295
+    },
+    {
+      "acc": 0.6879488,
+      "epoch": 0.4171740233384069,
+      "grad_norm": 3.125,
+      "learning_rate": 9.328028248142916e-06,
+      "loss": 1.34822302,
+      "memory(GiB)": 103.58,
+      "step": 16445,
+      "train_speed(iter/s)": 1.317339
+    },
+    {
+      "acc": 0.67208719,
+      "epoch": 0.41730086250634196,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.327503079475365e-06,
+      "loss": 1.4513176,
+      "memory(GiB)": 103.58,
+      "step": 16450,
+      "train_speed(iter/s)": 1.317384
+    },
+    {
+      "acc": 0.68564663,
+      "epoch": 0.417427701674277,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.326977720464908e-06,
+      "loss": 1.3746664,
+      "memory(GiB)": 103.58,
+      "step": 16455,
+      "train_speed(iter/s)": 1.31743
+    },
+    {
+      "acc": 0.69172068,
+      "epoch": 0.41755454084221205,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.326452171134652e-06,
+      "loss": 1.40662689,
+      "memory(GiB)": 103.58,
+      "step": 16460,
+      "train_speed(iter/s)": 1.317473
+    },
+    {
+      "acc": 0.67172117,
+      "epoch": 0.41768138001014715,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.325926431507714e-06,
+      "loss": 1.4080368,
+      "memory(GiB)": 103.58,
+      "step": 16465,
+      "train_speed(iter/s)": 1.317517
+    },
+    {
+      "acc": 0.68778315,
+      "epoch": 0.4178082191780822,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.325400501607218e-06,
+      "loss": 1.38357925,
+      "memory(GiB)": 103.58,
+      "step": 16470,
+      "train_speed(iter/s)": 1.317564
+    },
+    {
+      "acc": 0.67674737,
+      "epoch": 0.41793505834601724,
+      "grad_norm": 3.25,
+      "learning_rate": 9.324874381456295e-06,
+      "loss": 1.44242191,
+      "memory(GiB)": 103.58,
+      "step": 16475,
+      "train_speed(iter/s)": 1.31761
+    },
+    {
+      "acc": 0.68369942,
+      "epoch": 0.4180618975139523,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.324348071078088e-06,
+      "loss": 1.3914814,
+      "memory(GiB)": 103.58,
+      "step": 16480,
+      "train_speed(iter/s)": 1.317656
+    },
+    {
+      "acc": 0.66779346,
+      "epoch": 0.4181887366818874,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.323821570495748e-06,
+      "loss": 1.44414892,
+      "memory(GiB)": 103.58,
+      "step": 16485,
+      "train_speed(iter/s)": 1.317701
+    },
+    {
+      "acc": 0.67716656,
+      "epoch": 0.41831557584982243,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.32329487973243e-06,
+      "loss": 1.42911949,
+      "memory(GiB)": 103.58,
+      "step": 16490,
+      "train_speed(iter/s)": 1.317746
+    },
+    {
+      "acc": 0.69064913,
+      "epoch": 0.4184424150177575,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.3227679988113e-06,
+      "loss": 1.42438564,
+      "memory(GiB)": 103.58,
+      "step": 16495,
+      "train_speed(iter/s)": 1.317785
+    },
+    {
+      "acc": 0.69430151,
+      "epoch": 0.4185692541856925,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.322240927755534e-06,
+      "loss": 1.40087137,
+      "memory(GiB)": 103.58,
+      "step": 16500,
+      "train_speed(iter/s)": 1.317829
+    },
+    {
+      "acc": 0.68589463,
+      "epoch": 0.4186960933536276,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.321713666588314e-06,
+      "loss": 1.404076,
+      "memory(GiB)": 103.58,
+      "step": 16505,
+      "train_speed(iter/s)": 1.317875
+    },
+    {
+      "acc": 0.68646841,
+      "epoch": 0.41882293252156266,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.321186215332833e-06,
+      "loss": 1.45289726,
+      "memory(GiB)": 103.58,
+      "step": 16510,
+      "train_speed(iter/s)": 1.31792
+    },
+    {
+      "acc": 0.68682728,
+      "epoch": 0.4189497716894977,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.320658574012289e-06,
+      "loss": 1.41669464,
+      "memory(GiB)": 103.58,
+      "step": 16515,
+      "train_speed(iter/s)": 1.317966
+    },
+    {
+      "acc": 0.68675203,
+      "epoch": 0.41907661085743275,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.32013074264989e-06,
+      "loss": 1.38923168,
+      "memory(GiB)": 103.58,
+      "step": 16520,
+      "train_speed(iter/s)": 1.318009
+    },
+    {
+      "acc": 0.69197683,
+      "epoch": 0.41920345002536785,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.319602721268853e-06,
+      "loss": 1.37501392,
+      "memory(GiB)": 103.58,
+      "step": 16525,
+      "train_speed(iter/s)": 1.318054
+    },
+    {
+      "acc": 0.70885534,
+      "epoch": 0.4193302891933029,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.319074509892403e-06,
+      "loss": 1.3096344,
+      "memory(GiB)": 103.58,
+      "step": 16530,
+      "train_speed(iter/s)": 1.318099
+    },
+    {
+      "acc": 0.68588619,
+      "epoch": 0.41945712836123794,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.318546108543774e-06,
+      "loss": 1.34696331,
+      "memory(GiB)": 103.58,
+      "step": 16535,
+      "train_speed(iter/s)": 1.318143
+    },
+    {
+      "acc": 0.66015196,
+      "epoch": 0.419583967529173,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.318017517246205e-06,
+      "loss": 1.4700449,
+      "memory(GiB)": 103.58,
+      "step": 16540,
+      "train_speed(iter/s)": 1.318186
+    },
+    {
+      "acc": 0.66654339,
+      "epoch": 0.4197108066971081,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.317488736022948e-06,
+      "loss": 1.46991386,
+      "memory(GiB)": 103.58,
+      "step": 16545,
+      "train_speed(iter/s)": 1.31823
+    },
+    {
+      "acc": 0.67942286,
+      "epoch": 0.41983764586504313,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.316959764897259e-06,
+      "loss": 1.42567453,
+      "memory(GiB)": 103.58,
+      "step": 16550,
+      "train_speed(iter/s)": 1.318275
+    },
+    {
+      "acc": 0.67821975,
+      "epoch": 0.4199644850329782,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.316430603892406e-06,
+      "loss": 1.36963272,
+      "memory(GiB)": 103.58,
+      "step": 16555,
+      "train_speed(iter/s)": 1.318318
+    },
+    {
+      "acc": 0.67736812,
+      "epoch": 0.4200913242009132,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.315901253031663e-06,
+      "loss": 1.42078037,
+      "memory(GiB)": 103.58,
+      "step": 16560,
+      "train_speed(iter/s)": 1.318361
+    },
+    {
+      "acc": 0.67427931,
+      "epoch": 0.4202181633688483,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.315371712338315e-06,
+      "loss": 1.54262924,
+      "memory(GiB)": 103.58,
+      "step": 16565,
+      "train_speed(iter/s)": 1.318407
+    },
+    {
+      "acc": 0.68448529,
+      "epoch": 0.42034500253678336,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.314841981835652e-06,
+      "loss": 1.38290396,
+      "memory(GiB)": 103.58,
+      "step": 16570,
+      "train_speed(iter/s)": 1.31844
+    },
+    {
+      "acc": 0.67004547,
+      "epoch": 0.4204718417047184,
+      "grad_norm": 3.125,
+      "learning_rate": 9.314312061546974e-06,
+      "loss": 1.47773914,
+      "memory(GiB)": 103.58,
+      "step": 16575,
+      "train_speed(iter/s)": 1.318485
+    },
+    {
+      "acc": 0.6814559,
+      "epoch": 0.42059868087265345,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.313781951495588e-06,
+      "loss": 1.44314785,
+      "memory(GiB)": 103.58,
+      "step": 16580,
+      "train_speed(iter/s)": 1.318528
+    },
+    {
+      "acc": 0.68237753,
+      "epoch": 0.42072552004058855,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.313251651704816e-06,
+      "loss": 1.44193897,
+      "memory(GiB)": 103.58,
+      "step": 16585,
+      "train_speed(iter/s)": 1.318574
+    },
+    {
+      "acc": 0.68980856,
+      "epoch": 0.4208523592085236,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.312721162197975e-06,
+      "loss": 1.40276175,
+      "memory(GiB)": 103.58,
+      "step": 16590,
+      "train_speed(iter/s)": 1.318618
+    },
+    {
+      "acc": 0.6957901,
+      "epoch": 0.42097919837645864,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.312190482998405e-06,
+      "loss": 1.38612862,
+      "memory(GiB)": 103.58,
+      "step": 16595,
+      "train_speed(iter/s)": 1.318664
+    },
+    {
+      "acc": 0.68510695,
+      "epoch": 0.4211060375443937,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.311659614129443e-06,
+      "loss": 1.43384418,
+      "memory(GiB)": 103.58,
+      "step": 16600,
+      "train_speed(iter/s)": 1.318707
+    },
+    {
+      "acc": 0.67975855,
+      "epoch": 0.4212328767123288,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.311128555614443e-06,
+      "loss": 1.42701454,
+      "memory(GiB)": 103.58,
+      "step": 16605,
+      "train_speed(iter/s)": 1.318753
+    },
+    {
+      "acc": 0.69048157,
+      "epoch": 0.42135971588026383,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.31059730747676e-06,
+      "loss": 1.35326519,
+      "memory(GiB)": 103.58,
+      "step": 16610,
+      "train_speed(iter/s)": 1.318797
+    },
+    {
+      "acc": 0.69203339,
+      "epoch": 0.4214865550481989,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.310065869739763e-06,
+      "loss": 1.40782776,
+      "memory(GiB)": 103.58,
+      "step": 16615,
+      "train_speed(iter/s)": 1.318842
+    },
+    {
+      "acc": 0.68274326,
+      "epoch": 0.4216133942161339,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.309534242426826e-06,
+      "loss": 1.37524967,
+      "memory(GiB)": 103.58,
+      "step": 16620,
+      "train_speed(iter/s)": 1.318887
+    },
+    {
+      "acc": 0.69337988,
+      "epoch": 0.421740233384069,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.30900242556133e-06,
+      "loss": 1.37730675,
+      "memory(GiB)": 103.58,
+      "step": 16625,
+      "train_speed(iter/s)": 1.318932
+    },
+    {
+      "acc": 0.69422131,
+      "epoch": 0.42186707255200406,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.308470419166672e-06,
+      "loss": 1.33483448,
+      "memory(GiB)": 103.58,
+      "step": 16630,
+      "train_speed(iter/s)": 1.318976
+    },
+    {
+      "acc": 0.68123217,
+      "epoch": 0.4219939117199391,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.307938223266247e-06,
+      "loss": 1.39968414,
+      "memory(GiB)": 103.58,
+      "step": 16635,
+      "train_speed(iter/s)": 1.319019
+    },
+    {
+      "acc": 0.6712142,
+      "epoch": 0.42212075088787415,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.307405837883467e-06,
+      "loss": 1.47439404,
+      "memory(GiB)": 103.58,
+      "step": 16640,
+      "train_speed(iter/s)": 1.319063
+    },
+    {
+      "acc": 0.67609501,
+      "epoch": 0.42224759005580925,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.306873263041745e-06,
+      "loss": 1.41708908,
+      "memory(GiB)": 103.58,
+      "step": 16645,
+      "train_speed(iter/s)": 1.3191
+    },
+    {
+      "acc": 0.68545508,
+      "epoch": 0.4223744292237443,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.30634049876451e-06,
+      "loss": 1.39732828,
+      "memory(GiB)": 103.58,
+      "step": 16650,
+      "train_speed(iter/s)": 1.319143
+    },
+    {
+      "acc": 0.67827187,
+      "epoch": 0.42250126839167934,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.305807545075194e-06,
+      "loss": 1.41008158,
+      "memory(GiB)": 103.58,
+      "step": 16655,
+      "train_speed(iter/s)": 1.319186
+    },
+    {
+      "acc": 0.68246975,
+      "epoch": 0.4226281075596144,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.305274401997237e-06,
+      "loss": 1.41631546,
+      "memory(GiB)": 103.58,
+      "step": 16660,
+      "train_speed(iter/s)": 1.31923
+    },
+    {
+      "acc": 0.68556757,
+      "epoch": 0.4227549467275495,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.304741069554088e-06,
+      "loss": 1.37863846,
+      "memory(GiB)": 103.58,
+      "step": 16665,
+      "train_speed(iter/s)": 1.319272
+    },
+    {
+      "acc": 0.69803829,
+      "epoch": 0.42288178589548453,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.304207547769211e-06,
+      "loss": 1.35901203,
+      "memory(GiB)": 103.58,
+      "step": 16670,
+      "train_speed(iter/s)": 1.319314
+    },
+    {
+      "acc": 0.69889293,
+      "epoch": 0.4230086250634196,
+      "grad_norm": 3.5,
+      "learning_rate": 9.303673836666067e-06,
+      "loss": 1.39901752,
+      "memory(GiB)": 103.58,
+      "step": 16675,
+      "train_speed(iter/s)": 1.319358
+    },
+    {
+      "acc": 0.67570314,
+      "epoch": 0.4231354642313546,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.303139936268133e-06,
+      "loss": 1.43600283,
+      "memory(GiB)": 103.58,
+      "step": 16680,
+      "train_speed(iter/s)": 1.319402
+    },
+    {
+      "acc": 0.67480416,
+      "epoch": 0.4232623033992897,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.302605846598894e-06,
+      "loss": 1.45882721,
+      "memory(GiB)": 103.58,
+      "step": 16685,
+      "train_speed(iter/s)": 1.319447
+    },
+    {
+      "acc": 0.67832747,
+      "epoch": 0.42338914256722476,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.30207156768184e-06,
+      "loss": 1.41402321,
+      "memory(GiB)": 103.58,
+      "step": 16690,
+      "train_speed(iter/s)": 1.31949
+    },
+    {
+      "acc": 0.69028697,
+      "epoch": 0.4235159817351598,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.30153709954047e-06,
+      "loss": 1.34153156,
+      "memory(GiB)": 103.58,
+      "step": 16695,
+      "train_speed(iter/s)": 1.319533
+    },
+    {
+      "acc": 0.68665714,
+      "epoch": 0.42364282090309485,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.301002442198294e-06,
+      "loss": 1.35812654,
+      "memory(GiB)": 103.58,
+      "step": 16700,
+      "train_speed(iter/s)": 1.319576
+    },
+    {
+      "acc": 0.68085904,
+      "epoch": 0.42376966007102995,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.300467595678829e-06,
+      "loss": 1.41429443,
+      "memory(GiB)": 103.58,
+      "step": 16705,
+      "train_speed(iter/s)": 1.31962
+    },
+    {
+      "acc": 0.6870326,
+      "epoch": 0.423896499238965,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.299932560005596e-06,
+      "loss": 1.42147188,
+      "memory(GiB)": 103.58,
+      "step": 16710,
+      "train_speed(iter/s)": 1.319662
+    },
+    {
+      "acc": 0.68571005,
+      "epoch": 0.42402333840690004,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.299397335202133e-06,
+      "loss": 1.41787529,
+      "memory(GiB)": 103.58,
+      "step": 16715,
+      "train_speed(iter/s)": 1.319706
+    },
+    {
+      "acc": 0.69257259,
+      "epoch": 0.4241501775748351,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.29886192129198e-06,
+      "loss": 1.39792786,
+      "memory(GiB)": 103.58,
+      "step": 16720,
+      "train_speed(iter/s)": 1.319743
+    },
+    {
+      "acc": 0.67686243,
+      "epoch": 0.4242770167427702,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.298326318298688e-06,
+      "loss": 1.44470596,
+      "memory(GiB)": 103.58,
+      "step": 16725,
+      "train_speed(iter/s)": 1.319786
+    },
+    {
+      "acc": 0.6793849,
+      "epoch": 0.42440385591070523,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.29779052624581e-06,
+      "loss": 1.37488499,
+      "memory(GiB)": 103.58,
+      "step": 16730,
+      "train_speed(iter/s)": 1.31983
+    },
+    {
+      "acc": 0.68464508,
+      "epoch": 0.4245306950786403,
+      "grad_norm": 3.75,
+      "learning_rate": 9.29725454515692e-06,
+      "loss": 1.4115859,
+      "memory(GiB)": 103.58,
+      "step": 16735,
+      "train_speed(iter/s)": 1.319872
+    },
+    {
+      "acc": 0.68825626,
+      "epoch": 0.4246575342465753,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.296718375055587e-06,
+      "loss": 1.33890305,
+      "memory(GiB)": 103.58,
+      "step": 16740,
+      "train_speed(iter/s)": 1.319915
+    },
+    {
+      "acc": 0.69042101,
+      "epoch": 0.4247843734145104,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.296182015965399e-06,
+      "loss": 1.4034153,
+      "memory(GiB)": 103.58,
+      "step": 16745,
+      "train_speed(iter/s)": 1.319959
+    },
+    {
+      "acc": 0.68800783,
+      "epoch": 0.42491121258244546,
+      "grad_norm": 4.125,
+      "learning_rate": 9.295645467909942e-06,
+      "loss": 1.3833025,
+      "memory(GiB)": 103.58,
+      "step": 16750,
+      "train_speed(iter/s)": 1.320002
+    },
+    {
+      "acc": 0.67880077,
+      "epoch": 0.4250380517503805,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.29510873091282e-06,
+      "loss": 1.35150414,
+      "memory(GiB)": 103.58,
+      "step": 16755,
+      "train_speed(iter/s)": 1.320042
+    },
+    {
+      "acc": 0.68124242,
+      "epoch": 0.42516489091831555,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.29457180499764e-06,
+      "loss": 1.44315863,
+      "memory(GiB)": 103.58,
+      "step": 16760,
+      "train_speed(iter/s)": 1.320086
+    },
+    {
+      "acc": 0.68121901,
+      "epoch": 0.42529173008625065,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.294034690188016e-06,
+      "loss": 1.42672558,
+      "memory(GiB)": 103.58,
+      "step": 16765,
+      "train_speed(iter/s)": 1.320125
+    },
+    {
+      "acc": 0.67956638,
+      "epoch": 0.4254185692541857,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.293497386507577e-06,
+      "loss": 1.43991413,
+      "memory(GiB)": 103.58,
+      "step": 16770,
+      "train_speed(iter/s)": 1.320169
+    },
+    {
+      "acc": 0.6914515,
+      "epoch": 0.42554540842212074,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.292959893979953e-06,
+      "loss": 1.36857557,
+      "memory(GiB)": 103.58,
+      "step": 16775,
+      "train_speed(iter/s)": 1.320214
+    },
+    {
+      "acc": 0.68528938,
+      "epoch": 0.4256722475900558,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.292422212628786e-06,
+      "loss": 1.42460594,
+      "memory(GiB)": 103.58,
+      "step": 16780,
+      "train_speed(iter/s)": 1.32026
+    },
+    {
+      "acc": 0.69118147,
+      "epoch": 0.4257990867579909,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.291884342477728e-06,
+      "loss": 1.39679775,
+      "memory(GiB)": 103.58,
+      "step": 16785,
+      "train_speed(iter/s)": 1.320305
+    },
+    {
+      "acc": 0.67855492,
+      "epoch": 0.42592592592592593,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.291346283550433e-06,
+      "loss": 1.36122799,
+      "memory(GiB)": 103.58,
+      "step": 16790,
+      "train_speed(iter/s)": 1.320351
+    },
+    {
+      "acc": 0.68331375,
+      "epoch": 0.426052765093861,
+      "grad_norm": 3.375,
+      "learning_rate": 9.290808035870569e-06,
+      "loss": 1.45305233,
+      "memory(GiB)": 103.58,
+      "step": 16795,
+      "train_speed(iter/s)": 1.320395
+    },
+    {
+      "acc": 0.68381634,
+      "epoch": 0.426179604261796,
+      "grad_norm": 3.375,
+      "learning_rate": 9.29026959946181e-06,
+      "loss": 1.45682268,
+      "memory(GiB)": 103.58,
+      "step": 16800,
+      "train_speed(iter/s)": 1.320438
+    },
+    {
+      "acc": 0.68459163,
+      "epoch": 0.4263064434297311,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.289730974347841e-06,
+      "loss": 1.37615719,
+      "memory(GiB)": 103.58,
+      "step": 16805,
+      "train_speed(iter/s)": 1.320483
+    },
+    {
+      "acc": 0.70374856,
+      "epoch": 0.42643328259766616,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.28919216055235e-06,
+      "loss": 1.35844793,
+      "memory(GiB)": 103.58,
+      "step": 16810,
+      "train_speed(iter/s)": 1.320528
+    },
+    {
+      "acc": 0.69419699,
+      "epoch": 0.4265601217656012,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.288653158099038e-06,
+      "loss": 1.33143806,
+      "memory(GiB)": 103.58,
+      "step": 16815,
+      "train_speed(iter/s)": 1.320571
+    },
+    {
+      "acc": 0.69008107,
+      "epoch": 0.42668696093353625,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.288113967011612e-06,
+      "loss": 1.40730095,
+      "memory(GiB)": 103.58,
+      "step": 16820,
+      "train_speed(iter/s)": 1.320614
+    },
+    {
+      "acc": 0.69674854,
+      "epoch": 0.42681380010147135,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.28757458731379e-06,
+      "loss": 1.34939728,
+      "memory(GiB)": 103.58,
+      "step": 16825,
+      "train_speed(iter/s)": 1.320658
+    },
+    {
+      "acc": 0.67530842,
+      "epoch": 0.4269406392694064,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.287035019029295e-06,
+      "loss": 1.46061277,
+      "memory(GiB)": 103.58,
+      "step": 16830,
+      "train_speed(iter/s)": 1.320702
+    },
+    {
+      "acc": 0.6899971,
+      "epoch": 0.42706747843734144,
+      "grad_norm": 3.0,
+      "learning_rate": 9.286495262181859e-06,
+      "loss": 1.34047928,
+      "memory(GiB)": 103.58,
+      "step": 16835,
+      "train_speed(iter/s)": 1.320746
+    },
+    {
+      "acc": 0.68159437,
+      "epoch": 0.4271943176052765,
+      "grad_norm": 4.0,
+      "learning_rate": 9.285955316795224e-06,
+      "loss": 1.45396166,
+      "memory(GiB)": 103.58,
+      "step": 16840,
+      "train_speed(iter/s)": 1.320791
+    },
+    {
+      "acc": 0.68344069,
+      "epoch": 0.4273211567732116,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.285415182893138e-06,
+      "loss": 1.42150545,
+      "memory(GiB)": 103.58,
+      "step": 16845,
+      "train_speed(iter/s)": 1.320834
+    },
+    {
+      "acc": 0.67744122,
+      "epoch": 0.42744799594114663,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.28487486049936e-06,
+      "loss": 1.43001518,
+      "memory(GiB)": 103.58,
+      "step": 16850,
+      "train_speed(iter/s)": 1.320878
+    },
+    {
+      "acc": 0.68347316,
+      "epoch": 0.4275748351090817,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.284334349637655e-06,
+      "loss": 1.46999626,
+      "memory(GiB)": 103.58,
+      "step": 16855,
+      "train_speed(iter/s)": 1.320923
+    },
+    {
+      "acc": 0.70204968,
+      "epoch": 0.4277016742770167,
+      "grad_norm": 2.875,
+      "learning_rate": 9.283793650331798e-06,
+      "loss": 1.33210659,
+      "memory(GiB)": 103.58,
+      "step": 16860,
+      "train_speed(iter/s)": 1.320967
+    },
+    {
+      "acc": 0.69562759,
+      "epoch": 0.4278285134449518,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.283252762605568e-06,
+      "loss": 1.38911695,
+      "memory(GiB)": 103.58,
+      "step": 16865,
+      "train_speed(iter/s)": 1.321012
+    },
+    {
+      "acc": 0.68373709,
+      "epoch": 0.42795535261288686,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.28271168648276e-06,
+      "loss": 1.42851925,
+      "memory(GiB)": 103.58,
+      "step": 16870,
+      "train_speed(iter/s)": 1.321053
+    },
+    {
+      "acc": 0.67998085,
+      "epoch": 0.4280821917808219,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.282170421987171e-06,
+      "loss": 1.42442255,
+      "memory(GiB)": 103.58,
+      "step": 16875,
+      "train_speed(iter/s)": 1.321097
+    },
+    {
+      "acc": 0.67510343,
+      "epoch": 0.42820903094875695,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.281628969142609e-06,
+      "loss": 1.52305355,
+      "memory(GiB)": 103.58,
+      "step": 16880,
+      "train_speed(iter/s)": 1.321142
+    },
+    {
+      "acc": 0.66817746,
+      "epoch": 0.42833587011669205,
+      "grad_norm": 3.25,
+      "learning_rate": 9.281087327972886e-06,
+      "loss": 1.47247152,
+      "memory(GiB)": 103.58,
+      "step": 16885,
+      "train_speed(iter/s)": 1.321187
+    },
+    {
+      "acc": 0.68618832,
+      "epoch": 0.4284627092846271,
+      "grad_norm": 3.25,
+      "learning_rate": 9.280545498501832e-06,
+      "loss": 1.34135475,
+      "memory(GiB)": 103.58,
+      "step": 16890,
+      "train_speed(iter/s)": 1.32123
+    },
+    {
+      "acc": 0.6774334,
+      "epoch": 0.42858954845256214,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.280003480753274e-06,
+      "loss": 1.44889641,
+      "memory(GiB)": 103.58,
+      "step": 16895,
+      "train_speed(iter/s)": 1.321273
+    },
+    {
+      "acc": 0.66368389,
+      "epoch": 0.4287163876204972,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.279461274751054e-06,
+      "loss": 1.49162807,
+      "memory(GiB)": 103.58,
+      "step": 16900,
+      "train_speed(iter/s)": 1.321316
+    },
+    {
+      "acc": 0.67480373,
+      "epoch": 0.4288432267884323,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.27891888051902e-06,
+      "loss": 1.3997427,
+      "memory(GiB)": 103.58,
+      "step": 16905,
+      "train_speed(iter/s)": 1.321351
+    },
+    {
+      "acc": 0.68444662,
+      "epoch": 0.42897006595636733,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.278376298081032e-06,
+      "loss": 1.43496876,
+      "memory(GiB)": 103.58,
+      "step": 16910,
+      "train_speed(iter/s)": 1.321393
+    },
+    {
+      "acc": 0.68670273,
+      "epoch": 0.4290969051243024,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.277833527460952e-06,
+      "loss": 1.4230917,
+      "memory(GiB)": 103.58,
+      "step": 16915,
+      "train_speed(iter/s)": 1.321438
+    },
+    {
+      "acc": 0.69157119,
+      "epoch": 0.4292237442922374,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.277290568682653e-06,
+      "loss": 1.42471647,
+      "memory(GiB)": 103.58,
+      "step": 16920,
+      "train_speed(iter/s)": 1.321481
+    },
+    {
+      "acc": 0.66978054,
+      "epoch": 0.4293505834601725,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.27674742177002e-06,
+      "loss": 1.50041122,
+      "memory(GiB)": 103.58,
+      "step": 16925,
+      "train_speed(iter/s)": 1.321524
+    },
+    {
+      "acc": 0.67788115,
+      "epoch": 0.42947742262810756,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.27620408674694e-06,
+      "loss": 1.3841114,
+      "memory(GiB)": 103.58,
+      "step": 16930,
+      "train_speed(iter/s)": 1.321564
+    },
+    {
+      "acc": 0.67373018,
+      "epoch": 0.4296042617960426,
+      "grad_norm": 4.875,
+      "learning_rate": 9.275660563637313e-06,
+      "loss": 1.4011343,
+      "memory(GiB)": 103.58,
+      "step": 16935,
+      "train_speed(iter/s)": 1.321605
+    },
+    {
+      "acc": 0.69166894,
+      "epoch": 0.42973110096397765,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.275116852465043e-06,
+      "loss": 1.39903412,
+      "memory(GiB)": 103.58,
+      "step": 16940,
+      "train_speed(iter/s)": 1.321648
+    },
+    {
+      "acc": 0.68393049,
+      "epoch": 0.42985794013191275,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.274572953254048e-06,
+      "loss": 1.37754688,
+      "memory(GiB)": 103.58,
+      "step": 16945,
+      "train_speed(iter/s)": 1.321689
+    },
+    {
+      "acc": 0.69997835,
+      "epoch": 0.4299847792998478,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.27402886602825e-06,
+      "loss": 1.29537697,
+      "memory(GiB)": 103.58,
+      "step": 16950,
+      "train_speed(iter/s)": 1.321732
+    },
+    {
+      "acc": 0.67326217,
+      "epoch": 0.43011161846778284,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.27348459081158e-06,
+      "loss": 1.4280777,
+      "memory(GiB)": 103.58,
+      "step": 16955,
+      "train_speed(iter/s)": 1.321774
+    },
+    {
+      "acc": 0.68284359,
+      "epoch": 0.4302384576357179,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.272940127627979e-06,
+      "loss": 1.37643642,
+      "memory(GiB)": 103.58,
+      "step": 16960,
+      "train_speed(iter/s)": 1.321815
+    },
+    {
+      "acc": 0.68587103,
+      "epoch": 0.430365296803653,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.272395476501392e-06,
+      "loss": 1.40388126,
+      "memory(GiB)": 103.58,
+      "step": 16965,
+      "train_speed(iter/s)": 1.321859
+    },
+    {
+      "acc": 0.66638784,
+      "epoch": 0.43049213597158803,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.27185063745578e-06,
+      "loss": 1.48788242,
+      "memory(GiB)": 103.58,
+      "step": 16970,
+      "train_speed(iter/s)": 1.321901
+    },
+    {
+      "acc": 0.68911629,
+      "epoch": 0.4306189751395231,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.271305610515103e-06,
+      "loss": 1.3603651,
+      "memory(GiB)": 103.58,
+      "step": 16975,
+      "train_speed(iter/s)": 1.321942
+    },
+    {
+      "acc": 0.68104939,
+      "epoch": 0.4307458143074581,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.270760395703334e-06,
+      "loss": 1.44668608,
+      "memory(GiB)": 103.58,
+      "step": 16980,
+      "train_speed(iter/s)": 1.321978
+    },
+    {
+      "acc": 0.68447847,
+      "epoch": 0.4308726534753932,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.270214993044456e-06,
+      "loss": 1.42972813,
+      "memory(GiB)": 103.58,
+      "step": 16985,
+      "train_speed(iter/s)": 1.322019
+    },
+    {
+      "acc": 0.68030896,
+      "epoch": 0.43099949264332826,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.269669402562458e-06,
+      "loss": 1.41514759,
+      "memory(GiB)": 103.58,
+      "step": 16990,
+      "train_speed(iter/s)": 1.322056
+    },
+    {
+      "acc": 0.69106593,
+      "epoch": 0.4311263318112633,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.269123624281336e-06,
+      "loss": 1.39015083,
+      "memory(GiB)": 103.58,
+      "step": 16995,
+      "train_speed(iter/s)": 1.322099
+    },
+    {
+      "acc": 0.70567389,
+      "epoch": 0.43125317097919835,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.268577658225097e-06,
+      "loss": 1.29705124,
+      "memory(GiB)": 103.58,
+      "step": 17000,
+      "train_speed(iter/s)": 1.322141
+    },
+    {
+      "epoch": 0.43125317097919835,
+      "eval_acc": 0.6727185777800053,
+      "eval_loss": 1.370255470275879,
+      "eval_runtime": 70.0183,
+      "eval_samples_per_second": 90.976,
+      "eval_steps_per_second": 22.751,
+      "step": 17000
+    },
+    {
+      "acc": 0.6933558,
+      "epoch": 0.43138001014713345,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.268031504417756e-06,
+      "loss": 1.37967739,
+      "memory(GiB)": 103.58,
+      "step": 17005,
+      "train_speed(iter/s)": 1.313749
+    },
+    {
+      "acc": 0.67756968,
+      "epoch": 0.4315068493150685,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.267485162883334e-06,
+      "loss": 1.41726122,
+      "memory(GiB)": 103.58,
+      "step": 17010,
+      "train_speed(iter/s)": 1.31379
+    },
+    {
+      "acc": 0.67593207,
+      "epoch": 0.43163368848300354,
+      "grad_norm": 3.25,
+      "learning_rate": 9.266938633645861e-06,
+      "loss": 1.49772568,
+      "memory(GiB)": 103.58,
+      "step": 17015,
+      "train_speed(iter/s)": 1.313835
+    },
+    {
+      "acc": 0.69861298,
+      "epoch": 0.4317605276509386,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.266391916729376e-06,
+      "loss": 1.31940002,
+      "memory(GiB)": 103.58,
+      "step": 17020,
+      "train_speed(iter/s)": 1.31388
+    },
+    {
+      "acc": 0.67749267,
+      "epoch": 0.4318873668188737,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.265845012157926e-06,
+      "loss": 1.43340149,
+      "memory(GiB)": 103.58,
+      "step": 17025,
+      "train_speed(iter/s)": 1.313923
+    },
+    {
+      "acc": 0.67911005,
+      "epoch": 0.43201420598680873,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.265297919955566e-06,
+      "loss": 1.40188026,
+      "memory(GiB)": 103.58,
+      "step": 17030,
+      "train_speed(iter/s)": 1.313968
+    },
+    {
+      "acc": 0.68363571,
+      "epoch": 0.4321410451547438,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.264750640146363e-06,
+      "loss": 1.41996002,
+      "memory(GiB)": 103.58,
+      "step": 17035,
+      "train_speed(iter/s)": 1.314013
+    },
+    {
+      "acc": 0.6825182,
+      "epoch": 0.4322678843226788,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.264203172754384e-06,
+      "loss": 1.39577255,
+      "memory(GiB)": 103.58,
+      "step": 17040,
+      "train_speed(iter/s)": 1.314057
+    },
+    {
+      "acc": 0.68927674,
+      "epoch": 0.4323947234906139,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.263655517803713e-06,
+      "loss": 1.43524179,
+      "memory(GiB)": 103.58,
+      "step": 17045,
+      "train_speed(iter/s)": 1.314103
+    },
+    {
+      "acc": 0.67443409,
+      "epoch": 0.43252156265854896,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.263107675318434e-06,
+      "loss": 1.47168255,
+      "memory(GiB)": 103.58,
+      "step": 17050,
+      "train_speed(iter/s)": 1.314142
+    },
+    {
+      "acc": 0.69048119,
+      "epoch": 0.432648401826484,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.262559645322648e-06,
+      "loss": 1.3842432,
+      "memory(GiB)": 103.58,
+      "step": 17055,
+      "train_speed(iter/s)": 1.314188
+    },
+    {
+      "acc": 0.68718681,
+      "epoch": 0.43277524099441905,
+      "grad_norm": 3.25,
+      "learning_rate": 9.262011427840459e-06,
+      "loss": 1.40618,
+      "memory(GiB)": 103.58,
+      "step": 17060,
+      "train_speed(iter/s)": 1.314231
+    },
+    {
+      "acc": 0.68982081,
+      "epoch": 0.43290208016235415,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.261463022895976e-06,
+      "loss": 1.32220917,
+      "memory(GiB)": 103.58,
+      "step": 17065,
+      "train_speed(iter/s)": 1.314276
+    },
+    {
+      "acc": 0.68860488,
+      "epoch": 0.4330289193302892,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.260914430513325e-06,
+      "loss": 1.37908649,
+      "memory(GiB)": 103.58,
+      "step": 17070,
+      "train_speed(iter/s)": 1.314321
+    },
+    {
+      "acc": 0.66862597,
+      "epoch": 0.43315575849822424,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.260365650716632e-06,
+      "loss": 1.4804018,
+      "memory(GiB)": 103.58,
+      "step": 17075,
+      "train_speed(iter/s)": 1.314366
+    },
+    {
+      "acc": 0.68194885,
+      "epoch": 0.4332825976661593,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.259816683530038e-06,
+      "loss": 1.4397892,
+      "memory(GiB)": 103.58,
+      "step": 17080,
+      "train_speed(iter/s)": 1.314411
+    },
+    {
+      "acc": 0.69390163,
+      "epoch": 0.4334094368340944,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.259267528977687e-06,
+      "loss": 1.3201726,
+      "memory(GiB)": 103.58,
+      "step": 17085,
+      "train_speed(iter/s)": 1.314456
+    },
+    {
+      "acc": 0.68563824,
+      "epoch": 0.43353627600202943,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.258718187083734e-06,
+      "loss": 1.38698673,
+      "memory(GiB)": 103.58,
+      "step": 17090,
+      "train_speed(iter/s)": 1.314501
+    },
+    {
+      "acc": 0.69489174,
+      "epoch": 0.4336631151699645,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.258168657872341e-06,
+      "loss": 1.37867756,
+      "memory(GiB)": 103.58,
+      "step": 17095,
+      "train_speed(iter/s)": 1.314546
+    },
+    {
+      "acc": 0.69149737,
+      "epoch": 0.4337899543378995,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.25761894136768e-06,
+      "loss": 1.44616642,
+      "memory(GiB)": 103.58,
+      "step": 17100,
+      "train_speed(iter/s)": 1.314591
+    },
+    {
+      "acc": 0.67662191,
+      "epoch": 0.4339167935058346,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.25706903759393e-06,
+      "loss": 1.43873978,
+      "memory(GiB)": 103.58,
+      "step": 17105,
+      "train_speed(iter/s)": 1.314637
+    },
+    {
+      "acc": 0.69601622,
+      "epoch": 0.43404363267376966,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.256518946575274e-06,
+      "loss": 1.3451705,
+      "memory(GiB)": 103.58,
+      "step": 17110,
+      "train_speed(iter/s)": 1.314682
+    },
+    {
+      "acc": 0.68694391,
+      "epoch": 0.4341704718417047,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.255968668335912e-06,
+      "loss": 1.43521767,
+      "memory(GiB)": 103.58,
+      "step": 17115,
+      "train_speed(iter/s)": 1.314727
+    },
+    {
+      "acc": 0.68786898,
+      "epoch": 0.43429731100963975,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.255418202900048e-06,
+      "loss": 1.37950411,
+      "memory(GiB)": 103.58,
+      "step": 17120,
+      "train_speed(iter/s)": 1.31477
+    },
+    {
+      "acc": 0.677075,
+      "epoch": 0.43442415017757485,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.254867550291891e-06,
+      "loss": 1.42185764,
+      "memory(GiB)": 103.58,
+      "step": 17125,
+      "train_speed(iter/s)": 1.314816
+    },
+    {
+      "acc": 0.68422661,
+      "epoch": 0.4345509893455099,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.254316710535662e-06,
+      "loss": 1.34018049,
+      "memory(GiB)": 103.58,
+      "step": 17130,
+      "train_speed(iter/s)": 1.314861
+    },
+    {
+      "acc": 0.69548903,
+      "epoch": 0.43467782851344494,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.253765683655591e-06,
+      "loss": 1.3925602,
+      "memory(GiB)": 103.58,
+      "step": 17135,
+      "train_speed(iter/s)": 1.314906
+    },
+    {
+      "acc": 0.70385962,
+      "epoch": 0.43480466768138,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.253214469675913e-06,
+      "loss": 1.34826393,
+      "memory(GiB)": 103.58,
+      "step": 17140,
+      "train_speed(iter/s)": 1.314952
+    },
+    {
+      "acc": 0.69322519,
+      "epoch": 0.4349315068493151,
+      "grad_norm": 3.25,
+      "learning_rate": 9.252663068620874e-06,
+      "loss": 1.39987831,
+      "memory(GiB)": 103.58,
+      "step": 17145,
+      "train_speed(iter/s)": 1.314997
+    },
+    {
+      "acc": 0.67558603,
+      "epoch": 0.43505834601725013,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.252111480514726e-06,
+      "loss": 1.46113319,
+      "memory(GiB)": 103.58,
+      "step": 17150,
+      "train_speed(iter/s)": 1.315042
+    },
+    {
+      "acc": 0.68469372,
+      "epoch": 0.4351851851851852,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.251559705381731e-06,
+      "loss": 1.431674,
+      "memory(GiB)": 103.58,
+      "step": 17155,
+      "train_speed(iter/s)": 1.315086
+    },
+    {
+      "acc": 0.65929661,
+      "epoch": 0.4353120243531202,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.251007743246159e-06,
+      "loss": 1.45580139,
+      "memory(GiB)": 103.58,
+      "step": 17160,
+      "train_speed(iter/s)": 1.315129
+    },
+    {
+      "acc": 0.69370875,
+      "epoch": 0.4354388635210553,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.250455594132286e-06,
+      "loss": 1.38543682,
+      "memory(GiB)": 103.58,
+      "step": 17165,
+      "train_speed(iter/s)": 1.315174
+    },
+    {
+      "acc": 0.68739343,
+      "epoch": 0.43556570268899036,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.249903258064399e-06,
+      "loss": 1.45889969,
+      "memory(GiB)": 103.58,
+      "step": 17170,
+      "train_speed(iter/s)": 1.315217
+    },
+    {
+      "acc": 0.69775529,
+      "epoch": 0.4356925418569254,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.249350735066792e-06,
+      "loss": 1.32243195,
+      "memory(GiB)": 103.58,
+      "step": 17175,
+      "train_speed(iter/s)": 1.315263
+    },
+    {
+      "acc": 0.6681366,
+      "epoch": 0.43581938102486045,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.248798025163768e-06,
+      "loss": 1.41158199,
+      "memory(GiB)": 103.58,
+      "step": 17180,
+      "train_speed(iter/s)": 1.315306
+    },
+    {
+      "acc": 0.69492731,
+      "epoch": 0.43594622019279555,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.248245128379638e-06,
+      "loss": 1.34874535,
+      "memory(GiB)": 103.58,
+      "step": 17185,
+      "train_speed(iter/s)": 1.315351
+    },
+    {
+      "acc": 0.68620372,
+      "epoch": 0.4360730593607306,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.24769204473872e-06,
+      "loss": 1.39854536,
+      "memory(GiB)": 103.58,
+      "step": 17190,
+      "train_speed(iter/s)": 1.315394
+    },
+    {
+      "acc": 0.68085613,
+      "epoch": 0.43619989852866564,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.24713877426534e-06,
+      "loss": 1.39148455,
+      "memory(GiB)": 103.58,
+      "step": 17195,
+      "train_speed(iter/s)": 1.315439
+    },
+    {
+      "acc": 0.68815494,
+      "epoch": 0.4363267376966007,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.246585316983837e-06,
+      "loss": 1.41023951,
+      "memory(GiB)": 103.58,
+      "step": 17200,
+      "train_speed(iter/s)": 1.315484
+    },
+    {
+      "acc": 0.67436767,
+      "epoch": 0.4364535768645358,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.24603167291855e-06,
+      "loss": 1.47663803,
+      "memory(GiB)": 103.58,
+      "step": 17205,
+      "train_speed(iter/s)": 1.315529
+    },
+    {
+      "acc": 0.6844768,
+      "epoch": 0.43658041603247083,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.245477842093832e-06,
+      "loss": 1.40478668,
+      "memory(GiB)": 103.58,
+      "step": 17210,
+      "train_speed(iter/s)": 1.315572
+    },
+    {
+      "acc": 0.68215046,
+      "epoch": 0.4367072552004059,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.244923824534046e-06,
+      "loss": 1.42787609,
+      "memory(GiB)": 103.58,
+      "step": 17215,
+      "train_speed(iter/s)": 1.315617
+    },
+    {
+      "acc": 0.67433701,
+      "epoch": 0.4368340943683409,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.244369620263558e-06,
+      "loss": 1.41396313,
+      "memory(GiB)": 103.58,
+      "step": 17220,
+      "train_speed(iter/s)": 1.315662
+    },
+    {
+      "acc": 0.68591833,
+      "epoch": 0.436960933536276,
+      "grad_norm": 3.25,
+      "learning_rate": 9.243815229306746e-06,
+      "loss": 1.35870171,
+      "memory(GiB)": 103.58,
+      "step": 17225,
+      "train_speed(iter/s)": 1.315703
+    },
+    {
+      "acc": 0.67937317,
+      "epoch": 0.43708777270421106,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.243260651687989e-06,
+      "loss": 1.43433876,
+      "memory(GiB)": 103.58,
+      "step": 17230,
+      "train_speed(iter/s)": 1.315747
+    },
+    {
+      "acc": 0.68059897,
+      "epoch": 0.4372146118721461,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.242705887431685e-06,
+      "loss": 1.42952328,
+      "memory(GiB)": 103.58,
+      "step": 17235,
+      "train_speed(iter/s)": 1.315791
+    },
+    {
+      "acc": 0.68805566,
+      "epoch": 0.43734145104008115,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.242150936562235e-06,
+      "loss": 1.46332893,
+      "memory(GiB)": 103.58,
+      "step": 17240,
+      "train_speed(iter/s)": 1.315834
+    },
+    {
+      "acc": 0.6911005,
+      "epoch": 0.43746829020801625,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.241595799104046e-06,
+      "loss": 1.3624052,
+      "memory(GiB)": 103.58,
+      "step": 17245,
+      "train_speed(iter/s)": 1.315878
+    },
+    {
+      "acc": 0.68046689,
+      "epoch": 0.4375951293759513,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.241040475081537e-06,
+      "loss": 1.41621904,
+      "memory(GiB)": 103.58,
+      "step": 17250,
+      "train_speed(iter/s)": 1.31592
+    },
+    {
+      "acc": 0.6782268,
+      "epoch": 0.43772196854388634,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.240484964519131e-06,
+      "loss": 1.43225441,
+      "memory(GiB)": 103.58,
+      "step": 17255,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.69316192,
+      "epoch": 0.4378488077118214,
+      "grad_norm": 3.625,
+      "learning_rate": 9.239929267441267e-06,
+      "loss": 1.38857231,
+      "memory(GiB)": 103.58,
+      "step": 17260,
+      "train_speed(iter/s)": 1.315984
+    },
+    {
+      "acc": 0.69837956,
+      "epoch": 0.4379756468797565,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.239373383872382e-06,
+      "loss": 1.46143742,
+      "memory(GiB)": 103.58,
+      "step": 17265,
+      "train_speed(iter/s)": 1.316028
+    },
+    {
+      "acc": 0.67489986,
+      "epoch": 0.43810248604769153,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.238817313836927e-06,
+      "loss": 1.38328876,
+      "memory(GiB)": 103.58,
+      "step": 17270,
+      "train_speed(iter/s)": 1.31607
+    },
+    {
+      "acc": 0.6849637,
+      "epoch": 0.4382293252156266,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.238261057359365e-06,
+      "loss": 1.38372822,
+      "memory(GiB)": 103.58,
+      "step": 17275,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.68073149,
+      "epoch": 0.4383561643835616,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.237704614464157e-06,
+      "loss": 1.41254005,
+      "memory(GiB)": 103.58,
+      "step": 17280,
+      "train_speed(iter/s)": 1.316155
+    },
+    {
+      "acc": 0.67992206,
+      "epoch": 0.4384830035514967,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.237147985175781e-06,
+      "loss": 1.45997829,
+      "memory(GiB)": 103.58,
+      "step": 17285,
+      "train_speed(iter/s)": 1.316196
+    },
+    {
+      "acc": 0.68655653,
+      "epoch": 0.43860984271943176,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.236591169518717e-06,
+      "loss": 1.41195555,
+      "memory(GiB)": 103.58,
+      "step": 17290,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.68285551,
+      "epoch": 0.4387366818873668,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.236034167517461e-06,
+      "loss": 1.4501915,
+      "memory(GiB)": 103.58,
+      "step": 17295,
+      "train_speed(iter/s)": 1.316283
+    },
+    {
+      "acc": 0.69110041,
+      "epoch": 0.43886352105530185,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.235476979196507e-06,
+      "loss": 1.39655323,
+      "memory(GiB)": 103.58,
+      "step": 17300,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.69479446,
+      "epoch": 0.43899036022323695,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.234919604580368e-06,
+      "loss": 1.42478056,
+      "memory(GiB)": 103.58,
+      "step": 17305,
+      "train_speed(iter/s)": 1.31637
+    },
+    {
+      "acc": 0.69980326,
+      "epoch": 0.439117199391172,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.234362043693556e-06,
+      "loss": 1.34255257,
+      "memory(GiB)": 103.58,
+      "step": 17310,
+      "train_speed(iter/s)": 1.316413
+    },
+    {
+      "acc": 0.67072344,
+      "epoch": 0.43924403855910704,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.233804296560596e-06,
+      "loss": 1.43284149,
+      "memory(GiB)": 103.58,
+      "step": 17315,
+      "train_speed(iter/s)": 1.316456
+    },
+    {
+      "acc": 0.67209673,
+      "epoch": 0.4393708777270421,
+      "grad_norm": 3.25,
+      "learning_rate": 9.233246363206021e-06,
+      "loss": 1.47063847,
+      "memory(GiB)": 103.58,
+      "step": 17320,
+      "train_speed(iter/s)": 1.316499
+    },
+    {
+      "acc": 0.68243008,
+      "epoch": 0.4394977168949772,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.232688243654371e-06,
+      "loss": 1.4527173,
+      "memory(GiB)": 103.58,
+      "step": 17325,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.6724328,
+      "epoch": 0.43962455606291223,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.232129937930194e-06,
+      "loss": 1.47875509,
+      "memory(GiB)": 103.58,
+      "step": 17330,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.68906021,
+      "epoch": 0.4397513952308473,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.231571446058047e-06,
+      "loss": 1.44031124,
+      "memory(GiB)": 103.58,
+      "step": 17335,
+      "train_speed(iter/s)": 1.316629
+    },
+    {
+      "acc": 0.68762708,
+      "epoch": 0.4398782343987823,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.231012768062497e-06,
+      "loss": 1.40623484,
+      "memory(GiB)": 103.58,
+      "step": 17340,
+      "train_speed(iter/s)": 1.316671
+    },
+    {
+      "acc": 0.67949886,
+      "epoch": 0.4400050735667174,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.230453903968112e-06,
+      "loss": 1.43782864,
+      "memory(GiB)": 103.58,
+      "step": 17345,
+      "train_speed(iter/s)": 1.316713
+    },
+    {
+      "acc": 0.69101048,
+      "epoch": 0.44013191273465246,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.22989485379948e-06,
+      "loss": 1.35737419,
+      "memory(GiB)": 103.58,
+      "step": 17350,
+      "train_speed(iter/s)": 1.316757
+    },
+    {
+      "acc": 0.68966713,
+      "epoch": 0.4402587519025875,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.229335617581187e-06,
+      "loss": 1.3233202,
+      "memory(GiB)": 103.58,
+      "step": 17355,
+      "train_speed(iter/s)": 1.3168
+    },
+    {
+      "acc": 0.69461894,
+      "epoch": 0.44038559107052255,
+      "grad_norm": 2.875,
+      "learning_rate": 9.22877619533783e-06,
+      "loss": 1.34667349,
+      "memory(GiB)": 103.58,
+      "step": 17360,
+      "train_speed(iter/s)": 1.316844
+    },
+    {
+      "acc": 0.68709092,
+      "epoch": 0.44051243023845765,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.228216587094014e-06,
+      "loss": 1.38291759,
+      "memory(GiB)": 103.58,
+      "step": 17365,
+      "train_speed(iter/s)": 1.316888
+    },
+    {
+      "acc": 0.70009232,
+      "epoch": 0.4406392694063927,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.227656792874358e-06,
+      "loss": 1.30372486,
+      "memory(GiB)": 103.58,
+      "step": 17370,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.68795681,
+      "epoch": 0.44076610857432774,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.227096812703479e-06,
+      "loss": 1.4565424,
+      "memory(GiB)": 103.58,
+      "step": 17375,
+      "train_speed(iter/s)": 1.316974
+    },
+    {
+      "acc": 0.69329214,
+      "epoch": 0.4408929477422628,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.22653664660601e-06,
+      "loss": 1.39974251,
+      "memory(GiB)": 103.58,
+      "step": 17380,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.69044323,
+      "epoch": 0.4410197869101979,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.225976294606589e-06,
+      "loss": 1.41651268,
+      "memory(GiB)": 103.58,
+      "step": 17385,
+      "train_speed(iter/s)": 1.31706
+    },
+    {
+      "acc": 0.680937,
+      "epoch": 0.44114662607813293,
+      "grad_norm": 3.375,
+      "learning_rate": 9.225415756729863e-06,
+      "loss": 1.40972729,
+      "memory(GiB)": 103.58,
+      "step": 17390,
+      "train_speed(iter/s)": 1.317103
+    },
+    {
+      "acc": 0.67087402,
+      "epoch": 0.441273465246068,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.224855033000489e-06,
+      "loss": 1.43070793,
+      "memory(GiB)": 103.58,
+      "step": 17395,
+      "train_speed(iter/s)": 1.317148
+    },
+    {
+      "acc": 0.69942808,
+      "epoch": 0.441400304414003,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.224294123443125e-06,
+      "loss": 1.35707645,
+      "memory(GiB)": 103.58,
+      "step": 17400,
+      "train_speed(iter/s)": 1.317189
+    },
+    {
+      "acc": 0.69414043,
+      "epoch": 0.4415271435819381,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.223733028082447e-06,
+      "loss": 1.36811543,
+      "memory(GiB)": 103.58,
+      "step": 17405,
+      "train_speed(iter/s)": 1.317231
+    },
+    {
+      "acc": 0.67868242,
+      "epoch": 0.44165398274987316,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.223171746943132e-06,
+      "loss": 1.39343386,
+      "memory(GiB)": 103.58,
+      "step": 17410,
+      "train_speed(iter/s)": 1.317276
+    },
+    {
+      "acc": 0.69333858,
+      "epoch": 0.4417808219178082,
+      "grad_norm": 2.875,
+      "learning_rate": 9.222610280049868e-06,
+      "loss": 1.43934765,
+      "memory(GiB)": 103.58,
+      "step": 17415,
+      "train_speed(iter/s)": 1.317317
+    },
+    {
+      "acc": 0.67722697,
+      "epoch": 0.44190766108574325,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.222048627427352e-06,
+      "loss": 1.46140957,
+      "memory(GiB)": 103.58,
+      "step": 17420,
+      "train_speed(iter/s)": 1.31736
+    },
+    {
+      "acc": 0.68870449,
+      "epoch": 0.44203450025367835,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.221486789100288e-06,
+      "loss": 1.3942935,
+      "memory(GiB)": 103.58,
+      "step": 17425,
+      "train_speed(iter/s)": 1.317402
+    },
+    {
+      "acc": 0.67440271,
+      "epoch": 0.4421613394216134,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.220924765093386e-06,
+      "loss": 1.39044218,
+      "memory(GiB)": 103.58,
+      "step": 17430,
+      "train_speed(iter/s)": 1.317444
+    },
+    {
+      "acc": 0.6736104,
+      "epoch": 0.44228817858954844,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.220362555431369e-06,
+      "loss": 1.42132721,
+      "memory(GiB)": 103.58,
+      "step": 17435,
+      "train_speed(iter/s)": 1.317488
+    },
+    {
+      "acc": 0.67687483,
+      "epoch": 0.4424150177574835,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.219800160138964e-06,
+      "loss": 1.35640631,
+      "memory(GiB)": 103.58,
+      "step": 17440,
+      "train_speed(iter/s)": 1.317532
+    },
+    {
+      "acc": 0.68083453,
+      "epoch": 0.4425418569254186,
+      "grad_norm": 3.375,
+      "learning_rate": 9.219237579240907e-06,
+      "loss": 1.41440659,
+      "memory(GiB)": 103.58,
+      "step": 17445,
+      "train_speed(iter/s)": 1.317575
+    },
+    {
+      "acc": 0.69604292,
+      "epoch": 0.44266869609335363,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.218674812761946e-06,
+      "loss": 1.36945534,
+      "memory(GiB)": 103.58,
+      "step": 17450,
+      "train_speed(iter/s)": 1.317619
+    },
+    {
+      "acc": 0.68646102,
+      "epoch": 0.4427955352612887,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.21811186072683e-06,
+      "loss": 1.37772141,
+      "memory(GiB)": 103.58,
+      "step": 17455,
+      "train_speed(iter/s)": 1.317662
+    },
+    {
+      "acc": 0.69309397,
+      "epoch": 0.4429223744292237,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.21754872316032e-06,
+      "loss": 1.38533211,
+      "memory(GiB)": 103.58,
+      "step": 17460,
+      "train_speed(iter/s)": 1.317705
+    },
+    {
+      "acc": 0.70369229,
+      "epoch": 0.4430492135971588,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.21698540008719e-06,
+      "loss": 1.35472221,
+      "memory(GiB)": 103.58,
+      "step": 17465,
+      "train_speed(iter/s)": 1.317748
+    },
+    {
+      "acc": 0.68386111,
+      "epoch": 0.44317605276509386,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.216421891532214e-06,
+      "loss": 1.39204388,
+      "memory(GiB)": 103.58,
+      "step": 17470,
+      "train_speed(iter/s)": 1.317786
+    },
+    {
+      "acc": 0.70444384,
+      "epoch": 0.4433028919330289,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.215858197520178e-06,
+      "loss": 1.38780794,
+      "memory(GiB)": 103.58,
+      "step": 17475,
+      "train_speed(iter/s)": 1.317829
+    },
+    {
+      "acc": 0.69196992,
+      "epoch": 0.44342973110096395,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.215294318075876e-06,
+      "loss": 1.40978832,
+      "memory(GiB)": 103.58,
+      "step": 17480,
+      "train_speed(iter/s)": 1.317872
+    },
+    {
+      "acc": 0.69360428,
+      "epoch": 0.44355657026889905,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.21473025322411e-06,
+      "loss": 1.3486145,
+      "memory(GiB)": 103.58,
+      "step": 17485,
+      "train_speed(iter/s)": 1.317917
+    },
+    {
+      "acc": 0.67703781,
+      "epoch": 0.4436834094368341,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.21416600298969e-06,
+      "loss": 1.44370317,
+      "memory(GiB)": 103.58,
+      "step": 17490,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.6858469,
+      "epoch": 0.44381024860476914,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.213601567397434e-06,
+      "loss": 1.3260994,
+      "memory(GiB)": 103.58,
+      "step": 17495,
+      "train_speed(iter/s)": 1.317997
+    },
+    {
+      "acc": 0.67190075,
+      "epoch": 0.4439370877727042,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.213036946472169e-06,
+      "loss": 1.4851841,
+      "memory(GiB)": 103.58,
+      "step": 17500,
+      "train_speed(iter/s)": 1.31804
+    },
+    {
+      "acc": 0.67153311,
+      "epoch": 0.4440639269406393,
+      "grad_norm": 3.125,
+      "learning_rate": 9.212472140238729e-06,
+      "loss": 1.40444183,
+      "memory(GiB)": 103.58,
+      "step": 17505,
+      "train_speed(iter/s)": 1.318082
+    },
+    {
+      "acc": 0.68531508,
+      "epoch": 0.44419076610857433,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.211907148721958e-06,
+      "loss": 1.39016933,
+      "memory(GiB)": 103.58,
+      "step": 17510,
+      "train_speed(iter/s)": 1.318125
+    },
+    {
+      "acc": 0.68975706,
+      "epoch": 0.4443176052765094,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.211341971946705e-06,
+      "loss": 1.41874218,
+      "memory(GiB)": 103.58,
+      "step": 17515,
+      "train_speed(iter/s)": 1.318167
+    },
+    {
+      "acc": 0.69754667,
+      "epoch": 0.4444444444444444,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.21077660993783e-06,
+      "loss": 1.39080105,
+      "memory(GiB)": 103.58,
+      "step": 17520,
+      "train_speed(iter/s)": 1.318209
+    },
+    {
+      "acc": 0.68893585,
+      "epoch": 0.4445712836123795,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.210211062720198e-06,
+      "loss": 1.43290348,
+      "memory(GiB)": 103.58,
+      "step": 17525,
+      "train_speed(iter/s)": 1.318244
+    },
+    {
+      "acc": 0.66763101,
+      "epoch": 0.44469812278031456,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.209645330318689e-06,
+      "loss": 1.46232052,
+      "memory(GiB)": 103.58,
+      "step": 17530,
+      "train_speed(iter/s)": 1.318286
+    },
+    {
+      "acc": 0.69106417,
+      "epoch": 0.4448249619482496,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.209079412758183e-06,
+      "loss": 1.42004261,
+      "memory(GiB)": 103.58,
+      "step": 17535,
+      "train_speed(iter/s)": 1.318327
+    },
+    {
+      "acc": 0.67940063,
+      "epoch": 0.44495180111618465,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.208513310063572e-06,
+      "loss": 1.41828289,
+      "memory(GiB)": 103.58,
+      "step": 17540,
+      "train_speed(iter/s)": 1.318368
+    },
+    {
+      "acc": 0.69168062,
+      "epoch": 0.44507864028411975,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.207947022259755e-06,
+      "loss": 1.39349403,
+      "memory(GiB)": 103.58,
+      "step": 17545,
+      "train_speed(iter/s)": 1.318413
+    },
+    {
+      "acc": 0.68847713,
+      "epoch": 0.4452054794520548,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.207380549371642e-06,
+      "loss": 1.35426197,
+      "memory(GiB)": 103.58,
+      "step": 17550,
+      "train_speed(iter/s)": 1.318455
+    },
+    {
+      "acc": 0.67557764,
+      "epoch": 0.44533231861998984,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.206813891424147e-06,
+      "loss": 1.42605114,
+      "memory(GiB)": 103.58,
+      "step": 17555,
+      "train_speed(iter/s)": 1.318496
+    },
+    {
+      "acc": 0.6762773,
+      "epoch": 0.4454591577879249,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.206247048442196e-06,
+      "loss": 1.45389709,
+      "memory(GiB)": 103.58,
+      "step": 17560,
+      "train_speed(iter/s)": 1.318538
+    },
+    {
+      "acc": 0.68900309,
+      "epoch": 0.44558599695586,
+      "grad_norm": 2.875,
+      "learning_rate": 9.20568002045072e-06,
+      "loss": 1.39267588,
+      "memory(GiB)": 103.58,
+      "step": 17565,
+      "train_speed(iter/s)": 1.318581
+    },
+    {
+      "acc": 0.67346668,
+      "epoch": 0.44571283612379503,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.20511280747466e-06,
+      "loss": 1.41365604,
+      "memory(GiB)": 103.58,
+      "step": 17570,
+      "train_speed(iter/s)": 1.318624
+    },
+    {
+      "acc": 0.69779296,
+      "epoch": 0.4458396752917301,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.204545409538962e-06,
+      "loss": 1.31486111,
+      "memory(GiB)": 103.58,
+      "step": 17575,
+      "train_speed(iter/s)": 1.318669
+    },
+    {
+      "acc": 0.6690877,
+      "epoch": 0.4459665144596651,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.203977826668587e-06,
+      "loss": 1.50983648,
+      "memory(GiB)": 103.58,
+      "step": 17580,
+      "train_speed(iter/s)": 1.318712
+    },
+    {
+      "acc": 0.70015793,
+      "epoch": 0.4460933536276002,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.203410058888498e-06,
+      "loss": 1.33080072,
+      "memory(GiB)": 103.58,
+      "step": 17585,
+      "train_speed(iter/s)": 1.318751
+    },
+    {
+      "acc": 0.68879023,
+      "epoch": 0.44622019279553526,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.202842106223667e-06,
+      "loss": 1.39588509,
+      "memory(GiB)": 103.58,
+      "step": 17590,
+      "train_speed(iter/s)": 1.318791
+    },
+    {
+      "acc": 0.67209802,
+      "epoch": 0.4463470319634703,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.202273968699075e-06,
+      "loss": 1.46674681,
+      "memory(GiB)": 103.58,
+      "step": 17595,
+      "train_speed(iter/s)": 1.318833
+    },
+    {
+      "acc": 0.69776468,
+      "epoch": 0.44647387113140535,
+      "grad_norm": 4.125,
+      "learning_rate": 9.201705646339714e-06,
+      "loss": 1.40134211,
+      "memory(GiB)": 103.58,
+      "step": 17600,
+      "train_speed(iter/s)": 1.318875
+    },
+    {
+      "acc": 0.68127232,
+      "epoch": 0.44660071029934045,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.201137139170578e-06,
+      "loss": 1.36567383,
+      "memory(GiB)": 103.58,
+      "step": 17605,
+      "train_speed(iter/s)": 1.318916
+    },
+    {
+      "acc": 0.69463549,
+      "epoch": 0.4467275494672755,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.200568447216673e-06,
+      "loss": 1.36935787,
+      "memory(GiB)": 103.58,
+      "step": 17610,
+      "train_speed(iter/s)": 1.318957
+    },
+    {
+      "acc": 0.68284011,
+      "epoch": 0.44685438863521054,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.199999570503015e-06,
+      "loss": 1.345714,
+      "memory(GiB)": 103.58,
+      "step": 17615,
+      "train_speed(iter/s)": 1.318998
+    },
+    {
+      "acc": 0.66196661,
+      "epoch": 0.4469812278031456,
+      "grad_norm": 2.40625,
+      "learning_rate": 9.199430509054625e-06,
+      "loss": 1.51028061,
+      "memory(GiB)": 103.58,
+      "step": 17620,
+      "train_speed(iter/s)": 1.319041
+    },
+    {
+      "acc": 0.68462143,
+      "epoch": 0.4471080669710807,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.19886126289653e-06,
+      "loss": 1.40374527,
+      "memory(GiB)": 103.58,
+      "step": 17625,
+      "train_speed(iter/s)": 1.319082
+    },
+    {
+      "acc": 0.67850461,
+      "epoch": 0.44723490613901573,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.198291832053771e-06,
+      "loss": 1.43876562,
+      "memory(GiB)": 103.58,
+      "step": 17630,
+      "train_speed(iter/s)": 1.319124
+    },
+    {
+      "acc": 0.6912683,
+      "epoch": 0.4473617453069508,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.197722216551395e-06,
+      "loss": 1.41209612,
+      "memory(GiB)": 103.58,
+      "step": 17635,
+      "train_speed(iter/s)": 1.319167
+    },
+    {
+      "acc": 0.69275727,
+      "epoch": 0.4474885844748858,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.197152416414452e-06,
+      "loss": 1.34605827,
+      "memory(GiB)": 103.58,
+      "step": 17640,
+      "train_speed(iter/s)": 1.31921
+    },
+    {
+      "acc": 0.67436275,
+      "epoch": 0.4476154236428209,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.196582431668007e-06,
+      "loss": 1.49487057,
+      "memory(GiB)": 103.58,
+      "step": 17645,
+      "train_speed(iter/s)": 1.31925
+    },
+    {
+      "acc": 0.68231511,
+      "epoch": 0.44774226281075596,
+      "grad_norm": 3.375,
+      "learning_rate": 9.196012262337131e-06,
+      "loss": 1.36661358,
+      "memory(GiB)": 103.58,
+      "step": 17650,
+      "train_speed(iter/s)": 1.31929
+    },
+    {
+      "acc": 0.67640371,
+      "epoch": 0.447869101978691,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.195441908446902e-06,
+      "loss": 1.47019711,
+      "memory(GiB)": 103.58,
+      "step": 17655,
+      "train_speed(iter/s)": 1.319333
+    },
+    {
+      "acc": 0.68482008,
+      "epoch": 0.44799594114662605,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.194871370022407e-06,
+      "loss": 1.41426964,
+      "memory(GiB)": 103.58,
+      "step": 17660,
+      "train_speed(iter/s)": 1.319375
+    },
+    {
+      "acc": 0.69457741,
+      "epoch": 0.44812278031456115,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.194300647088739e-06,
+      "loss": 1.3839963,
+      "memory(GiB)": 103.58,
+      "step": 17665,
+      "train_speed(iter/s)": 1.319415
+    },
+    {
+      "acc": 0.67149639,
+      "epoch": 0.4482496194824962,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.193729739671002e-06,
+      "loss": 1.43540726,
+      "memory(GiB)": 103.58,
+      "step": 17670,
+      "train_speed(iter/s)": 1.319456
+    },
+    {
+      "acc": 0.67893152,
+      "epoch": 0.44837645865043124,
+      "grad_norm": 2.5,
+      "learning_rate": 9.193158647794308e-06,
+      "loss": 1.40055857,
+      "memory(GiB)": 103.58,
+      "step": 17675,
+      "train_speed(iter/s)": 1.319496
+    },
+    {
+      "acc": 0.67272105,
+      "epoch": 0.4485032978183663,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.192587371483777e-06,
+      "loss": 1.44613829,
+      "memory(GiB)": 103.58,
+      "step": 17680,
+      "train_speed(iter/s)": 1.319538
+    },
+    {
+      "acc": 0.66888924,
+      "epoch": 0.4486301369863014,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.192015910764535e-06,
+      "loss": 1.48117237,
+      "memory(GiB)": 103.58,
+      "step": 17685,
+      "train_speed(iter/s)": 1.319579
+    },
+    {
+      "acc": 0.69020109,
+      "epoch": 0.44875697615423643,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.191444265661715e-06,
+      "loss": 1.43210011,
+      "memory(GiB)": 103.58,
+      "step": 17690,
+      "train_speed(iter/s)": 1.319622
+    },
+    {
+      "acc": 0.68343925,
+      "epoch": 0.4488838153221715,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.190872436200464e-06,
+      "loss": 1.42774544,
+      "memory(GiB)": 103.58,
+      "step": 17695,
+      "train_speed(iter/s)": 1.319664
+    },
+    {
+      "acc": 0.67378168,
+      "epoch": 0.4490106544901065,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.190300422405933e-06,
+      "loss": 1.39493294,
+      "memory(GiB)": 103.58,
+      "step": 17700,
+      "train_speed(iter/s)": 1.319707
+    },
+    {
+      "acc": 0.68312006,
+      "epoch": 0.4491374936580416,
+      "grad_norm": 3.125,
+      "learning_rate": 9.18972822430328e-06,
+      "loss": 1.43546705,
+      "memory(GiB)": 103.58,
+      "step": 17705,
+      "train_speed(iter/s)": 1.31975
+    },
+    {
+      "acc": 0.68971238,
+      "epoch": 0.44926433282597666,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.189155841917675e-06,
+      "loss": 1.33979874,
+      "memory(GiB)": 103.58,
+      "step": 17710,
+      "train_speed(iter/s)": 1.319792
+    },
+    {
+      "acc": 0.67359247,
+      "epoch": 0.4493911719939117,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.18858327527429e-06,
+      "loss": 1.44330978,
+      "memory(GiB)": 103.58,
+      "step": 17715,
+      "train_speed(iter/s)": 1.319835
+    },
+    {
+      "acc": 0.67396154,
+      "epoch": 0.44951801116184675,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.188010524398314e-06,
+      "loss": 1.41264458,
+      "memory(GiB)": 103.58,
+      "step": 17720,
+      "train_speed(iter/s)": 1.319875
+    },
+    {
+      "acc": 0.7077301,
+      "epoch": 0.44964485032978185,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.187437589314939e-06,
+      "loss": 1.30978146,
+      "memory(GiB)": 103.58,
+      "step": 17725,
+      "train_speed(iter/s)": 1.319919
+    },
+    {
+      "acc": 0.68225622,
+      "epoch": 0.4497716894977169,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.186864470049358e-06,
+      "loss": 1.39009838,
+      "memory(GiB)": 103.58,
+      "step": 17730,
+      "train_speed(iter/s)": 1.319962
+    },
+    {
+      "acc": 0.69565201,
+      "epoch": 0.44989852866565194,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.186291166626789e-06,
+      "loss": 1.37295904,
+      "memory(GiB)": 103.58,
+      "step": 17735,
+      "train_speed(iter/s)": 1.320007
+    },
+    {
+      "acc": 0.68909283,
+      "epoch": 0.450025367833587,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.185717679072444e-06,
+      "loss": 1.4000309,
+      "memory(GiB)": 103.58,
+      "step": 17740,
+      "train_speed(iter/s)": 1.320049
+    },
+    {
+      "acc": 0.68017559,
+      "epoch": 0.4501522070015221,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.185144007411547e-06,
+      "loss": 1.40386333,
+      "memory(GiB)": 103.58,
+      "step": 17745,
+      "train_speed(iter/s)": 1.320089
+    },
+    {
+      "acc": 0.66254711,
+      "epoch": 0.45027904616945713,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.18457015166933e-06,
+      "loss": 1.410709,
+      "memory(GiB)": 103.58,
+      "step": 17750,
+      "train_speed(iter/s)": 1.320132
+    },
+    {
+      "acc": 0.69049087,
+      "epoch": 0.4504058853373922,
+      "grad_norm": 3.625,
+      "learning_rate": 9.183996111871034e-06,
+      "loss": 1.3634037,
+      "memory(GiB)": 103.58,
+      "step": 17755,
+      "train_speed(iter/s)": 1.320169
+    },
+    {
+      "acc": 0.69058895,
+      "epoch": 0.4505327245053272,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.18342188804191e-06,
+      "loss": 1.4515707,
+      "memory(GiB)": 103.58,
+      "step": 17760,
+      "train_speed(iter/s)": 1.320211
+    },
+    {
+      "acc": 0.68497,
+      "epoch": 0.4506595636732623,
+      "grad_norm": 5.5625,
+      "learning_rate": 9.182847480207215e-06,
+      "loss": 1.38735838,
+      "memory(GiB)": 103.58,
+      "step": 17765,
+      "train_speed(iter/s)": 1.320252
+    },
+    {
+      "acc": 0.69526362,
+      "epoch": 0.45078640284119736,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.182272888392211e-06,
+      "loss": 1.37652359,
+      "memory(GiB)": 103.58,
+      "step": 17770,
+      "train_speed(iter/s)": 1.320295
+    },
+    {
+      "acc": 0.6780427,
+      "epoch": 0.4509132420091324,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.181698112622175e-06,
+      "loss": 1.41378374,
+      "memory(GiB)": 103.58,
+      "step": 17775,
+      "train_speed(iter/s)": 1.320338
+    },
+    {
+      "acc": 0.65985975,
+      "epoch": 0.45104008117706745,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.181123152922384e-06,
+      "loss": 1.44863529,
+      "memory(GiB)": 103.58,
+      "step": 17780,
+      "train_speed(iter/s)": 1.32038
+    },
+    {
+      "acc": 0.67656136,
+      "epoch": 0.45116692034500255,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.18054800931813e-06,
+      "loss": 1.42198715,
+      "memory(GiB)": 103.58,
+      "step": 17785,
+      "train_speed(iter/s)": 1.320421
+    },
+    {
+      "acc": 0.67940812,
+      "epoch": 0.4512937595129376,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.17997268183471e-06,
+      "loss": 1.37274027,
+      "memory(GiB)": 103.58,
+      "step": 17790,
+      "train_speed(iter/s)": 1.320463
+    },
+    {
+      "acc": 0.67132754,
+      "epoch": 0.45142059868087264,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.17939717049743e-06,
+      "loss": 1.45173883,
+      "memory(GiB)": 103.58,
+      "step": 17795,
+      "train_speed(iter/s)": 1.320507
+    },
+    {
+      "acc": 0.68497314,
+      "epoch": 0.4515474378488077,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.1788214753316e-06,
+      "loss": 1.44672661,
+      "memory(GiB)": 103.58,
+      "step": 17800,
+      "train_speed(iter/s)": 1.320549
+    },
+    {
+      "acc": 0.69727302,
+      "epoch": 0.4516742770167428,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.178245596362546e-06,
+      "loss": 1.38922739,
+      "memory(GiB)": 103.58,
+      "step": 17805,
+      "train_speed(iter/s)": 1.32059
+    },
+    {
+      "acc": 0.6833178,
+      "epoch": 0.45180111618467783,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.177669533615599e-06,
+      "loss": 1.38823957,
+      "memory(GiB)": 103.58,
+      "step": 17810,
+      "train_speed(iter/s)": 1.320633
+    },
+    {
+      "acc": 0.69181042,
+      "epoch": 0.4519279553526129,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.17709328711609e-06,
+      "loss": 1.33911896,
+      "memory(GiB)": 103.58,
+      "step": 17815,
+      "train_speed(iter/s)": 1.320676
+    },
+    {
+      "acc": 0.67371435,
+      "epoch": 0.4520547945205479,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.17651685688937e-06,
+      "loss": 1.44330692,
+      "memory(GiB)": 103.58,
+      "step": 17820,
+      "train_speed(iter/s)": 1.320718
+    },
+    {
+      "acc": 0.68960371,
+      "epoch": 0.452181633688483,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.175940242960792e-06,
+      "loss": 1.43128681,
+      "memory(GiB)": 103.58,
+      "step": 17825,
+      "train_speed(iter/s)": 1.320761
+    },
+    {
+      "acc": 0.67843833,
+      "epoch": 0.45230847285641806,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.175363445355718e-06,
+      "loss": 1.46179504,
+      "memory(GiB)": 103.58,
+      "step": 17830,
+      "train_speed(iter/s)": 1.320803
+    },
+    {
+      "acc": 0.69433117,
+      "epoch": 0.4524353120243531,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.174786464099519e-06,
+      "loss": 1.39372807,
+      "memory(GiB)": 103.58,
+      "step": 17835,
+      "train_speed(iter/s)": 1.320844
+    },
+    {
+      "acc": 0.67555037,
+      "epoch": 0.45256215119228815,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.17420929921757e-06,
+      "loss": 1.42903585,
+      "memory(GiB)": 103.58,
+      "step": 17840,
+      "train_speed(iter/s)": 1.320884
+    },
+    {
+      "acc": 0.69934902,
+      "epoch": 0.45268899036022325,
+      "grad_norm": 3.75,
+      "learning_rate": 9.173631950735262e-06,
+      "loss": 1.35437756,
+      "memory(GiB)": 103.58,
+      "step": 17845,
+      "train_speed(iter/s)": 1.320924
+    },
+    {
+      "acc": 0.69121547,
+      "epoch": 0.4528158295281583,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.173054418677985e-06,
+      "loss": 1.4291256,
+      "memory(GiB)": 103.58,
+      "step": 17850,
+      "train_speed(iter/s)": 1.320966
+    },
+    {
+      "acc": 0.69472399,
+      "epoch": 0.45294266869609334,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.172476703071145e-06,
+      "loss": 1.37598019,
+      "memory(GiB)": 103.58,
+      "step": 17855,
+      "train_speed(iter/s)": 1.321006
+    },
+    {
+      "acc": 0.68909426,
+      "epoch": 0.4530695078640284,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.171898803940148e-06,
+      "loss": 1.37505016,
+      "memory(GiB)": 103.58,
+      "step": 17860,
+      "train_speed(iter/s)": 1.321048
+    },
+    {
+      "acc": 0.6917871,
+      "epoch": 0.4531963470319635,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.17132072131042e-06,
+      "loss": 1.34060173,
+      "memory(GiB)": 103.58,
+      "step": 17865,
+      "train_speed(iter/s)": 1.321088
+    },
+    {
+      "acc": 0.68540583,
+      "epoch": 0.45332318619989853,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.170742455207378e-06,
+      "loss": 1.39119968,
+      "memory(GiB)": 103.58,
+      "step": 17870,
+      "train_speed(iter/s)": 1.321127
+    },
+    {
+      "acc": 0.68525386,
+      "epoch": 0.4534500253678336,
+      "grad_norm": 4.0,
+      "learning_rate": 9.170164005656465e-06,
+      "loss": 1.39620886,
+      "memory(GiB)": 103.58,
+      "step": 17875,
+      "train_speed(iter/s)": 1.321164
+    },
+    {
+      "acc": 0.70204339,
+      "epoch": 0.4535768645357686,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.169585372683118e-06,
+      "loss": 1.29525557,
+      "memory(GiB)": 103.58,
+      "step": 17880,
+      "train_speed(iter/s)": 1.321203
+    },
+    {
+      "acc": 0.69329534,
+      "epoch": 0.4537037037037037,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.169006556312794e-06,
+      "loss": 1.38219452,
+      "memory(GiB)": 103.58,
+      "step": 17885,
+      "train_speed(iter/s)": 1.321243
+    },
+    {
+      "acc": 0.67504816,
+      "epoch": 0.45383054287163876,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.168427556570946e-06,
+      "loss": 1.45182295,
+      "memory(GiB)": 103.58,
+      "step": 17890,
+      "train_speed(iter/s)": 1.321283
+    },
+    {
+      "acc": 0.69706478,
+      "epoch": 0.4539573820395738,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.167848373483044e-06,
+      "loss": 1.39305019,
+      "memory(GiB)": 103.58,
+      "step": 17895,
+      "train_speed(iter/s)": 1.321325
+    },
+    {
+      "acc": 0.68144464,
+      "epoch": 0.45408422120750885,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.167269007074561e-06,
+      "loss": 1.43414545,
+      "memory(GiB)": 103.58,
+      "step": 17900,
+      "train_speed(iter/s)": 1.32136
+    },
+    {
+      "acc": 0.68458014,
+      "epoch": 0.45421106037544395,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.166689457370983e-06,
+      "loss": 1.40844975,
+      "memory(GiB)": 103.58,
+      "step": 17905,
+      "train_speed(iter/s)": 1.321401
+    },
+    {
+      "acc": 0.66466703,
+      "epoch": 0.454337899543379,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.166109724397801e-06,
+      "loss": 1.43098469,
+      "memory(GiB)": 103.58,
+      "step": 17910,
+      "train_speed(iter/s)": 1.321442
+    },
+    {
+      "acc": 0.68783321,
+      "epoch": 0.45446473871131404,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.165529808180511e-06,
+      "loss": 1.37401772,
+      "memory(GiB)": 103.58,
+      "step": 17915,
+      "train_speed(iter/s)": 1.321484
+    },
+    {
+      "acc": 0.68451595,
+      "epoch": 0.4545915778792491,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.164949708744622e-06,
+      "loss": 1.40320988,
+      "memory(GiB)": 103.58,
+      "step": 17920,
+      "train_speed(iter/s)": 1.321523
+    },
+    {
+      "acc": 0.68885412,
+      "epoch": 0.4547184170471842,
+      "grad_norm": 4.125,
+      "learning_rate": 9.164369426115652e-06,
+      "loss": 1.3587925,
+      "memory(GiB)": 103.58,
+      "step": 17925,
+      "train_speed(iter/s)": 1.321562
+    },
+    {
+      "acc": 0.68035097,
+      "epoch": 0.45484525621511923,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.16378896031912e-06,
+      "loss": 1.44144917,
+      "memory(GiB)": 103.58,
+      "step": 17930,
+      "train_speed(iter/s)": 1.321604
+    },
+    {
+      "acc": 0.68712769,
+      "epoch": 0.4549720953830543,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.163208311380561e-06,
+      "loss": 1.38010731,
+      "memory(GiB)": 103.58,
+      "step": 17935,
+      "train_speed(iter/s)": 1.321645
+    },
+    {
+      "acc": 0.68799658,
+      "epoch": 0.4550989345509893,
+      "grad_norm": 2.75,
+      "learning_rate": 9.162627479325511e-06,
+      "loss": 1.38739624,
+      "memory(GiB)": 103.58,
+      "step": 17940,
+      "train_speed(iter/s)": 1.321684
+    },
+    {
+      "acc": 0.69453111,
+      "epoch": 0.4552257737189244,
+      "grad_norm": 3.5,
+      "learning_rate": 9.16204646417952e-06,
+      "loss": 1.39729853,
+      "memory(GiB)": 103.58,
+      "step": 17945,
+      "train_speed(iter/s)": 1.321725
+    },
+    {
+      "acc": 0.68631268,
+      "epoch": 0.45535261288685946,
+      "grad_norm": 3.0,
+      "learning_rate": 9.161465265968145e-06,
+      "loss": 1.35300159,
+      "memory(GiB)": 103.58,
+      "step": 17950,
+      "train_speed(iter/s)": 1.321764
+    },
+    {
+      "acc": 0.70059166,
+      "epoch": 0.4554794520547945,
+      "grad_norm": 3.625,
+      "learning_rate": 9.160883884716948e-06,
+      "loss": 1.34310551,
+      "memory(GiB)": 103.58,
+      "step": 17955,
+      "train_speed(iter/s)": 1.321805
+    },
+    {
+      "acc": 0.68287349,
+      "epoch": 0.45560629122272955,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.1603023204515e-06,
+      "loss": 1.43561049,
+      "memory(GiB)": 103.58,
+      "step": 17960,
+      "train_speed(iter/s)": 1.321847
+    },
+    {
+      "acc": 0.67481918,
+      "epoch": 0.45573313039066465,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.15972057319738e-06,
+      "loss": 1.39595337,
+      "memory(GiB)": 103.58,
+      "step": 17965,
+      "train_speed(iter/s)": 1.321886
+    },
+    {
+      "acc": 0.69389424,
+      "epoch": 0.4558599695585997,
+      "grad_norm": 4.25,
+      "learning_rate": 9.159138642980178e-06,
+      "loss": 1.4195528,
+      "memory(GiB)": 103.58,
+      "step": 17970,
+      "train_speed(iter/s)": 1.321927
+    },
+    {
+      "acc": 0.68090267,
+      "epoch": 0.45598680872653474,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.15855652982549e-06,
+      "loss": 1.45416317,
+      "memory(GiB)": 103.58,
+      "step": 17975,
+      "train_speed(iter/s)": 1.321967
+    },
+    {
+      "acc": 0.69318781,
+      "epoch": 0.4561136478944698,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.15797423375892e-06,
+      "loss": 1.40834732,
+      "memory(GiB)": 103.58,
+      "step": 17980,
+      "train_speed(iter/s)": 1.322008
+    },
+    {
+      "acc": 0.68651714,
+      "epoch": 0.4562404870624049,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.157391754806079e-06,
+      "loss": 1.42432823,
+      "memory(GiB)": 103.58,
+      "step": 17985,
+      "train_speed(iter/s)": 1.322049
+    },
+    {
+      "acc": 0.6920042,
+      "epoch": 0.45636732623033993,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.156809092992588e-06,
+      "loss": 1.39915009,
+      "memory(GiB)": 103.58,
+      "step": 17990,
+      "train_speed(iter/s)": 1.32209
+    },
+    {
+      "acc": 0.68187008,
+      "epoch": 0.456494165398275,
+      "grad_norm": 3.625,
+      "learning_rate": 9.156226248344072e-06,
+      "loss": 1.39225407,
+      "memory(GiB)": 103.58,
+      "step": 17995,
+      "train_speed(iter/s)": 1.32213
+    },
+    {
+      "acc": 0.6952776,
+      "epoch": 0.45662100456621,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.15564322088617e-06,
+      "loss": 1.38610535,
+      "memory(GiB)": 103.58,
+      "step": 18000,
+      "train_speed(iter/s)": 1.32217
+    },
+    {
+      "epoch": 0.45662100456621,
+      "eval_acc": 0.6729228141177847,
+      "eval_loss": 1.368478536605835,
+      "eval_runtime": 70.0784,
+      "eval_samples_per_second": 90.898,
+      "eval_steps_per_second": 22.732,
+      "step": 18000
+    },
+    {
+      "acc": 0.68472939,
+      "epoch": 0.4567478437341451,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.155060010644525e-06,
+      "loss": 1.41576557,
+      "memory(GiB)": 103.58,
+      "step": 18005,
+      "train_speed(iter/s)": 1.314226
+    },
+    {
+      "acc": 0.69394345,
+      "epoch": 0.45687468290208016,
+      "grad_norm": 4.125,
+      "learning_rate": 9.154476617644792e-06,
+      "loss": 1.35836878,
+      "memory(GiB)": 103.58,
+      "step": 18010,
+      "train_speed(iter/s)": 1.314267
+    },
+    {
+      "acc": 0.68868647,
+      "epoch": 0.4570015220700152,
+      "grad_norm": 5.21875,
+      "learning_rate": 9.153893041912627e-06,
+      "loss": 1.34905281,
+      "memory(GiB)": 103.58,
+      "step": 18015,
+      "train_speed(iter/s)": 1.314309
+    },
+    {
+      "acc": 0.69638824,
+      "epoch": 0.45712836123795025,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.1533092834737e-06,
+      "loss": 1.36347847,
+      "memory(GiB)": 103.58,
+      "step": 18020,
+      "train_speed(iter/s)": 1.314351
+    },
+    {
+      "acc": 0.67651405,
+      "epoch": 0.45725520040588535,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.152725342353688e-06,
+      "loss": 1.43596134,
+      "memory(GiB)": 103.58,
+      "step": 18025,
+      "train_speed(iter/s)": 1.314393
+    },
+    {
+      "acc": 0.67475224,
+      "epoch": 0.4573820395738204,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.152141218578276e-06,
+      "loss": 1.46595612,
+      "memory(GiB)": 103.58,
+      "step": 18030,
+      "train_speed(iter/s)": 1.314426
+    },
+    {
+      "acc": 0.66989899,
+      "epoch": 0.45750887874175544,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.151556912173154e-06,
+      "loss": 1.48040333,
+      "memory(GiB)": 103.58,
+      "step": 18035,
+      "train_speed(iter/s)": 1.314465
+    },
+    {
+      "acc": 0.68059087,
+      "epoch": 0.4576357179096905,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.150972423164024e-06,
+      "loss": 1.40852718,
+      "memory(GiB)": 103.58,
+      "step": 18040,
+      "train_speed(iter/s)": 1.314507
+    },
+    {
+      "acc": 0.68294024,
+      "epoch": 0.4577625570776256,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.150387751576594e-06,
+      "loss": 1.45418606,
+      "memory(GiB)": 103.58,
+      "step": 18045,
+      "train_speed(iter/s)": 1.314549
+    },
+    {
+      "acc": 0.67627115,
+      "epoch": 0.45788939624556063,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.14980289743658e-06,
+      "loss": 1.41898556,
+      "memory(GiB)": 103.58,
+      "step": 18050,
+      "train_speed(iter/s)": 1.31459
+    },
+    {
+      "acc": 0.67407637,
+      "epoch": 0.4580162354134957,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.149217860769708e-06,
+      "loss": 1.45032806,
+      "memory(GiB)": 103.58,
+      "step": 18055,
+      "train_speed(iter/s)": 1.314631
+    },
+    {
+      "acc": 0.69466133,
+      "epoch": 0.4581430745814307,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.14863264160171e-06,
+      "loss": 1.36533165,
+      "memory(GiB)": 103.58,
+      "step": 18060,
+      "train_speed(iter/s)": 1.314673
+    },
+    {
+      "acc": 0.67270641,
+      "epoch": 0.4582699137493658,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.148047239958325e-06,
+      "loss": 1.48234005,
+      "memory(GiB)": 103.58,
+      "step": 18065,
+      "train_speed(iter/s)": 1.314716
+    },
+    {
+      "acc": 0.67643428,
+      "epoch": 0.45839675291730086,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.147461655865302e-06,
+      "loss": 1.41364489,
+      "memory(GiB)": 103.58,
+      "step": 18070,
+      "train_speed(iter/s)": 1.314757
+    },
+    {
+      "acc": 0.6906045,
+      "epoch": 0.4585235920852359,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.1468758893484e-06,
+      "loss": 1.43159227,
+      "memory(GiB)": 103.58,
+      "step": 18075,
+      "train_speed(iter/s)": 1.314799
+    },
+    {
+      "acc": 0.69659758,
+      "epoch": 0.45865043125317095,
+      "grad_norm": 2.75,
+      "learning_rate": 9.14628994043338e-06,
+      "loss": 1.43379135,
+      "memory(GiB)": 103.58,
+      "step": 18080,
+      "train_speed(iter/s)": 1.31484
+    },
+    {
+      "acc": 0.68272381,
+      "epoch": 0.45877727042110605,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.145703809146018e-06,
+      "loss": 1.41149673,
+      "memory(GiB)": 103.58,
+      "step": 18085,
+      "train_speed(iter/s)": 1.31488
+    },
+    {
+      "acc": 0.67127776,
+      "epoch": 0.4589041095890411,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.145117495512092e-06,
+      "loss": 1.42871103,
+      "memory(GiB)": 103.58,
+      "step": 18090,
+      "train_speed(iter/s)": 1.314924
+    },
+    {
+      "acc": 0.6839468,
+      "epoch": 0.45903094875697614,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.144530999557393e-06,
+      "loss": 1.48849106,
+      "memory(GiB)": 103.58,
+      "step": 18095,
+      "train_speed(iter/s)": 1.314967
+    },
+    {
+      "acc": 0.68413982,
+      "epoch": 0.4591577879249112,
+      "grad_norm": 3.75,
+      "learning_rate": 9.143944321307718e-06,
+      "loss": 1.41560173,
+      "memory(GiB)": 103.58,
+      "step": 18100,
+      "train_speed(iter/s)": 1.315009
+    },
+    {
+      "acc": 0.68047142,
+      "epoch": 0.4592846270928463,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.143357460788868e-06,
+      "loss": 1.43898869,
+      "memory(GiB)": 103.58,
+      "step": 18105,
+      "train_speed(iter/s)": 1.315049
+    },
+    {
+      "acc": 0.67363043,
+      "epoch": 0.45941146626078133,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.14277041802666e-06,
+      "loss": 1.49375095,
+      "memory(GiB)": 103.58,
+      "step": 18110,
+      "train_speed(iter/s)": 1.315091
+    },
+    {
+      "acc": 0.68517532,
+      "epoch": 0.4595383054287164,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.142183193046912e-06,
+      "loss": 1.39226198,
+      "memory(GiB)": 103.58,
+      "step": 18115,
+      "train_speed(iter/s)": 1.315132
+    },
+    {
+      "acc": 0.68850737,
+      "epoch": 0.4596651445966514,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.141595785875453e-06,
+      "loss": 1.32751236,
+      "memory(GiB)": 103.58,
+      "step": 18120,
+      "train_speed(iter/s)": 1.315173
+    },
+    {
+      "acc": 0.68520603,
+      "epoch": 0.4597919837645865,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.141008196538122e-06,
+      "loss": 1.38165731,
+      "memory(GiB)": 103.58,
+      "step": 18125,
+      "train_speed(iter/s)": 1.315217
+    },
+    {
+      "acc": 0.67946949,
+      "epoch": 0.45991882293252156,
+      "grad_norm": 3.125,
+      "learning_rate": 9.14042042506076e-06,
+      "loss": 1.42389126,
+      "memory(GiB)": 103.58,
+      "step": 18130,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.68872051,
+      "epoch": 0.4600456621004566,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.139832471469224e-06,
+      "loss": 1.40765018,
+      "memory(GiB)": 103.58,
+      "step": 18135,
+      "train_speed(iter/s)": 1.315301
+    },
+    {
+      "acc": 0.69110374,
+      "epoch": 0.46017250126839165,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.13924433578937e-06,
+      "loss": 1.38906717,
+      "memory(GiB)": 103.58,
+      "step": 18140,
+      "train_speed(iter/s)": 1.315342
+    },
+    {
+      "acc": 0.68910761,
+      "epoch": 0.46029934043632675,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.138656018047074e-06,
+      "loss": 1.39672127,
+      "memory(GiB)": 103.58,
+      "step": 18145,
+      "train_speed(iter/s)": 1.315385
+    },
+    {
+      "acc": 0.69107046,
+      "epoch": 0.4604261796042618,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.138067518268206e-06,
+      "loss": 1.39255447,
+      "memory(GiB)": 103.58,
+      "step": 18150,
+      "train_speed(iter/s)": 1.315422
+    },
+    {
+      "acc": 0.69321299,
+      "epoch": 0.46055301877219684,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.137478836478654e-06,
+      "loss": 1.37264481,
+      "memory(GiB)": 103.58,
+      "step": 18155,
+      "train_speed(iter/s)": 1.315464
+    },
+    {
+      "acc": 0.687293,
+      "epoch": 0.4606798579401319,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.13688997270431e-06,
+      "loss": 1.38487558,
+      "memory(GiB)": 103.58,
+      "step": 18160,
+      "train_speed(iter/s)": 1.315506
+    },
+    {
+      "acc": 0.71243944,
+      "epoch": 0.460806697108067,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.136300926971076e-06,
+      "loss": 1.32473412,
+      "memory(GiB)": 103.58,
+      "step": 18165,
+      "train_speed(iter/s)": 1.315547
+    },
+    {
+      "acc": 0.68727808,
+      "epoch": 0.46093353627600203,
+      "grad_norm": 2.75,
+      "learning_rate": 9.135711699304858e-06,
+      "loss": 1.46056938,
+      "memory(GiB)": 103.58,
+      "step": 18170,
+      "train_speed(iter/s)": 1.315574
+    },
+    {
+      "acc": 0.67352161,
+      "epoch": 0.4610603754439371,
+      "grad_norm": 3.0,
+      "learning_rate": 9.135122289731575e-06,
+      "loss": 1.40830498,
+      "memory(GiB)": 103.58,
+      "step": 18175,
+      "train_speed(iter/s)": 1.315618
+    },
+    {
+      "acc": 0.68101525,
+      "epoch": 0.4611872146118721,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.134532698277154e-06,
+      "loss": 1.36121788,
+      "memory(GiB)": 103.58,
+      "step": 18180,
+      "train_speed(iter/s)": 1.315661
+    },
+    {
+      "acc": 0.68063078,
+      "epoch": 0.4613140537798072,
+      "grad_norm": 3.875,
+      "learning_rate": 9.133942924967524e-06,
+      "loss": 1.42952023,
+      "memory(GiB)": 103.58,
+      "step": 18185,
+      "train_speed(iter/s)": 1.315704
+    },
+    {
+      "acc": 0.69900689,
+      "epoch": 0.46144089294774226,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.133352969828628e-06,
+      "loss": 1.35466957,
+      "memory(GiB)": 103.58,
+      "step": 18190,
+      "train_speed(iter/s)": 1.315745
+    },
+    {
+      "acc": 0.68315015,
+      "epoch": 0.4615677321156773,
+      "grad_norm": 2.875,
+      "learning_rate": 9.132762832886416e-06,
+      "loss": 1.36158524,
+      "memory(GiB)": 103.58,
+      "step": 18195,
+      "train_speed(iter/s)": 1.315787
+    },
+    {
+      "acc": 0.68439093,
+      "epoch": 0.46169457128361235,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.13217251416684e-06,
+      "loss": 1.45373249,
+      "memory(GiB)": 103.58,
+      "step": 18200,
+      "train_speed(iter/s)": 1.315828
+    },
+    {
+      "acc": 0.68878069,
+      "epoch": 0.46182141045154745,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.131582013695867e-06,
+      "loss": 1.40632296,
+      "memory(GiB)": 103.58,
+      "step": 18205,
+      "train_speed(iter/s)": 1.31587
+    },
+    {
+      "acc": 0.69749541,
+      "epoch": 0.4619482496194825,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.130991331499474e-06,
+      "loss": 1.43480511,
+      "memory(GiB)": 103.58,
+      "step": 18210,
+      "train_speed(iter/s)": 1.315914
+    },
+    {
+      "acc": 0.70287976,
+      "epoch": 0.46207508878741754,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.130400467603637e-06,
+      "loss": 1.32913818,
+      "memory(GiB)": 103.58,
+      "step": 18215,
+      "train_speed(iter/s)": 1.315956
+    },
+    {
+      "acc": 0.67182007,
+      "epoch": 0.4622019279553526,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.129809422034349e-06,
+      "loss": 1.439077,
+      "memory(GiB)": 103.58,
+      "step": 18220,
+      "train_speed(iter/s)": 1.315998
+    },
+    {
+      "acc": 0.68370657,
+      "epoch": 0.4623287671232877,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.129218194817601e-06,
+      "loss": 1.46313591,
+      "memory(GiB)": 103.58,
+      "step": 18225,
+      "train_speed(iter/s)": 1.316039
+    },
+    {
+      "acc": 0.67893887,
+      "epoch": 0.46245560629122273,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.128626785979404e-06,
+      "loss": 1.47171612,
+      "memory(GiB)": 103.58,
+      "step": 18230,
+      "train_speed(iter/s)": 1.316082
+    },
+    {
+      "acc": 0.67823648,
+      "epoch": 0.4625824454591578,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.128035195545766e-06,
+      "loss": 1.42482481,
+      "memory(GiB)": 103.58,
+      "step": 18235,
+      "train_speed(iter/s)": 1.316123
+    },
+    {
+      "acc": 0.66764612,
+      "epoch": 0.4627092846270928,
+      "grad_norm": 2.875,
+      "learning_rate": 9.12744342354271e-06,
+      "loss": 1.53531065,
+      "memory(GiB)": 103.58,
+      "step": 18240,
+      "train_speed(iter/s)": 1.316164
+    },
+    {
+      "acc": 0.68743882,
+      "epoch": 0.4628361237950279,
+      "grad_norm": 3.375,
+      "learning_rate": 9.126851469996265e-06,
+      "loss": 1.38940048,
+      "memory(GiB)": 103.58,
+      "step": 18245,
+      "train_speed(iter/s)": 1.316207
+    },
+    {
+      "acc": 0.68202047,
+      "epoch": 0.46296296296296297,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.126259334932467e-06,
+      "loss": 1.39646988,
+      "memory(GiB)": 103.58,
+      "step": 18250,
+      "train_speed(iter/s)": 1.31625
+    },
+    {
+      "acc": 0.66930833,
+      "epoch": 0.463089802130898,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.125667018377362e-06,
+      "loss": 1.47779512,
+      "memory(GiB)": 103.58,
+      "step": 18255,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.68491073,
+      "epoch": 0.46321664129883305,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.125074520357002e-06,
+      "loss": 1.41021652,
+      "memory(GiB)": 103.58,
+      "step": 18260,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.6837913,
+      "epoch": 0.46334348046676815,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.124481840897446e-06,
+      "loss": 1.3876379,
+      "memory(GiB)": 103.58,
+      "step": 18265,
+      "train_speed(iter/s)": 1.316377
+    },
+    {
+      "acc": 0.6808753,
+      "epoch": 0.4634703196347032,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.123888980024765e-06,
+      "loss": 1.35541573,
+      "memory(GiB)": 103.58,
+      "step": 18270,
+      "train_speed(iter/s)": 1.316417
+    },
+    {
+      "acc": 0.7033545,
+      "epoch": 0.46359715880263824,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.123295937765034e-06,
+      "loss": 1.35950756,
+      "memory(GiB)": 103.58,
+      "step": 18275,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.71100082,
+      "epoch": 0.4637239979705733,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.122702714144339e-06,
+      "loss": 1.33684273,
+      "memory(GiB)": 103.58,
+      "step": 18280,
+      "train_speed(iter/s)": 1.3165
+    },
+    {
+      "acc": 0.68099213,
+      "epoch": 0.4638508371385084,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.122109309188772e-06,
+      "loss": 1.39773197,
+      "memory(GiB)": 103.58,
+      "step": 18285,
+      "train_speed(iter/s)": 1.316542
+    },
+    {
+      "acc": 0.70946617,
+      "epoch": 0.46397767630644343,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.121515722924435e-06,
+      "loss": 1.35177832,
+      "memory(GiB)": 103.58,
+      "step": 18290,
+      "train_speed(iter/s)": 1.316584
+    },
+    {
+      "acc": 0.69465866,
+      "epoch": 0.4641045154743785,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.120921955377433e-06,
+      "loss": 1.36782513,
+      "memory(GiB)": 103.58,
+      "step": 18295,
+      "train_speed(iter/s)": 1.316626
+    },
+    {
+      "acc": 0.6745595,
+      "epoch": 0.4642313546423135,
+      "grad_norm": 3.0,
+      "learning_rate": 9.120328006573887e-06,
+      "loss": 1.40926113,
+      "memory(GiB)": 103.58,
+      "step": 18300,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.68727274,
+      "epoch": 0.4643581938102486,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.119733876539916e-06,
+      "loss": 1.35910778,
+      "memory(GiB)": 103.58,
+      "step": 18305,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.68775678,
+      "epoch": 0.46448503297818367,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.119139565301658e-06,
+      "loss": 1.37626896,
+      "memory(GiB)": 103.58,
+      "step": 18310,
+      "train_speed(iter/s)": 1.316747
+    },
+    {
+      "acc": 0.68176732,
+      "epoch": 0.4646118721461187,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.118545072885253e-06,
+      "loss": 1.44852676,
+      "memory(GiB)": 103.58,
+      "step": 18315,
+      "train_speed(iter/s)": 1.31679
+    },
+    {
+      "acc": 0.69190111,
+      "epoch": 0.46473871131405375,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.117950399316845e-06,
+      "loss": 1.36149416,
+      "memory(GiB)": 103.58,
+      "step": 18320,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.69213881,
+      "epoch": 0.46486555048198885,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.117355544622595e-06,
+      "loss": 1.37450428,
+      "memory(GiB)": 103.58,
+      "step": 18325,
+      "train_speed(iter/s)": 1.316875
+    },
+    {
+      "acc": 0.67863216,
+      "epoch": 0.4649923896499239,
+      "grad_norm": 2.875,
+      "learning_rate": 9.116760508828664e-06,
+      "loss": 1.35202198,
+      "memory(GiB)": 103.58,
+      "step": 18330,
+      "train_speed(iter/s)": 1.316917
+    },
+    {
+      "acc": 0.70329032,
+      "epoch": 0.46511922881785894,
+      "grad_norm": 2.875,
+      "learning_rate": 9.116165291961225e-06,
+      "loss": 1.37371235,
+      "memory(GiB)": 103.58,
+      "step": 18335,
+      "train_speed(iter/s)": 1.316959
+    },
+    {
+      "acc": 0.68390026,
+      "epoch": 0.465246067985794,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.11556989404646e-06,
+      "loss": 1.37284718,
+      "memory(GiB)": 103.58,
+      "step": 18340,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.6743804,
+      "epoch": 0.4653729071537291,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.114974315110558e-06,
+      "loss": 1.47121239,
+      "memory(GiB)": 103.58,
+      "step": 18345,
+      "train_speed(iter/s)": 1.317042
+    },
+    {
+      "acc": 0.69107099,
+      "epoch": 0.46549974632166413,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.114378555179712e-06,
+      "loss": 1.41305428,
+      "memory(GiB)": 103.58,
+      "step": 18350,
+      "train_speed(iter/s)": 1.31708
+    },
+    {
+      "acc": 0.69406343,
+      "epoch": 0.4656265854895992,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.11378261428013e-06,
+      "loss": 1.35997124,
+      "memory(GiB)": 103.58,
+      "step": 18355,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.6858036,
+      "epoch": 0.4657534246575342,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.11318649243802e-06,
+      "loss": 1.39628859,
+      "memory(GiB)": 103.58,
+      "step": 18360,
+      "train_speed(iter/s)": 1.31716
+    },
+    {
+      "acc": 0.69376717,
+      "epoch": 0.4658802638254693,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.112590189679604e-06,
+      "loss": 1.37847042,
+      "memory(GiB)": 103.58,
+      "step": 18365,
+      "train_speed(iter/s)": 1.317201
+    },
+    {
+      "acc": 0.68246555,
+      "epoch": 0.46600710299340437,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.111993706031109e-06,
+      "loss": 1.37873106,
+      "memory(GiB)": 103.58,
+      "step": 18370,
+      "train_speed(iter/s)": 1.317241
+    },
+    {
+      "acc": 0.69245925,
+      "epoch": 0.4661339421613394,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.111397041518774e-06,
+      "loss": 1.396661,
+      "memory(GiB)": 103.58,
+      "step": 18375,
+      "train_speed(iter/s)": 1.317283
+    },
+    {
+      "acc": 0.6860919,
+      "epoch": 0.46626078132927445,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.11080019616884e-06,
+      "loss": 1.36356554,
+      "memory(GiB)": 103.58,
+      "step": 18380,
+      "train_speed(iter/s)": 1.317325
+    },
+    {
+      "acc": 0.69591112,
+      "epoch": 0.46638762049720955,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.11020317000756e-06,
+      "loss": 1.41943064,
+      "memory(GiB)": 103.58,
+      "step": 18385,
+      "train_speed(iter/s)": 1.317362
+    },
+    {
+      "acc": 0.65961704,
+      "epoch": 0.4665144596651446,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.109605963061194e-06,
+      "loss": 1.50562735,
+      "memory(GiB)": 103.58,
+      "step": 18390,
+      "train_speed(iter/s)": 1.317397
+    },
+    {
+      "acc": 0.69818568,
+      "epoch": 0.46664129883307964,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.10900857535601e-06,
+      "loss": 1.36371202,
+      "memory(GiB)": 103.58,
+      "step": 18395,
+      "train_speed(iter/s)": 1.317439
+    },
+    {
+      "acc": 0.6728374,
+      "epoch": 0.4667681380010147,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.108411006918283e-06,
+      "loss": 1.45886402,
+      "memory(GiB)": 103.58,
+      "step": 18400,
+      "train_speed(iter/s)": 1.31748
+    },
+    {
+      "acc": 0.68950768,
+      "epoch": 0.4668949771689498,
+      "grad_norm": 3.75,
+      "learning_rate": 9.107813257774298e-06,
+      "loss": 1.41337223,
+      "memory(GiB)": 103.58,
+      "step": 18405,
+      "train_speed(iter/s)": 1.317522
+    },
+    {
+      "acc": 0.67424383,
+      "epoch": 0.46702181633688483,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.107215327950345e-06,
+      "loss": 1.41721973,
+      "memory(GiB)": 103.58,
+      "step": 18410,
+      "train_speed(iter/s)": 1.317563
+    },
+    {
+      "acc": 0.69026403,
+      "epoch": 0.4671486555048199,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.106617217472724e-06,
+      "loss": 1.40888786,
+      "memory(GiB)": 103.58,
+      "step": 18415,
+      "train_speed(iter/s)": 1.317606
+    },
+    {
+      "acc": 0.67638292,
+      "epoch": 0.4672754946727549,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.106018926367744e-06,
+      "loss": 1.43350582,
+      "memory(GiB)": 103.58,
+      "step": 18420,
+      "train_speed(iter/s)": 1.317649
+    },
+    {
+      "acc": 0.6839592,
+      "epoch": 0.46740233384069,
+      "grad_norm": 3.0,
+      "learning_rate": 9.10542045466172e-06,
+      "loss": 1.42158909,
+      "memory(GiB)": 103.58,
+      "step": 18425,
+      "train_speed(iter/s)": 1.317689
+    },
+    {
+      "acc": 0.69008794,
+      "epoch": 0.46752917300862507,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.104821802380974e-06,
+      "loss": 1.44977474,
+      "memory(GiB)": 103.58,
+      "step": 18430,
+      "train_speed(iter/s)": 1.317729
+    },
+    {
+      "acc": 0.69120216,
+      "epoch": 0.4676560121765601,
+      "grad_norm": 2.75,
+      "learning_rate": 9.104222969551838e-06,
+      "loss": 1.32885303,
+      "memory(GiB)": 103.58,
+      "step": 18435,
+      "train_speed(iter/s)": 1.31777
+    },
+    {
+      "acc": 0.68762751,
+      "epoch": 0.46778285134449515,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.103623956200654e-06,
+      "loss": 1.41555758,
+      "memory(GiB)": 103.58,
+      "step": 18440,
+      "train_speed(iter/s)": 1.317811
+    },
+    {
+      "acc": 0.67751966,
+      "epoch": 0.46790969051243025,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.103024762353766e-06,
+      "loss": 1.48271141,
+      "memory(GiB)": 103.58,
+      "step": 18445,
+      "train_speed(iter/s)": 1.317851
+    },
+    {
+      "acc": 0.69722805,
+      "epoch": 0.4680365296803653,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.102425388037527e-06,
+      "loss": 1.36015453,
+      "memory(GiB)": 103.58,
+      "step": 18450,
+      "train_speed(iter/s)": 1.31789
+    },
+    {
+      "acc": 0.68650341,
+      "epoch": 0.46816336884830034,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.101825833278308e-06,
+      "loss": 1.40113096,
+      "memory(GiB)": 103.58,
+      "step": 18455,
+      "train_speed(iter/s)": 1.317931
+    },
+    {
+      "acc": 0.69386182,
+      "epoch": 0.4682902080162354,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.101226098102473e-06,
+      "loss": 1.35056343,
+      "memory(GiB)": 103.58,
+      "step": 18460,
+      "train_speed(iter/s)": 1.317971
+    },
+    {
+      "acc": 0.69422131,
+      "epoch": 0.4684170471841705,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.100626182536405e-06,
+      "loss": 1.42733221,
+      "memory(GiB)": 103.58,
+      "step": 18465,
+      "train_speed(iter/s)": 1.318008
+    },
+    {
+      "acc": 0.69250059,
+      "epoch": 0.46854388635210553,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.100026086606488e-06,
+      "loss": 1.37561884,
+      "memory(GiB)": 103.58,
+      "step": 18470,
+      "train_speed(iter/s)": 1.318049
+    },
+    {
+      "acc": 0.70774083,
+      "epoch": 0.4686707255200406,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.09942581033912e-06,
+      "loss": 1.32776794,
+      "memory(GiB)": 103.58,
+      "step": 18475,
+      "train_speed(iter/s)": 1.318088
+    },
+    {
+      "acc": 0.70238132,
+      "epoch": 0.4687975646879756,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.0988253537607e-06,
+      "loss": 1.30904961,
+      "memory(GiB)": 103.58,
+      "step": 18480,
+      "train_speed(iter/s)": 1.318127
+    },
+    {
+      "acc": 0.69367108,
+      "epoch": 0.4689244038559107,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.098224716897644e-06,
+      "loss": 1.37608337,
+      "memory(GiB)": 103.58,
+      "step": 18485,
+      "train_speed(iter/s)": 1.318163
+    },
+    {
+      "acc": 0.69572258,
+      "epoch": 0.46905124302384577,
+      "grad_norm": 3.0,
+      "learning_rate": 9.097623899776366e-06,
+      "loss": 1.3634984,
+      "memory(GiB)": 103.58,
+      "step": 18490,
+      "train_speed(iter/s)": 1.318203
+    },
+    {
+      "acc": 0.67410069,
+      "epoch": 0.4691780821917808,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.097022902423294e-06,
+      "loss": 1.38468685,
+      "memory(GiB)": 103.58,
+      "step": 18495,
+      "train_speed(iter/s)": 1.318242
+    },
+    {
+      "acc": 0.68119469,
+      "epoch": 0.46930492135971585,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.096421724864864e-06,
+      "loss": 1.41594791,
+      "memory(GiB)": 103.58,
+      "step": 18500,
+      "train_speed(iter/s)": 1.31828
+    },
+    {
+      "acc": 0.68359919,
+      "epoch": 0.46943176052765095,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.095820367127517e-06,
+      "loss": 1.42956219,
+      "memory(GiB)": 103.58,
+      "step": 18505,
+      "train_speed(iter/s)": 1.318314
+    },
+    {
+      "acc": 0.67483616,
+      "epoch": 0.469558599695586,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.095218829237703e-06,
+      "loss": 1.44184561,
+      "memory(GiB)": 103.58,
+      "step": 18510,
+      "train_speed(iter/s)": 1.318353
+    },
+    {
+      "acc": 0.67478166,
+      "epoch": 0.46968543886352104,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.094617111221881e-06,
+      "loss": 1.46232662,
+      "memory(GiB)": 103.58,
+      "step": 18515,
+      "train_speed(iter/s)": 1.318391
+    },
+    {
+      "acc": 0.67986712,
+      "epoch": 0.4698122780314561,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.09401521310652e-06,
+      "loss": 1.36119757,
+      "memory(GiB)": 103.58,
+      "step": 18520,
+      "train_speed(iter/s)": 1.318431
+    },
+    {
+      "acc": 0.69663458,
+      "epoch": 0.4699391171993912,
+      "grad_norm": 2.875,
+      "learning_rate": 9.093413134918088e-06,
+      "loss": 1.35447674,
+      "memory(GiB)": 103.58,
+      "step": 18525,
+      "train_speed(iter/s)": 1.31847
+    },
+    {
+      "acc": 0.69073792,
+      "epoch": 0.47006595636732623,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.09281087668307e-06,
+      "loss": 1.39803753,
+      "memory(GiB)": 103.58,
+      "step": 18530,
+      "train_speed(iter/s)": 1.318509
+    },
+    {
+      "acc": 0.69893723,
+      "epoch": 0.4701927955352613,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.09220843842796e-06,
+      "loss": 1.35723801,
+      "memory(GiB)": 103.58,
+      "step": 18535,
+      "train_speed(iter/s)": 1.318547
+    },
+    {
+      "acc": 0.68323641,
+      "epoch": 0.4703196347031963,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.09160582017925e-06,
+      "loss": 1.38463316,
+      "memory(GiB)": 103.58,
+      "step": 18540,
+      "train_speed(iter/s)": 1.318585
+    },
+    {
+      "acc": 0.66238108,
+      "epoch": 0.4704464738711314,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.091003021963449e-06,
+      "loss": 1.47829838,
+      "memory(GiB)": 103.58,
+      "step": 18545,
+      "train_speed(iter/s)": 1.318623
+    },
+    {
+      "acc": 0.67937622,
+      "epoch": 0.47057331303906647,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.09040004380707e-06,
+      "loss": 1.43221931,
+      "memory(GiB)": 103.58,
+      "step": 18550,
+      "train_speed(iter/s)": 1.318659
+    },
+    {
+      "acc": 0.67707243,
+      "epoch": 0.4707001522070015,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.089796885736637e-06,
+      "loss": 1.37110233,
+      "memory(GiB)": 103.58,
+      "step": 18555,
+      "train_speed(iter/s)": 1.318697
+    },
+    {
+      "acc": 0.69709063,
+      "epoch": 0.47082699137493655,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.089193547778674e-06,
+      "loss": 1.37416563,
+      "memory(GiB)": 103.58,
+      "step": 18560,
+      "train_speed(iter/s)": 1.318726
+    },
+    {
+      "acc": 0.68469729,
+      "epoch": 0.47095383054287165,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.088590029959724e-06,
+      "loss": 1.3847208,
+      "memory(GiB)": 103.58,
+      "step": 18565,
+      "train_speed(iter/s)": 1.318764
+    },
+    {
+      "acc": 0.67783298,
+      "epoch": 0.4710806697108067,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.08798633230633e-06,
+      "loss": 1.50742054,
+      "memory(GiB)": 103.58,
+      "step": 18570,
+      "train_speed(iter/s)": 1.318803
+    },
+    {
+      "acc": 0.69248462,
+      "epoch": 0.47120750887874174,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.087382454845044e-06,
+      "loss": 1.46009903,
+      "memory(GiB)": 103.58,
+      "step": 18575,
+      "train_speed(iter/s)": 1.318841
+    },
+    {
+      "acc": 0.67121596,
+      "epoch": 0.4713343480466768,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.08677839760243e-06,
+      "loss": 1.45608807,
+      "memory(GiB)": 103.58,
+      "step": 18580,
+      "train_speed(iter/s)": 1.318877
+    },
+    {
+      "acc": 0.67141757,
+      "epoch": 0.4714611872146119,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.086174160605055e-06,
+      "loss": 1.40652771,
+      "memory(GiB)": 103.58,
+      "step": 18585,
+      "train_speed(iter/s)": 1.318915
+    },
+    {
+      "acc": 0.67328253,
+      "epoch": 0.47158802638254693,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.085569743879498e-06,
+      "loss": 1.44002342,
+      "memory(GiB)": 103.58,
+      "step": 18590,
+      "train_speed(iter/s)": 1.318953
+    },
+    {
+      "acc": 0.68477678,
+      "epoch": 0.471714865550482,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.084965147452342e-06,
+      "loss": 1.3863245,
+      "memory(GiB)": 103.58,
+      "step": 18595,
+      "train_speed(iter/s)": 1.318993
+    },
+    {
+      "acc": 0.68059049,
+      "epoch": 0.471841704718417,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.084360371350182e-06,
+      "loss": 1.41247768,
+      "memory(GiB)": 103.58,
+      "step": 18600,
+      "train_speed(iter/s)": 1.31903
+    },
+    {
+      "acc": 0.68016968,
+      "epoch": 0.4719685438863521,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.083755415599617e-06,
+      "loss": 1.40592098,
+      "memory(GiB)": 103.58,
+      "step": 18605,
+      "train_speed(iter/s)": 1.319068
+    },
+    {
+      "acc": 0.68816681,
+      "epoch": 0.47209538305428717,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.083150280227255e-06,
+      "loss": 1.36198082,
+      "memory(GiB)": 103.58,
+      "step": 18610,
+      "train_speed(iter/s)": 1.319107
+    },
+    {
+      "acc": 0.6820456,
+      "epoch": 0.4722222222222222,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.082544965259716e-06,
+      "loss": 1.38555861,
+      "memory(GiB)": 103.58,
+      "step": 18615,
+      "train_speed(iter/s)": 1.319145
+    },
+    {
+      "acc": 0.68901396,
+      "epoch": 0.47234906139015725,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.081939470723619e-06,
+      "loss": 1.39495029,
+      "memory(GiB)": 103.58,
+      "step": 18620,
+      "train_speed(iter/s)": 1.319184
+    },
+    {
+      "acc": 0.67965012,
+      "epoch": 0.47247590055809235,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.081333796645603e-06,
+      "loss": 1.41470032,
+      "memory(GiB)": 103.58,
+      "step": 18625,
+      "train_speed(iter/s)": 1.319225
+    },
+    {
+      "acc": 0.68874211,
+      "epoch": 0.4726027397260274,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.080727943052304e-06,
+      "loss": 1.35649052,
+      "memory(GiB)": 103.58,
+      "step": 18630,
+      "train_speed(iter/s)": 1.319266
+    },
+    {
+      "acc": 0.68070269,
+      "epoch": 0.47272957889396244,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.080121909970369e-06,
+      "loss": 1.42022095,
+      "memory(GiB)": 103.58,
+      "step": 18635,
+      "train_speed(iter/s)": 1.319304
+    },
+    {
+      "acc": 0.69254489,
+      "epoch": 0.4728564180618975,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.079515697426457e-06,
+      "loss": 1.36278772,
+      "memory(GiB)": 103.58,
+      "step": 18640,
+      "train_speed(iter/s)": 1.319343
+    },
+    {
+      "acc": 0.68416271,
+      "epoch": 0.4729832572298326,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.078909305447231e-06,
+      "loss": 1.40863171,
+      "memory(GiB)": 103.58,
+      "step": 18645,
+      "train_speed(iter/s)": 1.319382
+    },
+    {
+      "acc": 0.70422101,
+      "epoch": 0.47311009639776763,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.078302734059363e-06,
+      "loss": 1.32107334,
+      "memory(GiB)": 103.58,
+      "step": 18650,
+      "train_speed(iter/s)": 1.319421
+    },
+    {
+      "acc": 0.69338484,
+      "epoch": 0.4732369355657027,
+      "grad_norm": 3.5,
+      "learning_rate": 9.077695983289531e-06,
+      "loss": 1.36578045,
+      "memory(GiB)": 103.58,
+      "step": 18655,
+      "train_speed(iter/s)": 1.319459
+    },
+    {
+      "acc": 0.6816783,
+      "epoch": 0.4733637747336377,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.077089053164426e-06,
+      "loss": 1.40120144,
+      "memory(GiB)": 103.58,
+      "step": 18660,
+      "train_speed(iter/s)": 1.319498
+    },
+    {
+      "acc": 0.69826412,
+      "epoch": 0.4734906139015728,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.076481943710742e-06,
+      "loss": 1.33756943,
+      "memory(GiB)": 103.58,
+      "step": 18665,
+      "train_speed(iter/s)": 1.319537
+    },
+    {
+      "acc": 0.68349228,
+      "epoch": 0.47361745306950787,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.075874654955178e-06,
+      "loss": 1.40167809,
+      "memory(GiB)": 103.58,
+      "step": 18670,
+      "train_speed(iter/s)": 1.319576
+    },
+    {
+      "acc": 0.68551445,
+      "epoch": 0.4737442922374429,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.075267186924453e-06,
+      "loss": 1.4105814,
+      "memory(GiB)": 103.58,
+      "step": 18675,
+      "train_speed(iter/s)": 1.319615
+    },
+    {
+      "acc": 0.67841711,
+      "epoch": 0.47387113140537795,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.074659539645281e-06,
+      "loss": 1.38588381,
+      "memory(GiB)": 103.58,
+      "step": 18680,
+      "train_speed(iter/s)": 1.319654
+    },
+    {
+      "acc": 0.67864428,
+      "epoch": 0.47399797057331305,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.074051713144392e-06,
+      "loss": 1.47656279,
+      "memory(GiB)": 103.58,
+      "step": 18685,
+      "train_speed(iter/s)": 1.319693
+    },
+    {
+      "acc": 0.69257622,
+      "epoch": 0.4741248097412481,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.073443707448517e-06,
+      "loss": 1.42322283,
+      "memory(GiB)": 103.58,
+      "step": 18690,
+      "train_speed(iter/s)": 1.31973
+    },
+    {
+      "acc": 0.67179918,
+      "epoch": 0.47425164890918314,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.072835522584402e-06,
+      "loss": 1.44540873,
+      "memory(GiB)": 103.58,
+      "step": 18695,
+      "train_speed(iter/s)": 1.319769
+    },
+    {
+      "acc": 0.69661727,
+      "epoch": 0.4743784880771182,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.072227158578798e-06,
+      "loss": 1.3479085,
+      "memory(GiB)": 103.58,
+      "step": 18700,
+      "train_speed(iter/s)": 1.319806
+    },
+    {
+      "acc": 0.67166543,
+      "epoch": 0.4745053272450533,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.07161861545846e-06,
+      "loss": 1.3836462,
+      "memory(GiB)": 103.58,
+      "step": 18705,
+      "train_speed(iter/s)": 1.319843
+    },
+    {
+      "acc": 0.6838522,
+      "epoch": 0.47463216641298833,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.071009893250158e-06,
+      "loss": 1.47533379,
+      "memory(GiB)": 103.58,
+      "step": 18710,
+      "train_speed(iter/s)": 1.319882
+    },
+    {
+      "acc": 0.68143396,
+      "epoch": 0.4747590055809234,
+      "grad_norm": 3.25,
+      "learning_rate": 9.070400991980666e-06,
+      "loss": 1.42765274,
+      "memory(GiB)": 103.58,
+      "step": 18715,
+      "train_speed(iter/s)": 1.319918
+    },
+    {
+      "acc": 0.69891148,
+      "epoch": 0.4748858447488584,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.069791911676765e-06,
+      "loss": 1.38244143,
+      "memory(GiB)": 103.58,
+      "step": 18720,
+      "train_speed(iter/s)": 1.319957
+    },
+    {
+      "acc": 0.7012064,
+      "epoch": 0.4750126839167935,
+      "grad_norm": 3.125,
+      "learning_rate": 9.069182652365245e-06,
+      "loss": 1.37186451,
+      "memory(GiB)": 103.58,
+      "step": 18725,
+      "train_speed(iter/s)": 1.319997
+    },
+    {
+      "acc": 0.67570558,
+      "epoch": 0.47513952308472857,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.068573214072904e-06,
+      "loss": 1.51373348,
+      "memory(GiB)": 103.58,
+      "step": 18730,
+      "train_speed(iter/s)": 1.320035
+    },
+    {
+      "acc": 0.67079172,
+      "epoch": 0.4752663622526636,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.067963596826547e-06,
+      "loss": 1.44170017,
+      "memory(GiB)": 103.58,
+      "step": 18735,
+      "train_speed(iter/s)": 1.320074
+    },
+    {
+      "acc": 0.68760157,
+      "epoch": 0.47539320142059865,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.067353800652991e-06,
+      "loss": 1.41090813,
+      "memory(GiB)": 103.58,
+      "step": 18740,
+      "train_speed(iter/s)": 1.320113
+    },
+    {
+      "acc": 0.68874617,
+      "epoch": 0.47552004058853375,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.066743825579056e-06,
+      "loss": 1.39497604,
+      "memory(GiB)": 103.58,
+      "step": 18745,
+      "train_speed(iter/s)": 1.320152
+    },
+    {
+      "acc": 0.6950726,
+      "epoch": 0.4756468797564688,
+      "grad_norm": 3.375,
+      "learning_rate": 9.06613367163157e-06,
+      "loss": 1.3340538,
+      "memory(GiB)": 103.58,
+      "step": 18750,
+      "train_speed(iter/s)": 1.320189
+    },
+    {
+      "acc": 0.68294673,
+      "epoch": 0.47577371892440384,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.06552333883737e-06,
+      "loss": 1.38902073,
+      "memory(GiB)": 103.58,
+      "step": 18755,
+      "train_speed(iter/s)": 1.320226
+    },
+    {
+      "acc": 0.68211746,
+      "epoch": 0.4759005580923389,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.064912827223303e-06,
+      "loss": 1.40830832,
+      "memory(GiB)": 103.58,
+      "step": 18760,
+      "train_speed(iter/s)": 1.320263
+    },
+    {
+      "acc": 0.68463402,
+      "epoch": 0.476027397260274,
+      "grad_norm": 2.75,
+      "learning_rate": 9.06430213681622e-06,
+      "loss": 1.39204416,
+      "memory(GiB)": 103.58,
+      "step": 18765,
+      "train_speed(iter/s)": 1.320299
+    },
+    {
+      "acc": 0.66714849,
+      "epoch": 0.47615423642820903,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.063691267642987e-06,
+      "loss": 1.48596287,
+      "memory(GiB)": 103.58,
+      "step": 18770,
+      "train_speed(iter/s)": 1.320336
+    },
+    {
+      "acc": 0.68315115,
+      "epoch": 0.4762810755961441,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.063080219730467e-06,
+      "loss": 1.42450733,
+      "memory(GiB)": 103.58,
+      "step": 18775,
+      "train_speed(iter/s)": 1.320374
+    },
+    {
+      "acc": 0.67680426,
+      "epoch": 0.4764079147640791,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.062468993105538e-06,
+      "loss": 1.39634399,
+      "memory(GiB)": 103.58,
+      "step": 18780,
+      "train_speed(iter/s)": 1.320412
+    },
+    {
+      "acc": 0.70220194,
+      "epoch": 0.4765347539320142,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.061857587795084e-06,
+      "loss": 1.30595837,
+      "memory(GiB)": 103.58,
+      "step": 18785,
+      "train_speed(iter/s)": 1.320451
+    },
+    {
+      "acc": 0.68454051,
+      "epoch": 0.47666159309994927,
+      "grad_norm": 3.375,
+      "learning_rate": 9.061246003826e-06,
+      "loss": 1.47164841,
+      "memory(GiB)": 103.58,
+      "step": 18790,
+      "train_speed(iter/s)": 1.320485
+    },
+    {
+      "acc": 0.67807426,
+      "epoch": 0.4767884322678843,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.060634241225184e-06,
+      "loss": 1.44097071,
+      "memory(GiB)": 103.58,
+      "step": 18795,
+      "train_speed(iter/s)": 1.320516
+    },
+    {
+      "acc": 0.68563495,
+      "epoch": 0.47691527143581935,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.060022300019546e-06,
+      "loss": 1.39773235,
+      "memory(GiB)": 103.58,
+      "step": 18800,
+      "train_speed(iter/s)": 1.320556
+    },
+    {
+      "acc": 0.69292593,
+      "epoch": 0.47704211060375445,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.059410180236e-06,
+      "loss": 1.35888853,
+      "memory(GiB)": 103.58,
+      "step": 18805,
+      "train_speed(iter/s)": 1.320592
+    },
+    {
+      "acc": 0.69567652,
+      "epoch": 0.4771689497716895,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.058797881901469e-06,
+      "loss": 1.36771774,
+      "memory(GiB)": 103.58,
+      "step": 18810,
+      "train_speed(iter/s)": 1.320629
+    },
+    {
+      "acc": 0.69977674,
+      "epoch": 0.47729578893962454,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.058185405042886e-06,
+      "loss": 1.30052414,
+      "memory(GiB)": 103.58,
+      "step": 18815,
+      "train_speed(iter/s)": 1.320667
+    },
+    {
+      "acc": 0.68218918,
+      "epoch": 0.4774226281075596,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.05757274968719e-06,
+      "loss": 1.46027737,
+      "memory(GiB)": 103.58,
+      "step": 18820,
+      "train_speed(iter/s)": 1.320706
+    },
+    {
+      "acc": 0.68356199,
+      "epoch": 0.4775494672754947,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.056959915861331e-06,
+      "loss": 1.36388855,
+      "memory(GiB)": 103.58,
+      "step": 18825,
+      "train_speed(iter/s)": 1.320745
+    },
+    {
+      "acc": 0.68107591,
+      "epoch": 0.47767630644342973,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.056346903592262e-06,
+      "loss": 1.43494482,
+      "memory(GiB)": 103.58,
+      "step": 18830,
+      "train_speed(iter/s)": 1.320783
+    },
+    {
+      "acc": 0.675279,
+      "epoch": 0.4778031456113648,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.055733712906943e-06,
+      "loss": 1.40753794,
+      "memory(GiB)": 103.58,
+      "step": 18835,
+      "train_speed(iter/s)": 1.32082
+    },
+    {
+      "acc": 0.66922989,
+      "epoch": 0.4779299847792998,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.05512034383235e-06,
+      "loss": 1.41080675,
+      "memory(GiB)": 103.58,
+      "step": 18840,
+      "train_speed(iter/s)": 1.320857
+    },
+    {
+      "acc": 0.67349634,
+      "epoch": 0.4780568239472349,
+      "grad_norm": 3.0,
+      "learning_rate": 9.054506796395458e-06,
+      "loss": 1.39216614,
+      "memory(GiB)": 103.58,
+      "step": 18845,
+      "train_speed(iter/s)": 1.320896
+    },
+    {
+      "acc": 0.68647428,
+      "epoch": 0.47818366311516997,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.053893070623256e-06,
+      "loss": 1.44026642,
+      "memory(GiB)": 103.58,
+      "step": 18850,
+      "train_speed(iter/s)": 1.320933
+    },
+    {
+      "acc": 0.69244995,
+      "epoch": 0.478310502283105,
+      "grad_norm": 3.25,
+      "learning_rate": 9.053279166542738e-06,
+      "loss": 1.4293829,
+      "memory(GiB)": 103.58,
+      "step": 18855,
+      "train_speed(iter/s)": 1.32097
+    },
+    {
+      "acc": 0.70210152,
+      "epoch": 0.47843734145104005,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.052665084180906e-06,
+      "loss": 1.37433243,
+      "memory(GiB)": 103.58,
+      "step": 18860,
+      "train_speed(iter/s)": 1.321008
+    },
+    {
+      "acc": 0.68534617,
+      "epoch": 0.47856418061897515,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.052050823564767e-06,
+      "loss": 1.40738964,
+      "memory(GiB)": 103.58,
+      "step": 18865,
+      "train_speed(iter/s)": 1.321046
+    },
+    {
+      "acc": 0.69478521,
+      "epoch": 0.4786910197869102,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.051436384721344e-06,
+      "loss": 1.396311,
+      "memory(GiB)": 103.58,
+      "step": 18870,
+      "train_speed(iter/s)": 1.321082
+    },
+    {
+      "acc": 0.7122489,
+      "epoch": 0.47881785895484524,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.05082176767766e-06,
+      "loss": 1.30196333,
+      "memory(GiB)": 103.58,
+      "step": 18875,
+      "train_speed(iter/s)": 1.32112
+    },
+    {
+      "acc": 0.68459244,
+      "epoch": 0.4789446981227803,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.050206972460749e-06,
+      "loss": 1.40288029,
+      "memory(GiB)": 103.58,
+      "step": 18880,
+      "train_speed(iter/s)": 1.321157
+    },
+    {
+      "acc": 0.68960514,
+      "epoch": 0.4790715372907154,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.049591999097651e-06,
+      "loss": 1.41515455,
+      "memory(GiB)": 103.58,
+      "step": 18885,
+      "train_speed(iter/s)": 1.321194
+    },
+    {
+      "acc": 0.67584,
+      "epoch": 0.47919837645865043,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.048976847615418e-06,
+      "loss": 1.43499565,
+      "memory(GiB)": 103.58,
+      "step": 18890,
+      "train_speed(iter/s)": 1.321233
+    },
+    {
+      "acc": 0.68345957,
+      "epoch": 0.4793252156265855,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.048361518041107e-06,
+      "loss": 1.42118092,
+      "memory(GiB)": 103.58,
+      "step": 18895,
+      "train_speed(iter/s)": 1.32127
+    },
+    {
+      "acc": 0.69404058,
+      "epoch": 0.4794520547945205,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.04774601040178e-06,
+      "loss": 1.33283749,
+      "memory(GiB)": 103.58,
+      "step": 18900,
+      "train_speed(iter/s)": 1.321307
+    },
+    {
+      "acc": 0.67677393,
+      "epoch": 0.4795788939624556,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.047130324724513e-06,
+      "loss": 1.39266644,
+      "memory(GiB)": 103.58,
+      "step": 18905,
+      "train_speed(iter/s)": 1.321344
+    },
+    {
+      "acc": 0.68432217,
+      "epoch": 0.47970573313039067,
+      "grad_norm": 5.0,
+      "learning_rate": 9.046514461036385e-06,
+      "loss": 1.40724649,
+      "memory(GiB)": 103.58,
+      "step": 18910,
+      "train_speed(iter/s)": 1.32138
+    },
+    {
+      "acc": 0.67714438,
+      "epoch": 0.4798325722983257,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.045898419364483e-06,
+      "loss": 1.43462267,
+      "memory(GiB)": 103.58,
+      "step": 18915,
+      "train_speed(iter/s)": 1.321417
+    },
+    {
+      "acc": 0.69178362,
+      "epoch": 0.47995941146626075,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.045282199735906e-06,
+      "loss": 1.3548769,
+      "memory(GiB)": 103.58,
+      "step": 18920,
+      "train_speed(iter/s)": 1.321454
+    },
+    {
+      "acc": 0.66762557,
+      "epoch": 0.48008625063419585,
+      "grad_norm": 3.25,
+      "learning_rate": 9.044665802177756e-06,
+      "loss": 1.5322607,
+      "memory(GiB)": 103.58,
+      "step": 18925,
+      "train_speed(iter/s)": 1.321492
+    },
+    {
+      "acc": 0.68568668,
+      "epoch": 0.4802130898021309,
+      "grad_norm": 4.875,
+      "learning_rate": 9.044049226717148e-06,
+      "loss": 1.3820013,
+      "memory(GiB)": 103.58,
+      "step": 18930,
+      "train_speed(iter/s)": 1.32153
+    },
+    {
+      "acc": 0.70043707,
+      "epoch": 0.48033992897006594,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.043432473381198e-06,
+      "loss": 1.38653126,
+      "memory(GiB)": 103.58,
+      "step": 18935,
+      "train_speed(iter/s)": 1.321566
+    },
+    {
+      "acc": 0.68504024,
+      "epoch": 0.480466768138001,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.042815542197037e-06,
+      "loss": 1.43598289,
+      "memory(GiB)": 103.58,
+      "step": 18940,
+      "train_speed(iter/s)": 1.3216
+    },
+    {
+      "acc": 0.68158855,
+      "epoch": 0.4805936073059361,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.042198433191796e-06,
+      "loss": 1.37148733,
+      "memory(GiB)": 103.58,
+      "step": 18945,
+      "train_speed(iter/s)": 1.321632
+    },
+    {
+      "acc": 0.67623663,
+      "epoch": 0.48072044647387113,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.041581146392621e-06,
+      "loss": 1.42195253,
+      "memory(GiB)": 103.58,
+      "step": 18950,
+      "train_speed(iter/s)": 1.321669
+    },
+    {
+      "acc": 0.68790684,
+      "epoch": 0.4808472856418062,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.040963681826665e-06,
+      "loss": 1.37742786,
+      "memory(GiB)": 103.58,
+      "step": 18955,
+      "train_speed(iter/s)": 1.321706
+    },
+    {
+      "acc": 0.68301992,
+      "epoch": 0.4809741248097412,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.040346039521085e-06,
+      "loss": 1.38918972,
+      "memory(GiB)": 103.58,
+      "step": 18960,
+      "train_speed(iter/s)": 1.321742
+    },
+    {
+      "acc": 0.69233952,
+      "epoch": 0.4811009639776763,
+      "grad_norm": 2.875,
+      "learning_rate": 9.039728219503044e-06,
+      "loss": 1.37599888,
+      "memory(GiB)": 103.58,
+      "step": 18965,
+      "train_speed(iter/s)": 1.321777
+    },
+    {
+      "acc": 0.69192815,
+      "epoch": 0.48122780314561137,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.039110221799721e-06,
+      "loss": 1.36013107,
+      "memory(GiB)": 103.58,
+      "step": 18970,
+      "train_speed(iter/s)": 1.321813
+    },
+    {
+      "acc": 0.68378758,
+      "epoch": 0.4813546423135464,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.038492046438298e-06,
+      "loss": 1.41812811,
+      "memory(GiB)": 103.58,
+      "step": 18975,
+      "train_speed(iter/s)": 1.32185
+    },
+    {
+      "acc": 0.68946342,
+      "epoch": 0.48148148148148145,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.037873693445965e-06,
+      "loss": 1.42467136,
+      "memory(GiB)": 103.58,
+      "step": 18980,
+      "train_speed(iter/s)": 1.321886
+    },
+    {
+      "acc": 0.69210815,
+      "epoch": 0.48160832064941655,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.037255162849918e-06,
+      "loss": 1.38696508,
+      "memory(GiB)": 103.58,
+      "step": 18985,
+      "train_speed(iter/s)": 1.321922
+    },
+    {
+      "acc": 0.68417296,
+      "epoch": 0.4817351598173516,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.036636454677363e-06,
+      "loss": 1.39669647,
+      "memory(GiB)": 103.58,
+      "step": 18990,
+      "train_speed(iter/s)": 1.321958
+    },
+    {
+      "acc": 0.67813058,
+      "epoch": 0.48186199898528664,
+      "grad_norm": 3.375,
+      "learning_rate": 9.036017568955516e-06,
+      "loss": 1.43244305,
+      "memory(GiB)": 103.58,
+      "step": 18995,
+      "train_speed(iter/s)": 1.321994
+    },
+    {
+      "acc": 0.68088503,
+      "epoch": 0.4819888381532217,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.035398505711597e-06,
+      "loss": 1.43847151,
+      "memory(GiB)": 103.58,
+      "step": 19000,
+      "train_speed(iter/s)": 1.32203
+    },
+    {
+      "epoch": 0.4819888381532217,
+      "eval_acc": 0.6730844490108738,
+      "eval_loss": 1.366981863975525,
+      "eval_runtime": 69.7397,
+      "eval_samples_per_second": 91.34,
+      "eval_steps_per_second": 22.842,
+      "step": 19000
+    },
+    {
+      "acc": 0.6876956,
+      "epoch": 0.4821156773211568,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.034779264972834e-06,
+      "loss": 1.3566391,
+      "memory(GiB)": 103.58,
+      "step": 19005,
+      "train_speed(iter/s)": 1.314537
+    },
+    {
+      "acc": 0.68694429,
+      "epoch": 0.48224251648909183,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.034159846766464e-06,
+      "loss": 1.38222771,
+      "memory(GiB)": 103.58,
+      "step": 19010,
+      "train_speed(iter/s)": 1.314575
+    },
+    {
+      "acc": 0.69295158,
+      "epoch": 0.4823693556570269,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.033540251119734e-06,
+      "loss": 1.29019384,
+      "memory(GiB)": 103.58,
+      "step": 19015,
+      "train_speed(iter/s)": 1.314615
+    },
+    {
+      "acc": 0.69342327,
+      "epoch": 0.4824961948249619,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.032920478059897e-06,
+      "loss": 1.37044811,
+      "memory(GiB)": 103.58,
+      "step": 19020,
+      "train_speed(iter/s)": 1.314653
+    },
+    {
+      "acc": 0.68355613,
+      "epoch": 0.482623033992897,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.032300527614209e-06,
+      "loss": 1.44340658,
+      "memory(GiB)": 103.58,
+      "step": 19025,
+      "train_speed(iter/s)": 1.314692
+    },
+    {
+      "acc": 0.67872028,
+      "epoch": 0.48274987316083207,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.031680399809941e-06,
+      "loss": 1.44171791,
+      "memory(GiB)": 103.58,
+      "step": 19030,
+      "train_speed(iter/s)": 1.314732
+    },
+    {
+      "acc": 0.67052226,
+      "epoch": 0.4828767123287671,
+      "grad_norm": 2.78125,
+      "learning_rate": 9.031060094674371e-06,
+      "loss": 1.49645939,
+      "memory(GiB)": 103.58,
+      "step": 19035,
+      "train_speed(iter/s)": 1.314771
+    },
+    {
+      "acc": 0.68518386,
+      "epoch": 0.48300355149670215,
+      "grad_norm": 4.0,
+      "learning_rate": 9.030439612234778e-06,
+      "loss": 1.3779521,
+      "memory(GiB)": 103.58,
+      "step": 19040,
+      "train_speed(iter/s)": 1.31481
+    },
+    {
+      "acc": 0.68625898,
+      "epoch": 0.48313039066463725,
+      "grad_norm": 2.75,
+      "learning_rate": 9.029818952518458e-06,
+      "loss": 1.38108072,
+      "memory(GiB)": 103.58,
+      "step": 19045,
+      "train_speed(iter/s)": 1.314848
+    },
+    {
+      "acc": 0.70222263,
+      "epoch": 0.4832572298325723,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.029198115552708e-06,
+      "loss": 1.32400341,
+      "memory(GiB)": 103.58,
+      "step": 19050,
+      "train_speed(iter/s)": 1.314887
+    },
+    {
+      "acc": 0.68252773,
+      "epoch": 0.48338406900050734,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.028577101364837e-06,
+      "loss": 1.37568073,
+      "memory(GiB)": 103.58,
+      "step": 19055,
+      "train_speed(iter/s)": 1.314926
+    },
+    {
+      "acc": 0.69028454,
+      "epoch": 0.4835109081684424,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.027955909982157e-06,
+      "loss": 1.37558794,
+      "memory(GiB)": 103.58,
+      "step": 19060,
+      "train_speed(iter/s)": 1.314967
+    },
+    {
+      "acc": 0.71357508,
+      "epoch": 0.4836377473363775,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.027334541431993e-06,
+      "loss": 1.28133011,
+      "memory(GiB)": 103.58,
+      "step": 19065,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.68319945,
+      "epoch": 0.48376458650431253,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.026712995741676e-06,
+      "loss": 1.36374111,
+      "memory(GiB)": 103.58,
+      "step": 19070,
+      "train_speed(iter/s)": 1.315046
+    },
+    {
+      "acc": 0.67084517,
+      "epoch": 0.4838914256722476,
+      "grad_norm": 3.0,
+      "learning_rate": 9.026091272938543e-06,
+      "loss": 1.44566593,
+      "memory(GiB)": 103.58,
+      "step": 19075,
+      "train_speed(iter/s)": 1.315071
+    },
+    {
+      "acc": 0.69248781,
+      "epoch": 0.4840182648401826,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.02546937304994e-06,
+      "loss": 1.3839015,
+      "memory(GiB)": 103.58,
+      "step": 19080,
+      "train_speed(iter/s)": 1.315113
+    },
+    {
+      "acc": 0.67716398,
+      "epoch": 0.4841451040081177,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.02484729610322e-06,
+      "loss": 1.43924646,
+      "memory(GiB)": 103.58,
+      "step": 19085,
+      "train_speed(iter/s)": 1.315153
+    },
+    {
+      "acc": 0.67551746,
+      "epoch": 0.48427194317605277,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.02422504212575e-06,
+      "loss": 1.46464901,
+      "memory(GiB)": 103.58,
+      "step": 19090,
+      "train_speed(iter/s)": 1.315195
+    },
+    {
+      "acc": 0.69333582,
+      "epoch": 0.4843987823439878,
+      "grad_norm": 2.625,
+      "learning_rate": 9.023602611144893e-06,
+      "loss": 1.3724287,
+      "memory(GiB)": 103.58,
+      "step": 19095,
+      "train_speed(iter/s)": 1.315235
+    },
+    {
+      "acc": 0.67581468,
+      "epoch": 0.48452562151192285,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.02298000318803e-06,
+      "loss": 1.39292469,
+      "memory(GiB)": 103.58,
+      "step": 19100,
+      "train_speed(iter/s)": 1.315273
+    },
+    {
+      "acc": 0.68232269,
+      "epoch": 0.48465246067985795,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.022357218282546e-06,
+      "loss": 1.39472904,
+      "memory(GiB)": 103.58,
+      "step": 19105,
+      "train_speed(iter/s)": 1.315311
+    },
+    {
+      "acc": 0.67999849,
+      "epoch": 0.484779299847793,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.021734256455832e-06,
+      "loss": 1.44115562,
+      "memory(GiB)": 103.58,
+      "step": 19110,
+      "train_speed(iter/s)": 1.31535
+    },
+    {
+      "acc": 0.7125391,
+      "epoch": 0.48490613901572804,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.02111111773529e-06,
+      "loss": 1.34007378,
+      "memory(GiB)": 103.58,
+      "step": 19115,
+      "train_speed(iter/s)": 1.315388
+    },
+    {
+      "acc": 0.68850513,
+      "epoch": 0.4850329781836631,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.020487802148328e-06,
+      "loss": 1.34823608,
+      "memory(GiB)": 103.58,
+      "step": 19120,
+      "train_speed(iter/s)": 1.315428
+    },
+    {
+      "acc": 0.67842178,
+      "epoch": 0.4851598173515982,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.019864309722362e-06,
+      "loss": 1.40566626,
+      "memory(GiB)": 103.58,
+      "step": 19125,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.69361506,
+      "epoch": 0.48528665651953323,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.019240640484816e-06,
+      "loss": 1.36162586,
+      "memory(GiB)": 103.58,
+      "step": 19130,
+      "train_speed(iter/s)": 1.315505
+    },
+    {
+      "acc": 0.68406153,
+      "epoch": 0.4854134956874683,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.018616794463124e-06,
+      "loss": 1.34850626,
+      "memory(GiB)": 103.58,
+      "step": 19135,
+      "train_speed(iter/s)": 1.315543
+    },
+    {
+      "acc": 0.67942891,
+      "epoch": 0.4855403348554033,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.017992771684722e-06,
+      "loss": 1.43746557,
+      "memory(GiB)": 103.58,
+      "step": 19140,
+      "train_speed(iter/s)": 1.315581
+    },
+    {
+      "acc": 0.6761045,
+      "epoch": 0.4856671740233384,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.017368572177058e-06,
+      "loss": 1.38847046,
+      "memory(GiB)": 103.58,
+      "step": 19145,
+      "train_speed(iter/s)": 1.315619
+    },
+    {
+      "acc": 0.68292122,
+      "epoch": 0.48579401319127347,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.016744195967588e-06,
+      "loss": 1.39128628,
+      "memory(GiB)": 103.58,
+      "step": 19150,
+      "train_speed(iter/s)": 1.315657
+    },
+    {
+      "acc": 0.69990354,
+      "epoch": 0.4859208523592085,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.016119643083777e-06,
+      "loss": 1.41899891,
+      "memory(GiB)": 103.58,
+      "step": 19155,
+      "train_speed(iter/s)": 1.315695
+    },
+    {
+      "acc": 0.68931046,
+      "epoch": 0.48604769152714355,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.015494913553091e-06,
+      "loss": 1.4066905,
+      "memory(GiB)": 103.58,
+      "step": 19160,
+      "train_speed(iter/s)": 1.315735
+    },
+    {
+      "acc": 0.67969069,
+      "epoch": 0.48617453069507865,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.014870007403012e-06,
+      "loss": 1.38889656,
+      "memory(GiB)": 103.58,
+      "step": 19165,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.68226891,
+      "epoch": 0.4863013698630137,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.014244924661026e-06,
+      "loss": 1.36892815,
+      "memory(GiB)": 103.58,
+      "step": 19170,
+      "train_speed(iter/s)": 1.315816
+    },
+    {
+      "acc": 0.67655535,
+      "epoch": 0.48642820903094874,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.013619665354626e-06,
+      "loss": 1.4225523,
+      "memory(GiB)": 103.58,
+      "step": 19175,
+      "train_speed(iter/s)": 1.315851
+    },
+    {
+      "acc": 0.66859694,
+      "epoch": 0.4865550481988838,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.01299422951131e-06,
+      "loss": 1.47639503,
+      "memory(GiB)": 103.58,
+      "step": 19180,
+      "train_speed(iter/s)": 1.31589
+    },
+    {
+      "acc": 0.67787876,
+      "epoch": 0.4866818873668189,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.012368617158593e-06,
+      "loss": 1.42139635,
+      "memory(GiB)": 103.58,
+      "step": 19185,
+      "train_speed(iter/s)": 1.31593
+    },
+    {
+      "acc": 0.67928948,
+      "epoch": 0.48680872653475393,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.01174282832399e-06,
+      "loss": 1.4239809,
+      "memory(GiB)": 103.58,
+      "step": 19190,
+      "train_speed(iter/s)": 1.31597
+    },
+    {
+      "acc": 0.69949665,
+      "epoch": 0.486935565702689,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.011116863035027e-06,
+      "loss": 1.34040537,
+      "memory(GiB)": 103.58,
+      "step": 19195,
+      "train_speed(iter/s)": 1.316004
+    },
+    {
+      "acc": 0.6798625,
+      "epoch": 0.487062404870624,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.010490721319237e-06,
+      "loss": 1.4125349,
+      "memory(GiB)": 103.58,
+      "step": 19200,
+      "train_speed(iter/s)": 1.316045
+    },
+    {
+      "acc": 0.68512869,
+      "epoch": 0.4871892440385591,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.009864403204157e-06,
+      "loss": 1.3732666,
+      "memory(GiB)": 103.58,
+      "step": 19205,
+      "train_speed(iter/s)": 1.316083
+    },
+    {
+      "acc": 0.68800058,
+      "epoch": 0.48731608320649417,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.00923790871734e-06,
+      "loss": 1.35925894,
+      "memory(GiB)": 103.58,
+      "step": 19210,
+      "train_speed(iter/s)": 1.316123
+    },
+    {
+      "acc": 0.68618698,
+      "epoch": 0.4874429223744292,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.008611237886339e-06,
+      "loss": 1.45369463,
+      "memory(GiB)": 103.58,
+      "step": 19215,
+      "train_speed(iter/s)": 1.316163
+    },
+    {
+      "acc": 0.68434186,
+      "epoch": 0.48756976154236426,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.00798439073872e-06,
+      "loss": 1.41438131,
+      "memory(GiB)": 103.58,
+      "step": 19220,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.67759838,
+      "epoch": 0.48769660071029935,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.007357367302052e-06,
+      "loss": 1.42844849,
+      "memory(GiB)": 103.58,
+      "step": 19225,
+      "train_speed(iter/s)": 1.316244
+    },
+    {
+      "acc": 0.68190193,
+      "epoch": 0.4878234398782344,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.006730167603914e-06,
+      "loss": 1.46543331,
+      "memory(GiB)": 103.58,
+      "step": 19230,
+      "train_speed(iter/s)": 1.316285
+    },
+    {
+      "acc": 0.67640543,
+      "epoch": 0.48795027904616944,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.006102791671896e-06,
+      "loss": 1.42278423,
+      "memory(GiB)": 103.58,
+      "step": 19235,
+      "train_speed(iter/s)": 1.316325
+    },
+    {
+      "acc": 0.68868084,
+      "epoch": 0.4880771182141045,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.005475239533591e-06,
+      "loss": 1.40680428,
+      "memory(GiB)": 103.58,
+      "step": 19240,
+      "train_speed(iter/s)": 1.316366
+    },
+    {
+      "acc": 0.67565365,
+      "epoch": 0.4882039573820396,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.0048475112166e-06,
+      "loss": 1.40352755,
+      "memory(GiB)": 103.58,
+      "step": 19245,
+      "train_speed(iter/s)": 1.316406
+    },
+    {
+      "acc": 0.67944727,
+      "epoch": 0.48833079654997463,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.00421960674854e-06,
+      "loss": 1.41135635,
+      "memory(GiB)": 103.58,
+      "step": 19250,
+      "train_speed(iter/s)": 1.316447
+    },
+    {
+      "acc": 0.67938709,
+      "epoch": 0.4884576357179097,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.003591526157021e-06,
+      "loss": 1.4315753,
+      "memory(GiB)": 103.58,
+      "step": 19255,
+      "train_speed(iter/s)": 1.316488
+    },
+    {
+      "acc": 0.67961593,
+      "epoch": 0.4885844748858447,
+      "grad_norm": 3.5,
+      "learning_rate": 9.002963269469672e-06,
+      "loss": 1.42473602,
+      "memory(GiB)": 103.58,
+      "step": 19260,
+      "train_speed(iter/s)": 1.316527
+    },
+    {
+      "acc": 0.70032907,
+      "epoch": 0.4887113140537798,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.002334836714126e-06,
+      "loss": 1.346696,
+      "memory(GiB)": 103.58,
+      "step": 19265,
+      "train_speed(iter/s)": 1.316568
+    },
+    {
+      "acc": 0.67868986,
+      "epoch": 0.48883815322171487,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.001706227918023e-06,
+      "loss": 1.40347176,
+      "memory(GiB)": 103.58,
+      "step": 19270,
+      "train_speed(iter/s)": 1.316605
+    },
+    {
+      "acc": 0.68042727,
+      "epoch": 0.4889649923896499,
+      "grad_norm": 2.859375,
+      "learning_rate": 9.001077443109016e-06,
+      "loss": 1.44891891,
+      "memory(GiB)": 103.58,
+      "step": 19275,
+      "train_speed(iter/s)": 1.316646
+    },
+    {
+      "acc": 0.6722342,
+      "epoch": 0.48909183155758496,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.00044848231476e-06,
+      "loss": 1.42517815,
+      "memory(GiB)": 103.58,
+      "step": 19280,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.70216045,
+      "epoch": 0.48921867072552006,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.999819345562919e-06,
+      "loss": 1.35501261,
+      "memory(GiB)": 103.58,
+      "step": 19285,
+      "train_speed(iter/s)": 1.316726
+    },
+    {
+      "acc": 0.68037462,
+      "epoch": 0.4893455098934551,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.999190032881165e-06,
+      "loss": 1.45962029,
+      "memory(GiB)": 103.58,
+      "step": 19290,
+      "train_speed(iter/s)": 1.316766
+    },
+    {
+      "acc": 0.69964819,
+      "epoch": 0.48947234906139014,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.998560544297176e-06,
+      "loss": 1.34114542,
+      "memory(GiB)": 103.58,
+      "step": 19295,
+      "train_speed(iter/s)": 1.316805
+    },
+    {
+      "acc": 0.68907747,
+      "epoch": 0.4895991882293252,
+      "grad_norm": 2.875,
+      "learning_rate": 8.997930879838646e-06,
+      "loss": 1.36978378,
+      "memory(GiB)": 103.58,
+      "step": 19300,
+      "train_speed(iter/s)": 1.31684
+    },
+    {
+      "acc": 0.68266768,
+      "epoch": 0.4897260273972603,
+      "grad_norm": 2.75,
+      "learning_rate": 8.997301039533264e-06,
+      "loss": 1.40924397,
+      "memory(GiB)": 103.58,
+      "step": 19305,
+      "train_speed(iter/s)": 1.316881
+    },
+    {
+      "acc": 0.68282547,
+      "epoch": 0.48985286656519533,
+      "grad_norm": 3.25,
+      "learning_rate": 8.996671023408737e-06,
+      "loss": 1.40925245,
+      "memory(GiB)": 103.58,
+      "step": 19310,
+      "train_speed(iter/s)": 1.316921
+    },
+    {
+      "acc": 0.6810616,
+      "epoch": 0.4899797057331304,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.996040831492772e-06,
+      "loss": 1.42377605,
+      "memory(GiB)": 103.58,
+      "step": 19315,
+      "train_speed(iter/s)": 1.316959
+    },
+    {
+      "acc": 0.66551237,
+      "epoch": 0.4901065449010654,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.995410463813093e-06,
+      "loss": 1.39030361,
+      "memory(GiB)": 103.58,
+      "step": 19320,
+      "train_speed(iter/s)": 1.317
+    },
+    {
+      "acc": 0.68812351,
+      "epoch": 0.4902333840690005,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.99477992039742e-06,
+      "loss": 1.35861721,
+      "memory(GiB)": 103.58,
+      "step": 19325,
+      "train_speed(iter/s)": 1.317039
+    },
+    {
+      "acc": 0.6853539,
+      "epoch": 0.49036022323693557,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.994149201273495e-06,
+      "loss": 1.39324942,
+      "memory(GiB)": 103.58,
+      "step": 19330,
+      "train_speed(iter/s)": 1.317081
+    },
+    {
+      "acc": 0.6802495,
+      "epoch": 0.4904870624048706,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.993518306469052e-06,
+      "loss": 1.35569201,
+      "memory(GiB)": 103.58,
+      "step": 19335,
+      "train_speed(iter/s)": 1.317121
+    },
+    {
+      "acc": 0.65835567,
+      "epoch": 0.49061390157280566,
+      "grad_norm": 3.625,
+      "learning_rate": 8.992887236011847e-06,
+      "loss": 1.49313049,
+      "memory(GiB)": 103.58,
+      "step": 19340,
+      "train_speed(iter/s)": 1.317162
+    },
+    {
+      "acc": 0.67710876,
+      "epoch": 0.49074074074074076,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.992255989929632e-06,
+      "loss": 1.43807707,
+      "memory(GiB)": 103.58,
+      "step": 19345,
+      "train_speed(iter/s)": 1.317202
+    },
+    {
+      "acc": 0.70179052,
+      "epoch": 0.4908675799086758,
+      "grad_norm": 2.75,
+      "learning_rate": 8.991624568250175e-06,
+      "loss": 1.3543725,
+      "memory(GiB)": 103.58,
+      "step": 19350,
+      "train_speed(iter/s)": 1.317237
+    },
+    {
+      "acc": 0.67970772,
+      "epoch": 0.49099441907661084,
+      "grad_norm": 3.375,
+      "learning_rate": 8.99099297100125e-06,
+      "loss": 1.46779137,
+      "memory(GiB)": 103.58,
+      "step": 19355,
+      "train_speed(iter/s)": 1.317276
+    },
+    {
+      "acc": 0.69388609,
+      "epoch": 0.4911212582445459,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.990361198210634e-06,
+      "loss": 1.39862156,
+      "memory(GiB)": 103.58,
+      "step": 19360,
+      "train_speed(iter/s)": 1.317313
+    },
+    {
+      "acc": 0.66461067,
+      "epoch": 0.491248097412481,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.989729249906116e-06,
+      "loss": 1.48926249,
+      "memory(GiB)": 103.58,
+      "step": 19365,
+      "train_speed(iter/s)": 1.317353
+    },
+    {
+      "acc": 0.68404007,
+      "epoch": 0.49137493658041603,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.989097126115493e-06,
+      "loss": 1.38580408,
+      "memory(GiB)": 103.58,
+      "step": 19370,
+      "train_speed(iter/s)": 1.317394
+    },
+    {
+      "acc": 0.68146987,
+      "epoch": 0.4915017757483511,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.98846482686657e-06,
+      "loss": 1.45679464,
+      "memory(GiB)": 103.58,
+      "step": 19375,
+      "train_speed(iter/s)": 1.317435
+    },
+    {
+      "acc": 0.69298916,
+      "epoch": 0.4916286149162861,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.987832352187156e-06,
+      "loss": 1.41692915,
+      "memory(GiB)": 103.58,
+      "step": 19380,
+      "train_speed(iter/s)": 1.317475
+    },
+    {
+      "acc": 0.68328781,
+      "epoch": 0.4917554540842212,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.987199702105071e-06,
+      "loss": 1.46424561,
+      "memory(GiB)": 103.58,
+      "step": 19385,
+      "train_speed(iter/s)": 1.317514
+    },
+    {
+      "acc": 0.67280254,
+      "epoch": 0.49188229325215627,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.986566876648141e-06,
+      "loss": 1.450875,
+      "memory(GiB)": 103.58,
+      "step": 19390,
+      "train_speed(iter/s)": 1.317555
+    },
+    {
+      "acc": 0.68136091,
+      "epoch": 0.4920091324200913,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.985933875844202e-06,
+      "loss": 1.40279255,
+      "memory(GiB)": 103.58,
+      "step": 19395,
+      "train_speed(iter/s)": 1.317594
+    },
+    {
+      "acc": 0.71107731,
+      "epoch": 0.49213597158802636,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.985300699721094e-06,
+      "loss": 1.33500195,
+      "memory(GiB)": 103.58,
+      "step": 19400,
+      "train_speed(iter/s)": 1.317633
+    },
+    {
+      "acc": 0.68769989,
+      "epoch": 0.49226281075596146,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.984667348306669e-06,
+      "loss": 1.42610493,
+      "memory(GiB)": 103.58,
+      "step": 19405,
+      "train_speed(iter/s)": 1.317674
+    },
+    {
+      "acc": 0.69374847,
+      "epoch": 0.4923896499238965,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.984033821628782e-06,
+      "loss": 1.36379566,
+      "memory(GiB)": 103.58,
+      "step": 19410,
+      "train_speed(iter/s)": 1.317713
+    },
+    {
+      "acc": 0.7024827,
+      "epoch": 0.49251648909183154,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.983400119715303e-06,
+      "loss": 1.31514235,
+      "memory(GiB)": 103.58,
+      "step": 19415,
+      "train_speed(iter/s)": 1.317752
+    },
+    {
+      "acc": 0.68715725,
+      "epoch": 0.4926433282597666,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.982766242594099e-06,
+      "loss": 1.38897028,
+      "memory(GiB)": 103.58,
+      "step": 19420,
+      "train_speed(iter/s)": 1.317786
+    },
+    {
+      "acc": 0.7044467,
+      "epoch": 0.4927701674277017,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.982132190293056e-06,
+      "loss": 1.40171452,
+      "memory(GiB)": 103.58,
+      "step": 19425,
+      "train_speed(iter/s)": 1.317826
+    },
+    {
+      "acc": 0.69510517,
+      "epoch": 0.49289700659563673,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.98149796284006e-06,
+      "loss": 1.36614828,
+      "memory(GiB)": 103.58,
+      "step": 19430,
+      "train_speed(iter/s)": 1.317864
+    },
+    {
+      "acc": 0.6927731,
+      "epoch": 0.4930238457635718,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.980863560263007e-06,
+      "loss": 1.38352585,
+      "memory(GiB)": 103.58,
+      "step": 19435,
+      "train_speed(iter/s)": 1.317903
+    },
+    {
+      "acc": 0.68345218,
+      "epoch": 0.4931506849315068,
+      "grad_norm": 3.0,
+      "learning_rate": 8.980228982589802e-06,
+      "loss": 1.39334698,
+      "memory(GiB)": 103.58,
+      "step": 19440,
+      "train_speed(iter/s)": 1.317934
+    },
+    {
+      "acc": 0.69563665,
+      "epoch": 0.4932775240994419,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.979594229848355e-06,
+      "loss": 1.39272194,
+      "memory(GiB)": 103.58,
+      "step": 19445,
+      "train_speed(iter/s)": 1.317973
+    },
+    {
+      "acc": 0.6885015,
+      "epoch": 0.49340436326737697,
+      "grad_norm": 3.0,
+      "learning_rate": 8.978959302066587e-06,
+      "loss": 1.42690268,
+      "memory(GiB)": 103.58,
+      "step": 19450,
+      "train_speed(iter/s)": 1.318013
+    },
+    {
+      "acc": 0.68070879,
+      "epoch": 0.493531202435312,
+      "grad_norm": 2.625,
+      "learning_rate": 8.978324199272423e-06,
+      "loss": 1.38103733,
+      "memory(GiB)": 103.58,
+      "step": 19455,
+      "train_speed(iter/s)": 1.318051
+    },
+    {
+      "acc": 0.69759583,
+      "epoch": 0.49365804160324706,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.977688921493799e-06,
+      "loss": 1.34965553,
+      "memory(GiB)": 103.58,
+      "step": 19460,
+      "train_speed(iter/s)": 1.31809
+    },
+    {
+      "acc": 0.68801155,
+      "epoch": 0.49378488077118216,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.977053468758659e-06,
+      "loss": 1.37086906,
+      "memory(GiB)": 103.58,
+      "step": 19465,
+      "train_speed(iter/s)": 1.318129
+    },
+    {
+      "acc": 0.68840899,
+      "epoch": 0.4939117199391172,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.976417841094949e-06,
+      "loss": 1.40433722,
+      "memory(GiB)": 103.58,
+      "step": 19470,
+      "train_speed(iter/s)": 1.318169
+    },
+    {
+      "acc": 0.69692612,
+      "epoch": 0.49403855910705224,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.97578203853063e-06,
+      "loss": 1.40716772,
+      "memory(GiB)": 103.58,
+      "step": 19475,
+      "train_speed(iter/s)": 1.318208
+    },
+    {
+      "acc": 0.68428626,
+      "epoch": 0.4941653982749873,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.975146061093667e-06,
+      "loss": 1.38553066,
+      "memory(GiB)": 103.58,
+      "step": 19480,
+      "train_speed(iter/s)": 1.318247
+    },
+    {
+      "acc": 0.67270117,
+      "epoch": 0.4942922374429224,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.97450990881203e-06,
+      "loss": 1.48180809,
+      "memory(GiB)": 103.58,
+      "step": 19485,
+      "train_speed(iter/s)": 1.318285
+    },
+    {
+      "acc": 0.70159292,
+      "epoch": 0.49441907661085743,
+      "grad_norm": 4.8125,
+      "learning_rate": 8.973873581713705e-06,
+      "loss": 1.3912961,
+      "memory(GiB)": 103.58,
+      "step": 19490,
+      "train_speed(iter/s)": 1.318326
+    },
+    {
+      "acc": 0.69510765,
+      "epoch": 0.4945459157787925,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.973237079826676e-06,
+      "loss": 1.42784367,
+      "memory(GiB)": 103.58,
+      "step": 19495,
+      "train_speed(iter/s)": 1.318364
+    },
+    {
+      "acc": 0.68524561,
+      "epoch": 0.4946727549467275,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.972600403178941e-06,
+      "loss": 1.41881914,
+      "memory(GiB)": 103.58,
+      "step": 19500,
+      "train_speed(iter/s)": 1.318401
+    },
+    {
+      "acc": 0.68760252,
+      "epoch": 0.4947995941146626,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.971963551798506e-06,
+      "loss": 1.34056883,
+      "memory(GiB)": 103.58,
+      "step": 19505,
+      "train_speed(iter/s)": 1.318438
+    },
+    {
+      "acc": 0.68277826,
+      "epoch": 0.49492643328259767,
+      "grad_norm": 3.25,
+      "learning_rate": 8.971326525713378e-06,
+      "loss": 1.33998365,
+      "memory(GiB)": 103.58,
+      "step": 19510,
+      "train_speed(iter/s)": 1.318476
+    },
+    {
+      "acc": 0.6856822,
+      "epoch": 0.4950532724505327,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.97068932495158e-06,
+      "loss": 1.41674442,
+      "memory(GiB)": 103.58,
+      "step": 19515,
+      "train_speed(iter/s)": 1.318515
+    },
+    {
+      "acc": 0.67743979,
+      "epoch": 0.49518011161846776,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.970051949541137e-06,
+      "loss": 1.41237478,
+      "memory(GiB)": 103.58,
+      "step": 19520,
+      "train_speed(iter/s)": 1.318554
+    },
+    {
+      "acc": 0.68390827,
+      "epoch": 0.49530695078640286,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.969414399510085e-06,
+      "loss": 1.40137901,
+      "memory(GiB)": 103.58,
+      "step": 19525,
+      "train_speed(iter/s)": 1.318592
+    },
+    {
+      "acc": 0.68789005,
+      "epoch": 0.4954337899543379,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.968776674886466e-06,
+      "loss": 1.42594023,
+      "memory(GiB)": 103.58,
+      "step": 19530,
+      "train_speed(iter/s)": 1.318631
+    },
+    {
+      "acc": 0.66975317,
+      "epoch": 0.49556062912227294,
+      "grad_norm": 4.625,
+      "learning_rate": 8.968138775698328e-06,
+      "loss": 1.47759476,
+      "memory(GiB)": 103.58,
+      "step": 19535,
+      "train_speed(iter/s)": 1.31867
+    },
+    {
+      "acc": 0.69720759,
+      "epoch": 0.495687468290208,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.96750070197373e-06,
+      "loss": 1.37110672,
+      "memory(GiB)": 103.58,
+      "step": 19540,
+      "train_speed(iter/s)": 1.318708
+    },
+    {
+      "acc": 0.67497325,
+      "epoch": 0.4958143074581431,
+      "grad_norm": 3.25,
+      "learning_rate": 8.966862453740738e-06,
+      "loss": 1.39673271,
+      "memory(GiB)": 103.58,
+      "step": 19545,
+      "train_speed(iter/s)": 1.318746
+    },
+    {
+      "acc": 0.68385825,
+      "epoch": 0.49594114662607813,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.966224031027426e-06,
+      "loss": 1.44566784,
+      "memory(GiB)": 103.58,
+      "step": 19550,
+      "train_speed(iter/s)": 1.318782
+    },
+    {
+      "acc": 0.68259764,
+      "epoch": 0.4960679857940132,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.965585433861871e-06,
+      "loss": 1.41458416,
+      "memory(GiB)": 103.58,
+      "step": 19555,
+      "train_speed(iter/s)": 1.318821
+    },
+    {
+      "acc": 0.68412428,
+      "epoch": 0.4961948249619482,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.964946662272167e-06,
+      "loss": 1.37541771,
+      "memory(GiB)": 103.58,
+      "step": 19560,
+      "train_speed(iter/s)": 1.31886
+    },
+    {
+      "acc": 0.69732237,
+      "epoch": 0.4963216641298833,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.964307716286404e-06,
+      "loss": 1.35616837,
+      "memory(GiB)": 103.58,
+      "step": 19565,
+      "train_speed(iter/s)": 1.318897
+    },
+    {
+      "acc": 0.68887806,
+      "epoch": 0.49644850329781837,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.963668595932689e-06,
+      "loss": 1.43641644,
+      "memory(GiB)": 103.58,
+      "step": 19570,
+      "train_speed(iter/s)": 1.318934
+    },
+    {
+      "acc": 0.68298254,
+      "epoch": 0.4965753424657534,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.96302930123913e-06,
+      "loss": 1.41335669,
+      "memory(GiB)": 103.58,
+      "step": 19575,
+      "train_speed(iter/s)": 1.318971
+    },
+    {
+      "acc": 0.69458122,
+      "epoch": 0.49670218163368846,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.962389832233853e-06,
+      "loss": 1.37801352,
+      "memory(GiB)": 103.58,
+      "step": 19580,
+      "train_speed(iter/s)": 1.319009
+    },
+    {
+      "acc": 0.68197231,
+      "epoch": 0.49682902080162356,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.961750188944978e-06,
+      "loss": 1.4730546,
+      "memory(GiB)": 103.58,
+      "step": 19585,
+      "train_speed(iter/s)": 1.319048
+    },
+    {
+      "acc": 0.67737236,
+      "epoch": 0.4969558599695586,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.96111037140064e-06,
+      "loss": 1.4003994,
+      "memory(GiB)": 103.58,
+      "step": 19590,
+      "train_speed(iter/s)": 1.319081
+    },
+    {
+      "acc": 0.68623056,
+      "epoch": 0.49708269913749364,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.960470379628986e-06,
+      "loss": 1.3513258,
+      "memory(GiB)": 103.58,
+      "step": 19595,
+      "train_speed(iter/s)": 1.31912
+    },
+    {
+      "acc": 0.67791386,
+      "epoch": 0.4972095383054287,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.959830213658161e-06,
+      "loss": 1.40269423,
+      "memory(GiB)": 103.58,
+      "step": 19600,
+      "train_speed(iter/s)": 1.319157
+    },
+    {
+      "acc": 0.67989788,
+      "epoch": 0.4973363774733638,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.959189873516324e-06,
+      "loss": 1.45767746,
+      "memory(GiB)": 103.58,
+      "step": 19605,
+      "train_speed(iter/s)": 1.319195
+    },
+    {
+      "acc": 0.67374582,
+      "epoch": 0.49746321664129883,
+      "grad_norm": 2.75,
+      "learning_rate": 8.95854935923164e-06,
+      "loss": 1.47187309,
+      "memory(GiB)": 103.58,
+      "step": 19610,
+      "train_speed(iter/s)": 1.319234
+    },
+    {
+      "acc": 0.6916873,
+      "epoch": 0.4975900558092339,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.95790867083228e-06,
+      "loss": 1.46126909,
+      "memory(GiB)": 103.58,
+      "step": 19615,
+      "train_speed(iter/s)": 1.319272
+    },
+    {
+      "acc": 0.67325554,
+      "epoch": 0.4977168949771689,
+      "grad_norm": 3.5,
+      "learning_rate": 8.957267808346428e-06,
+      "loss": 1.44178114,
+      "memory(GiB)": 103.58,
+      "step": 19620,
+      "train_speed(iter/s)": 1.319304
+    },
+    {
+      "acc": 0.69150772,
+      "epoch": 0.497843734145104,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.95662677180227e-06,
+      "loss": 1.36337175,
+      "memory(GiB)": 103.58,
+      "step": 19625,
+      "train_speed(iter/s)": 1.31934
+    },
+    {
+      "acc": 0.70146685,
+      "epoch": 0.49797057331303907,
+      "grad_norm": 3.0,
+      "learning_rate": 8.955985561228e-06,
+      "loss": 1.37223339,
+      "memory(GiB)": 103.58,
+      "step": 19630,
+      "train_speed(iter/s)": 1.319377
+    },
+    {
+      "acc": 0.69182301,
+      "epoch": 0.4980974124809741,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.955344176651824e-06,
+      "loss": 1.39446602,
+      "memory(GiB)": 103.58,
+      "step": 19635,
+      "train_speed(iter/s)": 1.319414
+    },
+    {
+      "acc": 0.69342904,
+      "epoch": 0.49822425164890916,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.954702618101952e-06,
+      "loss": 1.37401381,
+      "memory(GiB)": 103.58,
+      "step": 19640,
+      "train_speed(iter/s)": 1.319452
+    },
+    {
+      "acc": 0.68022637,
+      "epoch": 0.49835109081684426,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.9540608856066e-06,
+      "loss": 1.41476946,
+      "memory(GiB)": 103.58,
+      "step": 19645,
+      "train_speed(iter/s)": 1.319488
+    },
+    {
+      "acc": 0.6761425,
+      "epoch": 0.4984779299847793,
+      "grad_norm": 4.125,
+      "learning_rate": 8.953418979194e-06,
+      "loss": 1.49609098,
+      "memory(GiB)": 103.58,
+      "step": 19650,
+      "train_speed(iter/s)": 1.319525
+    },
+    {
+      "acc": 0.69029808,
+      "epoch": 0.49860476915271434,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.95277689889238e-06,
+      "loss": 1.37234573,
+      "memory(GiB)": 103.58,
+      "step": 19655,
+      "train_speed(iter/s)": 1.319562
+    },
+    {
+      "acc": 0.6808197,
+      "epoch": 0.4987316083206494,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.952134644729985e-06,
+      "loss": 1.42074394,
+      "memory(GiB)": 103.58,
+      "step": 19660,
+      "train_speed(iter/s)": 1.319599
+    },
+    {
+      "acc": 0.68729038,
+      "epoch": 0.4988584474885845,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.951492216735062e-06,
+      "loss": 1.40622873,
+      "memory(GiB)": 103.58,
+      "step": 19665,
+      "train_speed(iter/s)": 1.319636
+    },
+    {
+      "acc": 0.67343764,
+      "epoch": 0.49898528665651953,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.950849614935872e-06,
+      "loss": 1.45426569,
+      "memory(GiB)": 103.58,
+      "step": 19670,
+      "train_speed(iter/s)": 1.319673
+    },
+    {
+      "acc": 0.67748556,
+      "epoch": 0.4991121258244546,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.950206839360674e-06,
+      "loss": 1.48211575,
+      "memory(GiB)": 103.58,
+      "step": 19675,
+      "train_speed(iter/s)": 1.31971
+    },
+    {
+      "acc": 0.674544,
+      "epoch": 0.4992389649923896,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.949563890037745e-06,
+      "loss": 1.43992691,
+      "memory(GiB)": 103.58,
+      "step": 19680,
+      "train_speed(iter/s)": 1.319745
+    },
+    {
+      "acc": 0.67195482,
+      "epoch": 0.4993658041603247,
+      "grad_norm": 3.0,
+      "learning_rate": 8.948920766995362e-06,
+      "loss": 1.44942665,
+      "memory(GiB)": 103.58,
+      "step": 19685,
+      "train_speed(iter/s)": 1.319783
+    },
+    {
+      "acc": 0.68367524,
+      "epoch": 0.49949264332825977,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.948277470261812e-06,
+      "loss": 1.38880205,
+      "memory(GiB)": 103.58,
+      "step": 19690,
+      "train_speed(iter/s)": 1.31982
+    },
+    {
+      "acc": 0.67365532,
+      "epoch": 0.4996194824961948,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.94763399986539e-06,
+      "loss": 1.42582588,
+      "memory(GiB)": 103.58,
+      "step": 19695,
+      "train_speed(iter/s)": 1.319856
+    },
+    {
+      "acc": 0.68712244,
+      "epoch": 0.49974632166412986,
+      "grad_norm": 2.625,
+      "learning_rate": 8.946990355834401e-06,
+      "loss": 1.37043953,
+      "memory(GiB)": 103.58,
+      "step": 19700,
+      "train_speed(iter/s)": 1.319893
+    },
+    {
+      "acc": 0.69154692,
+      "epoch": 0.49987316083206496,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.946346538197156e-06,
+      "loss": 1.35028687,
+      "memory(GiB)": 103.58,
+      "step": 19705,
+      "train_speed(iter/s)": 1.319932
+    },
+    {
+      "acc": 0.6770916,
+      "epoch": 0.5,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.94570254698197e-06,
+      "loss": 1.44391556,
+      "memory(GiB)": 103.58,
+      "step": 19710,
+      "train_speed(iter/s)": 1.319964
+    },
+    {
+      "acc": 0.67713594,
+      "epoch": 0.5001268391679351,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.945058382217168e-06,
+      "loss": 1.42051735,
+      "memory(GiB)": 103.58,
+      "step": 19715,
+      "train_speed(iter/s)": 1.320001
+    },
+    {
+      "acc": 0.67100992,
+      "epoch": 0.5002536783358701,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.944414043931086e-06,
+      "loss": 1.46472263,
+      "memory(GiB)": 103.58,
+      "step": 19720,
+      "train_speed(iter/s)": 1.320038
+    },
+    {
+      "acc": 0.68932557,
+      "epoch": 0.5003805175038052,
+      "grad_norm": 2.4375,
+      "learning_rate": 8.943769532152065e-06,
+      "loss": 1.36088772,
+      "memory(GiB)": 103.58,
+      "step": 19725,
+      "train_speed(iter/s)": 1.320075
+    },
+    {
+      "acc": 0.67515159,
+      "epoch": 0.5005073566717403,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.94312484690845e-06,
+      "loss": 1.4052722,
+      "memory(GiB)": 103.58,
+      "step": 19730,
+      "train_speed(iter/s)": 1.320114
+    },
+    {
+      "acc": 0.68254595,
+      "epoch": 0.5006341958396753,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.9424799882286e-06,
+      "loss": 1.40450869,
+      "memory(GiB)": 103.58,
+      "step": 19735,
+      "train_speed(iter/s)": 1.320151
+    },
+    {
+      "acc": 0.69900846,
+      "epoch": 0.5007610350076104,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.94183495614088e-06,
+      "loss": 1.37697983,
+      "memory(GiB)": 103.58,
+      "step": 19740,
+      "train_speed(iter/s)": 1.320187
+    },
+    {
+      "acc": 0.69765244,
+      "epoch": 0.5008878741755454,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.941189750673658e-06,
+      "loss": 1.35210447,
+      "memory(GiB)": 103.58,
+      "step": 19745,
+      "train_speed(iter/s)": 1.320223
+    },
+    {
+      "acc": 0.68402472,
+      "epoch": 0.5010147133434805,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.940544371855315e-06,
+      "loss": 1.39390144,
+      "memory(GiB)": 103.58,
+      "step": 19750,
+      "train_speed(iter/s)": 1.32026
+    },
+    {
+      "acc": 0.6947814,
+      "epoch": 0.5011415525114156,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.939898819714237e-06,
+      "loss": 1.31111336,
+      "memory(GiB)": 103.58,
+      "step": 19755,
+      "train_speed(iter/s)": 1.320298
+    },
+    {
+      "acc": 0.67771034,
+      "epoch": 0.5012683916793506,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.93925309427882e-06,
+      "loss": 1.36431923,
+      "memory(GiB)": 103.58,
+      "step": 19760,
+      "train_speed(iter/s)": 1.320336
+    },
+    {
+      "acc": 0.68143802,
+      "epoch": 0.5013952308472857,
+      "grad_norm": 3.125,
+      "learning_rate": 8.938607195577462e-06,
+      "loss": 1.44861069,
+      "memory(GiB)": 103.58,
+      "step": 19765,
+      "train_speed(iter/s)": 1.320373
+    },
+    {
+      "acc": 0.67330923,
+      "epoch": 0.5015220700152208,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.937961123638577e-06,
+      "loss": 1.45914497,
+      "memory(GiB)": 103.58,
+      "step": 19770,
+      "train_speed(iter/s)": 1.320411
+    },
+    {
+      "acc": 0.67617602,
+      "epoch": 0.5016489091831557,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.93731487849058e-06,
+      "loss": 1.42710114,
+      "memory(GiB)": 103.58,
+      "step": 19775,
+      "train_speed(iter/s)": 1.320448
+    },
+    {
+      "acc": 0.6768569,
+      "epoch": 0.5017757483510908,
+      "grad_norm": 3.875,
+      "learning_rate": 8.936668460161895e-06,
+      "loss": 1.44174929,
+      "memory(GiB)": 103.58,
+      "step": 19780,
+      "train_speed(iter/s)": 1.320486
+    },
+    {
+      "acc": 0.68371572,
+      "epoch": 0.5019025875190258,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.936021868680956e-06,
+      "loss": 1.46516075,
+      "memory(GiB)": 103.58,
+      "step": 19785,
+      "train_speed(iter/s)": 1.320518
+    },
+    {
+      "acc": 0.67560477,
+      "epoch": 0.5020294266869609,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.935375104076201e-06,
+      "loss": 1.46395645,
+      "memory(GiB)": 103.58,
+      "step": 19790,
+      "train_speed(iter/s)": 1.320555
+    },
+    {
+      "acc": 0.69070458,
+      "epoch": 0.502156265854896,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.93472816637608e-06,
+      "loss": 1.39218521,
+      "memory(GiB)": 103.58,
+      "step": 19795,
+      "train_speed(iter/s)": 1.320593
+    },
+    {
+      "acc": 0.69181395,
+      "epoch": 0.502283105022831,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.934081055609046e-06,
+      "loss": 1.41391125,
+      "memory(GiB)": 103.58,
+      "step": 19800,
+      "train_speed(iter/s)": 1.32063
+    },
+    {
+      "acc": 0.69568396,
+      "epoch": 0.5024099441907661,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.933433771803562e-06,
+      "loss": 1.36580744,
+      "memory(GiB)": 103.58,
+      "step": 19805,
+      "train_speed(iter/s)": 1.320667
+    },
+    {
+      "acc": 0.68727455,
+      "epoch": 0.5025367833587012,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.932786314988099e-06,
+      "loss": 1.40004721,
+      "memory(GiB)": 103.58,
+      "step": 19810,
+      "train_speed(iter/s)": 1.320703
+    },
+    {
+      "acc": 0.68932033,
+      "epoch": 0.5026636225266362,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.932138685191136e-06,
+      "loss": 1.38613796,
+      "memory(GiB)": 103.58,
+      "step": 19815,
+      "train_speed(iter/s)": 1.320741
+    },
+    {
+      "acc": 0.69823637,
+      "epoch": 0.5027904616945713,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.931490882441159e-06,
+      "loss": 1.35943069,
+      "memory(GiB)": 103.58,
+      "step": 19820,
+      "train_speed(iter/s)": 1.320779
+    },
+    {
+      "acc": 0.67956586,
+      "epoch": 0.5029173008625063,
+      "grad_norm": 3.953125,
+      "learning_rate": 8.930842906766659e-06,
+      "loss": 1.44133654,
+      "memory(GiB)": 103.58,
+      "step": 19825,
+      "train_speed(iter/s)": 1.320816
+    },
+    {
+      "acc": 0.6897748,
+      "epoch": 0.5030441400304414,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.930194758196138e-06,
+      "loss": 1.38044262,
+      "memory(GiB)": 103.58,
+      "step": 19830,
+      "train_speed(iter/s)": 1.320852
+    },
+    {
+      "acc": 0.6773046,
+      "epoch": 0.5031709791983765,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.929546436758105e-06,
+      "loss": 1.45258989,
+      "memory(GiB)": 103.58,
+      "step": 19835,
+      "train_speed(iter/s)": 1.320889
+    },
+    {
+      "acc": 0.68502522,
+      "epoch": 0.5032978183663115,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.928897942481075e-06,
+      "loss": 1.36830702,
+      "memory(GiB)": 103.58,
+      "step": 19840,
+      "train_speed(iter/s)": 1.320926
+    },
+    {
+      "acc": 0.67670064,
+      "epoch": 0.5034246575342466,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.928249275393572e-06,
+      "loss": 1.38597984,
+      "memory(GiB)": 103.58,
+      "step": 19845,
+      "train_speed(iter/s)": 1.320965
+    },
+    {
+      "acc": 0.68603997,
+      "epoch": 0.5035514967021817,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.927600435524129e-06,
+      "loss": 1.35848036,
+      "memory(GiB)": 103.58,
+      "step": 19850,
+      "train_speed(iter/s)": 1.321001
+    },
+    {
+      "acc": 0.68135891,
+      "epoch": 0.5036783358701167,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.926951422901282e-06,
+      "loss": 1.4259819,
+      "memory(GiB)": 103.58,
+      "step": 19855,
+      "train_speed(iter/s)": 1.321038
+    },
+    {
+      "acc": 0.69628468,
+      "epoch": 0.5038051750380518,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.926302237553578e-06,
+      "loss": 1.36217995,
+      "memory(GiB)": 103.58,
+      "step": 19860,
+      "train_speed(iter/s)": 1.321076
+    },
+    {
+      "acc": 0.67604909,
+      "epoch": 0.5039320142059868,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.925652879509575e-06,
+      "loss": 1.39901466,
+      "memory(GiB)": 103.58,
+      "step": 19865,
+      "train_speed(iter/s)": 1.321113
+    },
+    {
+      "acc": 0.69357762,
+      "epoch": 0.5040588533739219,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.925003348797829e-06,
+      "loss": 1.45051012,
+      "memory(GiB)": 103.58,
+      "step": 19870,
+      "train_speed(iter/s)": 1.321151
+    },
+    {
+      "acc": 0.68117981,
+      "epoch": 0.504185692541857,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.924353645446912e-06,
+      "loss": 1.35221586,
+      "memory(GiB)": 103.58,
+      "step": 19875,
+      "train_speed(iter/s)": 1.321188
+    },
+    {
+      "acc": 0.68166676,
+      "epoch": 0.504312531709792,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.923703769485403e-06,
+      "loss": 1.4189971,
+      "memory(GiB)": 103.58,
+      "step": 19880,
+      "train_speed(iter/s)": 1.321225
+    },
+    {
+      "acc": 0.68821335,
+      "epoch": 0.5044393708777271,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.92305372094188e-06,
+      "loss": 1.42541542,
+      "memory(GiB)": 103.58,
+      "step": 19885,
+      "train_speed(iter/s)": 1.321263
+    },
+    {
+      "acc": 0.69985132,
+      "epoch": 0.5045662100456622,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.922403499844943e-06,
+      "loss": 1.35402412,
+      "memory(GiB)": 103.58,
+      "step": 19890,
+      "train_speed(iter/s)": 1.321299
+    },
+    {
+      "acc": 0.68068972,
+      "epoch": 0.5046930492135971,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.921753106223186e-06,
+      "loss": 1.37659225,
+      "memory(GiB)": 103.58,
+      "step": 19895,
+      "train_speed(iter/s)": 1.321335
+    },
+    {
+      "acc": 0.65821629,
+      "epoch": 0.5048198883815322,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.92110254010522e-06,
+      "loss": 1.55344334,
+      "memory(GiB)": 103.58,
+      "step": 19900,
+      "train_speed(iter/s)": 1.321371
+    },
+    {
+      "acc": 0.67048407,
+      "epoch": 0.5049467275494672,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.920451801519656e-06,
+      "loss": 1.47370749,
+      "memory(GiB)": 103.58,
+      "step": 19905,
+      "train_speed(iter/s)": 1.321407
+    },
+    {
+      "acc": 0.68560457,
+      "epoch": 0.5050735667174023,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.919800890495118e-06,
+      "loss": 1.39053745,
+      "memory(GiB)": 103.58,
+      "step": 19910,
+      "train_speed(iter/s)": 1.321443
+    },
+    {
+      "acc": 0.67931237,
+      "epoch": 0.5052004058853374,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.919149807060237e-06,
+      "loss": 1.4563693,
+      "memory(GiB)": 103.58,
+      "step": 19915,
+      "train_speed(iter/s)": 1.321479
+    },
+    {
+      "acc": 0.67678657,
+      "epoch": 0.5053272450532724,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.918498551243649e-06,
+      "loss": 1.43937826,
+      "memory(GiB)": 103.58,
+      "step": 19920,
+      "train_speed(iter/s)": 1.321515
+    },
+    {
+      "acc": 0.67964082,
+      "epoch": 0.5054540842212075,
+      "grad_norm": 2.75,
+      "learning_rate": 8.917847123073999e-06,
+      "loss": 1.40929546,
+      "memory(GiB)": 103.58,
+      "step": 19925,
+      "train_speed(iter/s)": 1.321551
+    },
+    {
+      "acc": 0.67695112,
+      "epoch": 0.5055809233891426,
+      "grad_norm": 5.09375,
+      "learning_rate": 8.917195522579943e-06,
+      "loss": 1.37833977,
+      "memory(GiB)": 103.58,
+      "step": 19930,
+      "train_speed(iter/s)": 1.321585
+    },
+    {
+      "acc": 0.68490372,
+      "epoch": 0.5057077625570776,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.916543749790139e-06,
+      "loss": 1.4179759,
+      "memory(GiB)": 103.58,
+      "step": 19935,
+      "train_speed(iter/s)": 1.321622
+    },
+    {
+      "acc": 0.68138323,
+      "epoch": 0.5058346017250127,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.915891804733253e-06,
+      "loss": 1.44227219,
+      "memory(GiB)": 103.58,
+      "step": 19940,
+      "train_speed(iter/s)": 1.321659
+    },
+    {
+      "acc": 0.69010615,
+      "epoch": 0.5059614408929477,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.915239687437963e-06,
+      "loss": 1.37203722,
+      "memory(GiB)": 103.58,
+      "step": 19945,
+      "train_speed(iter/s)": 1.321694
+    },
+    {
+      "acc": 0.68169112,
+      "epoch": 0.5060882800608828,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.91458739793295e-06,
+      "loss": 1.38727417,
+      "memory(GiB)": 103.58,
+      "step": 19950,
+      "train_speed(iter/s)": 1.321296
+    },
+    {
+      "acc": 0.68261094,
+      "epoch": 0.5062151192288179,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.913934936246907e-06,
+      "loss": 1.44867611,
+      "memory(GiB)": 103.58,
+      "step": 19955,
+      "train_speed(iter/s)": 1.321333
+    },
+    {
+      "acc": 0.67004213,
+      "epoch": 0.5063419583967529,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.91328230240853e-06,
+      "loss": 1.48601818,
+      "memory(GiB)": 103.58,
+      "step": 19960,
+      "train_speed(iter/s)": 1.321365
+    },
+    {
+      "acc": 0.6899467,
+      "epoch": 0.506468797564688,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.912629496446528e-06,
+      "loss": 1.3361002,
+      "memory(GiB)": 103.58,
+      "step": 19965,
+      "train_speed(iter/s)": 1.321402
+    },
+    {
+      "acc": 0.69107261,
+      "epoch": 0.5065956367326231,
+      "grad_norm": 4.25,
+      "learning_rate": 8.911976518389612e-06,
+      "loss": 1.3167161,
+      "memory(GiB)": 103.58,
+      "step": 19970,
+      "train_speed(iter/s)": 1.321438
+    },
+    {
+      "acc": 0.69911013,
+      "epoch": 0.5067224759005581,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.9113233682665e-06,
+      "loss": 1.36770706,
+      "memory(GiB)": 103.58,
+      "step": 19975,
+      "train_speed(iter/s)": 1.321475
+    },
+    {
+      "acc": 0.68481083,
+      "epoch": 0.5068493150684932,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.910670046105927e-06,
+      "loss": 1.39013805,
+      "memory(GiB)": 103.58,
+      "step": 19980,
+      "train_speed(iter/s)": 1.32151
+    },
+    {
+      "acc": 0.68348656,
+      "epoch": 0.5069761542364282,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.910016551936623e-06,
+      "loss": 1.42333355,
+      "memory(GiB)": 103.58,
+      "step": 19985,
+      "train_speed(iter/s)": 1.321544
+    },
+    {
+      "acc": 0.67442675,
+      "epoch": 0.5071029934043633,
+      "grad_norm": 3.25,
+      "learning_rate": 8.909362885787333e-06,
+      "loss": 1.43964691,
+      "memory(GiB)": 103.58,
+      "step": 19990,
+      "train_speed(iter/s)": 1.321579
+    },
+    {
+      "acc": 0.67087212,
+      "epoch": 0.5072298325722984,
+      "grad_norm": 2.875,
+      "learning_rate": 8.908709047686813e-06,
+      "loss": 1.47225695,
+      "memory(GiB)": 103.58,
+      "step": 19995,
+      "train_speed(iter/s)": 1.321612
+    },
+    {
+      "acc": 0.69526515,
+      "epoch": 0.5073566717402334,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.908055037663815e-06,
+      "loss": 1.39258652,
+      "memory(GiB)": 103.58,
+      "step": 20000,
+      "train_speed(iter/s)": 1.321648
+    },
+    {
+      "epoch": 0.5073566717402334,
+      "eval_acc": 0.6732945326057674,
+      "eval_loss": 1.3660004138946533,
+      "eval_runtime": 69.5257,
+      "eval_samples_per_second": 91.621,
+      "eval_steps_per_second": 22.912,
+      "step": 20000
+    },
+    {
+      "acc": 0.69536238,
+      "epoch": 0.5074835109081685,
+      "grad_norm": 3.5,
+      "learning_rate": 8.907400855747111e-06,
+      "loss": 1.40485535,
+      "memory(GiB)": 103.58,
+      "step": 20005,
+      "train_speed(iter/s)": 1.314547
+    },
+    {
+      "acc": 0.6547143,
+      "epoch": 0.5076103500761036,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.906746501965468e-06,
+      "loss": 1.52191963,
+      "memory(GiB)": 103.58,
+      "step": 20010,
+      "train_speed(iter/s)": 1.314584
+    },
+    {
+      "acc": 0.70307856,
+      "epoch": 0.5077371892440385,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.906091976347675e-06,
+      "loss": 1.32399635,
+      "memory(GiB)": 103.58,
+      "step": 20015,
+      "train_speed(iter/s)": 1.314622
+    },
+    {
+      "acc": 0.67099152,
+      "epoch": 0.5078640284119736,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.905437278922516e-06,
+      "loss": 1.40839691,
+      "memory(GiB)": 103.58,
+      "step": 20020,
+      "train_speed(iter/s)": 1.314661
+    },
+    {
+      "acc": 0.68867559,
+      "epoch": 0.5079908675799086,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.90478240971879e-06,
+      "loss": 1.37977867,
+      "memory(GiB)": 103.58,
+      "step": 20025,
+      "train_speed(iter/s)": 1.314698
+    },
+    {
+      "acc": 0.6880125,
+      "epoch": 0.5081177067478437,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.904127368765298e-06,
+      "loss": 1.41413507,
+      "memory(GiB)": 103.58,
+      "step": 20030,
+      "train_speed(iter/s)": 1.314734
+    },
+    {
+      "acc": 0.67695675,
+      "epoch": 0.5082445459157788,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.903472156090856e-06,
+      "loss": 1.4020896,
+      "memory(GiB)": 103.58,
+      "step": 20035,
+      "train_speed(iter/s)": 1.314773
+    },
+    {
+      "acc": 0.69192705,
+      "epoch": 0.5083713850837138,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.902816771724279e-06,
+      "loss": 1.35355463,
+      "memory(GiB)": 103.58,
+      "step": 20040,
+      "train_speed(iter/s)": 1.314812
+    },
+    {
+      "acc": 0.68043566,
+      "epoch": 0.5084982242516489,
+      "grad_norm": 3.0,
+      "learning_rate": 8.902161215694396e-06,
+      "loss": 1.39213524,
+      "memory(GiB)": 103.58,
+      "step": 20045,
+      "train_speed(iter/s)": 1.314851
+    },
+    {
+      "acc": 0.68923969,
+      "epoch": 0.508625063419584,
+      "grad_norm": 3.0,
+      "learning_rate": 8.901505488030042e-06,
+      "loss": 1.40127974,
+      "memory(GiB)": 103.58,
+      "step": 20050,
+      "train_speed(iter/s)": 1.314889
+    },
+    {
+      "acc": 0.66122713,
+      "epoch": 0.508751902587519,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.900849588760057e-06,
+      "loss": 1.3895731,
+      "memory(GiB)": 103.58,
+      "step": 20055,
+      "train_speed(iter/s)": 1.314927
+    },
+    {
+      "acc": 0.66581812,
+      "epoch": 0.5088787417554541,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.90019351791329e-06,
+      "loss": 1.44869165,
+      "memory(GiB)": 103.58,
+      "step": 20060,
+      "train_speed(iter/s)": 1.314964
+    },
+    {
+      "acc": 0.68517742,
+      "epoch": 0.5090055809233891,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.8995372755186e-06,
+      "loss": 1.3651947,
+      "memory(GiB)": 103.58,
+      "step": 20065,
+      "train_speed(iter/s)": 1.315003
+    },
+    {
+      "acc": 0.67956929,
+      "epoch": 0.5091324200913242,
+      "grad_norm": 4.5,
+      "learning_rate": 8.89888086160485e-06,
+      "loss": 1.37972097,
+      "memory(GiB)": 103.58,
+      "step": 20070,
+      "train_speed(iter/s)": 1.31504
+    },
+    {
+      "acc": 0.68211412,
+      "epoch": 0.5092592592592593,
+      "grad_norm": 3.0,
+      "learning_rate": 8.898224276200913e-06,
+      "loss": 1.43925905,
+      "memory(GiB)": 103.58,
+      "step": 20075,
+      "train_speed(iter/s)": 1.315078
+    },
+    {
+      "acc": 0.69111261,
+      "epoch": 0.5093860984271943,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.897567519335669e-06,
+      "loss": 1.3797308,
+      "memory(GiB)": 103.58,
+      "step": 20080,
+      "train_speed(iter/s)": 1.315114
+    },
+    {
+      "acc": 0.68760834,
+      "epoch": 0.5095129375951294,
+      "grad_norm": 3.25,
+      "learning_rate": 8.896910591038002e-06,
+      "loss": 1.36849413,
+      "memory(GiB)": 103.58,
+      "step": 20085,
+      "train_speed(iter/s)": 1.315138
+    },
+    {
+      "acc": 0.67603922,
+      "epoch": 0.5096397767630645,
+      "grad_norm": 4.375,
+      "learning_rate": 8.896253491336809e-06,
+      "loss": 1.46056557,
+      "memory(GiB)": 103.58,
+      "step": 20090,
+      "train_speed(iter/s)": 1.315177
+    },
+    {
+      "acc": 0.67441535,
+      "epoch": 0.5097666159309995,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.895596220260993e-06,
+      "loss": 1.41351757,
+      "memory(GiB)": 103.58,
+      "step": 20095,
+      "train_speed(iter/s)": 1.315213
+    },
+    {
+      "acc": 0.68093166,
+      "epoch": 0.5098934550989346,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.894938777839462e-06,
+      "loss": 1.44270792,
+      "memory(GiB)": 103.58,
+      "step": 20100,
+      "train_speed(iter/s)": 1.315247
+    },
+    {
+      "acc": 0.68876677,
+      "epoch": 0.5100202942668696,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.894281164101134e-06,
+      "loss": 1.39160347,
+      "memory(GiB)": 103.58,
+      "step": 20105,
+      "train_speed(iter/s)": 1.315282
+    },
+    {
+      "acc": 0.66614599,
+      "epoch": 0.5101471334348047,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.893623379074934e-06,
+      "loss": 1.42026443,
+      "memory(GiB)": 103.58,
+      "step": 20110,
+      "train_speed(iter/s)": 1.315318
+    },
+    {
+      "acc": 0.68642416,
+      "epoch": 0.5102739726027398,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.892965422789793e-06,
+      "loss": 1.42294083,
+      "memory(GiB)": 103.58,
+      "step": 20115,
+      "train_speed(iter/s)": 1.315357
+    },
+    {
+      "acc": 0.68859305,
+      "epoch": 0.5104008117706748,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.892307295274654e-06,
+      "loss": 1.4290699,
+      "memory(GiB)": 103.58,
+      "step": 20120,
+      "train_speed(iter/s)": 1.315394
+    },
+    {
+      "acc": 0.68349037,
+      "epoch": 0.5105276509386099,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.89164899655846e-06,
+      "loss": 1.41428242,
+      "memory(GiB)": 103.58,
+      "step": 20125,
+      "train_speed(iter/s)": 1.315432
+    },
+    {
+      "acc": 0.69983964,
+      "epoch": 0.510654490106545,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.89099052667017e-06,
+      "loss": 1.3179265,
+      "memory(GiB)": 103.58,
+      "step": 20130,
+      "train_speed(iter/s)": 1.31547
+    },
+    {
+      "acc": 0.679638,
+      "epoch": 0.51078132927448,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.890331885638744e-06,
+      "loss": 1.43334198,
+      "memory(GiB)": 103.58,
+      "step": 20135,
+      "train_speed(iter/s)": 1.315507
+    },
+    {
+      "acc": 0.68820705,
+      "epoch": 0.510908168442415,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.889673073493151e-06,
+      "loss": 1.37662144,
+      "memory(GiB)": 103.58,
+      "step": 20140,
+      "train_speed(iter/s)": 1.315543
+    },
+    {
+      "acc": 0.68162513,
+      "epoch": 0.51103500761035,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.889014090262371e-06,
+      "loss": 1.46433353,
+      "memory(GiB)": 103.58,
+      "step": 20145,
+      "train_speed(iter/s)": 1.31558
+    },
+    {
+      "acc": 0.6835958,
+      "epoch": 0.5111618467782851,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.888354935975388e-06,
+      "loss": 1.3520216,
+      "memory(GiB)": 103.58,
+      "step": 20150,
+      "train_speed(iter/s)": 1.315613
+    },
+    {
+      "acc": 0.68839836,
+      "epoch": 0.5112886859462202,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.887695610661196e-06,
+      "loss": 1.36175594,
+      "memory(GiB)": 103.58,
+      "step": 20155,
+      "train_speed(iter/s)": 1.315647
+    },
+    {
+      "acc": 0.68630581,
+      "epoch": 0.5114155251141552,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.887036114348792e-06,
+      "loss": 1.44853973,
+      "memory(GiB)": 103.58,
+      "step": 20160,
+      "train_speed(iter/s)": 1.315686
+    },
+    {
+      "acc": 0.69909716,
+      "epoch": 0.5115423642820903,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.886376447067186e-06,
+      "loss": 1.42078648,
+      "memory(GiB)": 103.58,
+      "step": 20165,
+      "train_speed(iter/s)": 1.315725
+    },
+    {
+      "acc": 0.67946591,
+      "epoch": 0.5116692034500254,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.885716608845394e-06,
+      "loss": 1.41552114,
+      "memory(GiB)": 103.58,
+      "step": 20170,
+      "train_speed(iter/s)": 1.315764
+    },
+    {
+      "acc": 0.68977203,
+      "epoch": 0.5117960426179604,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.885056599712436e-06,
+      "loss": 1.369133,
+      "memory(GiB)": 103.58,
+      "step": 20175,
+      "train_speed(iter/s)": 1.315802
+    },
+    {
+      "acc": 0.6799902,
+      "epoch": 0.5119228817858955,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.884396419697343e-06,
+      "loss": 1.42164469,
+      "memory(GiB)": 103.58,
+      "step": 20180,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.68438168,
+      "epoch": 0.5120497209538305,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.883736068829151e-06,
+      "loss": 1.3823103,
+      "memory(GiB)": 103.58,
+      "step": 20185,
+      "train_speed(iter/s)": 1.315877
+    },
+    {
+      "acc": 0.68367038,
+      "epoch": 0.5121765601217656,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.88307554713691e-06,
+      "loss": 1.37781391,
+      "memory(GiB)": 103.58,
+      "step": 20190,
+      "train_speed(iter/s)": 1.315915
+    },
+    {
+      "acc": 0.69900293,
+      "epoch": 0.5123033992897007,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.882414854649667e-06,
+      "loss": 1.37848206,
+      "memory(GiB)": 103.58,
+      "step": 20195,
+      "train_speed(iter/s)": 1.315952
+    },
+    {
+      "acc": 0.69032774,
+      "epoch": 0.5124302384576357,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.881753991396488e-06,
+      "loss": 1.4321352,
+      "memory(GiB)": 103.58,
+      "step": 20200,
+      "train_speed(iter/s)": 1.315991
+    },
+    {
+      "acc": 0.68949981,
+      "epoch": 0.5125570776255708,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.881092957406436e-06,
+      "loss": 1.35196886,
+      "memory(GiB)": 103.58,
+      "step": 20205,
+      "train_speed(iter/s)": 1.316029
+    },
+    {
+      "acc": 0.68976016,
+      "epoch": 0.5126839167935059,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.880431752708588e-06,
+      "loss": 1.34440804,
+      "memory(GiB)": 103.58,
+      "step": 20210,
+      "train_speed(iter/s)": 1.316058
+    },
+    {
+      "acc": 0.70081062,
+      "epoch": 0.5128107559614409,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.879770377332026e-06,
+      "loss": 1.33325005,
+      "memory(GiB)": 103.58,
+      "step": 20215,
+      "train_speed(iter/s)": 1.316095
+    },
+    {
+      "acc": 0.6867353,
+      "epoch": 0.512937595129376,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.879108831305842e-06,
+      "loss": 1.39134731,
+      "memory(GiB)": 103.58,
+      "step": 20220,
+      "train_speed(iter/s)": 1.31613
+    },
+    {
+      "acc": 0.67811284,
+      "epoch": 0.513064434297311,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.878447114659131e-06,
+      "loss": 1.41253767,
+      "memory(GiB)": 103.58,
+      "step": 20225,
+      "train_speed(iter/s)": 1.316165
+    },
+    {
+      "acc": 0.70546188,
+      "epoch": 0.5131912734652461,
+      "grad_norm": 4.125,
+      "learning_rate": 8.877785227421003e-06,
+      "loss": 1.29743137,
+      "memory(GiB)": 103.58,
+      "step": 20230,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.69788637,
+      "epoch": 0.5133181126331812,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.877123169620565e-06,
+      "loss": 1.39363842,
+      "memory(GiB)": 103.58,
+      "step": 20235,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.67797508,
+      "epoch": 0.5134449518011162,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.876460941286941e-06,
+      "loss": 1.41712332,
+      "memory(GiB)": 103.58,
+      "step": 20240,
+      "train_speed(iter/s)": 1.316277
+    },
+    {
+      "acc": 0.67344522,
+      "epoch": 0.5135717909690513,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.875798542449256e-06,
+      "loss": 1.44520645,
+      "memory(GiB)": 103.58,
+      "step": 20245,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.66770678,
+      "epoch": 0.5136986301369864,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.87513597313665e-06,
+      "loss": 1.45116882,
+      "memory(GiB)": 103.58,
+      "step": 20250,
+      "train_speed(iter/s)": 1.316336
+    },
+    {
+      "acc": 0.6757863,
+      "epoch": 0.5138254693049213,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.87447323337826e-06,
+      "loss": 1.45002422,
+      "memory(GiB)": 103.58,
+      "step": 20255,
+      "train_speed(iter/s)": 1.316375
+    },
+    {
+      "acc": 0.69771042,
+      "epoch": 0.5139523084728564,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.87381032320324e-06,
+      "loss": 1.36454506,
+      "memory(GiB)": 103.58,
+      "step": 20260,
+      "train_speed(iter/s)": 1.316413
+    },
+    {
+      "acc": 0.68212509,
+      "epoch": 0.5140791476407914,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.873147242640746e-06,
+      "loss": 1.40420723,
+      "memory(GiB)": 103.58,
+      "step": 20265,
+      "train_speed(iter/s)": 1.316451
+    },
+    {
+      "acc": 0.68877645,
+      "epoch": 0.5142059868087265,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.872483991719944e-06,
+      "loss": 1.43697214,
+      "memory(GiB)": 103.58,
+      "step": 20270,
+      "train_speed(iter/s)": 1.316489
+    },
+    {
+      "acc": 0.67461076,
+      "epoch": 0.5143328259766616,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.871820570470009e-06,
+      "loss": 1.44258862,
+      "memory(GiB)": 103.58,
+      "step": 20275,
+      "train_speed(iter/s)": 1.316529
+    },
+    {
+      "acc": 0.68933115,
+      "epoch": 0.5144596651445966,
+      "grad_norm": 5.96875,
+      "learning_rate": 8.871156978920116e-06,
+      "loss": 1.42969055,
+      "memory(GiB)": 103.58,
+      "step": 20280,
+      "train_speed(iter/s)": 1.316568
+    },
+    {
+      "acc": 0.69213657,
+      "epoch": 0.5145865043125317,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.870493217099456e-06,
+      "loss": 1.29365625,
+      "memory(GiB)": 103.58,
+      "step": 20285,
+      "train_speed(iter/s)": 1.316603
+    },
+    {
+      "acc": 0.68537364,
+      "epoch": 0.5147133434804668,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.869829285037224e-06,
+      "loss": 1.31943398,
+      "memory(GiB)": 103.58,
+      "step": 20290,
+      "train_speed(iter/s)": 1.316641
+    },
+    {
+      "acc": 0.68374376,
+      "epoch": 0.5148401826484018,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.869165182762623e-06,
+      "loss": 1.4167511,
+      "memory(GiB)": 103.58,
+      "step": 20295,
+      "train_speed(iter/s)": 1.316678
+    },
+    {
+      "acc": 0.68310757,
+      "epoch": 0.5149670218163369,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.868500910304863e-06,
+      "loss": 1.44296246,
+      "memory(GiB)": 103.58,
+      "step": 20300,
+      "train_speed(iter/s)": 1.316716
+    },
+    {
+      "acc": 0.69016638,
+      "epoch": 0.5150938609842719,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.86783646769316e-06,
+      "loss": 1.37693577,
+      "memory(GiB)": 103.58,
+      "step": 20305,
+      "train_speed(iter/s)": 1.316754
+    },
+    {
+      "acc": 0.69234047,
+      "epoch": 0.515220700152207,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.867171854956742e-06,
+      "loss": 1.39432297,
+      "memory(GiB)": 103.58,
+      "step": 20310,
+      "train_speed(iter/s)": 1.316792
+    },
+    {
+      "acc": 0.68440385,
+      "epoch": 0.5153475393201421,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.86650707212484e-06,
+      "loss": 1.40261765,
+      "memory(GiB)": 103.58,
+      "step": 20315,
+      "train_speed(iter/s)": 1.316829
+    },
+    {
+      "acc": 0.68806372,
+      "epoch": 0.5154743784880771,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.865842119226693e-06,
+      "loss": 1.40339413,
+      "memory(GiB)": 103.58,
+      "step": 20320,
+      "train_speed(iter/s)": 1.316867
+    },
+    {
+      "acc": 0.67050495,
+      "epoch": 0.5156012176560122,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.86517699629155e-06,
+      "loss": 1.45585289,
+      "memory(GiB)": 103.58,
+      "step": 20325,
+      "train_speed(iter/s)": 1.316904
+    },
+    {
+      "acc": 0.67311931,
+      "epoch": 0.5157280568239473,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.864511703348666e-06,
+      "loss": 1.44937611,
+      "memory(GiB)": 103.58,
+      "step": 20330,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.67910585,
+      "epoch": 0.5158548959918823,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.863846240427306e-06,
+      "loss": 1.45888367,
+      "memory(GiB)": 103.58,
+      "step": 20335,
+      "train_speed(iter/s)": 1.3165
+    },
+    {
+      "acc": 0.69625607,
+      "epoch": 0.5159817351598174,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.863180607556733e-06,
+      "loss": 1.30763063,
+      "memory(GiB)": 103.58,
+      "step": 20340,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.68276844,
+      "epoch": 0.5161085743277524,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.862514804766234e-06,
+      "loss": 1.36270218,
+      "memory(GiB)": 103.58,
+      "step": 20345,
+      "train_speed(iter/s)": 1.316573
+    },
+    {
+      "acc": 0.6792686,
+      "epoch": 0.5162354134956875,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.861848832085084e-06,
+      "loss": 1.42033005,
+      "memory(GiB)": 103.58,
+      "step": 20350,
+      "train_speed(iter/s)": 1.316612
+    },
+    {
+      "acc": 0.68328476,
+      "epoch": 0.5163622526636226,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.861182689542585e-06,
+      "loss": 1.39333935,
+      "memory(GiB)": 103.58,
+      "step": 20355,
+      "train_speed(iter/s)": 1.316649
+    },
+    {
+      "acc": 0.69229326,
+      "epoch": 0.5164890918315576,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.86051637716803e-06,
+      "loss": 1.32554588,
+      "memory(GiB)": 103.58,
+      "step": 20360,
+      "train_speed(iter/s)": 1.316686
+    },
+    {
+      "acc": 0.67434015,
+      "epoch": 0.5166159309994927,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.859849894990728e-06,
+      "loss": 1.45331268,
+      "memory(GiB)": 103.58,
+      "step": 20365,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.68050604,
+      "epoch": 0.5167427701674278,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.859183243039995e-06,
+      "loss": 1.43713055,
+      "memory(GiB)": 103.58,
+      "step": 20370,
+      "train_speed(iter/s)": 1.31676
+    },
+    {
+      "acc": 0.66868963,
+      "epoch": 0.5168696093353627,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.858516421345154e-06,
+      "loss": 1.43922768,
+      "memory(GiB)": 103.58,
+      "step": 20375,
+      "train_speed(iter/s)": 1.316797
+    },
+    {
+      "acc": 0.68454981,
+      "epoch": 0.5169964485032978,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.857849429935534e-06,
+      "loss": 1.43399496,
+      "memory(GiB)": 103.58,
+      "step": 20380,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.68879471,
+      "epoch": 0.5171232876712328,
+      "grad_norm": 3.625,
+      "learning_rate": 8.857182268840471e-06,
+      "loss": 1.42619791,
+      "memory(GiB)": 103.58,
+      "step": 20385,
+      "train_speed(iter/s)": 1.316869
+    },
+    {
+      "acc": 0.67793074,
+      "epoch": 0.5172501268391679,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.856514938089312e-06,
+      "loss": 1.3864892,
+      "memory(GiB)": 103.58,
+      "step": 20390,
+      "train_speed(iter/s)": 1.316904
+    },
+    {
+      "acc": 0.67615075,
+      "epoch": 0.517376966007103,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.855847437711407e-06,
+      "loss": 1.49084759,
+      "memory(GiB)": 103.58,
+      "step": 20395,
+      "train_speed(iter/s)": 1.316941
+    },
+    {
+      "acc": 0.68855028,
+      "epoch": 0.517503805175038,
+      "grad_norm": 3.5,
+      "learning_rate": 8.855179767736117e-06,
+      "loss": 1.42371674,
+      "memory(GiB)": 103.58,
+      "step": 20400,
+      "train_speed(iter/s)": 1.316977
+    },
+    {
+      "acc": 0.68928661,
+      "epoch": 0.5176306443429731,
+      "grad_norm": 2.875,
+      "learning_rate": 8.854511928192808e-06,
+      "loss": 1.38536758,
+      "memory(GiB)": 103.58,
+      "step": 20405,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.67741337,
+      "epoch": 0.5177574835109082,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.853843919110856e-06,
+      "loss": 1.36607218,
+      "memory(GiB)": 103.58,
+      "step": 20410,
+      "train_speed(iter/s)": 1.317039
+    },
+    {
+      "acc": 0.68060656,
+      "epoch": 0.5178843226788432,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.853175740519642e-06,
+      "loss": 1.47884502,
+      "memory(GiB)": 103.58,
+      "step": 20415,
+      "train_speed(iter/s)": 1.317076
+    },
+    {
+      "acc": 0.69094739,
+      "epoch": 0.5180111618467783,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.852507392448555e-06,
+      "loss": 1.38871155,
+      "memory(GiB)": 103.58,
+      "step": 20420,
+      "train_speed(iter/s)": 1.31711
+    },
+    {
+      "acc": 0.68177257,
+      "epoch": 0.5181380010147133,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.851838874926994e-06,
+      "loss": 1.4166708,
+      "memory(GiB)": 103.58,
+      "step": 20425,
+      "train_speed(iter/s)": 1.317146
+    },
+    {
+      "acc": 0.68187356,
+      "epoch": 0.5182648401826484,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.851170187984362e-06,
+      "loss": 1.38627424,
+      "memory(GiB)": 103.58,
+      "step": 20430,
+      "train_speed(iter/s)": 1.316728
+    },
+    {
+      "acc": 0.68902831,
+      "epoch": 0.5183916793505835,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.850501331650069e-06,
+      "loss": 1.37234631,
+      "memory(GiB)": 103.58,
+      "step": 20435,
+      "train_speed(iter/s)": 1.316759
+    },
+    {
+      "acc": 0.68641491,
+      "epoch": 0.5185185185185185,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.849832305953536e-06,
+      "loss": 1.44701309,
+      "memory(GiB)": 103.58,
+      "step": 20440,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.68399086,
+      "epoch": 0.5186453576864536,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.849163110924193e-06,
+      "loss": 1.40543833,
+      "memory(GiB)": 103.58,
+      "step": 20445,
+      "train_speed(iter/s)": 1.316827
+    },
+    {
+      "acc": 0.67761221,
+      "epoch": 0.5187721968543887,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.84849374659147e-06,
+      "loss": 1.45825424,
+      "memory(GiB)": 103.58,
+      "step": 20450,
+      "train_speed(iter/s)": 1.316858
+    },
+    {
+      "acc": 0.6936615,
+      "epoch": 0.5188990360223237,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.847824212984807e-06,
+      "loss": 1.32823429,
+      "memory(GiB)": 103.58,
+      "step": 20455,
+      "train_speed(iter/s)": 1.316891
+    },
+    {
+      "acc": 0.69149647,
+      "epoch": 0.5190258751902588,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.84715451013366e-06,
+      "loss": 1.30649414,
+      "memory(GiB)": 103.58,
+      "step": 20460,
+      "train_speed(iter/s)": 1.316921
+    },
+    {
+      "acc": 0.68625526,
+      "epoch": 0.5191527143581938,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.846484638067478e-06,
+      "loss": 1.35889349,
+      "memory(GiB)": 103.58,
+      "step": 20465,
+      "train_speed(iter/s)": 1.316954
+    },
+    {
+      "acc": 0.68057451,
+      "epoch": 0.5192795535261289,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.84581459681573e-06,
+      "loss": 1.40935583,
+      "memory(GiB)": 103.58,
+      "step": 20470,
+      "train_speed(iter/s)": 1.316553
+    },
+    {
+      "acc": 0.69858274,
+      "epoch": 0.519406392694064,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.845144386407884e-06,
+      "loss": 1.40032444,
+      "memory(GiB)": 103.58,
+      "step": 20475,
+      "train_speed(iter/s)": 1.316587
+    },
+    {
+      "acc": 0.68484039,
+      "epoch": 0.519533231861999,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.844474006873422e-06,
+      "loss": 1.40718107,
+      "memory(GiB)": 103.58,
+      "step": 20480,
+      "train_speed(iter/s)": 1.316622
+    },
+    {
+      "acc": 0.69286284,
+      "epoch": 0.5196600710299341,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.84380345824183e-06,
+      "loss": 1.33025999,
+      "memory(GiB)": 103.58,
+      "step": 20485,
+      "train_speed(iter/s)": 1.316656
+    },
+    {
+      "acc": 0.68335323,
+      "epoch": 0.5197869101978692,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.843132740542599e-06,
+      "loss": 1.42429543,
+      "memory(GiB)": 103.58,
+      "step": 20490,
+      "train_speed(iter/s)": 1.316691
+    },
+    {
+      "acc": 0.69738836,
+      "epoch": 0.5199137493658041,
+      "grad_norm": 7.40625,
+      "learning_rate": 8.842461853805232e-06,
+      "loss": 1.30457392,
+      "memory(GiB)": 103.58,
+      "step": 20495,
+      "train_speed(iter/s)": 1.316725
+    },
+    {
+      "acc": 0.69089727,
+      "epoch": 0.5200405885337392,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.841790798059237e-06,
+      "loss": 1.3737196,
+      "memory(GiB)": 103.58,
+      "step": 20500,
+      "train_speed(iter/s)": 1.31676
+    },
+    {
+      "acc": 0.70438576,
+      "epoch": 0.5201674277016742,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.84111957333413e-06,
+      "loss": 1.31912508,
+      "memory(GiB)": 103.58,
+      "step": 20505,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.67368941,
+      "epoch": 0.5202942668696093,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.840448179659436e-06,
+      "loss": 1.53373985,
+      "memory(GiB)": 103.58,
+      "step": 20510,
+      "train_speed(iter/s)": 1.316828
+    },
+    {
+      "acc": 0.68094645,
+      "epoch": 0.5204211060375444,
+      "grad_norm": 2.75,
+      "learning_rate": 8.839776617064683e-06,
+      "loss": 1.41656742,
+      "memory(GiB)": 103.58,
+      "step": 20515,
+      "train_speed(iter/s)": 1.31686
+    },
+    {
+      "acc": 0.68244596,
+      "epoch": 0.5205479452054794,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.839104885579413e-06,
+      "loss": 1.38979025,
+      "memory(GiB)": 103.58,
+      "step": 20520,
+      "train_speed(iter/s)": 1.316893
+    },
+    {
+      "acc": 0.68936648,
+      "epoch": 0.5206747843734145,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.83843298523317e-06,
+      "loss": 1.45474396,
+      "memory(GiB)": 103.58,
+      "step": 20525,
+      "train_speed(iter/s)": 1.316928
+    },
+    {
+      "acc": 0.68469858,
+      "epoch": 0.5208016235413496,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.837760916055505e-06,
+      "loss": 1.41921444,
+      "memory(GiB)": 103.58,
+      "step": 20530,
+      "train_speed(iter/s)": 1.316967
+    },
+    {
+      "acc": 0.66852932,
+      "epoch": 0.5209284627092846,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.837088678075983e-06,
+      "loss": 1.49082842,
+      "memory(GiB)": 103.58,
+      "step": 20535,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.67854528,
+      "epoch": 0.5210553018772197,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.836416271324166e-06,
+      "loss": 1.42383299,
+      "memory(GiB)": 103.58,
+      "step": 20540,
+      "train_speed(iter/s)": 1.317038
+    },
+    {
+      "acc": 0.70065236,
+      "epoch": 0.5211821410451547,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.835743695829635e-06,
+      "loss": 1.31106253,
+      "memory(GiB)": 103.58,
+      "step": 20545,
+      "train_speed(iter/s)": 1.317073
+    },
+    {
+      "acc": 0.6726758,
+      "epoch": 0.5213089802130898,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.835070951621971e-06,
+      "loss": 1.43750496,
+      "memory(GiB)": 103.58,
+      "step": 20550,
+      "train_speed(iter/s)": 1.31711
+    },
+    {
+      "acc": 0.68377967,
+      "epoch": 0.5214358193810249,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.834398038730765e-06,
+      "loss": 1.39583988,
+      "memory(GiB)": 103.58,
+      "step": 20555,
+      "train_speed(iter/s)": 1.317147
+    },
+    {
+      "acc": 0.71670542,
+      "epoch": 0.5215626585489599,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.833724957185612e-06,
+      "loss": 1.27208443,
+      "memory(GiB)": 103.58,
+      "step": 20560,
+      "train_speed(iter/s)": 1.317184
+    },
+    {
+      "acc": 0.67353983,
+      "epoch": 0.521689497716895,
+      "grad_norm": 3.625,
+      "learning_rate": 8.83305170701612e-06,
+      "loss": 1.45635729,
+      "memory(GiB)": 103.58,
+      "step": 20565,
+      "train_speed(iter/s)": 1.317221
+    },
+    {
+      "acc": 0.68799896,
+      "epoch": 0.5218163368848301,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.832378288251902e-06,
+      "loss": 1.40825958,
+      "memory(GiB)": 103.58,
+      "step": 20570,
+      "train_speed(iter/s)": 1.317258
+    },
+    {
+      "acc": 0.68033829,
+      "epoch": 0.5219431760527651,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.831704700922574e-06,
+      "loss": 1.42829504,
+      "memory(GiB)": 103.58,
+      "step": 20575,
+      "train_speed(iter/s)": 1.317294
+    },
+    {
+      "acc": 0.68989391,
+      "epoch": 0.5220700152207002,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.831030945057767e-06,
+      "loss": 1.39665184,
+      "memory(GiB)": 103.58,
+      "step": 20580,
+      "train_speed(iter/s)": 1.317329
+    },
+    {
+      "acc": 0.70248036,
+      "epoch": 0.5221968543886352,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.830357020687115e-06,
+      "loss": 1.31259699,
+      "memory(GiB)": 103.58,
+      "step": 20585,
+      "train_speed(iter/s)": 1.317363
+    },
+    {
+      "acc": 0.69447556,
+      "epoch": 0.5223236935565703,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.82968292784026e-06,
+      "loss": 1.4010993,
+      "memory(GiB)": 103.58,
+      "step": 20590,
+      "train_speed(iter/s)": 1.317399
+    },
+    {
+      "acc": 0.67756763,
+      "epoch": 0.5224505327245054,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.82900866654685e-06,
+      "loss": 1.41036625,
+      "memory(GiB)": 103.58,
+      "step": 20595,
+      "train_speed(iter/s)": 1.317435
+    },
+    {
+      "acc": 0.68869233,
+      "epoch": 0.5225773718924404,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.828334236836546e-06,
+      "loss": 1.39497566,
+      "memory(GiB)": 103.58,
+      "step": 20600,
+      "train_speed(iter/s)": 1.317471
+    },
+    {
+      "acc": 0.67979269,
+      "epoch": 0.5227042110603755,
+      "grad_norm": 2.75,
+      "learning_rate": 8.827659638739007e-06,
+      "loss": 1.44060249,
+      "memory(GiB)": 103.58,
+      "step": 20605,
+      "train_speed(iter/s)": 1.317506
+    },
+    {
+      "acc": 0.68842201,
+      "epoch": 0.5228310502283106,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.82698487228391e-06,
+      "loss": 1.37045279,
+      "memory(GiB)": 103.58,
+      "step": 20610,
+      "train_speed(iter/s)": 1.317541
+    },
+    {
+      "acc": 0.67471914,
+      "epoch": 0.5229578893962455,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.826309937500932e-06,
+      "loss": 1.38433857,
+      "memory(GiB)": 103.58,
+      "step": 20615,
+      "train_speed(iter/s)": 1.317577
+    },
+    {
+      "acc": 0.68557286,
+      "epoch": 0.5230847285641806,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.825634834419758e-06,
+      "loss": 1.45918245,
+      "memory(GiB)": 103.58,
+      "step": 20620,
+      "train_speed(iter/s)": 1.317612
+    },
+    {
+      "acc": 0.69357138,
+      "epoch": 0.5232115677321156,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.824959563070085e-06,
+      "loss": 1.35061531,
+      "memory(GiB)": 103.58,
+      "step": 20625,
+      "train_speed(iter/s)": 1.317648
+    },
+    {
+      "acc": 0.69025755,
+      "epoch": 0.5233384069000507,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.824284123481614e-06,
+      "loss": 1.37607737,
+      "memory(GiB)": 103.58,
+      "step": 20630,
+      "train_speed(iter/s)": 1.317684
+    },
+    {
+      "acc": 0.68104324,
+      "epoch": 0.5234652460679858,
+      "grad_norm": 3.125,
+      "learning_rate": 8.823608515684053e-06,
+      "loss": 1.4378582,
+      "memory(GiB)": 103.58,
+      "step": 20635,
+      "train_speed(iter/s)": 1.317719
+    },
+    {
+      "acc": 0.67106266,
+      "epoch": 0.5235920852359208,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.822932739707118e-06,
+      "loss": 1.45749092,
+      "memory(GiB)": 103.58,
+      "step": 20640,
+      "train_speed(iter/s)": 1.317756
+    },
+    {
+      "acc": 0.67553515,
+      "epoch": 0.5237189244038559,
+      "grad_norm": 3.25,
+      "learning_rate": 8.822256795580532e-06,
+      "loss": 1.44946375,
+      "memory(GiB)": 103.58,
+      "step": 20645,
+      "train_speed(iter/s)": 1.317792
+    },
+    {
+      "acc": 0.68933086,
+      "epoch": 0.523845763571791,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.821580683334027e-06,
+      "loss": 1.39596062,
+      "memory(GiB)": 103.58,
+      "step": 20650,
+      "train_speed(iter/s)": 1.317829
+    },
+    {
+      "acc": 0.67909594,
+      "epoch": 0.523972602739726,
+      "grad_norm": 3.625,
+      "learning_rate": 8.820904402997343e-06,
+      "loss": 1.42768431,
+      "memory(GiB)": 103.58,
+      "step": 20655,
+      "train_speed(iter/s)": 1.317866
+    },
+    {
+      "acc": 0.68513136,
+      "epoch": 0.5240994419076611,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.820227954600222e-06,
+      "loss": 1.34684048,
+      "memory(GiB)": 103.58,
+      "step": 20660,
+      "train_speed(iter/s)": 1.317894
+    },
+    {
+      "acc": 0.67099266,
+      "epoch": 0.5242262810755961,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.819551338172421e-06,
+      "loss": 1.49144182,
+      "memory(GiB)": 103.58,
+      "step": 20665,
+      "train_speed(iter/s)": 1.317928
+    },
+    {
+      "acc": 0.69379921,
+      "epoch": 0.5243531202435312,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.8188745537437e-06,
+      "loss": 1.39757175,
+      "memory(GiB)": 103.58,
+      "step": 20670,
+      "train_speed(iter/s)": 1.317962
+    },
+    {
+      "acc": 0.68960609,
+      "epoch": 0.5244799594114663,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.818197601343822e-06,
+      "loss": 1.37982149,
+      "memory(GiB)": 103.58,
+      "step": 20675,
+      "train_speed(iter/s)": 1.317998
+    },
+    {
+      "acc": 0.68519521,
+      "epoch": 0.5246067985794013,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.81752048100257e-06,
+      "loss": 1.39286823,
+      "memory(GiB)": 103.58,
+      "step": 20680,
+      "train_speed(iter/s)": 1.318034
+    },
+    {
+      "acc": 0.6619411,
+      "epoch": 0.5247336377473364,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.816843192749724e-06,
+      "loss": 1.47724991,
+      "memory(GiB)": 103.58,
+      "step": 20685,
+      "train_speed(iter/s)": 1.318069
+    },
+    {
+      "acc": 0.68612432,
+      "epoch": 0.5248604769152715,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.816165736615072e-06,
+      "loss": 1.39727907,
+      "memory(GiB)": 103.58,
+      "step": 20690,
+      "train_speed(iter/s)": 1.318105
+    },
+    {
+      "acc": 0.68938656,
+      "epoch": 0.5249873160832065,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.815488112628412e-06,
+      "loss": 1.40109892,
+      "memory(GiB)": 103.58,
+      "step": 20695,
+      "train_speed(iter/s)": 1.318143
+    },
+    {
+      "acc": 0.6990757,
+      "epoch": 0.5251141552511416,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.814810320819551e-06,
+      "loss": 1.34248066,
+      "memory(GiB)": 103.58,
+      "step": 20700,
+      "train_speed(iter/s)": 1.318179
+    },
+    {
+      "acc": 0.6917078,
+      "epoch": 0.5252409944190766,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.8141323612183e-06,
+      "loss": 1.4068634,
+      "memory(GiB)": 103.58,
+      "step": 20705,
+      "train_speed(iter/s)": 1.318215
+    },
+    {
+      "acc": 0.67382984,
+      "epoch": 0.5253678335870117,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.813454233854479e-06,
+      "loss": 1.41959209,
+      "memory(GiB)": 103.58,
+      "step": 20710,
+      "train_speed(iter/s)": 1.318249
+    },
+    {
+      "acc": 0.69533548,
+      "epoch": 0.5254946727549468,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.812775938757914e-06,
+      "loss": 1.37885771,
+      "memory(GiB)": 103.58,
+      "step": 20715,
+      "train_speed(iter/s)": 1.318286
+    },
+    {
+      "acc": 0.70300837,
+      "epoch": 0.5256215119228818,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.812097475958442e-06,
+      "loss": 1.34932146,
+      "memory(GiB)": 103.58,
+      "step": 20720,
+      "train_speed(iter/s)": 1.31832
+    },
+    {
+      "acc": 0.67430506,
+      "epoch": 0.5257483510908169,
+      "grad_norm": 3.125,
+      "learning_rate": 8.811418845485902e-06,
+      "loss": 1.42645245,
+      "memory(GiB)": 103.58,
+      "step": 20725,
+      "train_speed(iter/s)": 1.318357
+    },
+    {
+      "acc": 0.69363103,
+      "epoch": 0.525875190258752,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.810740047370146e-06,
+      "loss": 1.42626028,
+      "memory(GiB)": 103.58,
+      "step": 20730,
+      "train_speed(iter/s)": 1.318392
+    },
+    {
+      "acc": 0.70138712,
+      "epoch": 0.526002029426687,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.810061081641026e-06,
+      "loss": 1.39790621,
+      "memory(GiB)": 103.58,
+      "step": 20735,
+      "train_speed(iter/s)": 1.318429
+    },
+    {
+      "acc": 0.69270935,
+      "epoch": 0.526128868594622,
+      "grad_norm": 2.5,
+      "learning_rate": 8.809381948328412e-06,
+      "loss": 1.384795,
+      "memory(GiB)": 103.58,
+      "step": 20740,
+      "train_speed(iter/s)": 1.318465
+    },
+    {
+      "acc": 0.68616514,
+      "epoch": 0.526255707762557,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.80870264746217e-06,
+      "loss": 1.42740326,
+      "memory(GiB)": 103.58,
+      "step": 20745,
+      "train_speed(iter/s)": 1.3185
+    },
+    {
+      "acc": 0.68192,
+      "epoch": 0.5263825469304921,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.808023179072183e-06,
+      "loss": 1.393643,
+      "memory(GiB)": 103.58,
+      "step": 20750,
+      "train_speed(iter/s)": 1.318534
+    },
+    {
+      "acc": 0.68251324,
+      "epoch": 0.5265093860984272,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.807343543188333e-06,
+      "loss": 1.46217747,
+      "memory(GiB)": 103.58,
+      "step": 20755,
+      "train_speed(iter/s)": 1.31857
+    },
+    {
+      "acc": 0.6989481,
+      "epoch": 0.5266362252663622,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.806663739840515e-06,
+      "loss": 1.39038792,
+      "memory(GiB)": 103.58,
+      "step": 20760,
+      "train_speed(iter/s)": 1.318606
+    },
+    {
+      "acc": 0.67511888,
+      "epoch": 0.5267630644342973,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.805983769058633e-06,
+      "loss": 1.42612162,
+      "memory(GiB)": 103.58,
+      "step": 20765,
+      "train_speed(iter/s)": 1.318636
+    },
+    {
+      "acc": 0.69828572,
+      "epoch": 0.5268899036022324,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.80530363087259e-06,
+      "loss": 1.39116354,
+      "memory(GiB)": 103.58,
+      "step": 20770,
+      "train_speed(iter/s)": 1.318673
+    },
+    {
+      "acc": 0.68162341,
+      "epoch": 0.5270167427701674,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.804623325312305e-06,
+      "loss": 1.45618649,
+      "memory(GiB)": 103.58,
+      "step": 20775,
+      "train_speed(iter/s)": 1.318707
+    },
+    {
+      "acc": 0.67842369,
+      "epoch": 0.5271435819381025,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.8039428524077e-06,
+      "loss": 1.38383694,
+      "memory(GiB)": 103.58,
+      "step": 20780,
+      "train_speed(iter/s)": 1.318743
+    },
+    {
+      "acc": 0.69120073,
+      "epoch": 0.5272704211060375,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.803262212188703e-06,
+      "loss": 1.45739393,
+      "memory(GiB)": 103.58,
+      "step": 20785,
+      "train_speed(iter/s)": 1.318778
+    },
+    {
+      "acc": 0.69271193,
+      "epoch": 0.5273972602739726,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.802581404685255e-06,
+      "loss": 1.40440817,
+      "memory(GiB)": 103.58,
+      "step": 20790,
+      "train_speed(iter/s)": 1.318814
+    },
+    {
+      "acc": 0.66418343,
+      "epoch": 0.5275240994419077,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.8019004299273e-06,
+      "loss": 1.44612532,
+      "memory(GiB)": 103.58,
+      "step": 20795,
+      "train_speed(iter/s)": 1.31885
+    },
+    {
+      "acc": 0.68350587,
+      "epoch": 0.5276509386098427,
+      "grad_norm": 4.125,
+      "learning_rate": 8.801219287944788e-06,
+      "loss": 1.41543388,
+      "memory(GiB)": 103.58,
+      "step": 20800,
+      "train_speed(iter/s)": 1.318887
+    },
+    {
+      "acc": 0.68398495,
+      "epoch": 0.5277777777777778,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.800537978767683e-06,
+      "loss": 1.3935298,
+      "memory(GiB)": 103.58,
+      "step": 20805,
+      "train_speed(iter/s)": 1.318923
+    },
+    {
+      "acc": 0.70015178,
+      "epoch": 0.5279046169457129,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.79985650242595e-06,
+      "loss": 1.35075512,
+      "memory(GiB)": 103.58,
+      "step": 20810,
+      "train_speed(iter/s)": 1.318958
+    },
+    {
+      "acc": 0.6727982,
+      "epoch": 0.5280314561136479,
+      "grad_norm": 2.75,
+      "learning_rate": 8.79917485894956e-06,
+      "loss": 1.4479454,
+      "memory(GiB)": 103.58,
+      "step": 20815,
+      "train_speed(iter/s)": 1.318992
+    },
+    {
+      "acc": 0.67971873,
+      "epoch": 0.528158295281583,
+      "grad_norm": 2.625,
+      "learning_rate": 8.798493048368498e-06,
+      "loss": 1.40817013,
+      "memory(GiB)": 103.58,
+      "step": 20820,
+      "train_speed(iter/s)": 1.31903
+    },
+    {
+      "acc": 0.69013023,
+      "epoch": 0.528285134449518,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.797811070712754e-06,
+      "loss": 1.36590824,
+      "memory(GiB)": 103.58,
+      "step": 20825,
+      "train_speed(iter/s)": 1.319066
+    },
+    {
+      "acc": 0.68059072,
+      "epoch": 0.5284119736174531,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.797128926012323e-06,
+      "loss": 1.45285473,
+      "memory(GiB)": 103.58,
+      "step": 20830,
+      "train_speed(iter/s)": 1.319102
+    },
+    {
+      "acc": 0.68311453,
+      "epoch": 0.5285388127853882,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.796446614297208e-06,
+      "loss": 1.40636883,
+      "memory(GiB)": 103.58,
+      "step": 20835,
+      "train_speed(iter/s)": 1.31914
+    },
+    {
+      "acc": 0.68418808,
+      "epoch": 0.5286656519533232,
+      "grad_norm": 3.375,
+      "learning_rate": 8.795764135597421e-06,
+      "loss": 1.42230997,
+      "memory(GiB)": 103.58,
+      "step": 20840,
+      "train_speed(iter/s)": 1.319178
+    },
+    {
+      "acc": 0.68659077,
+      "epoch": 0.5287924911212583,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.79508148994298e-06,
+      "loss": 1.41089306,
+      "memory(GiB)": 103.58,
+      "step": 20845,
+      "train_speed(iter/s)": 1.319215
+    },
+    {
+      "acc": 0.6750432,
+      "epoch": 0.5289193302891934,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.794398677363913e-06,
+      "loss": 1.42400761,
+      "memory(GiB)": 103.58,
+      "step": 20850,
+      "train_speed(iter/s)": 1.319253
+    },
+    {
+      "acc": 0.70312643,
+      "epoch": 0.5290461694571283,
+      "grad_norm": 3.25,
+      "learning_rate": 8.79371569789025e-06,
+      "loss": 1.34358501,
+      "memory(GiB)": 103.58,
+      "step": 20855,
+      "train_speed(iter/s)": 1.31929
+    },
+    {
+      "acc": 0.68421831,
+      "epoch": 0.5291730086250634,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.793032551552034e-06,
+      "loss": 1.41911421,
+      "memory(GiB)": 103.58,
+      "step": 20860,
+      "train_speed(iter/s)": 1.319323
+    },
+    {
+      "acc": 0.67754488,
+      "epoch": 0.5292998477929984,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.792349238379311e-06,
+      "loss": 1.44706306,
+      "memory(GiB)": 103.58,
+      "step": 20865,
+      "train_speed(iter/s)": 1.31936
+    },
+    {
+      "acc": 0.66706829,
+      "epoch": 0.5294266869609335,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.791665758402137e-06,
+      "loss": 1.48792534,
+      "memory(GiB)": 103.58,
+      "step": 20870,
+      "train_speed(iter/s)": 1.319397
+    },
+    {
+      "acc": 0.69498787,
+      "epoch": 0.5295535261288686,
+      "grad_norm": 3.25,
+      "learning_rate": 8.790982111650574e-06,
+      "loss": 1.3901741,
+      "memory(GiB)": 103.58,
+      "step": 20875,
+      "train_speed(iter/s)": 1.319432
+    },
+    {
+      "acc": 0.68577662,
+      "epoch": 0.5296803652968036,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.790298298154694e-06,
+      "loss": 1.36060143,
+      "memory(GiB)": 103.58,
+      "step": 20880,
+      "train_speed(iter/s)": 1.319468
+    },
+    {
+      "acc": 0.68176522,
+      "epoch": 0.5298072044647387,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.78961431794457e-06,
+      "loss": 1.41622295,
+      "memory(GiB)": 103.58,
+      "step": 20885,
+      "train_speed(iter/s)": 1.319505
+    },
+    {
+      "acc": 0.6901825,
+      "epoch": 0.5299340436326738,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.78893017105029e-06,
+      "loss": 1.39201775,
+      "memory(GiB)": 103.58,
+      "step": 20890,
+      "train_speed(iter/s)": 1.319542
+    },
+    {
+      "acc": 0.6864419,
+      "epoch": 0.5300608828006088,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.788245857501944e-06,
+      "loss": 1.378549,
+      "memory(GiB)": 103.58,
+      "step": 20895,
+      "train_speed(iter/s)": 1.319579
+    },
+    {
+      "acc": 0.67834749,
+      "epoch": 0.5301877219685439,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.787561377329633e-06,
+      "loss": 1.46893902,
+      "memory(GiB)": 103.58,
+      "step": 20900,
+      "train_speed(iter/s)": 1.319615
+    },
+    {
+      "acc": 0.69257298,
+      "epoch": 0.5303145611364789,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.786876730563462e-06,
+      "loss": 1.41164579,
+      "memory(GiB)": 103.58,
+      "step": 20905,
+      "train_speed(iter/s)": 1.319651
+    },
+    {
+      "acc": 0.68360901,
+      "epoch": 0.530441400304414,
+      "grad_norm": 4.125,
+      "learning_rate": 8.786191917233545e-06,
+      "loss": 1.40068283,
+      "memory(GiB)": 103.58,
+      "step": 20910,
+      "train_speed(iter/s)": 1.319685
+    },
+    {
+      "acc": 0.69209065,
+      "epoch": 0.5305682394723491,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.785506937370003e-06,
+      "loss": 1.35768652,
+      "memory(GiB)": 103.58,
+      "step": 20915,
+      "train_speed(iter/s)": 1.319719
+    },
+    {
+      "acc": 0.6929791,
+      "epoch": 0.5306950786402841,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.784821791002965e-06,
+      "loss": 1.42295628,
+      "memory(GiB)": 103.58,
+      "step": 20920,
+      "train_speed(iter/s)": 1.319753
+    },
+    {
+      "acc": 0.69085813,
+      "epoch": 0.5308219178082192,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.784136478162567e-06,
+      "loss": 1.46093664,
+      "memory(GiB)": 103.58,
+      "step": 20925,
+      "train_speed(iter/s)": 1.319789
+    },
+    {
+      "acc": 0.69543734,
+      "epoch": 0.5309487569761543,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.783450998878951e-06,
+      "loss": 1.36938953,
+      "memory(GiB)": 103.58,
+      "step": 20930,
+      "train_speed(iter/s)": 1.319824
+    },
+    {
+      "acc": 0.68117609,
+      "epoch": 0.5310755961440893,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.78276535318227e-06,
+      "loss": 1.38727589,
+      "memory(GiB)": 103.58,
+      "step": 20935,
+      "train_speed(iter/s)": 1.319859
+    },
+    {
+      "acc": 0.66179028,
+      "epoch": 0.5312024353120244,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.782079541102678e-06,
+      "loss": 1.47397394,
+      "memory(GiB)": 103.58,
+      "step": 20940,
+      "train_speed(iter/s)": 1.319894
+    },
+    {
+      "acc": 0.68292036,
+      "epoch": 0.5313292744799594,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.781393562670342e-06,
+      "loss": 1.402001,
+      "memory(GiB)": 103.58,
+      "step": 20945,
+      "train_speed(iter/s)": 1.319931
+    },
+    {
+      "acc": 0.67607703,
+      "epoch": 0.5314561136478945,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.780707417915436e-06,
+      "loss": 1.43629284,
+      "memory(GiB)": 103.58,
+      "step": 20950,
+      "train_speed(iter/s)": 1.319966
+    },
+    {
+      "acc": 0.69941716,
+      "epoch": 0.5315829528158296,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.780021106868138e-06,
+      "loss": 1.28052139,
+      "memory(GiB)": 103.58,
+      "step": 20955,
+      "train_speed(iter/s)": 1.320001
+    },
+    {
+      "acc": 0.67922864,
+      "epoch": 0.5317097919837646,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.779334629558633e-06,
+      "loss": 1.46130524,
+      "memory(GiB)": 103.58,
+      "step": 20960,
+      "train_speed(iter/s)": 1.320033
+    },
+    {
+      "acc": 0.68130131,
+      "epoch": 0.5318366311516997,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.77864798601712e-06,
+      "loss": 1.40610275,
+      "memory(GiB)": 103.58,
+      "step": 20965,
+      "train_speed(iter/s)": 1.320067
+    },
+    {
+      "acc": 0.68424587,
+      "epoch": 0.5319634703196348,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.777961176273795e-06,
+      "loss": 1.41904583,
+      "memory(GiB)": 103.58,
+      "step": 20970,
+      "train_speed(iter/s)": 1.320101
+    },
+    {
+      "acc": 0.69035187,
+      "epoch": 0.5320903094875697,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.777274200358873e-06,
+      "loss": 1.42254877,
+      "memory(GiB)": 103.58,
+      "step": 20975,
+      "train_speed(iter/s)": 1.320137
+    },
+    {
+      "acc": 0.69987497,
+      "epoch": 0.5322171486555048,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.776587058302566e-06,
+      "loss": 1.39151974,
+      "memory(GiB)": 103.58,
+      "step": 20980,
+      "train_speed(iter/s)": 1.320171
+    },
+    {
+      "acc": 0.69306841,
+      "epoch": 0.5323439878234398,
+      "grad_norm": 2.75,
+      "learning_rate": 8.7758997501351e-06,
+      "loss": 1.38018675,
+      "memory(GiB)": 103.58,
+      "step": 20985,
+      "train_speed(iter/s)": 1.320206
+    },
+    {
+      "acc": 0.7007092,
+      "epoch": 0.5324708269913749,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.775212275886705e-06,
+      "loss": 1.37754574,
+      "memory(GiB)": 103.58,
+      "step": 20990,
+      "train_speed(iter/s)": 1.320234
+    },
+    {
+      "acc": 0.66488223,
+      "epoch": 0.53259766615931,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.774524635587617e-06,
+      "loss": 1.44321041,
+      "memory(GiB)": 103.58,
+      "step": 20995,
+      "train_speed(iter/s)": 1.320268
+    },
+    {
+      "acc": 0.68134871,
+      "epoch": 0.532724505327245,
+      "grad_norm": 2.875,
+      "learning_rate": 8.773836829268084e-06,
+      "loss": 1.41455746,
+      "memory(GiB)": 103.58,
+      "step": 21000,
+      "train_speed(iter/s)": 1.320303
+    },
+    {
+      "epoch": 0.532724505327245,
+      "eval_acc": 0.6733504992095761,
+      "eval_loss": 1.364862084388733,
+      "eval_runtime": 70.1686,
+      "eval_samples_per_second": 90.781,
+      "eval_steps_per_second": 22.702,
+      "step": 21000
+    },
+    {
+      "acc": 0.68678617,
+      "epoch": 0.5328513444951801,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.77314885695836e-06,
+      "loss": 1.38053274,
+      "memory(GiB)": 103.58,
+      "step": 21005,
+      "train_speed(iter/s)": 1.313516
+    },
+    {
+      "acc": 0.68094878,
+      "epoch": 0.5329781836631152,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.772460718688702e-06,
+      "loss": 1.38675766,
+      "memory(GiB)": 103.58,
+      "step": 21010,
+      "train_speed(iter/s)": 1.31355
+    },
+    {
+      "acc": 0.69588604,
+      "epoch": 0.5331050228310502,
+      "grad_norm": 3.5,
+      "learning_rate": 8.771772414489379e-06,
+      "loss": 1.35722427,
+      "memory(GiB)": 103.58,
+      "step": 21015,
+      "train_speed(iter/s)": 1.313586
+    },
+    {
+      "acc": 0.67658815,
+      "epoch": 0.5332318619989853,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.771083944390665e-06,
+      "loss": 1.44313354,
+      "memory(GiB)": 103.58,
+      "step": 21020,
+      "train_speed(iter/s)": 1.313622
+    },
+    {
+      "acc": 0.67214136,
+      "epoch": 0.5333587011669203,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.770395308422842e-06,
+      "loss": 1.45906458,
+      "memory(GiB)": 103.58,
+      "step": 21025,
+      "train_speed(iter/s)": 1.313656
+    },
+    {
+      "acc": 0.69079471,
+      "epoch": 0.5334855403348554,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.769706506616201e-06,
+      "loss": 1.40784435,
+      "memory(GiB)": 103.58,
+      "step": 21030,
+      "train_speed(iter/s)": 1.313694
+    },
+    {
+      "acc": 0.69745102,
+      "epoch": 0.5336123795027905,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.769017539001037e-06,
+      "loss": 1.36608505,
+      "memory(GiB)": 103.58,
+      "step": 21035,
+      "train_speed(iter/s)": 1.313729
+    },
+    {
+      "acc": 0.69085474,
+      "epoch": 0.5337392186707255,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.768328405607655e-06,
+      "loss": 1.42750416,
+      "memory(GiB)": 103.58,
+      "step": 21040,
+      "train_speed(iter/s)": 1.313765
+    },
+    {
+      "acc": 0.6717412,
+      "epoch": 0.5338660578386606,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.767639106466364e-06,
+      "loss": 1.48949366,
+      "memory(GiB)": 103.58,
+      "step": 21045,
+      "train_speed(iter/s)": 1.3138
+    },
+    {
+      "acc": 0.68486967,
+      "epoch": 0.5339928970065957,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.766949641607484e-06,
+      "loss": 1.37240486,
+      "memory(GiB)": 103.58,
+      "step": 21050,
+      "train_speed(iter/s)": 1.313835
+    },
+    {
+      "acc": 0.68787699,
+      "epoch": 0.5341197361745307,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.76626001106134e-06,
+      "loss": 1.39683886,
+      "memory(GiB)": 103.58,
+      "step": 21055,
+      "train_speed(iter/s)": 1.313865
+    },
+    {
+      "acc": 0.69185219,
+      "epoch": 0.5342465753424658,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.765570214858268e-06,
+      "loss": 1.31167841,
+      "memory(GiB)": 103.58,
+      "step": 21060,
+      "train_speed(iter/s)": 1.313895
+    },
+    {
+      "acc": 0.68527889,
+      "epoch": 0.5343734145104008,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.764880253028604e-06,
+      "loss": 1.38187523,
+      "memory(GiB)": 103.58,
+      "step": 21065,
+      "train_speed(iter/s)": 1.313928
+    },
+    {
+      "acc": 0.68907433,
+      "epoch": 0.5345002536783359,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.764190125602698e-06,
+      "loss": 1.35984259,
+      "memory(GiB)": 103.58,
+      "step": 21070,
+      "train_speed(iter/s)": 1.313962
+    },
+    {
+      "acc": 0.68811793,
+      "epoch": 0.534627092846271,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.763499832610904e-06,
+      "loss": 1.41808872,
+      "memory(GiB)": 103.58,
+      "step": 21075,
+      "train_speed(iter/s)": 1.313995
+    },
+    {
+      "acc": 0.7014019,
+      "epoch": 0.534753932014206,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.762809374083585e-06,
+      "loss": 1.33282423,
+      "memory(GiB)": 103.58,
+      "step": 21080,
+      "train_speed(iter/s)": 1.314026
+    },
+    {
+      "acc": 0.67853317,
+      "epoch": 0.5348807711821411,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.76211875005111e-06,
+      "loss": 1.44091339,
+      "memory(GiB)": 103.58,
+      "step": 21085,
+      "train_speed(iter/s)": 1.31406
+    },
+    {
+      "acc": 0.71329799,
+      "epoch": 0.5350076103500762,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.761427960543854e-06,
+      "loss": 1.33146238,
+      "memory(GiB)": 103.58,
+      "step": 21090,
+      "train_speed(iter/s)": 1.314094
+    },
+    {
+      "acc": 0.68674788,
+      "epoch": 0.5351344495180111,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.760737005592205e-06,
+      "loss": 1.40140638,
+      "memory(GiB)": 103.58,
+      "step": 21095,
+      "train_speed(iter/s)": 1.314127
+    },
+    {
+      "acc": 0.71186142,
+      "epoch": 0.5352612886859462,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.760045885226551e-06,
+      "loss": 1.36106968,
+      "memory(GiB)": 103.58,
+      "step": 21100,
+      "train_speed(iter/s)": 1.31416
+    },
+    {
+      "acc": 0.69378462,
+      "epoch": 0.5353881278538812,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.759354599477293e-06,
+      "loss": 1.41884098,
+      "memory(GiB)": 103.58,
+      "step": 21105,
+      "train_speed(iter/s)": 1.314194
+    },
+    {
+      "acc": 0.69889002,
+      "epoch": 0.5355149670218163,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.758663148374833e-06,
+      "loss": 1.3249094,
+      "memory(GiB)": 103.58,
+      "step": 21110,
+      "train_speed(iter/s)": 1.314226
+    },
+    {
+      "acc": 0.69128776,
+      "epoch": 0.5356418061897514,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.757971531949587e-06,
+      "loss": 1.42673264,
+      "memory(GiB)": 103.58,
+      "step": 21115,
+      "train_speed(iter/s)": 1.31426
+    },
+    {
+      "acc": 0.69440718,
+      "epoch": 0.5357686453576864,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.757279750231977e-06,
+      "loss": 1.35926762,
+      "memory(GiB)": 103.58,
+      "step": 21120,
+      "train_speed(iter/s)": 1.314295
+    },
+    {
+      "acc": 0.6971765,
+      "epoch": 0.5358954845256215,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.756587803252426e-06,
+      "loss": 1.34887962,
+      "memory(GiB)": 103.58,
+      "step": 21125,
+      "train_speed(iter/s)": 1.314329
+    },
+    {
+      "acc": 0.69337263,
+      "epoch": 0.5360223236935566,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.755895691041373e-06,
+      "loss": 1.39696941,
+      "memory(GiB)": 103.58,
+      "step": 21130,
+      "train_speed(iter/s)": 1.314362
+    },
+    {
+      "acc": 0.69202433,
+      "epoch": 0.5361491628614916,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.755203413629257e-06,
+      "loss": 1.40347052,
+      "memory(GiB)": 103.58,
+      "step": 21135,
+      "train_speed(iter/s)": 1.314398
+    },
+    {
+      "acc": 0.69433479,
+      "epoch": 0.5362760020294267,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.75451097104653e-06,
+      "loss": 1.30807819,
+      "memory(GiB)": 103.58,
+      "step": 21140,
+      "train_speed(iter/s)": 1.314433
+    },
+    {
+      "acc": 0.6880784,
+      "epoch": 0.5364028411973617,
+      "grad_norm": 2.875,
+      "learning_rate": 8.75381836332365e-06,
+      "loss": 1.43334475,
+      "memory(GiB)": 103.58,
+      "step": 21145,
+      "train_speed(iter/s)": 1.314466
+    },
+    {
+      "acc": 0.66783547,
+      "epoch": 0.5365296803652968,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.753125590491077e-06,
+      "loss": 1.42898569,
+      "memory(GiB)": 103.58,
+      "step": 21150,
+      "train_speed(iter/s)": 1.314483
+    },
+    {
+      "acc": 0.69138565,
+      "epoch": 0.5366565195332319,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.752432652579284e-06,
+      "loss": 1.36450491,
+      "memory(GiB)": 103.58,
+      "step": 21155,
+      "train_speed(iter/s)": 1.314517
+    },
+    {
+      "acc": 0.6805872,
+      "epoch": 0.5367833587011669,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.751739549618749e-06,
+      "loss": 1.42755527,
+      "memory(GiB)": 103.58,
+      "step": 21160,
+      "train_speed(iter/s)": 1.314552
+    },
+    {
+      "acc": 0.68875294,
+      "epoch": 0.536910197869102,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.751046281639958e-06,
+      "loss": 1.36447239,
+      "memory(GiB)": 103.58,
+      "step": 21165,
+      "train_speed(iter/s)": 1.314585
+    },
+    {
+      "acc": 0.68989353,
+      "epoch": 0.5370370370370371,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.750352848673405e-06,
+      "loss": 1.42316551,
+      "memory(GiB)": 103.58,
+      "step": 21170,
+      "train_speed(iter/s)": 1.314618
+    },
+    {
+      "acc": 0.67583642,
+      "epoch": 0.5371638762049721,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.749659250749589e-06,
+      "loss": 1.48863163,
+      "memory(GiB)": 103.58,
+      "step": 21175,
+      "train_speed(iter/s)": 1.314653
+    },
+    {
+      "acc": 0.69916854,
+      "epoch": 0.5372907153729072,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.748965487899019e-06,
+      "loss": 1.36683903,
+      "memory(GiB)": 103.58,
+      "step": 21180,
+      "train_speed(iter/s)": 1.314686
+    },
+    {
+      "acc": 0.69577732,
+      "epoch": 0.5374175545408422,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.748271560152208e-06,
+      "loss": 1.37867661,
+      "memory(GiB)": 103.58,
+      "step": 21185,
+      "train_speed(iter/s)": 1.314722
+    },
+    {
+      "acc": 0.68346825,
+      "epoch": 0.5375443937087773,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.74757746753968e-06,
+      "loss": 1.4295702,
+      "memory(GiB)": 103.58,
+      "step": 21190,
+      "train_speed(iter/s)": 1.314755
+    },
+    {
+      "acc": 0.68828497,
+      "epoch": 0.5376712328767124,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.746883210091963e-06,
+      "loss": 1.32826128,
+      "memory(GiB)": 103.58,
+      "step": 21195,
+      "train_speed(iter/s)": 1.314789
+    },
+    {
+      "acc": 0.70841274,
+      "epoch": 0.5377980720446474,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.746188787839593e-06,
+      "loss": 1.29850178,
+      "memory(GiB)": 103.58,
+      "step": 21200,
+      "train_speed(iter/s)": 1.314825
+    },
+    {
+      "acc": 0.68459339,
+      "epoch": 0.5379249112125825,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.745494200813116e-06,
+      "loss": 1.4219965,
+      "memory(GiB)": 103.58,
+      "step": 21205,
+      "train_speed(iter/s)": 1.314862
+    },
+    {
+      "acc": 0.68561535,
+      "epoch": 0.5380517503805176,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.74479944904308e-06,
+      "loss": 1.42761078,
+      "memory(GiB)": 103.58,
+      "step": 21210,
+      "train_speed(iter/s)": 1.314896
+    },
+    {
+      "acc": 0.68150845,
+      "epoch": 0.5381785895484525,
+      "grad_norm": 3.0,
+      "learning_rate": 8.744104532560047e-06,
+      "loss": 1.41497936,
+      "memory(GiB)": 103.58,
+      "step": 21215,
+      "train_speed(iter/s)": 1.31493
+    },
+    {
+      "acc": 0.6683341,
+      "epoch": 0.5383054287163876,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.74340945139458e-06,
+      "loss": 1.49645672,
+      "memory(GiB)": 103.58,
+      "step": 21220,
+      "train_speed(iter/s)": 1.314965
+    },
+    {
+      "acc": 0.67736883,
+      "epoch": 0.5384322678843226,
+      "grad_norm": 4.8125,
+      "learning_rate": 8.742714205577251e-06,
+      "loss": 1.41114769,
+      "memory(GiB)": 103.58,
+      "step": 21225,
+      "train_speed(iter/s)": 1.315001
+    },
+    {
+      "acc": 0.70033998,
+      "epoch": 0.5385591070522577,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.742018795138642e-06,
+      "loss": 1.29460344,
+      "memory(GiB)": 103.58,
+      "step": 21230,
+      "train_speed(iter/s)": 1.315035
+    },
+    {
+      "acc": 0.67958736,
+      "epoch": 0.5386859462201928,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.74132322010934e-06,
+      "loss": 1.40158215,
+      "memory(GiB)": 103.58,
+      "step": 21235,
+      "train_speed(iter/s)": 1.315069
+    },
+    {
+      "acc": 0.68369184,
+      "epoch": 0.5388127853881278,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.740627480519937e-06,
+      "loss": 1.36274605,
+      "memory(GiB)": 103.58,
+      "step": 21240,
+      "train_speed(iter/s)": 1.315103
+    },
+    {
+      "acc": 0.68317056,
+      "epoch": 0.5389396245560629,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.739931576401037e-06,
+      "loss": 1.41569662,
+      "memory(GiB)": 103.58,
+      "step": 21245,
+      "train_speed(iter/s)": 1.315137
+    },
+    {
+      "acc": 0.67937641,
+      "epoch": 0.539066463723998,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.73923550778325e-06,
+      "loss": 1.32122135,
+      "memory(GiB)": 103.58,
+      "step": 21250,
+      "train_speed(iter/s)": 1.315173
+    },
+    {
+      "acc": 0.69189105,
+      "epoch": 0.539193302891933,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.73853927469719e-06,
+      "loss": 1.42061977,
+      "memory(GiB)": 103.58,
+      "step": 21255,
+      "train_speed(iter/s)": 1.315206
+    },
+    {
+      "acc": 0.68428173,
+      "epoch": 0.5393201420598681,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.73784287717348e-06,
+      "loss": 1.39857025,
+      "memory(GiB)": 103.58,
+      "step": 21260,
+      "train_speed(iter/s)": 1.315242
+    },
+    {
+      "acc": 0.65490952,
+      "epoch": 0.5394469812278031,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.737146315242755e-06,
+      "loss": 1.50987034,
+      "memory(GiB)": 103.58,
+      "step": 21265,
+      "train_speed(iter/s)": 1.315278
+    },
+    {
+      "acc": 0.67512622,
+      "epoch": 0.5395738203957382,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.73644958893565e-06,
+      "loss": 1.45201817,
+      "memory(GiB)": 103.58,
+      "step": 21270,
+      "train_speed(iter/s)": 1.315314
+    },
+    {
+      "acc": 0.68945484,
+      "epoch": 0.5397006595636733,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.735752698282807e-06,
+      "loss": 1.36922674,
+      "memory(GiB)": 103.58,
+      "step": 21275,
+      "train_speed(iter/s)": 1.31535
+    },
+    {
+      "acc": 0.7031033,
+      "epoch": 0.5398274987316083,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.735055643314883e-06,
+      "loss": 1.43142872,
+      "memory(GiB)": 103.58,
+      "step": 21280,
+      "train_speed(iter/s)": 1.315386
+    },
+    {
+      "acc": 0.67630386,
+      "epoch": 0.5399543378995434,
+      "grad_norm": 2.875,
+      "learning_rate": 8.734358424062536e-06,
+      "loss": 1.43648653,
+      "memory(GiB)": 103.58,
+      "step": 21285,
+      "train_speed(iter/s)": 1.315421
+    },
+    {
+      "acc": 0.69538169,
+      "epoch": 0.5400811770674785,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.733661040556433e-06,
+      "loss": 1.36905451,
+      "memory(GiB)": 103.58,
+      "step": 21290,
+      "train_speed(iter/s)": 1.315456
+    },
+    {
+      "acc": 0.66706481,
+      "epoch": 0.5402080162354135,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.732963492827248e-06,
+      "loss": 1.42221928,
+      "memory(GiB)": 103.58,
+      "step": 21295,
+      "train_speed(iter/s)": 1.315491
+    },
+    {
+      "acc": 0.68236094,
+      "epoch": 0.5403348554033486,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.732265780905661e-06,
+      "loss": 1.41945496,
+      "memory(GiB)": 103.58,
+      "step": 21300,
+      "train_speed(iter/s)": 1.315526
+    },
+    {
+      "acc": 0.67294111,
+      "epoch": 0.5404616945712836,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.731567904822362e-06,
+      "loss": 1.45018311,
+      "memory(GiB)": 103.58,
+      "step": 21305,
+      "train_speed(iter/s)": 1.315561
+    },
+    {
+      "acc": 0.6924798,
+      "epoch": 0.5405885337392187,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.730869864608047e-06,
+      "loss": 1.28735981,
+      "memory(GiB)": 103.58,
+      "step": 21310,
+      "train_speed(iter/s)": 1.315596
+    },
+    {
+      "acc": 0.67728472,
+      "epoch": 0.5407153729071538,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.730171660293418e-06,
+      "loss": 1.47817507,
+      "memory(GiB)": 103.58,
+      "step": 21315,
+      "train_speed(iter/s)": 1.315632
+    },
+    {
+      "acc": 0.68907056,
+      "epoch": 0.5408422120750888,
+      "grad_norm": 2.5,
+      "learning_rate": 8.729473291909185e-06,
+      "loss": 1.38139124,
+      "memory(GiB)": 103.58,
+      "step": 21320,
+      "train_speed(iter/s)": 1.315666
+    },
+    {
+      "acc": 0.67110801,
+      "epoch": 0.5409690512430239,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.728774759486065e-06,
+      "loss": 1.41518936,
+      "memory(GiB)": 103.58,
+      "step": 21325,
+      "train_speed(iter/s)": 1.315701
+    },
+    {
+      "acc": 0.67401962,
+      "epoch": 0.541095890410959,
+      "grad_norm": 2.875,
+      "learning_rate": 8.728076063054786e-06,
+      "loss": 1.41158752,
+      "memory(GiB)": 103.58,
+      "step": 21330,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.66792092,
+      "epoch": 0.541222729578894,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.727377202646074e-06,
+      "loss": 1.49457302,
+      "memory(GiB)": 103.58,
+      "step": 21335,
+      "train_speed(iter/s)": 1.31577
+    },
+    {
+      "acc": 0.69820223,
+      "epoch": 0.541349568746829,
+      "grad_norm": 3.875,
+      "learning_rate": 8.726678178290673e-06,
+      "loss": 1.44100895,
+      "memory(GiB)": 103.58,
+      "step": 21340,
+      "train_speed(iter/s)": 1.315803
+    },
+    {
+      "acc": 0.69307461,
+      "epoch": 0.541476407914764,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.725978990019326e-06,
+      "loss": 1.38754511,
+      "memory(GiB)": 103.58,
+      "step": 21345,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.69370995,
+      "epoch": 0.5416032470826991,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.72527963786279e-06,
+      "loss": 1.37405128,
+      "memory(GiB)": 103.58,
+      "step": 21350,
+      "train_speed(iter/s)": 1.315873
+    },
+    {
+      "acc": 0.67907839,
+      "epoch": 0.5417300862506342,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.72458012185182e-06,
+      "loss": 1.4620656,
+      "memory(GiB)": 103.58,
+      "step": 21355,
+      "train_speed(iter/s)": 1.315908
+    },
+    {
+      "acc": 0.69427853,
+      "epoch": 0.5418569254185692,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.72388044201719e-06,
+      "loss": 1.35402203,
+      "memory(GiB)": 103.58,
+      "step": 21360,
+      "train_speed(iter/s)": 1.315945
+    },
+    {
+      "acc": 0.7012291,
+      "epoch": 0.5419837645865043,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.723180598389671e-06,
+      "loss": 1.28676586,
+      "memory(GiB)": 103.58,
+      "step": 21365,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.68042169,
+      "epoch": 0.5421106037544394,
+      "grad_norm": 2.359375,
+      "learning_rate": 8.722480591000046e-06,
+      "loss": 1.442169,
+      "memory(GiB)": 103.58,
+      "step": 21370,
+      "train_speed(iter/s)": 1.316015
+    },
+    {
+      "acc": 0.70350618,
+      "epoch": 0.5422374429223744,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.721780419879106e-06,
+      "loss": 1.32094669,
+      "memory(GiB)": 103.58,
+      "step": 21375,
+      "train_speed(iter/s)": 1.316052
+    },
+    {
+      "acc": 0.6780992,
+      "epoch": 0.5423642820903095,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.721080085057646e-06,
+      "loss": 1.41335735,
+      "memory(GiB)": 103.58,
+      "step": 21380,
+      "train_speed(iter/s)": 1.316087
+    },
+    {
+      "acc": 0.68083973,
+      "epoch": 0.5424911212582445,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.72037958656647e-06,
+      "loss": 1.49879665,
+      "memory(GiB)": 103.58,
+      "step": 21385,
+      "train_speed(iter/s)": 1.316122
+    },
+    {
+      "acc": 0.67618914,
+      "epoch": 0.5426179604261796,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.71967892443639e-06,
+      "loss": 1.43045759,
+      "memory(GiB)": 103.58,
+      "step": 21390,
+      "train_speed(iter/s)": 1.316157
+    },
+    {
+      "acc": 0.69878416,
+      "epoch": 0.5427447995941147,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.718978098698226e-06,
+      "loss": 1.3774621,
+      "memory(GiB)": 103.58,
+      "step": 21395,
+      "train_speed(iter/s)": 1.316191
+    },
+    {
+      "acc": 0.68166771,
+      "epoch": 0.5428716387620497,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.718277109382799e-06,
+      "loss": 1.42399673,
+      "memory(GiB)": 103.58,
+      "step": 21400,
+      "train_speed(iter/s)": 1.316227
+    },
+    {
+      "acc": 0.68556104,
+      "epoch": 0.5429984779299848,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.717575956520942e-06,
+      "loss": 1.46378441,
+      "memory(GiB)": 103.58,
+      "step": 21405,
+      "train_speed(iter/s)": 1.316262
+    },
+    {
+      "acc": 0.69678278,
+      "epoch": 0.5431253170979199,
+      "grad_norm": 2.625,
+      "learning_rate": 8.716874640143498e-06,
+      "loss": 1.35134144,
+      "memory(GiB)": 103.58,
+      "step": 21410,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.68842077,
+      "epoch": 0.5432521562658549,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.716173160281315e-06,
+      "loss": 1.44391994,
+      "memory(GiB)": 103.58,
+      "step": 21415,
+      "train_speed(iter/s)": 1.316331
+    },
+    {
+      "acc": 0.67255564,
+      "epoch": 0.54337899543379,
+      "grad_norm": 3.953125,
+      "learning_rate": 8.715471516965242e-06,
+      "loss": 1.47558393,
+      "memory(GiB)": 103.58,
+      "step": 21420,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.6853941,
+      "epoch": 0.543505834601725,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.714769710226144e-06,
+      "loss": 1.43170719,
+      "memory(GiB)": 103.58,
+      "step": 21425,
+      "train_speed(iter/s)": 1.316402
+    },
+    {
+      "acc": 0.68471484,
+      "epoch": 0.5436326737696601,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.714067740094888e-06,
+      "loss": 1.41770172,
+      "memory(GiB)": 103.58,
+      "step": 21430,
+      "train_speed(iter/s)": 1.316437
+    },
+    {
+      "acc": 0.69549303,
+      "epoch": 0.5437595129375952,
+      "grad_norm": 4.40625,
+      "learning_rate": 8.713365606602353e-06,
+      "loss": 1.38129511,
+      "memory(GiB)": 103.58,
+      "step": 21435,
+      "train_speed(iter/s)": 1.316472
+    },
+    {
+      "acc": 0.69997654,
+      "epoch": 0.5438863521055302,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.71266330977942e-06,
+      "loss": 1.3177742,
+      "memory(GiB)": 103.58,
+      "step": 21440,
+      "train_speed(iter/s)": 1.316505
+    },
+    {
+      "acc": 0.70714817,
+      "epoch": 0.5440131912734653,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.711960849656975e-06,
+      "loss": 1.32433367,
+      "memory(GiB)": 103.58,
+      "step": 21445,
+      "train_speed(iter/s)": 1.316539
+    },
+    {
+      "acc": 0.67840433,
+      "epoch": 0.5441400304414004,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.711258226265922e-06,
+      "loss": 1.41349249,
+      "memory(GiB)": 103.58,
+      "step": 21450,
+      "train_speed(iter/s)": 1.316574
+    },
+    {
+      "acc": 0.68559055,
+      "epoch": 0.5442668696093353,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.710555439637163e-06,
+      "loss": 1.43964787,
+      "memory(GiB)": 103.58,
+      "step": 21455,
+      "train_speed(iter/s)": 1.316609
+    },
+    {
+      "acc": 0.6845149,
+      "epoch": 0.5443937087772704,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.709852489801608e-06,
+      "loss": 1.33580046,
+      "memory(GiB)": 103.58,
+      "step": 21460,
+      "train_speed(iter/s)": 1.316645
+    },
+    {
+      "acc": 0.67897263,
+      "epoch": 0.5445205479452054,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.709149376790177e-06,
+      "loss": 1.42329788,
+      "memory(GiB)": 103.58,
+      "step": 21465,
+      "train_speed(iter/s)": 1.31668
+    },
+    {
+      "acc": 0.68719339,
+      "epoch": 0.5446473871131405,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.708446100633796e-06,
+      "loss": 1.41544094,
+      "memory(GiB)": 103.58,
+      "step": 21470,
+      "train_speed(iter/s)": 1.316712
+    },
+    {
+      "acc": 0.7165205,
+      "epoch": 0.5447742262810756,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.707742661363401e-06,
+      "loss": 1.27734756,
+      "memory(GiB)": 103.58,
+      "step": 21475,
+      "train_speed(iter/s)": 1.316748
+    },
+    {
+      "acc": 0.67025146,
+      "epoch": 0.5449010654490106,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.707039059009927e-06,
+      "loss": 1.50134506,
+      "memory(GiB)": 103.58,
+      "step": 21480,
+      "train_speed(iter/s)": 1.316782
+    },
+    {
+      "acc": 0.69160013,
+      "epoch": 0.5450279046169457,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.706335293604326e-06,
+      "loss": 1.39394703,
+      "memory(GiB)": 103.58,
+      "step": 21485,
+      "train_speed(iter/s)": 1.316816
+    },
+    {
+      "acc": 0.68664351,
+      "epoch": 0.5451547437848808,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.705631365177552e-06,
+      "loss": 1.43678226,
+      "memory(GiB)": 103.58,
+      "step": 21490,
+      "train_speed(iter/s)": 1.316852
+    },
+    {
+      "acc": 0.68673892,
+      "epoch": 0.5452815829528158,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.704927273760563e-06,
+      "loss": 1.37427263,
+      "memory(GiB)": 103.58,
+      "step": 21495,
+      "train_speed(iter/s)": 1.316887
+    },
+    {
+      "acc": 0.68149929,
+      "epoch": 0.5454084221207509,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.704223019384334e-06,
+      "loss": 1.43104954,
+      "memory(GiB)": 103.58,
+      "step": 21500,
+      "train_speed(iter/s)": 1.316923
+    },
+    {
+      "acc": 0.69609838,
+      "epoch": 0.5455352612886859,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.703518602079836e-06,
+      "loss": 1.35936346,
+      "memory(GiB)": 103.58,
+      "step": 21505,
+      "train_speed(iter/s)": 1.316958
+    },
+    {
+      "acc": 0.68201265,
+      "epoch": 0.545662100456621,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.702814021878057e-06,
+      "loss": 1.411551,
+      "memory(GiB)": 103.58,
+      "step": 21510,
+      "train_speed(iter/s)": 1.316993
+    },
+    {
+      "acc": 0.67950954,
+      "epoch": 0.5457889396245561,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.702109278809985e-06,
+      "loss": 1.44764833,
+      "memory(GiB)": 103.58,
+      "step": 21515,
+      "train_speed(iter/s)": 1.317029
+    },
+    {
+      "acc": 0.68485532,
+      "epoch": 0.5459157787924911,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.70140437290662e-06,
+      "loss": 1.41560574,
+      "memory(GiB)": 103.58,
+      "step": 21520,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.67545185,
+      "epoch": 0.5460426179604262,
+      "grad_norm": 3.125,
+      "learning_rate": 8.700699304198963e-06,
+      "loss": 1.47755146,
+      "memory(GiB)": 103.58,
+      "step": 21525,
+      "train_speed(iter/s)": 1.317099
+    },
+    {
+      "acc": 0.67977266,
+      "epoch": 0.5461694571283613,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.699994072718026e-06,
+      "loss": 1.43645325,
+      "memory(GiB)": 103.58,
+      "step": 21530,
+      "train_speed(iter/s)": 1.31713
+    },
+    {
+      "acc": 0.69608855,
+      "epoch": 0.5462962962962963,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.699288678494833e-06,
+      "loss": 1.34875984,
+      "memory(GiB)": 103.58,
+      "step": 21535,
+      "train_speed(iter/s)": 1.317166
+    },
+    {
+      "acc": 0.67769041,
+      "epoch": 0.5464231354642314,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.698583121560407e-06,
+      "loss": 1.45286484,
+      "memory(GiB)": 103.58,
+      "step": 21540,
+      "train_speed(iter/s)": 1.317202
+    },
+    {
+      "acc": 0.67740555,
+      "epoch": 0.5465499746321664,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.697877401945784e-06,
+      "loss": 1.40560408,
+      "memory(GiB)": 103.58,
+      "step": 21545,
+      "train_speed(iter/s)": 1.317237
+    },
+    {
+      "acc": 0.69763069,
+      "epoch": 0.5466768138001015,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.697171519682002e-06,
+      "loss": 1.33178234,
+      "memory(GiB)": 103.58,
+      "step": 21550,
+      "train_speed(iter/s)": 1.317272
+    },
+    {
+      "acc": 0.68587961,
+      "epoch": 0.5468036529680366,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.696465474800109e-06,
+      "loss": 1.38510551,
+      "memory(GiB)": 103.58,
+      "step": 21555,
+      "train_speed(iter/s)": 1.317308
+    },
+    {
+      "acc": 0.68334718,
+      "epoch": 0.5469304921359716,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.695759267331162e-06,
+      "loss": 1.41050606,
+      "memory(GiB)": 103.58,
+      "step": 21560,
+      "train_speed(iter/s)": 1.317342
+    },
+    {
+      "acc": 0.6817318,
+      "epoch": 0.5470573313039067,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.69505289730622e-06,
+      "loss": 1.41331959,
+      "memory(GiB)": 103.58,
+      "step": 21565,
+      "train_speed(iter/s)": 1.317377
+    },
+    {
+      "acc": 0.69215994,
+      "epoch": 0.5471841704718418,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.694346364756356e-06,
+      "loss": 1.40290823,
+      "memory(GiB)": 103.58,
+      "step": 21570,
+      "train_speed(iter/s)": 1.317413
+    },
+    {
+      "acc": 0.69362173,
+      "epoch": 0.5473110096397767,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.693639669712645e-06,
+      "loss": 1.39848928,
+      "memory(GiB)": 103.58,
+      "step": 21575,
+      "train_speed(iter/s)": 1.317448
+    },
+    {
+      "acc": 0.69803982,
+      "epoch": 0.5474378488077118,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.692932812206171e-06,
+      "loss": 1.34439735,
+      "memory(GiB)": 103.58,
+      "step": 21580,
+      "train_speed(iter/s)": 1.317484
+    },
+    {
+      "acc": 0.67893238,
+      "epoch": 0.5475646879756468,
+      "grad_norm": 2.875,
+      "learning_rate": 8.692225792268023e-06,
+      "loss": 1.4152832,
+      "memory(GiB)": 103.58,
+      "step": 21585,
+      "train_speed(iter/s)": 1.317518
+    },
+    {
+      "acc": 0.68051066,
+      "epoch": 0.5476915271435819,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.691518609929302e-06,
+      "loss": 1.44598961,
+      "memory(GiB)": 103.58,
+      "step": 21590,
+      "train_speed(iter/s)": 1.317554
+    },
+    {
+      "acc": 0.68673372,
+      "epoch": 0.547818366311517,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.690811265221108e-06,
+      "loss": 1.38302603,
+      "memory(GiB)": 103.58,
+      "step": 21595,
+      "train_speed(iter/s)": 1.317588
+    },
+    {
+      "acc": 0.69380851,
+      "epoch": 0.547945205479452,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.690103758174558e-06,
+      "loss": 1.40402088,
+      "memory(GiB)": 103.58,
+      "step": 21600,
+      "train_speed(iter/s)": 1.317624
+    },
+    {
+      "acc": 0.66138687,
+      "epoch": 0.5480720446473871,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.68939608882077e-06,
+      "loss": 1.4334857,
+      "memory(GiB)": 103.58,
+      "step": 21605,
+      "train_speed(iter/s)": 1.317661
+    },
+    {
+      "acc": 0.67458992,
+      "epoch": 0.5481988838153222,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.688688257190869e-06,
+      "loss": 1.42300396,
+      "memory(GiB)": 103.58,
+      "step": 21610,
+      "train_speed(iter/s)": 1.317696
+    },
+    {
+      "acc": 0.7037241,
+      "epoch": 0.5483257229832572,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.68798026331599e-06,
+      "loss": 1.30981369,
+      "memory(GiB)": 103.58,
+      "step": 21615,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "acc": 0.67808676,
+      "epoch": 0.5484525621511923,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.687272107227274e-06,
+      "loss": 1.45087452,
+      "memory(GiB)": 103.58,
+      "step": 21620,
+      "train_speed(iter/s)": 1.317766
+    },
+    {
+      "acc": 0.66751728,
+      "epoch": 0.5485794013191273,
+      "grad_norm": 2.5,
+      "learning_rate": 8.686563788955867e-06,
+      "loss": 1.4407074,
+      "memory(GiB)": 103.58,
+      "step": 21625,
+      "train_speed(iter/s)": 1.317802
+    },
+    {
+      "acc": 0.69195385,
+      "epoch": 0.5487062404870624,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.685855308532926e-06,
+      "loss": 1.35114403,
+      "memory(GiB)": 103.58,
+      "step": 21630,
+      "train_speed(iter/s)": 1.317836
+    },
+    {
+      "acc": 0.69654956,
+      "epoch": 0.5488330796549975,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.685146665989613e-06,
+      "loss": 1.38646421,
+      "memory(GiB)": 103.58,
+      "step": 21635,
+      "train_speed(iter/s)": 1.317871
+    },
+    {
+      "acc": 0.67954645,
+      "epoch": 0.5489599188229325,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.684437861357095e-06,
+      "loss": 1.38127098,
+      "memory(GiB)": 103.58,
+      "step": 21640,
+      "train_speed(iter/s)": 1.317906
+    },
+    {
+      "acc": 0.68796778,
+      "epoch": 0.5490867579908676,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.683728894666551e-06,
+      "loss": 1.34233704,
+      "memory(GiB)": 103.58,
+      "step": 21645,
+      "train_speed(iter/s)": 1.317942
+    },
+    {
+      "acc": 0.68611774,
+      "epoch": 0.5492135971588027,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.683019765949163e-06,
+      "loss": 1.41957378,
+      "memory(GiB)": 103.58,
+      "step": 21650,
+      "train_speed(iter/s)": 1.317977
+    },
+    {
+      "acc": 0.67939382,
+      "epoch": 0.5493404363267377,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.682310475236123e-06,
+      "loss": 1.39267759,
+      "memory(GiB)": 103.58,
+      "step": 21655,
+      "train_speed(iter/s)": 1.318013
+    },
+    {
+      "acc": 0.68220077,
+      "epoch": 0.5494672754946728,
+      "grad_norm": 3.25,
+      "learning_rate": 8.681601022558628e-06,
+      "loss": 1.43452892,
+      "memory(GiB)": 103.58,
+      "step": 21660,
+      "train_speed(iter/s)": 1.318049
+    },
+    {
+      "acc": 0.68735232,
+      "epoch": 0.5495941146626078,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.680891407947882e-06,
+      "loss": 1.38080406,
+      "memory(GiB)": 103.58,
+      "step": 21665,
+      "train_speed(iter/s)": 1.318083
+    },
+    {
+      "acc": 0.67715192,
+      "epoch": 0.5497209538305429,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.680181631435098e-06,
+      "loss": 1.50328751,
+      "memory(GiB)": 103.58,
+      "step": 21670,
+      "train_speed(iter/s)": 1.318118
+    },
+    {
+      "acc": 0.68684998,
+      "epoch": 0.549847792998478,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.679471693051495e-06,
+      "loss": 1.39962263,
+      "memory(GiB)": 103.58,
+      "step": 21675,
+      "train_speed(iter/s)": 1.318152
+    },
+    {
+      "acc": 0.6981462,
+      "epoch": 0.549974632166413,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.678761592828301e-06,
+      "loss": 1.38240967,
+      "memory(GiB)": 103.58,
+      "step": 21680,
+      "train_speed(iter/s)": 1.318188
+    },
+    {
+      "acc": 0.68588867,
+      "epoch": 0.5501014713343481,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.678051330796746e-06,
+      "loss": 1.36021376,
+      "memory(GiB)": 103.58,
+      "step": 21685,
+      "train_speed(iter/s)": 1.318216
+    },
+    {
+      "acc": 0.67108612,
+      "epoch": 0.5502283105022832,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.677340906988072e-06,
+      "loss": 1.40763073,
+      "memory(GiB)": 103.58,
+      "step": 21690,
+      "train_speed(iter/s)": 1.318249
+    },
+    {
+      "acc": 0.68945746,
+      "epoch": 0.5503551496702181,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.676630321433528e-06,
+      "loss": 1.35040207,
+      "memory(GiB)": 103.58,
+      "step": 21695,
+      "train_speed(iter/s)": 1.318284
+    },
+    {
+      "acc": 0.68875222,
+      "epoch": 0.5504819888381532,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.675919574164366e-06,
+      "loss": 1.41395483,
+      "memory(GiB)": 103.58,
+      "step": 21700,
+      "train_speed(iter/s)": 1.318319
+    },
+    {
+      "acc": 0.68427572,
+      "epoch": 0.5506088280060882,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.675208665211851e-06,
+      "loss": 1.39028969,
+      "memory(GiB)": 103.58,
+      "step": 21705,
+      "train_speed(iter/s)": 1.318354
+    },
+    {
+      "acc": 0.67924085,
+      "epoch": 0.5507356671740233,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.674497594607249e-06,
+      "loss": 1.41348495,
+      "memory(GiB)": 103.58,
+      "step": 21710,
+      "train_speed(iter/s)": 1.31839
+    },
+    {
+      "acc": 0.6744801,
+      "epoch": 0.5508625063419584,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.673786362381837e-06,
+      "loss": 1.45496683,
+      "memory(GiB)": 103.58,
+      "step": 21715,
+      "train_speed(iter/s)": 1.318425
+    },
+    {
+      "acc": 0.6846478,
+      "epoch": 0.5509893455098934,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.673074968566899e-06,
+      "loss": 1.35211306,
+      "memory(GiB)": 103.58,
+      "step": 21720,
+      "train_speed(iter/s)": 1.318459
+    },
+    {
+      "acc": 0.66545563,
+      "epoch": 0.5511161846778285,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.672363413193724e-06,
+      "loss": 1.47779179,
+      "memory(GiB)": 103.58,
+      "step": 21725,
+      "train_speed(iter/s)": 1.318488
+    },
+    {
+      "acc": 0.67853909,
+      "epoch": 0.5512430238457636,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.671651696293613e-06,
+      "loss": 1.3925703,
+      "memory(GiB)": 103.58,
+      "step": 21730,
+      "train_speed(iter/s)": 1.318522
+    },
+    {
+      "acc": 0.68203959,
+      "epoch": 0.5513698630136986,
+      "grad_norm": 3.25,
+      "learning_rate": 8.670939817897865e-06,
+      "loss": 1.44375753,
+      "memory(GiB)": 103.58,
+      "step": 21735,
+      "train_speed(iter/s)": 1.318556
+    },
+    {
+      "acc": 0.67495718,
+      "epoch": 0.5514967021816337,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.670227778037796e-06,
+      "loss": 1.41909351,
+      "memory(GiB)": 103.58,
+      "step": 21740,
+      "train_speed(iter/s)": 1.31859
+    },
+    {
+      "acc": 0.68948469,
+      "epoch": 0.5516235413495687,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.669515576744722e-06,
+      "loss": 1.39302549,
+      "memory(GiB)": 103.58,
+      "step": 21745,
+      "train_speed(iter/s)": 1.318623
+    },
+    {
+      "acc": 0.67842617,
+      "epoch": 0.5517503805175038,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.66880321404997e-06,
+      "loss": 1.43328972,
+      "memory(GiB)": 103.58,
+      "step": 21750,
+      "train_speed(iter/s)": 1.318657
+    },
+    {
+      "acc": 0.70428152,
+      "epoch": 0.5518772196854389,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.668090689984872e-06,
+      "loss": 1.36943541,
+      "memory(GiB)": 103.58,
+      "step": 21755,
+      "train_speed(iter/s)": 1.31869
+    },
+    {
+      "acc": 0.68307977,
+      "epoch": 0.5520040588533739,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.667378004580769e-06,
+      "loss": 1.37062597,
+      "memory(GiB)": 103.58,
+      "step": 21760,
+      "train_speed(iter/s)": 1.318724
+    },
+    {
+      "acc": 0.69013653,
+      "epoch": 0.552130898021309,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.666665157869007e-06,
+      "loss": 1.3653019,
+      "memory(GiB)": 103.58,
+      "step": 21765,
+      "train_speed(iter/s)": 1.318759
+    },
+    {
+      "acc": 0.68256741,
+      "epoch": 0.5522577371892441,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.665952149880942e-06,
+      "loss": 1.44375362,
+      "memory(GiB)": 103.58,
+      "step": 21770,
+      "train_speed(iter/s)": 1.318793
+    },
+    {
+      "acc": 0.69917784,
+      "epoch": 0.5523845763571791,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.665238980647934e-06,
+      "loss": 1.30664482,
+      "memory(GiB)": 103.58,
+      "step": 21775,
+      "train_speed(iter/s)": 1.318828
+    },
+    {
+      "acc": 0.67630672,
+      "epoch": 0.5525114155251142,
+      "grad_norm": 3.25,
+      "learning_rate": 8.66452565020135e-06,
+      "loss": 1.42955942,
+      "memory(GiB)": 103.58,
+      "step": 21780,
+      "train_speed(iter/s)": 1.318854
+    },
+    {
+      "acc": 0.68162088,
+      "epoch": 0.5526382546930492,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.663812158572568e-06,
+      "loss": 1.39275665,
+      "memory(GiB)": 103.58,
+      "step": 21785,
+      "train_speed(iter/s)": 1.318882
+    },
+    {
+      "acc": 0.69879026,
+      "epoch": 0.5527650938609843,
+      "grad_norm": 3.953125,
+      "learning_rate": 8.663098505792971e-06,
+      "loss": 1.34660683,
+      "memory(GiB)": 103.58,
+      "step": 21790,
+      "train_speed(iter/s)": 1.318915
+    },
+    {
+      "acc": 0.67512097,
+      "epoch": 0.5528919330289194,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.662384691893947e-06,
+      "loss": 1.38624496,
+      "memory(GiB)": 103.58,
+      "step": 21795,
+      "train_speed(iter/s)": 1.318945
+    },
+    {
+      "acc": 0.68687897,
+      "epoch": 0.5530187721968544,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.661670716906889e-06,
+      "loss": 1.33505497,
+      "memory(GiB)": 103.58,
+      "step": 21800,
+      "train_speed(iter/s)": 1.318976
+    },
+    {
+      "acc": 0.68283324,
+      "epoch": 0.5531456113647895,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.66095658086321e-06,
+      "loss": 1.47881718,
+      "memory(GiB)": 103.58,
+      "step": 21805,
+      "train_speed(iter/s)": 1.319008
+    },
+    {
+      "acc": 0.67644129,
+      "epoch": 0.5532724505327246,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.660242283794312e-06,
+      "loss": 1.38844738,
+      "memory(GiB)": 103.58,
+      "step": 21810,
+      "train_speed(iter/s)": 1.319039
+    },
+    {
+      "acc": 0.67326083,
+      "epoch": 0.5533992897006595,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.659527825731617e-06,
+      "loss": 1.45902071,
+      "memory(GiB)": 103.58,
+      "step": 21815,
+      "train_speed(iter/s)": 1.319071
+    },
+    {
+      "acc": 0.68006086,
+      "epoch": 0.5535261288685946,
+      "grad_norm": 3.375,
+      "learning_rate": 8.65881320670655e-06,
+      "loss": 1.39189339,
+      "memory(GiB)": 103.58,
+      "step": 21820,
+      "train_speed(iter/s)": 1.319103
+    },
+    {
+      "acc": 0.69261656,
+      "epoch": 0.5536529680365296,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.658098426750543e-06,
+      "loss": 1.40551357,
+      "memory(GiB)": 103.58,
+      "step": 21825,
+      "train_speed(iter/s)": 1.319135
+    },
+    {
+      "acc": 0.67593026,
+      "epoch": 0.5537798072044647,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.657383485895034e-06,
+      "loss": 1.47328453,
+      "memory(GiB)": 103.58,
+      "step": 21830,
+      "train_speed(iter/s)": 1.319166
+    },
+    {
+      "acc": 0.70422468,
+      "epoch": 0.5539066463723998,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.656668384171472e-06,
+      "loss": 1.41578026,
+      "memory(GiB)": 103.58,
+      "step": 21835,
+      "train_speed(iter/s)": 1.3192
+    },
+    {
+      "acc": 0.69902172,
+      "epoch": 0.5540334855403348,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.655953121611307e-06,
+      "loss": 1.37696724,
+      "memory(GiB)": 103.58,
+      "step": 21840,
+      "train_speed(iter/s)": 1.319232
+    },
+    {
+      "acc": 0.68047905,
+      "epoch": 0.5541603247082699,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.655237698246002e-06,
+      "loss": 1.41529198,
+      "memory(GiB)": 103.58,
+      "step": 21845,
+      "train_speed(iter/s)": 1.319262
+    },
+    {
+      "acc": 0.68876953,
+      "epoch": 0.554287163876205,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.654522114107024e-06,
+      "loss": 1.34782715,
+      "memory(GiB)": 103.58,
+      "step": 21850,
+      "train_speed(iter/s)": 1.319295
+    },
+    {
+      "acc": 0.69058714,
+      "epoch": 0.55441400304414,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.653806369225846e-06,
+      "loss": 1.38566494,
+      "memory(GiB)": 103.58,
+      "step": 21855,
+      "train_speed(iter/s)": 1.319326
+    },
+    {
+      "acc": 0.67612371,
+      "epoch": 0.5545408422120751,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.65309046363395e-06,
+      "loss": 1.37507133,
+      "memory(GiB)": 103.58,
+      "step": 21860,
+      "train_speed(iter/s)": 1.319358
+    },
+    {
+      "acc": 0.6813529,
+      "epoch": 0.5546676813800101,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.652374397362828e-06,
+      "loss": 1.3775034,
+      "memory(GiB)": 103.58,
+      "step": 21865,
+      "train_speed(iter/s)": 1.319388
+    },
+    {
+      "acc": 0.6926085,
+      "epoch": 0.5547945205479452,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.651658170443972e-06,
+      "loss": 1.31742859,
+      "memory(GiB)": 103.58,
+      "step": 21870,
+      "train_speed(iter/s)": 1.31942
+    },
+    {
+      "acc": 0.690943,
+      "epoch": 0.5549213597158803,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.650941782908886e-06,
+      "loss": 1.35794697,
+      "memory(GiB)": 103.58,
+      "step": 21875,
+      "train_speed(iter/s)": 1.319452
+    },
+    {
+      "acc": 0.70451117,
+      "epoch": 0.5550481988838153,
+      "grad_norm": 2.4375,
+      "learning_rate": 8.65022523478908e-06,
+      "loss": 1.31586504,
+      "memory(GiB)": 103.58,
+      "step": 21880,
+      "train_speed(iter/s)": 1.319484
+    },
+    {
+      "acc": 0.6874918,
+      "epoch": 0.5551750380517504,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.649508526116073e-06,
+      "loss": 1.37494278,
+      "memory(GiB)": 103.58,
+      "step": 21885,
+      "train_speed(iter/s)": 1.319515
+    },
+    {
+      "acc": 0.67641525,
+      "epoch": 0.5553018772196855,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.648791656921384e-06,
+      "loss": 1.37199564,
+      "memory(GiB)": 103.58,
+      "step": 21890,
+      "train_speed(iter/s)": 1.319547
+    },
+    {
+      "acc": 0.69327736,
+      "epoch": 0.5554287163876205,
+      "grad_norm": 3.0,
+      "learning_rate": 8.648074627236549e-06,
+      "loss": 1.37664852,
+      "memory(GiB)": 103.58,
+      "step": 21895,
+      "train_speed(iter/s)": 1.319576
+    },
+    {
+      "acc": 0.69229002,
+      "epoch": 0.5555555555555556,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.647357437093104e-06,
+      "loss": 1.40322418,
+      "memory(GiB)": 103.58,
+      "step": 21900,
+      "train_speed(iter/s)": 1.319607
+    },
+    {
+      "acc": 0.69155636,
+      "epoch": 0.5556823947234906,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.646640086522595e-06,
+      "loss": 1.41391172,
+      "memory(GiB)": 103.58,
+      "step": 21905,
+      "train_speed(iter/s)": 1.31964
+    },
+    {
+      "acc": 0.68392267,
+      "epoch": 0.5558092338914257,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.645922575556575e-06,
+      "loss": 1.45915279,
+      "memory(GiB)": 103.58,
+      "step": 21910,
+      "train_speed(iter/s)": 1.319673
+    },
+    {
+      "acc": 0.68749952,
+      "epoch": 0.5559360730593608,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.645204904226601e-06,
+      "loss": 1.41018496,
+      "memory(GiB)": 103.58,
+      "step": 21915,
+      "train_speed(iter/s)": 1.319704
+    },
+    {
+      "acc": 0.70249777,
+      "epoch": 0.5560629122272958,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.64448707256424e-06,
+      "loss": 1.32205467,
+      "memory(GiB)": 103.58,
+      "step": 21920,
+      "train_speed(iter/s)": 1.319738
+    },
+    {
+      "acc": 0.69245977,
+      "epoch": 0.5561897513952309,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.643769080601067e-06,
+      "loss": 1.37566595,
+      "memory(GiB)": 103.58,
+      "step": 21925,
+      "train_speed(iter/s)": 1.31977
+    },
+    {
+      "acc": 0.70213933,
+      "epoch": 0.556316590563166,
+      "grad_norm": 3.375,
+      "learning_rate": 8.643050928368661e-06,
+      "loss": 1.3180913,
+      "memory(GiB)": 103.58,
+      "step": 21930,
+      "train_speed(iter/s)": 1.319802
+    },
+    {
+      "acc": 0.69639072,
+      "epoch": 0.556443429731101,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.642332615898611e-06,
+      "loss": 1.37450991,
+      "memory(GiB)": 103.58,
+      "step": 21935,
+      "train_speed(iter/s)": 1.319833
+    },
+    {
+      "acc": 0.69697323,
+      "epoch": 0.556570268899036,
+      "grad_norm": 3.125,
+      "learning_rate": 8.64161414322251e-06,
+      "loss": 1.33812685,
+      "memory(GiB)": 103.58,
+      "step": 21940,
+      "train_speed(iter/s)": 1.319864
+    },
+    {
+      "acc": 0.70538177,
+      "epoch": 0.556697108066971,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.64089551037196e-06,
+      "loss": 1.32461243,
+      "memory(GiB)": 103.58,
+      "step": 21945,
+      "train_speed(iter/s)": 1.319897
+    },
+    {
+      "acc": 0.68024015,
+      "epoch": 0.5568239472349061,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.640176717378573e-06,
+      "loss": 1.37821217,
+      "memory(GiB)": 103.58,
+      "step": 21950,
+      "train_speed(iter/s)": 1.319926
+    },
+    {
+      "acc": 0.6900259,
+      "epoch": 0.5569507864028412,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.639457764273957e-06,
+      "loss": 1.42845821,
+      "memory(GiB)": 103.58,
+      "step": 21955,
+      "train_speed(iter/s)": 1.319958
+    },
+    {
+      "acc": 0.6677249,
+      "epoch": 0.5570776255707762,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.638738651089744e-06,
+      "loss": 1.41156883,
+      "memory(GiB)": 103.58,
+      "step": 21960,
+      "train_speed(iter/s)": 1.319988
+    },
+    {
+      "acc": 0.67455978,
+      "epoch": 0.5572044647387113,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.638019377857555e-06,
+      "loss": 1.42325926,
+      "memory(GiB)": 103.58,
+      "step": 21965,
+      "train_speed(iter/s)": 1.320017
+    },
+    {
+      "acc": 0.68378096,
+      "epoch": 0.5573313039066464,
+      "grad_norm": 4.25,
+      "learning_rate": 8.637299944609034e-06,
+      "loss": 1.45616894,
+      "memory(GiB)": 103.58,
+      "step": 21970,
+      "train_speed(iter/s)": 1.320048
+    },
+    {
+      "acc": 0.68750381,
+      "epoch": 0.5574581430745814,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.636580351375821e-06,
+      "loss": 1.40767565,
+      "memory(GiB)": 103.58,
+      "step": 21975,
+      "train_speed(iter/s)": 1.320079
+    },
+    {
+      "acc": 0.67711287,
+      "epoch": 0.5575849822425165,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.635860598189569e-06,
+      "loss": 1.47237701,
+      "memory(GiB)": 103.58,
+      "step": 21980,
+      "train_speed(iter/s)": 1.320109
+    },
+    {
+      "acc": 0.67614717,
+      "epoch": 0.5577118214104515,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.635140685081936e-06,
+      "loss": 1.45283766,
+      "memory(GiB)": 103.58,
+      "step": 21985,
+      "train_speed(iter/s)": 1.32014
+    },
+    {
+      "acc": 0.68812895,
+      "epoch": 0.5578386605783866,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.634420612084583e-06,
+      "loss": 1.37474861,
+      "memory(GiB)": 103.58,
+      "step": 21990,
+      "train_speed(iter/s)": 1.320171
+    },
+    {
+      "acc": 0.68077087,
+      "epoch": 0.5579654997463217,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.633700379229187e-06,
+      "loss": 1.4370657,
+      "memory(GiB)": 103.58,
+      "step": 21995,
+      "train_speed(iter/s)": 1.320201
+    },
+    {
+      "acc": 0.68439646,
+      "epoch": 0.5580923389142567,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.632979986547423e-06,
+      "loss": 1.48154488,
+      "memory(GiB)": 103.58,
+      "step": 22000,
+      "train_speed(iter/s)": 1.320232
+    },
+    {
+      "epoch": 0.5580923389142567,
+      "eval_acc": 0.6737330768893427,
+      "eval_loss": 1.3639215230941772,
+      "eval_runtime": 70.0673,
+      "eval_samples_per_second": 90.913,
+      "eval_steps_per_second": 22.735,
+      "step": 22000
+    },
+    {
+      "acc": 0.67670536,
+      "epoch": 0.5582191780821918,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.632259434070982e-06,
+      "loss": 1.43073139,
+      "memory(GiB)": 103.58,
+      "step": 22005,
+      "train_speed(iter/s)": 1.313755
+    },
+    {
+      "acc": 0.70416508,
+      "epoch": 0.5583460172501269,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.631538721831551e-06,
+      "loss": 1.39402742,
+      "memory(GiB)": 103.58,
+      "step": 22010,
+      "train_speed(iter/s)": 1.313791
+    },
+    {
+      "acc": 0.67957025,
+      "epoch": 0.5584728564180619,
+      "grad_norm": 2.75,
+      "learning_rate": 8.630817849860835e-06,
+      "loss": 1.45598469,
+      "memory(GiB)": 103.58,
+      "step": 22015,
+      "train_speed(iter/s)": 1.313825
+    },
+    {
+      "acc": 0.68967056,
+      "epoch": 0.558599695585997,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.63009681819054e-06,
+      "loss": 1.36878757,
+      "memory(GiB)": 103.58,
+      "step": 22020,
+      "train_speed(iter/s)": 1.313859
+    },
+    {
+      "acc": 0.68768296,
+      "epoch": 0.558726534753932,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.629375626852378e-06,
+      "loss": 1.36794891,
+      "memory(GiB)": 103.58,
+      "step": 22025,
+      "train_speed(iter/s)": 1.313892
+    },
+    {
+      "acc": 0.67991686,
+      "epoch": 0.5588533739218671,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.628654275878074e-06,
+      "loss": 1.44243021,
+      "memory(GiB)": 103.58,
+      "step": 22030,
+      "train_speed(iter/s)": 1.313927
+    },
+    {
+      "acc": 0.68723288,
+      "epoch": 0.5589802130898022,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.627932765299353e-06,
+      "loss": 1.36518459,
+      "memory(GiB)": 103.58,
+      "step": 22035,
+      "train_speed(iter/s)": 1.313961
+    },
+    {
+      "acc": 0.69151793,
+      "epoch": 0.5591070522577372,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.627211095147952e-06,
+      "loss": 1.33951359,
+      "memory(GiB)": 103.58,
+      "step": 22040,
+      "train_speed(iter/s)": 1.31399
+    },
+    {
+      "acc": 0.6961297,
+      "epoch": 0.5592338914256723,
+      "grad_norm": 3.5,
+      "learning_rate": 8.626489265455614e-06,
+      "loss": 1.45995693,
+      "memory(GiB)": 103.58,
+      "step": 22045,
+      "train_speed(iter/s)": 1.314025
+    },
+    {
+      "acc": 0.67792072,
+      "epoch": 0.5593607305936074,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.625767276254084e-06,
+      "loss": 1.44753208,
+      "memory(GiB)": 103.58,
+      "step": 22050,
+      "train_speed(iter/s)": 1.314059
+    },
+    {
+      "acc": 0.67676506,
+      "epoch": 0.5594875697615423,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.625045127575123e-06,
+      "loss": 1.44277239,
+      "memory(GiB)": 103.58,
+      "step": 22055,
+      "train_speed(iter/s)": 1.314093
+    },
+    {
+      "acc": 0.6965497,
+      "epoch": 0.5596144089294774,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.624322819450493e-06,
+      "loss": 1.42375412,
+      "memory(GiB)": 103.58,
+      "step": 22060,
+      "train_speed(iter/s)": 1.314114
+    },
+    {
+      "acc": 0.70092344,
+      "epoch": 0.5597412480974124,
+      "grad_norm": 4.625,
+      "learning_rate": 8.623600351911962e-06,
+      "loss": 1.33233976,
+      "memory(GiB)": 103.58,
+      "step": 22065,
+      "train_speed(iter/s)": 1.314148
+    },
+    {
+      "acc": 0.69446487,
+      "epoch": 0.5598680872653475,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.622877724991312e-06,
+      "loss": 1.4128768,
+      "memory(GiB)": 103.58,
+      "step": 22070,
+      "train_speed(iter/s)": 1.31418
+    },
+    {
+      "acc": 0.69172239,
+      "epoch": 0.5599949264332826,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.622154938720323e-06,
+      "loss": 1.37247095,
+      "memory(GiB)": 103.58,
+      "step": 22075,
+      "train_speed(iter/s)": 1.314214
+    },
+    {
+      "acc": 0.67437234,
+      "epoch": 0.5601217656012176,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.621431993130787e-06,
+      "loss": 1.44269981,
+      "memory(GiB)": 103.58,
+      "step": 22080,
+      "train_speed(iter/s)": 1.314246
+    },
+    {
+      "acc": 0.69657927,
+      "epoch": 0.5602486047691527,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.620708888254506e-06,
+      "loss": 1.37394848,
+      "memory(GiB)": 103.58,
+      "step": 22085,
+      "train_speed(iter/s)": 1.31428
+    },
+    {
+      "acc": 0.6904994,
+      "epoch": 0.5603754439370878,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.619985624123282e-06,
+      "loss": 1.35684299,
+      "memory(GiB)": 103.58,
+      "step": 22090,
+      "train_speed(iter/s)": 1.314314
+    },
+    {
+      "acc": 0.6953063,
+      "epoch": 0.5605022831050228,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.619262200768928e-06,
+      "loss": 1.3169735,
+      "memory(GiB)": 103.58,
+      "step": 22095,
+      "train_speed(iter/s)": 1.314348
+    },
+    {
+      "acc": 0.69508677,
+      "epoch": 0.5606291222729579,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.618538618223262e-06,
+      "loss": 1.37219934,
+      "memory(GiB)": 103.58,
+      "step": 22100,
+      "train_speed(iter/s)": 1.314381
+    },
+    {
+      "acc": 0.69026985,
+      "epoch": 0.5607559614408929,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.617814876518114e-06,
+      "loss": 1.40087242,
+      "memory(GiB)": 103.58,
+      "step": 22105,
+      "train_speed(iter/s)": 1.314414
+    },
+    {
+      "acc": 0.68966417,
+      "epoch": 0.560882800608828,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.617090975685314e-06,
+      "loss": 1.42481966,
+      "memory(GiB)": 103.58,
+      "step": 22110,
+      "train_speed(iter/s)": 1.314446
+    },
+    {
+      "acc": 0.68518486,
+      "epoch": 0.5610096397767631,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.616366915756704e-06,
+      "loss": 1.41290369,
+      "memory(GiB)": 103.58,
+      "step": 22115,
+      "train_speed(iter/s)": 1.314478
+    },
+    {
+      "acc": 0.68784838,
+      "epoch": 0.5611364789446981,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.615642696764131e-06,
+      "loss": 1.41080761,
+      "memory(GiB)": 103.58,
+      "step": 22120,
+      "train_speed(iter/s)": 1.314512
+    },
+    {
+      "acc": 0.68450689,
+      "epoch": 0.5612633181126332,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.614918318739452e-06,
+      "loss": 1.41795979,
+      "memory(GiB)": 103.58,
+      "step": 22125,
+      "train_speed(iter/s)": 1.314547
+    },
+    {
+      "acc": 0.69116921,
+      "epoch": 0.5613901572805683,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.614193781714522e-06,
+      "loss": 1.40302544,
+      "memory(GiB)": 103.58,
+      "step": 22130,
+      "train_speed(iter/s)": 1.314582
+    },
+    {
+      "acc": 0.68341866,
+      "epoch": 0.5615169964485033,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.613469085721215e-06,
+      "loss": 1.39569778,
+      "memory(GiB)": 103.58,
+      "step": 22135,
+      "train_speed(iter/s)": 1.314617
+    },
+    {
+      "acc": 0.68743334,
+      "epoch": 0.5616438356164384,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.612744230791406e-06,
+      "loss": 1.44922428,
+      "memory(GiB)": 103.58,
+      "step": 22140,
+      "train_speed(iter/s)": 1.314642
+    },
+    {
+      "acc": 0.69169493,
+      "epoch": 0.5617706747843734,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.612019216956975e-06,
+      "loss": 1.35267258,
+      "memory(GiB)": 103.58,
+      "step": 22145,
+      "train_speed(iter/s)": 1.314677
+    },
+    {
+      "acc": 0.67839465,
+      "epoch": 0.5618975139523085,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.611294044249811e-06,
+      "loss": 1.39433002,
+      "memory(GiB)": 103.58,
+      "step": 22150,
+      "train_speed(iter/s)": 1.314712
+    },
+    {
+      "acc": 0.68209238,
+      "epoch": 0.5620243531202436,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.610568712701814e-06,
+      "loss": 1.39559574,
+      "memory(GiB)": 103.58,
+      "step": 22155,
+      "train_speed(iter/s)": 1.314745
+    },
+    {
+      "acc": 0.67739353,
+      "epoch": 0.5621511922881786,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.609843222344883e-06,
+      "loss": 1.38974924,
+      "memory(GiB)": 103.58,
+      "step": 22160,
+      "train_speed(iter/s)": 1.314778
+    },
+    {
+      "acc": 0.70033498,
+      "epoch": 0.5622780314561137,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.609117573210931e-06,
+      "loss": 1.29239836,
+      "memory(GiB)": 103.58,
+      "step": 22165,
+      "train_speed(iter/s)": 1.314812
+    },
+    {
+      "acc": 0.68533311,
+      "epoch": 0.5624048706240488,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.608391765331875e-06,
+      "loss": 1.38021412,
+      "memory(GiB)": 103.58,
+      "step": 22170,
+      "train_speed(iter/s)": 1.314845
+    },
+    {
+      "acc": 0.69824524,
+      "epoch": 0.5625317097919837,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.607665798739638e-06,
+      "loss": 1.37819595,
+      "memory(GiB)": 103.58,
+      "step": 22175,
+      "train_speed(iter/s)": 1.314879
+    },
+    {
+      "acc": 0.69016714,
+      "epoch": 0.5626585489599188,
+      "grad_norm": 3.25,
+      "learning_rate": 8.606939673466153e-06,
+      "loss": 1.37345953,
+      "memory(GiB)": 103.58,
+      "step": 22180,
+      "train_speed(iter/s)": 1.314912
+    },
+    {
+      "acc": 0.69194822,
+      "epoch": 0.5627853881278538,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.606213389543356e-06,
+      "loss": 1.3654398,
+      "memory(GiB)": 103.58,
+      "step": 22185,
+      "train_speed(iter/s)": 1.314943
+    },
+    {
+      "acc": 0.68145275,
+      "epoch": 0.5629122272957889,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.605486947003194e-06,
+      "loss": 1.45338411,
+      "memory(GiB)": 103.58,
+      "step": 22190,
+      "train_speed(iter/s)": 1.314973
+    },
+    {
+      "acc": 0.67328143,
+      "epoch": 0.563039066463724,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.60476034587762e-06,
+      "loss": 1.46318741,
+      "memory(GiB)": 103.58,
+      "step": 22195,
+      "train_speed(iter/s)": 1.315005
+    },
+    {
+      "acc": 0.69285536,
+      "epoch": 0.563165905631659,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.604033586198592e-06,
+      "loss": 1.36712265,
+      "memory(GiB)": 103.58,
+      "step": 22200,
+      "train_speed(iter/s)": 1.315037
+    },
+    {
+      "acc": 0.6700439,
+      "epoch": 0.5632927447995941,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.603306667998074e-06,
+      "loss": 1.43961611,
+      "memory(GiB)": 103.58,
+      "step": 22205,
+      "train_speed(iter/s)": 1.315069
+    },
+    {
+      "acc": 0.67749567,
+      "epoch": 0.5634195839675292,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.602579591308043e-06,
+      "loss": 1.42536697,
+      "memory(GiB)": 103.58,
+      "step": 22210,
+      "train_speed(iter/s)": 1.315097
+    },
+    {
+      "acc": 0.66245956,
+      "epoch": 0.5635464231354642,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.601852356160476e-06,
+      "loss": 1.49403725,
+      "memory(GiB)": 103.58,
+      "step": 22215,
+      "train_speed(iter/s)": 1.315127
+    },
+    {
+      "acc": 0.68984466,
+      "epoch": 0.5636732623033993,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.60112496258736e-06,
+      "loss": 1.39255142,
+      "memory(GiB)": 103.58,
+      "step": 22220,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.67460222,
+      "epoch": 0.5638001014713343,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.600397410620693e-06,
+      "loss": 1.42522879,
+      "memory(GiB)": 103.58,
+      "step": 22225,
+      "train_speed(iter/s)": 1.315191
+    },
+    {
+      "acc": 0.67878461,
+      "epoch": 0.5639269406392694,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.599669700292472e-06,
+      "loss": 1.41297903,
+      "memory(GiB)": 103.58,
+      "step": 22230,
+      "train_speed(iter/s)": 1.315223
+    },
+    {
+      "acc": 0.69519453,
+      "epoch": 0.5640537798072045,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.598941831634707e-06,
+      "loss": 1.43512764,
+      "memory(GiB)": 103.58,
+      "step": 22235,
+      "train_speed(iter/s)": 1.315256
+    },
+    {
+      "acc": 0.67032118,
+      "epoch": 0.5641806189751395,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.598213804679412e-06,
+      "loss": 1.43162794,
+      "memory(GiB)": 103.58,
+      "step": 22240,
+      "train_speed(iter/s)": 1.315289
+    },
+    {
+      "acc": 0.68137221,
+      "epoch": 0.5643074581430746,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.597485619458609e-06,
+      "loss": 1.45515289,
+      "memory(GiB)": 103.58,
+      "step": 22245,
+      "train_speed(iter/s)": 1.315323
+    },
+    {
+      "acc": 0.6802403,
+      "epoch": 0.5644342973110097,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.596757276004327e-06,
+      "loss": 1.40545044,
+      "memory(GiB)": 103.58,
+      "step": 22250,
+      "train_speed(iter/s)": 1.315356
+    },
+    {
+      "acc": 0.70230427,
+      "epoch": 0.5645611364789447,
+      "grad_norm": 3.5,
+      "learning_rate": 8.5960287743486e-06,
+      "loss": 1.44522018,
+      "memory(GiB)": 103.58,
+      "step": 22255,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.68975143,
+      "epoch": 0.5646879756468798,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.595300114523473e-06,
+      "loss": 1.35846024,
+      "memory(GiB)": 103.58,
+      "step": 22260,
+      "train_speed(iter/s)": 1.315423
+    },
+    {
+      "acc": 0.68982787,
+      "epoch": 0.5648148148148148,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.594571296560997e-06,
+      "loss": 1.45822639,
+      "memory(GiB)": 103.58,
+      "step": 22265,
+      "train_speed(iter/s)": 1.315456
+    },
+    {
+      "acc": 0.67856035,
+      "epoch": 0.5649416539827499,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.593842320493224e-06,
+      "loss": 1.47334623,
+      "memory(GiB)": 103.58,
+      "step": 22270,
+      "train_speed(iter/s)": 1.315489
+    },
+    {
+      "acc": 0.68532152,
+      "epoch": 0.565068493150685,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.593113186352222e-06,
+      "loss": 1.45072479,
+      "memory(GiB)": 103.58,
+      "step": 22275,
+      "train_speed(iter/s)": 1.315526
+    },
+    {
+      "acc": 0.66880827,
+      "epoch": 0.56519533231862,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.592383894170059e-06,
+      "loss": 1.44978619,
+      "memory(GiB)": 103.58,
+      "step": 22280,
+      "train_speed(iter/s)": 1.315559
+    },
+    {
+      "acc": 0.69491382,
+      "epoch": 0.5653221714865551,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.591654443978815e-06,
+      "loss": 1.37982616,
+      "memory(GiB)": 103.58,
+      "step": 22285,
+      "train_speed(iter/s)": 1.315592
+    },
+    {
+      "acc": 0.68551474,
+      "epoch": 0.5654490106544902,
+      "grad_norm": 3.125,
+      "learning_rate": 8.590924835810572e-06,
+      "loss": 1.39487133,
+      "memory(GiB)": 103.58,
+      "step": 22290,
+      "train_speed(iter/s)": 1.315624
+    },
+    {
+      "acc": 0.69383621,
+      "epoch": 0.5655758498224251,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.590195069697423e-06,
+      "loss": 1.36038055,
+      "memory(GiB)": 103.58,
+      "step": 22295,
+      "train_speed(iter/s)": 1.315657
+    },
+    {
+      "acc": 0.6802803,
+      "epoch": 0.5657026889903602,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.589465145671465e-06,
+      "loss": 1.39665928,
+      "memory(GiB)": 103.58,
+      "step": 22300,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.67497845,
+      "epoch": 0.5658295281582952,
+      "grad_norm": 2.875,
+      "learning_rate": 8.588735063764803e-06,
+      "loss": 1.40972195,
+      "memory(GiB)": 103.58,
+      "step": 22305,
+      "train_speed(iter/s)": 1.315722
+    },
+    {
+      "acc": 0.68772907,
+      "epoch": 0.5659563673262303,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.588004824009552e-06,
+      "loss": 1.46027517,
+      "memory(GiB)": 103.58,
+      "step": 22310,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.67897553,
+      "epoch": 0.5660832064941654,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.58727442643783e-06,
+      "loss": 1.47205524,
+      "memory(GiB)": 103.58,
+      "step": 22315,
+      "train_speed(iter/s)": 1.315783
+    },
+    {
+      "acc": 0.67366457,
+      "epoch": 0.5662100456621004,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.586543871081764e-06,
+      "loss": 1.41295843,
+      "memory(GiB)": 103.58,
+      "step": 22320,
+      "train_speed(iter/s)": 1.315814
+    },
+    {
+      "acc": 0.70038419,
+      "epoch": 0.5663368848300355,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.585813157973482e-06,
+      "loss": 1.32712326,
+      "memory(GiB)": 103.58,
+      "step": 22325,
+      "train_speed(iter/s)": 1.315846
+    },
+    {
+      "acc": 0.67047396,
+      "epoch": 0.5664637239979706,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.58508228714513e-06,
+      "loss": 1.37684174,
+      "memory(GiB)": 103.58,
+      "step": 22330,
+      "train_speed(iter/s)": 1.315878
+    },
+    {
+      "acc": 0.692483,
+      "epoch": 0.5665905631659056,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.584351258628852e-06,
+      "loss": 1.36952658,
+      "memory(GiB)": 103.58,
+      "step": 22335,
+      "train_speed(iter/s)": 1.315909
+    },
+    {
+      "acc": 0.70516009,
+      "epoch": 0.5667174023338407,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.583620072456803e-06,
+      "loss": 1.26512842,
+      "memory(GiB)": 103.58,
+      "step": 22340,
+      "train_speed(iter/s)": 1.31594
+    },
+    {
+      "acc": 0.69757152,
+      "epoch": 0.5668442415017757,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.582888728661142e-06,
+      "loss": 1.37533522,
+      "memory(GiB)": 103.58,
+      "step": 22345,
+      "train_speed(iter/s)": 1.31597
+    },
+    {
+      "acc": 0.6796021,
+      "epoch": 0.5669710806697108,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.582157227274042e-06,
+      "loss": 1.44193592,
+      "memory(GiB)": 103.58,
+      "step": 22350,
+      "train_speed(iter/s)": 1.316001
+    },
+    {
+      "acc": 0.68414264,
+      "epoch": 0.5670979198376459,
+      "grad_norm": 3.5,
+      "learning_rate": 8.581425568327671e-06,
+      "loss": 1.40216579,
+      "memory(GiB)": 103.58,
+      "step": 22355,
+      "train_speed(iter/s)": 1.316034
+    },
+    {
+      "acc": 0.67027936,
+      "epoch": 0.5672247590055809,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.580693751854215e-06,
+      "loss": 1.4400815,
+      "memory(GiB)": 103.58,
+      "step": 22360,
+      "train_speed(iter/s)": 1.316067
+    },
+    {
+      "acc": 0.69171529,
+      "epoch": 0.567351598173516,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.57996177788586e-06,
+      "loss": 1.37898169,
+      "memory(GiB)": 103.58,
+      "step": 22365,
+      "train_speed(iter/s)": 1.316099
+    },
+    {
+      "acc": 0.68049021,
+      "epoch": 0.5674784373414511,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.579229646454803e-06,
+      "loss": 1.40231438,
+      "memory(GiB)": 103.58,
+      "step": 22370,
+      "train_speed(iter/s)": 1.316131
+    },
+    {
+      "acc": 0.70239649,
+      "epoch": 0.5676052765093861,
+      "grad_norm": 4.5,
+      "learning_rate": 8.578497357593246e-06,
+      "loss": 1.32684326,
+      "memory(GiB)": 103.58,
+      "step": 22375,
+      "train_speed(iter/s)": 1.316164
+    },
+    {
+      "acc": 0.69832478,
+      "epoch": 0.5677321156773212,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.5777649113334e-06,
+      "loss": 1.29630547,
+      "memory(GiB)": 103.58,
+      "step": 22380,
+      "train_speed(iter/s)": 1.316192
+    },
+    {
+      "acc": 0.68836079,
+      "epoch": 0.5678589548452562,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.577032307707476e-06,
+      "loss": 1.36448555,
+      "memory(GiB)": 103.58,
+      "step": 22385,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.66776376,
+      "epoch": 0.5679857940131913,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.576299546747704e-06,
+      "loss": 1.49165401,
+      "memory(GiB)": 103.58,
+      "step": 22390,
+      "train_speed(iter/s)": 1.31626
+    },
+    {
+      "acc": 0.68910141,
+      "epoch": 0.5681126331811264,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.575566628486309e-06,
+      "loss": 1.41611328,
+      "memory(GiB)": 103.58,
+      "step": 22395,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.66652031,
+      "epoch": 0.5682394723490614,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.574833552955532e-06,
+      "loss": 1.44135876,
+      "memory(GiB)": 103.58,
+      "step": 22400,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.68174224,
+      "epoch": 0.5683663115169965,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.574100320187612e-06,
+      "loss": 1.44030952,
+      "memory(GiB)": 103.58,
+      "step": 22405,
+      "train_speed(iter/s)": 1.316357
+    },
+    {
+      "acc": 0.6834734,
+      "epoch": 0.5684931506849316,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.573366930214807e-06,
+      "loss": 1.45306911,
+      "memory(GiB)": 103.58,
+      "step": 22410,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.66188965,
+      "epoch": 0.5686199898528665,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.572633383069366e-06,
+      "loss": 1.46069546,
+      "memory(GiB)": 103.58,
+      "step": 22415,
+      "train_speed(iter/s)": 1.31642
+    },
+    {
+      "acc": 0.68169231,
+      "epoch": 0.5687468290208016,
+      "grad_norm": 5.78125,
+      "learning_rate": 8.571899678783561e-06,
+      "loss": 1.40892811,
+      "memory(GiB)": 103.58,
+      "step": 22420,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.68211603,
+      "epoch": 0.5688736681887366,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.57116581738966e-06,
+      "loss": 1.40922565,
+      "memory(GiB)": 103.58,
+      "step": 22425,
+      "train_speed(iter/s)": 1.316484
+    },
+    {
+      "acc": 0.69867277,
+      "epoch": 0.5690005073566717,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.570431798919941e-06,
+      "loss": 1.37000322,
+      "memory(GiB)": 103.58,
+      "step": 22430,
+      "train_speed(iter/s)": 1.316514
+    },
+    {
+      "acc": 0.69596953,
+      "epoch": 0.5691273465246068,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.569697623406692e-06,
+      "loss": 1.33943501,
+      "memory(GiB)": 103.58,
+      "step": 22435,
+      "train_speed(iter/s)": 1.316544
+    },
+    {
+      "acc": 0.67701931,
+      "epoch": 0.5692541856925418,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.568963290882204e-06,
+      "loss": 1.41908112,
+      "memory(GiB)": 103.58,
+      "step": 22440,
+      "train_speed(iter/s)": 1.316576
+    },
+    {
+      "acc": 0.69199104,
+      "epoch": 0.5693810248604769,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.568228801378775e-06,
+      "loss": 1.31647663,
+      "memory(GiB)": 103.58,
+      "step": 22445,
+      "train_speed(iter/s)": 1.316608
+    },
+    {
+      "acc": 0.6883152,
+      "epoch": 0.569507864028412,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.567494154928713e-06,
+      "loss": 1.37554798,
+      "memory(GiB)": 103.58,
+      "step": 22450,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.67274694,
+      "epoch": 0.569634703196347,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.566759351564332e-06,
+      "loss": 1.40477886,
+      "memory(GiB)": 103.58,
+      "step": 22455,
+      "train_speed(iter/s)": 1.316672
+    },
+    {
+      "acc": 0.67578583,
+      "epoch": 0.5697615423642821,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.566024391317947e-06,
+      "loss": 1.44917803,
+      "memory(GiB)": 103.58,
+      "step": 22460,
+      "train_speed(iter/s)": 1.316702
+    },
+    {
+      "acc": 0.68651385,
+      "epoch": 0.5698883815322171,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.565289274221891e-06,
+      "loss": 1.38788052,
+      "memory(GiB)": 103.58,
+      "step": 22465,
+      "train_speed(iter/s)": 1.316736
+    },
+    {
+      "acc": 0.69504766,
+      "epoch": 0.5700152207001522,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.564554000308493e-06,
+      "loss": 1.37280579,
+      "memory(GiB)": 103.58,
+      "step": 22470,
+      "train_speed(iter/s)": 1.316769
+    },
+    {
+      "acc": 0.68025961,
+      "epoch": 0.5701420598680873,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.563818569610096e-06,
+      "loss": 1.39694471,
+      "memory(GiB)": 103.58,
+      "step": 22475,
+      "train_speed(iter/s)": 1.316801
+    },
+    {
+      "acc": 0.67794375,
+      "epoch": 0.5702688990360223,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.563082982159048e-06,
+      "loss": 1.40016174,
+      "memory(GiB)": 103.58,
+      "step": 22480,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.69289107,
+      "epoch": 0.5703957382039574,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.562347237987701e-06,
+      "loss": 1.41540489,
+      "memory(GiB)": 103.58,
+      "step": 22485,
+      "train_speed(iter/s)": 1.316865
+    },
+    {
+      "acc": 0.6692091,
+      "epoch": 0.5705225773718925,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.561611337128418e-06,
+      "loss": 1.45408087,
+      "memory(GiB)": 103.58,
+      "step": 22490,
+      "train_speed(iter/s)": 1.316897
+    },
+    {
+      "acc": 0.69023523,
+      "epoch": 0.5706494165398275,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.560875279613568e-06,
+      "loss": 1.36794996,
+      "memory(GiB)": 103.58,
+      "step": 22495,
+      "train_speed(iter/s)": 1.316929
+    },
+    {
+      "acc": 0.67681646,
+      "epoch": 0.5707762557077626,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.560139065475523e-06,
+      "loss": 1.45121107,
+      "memory(GiB)": 103.58,
+      "step": 22500,
+      "train_speed(iter/s)": 1.316959
+    },
+    {
+      "acc": 0.70125952,
+      "epoch": 0.5709030948756976,
+      "grad_norm": 3.375,
+      "learning_rate": 8.559402694746671e-06,
+      "loss": 1.37287998,
+      "memory(GiB)": 103.58,
+      "step": 22505,
+      "train_speed(iter/s)": 1.316991
+    },
+    {
+      "acc": 0.69590445,
+      "epoch": 0.5710299340436327,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.558666167459393e-06,
+      "loss": 1.30867023,
+      "memory(GiB)": 103.58,
+      "step": 22510,
+      "train_speed(iter/s)": 1.317024
+    },
+    {
+      "acc": 0.67869415,
+      "epoch": 0.5711567732115678,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.55792948364609e-06,
+      "loss": 1.51617165,
+      "memory(GiB)": 103.58,
+      "step": 22515,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.70391002,
+      "epoch": 0.5712836123795028,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.557192643339164e-06,
+      "loss": 1.43244667,
+      "memory(GiB)": 103.58,
+      "step": 22520,
+      "train_speed(iter/s)": 1.317087
+    },
+    {
+      "acc": 0.69364109,
+      "epoch": 0.5714104515474379,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.556455646571022e-06,
+      "loss": 1.389641,
+      "memory(GiB)": 103.58,
+      "step": 22525,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.69410934,
+      "epoch": 0.571537290715373,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.555718493374084e-06,
+      "loss": 1.43295612,
+      "memory(GiB)": 103.58,
+      "step": 22530,
+      "train_speed(iter/s)": 1.317147
+    },
+    {
+      "acc": 0.68920794,
+      "epoch": 0.571664129883308,
+      "grad_norm": 3.0,
+      "learning_rate": 8.55498118378077e-06,
+      "loss": 1.42842884,
+      "memory(GiB)": 103.58,
+      "step": 22535,
+      "train_speed(iter/s)": 1.317179
+    },
+    {
+      "acc": 0.70578671,
+      "epoch": 0.571790969051243,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.554243717823512e-06,
+      "loss": 1.29794064,
+      "memory(GiB)": 103.58,
+      "step": 22540,
+      "train_speed(iter/s)": 1.317211
+    },
+    {
+      "acc": 0.68910089,
+      "epoch": 0.571917808219178,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.553506095534747e-06,
+      "loss": 1.42111511,
+      "memory(GiB)": 103.58,
+      "step": 22545,
+      "train_speed(iter/s)": 1.317244
+    },
+    {
+      "acc": 0.68680038,
+      "epoch": 0.5720446473871131,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.55276831694692e-06,
+      "loss": 1.34351826,
+      "memory(GiB)": 103.58,
+      "step": 22550,
+      "train_speed(iter/s)": 1.317276
+    },
+    {
+      "acc": 0.69124632,
+      "epoch": 0.5721714865550482,
+      "grad_norm": 3.125,
+      "learning_rate": 8.552030382092477e-06,
+      "loss": 1.40220394,
+      "memory(GiB)": 103.58,
+      "step": 22555,
+      "train_speed(iter/s)": 1.317308
+    },
+    {
+      "acc": 0.69898829,
+      "epoch": 0.5722983257229832,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.551292291003884e-06,
+      "loss": 1.40948639,
+      "memory(GiB)": 103.58,
+      "step": 22560,
+      "train_speed(iter/s)": 1.317341
+    },
+    {
+      "acc": 0.68655305,
+      "epoch": 0.5724251648909183,
+      "grad_norm": 4.25,
+      "learning_rate": 8.550554043713597e-06,
+      "loss": 1.40371819,
+      "memory(GiB)": 103.58,
+      "step": 22565,
+      "train_speed(iter/s)": 1.317374
+    },
+    {
+      "acc": 0.69261394,
+      "epoch": 0.5725520040588534,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.549815640254092e-06,
+      "loss": 1.3510047,
+      "memory(GiB)": 103.58,
+      "step": 22570,
+      "train_speed(iter/s)": 1.317406
+    },
+    {
+      "acc": 0.69140978,
+      "epoch": 0.5726788432267884,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.549077080657846e-06,
+      "loss": 1.39633484,
+      "memory(GiB)": 103.58,
+      "step": 22575,
+      "train_speed(iter/s)": 1.317439
+    },
+    {
+      "acc": 0.6733027,
+      "epoch": 0.5728056823947235,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.548338364957345e-06,
+      "loss": 1.46119957,
+      "memory(GiB)": 103.58,
+      "step": 22580,
+      "train_speed(iter/s)": 1.317469
+    },
+    {
+      "acc": 0.68008537,
+      "epoch": 0.5729325215626585,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.54759949318508e-06,
+      "loss": 1.41135273,
+      "memory(GiB)": 103.58,
+      "step": 22585,
+      "train_speed(iter/s)": 1.317502
+    },
+    {
+      "acc": 0.68118801,
+      "epoch": 0.5730593607305936,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.546860465373552e-06,
+      "loss": 1.37119751,
+      "memory(GiB)": 103.58,
+      "step": 22590,
+      "train_speed(iter/s)": 1.317533
+    },
+    {
+      "acc": 0.68806405,
+      "epoch": 0.5731861998985287,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.546121281555265e-06,
+      "loss": 1.35836897,
+      "memory(GiB)": 103.58,
+      "step": 22595,
+      "train_speed(iter/s)": 1.317563
+    },
+    {
+      "acc": 0.6864778,
+      "epoch": 0.5733130390664637,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.54538194176273e-06,
+      "loss": 1.353545,
+      "memory(GiB)": 103.58,
+      "step": 22600,
+      "train_speed(iter/s)": 1.317595
+    },
+    {
+      "acc": 0.68520694,
+      "epoch": 0.5734398782343988,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.544642446028469e-06,
+      "loss": 1.38403797,
+      "memory(GiB)": 103.58,
+      "step": 22605,
+      "train_speed(iter/s)": 1.317628
+    },
+    {
+      "acc": 0.66016979,
+      "epoch": 0.5735667174023339,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.543902794385008e-06,
+      "loss": 1.51416664,
+      "memory(GiB)": 103.58,
+      "step": 22610,
+      "train_speed(iter/s)": 1.31766
+    },
+    {
+      "acc": 0.67661772,
+      "epoch": 0.5736935565702689,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.543162986864879e-06,
+      "loss": 1.45265903,
+      "memory(GiB)": 103.58,
+      "step": 22615,
+      "train_speed(iter/s)": 1.317692
+    },
+    {
+      "acc": 0.70183229,
+      "epoch": 0.573820395738204,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.542423023500623e-06,
+      "loss": 1.34232855,
+      "memory(GiB)": 103.58,
+      "step": 22620,
+      "train_speed(iter/s)": 1.317724
+    },
+    {
+      "acc": 0.68229547,
+      "epoch": 0.573947234906139,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.541682904324786e-06,
+      "loss": 1.4139534,
+      "memory(GiB)": 103.58,
+      "step": 22625,
+      "train_speed(iter/s)": 1.317756
+    },
+    {
+      "acc": 0.6865922,
+      "epoch": 0.5740740740740741,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.540942629369923e-06,
+      "loss": 1.44951668,
+      "memory(GiB)": 103.58,
+      "step": 22630,
+      "train_speed(iter/s)": 1.317789
+    },
+    {
+      "acc": 0.6598402,
+      "epoch": 0.5742009132420092,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.540202198668595e-06,
+      "loss": 1.42304773,
+      "memory(GiB)": 103.58,
+      "step": 22635,
+      "train_speed(iter/s)": 1.317823
+    },
+    {
+      "acc": 0.69081306,
+      "epoch": 0.5743277524099442,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.539461612253368e-06,
+      "loss": 1.36454878,
+      "memory(GiB)": 103.58,
+      "step": 22640,
+      "train_speed(iter/s)": 1.317854
+    },
+    {
+      "acc": 0.68329759,
+      "epoch": 0.5744545915778793,
+      "grad_norm": 3.125,
+      "learning_rate": 8.538720870156816e-06,
+      "loss": 1.39249449,
+      "memory(GiB)": 103.58,
+      "step": 22645,
+      "train_speed(iter/s)": 1.317886
+    },
+    {
+      "acc": 0.68559194,
+      "epoch": 0.5745814307458144,
+      "grad_norm": 2.875,
+      "learning_rate": 8.53797997241152e-06,
+      "loss": 1.346562,
+      "memory(GiB)": 103.58,
+      "step": 22650,
+      "train_speed(iter/s)": 1.317918
+    },
+    {
+      "acc": 0.69659696,
+      "epoch": 0.5747082699137493,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.537238919050071e-06,
+      "loss": 1.37031517,
+      "memory(GiB)": 103.58,
+      "step": 22655,
+      "train_speed(iter/s)": 1.317951
+    },
+    {
+      "acc": 0.69836645,
+      "epoch": 0.5748351090816844,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.53649771010506e-06,
+      "loss": 1.34932232,
+      "memory(GiB)": 103.58,
+      "step": 22660,
+      "train_speed(iter/s)": 1.317983
+    },
+    {
+      "acc": 0.6786891,
+      "epoch": 0.5749619482496194,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.535756345609092e-06,
+      "loss": 1.39417238,
+      "memory(GiB)": 103.58,
+      "step": 22665,
+      "train_speed(iter/s)": 1.318015
+    },
+    {
+      "acc": 0.68958988,
+      "epoch": 0.5750887874175545,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.535014825594772e-06,
+      "loss": 1.38276291,
+      "memory(GiB)": 103.58,
+      "step": 22670,
+      "train_speed(iter/s)": 1.318045
+    },
+    {
+      "acc": 0.69009175,
+      "epoch": 0.5752156265854896,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.534273150094718e-06,
+      "loss": 1.38707867,
+      "memory(GiB)": 103.58,
+      "step": 22675,
+      "train_speed(iter/s)": 1.318074
+    },
+    {
+      "acc": 0.67197318,
+      "epoch": 0.5753424657534246,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.533531319141552e-06,
+      "loss": 1.36708584,
+      "memory(GiB)": 103.58,
+      "step": 22680,
+      "train_speed(iter/s)": 1.318106
+    },
+    {
+      "acc": 0.6679081,
+      "epoch": 0.5754693049213597,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.532789332767902e-06,
+      "loss": 1.48022423,
+      "memory(GiB)": 103.58,
+      "step": 22685,
+      "train_speed(iter/s)": 1.318139
+    },
+    {
+      "acc": 0.68914471,
+      "epoch": 0.5755961440892948,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.532047191006405e-06,
+      "loss": 1.3779705,
+      "memory(GiB)": 103.58,
+      "step": 22690,
+      "train_speed(iter/s)": 1.318171
+    },
+    {
+      "acc": 0.67586756,
+      "epoch": 0.5757229832572298,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.531304893889702e-06,
+      "loss": 1.41243458,
+      "memory(GiB)": 103.58,
+      "step": 22695,
+      "train_speed(iter/s)": 1.318204
+    },
+    {
+      "acc": 0.6944953,
+      "epoch": 0.5758498224251649,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.530562441450445e-06,
+      "loss": 1.40052681,
+      "memory(GiB)": 103.58,
+      "step": 22700,
+      "train_speed(iter/s)": 1.318235
+    },
+    {
+      "acc": 0.68782973,
+      "epoch": 0.5759766615930999,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.529819833721289e-06,
+      "loss": 1.39248085,
+      "memory(GiB)": 103.58,
+      "step": 22705,
+      "train_speed(iter/s)": 1.318266
+    },
+    {
+      "acc": 0.67965975,
+      "epoch": 0.576103500761035,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.529077070734896e-06,
+      "loss": 1.4469389,
+      "memory(GiB)": 103.58,
+      "step": 22710,
+      "train_speed(iter/s)": 1.318298
+    },
+    {
+      "acc": 0.68033724,
+      "epoch": 0.5762303399289701,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.528334152523938e-06,
+      "loss": 1.45253849,
+      "memory(GiB)": 103.58,
+      "step": 22715,
+      "train_speed(iter/s)": 1.318327
+    },
+    {
+      "acc": 0.70252595,
+      "epoch": 0.5763571790969051,
+      "grad_norm": 3.25,
+      "learning_rate": 8.52759107912109e-06,
+      "loss": 1.33818083,
+      "memory(GiB)": 103.58,
+      "step": 22720,
+      "train_speed(iter/s)": 1.318357
+    },
+    {
+      "acc": 0.69085312,
+      "epoch": 0.5764840182648402,
+      "grad_norm": 3.0,
+      "learning_rate": 8.526847850559037e-06,
+      "loss": 1.36938391,
+      "memory(GiB)": 103.58,
+      "step": 22725,
+      "train_speed(iter/s)": 1.318387
+    },
+    {
+      "acc": 0.67822714,
+      "epoch": 0.5766108574327753,
+      "grad_norm": 2.5,
+      "learning_rate": 8.526104466870472e-06,
+      "loss": 1.43836184,
+      "memory(GiB)": 103.58,
+      "step": 22730,
+      "train_speed(iter/s)": 1.318416
+    },
+    {
+      "acc": 0.69662476,
+      "epoch": 0.5767376966007103,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.525360928088087e-06,
+      "loss": 1.35514126,
+      "memory(GiB)": 103.58,
+      "step": 22735,
+      "train_speed(iter/s)": 1.318443
+    },
+    {
+      "acc": 0.69485478,
+      "epoch": 0.5768645357686454,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.524617234244588e-06,
+      "loss": 1.38328686,
+      "memory(GiB)": 103.58,
+      "step": 22740,
+      "train_speed(iter/s)": 1.318472
+    },
+    {
+      "acc": 0.68198972,
+      "epoch": 0.5769913749365804,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.523873385372687e-06,
+      "loss": 1.31504192,
+      "memory(GiB)": 103.58,
+      "step": 22745,
+      "train_speed(iter/s)": 1.318495
+    },
+    {
+      "acc": 0.68203735,
+      "epoch": 0.5771182141045155,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.523129381505104e-06,
+      "loss": 1.48474121,
+      "memory(GiB)": 103.58,
+      "step": 22750,
+      "train_speed(iter/s)": 1.318526
+    },
+    {
+      "acc": 0.70019255,
+      "epoch": 0.5772450532724506,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.522385222674559e-06,
+      "loss": 1.33784771,
+      "memory(GiB)": 103.58,
+      "step": 22755,
+      "train_speed(iter/s)": 1.318557
+    },
+    {
+      "acc": 0.6924468,
+      "epoch": 0.5773718924403856,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.521640908913787e-06,
+      "loss": 1.34045362,
+      "memory(GiB)": 103.58,
+      "step": 22760,
+      "train_speed(iter/s)": 1.31859
+    },
+    {
+      "acc": 0.69122066,
+      "epoch": 0.5774987316083207,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.520896440255524e-06,
+      "loss": 1.41776447,
+      "memory(GiB)": 103.58,
+      "step": 22765,
+      "train_speed(iter/s)": 1.318617
+    },
+    {
+      "acc": 0.69691982,
+      "epoch": 0.5776255707762558,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.520151816732517e-06,
+      "loss": 1.3593895,
+      "memory(GiB)": 103.58,
+      "step": 22770,
+      "train_speed(iter/s)": 1.31865
+    },
+    {
+      "acc": 0.6767189,
+      "epoch": 0.5777524099441907,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.519407038377515e-06,
+      "loss": 1.49625988,
+      "memory(GiB)": 103.58,
+      "step": 22775,
+      "train_speed(iter/s)": 1.318682
+    },
+    {
+      "acc": 0.68398275,
+      "epoch": 0.5778792491121258,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.518662105223279e-06,
+      "loss": 1.45057077,
+      "memory(GiB)": 103.58,
+      "step": 22780,
+      "train_speed(iter/s)": 1.318714
+    },
+    {
+      "acc": 0.68670979,
+      "epoch": 0.5780060882800608,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.517917017302574e-06,
+      "loss": 1.42375345,
+      "memory(GiB)": 103.58,
+      "step": 22785,
+      "train_speed(iter/s)": 1.318747
+    },
+    {
+      "acc": 0.68529229,
+      "epoch": 0.5781329274479959,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.517171774648172e-06,
+      "loss": 1.44700279,
+      "memory(GiB)": 103.58,
+      "step": 22790,
+      "train_speed(iter/s)": 1.318774
+    },
+    {
+      "acc": 0.67760234,
+      "epoch": 0.578259766615931,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.516426377292854e-06,
+      "loss": 1.40523119,
+      "memory(GiB)": 103.58,
+      "step": 22795,
+      "train_speed(iter/s)": 1.318806
+    },
+    {
+      "acc": 0.68498087,
+      "epoch": 0.578386605783866,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.515680825269404e-06,
+      "loss": 1.39658766,
+      "memory(GiB)": 103.58,
+      "step": 22800,
+      "train_speed(iter/s)": 1.318839
+    },
+    {
+      "acc": 0.68511329,
+      "epoch": 0.5785134449518011,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.514935118610613e-06,
+      "loss": 1.46241293,
+      "memory(GiB)": 103.58,
+      "step": 22805,
+      "train_speed(iter/s)": 1.318871
+    },
+    {
+      "acc": 0.68872557,
+      "epoch": 0.5786402841197362,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.514189257349283e-06,
+      "loss": 1.39911098,
+      "memory(GiB)": 103.58,
+      "step": 22810,
+      "train_speed(iter/s)": 1.318903
+    },
+    {
+      "acc": 0.69405851,
+      "epoch": 0.5787671232876712,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.51344324151822e-06,
+      "loss": 1.34914913,
+      "memory(GiB)": 103.58,
+      "step": 22815,
+      "train_speed(iter/s)": 1.318935
+    },
+    {
+      "acc": 0.67759485,
+      "epoch": 0.5788939624556063,
+      "grad_norm": 3.5,
+      "learning_rate": 8.512697071150235e-06,
+      "loss": 1.38989258,
+      "memory(GiB)": 103.58,
+      "step": 22820,
+      "train_speed(iter/s)": 1.318965
+    },
+    {
+      "acc": 0.6804492,
+      "epoch": 0.5790208016235413,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.511950746278152e-06,
+      "loss": 1.29403906,
+      "memory(GiB)": 103.58,
+      "step": 22825,
+      "train_speed(iter/s)": 1.318998
+    },
+    {
+      "acc": 0.67978411,
+      "epoch": 0.5791476407914764,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.511204266934797e-06,
+      "loss": 1.41653996,
+      "memory(GiB)": 103.58,
+      "step": 22830,
+      "train_speed(iter/s)": 1.319029
+    },
+    {
+      "acc": 0.67136068,
+      "epoch": 0.5792744799594115,
+      "grad_norm": 3.125,
+      "learning_rate": 8.510457633152998e-06,
+      "loss": 1.48808441,
+      "memory(GiB)": 103.58,
+      "step": 22835,
+      "train_speed(iter/s)": 1.319061
+    },
+    {
+      "acc": 0.67881479,
+      "epoch": 0.5794013191273465,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.509710844965602e-06,
+      "loss": 1.41297483,
+      "memory(GiB)": 103.58,
+      "step": 22840,
+      "train_speed(iter/s)": 1.319092
+    },
+    {
+      "acc": 0.67582154,
+      "epoch": 0.5795281582952816,
+      "grad_norm": 4.5625,
+      "learning_rate": 8.508963902405451e-06,
+      "loss": 1.47056503,
+      "memory(GiB)": 103.58,
+      "step": 22845,
+      "train_speed(iter/s)": 1.319125
+    },
+    {
+      "acc": 0.68740954,
+      "epoch": 0.5796549974632167,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.508216805505403e-06,
+      "loss": 1.38177719,
+      "memory(GiB)": 103.58,
+      "step": 22850,
+      "train_speed(iter/s)": 1.319157
+    },
+    {
+      "acc": 0.6868916,
+      "epoch": 0.5797818366311517,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.507469554298318e-06,
+      "loss": 1.41726093,
+      "memory(GiB)": 103.58,
+      "step": 22855,
+      "train_speed(iter/s)": 1.319188
+    },
+    {
+      "acc": 0.69638119,
+      "epoch": 0.5799086757990868,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.506722148817061e-06,
+      "loss": 1.38258009,
+      "memory(GiB)": 103.58,
+      "step": 22860,
+      "train_speed(iter/s)": 1.31922
+    },
+    {
+      "acc": 0.69530811,
+      "epoch": 0.5800355149670218,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.505974589094505e-06,
+      "loss": 1.34813271,
+      "memory(GiB)": 103.58,
+      "step": 22865,
+      "train_speed(iter/s)": 1.319252
+    },
+    {
+      "acc": 0.69425626,
+      "epoch": 0.5801623541349569,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.505226875163537e-06,
+      "loss": 1.41670399,
+      "memory(GiB)": 103.58,
+      "step": 22870,
+      "train_speed(iter/s)": 1.319284
+    },
+    {
+      "acc": 0.69312105,
+      "epoch": 0.580289193302892,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.50447900705704e-06,
+      "loss": 1.37746668,
+      "memory(GiB)": 103.58,
+      "step": 22875,
+      "train_speed(iter/s)": 1.319317
+    },
+    {
+      "acc": 0.68998575,
+      "epoch": 0.580416032470827,
+      "grad_norm": 2.34375,
+      "learning_rate": 8.503730984807911e-06,
+      "loss": 1.38452663,
+      "memory(GiB)": 103.58,
+      "step": 22880,
+      "train_speed(iter/s)": 1.319348
+    },
+    {
+      "acc": 0.69863791,
+      "epoch": 0.5805428716387621,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.502982808449049e-06,
+      "loss": 1.33629856,
+      "memory(GiB)": 103.58,
+      "step": 22885,
+      "train_speed(iter/s)": 1.319382
+    },
+    {
+      "acc": 0.67753596,
+      "epoch": 0.5806697108066972,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.502234478013363e-06,
+      "loss": 1.46268158,
+      "memory(GiB)": 103.58,
+      "step": 22890,
+      "train_speed(iter/s)": 1.319414
+    },
+    {
+      "acc": 0.68122473,
+      "epoch": 0.5807965499746321,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.501485993533769e-06,
+      "loss": 1.43330164,
+      "memory(GiB)": 103.58,
+      "step": 22895,
+      "train_speed(iter/s)": 1.319446
+    },
+    {
+      "acc": 0.69275961,
+      "epoch": 0.5809233891425672,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.500737355043188e-06,
+      "loss": 1.41019382,
+      "memory(GiB)": 103.58,
+      "step": 22900,
+      "train_speed(iter/s)": 1.319478
+    },
+    {
+      "acc": 0.69568939,
+      "epoch": 0.5810502283105022,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.499988562574549e-06,
+      "loss": 1.39345188,
+      "memory(GiB)": 103.58,
+      "step": 22905,
+      "train_speed(iter/s)": 1.31951
+    },
+    {
+      "acc": 0.69414926,
+      "epoch": 0.5811770674784373,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.499239616160787e-06,
+      "loss": 1.43854599,
+      "memory(GiB)": 103.58,
+      "step": 22910,
+      "train_speed(iter/s)": 1.319544
+    },
+    {
+      "acc": 0.69149656,
+      "epoch": 0.5813039066463724,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.498490515834841e-06,
+      "loss": 1.4001421,
+      "memory(GiB)": 103.58,
+      "step": 22915,
+      "train_speed(iter/s)": 1.319578
+    },
+    {
+      "acc": 0.67875133,
+      "epoch": 0.5814307458143074,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.497741261629664e-06,
+      "loss": 1.41006069,
+      "memory(GiB)": 103.58,
+      "step": 22920,
+      "train_speed(iter/s)": 1.31961
+    },
+    {
+      "acc": 0.68473282,
+      "epoch": 0.5815575849822425,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.496991853578212e-06,
+      "loss": 1.3932023,
+      "memory(GiB)": 103.58,
+      "step": 22925,
+      "train_speed(iter/s)": 1.319643
+    },
+    {
+      "acc": 0.68383379,
+      "epoch": 0.5816844241501776,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.496242291713444e-06,
+      "loss": 1.40274363,
+      "memory(GiB)": 103.58,
+      "step": 22930,
+      "train_speed(iter/s)": 1.319676
+    },
+    {
+      "acc": 0.69678955,
+      "epoch": 0.5818112633181126,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.495492576068329e-06,
+      "loss": 1.38756275,
+      "memory(GiB)": 103.58,
+      "step": 22935,
+      "train_speed(iter/s)": 1.319707
+    },
+    {
+      "acc": 0.68553858,
+      "epoch": 0.5819381024860477,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.494742706675844e-06,
+      "loss": 1.43461142,
+      "memory(GiB)": 103.58,
+      "step": 22940,
+      "train_speed(iter/s)": 1.319738
+    },
+    {
+      "acc": 0.67584343,
+      "epoch": 0.5820649416539827,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.493992683568975e-06,
+      "loss": 1.47610035,
+      "memory(GiB)": 103.58,
+      "step": 22945,
+      "train_speed(iter/s)": 1.319769
+    },
+    {
+      "acc": 0.69120197,
+      "epoch": 0.5821917808219178,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.493242506780705e-06,
+      "loss": 1.39326305,
+      "memory(GiB)": 103.58,
+      "step": 22950,
+      "train_speed(iter/s)": 1.319799
+    },
+    {
+      "acc": 0.67452369,
+      "epoch": 0.5823186199898529,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.492492176344035e-06,
+      "loss": 1.47309628,
+      "memory(GiB)": 103.58,
+      "step": 22955,
+      "train_speed(iter/s)": 1.319829
+    },
+    {
+      "acc": 0.69235067,
+      "epoch": 0.5824454591577879,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.491741692291967e-06,
+      "loss": 1.37161427,
+      "memory(GiB)": 103.58,
+      "step": 22960,
+      "train_speed(iter/s)": 1.319859
+    },
+    {
+      "acc": 0.67026329,
+      "epoch": 0.582572298325723,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.490991054657507e-06,
+      "loss": 1.47554884,
+      "memory(GiB)": 103.58,
+      "step": 22965,
+      "train_speed(iter/s)": 1.319889
+    },
+    {
+      "acc": 0.6851923,
+      "epoch": 0.5826991374936581,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.490240263473677e-06,
+      "loss": 1.45192165,
+      "memory(GiB)": 103.58,
+      "step": 22970,
+      "train_speed(iter/s)": 1.319914
+    },
+    {
+      "acc": 0.68304224,
+      "epoch": 0.5828259766615931,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.489489318773496e-06,
+      "loss": 1.42613688,
+      "memory(GiB)": 103.58,
+      "step": 22975,
+      "train_speed(iter/s)": 1.319944
+    },
+    {
+      "acc": 0.67620769,
+      "epoch": 0.5829528158295282,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.488738220589996e-06,
+      "loss": 1.47837524,
+      "memory(GiB)": 103.58,
+      "step": 22980,
+      "train_speed(iter/s)": 1.319974
+    },
+    {
+      "acc": 0.6965755,
+      "epoch": 0.5830796549974632,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.487986968956212e-06,
+      "loss": 1.35159626,
+      "memory(GiB)": 103.58,
+      "step": 22985,
+      "train_speed(iter/s)": 1.320004
+    },
+    {
+      "acc": 0.69445763,
+      "epoch": 0.5832064941653983,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.487235563905191e-06,
+      "loss": 1.38485126,
+      "memory(GiB)": 103.58,
+      "step": 22990,
+      "train_speed(iter/s)": 1.320034
+    },
+    {
+      "acc": 0.68775425,
+      "epoch": 0.5833333333333334,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.486484005469977e-06,
+      "loss": 1.38442211,
+      "memory(GiB)": 103.58,
+      "step": 22995,
+      "train_speed(iter/s)": 1.320065
+    },
+    {
+      "acc": 0.68948126,
+      "epoch": 0.5834601725012684,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.485732293683633e-06,
+      "loss": 1.36803799,
+      "memory(GiB)": 103.58,
+      "step": 23000,
+      "train_speed(iter/s)": 1.320094
+    },
+    {
+      "epoch": 0.5834601725012684,
+      "eval_acc": 0.6738993060558789,
+      "eval_loss": 1.363337516784668,
+      "eval_runtime": 69.7744,
+      "eval_samples_per_second": 91.294,
+      "eval_steps_per_second": 22.831,
+      "step": 23000
+    },
+    {
+      "acc": 0.68242836,
+      "epoch": 0.5835870116692035,
+      "grad_norm": 3.625,
+      "learning_rate": 8.48498042857922e-06,
+      "loss": 1.4184906,
+      "memory(GiB)": 103.58,
+      "step": 23005,
+      "train_speed(iter/s)": 1.313924
+    },
+    {
+      "acc": 0.67455225,
+      "epoch": 0.5837138508371386,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.484228410189807e-06,
+      "loss": 1.41156311,
+      "memory(GiB)": 103.58,
+      "step": 23010,
+      "train_speed(iter/s)": 1.313955
+    },
+    {
+      "acc": 0.66721306,
+      "epoch": 0.5838406900050735,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.483476238548473e-06,
+      "loss": 1.44953737,
+      "memory(GiB)": 103.58,
+      "step": 23015,
+      "train_speed(iter/s)": 1.313987
+    },
+    {
+      "acc": 0.67747955,
+      "epoch": 0.5839675291730086,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.482723913688301e-06,
+      "loss": 1.43191471,
+      "memory(GiB)": 103.58,
+      "step": 23020,
+      "train_speed(iter/s)": 1.31402
+    },
+    {
+      "acc": 0.6929646,
+      "epoch": 0.5840943683409436,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.481971435642382e-06,
+      "loss": 1.37644987,
+      "memory(GiB)": 103.58,
+      "step": 23025,
+      "train_speed(iter/s)": 1.314051
+    },
+    {
+      "acc": 0.69515128,
+      "epoch": 0.5842212075088787,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.481218804443814e-06,
+      "loss": 1.35249901,
+      "memory(GiB)": 103.58,
+      "step": 23030,
+      "train_speed(iter/s)": 1.314083
+    },
+    {
+      "acc": 0.67634568,
+      "epoch": 0.5843480466768138,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.480466020125701e-06,
+      "loss": 1.41483545,
+      "memory(GiB)": 103.58,
+      "step": 23035,
+      "train_speed(iter/s)": 1.314116
+    },
+    {
+      "acc": 0.68641849,
+      "epoch": 0.5844748858447488,
+      "grad_norm": 2.625,
+      "learning_rate": 8.479713082721153e-06,
+      "loss": 1.44392509,
+      "memory(GiB)": 103.58,
+      "step": 23040,
+      "train_speed(iter/s)": 1.314147
+    },
+    {
+      "acc": 0.70150776,
+      "epoch": 0.5846017250126839,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.478959992263288e-06,
+      "loss": 1.33517075,
+      "memory(GiB)": 103.58,
+      "step": 23045,
+      "train_speed(iter/s)": 1.314179
+    },
+    {
+      "acc": 0.67653637,
+      "epoch": 0.584728564180619,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.478206748785229e-06,
+      "loss": 1.45533419,
+      "memory(GiB)": 103.58,
+      "step": 23050,
+      "train_speed(iter/s)": 1.314212
+    },
+    {
+      "acc": 0.68660774,
+      "epoch": 0.584855403348554,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.477453352320108e-06,
+      "loss": 1.35963755,
+      "memory(GiB)": 103.58,
+      "step": 23055,
+      "train_speed(iter/s)": 1.314244
+    },
+    {
+      "acc": 0.69018545,
+      "epoch": 0.5849822425164891,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.476699802901066e-06,
+      "loss": 1.38137903,
+      "memory(GiB)": 103.58,
+      "step": 23060,
+      "train_speed(iter/s)": 1.314277
+    },
+    {
+      "acc": 0.68628697,
+      "epoch": 0.5851090816844241,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.47594610056124e-06,
+      "loss": 1.37240658,
+      "memory(GiB)": 103.58,
+      "step": 23065,
+      "train_speed(iter/s)": 1.31431
+    },
+    {
+      "acc": 0.68608627,
+      "epoch": 0.5852359208523592,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.475192245333787e-06,
+      "loss": 1.38988514,
+      "memory(GiB)": 103.58,
+      "step": 23070,
+      "train_speed(iter/s)": 1.314343
+    },
+    {
+      "acc": 0.6925406,
+      "epoch": 0.5853627600202943,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.474438237251864e-06,
+      "loss": 1.3485754,
+      "memory(GiB)": 103.58,
+      "step": 23075,
+      "train_speed(iter/s)": 1.314374
+    },
+    {
+      "acc": 0.69463902,
+      "epoch": 0.5854895991882293,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.473684076348635e-06,
+      "loss": 1.35083284,
+      "memory(GiB)": 103.58,
+      "step": 23080,
+      "train_speed(iter/s)": 1.314407
+    },
+    {
+      "acc": 0.68704948,
+      "epoch": 0.5856164383561644,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.472929762657272e-06,
+      "loss": 1.3922452,
+      "memory(GiB)": 103.58,
+      "step": 23085,
+      "train_speed(iter/s)": 1.314439
+    },
+    {
+      "acc": 0.69852438,
+      "epoch": 0.5857432775240995,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.472175296210952e-06,
+      "loss": 1.37780371,
+      "memory(GiB)": 103.58,
+      "step": 23090,
+      "train_speed(iter/s)": 1.314472
+    },
+    {
+      "acc": 0.68478737,
+      "epoch": 0.5858701166920345,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.471420677042858e-06,
+      "loss": 1.41562538,
+      "memory(GiB)": 103.58,
+      "step": 23095,
+      "train_speed(iter/s)": 1.314504
+    },
+    {
+      "acc": 0.69154668,
+      "epoch": 0.5859969558599696,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.470665905186188e-06,
+      "loss": 1.34780664,
+      "memory(GiB)": 103.58,
+      "step": 23100,
+      "train_speed(iter/s)": 1.314537
+    },
+    {
+      "acc": 0.68291721,
+      "epoch": 0.5861237950279046,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.469910980674134e-06,
+      "loss": 1.44547215,
+      "memory(GiB)": 103.58,
+      "step": 23105,
+      "train_speed(iter/s)": 1.31457
+    },
+    {
+      "acc": 0.70865436,
+      "epoch": 0.5862506341958397,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.469155903539903e-06,
+      "loss": 1.27663517,
+      "memory(GiB)": 103.58,
+      "step": 23110,
+      "train_speed(iter/s)": 1.314602
+    },
+    {
+      "acc": 0.67836332,
+      "epoch": 0.5863774733637748,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.468400673816705e-06,
+      "loss": 1.42494831,
+      "memory(GiB)": 103.58,
+      "step": 23115,
+      "train_speed(iter/s)": 1.314634
+    },
+    {
+      "acc": 0.67745576,
+      "epoch": 0.5865043125317098,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.467645291537763e-06,
+      "loss": 1.47316246,
+      "memory(GiB)": 103.58,
+      "step": 23120,
+      "train_speed(iter/s)": 1.314667
+    },
+    {
+      "acc": 0.67648153,
+      "epoch": 0.5866311516996449,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.466889756736298e-06,
+      "loss": 1.37327442,
+      "memory(GiB)": 103.58,
+      "step": 23125,
+      "train_speed(iter/s)": 1.314701
+    },
+    {
+      "acc": 0.68216887,
+      "epoch": 0.58675799086758,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.466134069445544e-06,
+      "loss": 1.37402344,
+      "memory(GiB)": 103.58,
+      "step": 23130,
+      "train_speed(iter/s)": 1.314729
+    },
+    {
+      "acc": 0.67992945,
+      "epoch": 0.586884830035515,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.465378229698737e-06,
+      "loss": 1.42896385,
+      "memory(GiB)": 103.58,
+      "step": 23135,
+      "train_speed(iter/s)": 1.314763
+    },
+    {
+      "acc": 0.67275057,
+      "epoch": 0.58701166920345,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.464622237529123e-06,
+      "loss": 1.44980173,
+      "memory(GiB)": 103.58,
+      "step": 23140,
+      "train_speed(iter/s)": 1.314796
+    },
+    {
+      "acc": 0.69674835,
+      "epoch": 0.587138508371385,
+      "grad_norm": 3.25,
+      "learning_rate": 8.463866092969958e-06,
+      "loss": 1.34274349,
+      "memory(GiB)": 103.58,
+      "step": 23145,
+      "train_speed(iter/s)": 1.314829
+    },
+    {
+      "acc": 0.68815475,
+      "epoch": 0.5872653475393201,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.463109796054495e-06,
+      "loss": 1.41199303,
+      "memory(GiB)": 103.58,
+      "step": 23150,
+      "train_speed(iter/s)": 1.31486
+    },
+    {
+      "acc": 0.68578157,
+      "epoch": 0.5873921867072552,
+      "grad_norm": 3.875,
+      "learning_rate": 8.462353346815999e-06,
+      "loss": 1.39132833,
+      "memory(GiB)": 103.58,
+      "step": 23155,
+      "train_speed(iter/s)": 1.314893
+    },
+    {
+      "acc": 0.68732009,
+      "epoch": 0.5875190258751902,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.461596745287747e-06,
+      "loss": 1.38165674,
+      "memory(GiB)": 103.58,
+      "step": 23160,
+      "train_speed(iter/s)": 1.314926
+    },
+    {
+      "acc": 0.67224784,
+      "epoch": 0.5876458650431253,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.460839991503016e-06,
+      "loss": 1.46712627,
+      "memory(GiB)": 103.58,
+      "step": 23165,
+      "train_speed(iter/s)": 1.314959
+    },
+    {
+      "acc": 0.70358124,
+      "epoch": 0.5877727042110604,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.46008308549509e-06,
+      "loss": 1.33432407,
+      "memory(GiB)": 103.58,
+      "step": 23170,
+      "train_speed(iter/s)": 1.314993
+    },
+    {
+      "acc": 0.68530416,
+      "epoch": 0.5878995433789954,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.459326027297261e-06,
+      "loss": 1.3790226,
+      "memory(GiB)": 103.58,
+      "step": 23175,
+      "train_speed(iter/s)": 1.315027
+    },
+    {
+      "acc": 0.68518634,
+      "epoch": 0.5880263825469305,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.45856881694283e-06,
+      "loss": 1.3893177,
+      "memory(GiB)": 103.58,
+      "step": 23180,
+      "train_speed(iter/s)": 1.31506
+    },
+    {
+      "acc": 0.68349524,
+      "epoch": 0.5881532217148655,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.4578114544651e-06,
+      "loss": 1.36762266,
+      "memory(GiB)": 103.58,
+      "step": 23185,
+      "train_speed(iter/s)": 1.315092
+    },
+    {
+      "acc": 0.6910563,
+      "epoch": 0.5882800608828006,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.457053939897385e-06,
+      "loss": 1.39411364,
+      "memory(GiB)": 103.58,
+      "step": 23190,
+      "train_speed(iter/s)": 1.315127
+    },
+    {
+      "acc": 0.69307637,
+      "epoch": 0.5884069000507357,
+      "grad_norm": 2.875,
+      "learning_rate": 8.456296273273e-06,
+      "loss": 1.37019739,
+      "memory(GiB)": 103.58,
+      "step": 23195,
+      "train_speed(iter/s)": 1.315159
+    },
+    {
+      "acc": 0.68871617,
+      "epoch": 0.5885337392186707,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.455538454625276e-06,
+      "loss": 1.3874445,
+      "memory(GiB)": 103.58,
+      "step": 23200,
+      "train_speed(iter/s)": 1.315192
+    },
+    {
+      "acc": 0.68179836,
+      "epoch": 0.5886605783866058,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.454780483987544e-06,
+      "loss": 1.39822674,
+      "memory(GiB)": 103.58,
+      "step": 23205,
+      "train_speed(iter/s)": 1.315226
+    },
+    {
+      "acc": 0.69070687,
+      "epoch": 0.5887874175545409,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.45402236139314e-06,
+      "loss": 1.35748024,
+      "memory(GiB)": 103.58,
+      "step": 23210,
+      "train_speed(iter/s)": 1.31526
+    },
+    {
+      "acc": 0.68753147,
+      "epoch": 0.5889142567224759,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.453264086875411e-06,
+      "loss": 1.39953861,
+      "memory(GiB)": 103.58,
+      "step": 23215,
+      "train_speed(iter/s)": 1.315294
+    },
+    {
+      "acc": 0.68605976,
+      "epoch": 0.589041095890411,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.452505660467713e-06,
+      "loss": 1.3311636,
+      "memory(GiB)": 103.58,
+      "step": 23220,
+      "train_speed(iter/s)": 1.315327
+    },
+    {
+      "acc": 0.69111004,
+      "epoch": 0.589167935058346,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.451747082203398e-06,
+      "loss": 1.40163155,
+      "memory(GiB)": 103.58,
+      "step": 23225,
+      "train_speed(iter/s)": 1.31536
+    },
+    {
+      "acc": 0.69552116,
+      "epoch": 0.5892947742262811,
+      "grad_norm": 4.5625,
+      "learning_rate": 8.450988352115838e-06,
+      "loss": 1.34609804,
+      "memory(GiB)": 103.58,
+      "step": 23230,
+      "train_speed(iter/s)": 1.315382
+    },
+    {
+      "acc": 0.68558307,
+      "epoch": 0.5894216133942162,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.450229470238401e-06,
+      "loss": 1.39204664,
+      "memory(GiB)": 103.58,
+      "step": 23235,
+      "train_speed(iter/s)": 1.315415
+    },
+    {
+      "acc": 0.6821413,
+      "epoch": 0.5895484525621512,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.44947043660447e-06,
+      "loss": 1.39836311,
+      "memory(GiB)": 103.58,
+      "step": 23240,
+      "train_speed(iter/s)": 1.315447
+    },
+    {
+      "acc": 0.70045433,
+      "epoch": 0.5896752917300863,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.448711251247425e-06,
+      "loss": 1.33987799,
+      "memory(GiB)": 103.58,
+      "step": 23245,
+      "train_speed(iter/s)": 1.315482
+    },
+    {
+      "acc": 0.6830472,
+      "epoch": 0.5898021308980214,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.447951914200665e-06,
+      "loss": 1.41227684,
+      "memory(GiB)": 103.58,
+      "step": 23250,
+      "train_speed(iter/s)": 1.315516
+    },
+    {
+      "acc": 0.70218668,
+      "epoch": 0.5899289700659563,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.447192425497583e-06,
+      "loss": 1.35917559,
+      "memory(GiB)": 103.58,
+      "step": 23255,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.70102353,
+      "epoch": 0.5900558092338914,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.44643278517159e-06,
+      "loss": 1.34782581,
+      "memory(GiB)": 103.58,
+      "step": 23260,
+      "train_speed(iter/s)": 1.315582
+    },
+    {
+      "acc": 0.67610202,
+      "epoch": 0.5901826484018264,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.445672993256095e-06,
+      "loss": 1.49081411,
+      "memory(GiB)": 103.58,
+      "step": 23265,
+      "train_speed(iter/s)": 1.315615
+    },
+    {
+      "acc": 0.67402763,
+      "epoch": 0.5903094875697615,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.444913049784517e-06,
+      "loss": 1.45445175,
+      "memory(GiB)": 103.58,
+      "step": 23270,
+      "train_speed(iter/s)": 1.315648
+    },
+    {
+      "acc": 0.69420176,
+      "epoch": 0.5904363267376966,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.444152954790285e-06,
+      "loss": 1.38068781,
+      "memory(GiB)": 103.58,
+      "step": 23275,
+      "train_speed(iter/s)": 1.315682
+    },
+    {
+      "acc": 0.67985363,
+      "epoch": 0.5905631659056316,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.443392708306827e-06,
+      "loss": 1.41744404,
+      "memory(GiB)": 103.58,
+      "step": 23280,
+      "train_speed(iter/s)": 1.315713
+    },
+    {
+      "acc": 0.69665809,
+      "epoch": 0.5906900050735667,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.442632310367585e-06,
+      "loss": 1.35482292,
+      "memory(GiB)": 103.58,
+      "step": 23285,
+      "train_speed(iter/s)": 1.315748
+    },
+    {
+      "acc": 0.69368706,
+      "epoch": 0.5908168442415018,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.441871761006001e-06,
+      "loss": 1.34519024,
+      "memory(GiB)": 103.58,
+      "step": 23290,
+      "train_speed(iter/s)": 1.315781
+    },
+    {
+      "acc": 0.68080139,
+      "epoch": 0.5909436834094368,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.441111060255533e-06,
+      "loss": 1.43291883,
+      "memory(GiB)": 103.58,
+      "step": 23295,
+      "train_speed(iter/s)": 1.315814
+    },
+    {
+      "acc": 0.66781349,
+      "epoch": 0.5910705225773719,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.440350208149637e-06,
+      "loss": 1.48748894,
+      "memory(GiB)": 103.58,
+      "step": 23300,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.68794875,
+      "epoch": 0.5911973617453069,
+      "grad_norm": 4.25,
+      "learning_rate": 8.43958920472178e-06,
+      "loss": 1.36933498,
+      "memory(GiB)": 103.58,
+      "step": 23305,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.6902503,
+      "epoch": 0.591324200913242,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.43882805000543e-06,
+      "loss": 1.38134956,
+      "memory(GiB)": 103.58,
+      "step": 23310,
+      "train_speed(iter/s)": 1.315912
+    },
+    {
+      "acc": 0.67890878,
+      "epoch": 0.5914510400811771,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.43806674403407e-06,
+      "loss": 1.46557436,
+      "memory(GiB)": 103.58,
+      "step": 23315,
+      "train_speed(iter/s)": 1.315946
+    },
+    {
+      "acc": 0.69934416,
+      "epoch": 0.5915778792491121,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.437305286841187e-06,
+      "loss": 1.333389,
+      "memory(GiB)": 103.58,
+      "step": 23320,
+      "train_speed(iter/s)": 1.31598
+    },
+    {
+      "acc": 0.70157404,
+      "epoch": 0.5917047184170472,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.436543678460269e-06,
+      "loss": 1.37599525,
+      "memory(GiB)": 103.58,
+      "step": 23325,
+      "train_speed(iter/s)": 1.316013
+    },
+    {
+      "acc": 0.70046711,
+      "epoch": 0.5918315575849823,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.435781918924817e-06,
+      "loss": 1.35595627,
+      "memory(GiB)": 103.58,
+      "step": 23330,
+      "train_speed(iter/s)": 1.316046
+    },
+    {
+      "acc": 0.701017,
+      "epoch": 0.5919583967529173,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.435020008268335e-06,
+      "loss": 1.36324539,
+      "memory(GiB)": 103.58,
+      "step": 23335,
+      "train_speed(iter/s)": 1.316075
+    },
+    {
+      "acc": 0.68872781,
+      "epoch": 0.5920852359208524,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.43425794652434e-06,
+      "loss": 1.37889366,
+      "memory(GiB)": 103.58,
+      "step": 23340,
+      "train_speed(iter/s)": 1.316109
+    },
+    {
+      "acc": 0.67762384,
+      "epoch": 0.5922120750887874,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.433495733726345e-06,
+      "loss": 1.44428005,
+      "memory(GiB)": 103.58,
+      "step": 23345,
+      "train_speed(iter/s)": 1.316141
+    },
+    {
+      "acc": 0.67472329,
+      "epoch": 0.5923389142567225,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.43273336990788e-06,
+      "loss": 1.45045147,
+      "memory(GiB)": 103.58,
+      "step": 23350,
+      "train_speed(iter/s)": 1.316175
+    },
+    {
+      "acc": 0.69138222,
+      "epoch": 0.5924657534246576,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.431970855102475e-06,
+      "loss": 1.40197592,
+      "memory(GiB)": 103.58,
+      "step": 23355,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.68920155,
+      "epoch": 0.5925925925925926,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.43120818934367e-06,
+      "loss": 1.36545973,
+      "memory(GiB)": 103.58,
+      "step": 23360,
+      "train_speed(iter/s)": 1.316242
+    },
+    {
+      "acc": 0.68749223,
+      "epoch": 0.5927194317605277,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.430445372665008e-06,
+      "loss": 1.41887112,
+      "memory(GiB)": 103.58,
+      "step": 23365,
+      "train_speed(iter/s)": 1.316278
+    },
+    {
+      "acc": 0.68444538,
+      "epoch": 0.5928462709284628,
+      "grad_norm": 2.875,
+      "learning_rate": 8.429682405100042e-06,
+      "loss": 1.43141232,
+      "memory(GiB)": 103.58,
+      "step": 23370,
+      "train_speed(iter/s)": 1.316311
+    },
+    {
+      "acc": 0.70004334,
+      "epoch": 0.5929731100963977,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.428919286682333e-06,
+      "loss": 1.34130173,
+      "memory(GiB)": 103.58,
+      "step": 23375,
+      "train_speed(iter/s)": 1.316345
+    },
+    {
+      "acc": 0.69240918,
+      "epoch": 0.5930999492643328,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.428156017445443e-06,
+      "loss": 1.36211977,
+      "memory(GiB)": 103.58,
+      "step": 23380,
+      "train_speed(iter/s)": 1.316374
+    },
+    {
+      "acc": 0.69181681,
+      "epoch": 0.5932267884322678,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.427392597422947e-06,
+      "loss": 1.41372271,
+      "memory(GiB)": 103.58,
+      "step": 23385,
+      "train_speed(iter/s)": 1.316408
+    },
+    {
+      "acc": 0.6885911,
+      "epoch": 0.5933536276002029,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.426629026648423e-06,
+      "loss": 1.42282734,
+      "memory(GiB)": 103.58,
+      "step": 23390,
+      "train_speed(iter/s)": 1.316442
+    },
+    {
+      "acc": 0.68184214,
+      "epoch": 0.593480466768138,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.425865305155455e-06,
+      "loss": 1.43205318,
+      "memory(GiB)": 103.58,
+      "step": 23395,
+      "train_speed(iter/s)": 1.316475
+    },
+    {
+      "acc": 0.68801069,
+      "epoch": 0.593607305936073,
+      "grad_norm": 3.9375,
+      "learning_rate": 8.425101432977636e-06,
+      "loss": 1.42605877,
+      "memory(GiB)": 103.58,
+      "step": 23400,
+      "train_speed(iter/s)": 1.316509
+    },
+    {
+      "acc": 0.66681781,
+      "epoch": 0.5937341451040081,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.424337410148562e-06,
+      "loss": 1.4383358,
+      "memory(GiB)": 103.58,
+      "step": 23405,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.67694988,
+      "epoch": 0.5938609842719432,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.423573236701842e-06,
+      "loss": 1.43579369,
+      "memory(GiB)": 103.58,
+      "step": 23410,
+      "train_speed(iter/s)": 1.316577
+    },
+    {
+      "acc": 0.68256292,
+      "epoch": 0.5939878234398782,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.422808912671086e-06,
+      "loss": 1.4388113,
+      "memory(GiB)": 103.58,
+      "step": 23415,
+      "train_speed(iter/s)": 1.316611
+    },
+    {
+      "acc": 0.69445801,
+      "epoch": 0.5941146626078133,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.422044438089911e-06,
+      "loss": 1.41934462,
+      "memory(GiB)": 103.58,
+      "step": 23420,
+      "train_speed(iter/s)": 1.316642
+    },
+    {
+      "acc": 0.69548922,
+      "epoch": 0.5942415017757483,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.421279812991944e-06,
+      "loss": 1.36157131,
+      "memory(GiB)": 103.58,
+      "step": 23425,
+      "train_speed(iter/s)": 1.316675
+    },
+    {
+      "acc": 0.69099994,
+      "epoch": 0.5943683409436834,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.420515037410817e-06,
+      "loss": 1.370786,
+      "memory(GiB)": 103.58,
+      "step": 23430,
+      "train_speed(iter/s)": 1.31671
+    },
+    {
+      "acc": 0.67540998,
+      "epoch": 0.5944951801116185,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.419750111380166e-06,
+      "loss": 1.38823814,
+      "memory(GiB)": 103.58,
+      "step": 23435,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.68307037,
+      "epoch": 0.5946220192795535,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.418985034933637e-06,
+      "loss": 1.41272488,
+      "memory(GiB)": 103.58,
+      "step": 23440,
+      "train_speed(iter/s)": 1.316775
+    },
+    {
+      "acc": 0.6929522,
+      "epoch": 0.5947488584474886,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.418219808104882e-06,
+      "loss": 1.34912605,
+      "memory(GiB)": 103.58,
+      "step": 23445,
+      "train_speed(iter/s)": 1.316808
+    },
+    {
+      "acc": 0.6982399,
+      "epoch": 0.5948756976154237,
+      "grad_norm": 3.125,
+      "learning_rate": 8.417454430927559e-06,
+      "loss": 1.40504246,
+      "memory(GiB)": 103.58,
+      "step": 23450,
+      "train_speed(iter/s)": 1.316842
+    },
+    {
+      "acc": 0.68380003,
+      "epoch": 0.5950025367833587,
+      "grad_norm": 3.125,
+      "learning_rate": 8.41668890343533e-06,
+      "loss": 1.44835262,
+      "memory(GiB)": 103.58,
+      "step": 23455,
+      "train_speed(iter/s)": 1.316875
+    },
+    {
+      "acc": 0.67013588,
+      "epoch": 0.5951293759512938,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.41592322566187e-06,
+      "loss": 1.47240086,
+      "memory(GiB)": 103.58,
+      "step": 23460,
+      "train_speed(iter/s)": 1.316907
+    },
+    {
+      "acc": 0.68477001,
+      "epoch": 0.5952562151192288,
+      "grad_norm": 3.25,
+      "learning_rate": 8.415157397640857e-06,
+      "loss": 1.33525133,
+      "memory(GiB)": 103.58,
+      "step": 23465,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.67771769,
+      "epoch": 0.5953830542871639,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.414391419405972e-06,
+      "loss": 1.39277,
+      "memory(GiB)": 103.58,
+      "step": 23470,
+      "train_speed(iter/s)": 1.316969
+    },
+    {
+      "acc": 0.6848896,
+      "epoch": 0.595509893455099,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.413625290990909e-06,
+      "loss": 1.43754158,
+      "memory(GiB)": 103.58,
+      "step": 23475,
+      "train_speed(iter/s)": 1.317003
+    },
+    {
+      "acc": 0.70354314,
+      "epoch": 0.595636732623034,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.412859012429365e-06,
+      "loss": 1.32153492,
+      "memory(GiB)": 103.58,
+      "step": 23480,
+      "train_speed(iter/s)": 1.317035
+    },
+    {
+      "acc": 0.69404197,
+      "epoch": 0.5957635717909691,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.412092583755043e-06,
+      "loss": 1.35435963,
+      "memory(GiB)": 103.58,
+      "step": 23485,
+      "train_speed(iter/s)": 1.317067
+    },
+    {
+      "acc": 0.68186302,
+      "epoch": 0.5958904109589042,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.411326005001658e-06,
+      "loss": 1.404035,
+      "memory(GiB)": 103.58,
+      "step": 23490,
+      "train_speed(iter/s)": 1.3171
+    },
+    {
+      "acc": 0.67219906,
+      "epoch": 0.5960172501268391,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.410559276202922e-06,
+      "loss": 1.42042227,
+      "memory(GiB)": 103.58,
+      "step": 23495,
+      "train_speed(iter/s)": 1.317133
+    },
+    {
+      "acc": 0.6753818,
+      "epoch": 0.5961440892947742,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.409792397392565e-06,
+      "loss": 1.39980888,
+      "memory(GiB)": 103.58,
+      "step": 23500,
+      "train_speed(iter/s)": 1.317165
+    },
+    {
+      "acc": 0.68309774,
+      "epoch": 0.5962709284627092,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.40902536860431e-06,
+      "loss": 1.37821703,
+      "memory(GiB)": 103.58,
+      "step": 23505,
+      "train_speed(iter/s)": 1.317198
+    },
+    {
+      "acc": 0.6890192,
+      "epoch": 0.5963977676306443,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.408258189871904e-06,
+      "loss": 1.37746077,
+      "memory(GiB)": 103.58,
+      "step": 23510,
+      "train_speed(iter/s)": 1.317231
+    },
+    {
+      "acc": 0.69260225,
+      "epoch": 0.5965246067985794,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.407490861229084e-06,
+      "loss": 1.41943483,
+      "memory(GiB)": 103.58,
+      "step": 23515,
+      "train_speed(iter/s)": 1.317263
+    },
+    {
+      "acc": 0.67725172,
+      "epoch": 0.5966514459665144,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.406723382709603e-06,
+      "loss": 1.45175972,
+      "memory(GiB)": 103.58,
+      "step": 23520,
+      "train_speed(iter/s)": 1.317296
+    },
+    {
+      "acc": 0.69180088,
+      "epoch": 0.5967782851344495,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.405955754347216e-06,
+      "loss": 1.41398258,
+      "memory(GiB)": 103.58,
+      "step": 23525,
+      "train_speed(iter/s)": 1.317328
+    },
+    {
+      "acc": 0.6853817,
+      "epoch": 0.5969051243023846,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.40518797617569e-06,
+      "loss": 1.40352268,
+      "memory(GiB)": 103.58,
+      "step": 23530,
+      "train_speed(iter/s)": 1.317359
+    },
+    {
+      "acc": 0.70942078,
+      "epoch": 0.5970319634703196,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.404420048228794e-06,
+      "loss": 1.27399445,
+      "memory(GiB)": 103.58,
+      "step": 23535,
+      "train_speed(iter/s)": 1.317388
+    },
+    {
+      "acc": 0.67742491,
+      "epoch": 0.5971588026382547,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.403651970540305e-06,
+      "loss": 1.44449615,
+      "memory(GiB)": 103.58,
+      "step": 23540,
+      "train_speed(iter/s)": 1.31742
+    },
+    {
+      "acc": 0.68155408,
+      "epoch": 0.5972856418061897,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.402883743144005e-06,
+      "loss": 1.39199314,
+      "memory(GiB)": 103.58,
+      "step": 23545,
+      "train_speed(iter/s)": 1.317453
+    },
+    {
+      "acc": 0.67982607,
+      "epoch": 0.5974124809741248,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.402115366073686e-06,
+      "loss": 1.44691811,
+      "memory(GiB)": 103.58,
+      "step": 23550,
+      "train_speed(iter/s)": 1.317485
+    },
+    {
+      "acc": 0.68473582,
+      "epoch": 0.5975393201420599,
+      "grad_norm": 3.0,
+      "learning_rate": 8.401346839363143e-06,
+      "loss": 1.39073839,
+      "memory(GiB)": 103.58,
+      "step": 23555,
+      "train_speed(iter/s)": 1.317515
+    },
+    {
+      "acc": 0.68168921,
+      "epoch": 0.5976661593099949,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.40057816304618e-06,
+      "loss": 1.42406387,
+      "memory(GiB)": 103.58,
+      "step": 23560,
+      "train_speed(iter/s)": 1.317548
+    },
+    {
+      "acc": 0.68141623,
+      "epoch": 0.59779299847793,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.399809337156608e-06,
+      "loss": 1.46495838,
+      "memory(GiB)": 103.58,
+      "step": 23565,
+      "train_speed(iter/s)": 1.317581
+    },
+    {
+      "acc": 0.68476405,
+      "epoch": 0.5979198376458651,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.39904036172824e-06,
+      "loss": 1.41861382,
+      "memory(GiB)": 103.58,
+      "step": 23570,
+      "train_speed(iter/s)": 1.317614
+    },
+    {
+      "acc": 0.68725462,
+      "epoch": 0.5980466768138001,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.398271236794904e-06,
+      "loss": 1.35901995,
+      "memory(GiB)": 103.58,
+      "step": 23575,
+      "train_speed(iter/s)": 1.317646
+    },
+    {
+      "acc": 0.6806149,
+      "epoch": 0.5981735159817352,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.397501962390427e-06,
+      "loss": 1.4044219,
+      "memory(GiB)": 103.58,
+      "step": 23580,
+      "train_speed(iter/s)": 1.317678
+    },
+    {
+      "acc": 0.67185984,
+      "epoch": 0.5983003551496702,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.396732538548642e-06,
+      "loss": 1.43479328,
+      "memory(GiB)": 103.58,
+      "step": 23585,
+      "train_speed(iter/s)": 1.317709
+    },
+    {
+      "acc": 0.69037251,
+      "epoch": 0.5984271943176053,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.395962965303397e-06,
+      "loss": 1.44913702,
+      "memory(GiB)": 103.58,
+      "step": 23590,
+      "train_speed(iter/s)": 1.317738
+    },
+    {
+      "acc": 0.71142673,
+      "epoch": 0.5985540334855404,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.395193242688537e-06,
+      "loss": 1.36079235,
+      "memory(GiB)": 103.58,
+      "step": 23595,
+      "train_speed(iter/s)": 1.317769
+    },
+    {
+      "acc": 0.68914785,
+      "epoch": 0.5986808726534754,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.394423370737922e-06,
+      "loss": 1.30906305,
+      "memory(GiB)": 103.58,
+      "step": 23600,
+      "train_speed(iter/s)": 1.317801
+    },
+    {
+      "acc": 0.6889472,
+      "epoch": 0.5988077118214105,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.393653349485412e-06,
+      "loss": 1.36403732,
+      "memory(GiB)": 103.58,
+      "step": 23605,
+      "train_speed(iter/s)": 1.317834
+    },
+    {
+      "acc": 0.68269968,
+      "epoch": 0.5989345509893456,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.392883178964874e-06,
+      "loss": 1.43294086,
+      "memory(GiB)": 103.58,
+      "step": 23610,
+      "train_speed(iter/s)": 1.317864
+    },
+    {
+      "acc": 0.68880215,
+      "epoch": 0.5990613901572805,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.392112859210186e-06,
+      "loss": 1.37450094,
+      "memory(GiB)": 103.58,
+      "step": 23615,
+      "train_speed(iter/s)": 1.317897
+    },
+    {
+      "acc": 0.68477359,
+      "epoch": 0.5991882293252156,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.391342390255232e-06,
+      "loss": 1.42443218,
+      "memory(GiB)": 103.58,
+      "step": 23620,
+      "train_speed(iter/s)": 1.317929
+    },
+    {
+      "acc": 0.68438139,
+      "epoch": 0.5993150684931506,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.390571772133896e-06,
+      "loss": 1.42452412,
+      "memory(GiB)": 103.58,
+      "step": 23625,
+      "train_speed(iter/s)": 1.317961
+    },
+    {
+      "acc": 0.69199967,
+      "epoch": 0.5994419076610857,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.389801004880077e-06,
+      "loss": 1.4274353,
+      "memory(GiB)": 103.58,
+      "step": 23630,
+      "train_speed(iter/s)": 1.317993
+    },
+    {
+      "acc": 0.68041158,
+      "epoch": 0.5995687468290208,
+      "grad_norm": 3.0,
+      "learning_rate": 8.389030088527675e-06,
+      "loss": 1.39876461,
+      "memory(GiB)": 103.58,
+      "step": 23635,
+      "train_speed(iter/s)": 1.318025
+    },
+    {
+      "acc": 0.67278214,
+      "epoch": 0.5996955859969558,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.388259023110598e-06,
+      "loss": 1.42567444,
+      "memory(GiB)": 103.58,
+      "step": 23640,
+      "train_speed(iter/s)": 1.318056
+    },
+    {
+      "acc": 0.68671522,
+      "epoch": 0.5998224251648909,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.387487808662765e-06,
+      "loss": 1.41083984,
+      "memory(GiB)": 103.58,
+      "step": 23645,
+      "train_speed(iter/s)": 1.318088
+    },
+    {
+      "acc": 0.66724968,
+      "epoch": 0.599949264332826,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.38671644521809e-06,
+      "loss": 1.46559954,
+      "memory(GiB)": 103.58,
+      "step": 23650,
+      "train_speed(iter/s)": 1.31812
+    },
+    {
+      "acc": 0.68602753,
+      "epoch": 0.600076103500761,
+      "grad_norm": 2.875,
+      "learning_rate": 8.385944932810508e-06,
+      "loss": 1.41260004,
+      "memory(GiB)": 103.58,
+      "step": 23655,
+      "train_speed(iter/s)": 1.318152
+    },
+    {
+      "acc": 0.70026093,
+      "epoch": 0.6002029426686961,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.385173271473948e-06,
+      "loss": 1.36612053,
+      "memory(GiB)": 103.58,
+      "step": 23660,
+      "train_speed(iter/s)": 1.318183
+    },
+    {
+      "acc": 0.68996882,
+      "epoch": 0.6003297818366311,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.384401461242355e-06,
+      "loss": 1.42563686,
+      "memory(GiB)": 103.58,
+      "step": 23665,
+      "train_speed(iter/s)": 1.318216
+    },
+    {
+      "acc": 0.69024711,
+      "epoch": 0.6004566210045662,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.383629502149678e-06,
+      "loss": 1.38892317,
+      "memory(GiB)": 103.58,
+      "step": 23670,
+      "train_speed(iter/s)": 1.318248
+    },
+    {
+      "acc": 0.67932825,
+      "epoch": 0.6005834601725013,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.382857394229865e-06,
+      "loss": 1.40682449,
+      "memory(GiB)": 103.58,
+      "step": 23675,
+      "train_speed(iter/s)": 1.31828
+    },
+    {
+      "acc": 0.68339996,
+      "epoch": 0.6007102993404363,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.382085137516883e-06,
+      "loss": 1.43141937,
+      "memory(GiB)": 103.58,
+      "step": 23680,
+      "train_speed(iter/s)": 1.31831
+    },
+    {
+      "acc": 0.69578314,
+      "epoch": 0.6008371385083714,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.381312732044696e-06,
+      "loss": 1.40374432,
+      "memory(GiB)": 103.58,
+      "step": 23685,
+      "train_speed(iter/s)": 1.318342
+    },
+    {
+      "acc": 0.69889908,
+      "epoch": 0.6009639776763065,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.380540177847278e-06,
+      "loss": 1.35964527,
+      "memory(GiB)": 103.58,
+      "step": 23690,
+      "train_speed(iter/s)": 1.318374
+    },
+    {
+      "acc": 0.66827354,
+      "epoch": 0.6010908168442415,
+      "grad_norm": 3.125,
+      "learning_rate": 8.37976747495861e-06,
+      "loss": 1.46422176,
+      "memory(GiB)": 103.58,
+      "step": 23695,
+      "train_speed(iter/s)": 1.318407
+    },
+    {
+      "acc": 0.69331865,
+      "epoch": 0.6012176560121766,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.378994623412679e-06,
+      "loss": 1.32003841,
+      "memory(GiB)": 103.58,
+      "step": 23700,
+      "train_speed(iter/s)": 1.31844
+    },
+    {
+      "acc": 0.67517776,
+      "epoch": 0.6013444951801116,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.378221623243478e-06,
+      "loss": 1.37735405,
+      "memory(GiB)": 103.58,
+      "step": 23705,
+      "train_speed(iter/s)": 1.318473
+    },
+    {
+      "acc": 0.68609066,
+      "epoch": 0.6014713343480467,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.377448474485008e-06,
+      "loss": 1.34251442,
+      "memory(GiB)": 103.58,
+      "step": 23710,
+      "train_speed(iter/s)": 1.318505
+    },
+    {
+      "acc": 0.69057117,
+      "epoch": 0.6015981735159818,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.376675177171273e-06,
+      "loss": 1.3836132,
+      "memory(GiB)": 103.58,
+      "step": 23715,
+      "train_speed(iter/s)": 1.318538
+    },
+    {
+      "acc": 0.69397779,
+      "epoch": 0.6017250126839168,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.375901731336292e-06,
+      "loss": 1.35784035,
+      "memory(GiB)": 103.58,
+      "step": 23720,
+      "train_speed(iter/s)": 1.31857
+    },
+    {
+      "acc": 0.68565593,
+      "epoch": 0.6018518518518519,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.375128137014076e-06,
+      "loss": 1.38385849,
+      "memory(GiB)": 103.58,
+      "step": 23725,
+      "train_speed(iter/s)": 1.318602
+    },
+    {
+      "acc": 0.67943497,
+      "epoch": 0.601978691019787,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.374354394238658e-06,
+      "loss": 1.48674526,
+      "memory(GiB)": 103.58,
+      "step": 23730,
+      "train_speed(iter/s)": 1.318634
+    },
+    {
+      "acc": 0.68780413,
+      "epoch": 0.602105530187722,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.373580503044068e-06,
+      "loss": 1.37582684,
+      "memory(GiB)": 103.58,
+      "step": 23735,
+      "train_speed(iter/s)": 1.318665
+    },
+    {
+      "acc": 0.69015532,
+      "epoch": 0.602232369355657,
+      "grad_norm": 3.875,
+      "learning_rate": 8.372806463464347e-06,
+      "loss": 1.41166239,
+      "memory(GiB)": 103.58,
+      "step": 23740,
+      "train_speed(iter/s)": 1.318696
+    },
+    {
+      "acc": 0.69522095,
+      "epoch": 0.602359208523592,
+      "grad_norm": 3.125,
+      "learning_rate": 8.372032275533538e-06,
+      "loss": 1.3154726,
+      "memory(GiB)": 103.58,
+      "step": 23745,
+      "train_speed(iter/s)": 1.318728
+    },
+    {
+      "acc": 0.69044895,
+      "epoch": 0.6024860476915271,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.371257939285692e-06,
+      "loss": 1.41762714,
+      "memory(GiB)": 103.58,
+      "step": 23750,
+      "train_speed(iter/s)": 1.31876
+    },
+    {
+      "acc": 0.70050254,
+      "epoch": 0.6026128868594622,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.370483454754873e-06,
+      "loss": 1.39515018,
+      "memory(GiB)": 103.58,
+      "step": 23755,
+      "train_speed(iter/s)": 1.318789
+    },
+    {
+      "acc": 0.69300966,
+      "epoch": 0.6027397260273972,
+      "grad_norm": 2.875,
+      "learning_rate": 8.369708821975144e-06,
+      "loss": 1.36728249,
+      "memory(GiB)": 103.58,
+      "step": 23760,
+      "train_speed(iter/s)": 1.318821
+    },
+    {
+      "acc": 0.70562654,
+      "epoch": 0.6028665651953323,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.368934040980576e-06,
+      "loss": 1.31398506,
+      "memory(GiB)": 103.58,
+      "step": 23765,
+      "train_speed(iter/s)": 1.31885
+    },
+    {
+      "acc": 0.69163499,
+      "epoch": 0.6029934043632674,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.368159111805246e-06,
+      "loss": 1.45920353,
+      "memory(GiB)": 103.58,
+      "step": 23770,
+      "train_speed(iter/s)": 1.318881
+    },
+    {
+      "acc": 0.70076962,
+      "epoch": 0.6031202435312024,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.367384034483242e-06,
+      "loss": 1.36780548,
+      "memory(GiB)": 103.58,
+      "step": 23775,
+      "train_speed(iter/s)": 1.318914
+    },
+    {
+      "acc": 0.68713589,
+      "epoch": 0.6032470826991375,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.366608809048653e-06,
+      "loss": 1.44078503,
+      "memory(GiB)": 103.58,
+      "step": 23780,
+      "train_speed(iter/s)": 1.318945
+    },
+    {
+      "acc": 0.6844924,
+      "epoch": 0.6033739218670725,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.365833435535579e-06,
+      "loss": 1.39962759,
+      "memory(GiB)": 103.58,
+      "step": 23785,
+      "train_speed(iter/s)": 1.318978
+    },
+    {
+      "acc": 0.69268699,
+      "epoch": 0.6035007610350076,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.365057913978123e-06,
+      "loss": 1.4522625,
+      "memory(GiB)": 103.58,
+      "step": 23790,
+      "train_speed(iter/s)": 1.319009
+    },
+    {
+      "acc": 0.67419138,
+      "epoch": 0.6036276002029427,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.364282244410394e-06,
+      "loss": 1.46921883,
+      "memory(GiB)": 103.58,
+      "step": 23795,
+      "train_speed(iter/s)": 1.31904
+    },
+    {
+      "acc": 0.68587503,
+      "epoch": 0.6037544393708777,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.363506426866513e-06,
+      "loss": 1.39669333,
+      "memory(GiB)": 103.58,
+      "step": 23800,
+      "train_speed(iter/s)": 1.319071
+    },
+    {
+      "acc": 0.68444986,
+      "epoch": 0.6038812785388128,
+      "grad_norm": 2.75,
+      "learning_rate": 8.362730461380602e-06,
+      "loss": 1.38332195,
+      "memory(GiB)": 103.58,
+      "step": 23805,
+      "train_speed(iter/s)": 1.319103
+    },
+    {
+      "acc": 0.67817287,
+      "epoch": 0.6040081177067479,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.361954347986793e-06,
+      "loss": 1.39745178,
+      "memory(GiB)": 103.58,
+      "step": 23810,
+      "train_speed(iter/s)": 1.319134
+    },
+    {
+      "acc": 0.70008349,
+      "epoch": 0.6041349568746829,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.36117808671922e-06,
+      "loss": 1.30916729,
+      "memory(GiB)": 103.58,
+      "step": 23815,
+      "train_speed(iter/s)": 1.319166
+    },
+    {
+      "acc": 0.68613935,
+      "epoch": 0.604261796042618,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.36040167761203e-06,
+      "loss": 1.38754644,
+      "memory(GiB)": 103.58,
+      "step": 23820,
+      "train_speed(iter/s)": 1.319198
+    },
+    {
+      "acc": 0.68477907,
+      "epoch": 0.604388635210553,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.359625120699368e-06,
+      "loss": 1.40618267,
+      "memory(GiB)": 103.58,
+      "step": 23825,
+      "train_speed(iter/s)": 1.31923
+    },
+    {
+      "acc": 0.68886614,
+      "epoch": 0.6045154743784881,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.358848416015397e-06,
+      "loss": 1.36116924,
+      "memory(GiB)": 103.58,
+      "step": 23830,
+      "train_speed(iter/s)": 1.319261
+    },
+    {
+      "acc": 0.6714283,
+      "epoch": 0.6046423135464232,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.358071563594274e-06,
+      "loss": 1.44969139,
+      "memory(GiB)": 103.58,
+      "step": 23835,
+      "train_speed(iter/s)": 1.319293
+    },
+    {
+      "acc": 0.67546368,
+      "epoch": 0.6047691527143582,
+      "grad_norm": 3.0,
+      "learning_rate": 8.357294563470173e-06,
+      "loss": 1.40987549,
+      "memory(GiB)": 103.58,
+      "step": 23840,
+      "train_speed(iter/s)": 1.319326
+    },
+    {
+      "acc": 0.69847431,
+      "epoch": 0.6048959918822933,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.356517415677267e-06,
+      "loss": 1.33207932,
+      "memory(GiB)": 103.58,
+      "step": 23845,
+      "train_speed(iter/s)": 1.319357
+    },
+    {
+      "acc": 0.66207247,
+      "epoch": 0.6050228310502284,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.355740120249739e-06,
+      "loss": 1.46110268,
+      "memory(GiB)": 103.58,
+      "step": 23850,
+      "train_speed(iter/s)": 1.31939
+    },
+    {
+      "acc": 0.67622719,
+      "epoch": 0.6051496702181633,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.354962677221779e-06,
+      "loss": 1.40835104,
+      "memory(GiB)": 103.58,
+      "step": 23855,
+      "train_speed(iter/s)": 1.319423
+    },
+    {
+      "acc": 0.67959037,
+      "epoch": 0.6052765093860984,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.35418508662758e-06,
+      "loss": 1.37493362,
+      "memory(GiB)": 103.58,
+      "step": 23860,
+      "train_speed(iter/s)": 1.319456
+    },
+    {
+      "acc": 0.68635988,
+      "epoch": 0.6054033485540334,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.353407348501346e-06,
+      "loss": 1.430336,
+      "memory(GiB)": 103.58,
+      "step": 23865,
+      "train_speed(iter/s)": 1.319488
+    },
+    {
+      "acc": 0.68827081,
+      "epoch": 0.6055301877219685,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.352629462877286e-06,
+      "loss": 1.3874774,
+      "memory(GiB)": 103.58,
+      "step": 23870,
+      "train_speed(iter/s)": 1.319521
+    },
+    {
+      "acc": 0.6777297,
+      "epoch": 0.6056570268899036,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.351851429789613e-06,
+      "loss": 1.44862432,
+      "memory(GiB)": 103.58,
+      "step": 23875,
+      "train_speed(iter/s)": 1.319552
+    },
+    {
+      "acc": 0.66783953,
+      "epoch": 0.6057838660578386,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.35107324927255e-06,
+      "loss": 1.41206131,
+      "memory(GiB)": 103.58,
+      "step": 23880,
+      "train_speed(iter/s)": 1.319586
+    },
+    {
+      "acc": 0.68155231,
+      "epoch": 0.6059107052257737,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.350294921360323e-06,
+      "loss": 1.47051849,
+      "memory(GiB)": 103.58,
+      "step": 23885,
+      "train_speed(iter/s)": 1.319618
+    },
+    {
+      "acc": 0.67561054,
+      "epoch": 0.6060375443937088,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.349516446087168e-06,
+      "loss": 1.44209061,
+      "memory(GiB)": 103.58,
+      "step": 23890,
+      "train_speed(iter/s)": 1.319648
+    },
+    {
+      "acc": 0.66936483,
+      "epoch": 0.6061643835616438,
+      "grad_norm": 3.125,
+      "learning_rate": 8.348737823487325e-06,
+      "loss": 1.46133633,
+      "memory(GiB)": 103.58,
+      "step": 23895,
+      "train_speed(iter/s)": 1.31968
+    },
+    {
+      "acc": 0.68512034,
+      "epoch": 0.6062912227295789,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.347959053595042e-06,
+      "loss": 1.44385023,
+      "memory(GiB)": 103.58,
+      "step": 23900,
+      "train_speed(iter/s)": 1.31971
+    },
+    {
+      "acc": 0.67464352,
+      "epoch": 0.6064180618975139,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.347180136444572e-06,
+      "loss": 1.47862806,
+      "memory(GiB)": 103.58,
+      "step": 23905,
+      "train_speed(iter/s)": 1.319743
+    },
+    {
+      "acc": 0.69064894,
+      "epoch": 0.606544901065449,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.346401072070174e-06,
+      "loss": 1.39700613,
+      "memory(GiB)": 103.58,
+      "step": 23910,
+      "train_speed(iter/s)": 1.319775
+    },
+    {
+      "acc": 0.66710243,
+      "epoch": 0.6066717402333841,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.345621860506119e-06,
+      "loss": 1.46272745,
+      "memory(GiB)": 103.58,
+      "step": 23915,
+      "train_speed(iter/s)": 1.319807
+    },
+    {
+      "acc": 0.69358225,
+      "epoch": 0.6067985794013191,
+      "grad_norm": 3.25,
+      "learning_rate": 8.344842501786675e-06,
+      "loss": 1.32059155,
+      "memory(GiB)": 103.58,
+      "step": 23920,
+      "train_speed(iter/s)": 1.319839
+    },
+    {
+      "acc": 0.67481174,
+      "epoch": 0.6069254185692542,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.344062995946125e-06,
+      "loss": 1.46306448,
+      "memory(GiB)": 103.58,
+      "step": 23925,
+      "train_speed(iter/s)": 1.319871
+    },
+    {
+      "acc": 0.67994165,
+      "epoch": 0.6070522577371893,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.343283343018755e-06,
+      "loss": 1.39525995,
+      "memory(GiB)": 103.58,
+      "step": 23930,
+      "train_speed(iter/s)": 1.319898
+    },
+    {
+      "acc": 0.6828536,
+      "epoch": 0.6071790969051243,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.342503543038855e-06,
+      "loss": 1.41891279,
+      "memory(GiB)": 103.58,
+      "step": 23935,
+      "train_speed(iter/s)": 1.319925
+    },
+    {
+      "acc": 0.68095779,
+      "epoch": 0.6073059360730594,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.341723596040728e-06,
+      "loss": 1.47513647,
+      "memory(GiB)": 103.58,
+      "step": 23940,
+      "train_speed(iter/s)": 1.319956
+    },
+    {
+      "acc": 0.66612334,
+      "epoch": 0.6074327752409944,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.340943502058675e-06,
+      "loss": 1.47766457,
+      "memory(GiB)": 103.58,
+      "step": 23945,
+      "train_speed(iter/s)": 1.319988
+    },
+    {
+      "acc": 0.68486433,
+      "epoch": 0.6075596144089295,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.340163261127014e-06,
+      "loss": 1.43138618,
+      "memory(GiB)": 103.58,
+      "step": 23950,
+      "train_speed(iter/s)": 1.320021
+    },
+    {
+      "acc": 0.67171955,
+      "epoch": 0.6076864535768646,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.339382873280058e-06,
+      "loss": 1.45177889,
+      "memory(GiB)": 103.58,
+      "step": 23955,
+      "train_speed(iter/s)": 1.320053
+    },
+    {
+      "acc": 0.68113575,
+      "epoch": 0.6078132927447996,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.338602338552136e-06,
+      "loss": 1.43455505,
+      "memory(GiB)": 113.67,
+      "step": 23960,
+      "train_speed(iter/s)": 1.320076
+    },
+    {
+      "acc": 0.68180661,
+      "epoch": 0.6079401319127347,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.337821656977574e-06,
+      "loss": 1.38971062,
+      "memory(GiB)": 113.67,
+      "step": 23965,
+      "train_speed(iter/s)": 1.320106
+    },
+    {
+      "acc": 0.6960681,
+      "epoch": 0.6080669710806698,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.337040828590715e-06,
+      "loss": 1.38726444,
+      "memory(GiB)": 113.67,
+      "step": 23970,
+      "train_speed(iter/s)": 1.320138
+    },
+    {
+      "acc": 0.67581015,
+      "epoch": 0.6081938102486047,
+      "grad_norm": 3.375,
+      "learning_rate": 8.336259853425901e-06,
+      "loss": 1.41289005,
+      "memory(GiB)": 113.67,
+      "step": 23975,
+      "train_speed(iter/s)": 1.320169
+    },
+    {
+      "acc": 0.67696805,
+      "epoch": 0.6083206494165398,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.335478731517484e-06,
+      "loss": 1.39666185,
+      "memory(GiB)": 113.67,
+      "step": 23980,
+      "train_speed(iter/s)": 1.3202
+    },
+    {
+      "acc": 0.67044692,
+      "epoch": 0.6084474885844748,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.33469746289982e-06,
+      "loss": 1.44842539,
+      "memory(GiB)": 113.67,
+      "step": 23985,
+      "train_speed(iter/s)": 1.320225
+    },
+    {
+      "acc": 0.67307601,
+      "epoch": 0.6085743277524099,
+      "grad_norm": 3.25,
+      "learning_rate": 8.333916047607274e-06,
+      "loss": 1.42955599,
+      "memory(GiB)": 113.67,
+      "step": 23990,
+      "train_speed(iter/s)": 1.320258
+    },
+    {
+      "acc": 0.69759474,
+      "epoch": 0.608701166920345,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.333134485674214e-06,
+      "loss": 1.39822273,
+      "memory(GiB)": 113.67,
+      "step": 23995,
+      "train_speed(iter/s)": 1.320289
+    },
+    {
+      "acc": 0.67286906,
+      "epoch": 0.60882800608828,
+      "grad_norm": 3.0,
+      "learning_rate": 8.33235277713502e-06,
+      "loss": 1.4224555,
+      "memory(GiB)": 113.67,
+      "step": 24000,
+      "train_speed(iter/s)": 1.320321
+    },
+    {
+      "epoch": 0.60882800608828,
+      "eval_acc": 0.6739744850759204,
+      "eval_loss": 1.3625943660736084,
+      "eval_runtime": 69.8993,
+      "eval_samples_per_second": 91.131,
+      "eval_steps_per_second": 22.79,
+      "step": 24000
+    },
+    {
+      "acc": 0.69345646,
+      "epoch": 0.6089548452562151,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.33157092202407e-06,
+      "loss": 1.48116112,
+      "memory(GiB)": 113.67,
+      "step": 24005,
+      "train_speed(iter/s)": 1.314387
+    },
+    {
+      "acc": 0.68804383,
+      "epoch": 0.6090816844241502,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.33078892037576e-06,
+      "loss": 1.41402464,
+      "memory(GiB)": 113.67,
+      "step": 24010,
+      "train_speed(iter/s)": 1.314411
+    },
+    {
+      "acc": 0.68394146,
+      "epoch": 0.6092085235920852,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.33000677222448e-06,
+      "loss": 1.40689535,
+      "memory(GiB)": 113.67,
+      "step": 24015,
+      "train_speed(iter/s)": 1.314442
+    },
+    {
+      "acc": 0.68705463,
+      "epoch": 0.6093353627600203,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.329224477604635e-06,
+      "loss": 1.4103714,
+      "memory(GiB)": 113.67,
+      "step": 24020,
+      "train_speed(iter/s)": 1.314473
+    },
+    {
+      "acc": 0.67911282,
+      "epoch": 0.6094622019279553,
+      "grad_norm": 3.125,
+      "learning_rate": 8.328442036550633e-06,
+      "loss": 1.42971706,
+      "memory(GiB)": 113.67,
+      "step": 24025,
+      "train_speed(iter/s)": 1.314504
+    },
+    {
+      "acc": 0.68903561,
+      "epoch": 0.6095890410958904,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.327659449096892e-06,
+      "loss": 1.34558697,
+      "memory(GiB)": 113.67,
+      "step": 24030,
+      "train_speed(iter/s)": 1.314536
+    },
+    {
+      "acc": 0.70607729,
+      "epoch": 0.6097158802638255,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.32687671527783e-06,
+      "loss": 1.28343439,
+      "memory(GiB)": 113.67,
+      "step": 24035,
+      "train_speed(iter/s)": 1.314566
+    },
+    {
+      "acc": 0.68156843,
+      "epoch": 0.6098427194317605,
+      "grad_norm": 3.0,
+      "learning_rate": 8.326093835127878e-06,
+      "loss": 1.44165001,
+      "memory(GiB)": 113.67,
+      "step": 24040,
+      "train_speed(iter/s)": 1.314597
+    },
+    {
+      "acc": 0.67696352,
+      "epoch": 0.6099695585996956,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.325310808681466e-06,
+      "loss": 1.40975304,
+      "memory(GiB)": 113.67,
+      "step": 24045,
+      "train_speed(iter/s)": 1.314628
+    },
+    {
+      "acc": 0.68946671,
+      "epoch": 0.6100963977676307,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.32452763597304e-06,
+      "loss": 1.3224514,
+      "memory(GiB)": 113.67,
+      "step": 24050,
+      "train_speed(iter/s)": 1.314658
+    },
+    {
+      "acc": 0.68002539,
+      "epoch": 0.6102232369355657,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.323744317037048e-06,
+      "loss": 1.36028852,
+      "memory(GiB)": 113.67,
+      "step": 24055,
+      "train_speed(iter/s)": 1.314689
+    },
+    {
+      "acc": 0.67777214,
+      "epoch": 0.6103500761035008,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.322960851907937e-06,
+      "loss": 1.40723209,
+      "memory(GiB)": 113.67,
+      "step": 24060,
+      "train_speed(iter/s)": 1.31472
+    },
+    {
+      "acc": 0.66645894,
+      "epoch": 0.6104769152714358,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.322177240620175e-06,
+      "loss": 1.43500662,
+      "memory(GiB)": 113.67,
+      "step": 24065,
+      "train_speed(iter/s)": 1.314751
+    },
+    {
+      "acc": 0.6940124,
+      "epoch": 0.6106037544393709,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.321393483208224e-06,
+      "loss": 1.36561604,
+      "memory(GiB)": 113.67,
+      "step": 24070,
+      "train_speed(iter/s)": 1.314781
+    },
+    {
+      "acc": 0.69299407,
+      "epoch": 0.610730593607306,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.32060957970656e-06,
+      "loss": 1.3800539,
+      "memory(GiB)": 113.67,
+      "step": 24075,
+      "train_speed(iter/s)": 1.314812
+    },
+    {
+      "acc": 0.67874818,
+      "epoch": 0.610857432775241,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.319825530149661e-06,
+      "loss": 1.47326136,
+      "memory(GiB)": 113.67,
+      "step": 24080,
+      "train_speed(iter/s)": 1.314843
+    },
+    {
+      "acc": 0.67771788,
+      "epoch": 0.6109842719431761,
+      "grad_norm": 3.125,
+      "learning_rate": 8.319041334572012e-06,
+      "loss": 1.42370186,
+      "memory(GiB)": 113.67,
+      "step": 24085,
+      "train_speed(iter/s)": 1.314874
+    },
+    {
+      "acc": 0.68437281,
+      "epoch": 0.6111111111111112,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.318256993008108e-06,
+      "loss": 1.44338322,
+      "memory(GiB)": 113.67,
+      "step": 24090,
+      "train_speed(iter/s)": 1.314907
+    },
+    {
+      "acc": 0.67389216,
+      "epoch": 0.6112379502790461,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.317472505492446e-06,
+      "loss": 1.50183659,
+      "memory(GiB)": 113.67,
+      "step": 24095,
+      "train_speed(iter/s)": 1.314939
+    },
+    {
+      "acc": 0.69787135,
+      "epoch": 0.6113647894469812,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.31668787205953e-06,
+      "loss": 1.28734531,
+      "memory(GiB)": 113.67,
+      "step": 24100,
+      "train_speed(iter/s)": 1.314971
+    },
+    {
+      "acc": 0.69592428,
+      "epoch": 0.6114916286149162,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.315903092743876e-06,
+      "loss": 1.40484657,
+      "memory(GiB)": 113.67,
+      "step": 24105,
+      "train_speed(iter/s)": 1.315002
+    },
+    {
+      "acc": 0.69588933,
+      "epoch": 0.6116184677828513,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.315118167579999e-06,
+      "loss": 1.37396364,
+      "memory(GiB)": 113.67,
+      "step": 24110,
+      "train_speed(iter/s)": 1.315033
+    },
+    {
+      "acc": 0.69450169,
+      "epoch": 0.6117453069507864,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.314333096602423e-06,
+      "loss": 1.35060253,
+      "memory(GiB)": 113.67,
+      "step": 24115,
+      "train_speed(iter/s)": 1.315063
+    },
+    {
+      "acc": 0.6913651,
+      "epoch": 0.6118721461187214,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.313547879845682e-06,
+      "loss": 1.41917725,
+      "memory(GiB)": 113.67,
+      "step": 24120,
+      "train_speed(iter/s)": 1.315095
+    },
+    {
+      "acc": 0.67235231,
+      "epoch": 0.6119989852866565,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.312762517344308e-06,
+      "loss": 1.42473154,
+      "memory(GiB)": 113.67,
+      "step": 24125,
+      "train_speed(iter/s)": 1.315127
+    },
+    {
+      "acc": 0.6830174,
+      "epoch": 0.6121258244545916,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.311977009132851e-06,
+      "loss": 1.42532501,
+      "memory(GiB)": 113.67,
+      "step": 24130,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.68869658,
+      "epoch": 0.6122526636225266,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.311191355245858e-06,
+      "loss": 1.42708397,
+      "memory(GiB)": 113.67,
+      "step": 24135,
+      "train_speed(iter/s)": 1.315188
+    },
+    {
+      "acc": 0.68694496,
+      "epoch": 0.6123795027904617,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.310405555717884e-06,
+      "loss": 1.4063715,
+      "memory(GiB)": 113.67,
+      "step": 24140,
+      "train_speed(iter/s)": 1.315208
+    },
+    {
+      "acc": 0.68015184,
+      "epoch": 0.6125063419583967,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.309619610583495e-06,
+      "loss": 1.44179811,
+      "memory(GiB)": 113.67,
+      "step": 24145,
+      "train_speed(iter/s)": 1.315237
+    },
+    {
+      "acc": 0.71020432,
+      "epoch": 0.6126331811263318,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.30883351987726e-06,
+      "loss": 1.37945862,
+      "memory(GiB)": 113.67,
+      "step": 24150,
+      "train_speed(iter/s)": 1.315269
+    },
+    {
+      "acc": 0.68502407,
+      "epoch": 0.6127600202942669,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.30804728363375e-06,
+      "loss": 1.42034445,
+      "memory(GiB)": 113.67,
+      "step": 24155,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.69709849,
+      "epoch": 0.6128868594622019,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.307260901887556e-06,
+      "loss": 1.3843152,
+      "memory(GiB)": 113.67,
+      "step": 24160,
+      "train_speed(iter/s)": 1.315331
+    },
+    {
+      "acc": 0.69065356,
+      "epoch": 0.613013698630137,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.306474374673259e-06,
+      "loss": 1.32427492,
+      "memory(GiB)": 113.67,
+      "step": 24165,
+      "train_speed(iter/s)": 1.315362
+    },
+    {
+      "acc": 0.69328885,
+      "epoch": 0.6131405377980721,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.305687702025457e-06,
+      "loss": 1.36357174,
+      "memory(GiB)": 113.67,
+      "step": 24170,
+      "train_speed(iter/s)": 1.315394
+    },
+    {
+      "acc": 0.68299417,
+      "epoch": 0.6132673769660071,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.304900883978753e-06,
+      "loss": 1.37929716,
+      "memory(GiB)": 113.67,
+      "step": 24175,
+      "train_speed(iter/s)": 1.315425
+    },
+    {
+      "acc": 0.67260571,
+      "epoch": 0.6133942161339422,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.304113920567751e-06,
+      "loss": 1.45536308,
+      "memory(GiB)": 113.67,
+      "step": 24180,
+      "train_speed(iter/s)": 1.315456
+    },
+    {
+      "acc": 0.69024229,
+      "epoch": 0.6135210553018772,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.303326811827066e-06,
+      "loss": 1.38985147,
+      "memory(GiB)": 113.67,
+      "step": 24185,
+      "train_speed(iter/s)": 1.315487
+    },
+    {
+      "acc": 0.68872147,
+      "epoch": 0.6136478944698123,
+      "grad_norm": 2.875,
+      "learning_rate": 8.302539557791322e-06,
+      "loss": 1.41754618,
+      "memory(GiB)": 113.67,
+      "step": 24190,
+      "train_speed(iter/s)": 1.315519
+    },
+    {
+      "acc": 0.66962352,
+      "epoch": 0.6137747336377474,
+      "grad_norm": 2.75,
+      "learning_rate": 8.301752158495141e-06,
+      "loss": 1.42971535,
+      "memory(GiB)": 113.67,
+      "step": 24195,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.68577776,
+      "epoch": 0.6139015728056824,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.300964613973159e-06,
+      "loss": 1.44519882,
+      "memory(GiB)": 113.67,
+      "step": 24200,
+      "train_speed(iter/s)": 1.315579
+    },
+    {
+      "acc": 0.67864447,
+      "epoch": 0.6140284119736175,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.300176924260017e-06,
+      "loss": 1.38796663,
+      "memory(GiB)": 113.67,
+      "step": 24205,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.68489323,
+      "epoch": 0.6141552511415526,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.299389089390359e-06,
+      "loss": 1.39707222,
+      "memory(GiB)": 113.67,
+      "step": 24210,
+      "train_speed(iter/s)": 1.315638
+    },
+    {
+      "acc": 0.69319139,
+      "epoch": 0.6142820903094875,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.298601109398838e-06,
+      "loss": 1.3639679,
+      "memory(GiB)": 113.67,
+      "step": 24215,
+      "train_speed(iter/s)": 1.315669
+    },
+    {
+      "acc": 0.67987938,
+      "epoch": 0.6144089294774226,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.297812984320113e-06,
+      "loss": 1.42700806,
+      "memory(GiB)": 113.67,
+      "step": 24220,
+      "train_speed(iter/s)": 1.3157
+    },
+    {
+      "acc": 0.68710461,
+      "epoch": 0.6145357686453576,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.297024714188851e-06,
+      "loss": 1.3856185,
+      "memory(GiB)": 113.67,
+      "step": 24225,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.69245858,
+      "epoch": 0.6146626078132927,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.296236299039719e-06,
+      "loss": 1.43932562,
+      "memory(GiB)": 113.67,
+      "step": 24230,
+      "train_speed(iter/s)": 1.315756
+    },
+    {
+      "acc": 0.6841794,
+      "epoch": 0.6147894469812278,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.295447738907401e-06,
+      "loss": 1.34915171,
+      "memory(GiB)": 113.67,
+      "step": 24235,
+      "train_speed(iter/s)": 1.315787
+    },
+    {
+      "acc": 0.68615689,
+      "epoch": 0.6149162861491628,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.294659033826576e-06,
+      "loss": 1.42161217,
+      "memory(GiB)": 113.67,
+      "step": 24240,
+      "train_speed(iter/s)": 1.315819
+    },
+    {
+      "acc": 0.67778883,
+      "epoch": 0.6150431253170979,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.293870183831937e-06,
+      "loss": 1.42819576,
+      "memory(GiB)": 113.67,
+      "step": 24245,
+      "train_speed(iter/s)": 1.31585
+    },
+    {
+      "acc": 0.68927689,
+      "epoch": 0.615169964485033,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.293081188958183e-06,
+      "loss": 1.36966248,
+      "memory(GiB)": 113.67,
+      "step": 24250,
+      "train_speed(iter/s)": 1.315881
+    },
+    {
+      "acc": 0.69449029,
+      "epoch": 0.615296803652968,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.292292049240014e-06,
+      "loss": 1.3729249,
+      "memory(GiB)": 113.67,
+      "step": 24255,
+      "train_speed(iter/s)": 1.315911
+    },
+    {
+      "acc": 0.67435913,
+      "epoch": 0.6154236428209031,
+      "grad_norm": 3.0,
+      "learning_rate": 8.291502764712143e-06,
+      "loss": 1.45805502,
+      "memory(GiB)": 113.67,
+      "step": 24260,
+      "train_speed(iter/s)": 1.315942
+    },
+    {
+      "acc": 0.68931122,
+      "epoch": 0.6155504819888381,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.290713335409284e-06,
+      "loss": 1.38724728,
+      "memory(GiB)": 113.67,
+      "step": 24265,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.68408198,
+      "epoch": 0.6156773211567732,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.28992376136616e-06,
+      "loss": 1.37752733,
+      "memory(GiB)": 113.67,
+      "step": 24270,
+      "train_speed(iter/s)": 1.316002
+    },
+    {
+      "acc": 0.67503004,
+      "epoch": 0.6158041603247083,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.289134042617502e-06,
+      "loss": 1.45954151,
+      "memory(GiB)": 113.67,
+      "step": 24275,
+      "train_speed(iter/s)": 1.316032
+    },
+    {
+      "acc": 0.69108367,
+      "epoch": 0.6159309994926433,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.288344179198043e-06,
+      "loss": 1.35782204,
+      "memory(GiB)": 113.67,
+      "step": 24280,
+      "train_speed(iter/s)": 1.316063
+    },
+    {
+      "acc": 0.68716316,
+      "epoch": 0.6160578386605784,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.287554171142525e-06,
+      "loss": 1.41895323,
+      "memory(GiB)": 113.67,
+      "step": 24285,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.66752291,
+      "epoch": 0.6161846778285135,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.2867640184857e-06,
+      "loss": 1.46589203,
+      "memory(GiB)": 113.67,
+      "step": 24290,
+      "train_speed(iter/s)": 1.316125
+    },
+    {
+      "acc": 0.66402254,
+      "epoch": 0.6163115169964485,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.285973721262315e-06,
+      "loss": 1.45004368,
+      "memory(GiB)": 113.67,
+      "step": 24295,
+      "train_speed(iter/s)": 1.316155
+    },
+    {
+      "acc": 0.68560438,
+      "epoch": 0.6164383561643836,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.285183279507135e-06,
+      "loss": 1.39416571,
+      "memory(GiB)": 113.67,
+      "step": 24300,
+      "train_speed(iter/s)": 1.316186
+    },
+    {
+      "acc": 0.68271441,
+      "epoch": 0.6165651953323186,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.28439269325493e-06,
+      "loss": 1.38560238,
+      "memory(GiB)": 113.67,
+      "step": 24305,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.6793623,
+      "epoch": 0.6166920345002537,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.28360196254047e-06,
+      "loss": 1.45788689,
+      "memory(GiB)": 113.67,
+      "step": 24310,
+      "train_speed(iter/s)": 1.316245
+    },
+    {
+      "acc": 0.68186131,
+      "epoch": 0.6168188736681888,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.282811087398535e-06,
+      "loss": 1.40909128,
+      "memory(GiB)": 113.67,
+      "step": 24315,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.68008399,
+      "epoch": 0.6169457128361238,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.282020067863911e-06,
+      "loss": 1.41665888,
+      "memory(GiB)": 113.67,
+      "step": 24320,
+      "train_speed(iter/s)": 1.316307
+    },
+    {
+      "acc": 0.69982872,
+      "epoch": 0.6170725520040589,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.281228903971391e-06,
+      "loss": 1.35025749,
+      "memory(GiB)": 113.67,
+      "step": 24325,
+      "train_speed(iter/s)": 1.316337
+    },
+    {
+      "acc": 0.69369583,
+      "epoch": 0.617199391171994,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.280437595755774e-06,
+      "loss": 1.32799902,
+      "memory(GiB)": 113.67,
+      "step": 24330,
+      "train_speed(iter/s)": 1.316368
+    },
+    {
+      "acc": 0.68414955,
+      "epoch": 0.617326230339929,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.279646143251867e-06,
+      "loss": 1.31631193,
+      "memory(GiB)": 113.67,
+      "step": 24335,
+      "train_speed(iter/s)": 1.3164
+    },
+    {
+      "acc": 0.6978075,
+      "epoch": 0.617453069507864,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.278854546494479e-06,
+      "loss": 1.37739582,
+      "memory(GiB)": 113.67,
+      "step": 24340,
+      "train_speed(iter/s)": 1.316432
+    },
+    {
+      "acc": 0.68521929,
+      "epoch": 0.617579908675799,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.27806280551843e-06,
+      "loss": 1.33921356,
+      "memory(GiB)": 113.67,
+      "step": 24345,
+      "train_speed(iter/s)": 1.316464
+    },
+    {
+      "acc": 0.67403908,
+      "epoch": 0.6177067478437341,
+      "grad_norm": 3.125,
+      "learning_rate": 8.277270920358542e-06,
+      "loss": 1.40096684,
+      "memory(GiB)": 113.67,
+      "step": 24350,
+      "train_speed(iter/s)": 1.316494
+    },
+    {
+      "acc": 0.70389757,
+      "epoch": 0.6178335870116692,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.276478891049649e-06,
+      "loss": 1.37951603,
+      "memory(GiB)": 113.67,
+      "step": 24355,
+      "train_speed(iter/s)": 1.316526
+    },
+    {
+      "acc": 0.69299479,
+      "epoch": 0.6179604261796042,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.275686717626584e-06,
+      "loss": 1.3660429,
+      "memory(GiB)": 113.67,
+      "step": 24360,
+      "train_speed(iter/s)": 1.316556
+    },
+    {
+      "acc": 0.68615942,
+      "epoch": 0.6180872653475393,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.274894400124191e-06,
+      "loss": 1.38316126,
+      "memory(GiB)": 113.67,
+      "step": 24365,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.68149605,
+      "epoch": 0.6182141045154744,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.274101938577324e-06,
+      "loss": 1.43119221,
+      "memory(GiB)": 113.67,
+      "step": 24370,
+      "train_speed(iter/s)": 1.316615
+    },
+    {
+      "acc": 0.68716106,
+      "epoch": 0.6183409436834094,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.273309333020834e-06,
+      "loss": 1.34379835,
+      "memory(GiB)": 113.67,
+      "step": 24375,
+      "train_speed(iter/s)": 1.316644
+    },
+    {
+      "acc": 0.68401923,
+      "epoch": 0.6184677828513445,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.272516583489587e-06,
+      "loss": 1.44261742,
+      "memory(GiB)": 113.67,
+      "step": 24380,
+      "train_speed(iter/s)": 1.316674
+    },
+    {
+      "acc": 0.69873729,
+      "epoch": 0.6185946220192795,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.271723690018448e-06,
+      "loss": 1.34753075,
+      "memory(GiB)": 113.67,
+      "step": 24385,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.66646929,
+      "epoch": 0.6187214611872146,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.270930652642295e-06,
+      "loss": 1.42822075,
+      "memory(GiB)": 113.67,
+      "step": 24390,
+      "train_speed(iter/s)": 1.316735
+    },
+    {
+      "acc": 0.67857037,
+      "epoch": 0.6188483003551497,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.270137471396007e-06,
+      "loss": 1.42023792,
+      "memory(GiB)": 113.67,
+      "step": 24395,
+      "train_speed(iter/s)": 1.316766
+    },
+    {
+      "acc": 0.68715272,
+      "epoch": 0.6189751395230847,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.269344146314475e-06,
+      "loss": 1.40260048,
+      "memory(GiB)": 113.67,
+      "step": 24400,
+      "train_speed(iter/s)": 1.316796
+    },
+    {
+      "acc": 0.68633585,
+      "epoch": 0.6191019786910198,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.26855067743259e-06,
+      "loss": 1.46895742,
+      "memory(GiB)": 113.67,
+      "step": 24405,
+      "train_speed(iter/s)": 1.316825
+    },
+    {
+      "acc": 0.68971376,
+      "epoch": 0.6192288178589549,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.267757064785254e-06,
+      "loss": 1.38258724,
+      "memory(GiB)": 113.67,
+      "step": 24410,
+      "train_speed(iter/s)": 1.316855
+    },
+    {
+      "acc": 0.68405056,
+      "epoch": 0.6193556570268899,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.26696330840737e-06,
+      "loss": 1.40967979,
+      "memory(GiB)": 113.67,
+      "step": 24415,
+      "train_speed(iter/s)": 1.316886
+    },
+    {
+      "acc": 0.70530977,
+      "epoch": 0.619482496194825,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.266169408333856e-06,
+      "loss": 1.36876488,
+      "memory(GiB)": 113.67,
+      "step": 24420,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.69265699,
+      "epoch": 0.61960933536276,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.265375364599629e-06,
+      "loss": 1.39393997,
+      "memory(GiB)": 113.67,
+      "step": 24425,
+      "train_speed(iter/s)": 1.316947
+    },
+    {
+      "acc": 0.6989295,
+      "epoch": 0.6197361745306951,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.264581177239615e-06,
+      "loss": 1.38851318,
+      "memory(GiB)": 113.67,
+      "step": 24430,
+      "train_speed(iter/s)": 1.316976
+    },
+    {
+      "acc": 0.70307274,
+      "epoch": 0.6198630136986302,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.263786846288745e-06,
+      "loss": 1.32401323,
+      "memory(GiB)": 113.67,
+      "step": 24435,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.69068151,
+      "epoch": 0.6199898528665652,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.262992371781956e-06,
+      "loss": 1.3821538,
+      "memory(GiB)": 113.67,
+      "step": 24440,
+      "train_speed(iter/s)": 1.317038
+    },
+    {
+      "acc": 0.66219168,
+      "epoch": 0.6201166920345003,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.262197753754195e-06,
+      "loss": 1.47929058,
+      "memory(GiB)": 113.67,
+      "step": 24445,
+      "train_speed(iter/s)": 1.317069
+    },
+    {
+      "acc": 0.67920427,
+      "epoch": 0.6202435312024354,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.261402992240414e-06,
+      "loss": 1.44845762,
+      "memory(GiB)": 113.67,
+      "step": 24450,
+      "train_speed(iter/s)": 1.317101
+    },
+    {
+      "acc": 0.68044605,
+      "epoch": 0.6203703703703703,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.260608087275566e-06,
+      "loss": 1.44861946,
+      "memory(GiB)": 113.67,
+      "step": 24455,
+      "train_speed(iter/s)": 1.317132
+    },
+    {
+      "acc": 0.68511167,
+      "epoch": 0.6204972095383054,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.259813038894617e-06,
+      "loss": 1.40342188,
+      "memory(GiB)": 113.67,
+      "step": 24460,
+      "train_speed(iter/s)": 1.317162
+    },
+    {
+      "acc": 0.68348317,
+      "epoch": 0.6206240487062404,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.259017847132538e-06,
+      "loss": 1.36506481,
+      "memory(GiB)": 113.67,
+      "step": 24465,
+      "train_speed(iter/s)": 1.317191
+    },
+    {
+      "acc": 0.68136234,
+      "epoch": 0.6207508878741755,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.258222512024303e-06,
+      "loss": 1.44610624,
+      "memory(GiB)": 113.67,
+      "step": 24470,
+      "train_speed(iter/s)": 1.317222
+    },
+    {
+      "acc": 0.68962202,
+      "epoch": 0.6208777270421106,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.257427033604894e-06,
+      "loss": 1.33757563,
+      "memory(GiB)": 113.67,
+      "step": 24475,
+      "train_speed(iter/s)": 1.317253
+    },
+    {
+      "acc": 0.67183352,
+      "epoch": 0.6210045662100456,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.256631411909305e-06,
+      "loss": 1.44376144,
+      "memory(GiB)": 113.67,
+      "step": 24480,
+      "train_speed(iter/s)": 1.317283
+    },
+    {
+      "acc": 0.68905473,
+      "epoch": 0.6211314053779807,
+      "grad_norm": 3.25,
+      "learning_rate": 8.25583564697252e-06,
+      "loss": 1.33492374,
+      "memory(GiB)": 113.67,
+      "step": 24485,
+      "train_speed(iter/s)": 1.317314
+    },
+    {
+      "acc": 0.68939772,
+      "epoch": 0.6212582445459158,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.255039738829552e-06,
+      "loss": 1.40299568,
+      "memory(GiB)": 113.67,
+      "step": 24490,
+      "train_speed(iter/s)": 1.317344
+    },
+    {
+      "acc": 0.68539934,
+      "epoch": 0.6213850837138508,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.254243687515402e-06,
+      "loss": 1.39288616,
+      "memory(GiB)": 113.67,
+      "step": 24495,
+      "train_speed(iter/s)": 1.317374
+    },
+    {
+      "acc": 0.68243957,
+      "epoch": 0.6215119228817859,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.253447493065085e-06,
+      "loss": 1.39730949,
+      "memory(GiB)": 113.67,
+      "step": 24500,
+      "train_speed(iter/s)": 1.317404
+    },
+    {
+      "acc": 0.67582998,
+      "epoch": 0.6216387620497209,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.252651155513622e-06,
+      "loss": 1.42115259,
+      "memory(GiB)": 113.67,
+      "step": 24505,
+      "train_speed(iter/s)": 1.317434
+    },
+    {
+      "acc": 0.66696396,
+      "epoch": 0.621765601217656,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.251854674896039e-06,
+      "loss": 1.49542713,
+      "memory(GiB)": 113.67,
+      "step": 24510,
+      "train_speed(iter/s)": 1.317463
+    },
+    {
+      "acc": 0.6801734,
+      "epoch": 0.6218924403855911,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.251058051247368e-06,
+      "loss": 1.40630054,
+      "memory(GiB)": 113.67,
+      "step": 24515,
+      "train_speed(iter/s)": 1.317493
+    },
+    {
+      "acc": 0.6936451,
+      "epoch": 0.6220192795535261,
+      "grad_norm": 3.0,
+      "learning_rate": 8.250261284602651e-06,
+      "loss": 1.37871857,
+      "memory(GiB)": 113.67,
+      "step": 24520,
+      "train_speed(iter/s)": 1.317523
+    },
+    {
+      "acc": 0.69242091,
+      "epoch": 0.6221461187214612,
+      "grad_norm": 3.125,
+      "learning_rate": 8.249464374996932e-06,
+      "loss": 1.40150995,
+      "memory(GiB)": 113.67,
+      "step": 24525,
+      "train_speed(iter/s)": 1.317554
+    },
+    {
+      "acc": 0.68239179,
+      "epoch": 0.6222729578893963,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.24866732246526e-06,
+      "loss": 1.46351776,
+      "memory(GiB)": 113.67,
+      "step": 24530,
+      "train_speed(iter/s)": 1.317585
+    },
+    {
+      "acc": 0.6855372,
+      "epoch": 0.6223997970573313,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.247870127042695e-06,
+      "loss": 1.40218496,
+      "memory(GiB)": 113.67,
+      "step": 24535,
+      "train_speed(iter/s)": 1.317615
+    },
+    {
+      "acc": 0.680902,
+      "epoch": 0.6225266362252664,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.247072788764302e-06,
+      "loss": 1.39858999,
+      "memory(GiB)": 113.67,
+      "step": 24540,
+      "train_speed(iter/s)": 1.317643
+    },
+    {
+      "acc": 0.6705019,
+      "epoch": 0.6226534753932014,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.246275307665147e-06,
+      "loss": 1.45922155,
+      "memory(GiB)": 113.67,
+      "step": 24545,
+      "train_speed(iter/s)": 1.317674
+    },
+    {
+      "acc": 0.69337578,
+      "epoch": 0.6227803145611365,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.245477683780316e-06,
+      "loss": 1.33086081,
+      "memory(GiB)": 113.67,
+      "step": 24550,
+      "train_speed(iter/s)": 1.317702
+    },
+    {
+      "acc": 0.6837944,
+      "epoch": 0.6229071537290716,
+      "grad_norm": 4.0,
+      "learning_rate": 8.244679917144883e-06,
+      "loss": 1.40087872,
+      "memory(GiB)": 113.67,
+      "step": 24555,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "acc": 0.6733181,
+      "epoch": 0.6230339928970066,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.243882007793941e-06,
+      "loss": 1.47463903,
+      "memory(GiB)": 113.67,
+      "step": 24560,
+      "train_speed(iter/s)": 1.31776
+    },
+    {
+      "acc": 0.68631878,
+      "epoch": 0.6231608320649417,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.243083955762588e-06,
+      "loss": 1.40241642,
+      "memory(GiB)": 113.67,
+      "step": 24565,
+      "train_speed(iter/s)": 1.31779
+    },
+    {
+      "acc": 0.68959069,
+      "epoch": 0.6232876712328768,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.24228576108592e-06,
+      "loss": 1.39105434,
+      "memory(GiB)": 113.67,
+      "step": 24570,
+      "train_speed(iter/s)": 1.31782
+    },
+    {
+      "acc": 0.68554287,
+      "epoch": 0.6234145104008117,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.24148742379905e-06,
+      "loss": 1.34545345,
+      "memory(GiB)": 113.67,
+      "step": 24575,
+      "train_speed(iter/s)": 1.317849
+    },
+    {
+      "acc": 0.68885565,
+      "epoch": 0.6235413495687468,
+      "grad_norm": 3.125,
+      "learning_rate": 8.240688943937092e-06,
+      "loss": 1.38751678,
+      "memory(GiB)": 113.67,
+      "step": 24580,
+      "train_speed(iter/s)": 1.317878
+    },
+    {
+      "acc": 0.68943815,
+      "epoch": 0.6236681887366818,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.239890321535163e-06,
+      "loss": 1.40269299,
+      "memory(GiB)": 113.67,
+      "step": 24585,
+      "train_speed(iter/s)": 1.317905
+    },
+    {
+      "acc": 0.68103065,
+      "epoch": 0.6237950279046169,
+      "grad_norm": 3.125,
+      "learning_rate": 8.239091556628395e-06,
+      "loss": 1.44911261,
+      "memory(GiB)": 113.67,
+      "step": 24590,
+      "train_speed(iter/s)": 1.317932
+    },
+    {
+      "acc": 0.6810792,
+      "epoch": 0.623921867072552,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.238292649251918e-06,
+      "loss": 1.43460007,
+      "memory(GiB)": 113.67,
+      "step": 24595,
+      "train_speed(iter/s)": 1.317962
+    },
+    {
+      "acc": 0.69737902,
+      "epoch": 0.624048706240487,
+      "grad_norm": 3.0,
+      "learning_rate": 8.237493599440871e-06,
+      "loss": 1.39283476,
+      "memory(GiB)": 113.67,
+      "step": 24600,
+      "train_speed(iter/s)": 1.317992
+    },
+    {
+      "acc": 0.66769619,
+      "epoch": 0.6241755454084221,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.236694407230402e-06,
+      "loss": 1.54185162,
+      "memory(GiB)": 113.67,
+      "step": 24605,
+      "train_speed(iter/s)": 1.318021
+    },
+    {
+      "acc": 0.7061255,
+      "epoch": 0.6243023845763572,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.235895072655664e-06,
+      "loss": 1.28215513,
+      "memory(GiB)": 113.67,
+      "step": 24610,
+      "train_speed(iter/s)": 1.318049
+    },
+    {
+      "acc": 0.68178902,
+      "epoch": 0.6244292237442922,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.235095595751809e-06,
+      "loss": 1.40205126,
+      "memory(GiB)": 113.67,
+      "step": 24615,
+      "train_speed(iter/s)": 1.318079
+    },
+    {
+      "acc": 0.69555836,
+      "epoch": 0.6245560629122273,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.23429597655401e-06,
+      "loss": 1.39333019,
+      "memory(GiB)": 113.67,
+      "step": 24620,
+      "train_speed(iter/s)": 1.318107
+    },
+    {
+      "acc": 0.6683754,
+      "epoch": 0.6246829020801623,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.233496215097433e-06,
+      "loss": 1.47221909,
+      "memory(GiB)": 113.67,
+      "step": 24625,
+      "train_speed(iter/s)": 1.318137
+    },
+    {
+      "acc": 0.69233761,
+      "epoch": 0.6248097412480974,
+      "grad_norm": 3.0,
+      "learning_rate": 8.232696311417256e-06,
+      "loss": 1.34255056,
+      "memory(GiB)": 113.67,
+      "step": 24630,
+      "train_speed(iter/s)": 1.318165
+    },
+    {
+      "acc": 0.67399263,
+      "epoch": 0.6249365804160325,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.231896265548662e-06,
+      "loss": 1.45162086,
+      "memory(GiB)": 113.67,
+      "step": 24635,
+      "train_speed(iter/s)": 1.318191
+    },
+    {
+      "acc": 0.68279424,
+      "epoch": 0.6250634195839675,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.231096077526841e-06,
+      "loss": 1.43508568,
+      "memory(GiB)": 113.67,
+      "step": 24640,
+      "train_speed(iter/s)": 1.31822
+    },
+    {
+      "acc": 0.70261703,
+      "epoch": 0.6251902587519026,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.230295747386988e-06,
+      "loss": 1.34601974,
+      "memory(GiB)": 113.67,
+      "step": 24645,
+      "train_speed(iter/s)": 1.31825
+    },
+    {
+      "acc": 0.69289589,
+      "epoch": 0.6253170979198377,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.229495275164307e-06,
+      "loss": 1.31735001,
+      "memory(GiB)": 113.67,
+      "step": 24650,
+      "train_speed(iter/s)": 1.318279
+    },
+    {
+      "acc": 0.68638906,
+      "epoch": 0.6254439370877727,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.228694660894003e-06,
+      "loss": 1.44650517,
+      "memory(GiB)": 113.67,
+      "step": 24655,
+      "train_speed(iter/s)": 1.318309
+    },
+    {
+      "acc": 0.69611702,
+      "epoch": 0.6255707762557078,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.227893904611295e-06,
+      "loss": 1.37424784,
+      "memory(GiB)": 113.67,
+      "step": 24660,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "acc": 0.68972993,
+      "epoch": 0.6256976154236428,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.2270930063514e-06,
+      "loss": 1.38502235,
+      "memory(GiB)": 113.67,
+      "step": 24665,
+      "train_speed(iter/s)": 1.31837
+    },
+    {
+      "acc": 0.68220301,
+      "epoch": 0.6258244545915779,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.226291966149549e-06,
+      "loss": 1.43960857,
+      "memory(GiB)": 113.67,
+      "step": 24670,
+      "train_speed(iter/s)": 1.3184
+    },
+    {
+      "acc": 0.70051098,
+      "epoch": 0.625951293759513,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.225490784040971e-06,
+      "loss": 1.3559021,
+      "memory(GiB)": 113.67,
+      "step": 24675,
+      "train_speed(iter/s)": 1.318432
+    },
+    {
+      "acc": 0.68093181,
+      "epoch": 0.626078132927448,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.224689460060908e-06,
+      "loss": 1.42499704,
+      "memory(GiB)": 113.67,
+      "step": 24680,
+      "train_speed(iter/s)": 1.318463
+    },
+    {
+      "acc": 0.6952641,
+      "epoch": 0.6262049720953831,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.223887994244604e-06,
+      "loss": 1.36143389,
+      "memory(GiB)": 113.67,
+      "step": 24685,
+      "train_speed(iter/s)": 1.318493
+    },
+    {
+      "acc": 0.69548712,
+      "epoch": 0.6263318112633182,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.223086386627314e-06,
+      "loss": 1.35529861,
+      "memory(GiB)": 113.67,
+      "step": 24690,
+      "train_speed(iter/s)": 1.318524
+    },
+    {
+      "acc": 0.7032505,
+      "epoch": 0.6264586504312532,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.222284637244296e-06,
+      "loss": 1.31977339,
+      "memory(GiB)": 113.67,
+      "step": 24695,
+      "train_speed(iter/s)": 1.318554
+    },
+    {
+      "acc": 0.68525271,
+      "epoch": 0.6265854895991883,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.221482746130811e-06,
+      "loss": 1.36016216,
+      "memory(GiB)": 113.67,
+      "step": 24700,
+      "train_speed(iter/s)": 1.318586
+    },
+    {
+      "acc": 0.68526745,
+      "epoch": 0.6267123287671232,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.220680713322131e-06,
+      "loss": 1.42026997,
+      "memory(GiB)": 113.67,
+      "step": 24705,
+      "train_speed(iter/s)": 1.318617
+    },
+    {
+      "acc": 0.68399439,
+      "epoch": 0.6268391679350583,
+      "grad_norm": 3.0,
+      "learning_rate": 8.219878538853537e-06,
+      "loss": 1.36798763,
+      "memory(GiB)": 113.67,
+      "step": 24710,
+      "train_speed(iter/s)": 1.318647
+    },
+    {
+      "acc": 0.67040153,
+      "epoch": 0.6269660071029934,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.219076222760307e-06,
+      "loss": 1.46586723,
+      "memory(GiB)": 113.67,
+      "step": 24715,
+      "train_speed(iter/s)": 1.318677
+    },
+    {
+      "acc": 0.6649312,
+      "epoch": 0.6270928462709284,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.218273765077734e-06,
+      "loss": 1.40041265,
+      "memory(GiB)": 113.67,
+      "step": 24720,
+      "train_speed(iter/s)": 1.318708
+    },
+    {
+      "acc": 0.6877449,
+      "epoch": 0.6272196854388635,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.21747116584111e-06,
+      "loss": 1.39354763,
+      "memory(GiB)": 113.67,
+      "step": 24725,
+      "train_speed(iter/s)": 1.318739
+    },
+    {
+      "acc": 0.68551941,
+      "epoch": 0.6273465246067986,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.21666842508574e-06,
+      "loss": 1.34058628,
+      "memory(GiB)": 113.67,
+      "step": 24730,
+      "train_speed(iter/s)": 1.318769
+    },
+    {
+      "acc": 0.68164873,
+      "epoch": 0.6274733637747336,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.215865542846932e-06,
+      "loss": 1.47655611,
+      "memory(GiB)": 113.67,
+      "step": 24735,
+      "train_speed(iter/s)": 1.3188
+    },
+    {
+      "acc": 0.67554445,
+      "epoch": 0.6276002029426687,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.215062519160002e-06,
+      "loss": 1.42443848,
+      "memory(GiB)": 113.67,
+      "step": 24740,
+      "train_speed(iter/s)": 1.318831
+    },
+    {
+      "acc": 0.68462062,
+      "epoch": 0.6277270421106037,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.214259354060263e-06,
+      "loss": 1.41746283,
+      "memory(GiB)": 113.67,
+      "step": 24745,
+      "train_speed(iter/s)": 1.318861
+    },
+    {
+      "acc": 0.68289509,
+      "epoch": 0.6278538812785388,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.21345604758305e-06,
+      "loss": 1.37405424,
+      "memory(GiB)": 113.67,
+      "step": 24750,
+      "train_speed(iter/s)": 1.318892
+    },
+    {
+      "acc": 0.69359841,
+      "epoch": 0.6279807204464739,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.212652599763693e-06,
+      "loss": 1.37280369,
+      "memory(GiB)": 113.67,
+      "step": 24755,
+      "train_speed(iter/s)": 1.318923
+    },
+    {
+      "acc": 0.6804451,
+      "epoch": 0.6281075596144089,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.211849010637532e-06,
+      "loss": 1.39033403,
+      "memory(GiB)": 113.67,
+      "step": 24760,
+      "train_speed(iter/s)": 1.318955
+    },
+    {
+      "acc": 0.67523379,
+      "epoch": 0.628234398782344,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.211045280239908e-06,
+      "loss": 1.37049618,
+      "memory(GiB)": 113.67,
+      "step": 24765,
+      "train_speed(iter/s)": 1.318986
+    },
+    {
+      "acc": 0.68227782,
+      "epoch": 0.6283612379502791,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.210241408606182e-06,
+      "loss": 1.35722456,
+      "memory(GiB)": 113.67,
+      "step": 24770,
+      "train_speed(iter/s)": 1.319018
+    },
+    {
+      "acc": 0.69086385,
+      "epoch": 0.6284880771182141,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.2094373957717e-06,
+      "loss": 1.41099968,
+      "memory(GiB)": 113.67,
+      "step": 24775,
+      "train_speed(iter/s)": 1.319049
+    },
+    {
+      "acc": 0.68588753,
+      "epoch": 0.6286149162861492,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.208633241771836e-06,
+      "loss": 1.37459316,
+      "memory(GiB)": 113.67,
+      "step": 24780,
+      "train_speed(iter/s)": 1.319079
+    },
+    {
+      "acc": 0.68301811,
+      "epoch": 0.6287417554540842,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.207828946641956e-06,
+      "loss": 1.41251125,
+      "memory(GiB)": 113.67,
+      "step": 24785,
+      "train_speed(iter/s)": 1.319109
+    },
+    {
+      "acc": 0.69467726,
+      "epoch": 0.6288685946220193,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.207024510417436e-06,
+      "loss": 1.37117767,
+      "memory(GiB)": 113.67,
+      "step": 24790,
+      "train_speed(iter/s)": 1.319139
+    },
+    {
+      "acc": 0.68180351,
+      "epoch": 0.6289954337899544,
+      "grad_norm": 4.25,
+      "learning_rate": 8.20621993313366e-06,
+      "loss": 1.43577652,
+      "memory(GiB)": 113.67,
+      "step": 24795,
+      "train_speed(iter/s)": 1.31917
+    },
+    {
+      "acc": 0.69024558,
+      "epoch": 0.6291222729578894,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.205415214826018e-06,
+      "loss": 1.54861689,
+      "memory(GiB)": 113.67,
+      "step": 24800,
+      "train_speed(iter/s)": 1.319199
+    },
+    {
+      "acc": 0.68078547,
+      "epoch": 0.6292491121258245,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.204610355529901e-06,
+      "loss": 1.41384935,
+      "memory(GiB)": 113.67,
+      "step": 24805,
+      "train_speed(iter/s)": 1.319229
+    },
+    {
+      "acc": 0.67464914,
+      "epoch": 0.6293759512937596,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.203805355280715e-06,
+      "loss": 1.43326321,
+      "memory(GiB)": 113.67,
+      "step": 24810,
+      "train_speed(iter/s)": 1.31926
+    },
+    {
+      "acc": 0.68673339,
+      "epoch": 0.6295027904616946,
+      "grad_norm": 2.75,
+      "learning_rate": 8.203000214113865e-06,
+      "loss": 1.42129688,
+      "memory(GiB)": 113.67,
+      "step": 24815,
+      "train_speed(iter/s)": 1.319291
+    },
+    {
+      "acc": 0.6925951,
+      "epoch": 0.6296296296296297,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.202194932064767e-06,
+      "loss": 1.38160706,
+      "memory(GiB)": 113.67,
+      "step": 24820,
+      "train_speed(iter/s)": 1.319321
+    },
+    {
+      "acc": 0.68413205,
+      "epoch": 0.6297564687975646,
+      "grad_norm": 3.75,
+      "learning_rate": 8.201389509168836e-06,
+      "loss": 1.4184516,
+      "memory(GiB)": 113.67,
+      "step": 24825,
+      "train_speed(iter/s)": 1.319352
+    },
+    {
+      "acc": 0.6715333,
+      "epoch": 0.6298833079654997,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.200583945461502e-06,
+      "loss": 1.43021507,
+      "memory(GiB)": 113.67,
+      "step": 24830,
+      "train_speed(iter/s)": 1.319383
+    },
+    {
+      "acc": 0.68664246,
+      "epoch": 0.6300101471334348,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.199778240978197e-06,
+      "loss": 1.36713371,
+      "memory(GiB)": 113.67,
+      "step": 24835,
+      "train_speed(iter/s)": 1.319414
+    },
+    {
+      "acc": 0.67104197,
+      "epoch": 0.6301369863013698,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.19897239575436e-06,
+      "loss": 1.47270985,
+      "memory(GiB)": 113.67,
+      "step": 24840,
+      "train_speed(iter/s)": 1.319446
+    },
+    {
+      "acc": 0.67054873,
+      "epoch": 0.6302638254693049,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.198166409825434e-06,
+      "loss": 1.46529446,
+      "memory(GiB)": 113.67,
+      "step": 24845,
+      "train_speed(iter/s)": 1.319476
+    },
+    {
+      "acc": 0.68675489,
+      "epoch": 0.63039066463724,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.19736028322687e-06,
+      "loss": 1.40671301,
+      "memory(GiB)": 113.67,
+      "step": 24850,
+      "train_speed(iter/s)": 1.319507
+    },
+    {
+      "acc": 0.68443761,
+      "epoch": 0.630517503805175,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.196554015994126e-06,
+      "loss": 1.42146711,
+      "memory(GiB)": 113.67,
+      "step": 24855,
+      "train_speed(iter/s)": 1.319537
+    },
+    {
+      "acc": 0.6834537,
+      "epoch": 0.6306443429731101,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.195747608162665e-06,
+      "loss": 1.41778011,
+      "memory(GiB)": 113.67,
+      "step": 24860,
+      "train_speed(iter/s)": 1.319567
+    },
+    {
+      "acc": 0.68450069,
+      "epoch": 0.6307711821410451,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.194941059767957e-06,
+      "loss": 1.41181641,
+      "memory(GiB)": 113.67,
+      "step": 24865,
+      "train_speed(iter/s)": 1.319598
+    },
+    {
+      "acc": 0.68075085,
+      "epoch": 0.6308980213089802,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.194134370845474e-06,
+      "loss": 1.39376831,
+      "memory(GiB)": 113.67,
+      "step": 24870,
+      "train_speed(iter/s)": 1.31963
+    },
+    {
+      "acc": 0.68366027,
+      "epoch": 0.6310248604769153,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.193327541430703e-06,
+      "loss": 1.42000427,
+      "memory(GiB)": 113.67,
+      "step": 24875,
+      "train_speed(iter/s)": 1.319661
+    },
+    {
+      "acc": 0.69562864,
+      "epoch": 0.6311516996448503,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.192520571559128e-06,
+      "loss": 1.44497318,
+      "memory(GiB)": 113.67,
+      "step": 24880,
+      "train_speed(iter/s)": 1.319693
+    },
+    {
+      "acc": 0.6929893,
+      "epoch": 0.6312785388127854,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.191713461266246e-06,
+      "loss": 1.37815084,
+      "memory(GiB)": 113.67,
+      "step": 24885,
+      "train_speed(iter/s)": 1.319725
+    },
+    {
+      "acc": 0.68034563,
+      "epoch": 0.6314053779807205,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.190906210587555e-06,
+      "loss": 1.41174259,
+      "memory(GiB)": 113.67,
+      "step": 24890,
+      "train_speed(iter/s)": 1.319753
+    },
+    {
+      "acc": 0.69334817,
+      "epoch": 0.6315322171486555,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.190098819558562e-06,
+      "loss": 1.35507755,
+      "memory(GiB)": 113.67,
+      "step": 24895,
+      "train_speed(iter/s)": 1.319783
+    },
+    {
+      "acc": 0.68136339,
+      "epoch": 0.6316590563165906,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.189291288214782e-06,
+      "loss": 1.39417419,
+      "memory(GiB)": 113.67,
+      "step": 24900,
+      "train_speed(iter/s)": 1.319813
+    },
+    {
+      "acc": 0.67777557,
+      "epoch": 0.6317858954845256,
+      "grad_norm": 2.4375,
+      "learning_rate": 8.18848361659173e-06,
+      "loss": 1.41754951,
+      "memory(GiB)": 113.67,
+      "step": 24905,
+      "train_speed(iter/s)": 1.319844
+    },
+    {
+      "acc": 0.69854255,
+      "epoch": 0.6319127346524607,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.187675804724935e-06,
+      "loss": 1.36721106,
+      "memory(GiB)": 113.67,
+      "step": 24910,
+      "train_speed(iter/s)": 1.319874
+    },
+    {
+      "acc": 0.68605838,
+      "epoch": 0.6320395738203958,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.186867852649925e-06,
+      "loss": 1.39701748,
+      "memory(GiB)": 113.67,
+      "step": 24915,
+      "train_speed(iter/s)": 1.319903
+    },
+    {
+      "acc": 0.68633823,
+      "epoch": 0.6321664129883308,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.186059760402238e-06,
+      "loss": 1.3843874,
+      "memory(GiB)": 113.67,
+      "step": 24920,
+      "train_speed(iter/s)": 1.319933
+    },
+    {
+      "acc": 0.69484501,
+      "epoch": 0.6322932521562659,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.185251528017419e-06,
+      "loss": 1.39510889,
+      "memory(GiB)": 113.67,
+      "step": 24925,
+      "train_speed(iter/s)": 1.319963
+    },
+    {
+      "acc": 0.67991667,
+      "epoch": 0.632420091324201,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.184443155531016e-06,
+      "loss": 1.40221748,
+      "memory(GiB)": 113.67,
+      "step": 24930,
+      "train_speed(iter/s)": 1.319993
+    },
+    {
+      "acc": 0.69282894,
+      "epoch": 0.632546930492136,
+      "grad_norm": 2.875,
+      "learning_rate": 8.183634642978586e-06,
+      "loss": 1.37950182,
+      "memory(GiB)": 113.67,
+      "step": 24935,
+      "train_speed(iter/s)": 1.320022
+    },
+    {
+      "acc": 0.68992329,
+      "epoch": 0.632673769660071,
+      "grad_norm": 4.125,
+      "learning_rate": 8.18282599039569e-06,
+      "loss": 1.39747534,
+      "memory(GiB)": 113.67,
+      "step": 24940,
+      "train_speed(iter/s)": 1.320053
+    },
+    {
+      "acc": 0.6849411,
+      "epoch": 0.632800608828006,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.182017197817898e-06,
+      "loss": 1.43450575,
+      "memory(GiB)": 113.67,
+      "step": 24945,
+      "train_speed(iter/s)": 1.320083
+    },
+    {
+      "acc": 0.68920259,
+      "epoch": 0.6329274479959411,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.181208265280782e-06,
+      "loss": 1.39892044,
+      "memory(GiB)": 113.67,
+      "step": 24950,
+      "train_speed(iter/s)": 1.320112
+    },
+    {
+      "acc": 0.67520599,
+      "epoch": 0.6330542871638762,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.180399192819923e-06,
+      "loss": 1.46291142,
+      "memory(GiB)": 113.67,
+      "step": 24955,
+      "train_speed(iter/s)": 1.320142
+    },
+    {
+      "acc": 0.68179884,
+      "epoch": 0.6331811263318112,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.17958998047091e-06,
+      "loss": 1.42110481,
+      "memory(GiB)": 113.67,
+      "step": 24960,
+      "train_speed(iter/s)": 1.320172
+    },
+    {
+      "acc": 0.6925633,
+      "epoch": 0.6333079654997463,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.178780628269332e-06,
+      "loss": 1.34022713,
+      "memory(GiB)": 113.67,
+      "step": 24965,
+      "train_speed(iter/s)": 1.320202
+    },
+    {
+      "acc": 0.6860662,
+      "epoch": 0.6334348046676814,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.177971136250788e-06,
+      "loss": 1.45435867,
+      "memory(GiB)": 113.67,
+      "step": 24970,
+      "train_speed(iter/s)": 1.320229
+    },
+    {
+      "acc": 0.69132462,
+      "epoch": 0.6335616438356164,
+      "grad_norm": 2.75,
+      "learning_rate": 8.177161504450887e-06,
+      "loss": 1.37364225,
+      "memory(GiB)": 113.67,
+      "step": 24975,
+      "train_speed(iter/s)": 1.320259
+    },
+    {
+      "acc": 0.67624283,
+      "epoch": 0.6336884830035515,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.176351732905239e-06,
+      "loss": 1.4642168,
+      "memory(GiB)": 113.67,
+      "step": 24980,
+      "train_speed(iter/s)": 1.320289
+    },
+    {
+      "acc": 0.69012156,
+      "epoch": 0.6338153221714865,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.175541821649459e-06,
+      "loss": 1.40756912,
+      "memory(GiB)": 113.67,
+      "step": 24985,
+      "train_speed(iter/s)": 1.320318
+    },
+    {
+      "acc": 0.7084691,
+      "epoch": 0.6339421613394216,
+      "grad_norm": 3.25,
+      "learning_rate": 8.174731770719173e-06,
+      "loss": 1.32791634,
+      "memory(GiB)": 113.67,
+      "step": 24990,
+      "train_speed(iter/s)": 1.320346
+    },
+    {
+      "acc": 0.69135013,
+      "epoch": 0.6340690005073567,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.173921580150008e-06,
+      "loss": 1.41324959,
+      "memory(GiB)": 113.67,
+      "step": 24995,
+      "train_speed(iter/s)": 1.320375
+    },
+    {
+      "acc": 0.67378397,
+      "epoch": 0.6341958396752917,
+      "grad_norm": 3.25,
+      "learning_rate": 8.173111249977602e-06,
+      "loss": 1.46754971,
+      "memory(GiB)": 113.67,
+      "step": 25000,
+      "train_speed(iter/s)": 1.320405
+    },
+    {
+      "epoch": 0.6341958396752917,
+      "eval_acc": 0.6740429815164026,
+      "eval_loss": 1.361513614654541,
+      "eval_runtime": 69.5244,
+      "eval_samples_per_second": 91.622,
+      "eval_steps_per_second": 22.913,
+      "step": 25000
+    },
+    {
+      "acc": 0.69750948,
+      "epoch": 0.6343226788432268,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.172300780237596e-06,
+      "loss": 1.32421675,
+      "memory(GiB)": 113.67,
+      "step": 25005,
+      "train_speed(iter/s)": 1.314729
+    },
+    {
+      "acc": 0.68245392,
+      "epoch": 0.6344495180111619,
+      "grad_norm": 3.375,
+      "learning_rate": 8.171490170965639e-06,
+      "loss": 1.39134235,
+      "memory(GiB)": 113.67,
+      "step": 25010,
+      "train_speed(iter/s)": 1.31476
+    },
+    {
+      "acc": 0.7020031,
+      "epoch": 0.6345763571790969,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.170679422197385e-06,
+      "loss": 1.32624407,
+      "memory(GiB)": 113.67,
+      "step": 25015,
+      "train_speed(iter/s)": 1.314789
+    },
+    {
+      "acc": 0.68566523,
+      "epoch": 0.634703196347032,
+      "grad_norm": 3.25,
+      "learning_rate": 8.169868533968493e-06,
+      "loss": 1.38714151,
+      "memory(GiB)": 113.67,
+      "step": 25020,
+      "train_speed(iter/s)": 1.314819
+    },
+    {
+      "acc": 0.67359862,
+      "epoch": 0.634830035514967,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.16905750631463e-06,
+      "loss": 1.41877937,
+      "memory(GiB)": 113.67,
+      "step": 25025,
+      "train_speed(iter/s)": 1.314849
+    },
+    {
+      "acc": 0.69376035,
+      "epoch": 0.6349568746829021,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.168246339271471e-06,
+      "loss": 1.38843422,
+      "memory(GiB)": 113.67,
+      "step": 25030,
+      "train_speed(iter/s)": 1.314866
+    },
+    {
+      "acc": 0.66941748,
+      "epoch": 0.6350837138508372,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.16743503287469e-06,
+      "loss": 1.4165719,
+      "memory(GiB)": 113.67,
+      "step": 25035,
+      "train_speed(iter/s)": 1.314896
+    },
+    {
+      "acc": 0.68600445,
+      "epoch": 0.6352105530187722,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.166623587159978e-06,
+      "loss": 1.36639462,
+      "memory(GiB)": 113.67,
+      "step": 25040,
+      "train_speed(iter/s)": 1.314927
+    },
+    {
+      "acc": 0.68704562,
+      "epoch": 0.6353373921867073,
+      "grad_norm": 3.25,
+      "learning_rate": 8.16581200216302e-06,
+      "loss": 1.36148586,
+      "memory(GiB)": 113.67,
+      "step": 25045,
+      "train_speed(iter/s)": 1.314958
+    },
+    {
+      "acc": 0.67939124,
+      "epoch": 0.6354642313546424,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.165000277919517e-06,
+      "loss": 1.41767187,
+      "memory(GiB)": 113.67,
+      "step": 25050,
+      "train_speed(iter/s)": 1.314989
+    },
+    {
+      "acc": 0.68955669,
+      "epoch": 0.6355910705225774,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.16418841446517e-06,
+      "loss": 1.38883924,
+      "memory(GiB)": 113.67,
+      "step": 25055,
+      "train_speed(iter/s)": 1.31502
+    },
+    {
+      "acc": 0.68742232,
+      "epoch": 0.6357179096905125,
+      "grad_norm": 2.875,
+      "learning_rate": 8.163376411835691e-06,
+      "loss": 1.37079325,
+      "memory(GiB)": 113.67,
+      "step": 25060,
+      "train_speed(iter/s)": 1.315051
+    },
+    {
+      "acc": 0.70418134,
+      "epoch": 0.6358447488584474,
+      "grad_norm": 2.5,
+      "learning_rate": 8.162564270066793e-06,
+      "loss": 1.30899296,
+      "memory(GiB)": 113.67,
+      "step": 25065,
+      "train_speed(iter/s)": 1.31508
+    },
+    {
+      "acc": 0.69036479,
+      "epoch": 0.6359715880263825,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.1617519891942e-06,
+      "loss": 1.40943356,
+      "memory(GiB)": 113.67,
+      "step": 25070,
+      "train_speed(iter/s)": 1.315109
+    },
+    {
+      "acc": 0.68825922,
+      "epoch": 0.6360984271943176,
+      "grad_norm": 2.875,
+      "learning_rate": 8.160939569253637e-06,
+      "loss": 1.33983669,
+      "memory(GiB)": 113.67,
+      "step": 25075,
+      "train_speed(iter/s)": 1.315138
+    },
+    {
+      "acc": 0.68554983,
+      "epoch": 0.6362252663622526,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.160127010280838e-06,
+      "loss": 1.39553242,
+      "memory(GiB)": 113.67,
+      "step": 25080,
+      "train_speed(iter/s)": 1.315166
+    },
+    {
+      "acc": 0.68415661,
+      "epoch": 0.6363521055301877,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.159314312311546e-06,
+      "loss": 1.43861074,
+      "memory(GiB)": 113.67,
+      "step": 25085,
+      "train_speed(iter/s)": 1.315196
+    },
+    {
+      "acc": 0.68275433,
+      "epoch": 0.6364789446981228,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.158501475381505e-06,
+      "loss": 1.4534585,
+      "memory(GiB)": 113.67,
+      "step": 25090,
+      "train_speed(iter/s)": 1.315224
+    },
+    {
+      "acc": 0.68245316,
+      "epoch": 0.6366057838660578,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.157688499526466e-06,
+      "loss": 1.32232265,
+      "memory(GiB)": 113.67,
+      "step": 25095,
+      "train_speed(iter/s)": 1.315254
+    },
+    {
+      "acc": 0.6855094,
+      "epoch": 0.6367326230339929,
+      "grad_norm": 3.25,
+      "learning_rate": 8.15687538478219e-06,
+      "loss": 1.39811106,
+      "memory(GiB)": 113.67,
+      "step": 25100,
+      "train_speed(iter/s)": 1.315275
+    },
+    {
+      "acc": 0.67731371,
+      "epoch": 0.6368594622019279,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.156062131184439e-06,
+      "loss": 1.43376417,
+      "memory(GiB)": 113.67,
+      "step": 25105,
+      "train_speed(iter/s)": 1.315307
+    },
+    {
+      "acc": 0.68610296,
+      "epoch": 0.636986301369863,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.155248738768986e-06,
+      "loss": 1.40342655,
+      "memory(GiB)": 113.67,
+      "step": 25110,
+      "train_speed(iter/s)": 1.315336
+    },
+    {
+      "acc": 0.68980408,
+      "epoch": 0.6371131405377981,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.154435207571606e-06,
+      "loss": 1.3179966,
+      "memory(GiB)": 113.67,
+      "step": 25115,
+      "train_speed(iter/s)": 1.315367
+    },
+    {
+      "acc": 0.69348612,
+      "epoch": 0.6372399797057331,
+      "grad_norm": 4.4375,
+      "learning_rate": 8.153621537628083e-06,
+      "loss": 1.3651536,
+      "memory(GiB)": 113.67,
+      "step": 25120,
+      "train_speed(iter/s)": 1.315398
+    },
+    {
+      "acc": 0.68471661,
+      "epoch": 0.6373668188736682,
+      "grad_norm": 3.375,
+      "learning_rate": 8.152807728974203e-06,
+      "loss": 1.39159126,
+      "memory(GiB)": 113.67,
+      "step": 25125,
+      "train_speed(iter/s)": 1.315428
+    },
+    {
+      "acc": 0.69991241,
+      "epoch": 0.6374936580416033,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.151993781645765e-06,
+      "loss": 1.35098553,
+      "memory(GiB)": 113.67,
+      "step": 25130,
+      "train_speed(iter/s)": 1.315458
+    },
+    {
+      "acc": 0.67186403,
+      "epoch": 0.6376204972095383,
+      "grad_norm": 4.625,
+      "learning_rate": 8.151179695678565e-06,
+      "loss": 1.4287221,
+      "memory(GiB)": 113.67,
+      "step": 25135,
+      "train_speed(iter/s)": 1.315489
+    },
+    {
+      "acc": 0.69732933,
+      "epoch": 0.6377473363774734,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.150365471108414e-06,
+      "loss": 1.33355026,
+      "memory(GiB)": 113.67,
+      "step": 25140,
+      "train_speed(iter/s)": 1.31552
+    },
+    {
+      "acc": 0.67529244,
+      "epoch": 0.6378741755454084,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.149551107971125e-06,
+      "loss": 1.46495085,
+      "memory(GiB)": 113.67,
+      "step": 25145,
+      "train_speed(iter/s)": 1.31555
+    },
+    {
+      "acc": 0.67799482,
+      "epoch": 0.6380010147133435,
+      "grad_norm": 3.25,
+      "learning_rate": 8.148736606302517e-06,
+      "loss": 1.43355551,
+      "memory(GiB)": 113.67,
+      "step": 25150,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.67592793,
+      "epoch": 0.6381278538812786,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.147921966138412e-06,
+      "loss": 1.44158611,
+      "memory(GiB)": 113.67,
+      "step": 25155,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.70718923,
+      "epoch": 0.6382546930492136,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.147107187514647e-06,
+      "loss": 1.33894768,
+      "memory(GiB)": 113.67,
+      "step": 25160,
+      "train_speed(iter/s)": 1.315636
+    },
+    {
+      "acc": 0.69049835,
+      "epoch": 0.6383815322171487,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.146292270467056e-06,
+      "loss": 1.28380852,
+      "memory(GiB)": 113.67,
+      "step": 25165,
+      "train_speed(iter/s)": 1.315664
+    },
+    {
+      "acc": 0.69271846,
+      "epoch": 0.6385083713850838,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.145477215031486e-06,
+      "loss": 1.36462307,
+      "memory(GiB)": 113.67,
+      "step": 25170,
+      "train_speed(iter/s)": 1.315693
+    },
+    {
+      "acc": 0.68509169,
+      "epoch": 0.6386352105530188,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.144662021243782e-06,
+      "loss": 1.33677006,
+      "memory(GiB)": 113.67,
+      "step": 25175,
+      "train_speed(iter/s)": 1.315724
+    },
+    {
+      "acc": 0.68394938,
+      "epoch": 0.6387620497209539,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.143846689139805e-06,
+      "loss": 1.45861597,
+      "memory(GiB)": 113.67,
+      "step": 25180,
+      "train_speed(iter/s)": 1.315755
+    },
+    {
+      "acc": 0.67564869,
+      "epoch": 0.6388888888888888,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.143031218755411e-06,
+      "loss": 1.43424435,
+      "memory(GiB)": 113.67,
+      "step": 25185,
+      "train_speed(iter/s)": 1.315784
+    },
+    {
+      "acc": 0.66625009,
+      "epoch": 0.6390157280568239,
+      "grad_norm": 3.125,
+      "learning_rate": 8.142215610126474e-06,
+      "loss": 1.47693186,
+      "memory(GiB)": 113.67,
+      "step": 25190,
+      "train_speed(iter/s)": 1.315814
+    },
+    {
+      "acc": 0.70401726,
+      "epoch": 0.639142567224759,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.141399863288863e-06,
+      "loss": 1.35549192,
+      "memory(GiB)": 113.67,
+      "step": 25195,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.67772417,
+      "epoch": 0.639269406392694,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.140583978278463e-06,
+      "loss": 1.38116302,
+      "memory(GiB)": 113.67,
+      "step": 25200,
+      "train_speed(iter/s)": 1.315869
+    },
+    {
+      "acc": 0.70902138,
+      "epoch": 0.6393962455606291,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.139767955131157e-06,
+      "loss": 1.21811495,
+      "memory(GiB)": 113.67,
+      "step": 25205,
+      "train_speed(iter/s)": 1.315899
+    },
+    {
+      "acc": 0.70336685,
+      "epoch": 0.6395230847285642,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.138951793882838e-06,
+      "loss": 1.33366098,
+      "memory(GiB)": 113.67,
+      "step": 25210,
+      "train_speed(iter/s)": 1.315929
+    },
+    {
+      "acc": 0.69655008,
+      "epoch": 0.6396499238964992,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.138135494569405e-06,
+      "loss": 1.35065479,
+      "memory(GiB)": 113.67,
+      "step": 25215,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.6913362,
+      "epoch": 0.6397767630644343,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.137319057226763e-06,
+      "loss": 1.33208113,
+      "memory(GiB)": 113.67,
+      "step": 25220,
+      "train_speed(iter/s)": 1.31599
+    },
+    {
+      "acc": 0.69197445,
+      "epoch": 0.6399036022323693,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.136502481890821e-06,
+      "loss": 1.40611429,
+      "memory(GiB)": 113.67,
+      "step": 25225,
+      "train_speed(iter/s)": 1.31602
+    },
+    {
+      "acc": 0.69600468,
+      "epoch": 0.6400304414003044,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.135685768597496e-06,
+      "loss": 1.3738121,
+      "memory(GiB)": 113.67,
+      "step": 25230,
+      "train_speed(iter/s)": 1.316049
+    },
+    {
+      "acc": 0.68199539,
+      "epoch": 0.6401572805682395,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.134868917382713e-06,
+      "loss": 1.36942625,
+      "memory(GiB)": 113.67,
+      "step": 25235,
+      "train_speed(iter/s)": 1.316078
+    },
+    {
+      "acc": 0.68470292,
+      "epoch": 0.6402841197361745,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.134051928282396e-06,
+      "loss": 1.40023022,
+      "memory(GiB)": 113.67,
+      "step": 25240,
+      "train_speed(iter/s)": 1.316107
+    },
+    {
+      "acc": 0.68826199,
+      "epoch": 0.6404109589041096,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.133234801332484e-06,
+      "loss": 1.43491678,
+      "memory(GiB)": 113.67,
+      "step": 25245,
+      "train_speed(iter/s)": 1.316137
+    },
+    {
+      "acc": 0.69326029,
+      "epoch": 0.6405377980720447,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.132417536568918e-06,
+      "loss": 1.34466839,
+      "memory(GiB)": 113.67,
+      "step": 25250,
+      "train_speed(iter/s)": 1.316167
+    },
+    {
+      "acc": 0.67309418,
+      "epoch": 0.6406646372399797,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.131600134027641e-06,
+      "loss": 1.44307432,
+      "memory(GiB)": 113.67,
+      "step": 25255,
+      "train_speed(iter/s)": 1.316197
+    },
+    {
+      "acc": 0.68354888,
+      "epoch": 0.6407914764079148,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.13078259374461e-06,
+      "loss": 1.39933252,
+      "memory(GiB)": 113.67,
+      "step": 25260,
+      "train_speed(iter/s)": 1.316223
+    },
+    {
+      "acc": 0.6804142,
+      "epoch": 0.6409183155758498,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.129964915755781e-06,
+      "loss": 1.39734497,
+      "memory(GiB)": 113.67,
+      "step": 25265,
+      "train_speed(iter/s)": 1.316254
+    },
+    {
+      "acc": 0.69708548,
+      "epoch": 0.6410451547437849,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.129147100097122e-06,
+      "loss": 1.31517429,
+      "memory(GiB)": 113.67,
+      "step": 25270,
+      "train_speed(iter/s)": 1.316282
+    },
+    {
+      "acc": 0.68720593,
+      "epoch": 0.64117199391172,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.128329146804604e-06,
+      "loss": 1.40593634,
+      "memory(GiB)": 113.67,
+      "step": 25275,
+      "train_speed(iter/s)": 1.316311
+    },
+    {
+      "acc": 0.69078026,
+      "epoch": 0.641298833079655,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.127511055914201e-06,
+      "loss": 1.37713175,
+      "memory(GiB)": 113.67,
+      "step": 25280,
+      "train_speed(iter/s)": 1.316342
+    },
+    {
+      "acc": 0.71220155,
+      "epoch": 0.6414256722475901,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.1266928274619e-06,
+      "loss": 1.35392246,
+      "memory(GiB)": 113.67,
+      "step": 25285,
+      "train_speed(iter/s)": 1.316373
+    },
+    {
+      "acc": 0.68803868,
+      "epoch": 0.6415525114155252,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.125874461483687e-06,
+      "loss": 1.37766876,
+      "memory(GiB)": 113.67,
+      "step": 25290,
+      "train_speed(iter/s)": 1.316404
+    },
+    {
+      "acc": 0.66024923,
+      "epoch": 0.6416793505834602,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.12505595801556e-06,
+      "loss": 1.47370872,
+      "memory(GiB)": 113.67,
+      "step": 25295,
+      "train_speed(iter/s)": 1.316435
+    },
+    {
+      "acc": 0.67457733,
+      "epoch": 0.6418061897513953,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.12423731709352e-06,
+      "loss": 1.46985741,
+      "memory(GiB)": 113.67,
+      "step": 25300,
+      "train_speed(iter/s)": 1.316465
+    },
+    {
+      "acc": 0.67728262,
+      "epoch": 0.6419330289193302,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.123418538753573e-06,
+      "loss": 1.36092281,
+      "memory(GiB)": 113.67,
+      "step": 25305,
+      "train_speed(iter/s)": 1.316496
+    },
+    {
+      "acc": 0.69134893,
+      "epoch": 0.6420598680872653,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.122599623031735e-06,
+      "loss": 1.35727577,
+      "memory(GiB)": 113.67,
+      "step": 25310,
+      "train_speed(iter/s)": 1.316526
+    },
+    {
+      "acc": 0.67909236,
+      "epoch": 0.6421867072552004,
+      "grad_norm": 2.75,
+      "learning_rate": 8.121780569964024e-06,
+      "loss": 1.38625584,
+      "memory(GiB)": 113.67,
+      "step": 25315,
+      "train_speed(iter/s)": 1.316553
+    },
+    {
+      "acc": 0.69410219,
+      "epoch": 0.6423135464231354,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.120961379586466e-06,
+      "loss": 1.36490393,
+      "memory(GiB)": 113.67,
+      "step": 25320,
+      "train_speed(iter/s)": 1.316583
+    },
+    {
+      "acc": 0.672539,
+      "epoch": 0.6424403855910705,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.120142051935092e-06,
+      "loss": 1.44884586,
+      "memory(GiB)": 113.67,
+      "step": 25325,
+      "train_speed(iter/s)": 1.316613
+    },
+    {
+      "acc": 0.66245584,
+      "epoch": 0.6425672247590056,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.11932258704594e-06,
+      "loss": 1.49394093,
+      "memory(GiB)": 113.67,
+      "step": 25330,
+      "train_speed(iter/s)": 1.316643
+    },
+    {
+      "acc": 0.66193581,
+      "epoch": 0.6426940639269406,
+      "grad_norm": 4.625,
+      "learning_rate": 8.118502984955053e-06,
+      "loss": 1.51513119,
+      "memory(GiB)": 113.67,
+      "step": 25335,
+      "train_speed(iter/s)": 1.316673
+    },
+    {
+      "acc": 0.68183227,
+      "epoch": 0.6428209030948757,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.117683245698483e-06,
+      "loss": 1.39808302,
+      "memory(GiB)": 113.67,
+      "step": 25340,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.68482199,
+      "epoch": 0.6429477422628107,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.116863369312283e-06,
+      "loss": 1.44483051,
+      "memory(GiB)": 113.67,
+      "step": 25345,
+      "train_speed(iter/s)": 1.316733
+    },
+    {
+      "acc": 0.68744712,
+      "epoch": 0.6430745814307458,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.116043355832518e-06,
+      "loss": 1.43690472,
+      "memory(GiB)": 113.67,
+      "step": 25350,
+      "train_speed(iter/s)": 1.316763
+    },
+    {
+      "acc": 0.684093,
+      "epoch": 0.6432014205986809,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.115223205295253e-06,
+      "loss": 1.42597322,
+      "memory(GiB)": 113.67,
+      "step": 25355,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.69258747,
+      "epoch": 0.6433282597666159,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.114402917736563e-06,
+      "loss": 1.39445286,
+      "memory(GiB)": 113.67,
+      "step": 25360,
+      "train_speed(iter/s)": 1.316824
+    },
+    {
+      "acc": 0.68611898,
+      "epoch": 0.643455098934551,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.113582493192529e-06,
+      "loss": 1.33456287,
+      "memory(GiB)": 113.67,
+      "step": 25365,
+      "train_speed(iter/s)": 1.316854
+    },
+    {
+      "acc": 0.69634123,
+      "epoch": 0.6435819381024861,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.112761931699235e-06,
+      "loss": 1.39957676,
+      "memory(GiB)": 113.67,
+      "step": 25370,
+      "train_speed(iter/s)": 1.316883
+    },
+    {
+      "acc": 0.68080263,
+      "epoch": 0.6437087772704211,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.111941233292772e-06,
+      "loss": 1.4421731,
+      "memory(GiB)": 113.67,
+      "step": 25375,
+      "train_speed(iter/s)": 1.316912
+    },
+    {
+      "acc": 0.69846969,
+      "epoch": 0.6438356164383562,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.111120398009243e-06,
+      "loss": 1.34980536,
+      "memory(GiB)": 113.67,
+      "step": 25380,
+      "train_speed(iter/s)": 1.316941
+    },
+    {
+      "acc": 0.67604055,
+      "epoch": 0.6439624556062912,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.110299425884745e-06,
+      "loss": 1.41835365,
+      "memory(GiB)": 113.67,
+      "step": 25385,
+      "train_speed(iter/s)": 1.316971
+    },
+    {
+      "acc": 0.67424164,
+      "epoch": 0.6440892947742263,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.109478316955394e-06,
+      "loss": 1.45284891,
+      "memory(GiB)": 113.67,
+      "step": 25390,
+      "train_speed(iter/s)": 1.317
+    },
+    {
+      "acc": 0.6816256,
+      "epoch": 0.6442161339421614,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.108657071257304e-06,
+      "loss": 1.41657019,
+      "memory(GiB)": 113.67,
+      "step": 25395,
+      "train_speed(iter/s)": 1.317029
+    },
+    {
+      "acc": 0.68065166,
+      "epoch": 0.6443429731100964,
+      "grad_norm": 3.25,
+      "learning_rate": 8.107835688826598e-06,
+      "loss": 1.41824417,
+      "memory(GiB)": 113.67,
+      "step": 25400,
+      "train_speed(iter/s)": 1.317058
+    },
+    {
+      "acc": 0.69447403,
+      "epoch": 0.6444698122780315,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.1070141696994e-06,
+      "loss": 1.35815077,
+      "memory(GiB)": 113.67,
+      "step": 25405,
+      "train_speed(iter/s)": 1.317088
+    },
+    {
+      "acc": 0.69544883,
+      "epoch": 0.6445966514459666,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.106192513911849e-06,
+      "loss": 1.3947237,
+      "memory(GiB)": 113.67,
+      "step": 25410,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.68358183,
+      "epoch": 0.6447234906139016,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.105370721500083e-06,
+      "loss": 1.4705368,
+      "memory(GiB)": 113.67,
+      "step": 25415,
+      "train_speed(iter/s)": 1.317146
+    },
+    {
+      "acc": 0.68641968,
+      "epoch": 0.6448503297818367,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.104548792500246e-06,
+      "loss": 1.35772667,
+      "memory(GiB)": 113.67,
+      "step": 25420,
+      "train_speed(iter/s)": 1.317173
+    },
+    {
+      "acc": 0.68310061,
+      "epoch": 0.6449771689497716,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.103726726948495e-06,
+      "loss": 1.41775732,
+      "memory(GiB)": 113.67,
+      "step": 25425,
+      "train_speed(iter/s)": 1.317203
+    },
+    {
+      "acc": 0.68262324,
+      "epoch": 0.6451040081177067,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.102904524880985e-06,
+      "loss": 1.39452238,
+      "memory(GiB)": 113.67,
+      "step": 25430,
+      "train_speed(iter/s)": 1.317232
+    },
+    {
+      "acc": 0.67801752,
+      "epoch": 0.6452308472856418,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.10208218633388e-06,
+      "loss": 1.41539478,
+      "memory(GiB)": 113.67,
+      "step": 25435,
+      "train_speed(iter/s)": 1.317262
+    },
+    {
+      "acc": 0.68192315,
+      "epoch": 0.6453576864535768,
+      "grad_norm": 5.59375,
+      "learning_rate": 8.10125971134335e-06,
+      "loss": 1.40402689,
+      "memory(GiB)": 113.67,
+      "step": 25440,
+      "train_speed(iter/s)": 1.317292
+    },
+    {
+      "acc": 0.69022083,
+      "epoch": 0.6454845256215119,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.100437099945572e-06,
+      "loss": 1.36222172,
+      "memory(GiB)": 113.67,
+      "step": 25445,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.68752036,
+      "epoch": 0.645611364789447,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.099614352176727e-06,
+      "loss": 1.3615612,
+      "memory(GiB)": 113.67,
+      "step": 25450,
+      "train_speed(iter/s)": 1.317351
+    },
+    {
+      "acc": 0.68965845,
+      "epoch": 0.645738203957382,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.098791468073007e-06,
+      "loss": 1.39815569,
+      "memory(GiB)": 113.67,
+      "step": 25455,
+      "train_speed(iter/s)": 1.317381
+    },
+    {
+      "acc": 0.69576602,
+      "epoch": 0.6458650431253171,
+      "grad_norm": 3.625,
+      "learning_rate": 8.097968447670601e-06,
+      "loss": 1.39696836,
+      "memory(GiB)": 113.67,
+      "step": 25460,
+      "train_speed(iter/s)": 1.317411
+    },
+    {
+      "acc": 0.69166927,
+      "epoch": 0.6459918822932521,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.09714529100571e-06,
+      "loss": 1.39295349,
+      "memory(GiB)": 113.67,
+      "step": 25465,
+      "train_speed(iter/s)": 1.317441
+    },
+    {
+      "acc": 0.68799143,
+      "epoch": 0.6461187214611872,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.096321998114545e-06,
+      "loss": 1.40489216,
+      "memory(GiB)": 113.67,
+      "step": 25470,
+      "train_speed(iter/s)": 1.317469
+    },
+    {
+      "acc": 0.69704094,
+      "epoch": 0.6462455606291223,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.09549856903331e-06,
+      "loss": 1.32422667,
+      "memory(GiB)": 113.67,
+      "step": 25475,
+      "train_speed(iter/s)": 1.317499
+    },
+    {
+      "acc": 0.67656212,
+      "epoch": 0.6463723997970573,
+      "grad_norm": 2.875,
+      "learning_rate": 8.094675003798232e-06,
+      "loss": 1.44491701,
+      "memory(GiB)": 113.67,
+      "step": 25480,
+      "train_speed(iter/s)": 1.317529
+    },
+    {
+      "acc": 0.69939442,
+      "epoch": 0.6464992389649924,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.093851302445528e-06,
+      "loss": 1.33927956,
+      "memory(GiB)": 113.67,
+      "step": 25485,
+      "train_speed(iter/s)": 1.317555
+    },
+    {
+      "acc": 0.68554173,
+      "epoch": 0.6466260781329275,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.093027465011431e-06,
+      "loss": 1.37381248,
+      "memory(GiB)": 113.67,
+      "step": 25490,
+      "train_speed(iter/s)": 1.317585
+    },
+    {
+      "acc": 0.68717842,
+      "epoch": 0.6467529173008625,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.092203491532178e-06,
+      "loss": 1.34764051,
+      "memory(GiB)": 113.67,
+      "step": 25495,
+      "train_speed(iter/s)": 1.317613
+    },
+    {
+      "acc": 0.68500609,
+      "epoch": 0.6468797564687976,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.091379382044009e-06,
+      "loss": 1.41932602,
+      "memory(GiB)": 113.67,
+      "step": 25500,
+      "train_speed(iter/s)": 1.317643
+    },
+    {
+      "acc": 0.69336929,
+      "epoch": 0.6470065956367326,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.090555136583172e-06,
+      "loss": 1.37360764,
+      "memory(GiB)": 113.67,
+      "step": 25505,
+      "train_speed(iter/s)": 1.317671
+    },
+    {
+      "acc": 0.683986,
+      "epoch": 0.6471334348046677,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.089730755185921e-06,
+      "loss": 1.38257751,
+      "memory(GiB)": 113.67,
+      "step": 25510,
+      "train_speed(iter/s)": 1.317699
+    },
+    {
+      "acc": 0.68442593,
+      "epoch": 0.6472602739726028,
+      "grad_norm": 3.25,
+      "learning_rate": 8.088906237888517e-06,
+      "loss": 1.4275197,
+      "memory(GiB)": 113.67,
+      "step": 25515,
+      "train_speed(iter/s)": 1.317726
+    },
+    {
+      "acc": 0.67830782,
+      "epoch": 0.6473871131405378,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.088081584727228e-06,
+      "loss": 1.40269814,
+      "memory(GiB)": 113.67,
+      "step": 25520,
+      "train_speed(iter/s)": 1.317755
+    },
+    {
+      "acc": 0.67806926,
+      "epoch": 0.6475139523084729,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.08725679573832e-06,
+      "loss": 1.38634682,
+      "memory(GiB)": 113.67,
+      "step": 25525,
+      "train_speed(iter/s)": 1.317785
+    },
+    {
+      "acc": 0.69318895,
+      "epoch": 0.647640791476408,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.086431870958078e-06,
+      "loss": 1.38200684,
+      "memory(GiB)": 113.67,
+      "step": 25530,
+      "train_speed(iter/s)": 1.317809
+    },
+    {
+      "acc": 0.67804751,
+      "epoch": 0.647767630644343,
+      "grad_norm": 2.4375,
+      "learning_rate": 8.085606810422781e-06,
+      "loss": 1.41039505,
+      "memory(GiB)": 113.67,
+      "step": 25535,
+      "train_speed(iter/s)": 1.317839
+    },
+    {
+      "acc": 0.67567263,
+      "epoch": 0.647894469812278,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.08478161416872e-06,
+      "loss": 1.47096882,
+      "memory(GiB)": 113.67,
+      "step": 25540,
+      "train_speed(iter/s)": 1.317868
+    },
+    {
+      "acc": 0.68462343,
+      "epoch": 0.648021308980213,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.083956282232192e-06,
+      "loss": 1.44728298,
+      "memory(GiB)": 113.67,
+      "step": 25545,
+      "train_speed(iter/s)": 1.317898
+    },
+    {
+      "acc": 0.68410306,
+      "epoch": 0.6481481481481481,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.083130814649498e-06,
+      "loss": 1.35747385,
+      "memory(GiB)": 113.67,
+      "step": 25550,
+      "train_speed(iter/s)": 1.317928
+    },
+    {
+      "acc": 0.70545325,
+      "epoch": 0.6482749873160832,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.082305211456943e-06,
+      "loss": 1.31017857,
+      "memory(GiB)": 113.67,
+      "step": 25555,
+      "train_speed(iter/s)": 1.317957
+    },
+    {
+      "acc": 0.67554474,
+      "epoch": 0.6484018264840182,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.081479472690846e-06,
+      "loss": 1.46944265,
+      "memory(GiB)": 113.67,
+      "step": 25560,
+      "train_speed(iter/s)": 1.317986
+    },
+    {
+      "acc": 0.68596745,
+      "epoch": 0.6485286656519533,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.080653598387522e-06,
+      "loss": 1.34772453,
+      "memory(GiB)": 113.67,
+      "step": 25565,
+      "train_speed(iter/s)": 1.318013
+    },
+    {
+      "acc": 0.70767117,
+      "epoch": 0.6486555048198884,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.0798275885833e-06,
+      "loss": 1.35157337,
+      "memory(GiB)": 113.67,
+      "step": 25570,
+      "train_speed(iter/s)": 1.318042
+    },
+    {
+      "acc": 0.69183464,
+      "epoch": 0.6487823439878234,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.07900144331451e-06,
+      "loss": 1.3395834,
+      "memory(GiB)": 113.67,
+      "step": 25575,
+      "train_speed(iter/s)": 1.318071
+    },
+    {
+      "acc": 0.68765554,
+      "epoch": 0.6489091831557585,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.07817516261749e-06,
+      "loss": 1.40484781,
+      "memory(GiB)": 113.67,
+      "step": 25580,
+      "train_speed(iter/s)": 1.3181
+    },
+    {
+      "acc": 0.68277693,
+      "epoch": 0.6490360223236935,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.077348746528583e-06,
+      "loss": 1.36913128,
+      "memory(GiB)": 113.67,
+      "step": 25585,
+      "train_speed(iter/s)": 1.318129
+    },
+    {
+      "acc": 0.6952024,
+      "epoch": 0.6491628614916286,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.076522195084139e-06,
+      "loss": 1.32180586,
+      "memory(GiB)": 113.67,
+      "step": 25590,
+      "train_speed(iter/s)": 1.318157
+    },
+    {
+      "acc": 0.6977396,
+      "epoch": 0.6492897006595637,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.075695508320512e-06,
+      "loss": 1.40151415,
+      "memory(GiB)": 113.67,
+      "step": 25595,
+      "train_speed(iter/s)": 1.318185
+    },
+    {
+      "acc": 0.68276381,
+      "epoch": 0.6494165398274987,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.074868686274065e-06,
+      "loss": 1.49810696,
+      "memory(GiB)": 113.67,
+      "step": 25600,
+      "train_speed(iter/s)": 1.318214
+    },
+    {
+      "acc": 0.69954591,
+      "epoch": 0.6495433789954338,
+      "grad_norm": 4.0,
+      "learning_rate": 8.074041728981166e-06,
+      "loss": 1.33509007,
+      "memory(GiB)": 113.67,
+      "step": 25605,
+      "train_speed(iter/s)": 1.318244
+    },
+    {
+      "acc": 0.68632045,
+      "epoch": 0.6496702181633689,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.073214636478186e-06,
+      "loss": 1.38512669,
+      "memory(GiB)": 113.67,
+      "step": 25610,
+      "train_speed(iter/s)": 1.318273
+    },
+    {
+      "acc": 0.68994217,
+      "epoch": 0.6497970573313039,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.072387408801506e-06,
+      "loss": 1.43838806,
+      "memory(GiB)": 113.67,
+      "step": 25615,
+      "train_speed(iter/s)": 1.318302
+    },
+    {
+      "acc": 0.67604208,
+      "epoch": 0.649923896499239,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.07156004598751e-06,
+      "loss": 1.38490582,
+      "memory(GiB)": 113.67,
+      "step": 25620,
+      "train_speed(iter/s)": 1.318331
+    },
+    {
+      "acc": 0.68536706,
+      "epoch": 0.650050735667174,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.07073254807259e-06,
+      "loss": 1.40627594,
+      "memory(GiB)": 113.67,
+      "step": 25625,
+      "train_speed(iter/s)": 1.318361
+    },
+    {
+      "acc": 0.68291426,
+      "epoch": 0.6501775748351091,
+      "grad_norm": 2.75,
+      "learning_rate": 8.069904915093144e-06,
+      "loss": 1.40468655,
+      "memory(GiB)": 113.67,
+      "step": 25630,
+      "train_speed(iter/s)": 1.318391
+    },
+    {
+      "acc": 0.67864962,
+      "epoch": 0.6503044140030442,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.069077147085571e-06,
+      "loss": 1.3861187,
+      "memory(GiB)": 113.67,
+      "step": 25635,
+      "train_speed(iter/s)": 1.31842
+    },
+    {
+      "acc": 0.69792199,
+      "epoch": 0.6504312531709792,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.068249244086283e-06,
+      "loss": 1.36303587,
+      "memory(GiB)": 113.67,
+      "step": 25640,
+      "train_speed(iter/s)": 1.31845
+    },
+    {
+      "acc": 0.67588234,
+      "epoch": 0.6505580923389143,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.067421206131696e-06,
+      "loss": 1.46540728,
+      "memory(GiB)": 113.67,
+      "step": 25645,
+      "train_speed(iter/s)": 1.31848
+    },
+    {
+      "acc": 0.68562679,
+      "epoch": 0.6506849315068494,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.06659303325823e-06,
+      "loss": 1.37948542,
+      "memory(GiB)": 113.67,
+      "step": 25650,
+      "train_speed(iter/s)": 1.31851
+    },
+    {
+      "acc": 0.68567896,
+      "epoch": 0.6508117706747844,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.06576472550231e-06,
+      "loss": 1.42848988,
+      "memory(GiB)": 113.67,
+      "step": 25655,
+      "train_speed(iter/s)": 1.318539
+    },
+    {
+      "acc": 0.68199158,
+      "epoch": 0.6509386098427195,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.064936282900368e-06,
+      "loss": 1.42309837,
+      "memory(GiB)": 113.67,
+      "step": 25660,
+      "train_speed(iter/s)": 1.318566
+    },
+    {
+      "acc": 0.68127389,
+      "epoch": 0.6510654490106544,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.064107705488846e-06,
+      "loss": 1.38399658,
+      "memory(GiB)": 113.67,
+      "step": 25665,
+      "train_speed(iter/s)": 1.318595
+    },
+    {
+      "acc": 0.69190426,
+      "epoch": 0.6511922881785895,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.063278993304188e-06,
+      "loss": 1.34412403,
+      "memory(GiB)": 113.67,
+      "step": 25670,
+      "train_speed(iter/s)": 1.318623
+    },
+    {
+      "acc": 0.68257146,
+      "epoch": 0.6513191273465246,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.06245014638284e-06,
+      "loss": 1.40570545,
+      "memory(GiB)": 113.67,
+      "step": 25675,
+      "train_speed(iter/s)": 1.31865
+    },
+    {
+      "acc": 0.69038043,
+      "epoch": 0.6514459665144596,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.061621164761266e-06,
+      "loss": 1.33663244,
+      "memory(GiB)": 113.67,
+      "step": 25680,
+      "train_speed(iter/s)": 1.318677
+    },
+    {
+      "acc": 0.67706156,
+      "epoch": 0.6515728056823947,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.06079204847592e-06,
+      "loss": 1.35792618,
+      "memory(GiB)": 113.67,
+      "step": 25685,
+      "train_speed(iter/s)": 1.318705
+    },
+    {
+      "acc": 0.69558058,
+      "epoch": 0.6516996448503298,
+      "grad_norm": 2.875,
+      "learning_rate": 8.059962797563277e-06,
+      "loss": 1.36407194,
+      "memory(GiB)": 113.67,
+      "step": 25690,
+      "train_speed(iter/s)": 1.318734
+    },
+    {
+      "acc": 0.69682322,
+      "epoch": 0.6518264840182648,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.059133412059808e-06,
+      "loss": 1.35442419,
+      "memory(GiB)": 113.67,
+      "step": 25695,
+      "train_speed(iter/s)": 1.318762
+    },
+    {
+      "acc": 0.67961617,
+      "epoch": 0.6519533231861999,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.058303892001993e-06,
+      "loss": 1.38238735,
+      "memory(GiB)": 113.67,
+      "step": 25700,
+      "train_speed(iter/s)": 1.318784
+    },
+    {
+      "acc": 0.68117509,
+      "epoch": 0.6520801623541349,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.057474237426318e-06,
+      "loss": 1.38005447,
+      "memory(GiB)": 113.67,
+      "step": 25705,
+      "train_speed(iter/s)": 1.318813
+    },
+    {
+      "acc": 0.67834449,
+      "epoch": 0.65220700152207,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.056644448369275e-06,
+      "loss": 1.47099762,
+      "memory(GiB)": 113.67,
+      "step": 25710,
+      "train_speed(iter/s)": 1.31884
+    },
+    {
+      "acc": 0.68277392,
+      "epoch": 0.6523338406900051,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.055814524867364e-06,
+      "loss": 1.43009262,
+      "memory(GiB)": 113.67,
+      "step": 25715,
+      "train_speed(iter/s)": 1.318867
+    },
+    {
+      "acc": 0.67240224,
+      "epoch": 0.6524606798579401,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.054984466957085e-06,
+      "loss": 1.46572247,
+      "memory(GiB)": 113.67,
+      "step": 25720,
+      "train_speed(iter/s)": 1.318895
+    },
+    {
+      "acc": 0.70338159,
+      "epoch": 0.6525875190258752,
+      "grad_norm": 2.875,
+      "learning_rate": 8.05415427467495e-06,
+      "loss": 1.25809002,
+      "memory(GiB)": 113.67,
+      "step": 25725,
+      "train_speed(iter/s)": 1.318923
+    },
+    {
+      "acc": 0.69374781,
+      "epoch": 0.6527143581938103,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.053323948057477e-06,
+      "loss": 1.37555523,
+      "memory(GiB)": 113.67,
+      "step": 25730,
+      "train_speed(iter/s)": 1.318952
+    },
+    {
+      "acc": 0.67704067,
+      "epoch": 0.6528411973617453,
+      "grad_norm": 3.0,
+      "learning_rate": 8.052493487141183e-06,
+      "loss": 1.4009306,
+      "memory(GiB)": 113.67,
+      "step": 25735,
+      "train_speed(iter/s)": 1.31898
+    },
+    {
+      "acc": 0.67004833,
+      "epoch": 0.6529680365296804,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.051662891962594e-06,
+      "loss": 1.43182058,
+      "memory(GiB)": 113.67,
+      "step": 25740,
+      "train_speed(iter/s)": 1.319009
+    },
+    {
+      "acc": 0.68649254,
+      "epoch": 0.6530948756976154,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.05083216255825e-06,
+      "loss": 1.40978489,
+      "memory(GiB)": 113.67,
+      "step": 25745,
+      "train_speed(iter/s)": 1.319036
+    },
+    {
+      "acc": 0.68765211,
+      "epoch": 0.6532217148655505,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.050001298964685e-06,
+      "loss": 1.44263325,
+      "memory(GiB)": 113.67,
+      "step": 25750,
+      "train_speed(iter/s)": 1.319062
+    },
+    {
+      "acc": 0.6926652,
+      "epoch": 0.6533485540334856,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.049170301218445e-06,
+      "loss": 1.40711355,
+      "memory(GiB)": 113.67,
+      "step": 25755,
+      "train_speed(iter/s)": 1.31909
+    },
+    {
+      "acc": 0.69464645,
+      "epoch": 0.6534753932014206,
+      "grad_norm": 3.75,
+      "learning_rate": 8.048339169356085e-06,
+      "loss": 1.38034678,
+      "memory(GiB)": 113.67,
+      "step": 25760,
+      "train_speed(iter/s)": 1.31912
+    },
+    {
+      "acc": 0.69821877,
+      "epoch": 0.6536022323693557,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.047507903414155e-06,
+      "loss": 1.31620693,
+      "memory(GiB)": 113.67,
+      "step": 25765,
+      "train_speed(iter/s)": 1.319149
+    },
+    {
+      "acc": 0.70376959,
+      "epoch": 0.6537290715372908,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.046676503429222e-06,
+      "loss": 1.34798737,
+      "memory(GiB)": 113.67,
+      "step": 25770,
+      "train_speed(iter/s)": 1.319177
+    },
+    {
+      "acc": 0.69286814,
+      "epoch": 0.6538559107052258,
+      "grad_norm": 3.25,
+      "learning_rate": 8.045844969437855e-06,
+      "loss": 1.36309357,
+      "memory(GiB)": 113.67,
+      "step": 25775,
+      "train_speed(iter/s)": 1.319205
+    },
+    {
+      "acc": 0.67380428,
+      "epoch": 0.6539827498731609,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.045013301476625e-06,
+      "loss": 1.40829306,
+      "memory(GiB)": 113.67,
+      "step": 25780,
+      "train_speed(iter/s)": 1.319233
+    },
+    {
+      "acc": 0.68919015,
+      "epoch": 0.6541095890410958,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.044181499582117e-06,
+      "loss": 1.40960808,
+      "memory(GiB)": 113.67,
+      "step": 25785,
+      "train_speed(iter/s)": 1.31926
+    },
+    {
+      "acc": 0.69559736,
+      "epoch": 0.6542364282090309,
+      "grad_norm": 3.25,
+      "learning_rate": 8.043349563790917e-06,
+      "loss": 1.39287796,
+      "memory(GiB)": 113.67,
+      "step": 25790,
+      "train_speed(iter/s)": 1.319286
+    },
+    {
+      "acc": 0.69109364,
+      "epoch": 0.654363267376966,
+      "grad_norm": 3.0,
+      "learning_rate": 8.042517494139612e-06,
+      "loss": 1.43661633,
+      "memory(GiB)": 113.67,
+      "step": 25795,
+      "train_speed(iter/s)": 1.319314
+    },
+    {
+      "acc": 0.69061623,
+      "epoch": 0.654490106544901,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.041685290664806e-06,
+      "loss": 1.31261606,
+      "memory(GiB)": 113.67,
+      "step": 25800,
+      "train_speed(iter/s)": 1.319342
+    },
+    {
+      "acc": 0.68004861,
+      "epoch": 0.6546169457128361,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.0408529534031e-06,
+      "loss": 1.38931675,
+      "memory(GiB)": 113.67,
+      "step": 25805,
+      "train_speed(iter/s)": 1.31937
+    },
+    {
+      "acc": 0.68423967,
+      "epoch": 0.6547437848807712,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.040020482391105e-06,
+      "loss": 1.38302917,
+      "memory(GiB)": 113.67,
+      "step": 25810,
+      "train_speed(iter/s)": 1.319399
+    },
+    {
+      "acc": 0.71667981,
+      "epoch": 0.6548706240487062,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.039187877665435e-06,
+      "loss": 1.29383831,
+      "memory(GiB)": 113.67,
+      "step": 25815,
+      "train_speed(iter/s)": 1.319428
+    },
+    {
+      "acc": 0.69374237,
+      "epoch": 0.6549974632166413,
+      "grad_norm": 4.9375,
+      "learning_rate": 8.038355139262716e-06,
+      "loss": 1.33774834,
+      "memory(GiB)": 113.67,
+      "step": 25820,
+      "train_speed(iter/s)": 1.319457
+    },
+    {
+      "acc": 0.69662738,
+      "epoch": 0.6551243023845763,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.037522267219571e-06,
+      "loss": 1.36716242,
+      "memory(GiB)": 113.67,
+      "step": 25825,
+      "train_speed(iter/s)": 1.319484
+    },
+    {
+      "acc": 0.67623072,
+      "epoch": 0.6552511415525114,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.036689261572636e-06,
+      "loss": 1.38759689,
+      "memory(GiB)": 113.67,
+      "step": 25830,
+      "train_speed(iter/s)": 1.319512
+    },
+    {
+      "acc": 0.68002596,
+      "epoch": 0.6553779807204465,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.035856122358548e-06,
+      "loss": 1.42601185,
+      "memory(GiB)": 113.67,
+      "step": 25835,
+      "train_speed(iter/s)": 1.319541
+    },
+    {
+      "acc": 0.68814163,
+      "epoch": 0.6555048198883815,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.035022849613954e-06,
+      "loss": 1.41455784,
+      "memory(GiB)": 113.67,
+      "step": 25840,
+      "train_speed(iter/s)": 1.319567
+    },
+    {
+      "acc": 0.68532529,
+      "epoch": 0.6556316590563166,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.034189443375505e-06,
+      "loss": 1.43183441,
+      "memory(GiB)": 113.67,
+      "step": 25845,
+      "train_speed(iter/s)": 1.319596
+    },
+    {
+      "acc": 0.68479748,
+      "epoch": 0.6557584982242517,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.033355903679858e-06,
+      "loss": 1.34904089,
+      "memory(GiB)": 113.67,
+      "step": 25850,
+      "train_speed(iter/s)": 1.319625
+    },
+    {
+      "acc": 0.69320045,
+      "epoch": 0.6558853373921867,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.032522230563676e-06,
+      "loss": 1.3930151,
+      "memory(GiB)": 113.67,
+      "step": 25855,
+      "train_speed(iter/s)": 1.319655
+    },
+    {
+      "acc": 0.6849844,
+      "epoch": 0.6560121765601218,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.031688424063625e-06,
+      "loss": 1.37387133,
+      "memory(GiB)": 113.67,
+      "step": 25860,
+      "train_speed(iter/s)": 1.319682
+    },
+    {
+      "acc": 0.68697639,
+      "epoch": 0.6561390157280568,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.030854484216381e-06,
+      "loss": 1.40365677,
+      "memory(GiB)": 113.67,
+      "step": 25865,
+      "train_speed(iter/s)": 1.319712
+    },
+    {
+      "acc": 0.68771434,
+      "epoch": 0.6562658548959919,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.030020411058627e-06,
+      "loss": 1.43342323,
+      "memory(GiB)": 113.67,
+      "step": 25870,
+      "train_speed(iter/s)": 1.31974
+    },
+    {
+      "acc": 0.69218845,
+      "epoch": 0.656392694063927,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.029186204627049e-06,
+      "loss": 1.36124878,
+      "memory(GiB)": 113.67,
+      "step": 25875,
+      "train_speed(iter/s)": 1.319769
+    },
+    {
+      "acc": 0.67999954,
+      "epoch": 0.656519533231862,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.028351864958335e-06,
+      "loss": 1.38438797,
+      "memory(GiB)": 113.67,
+      "step": 25880,
+      "train_speed(iter/s)": 1.319796
+    },
+    {
+      "acc": 0.69727097,
+      "epoch": 0.6566463723997971,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.027517392089185e-06,
+      "loss": 1.34764519,
+      "memory(GiB)": 113.67,
+      "step": 25885,
+      "train_speed(iter/s)": 1.319825
+    },
+    {
+      "acc": 0.68657465,
+      "epoch": 0.6567732115677322,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.026682786056304e-06,
+      "loss": 1.41882305,
+      "memory(GiB)": 113.67,
+      "step": 25890,
+      "train_speed(iter/s)": 1.319855
+    },
+    {
+      "acc": 0.66944265,
+      "epoch": 0.6569000507356672,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.025848046896401e-06,
+      "loss": 1.46008253,
+      "memory(GiB)": 113.67,
+      "step": 25895,
+      "train_speed(iter/s)": 1.319884
+    },
+    {
+      "acc": 0.68206825,
+      "epoch": 0.6570268899036023,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.02501317464619e-06,
+      "loss": 1.38821592,
+      "memory(GiB)": 113.67,
+      "step": 25900,
+      "train_speed(iter/s)": 1.319913
+    },
+    {
+      "acc": 0.68777328,
+      "epoch": 0.6571537290715372,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.024178169342396e-06,
+      "loss": 1.43683786,
+      "memory(GiB)": 113.67,
+      "step": 25905,
+      "train_speed(iter/s)": 1.319943
+    },
+    {
+      "acc": 0.69077177,
+      "epoch": 0.6572805682394723,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.023343031021744e-06,
+      "loss": 1.39085636,
+      "memory(GiB)": 113.67,
+      "step": 25910,
+      "train_speed(iter/s)": 1.319967
+    },
+    {
+      "acc": 0.67673721,
+      "epoch": 0.6574074074074074,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.022507759720966e-06,
+      "loss": 1.38281059,
+      "memory(GiB)": 113.67,
+      "step": 25915,
+      "train_speed(iter/s)": 1.319996
+    },
+    {
+      "acc": 0.68315449,
+      "epoch": 0.6575342465753424,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.021672355476802e-06,
+      "loss": 1.44375277,
+      "memory(GiB)": 113.67,
+      "step": 25920,
+      "train_speed(iter/s)": 1.320022
+    },
+    {
+      "acc": 0.67704482,
+      "epoch": 0.6576610857432775,
+      "grad_norm": 3.0,
+      "learning_rate": 8.020836818325997e-06,
+      "loss": 1.3829855,
+      "memory(GiB)": 113.67,
+      "step": 25925,
+      "train_speed(iter/s)": 1.320052
+    },
+    {
+      "acc": 0.68536315,
+      "epoch": 0.6577879249112126,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.020001148305304e-06,
+      "loss": 1.37248096,
+      "memory(GiB)": 113.67,
+      "step": 25930,
+      "train_speed(iter/s)": 1.320081
+    },
+    {
+      "acc": 0.7017303,
+      "epoch": 0.6579147640791476,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.019165345451475e-06,
+      "loss": 1.34489918,
+      "memory(GiB)": 113.67,
+      "step": 25935,
+      "train_speed(iter/s)": 1.320109
+    },
+    {
+      "acc": 0.6932025,
+      "epoch": 0.6580416032470827,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.018329409801276e-06,
+      "loss": 1.36639252,
+      "memory(GiB)": 113.67,
+      "step": 25940,
+      "train_speed(iter/s)": 1.320138
+    },
+    {
+      "acc": 0.69286633,
+      "epoch": 0.6581684424150177,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.017493341391471e-06,
+      "loss": 1.35692577,
+      "memory(GiB)": 113.67,
+      "step": 25945,
+      "train_speed(iter/s)": 1.320167
+    },
+    {
+      "acc": 0.68473711,
+      "epoch": 0.6582952815829528,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.016657140258839e-06,
+      "loss": 1.37418747,
+      "memory(GiB)": 113.67,
+      "step": 25950,
+      "train_speed(iter/s)": 1.320195
+    },
+    {
+      "acc": 0.6889832,
+      "epoch": 0.6584221207508879,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.015820806440157e-06,
+      "loss": 1.41142349,
+      "memory(GiB)": 113.67,
+      "step": 25955,
+      "train_speed(iter/s)": 1.320224
+    },
+    {
+      "acc": 0.6625906,
+      "epoch": 0.6585489599188229,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.014984339972211e-06,
+      "loss": 1.41675949,
+      "memory(GiB)": 113.67,
+      "step": 25960,
+      "train_speed(iter/s)": 1.320253
+    },
+    {
+      "acc": 0.68945704,
+      "epoch": 0.658675799086758,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.014147740891793e-06,
+      "loss": 1.38457661,
+      "memory(GiB)": 113.67,
+      "step": 25965,
+      "train_speed(iter/s)": 1.320281
+    },
+    {
+      "acc": 0.69247518,
+      "epoch": 0.6588026382546931,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.0133110092357e-06,
+      "loss": 1.35874186,
+      "memory(GiB)": 113.67,
+      "step": 25970,
+      "train_speed(iter/s)": 1.320309
+    },
+    {
+      "acc": 0.68554406,
+      "epoch": 0.6589294774226281,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.012474145040737e-06,
+      "loss": 1.46238766,
+      "memory(GiB)": 113.67,
+      "step": 25975,
+      "train_speed(iter/s)": 1.320337
+    },
+    {
+      "acc": 0.68472824,
+      "epoch": 0.6590563165905632,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.01163714834371e-06,
+      "loss": 1.41644726,
+      "memory(GiB)": 113.67,
+      "step": 25980,
+      "train_speed(iter/s)": 1.320364
+    },
+    {
+      "acc": 0.68186851,
+      "epoch": 0.6591831557584982,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.010800019181433e-06,
+      "loss": 1.46289415,
+      "memory(GiB)": 113.67,
+      "step": 25985,
+      "train_speed(iter/s)": 1.320393
+    },
+    {
+      "acc": 0.67694044,
+      "epoch": 0.6593099949264333,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.009962757590732e-06,
+      "loss": 1.42731323,
+      "memory(GiB)": 113.67,
+      "step": 25990,
+      "train_speed(iter/s)": 1.320421
+    },
+    {
+      "acc": 0.69449306,
+      "epoch": 0.6594368340943684,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.00912536360843e-06,
+      "loss": 1.37586498,
+      "memory(GiB)": 113.67,
+      "step": 25995,
+      "train_speed(iter/s)": 1.32045
+    },
+    {
+      "acc": 0.68378425,
+      "epoch": 0.6595636732623034,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.008287837271359e-06,
+      "loss": 1.36254473,
+      "memory(GiB)": 113.67,
+      "step": 26000,
+      "train_speed(iter/s)": 1.320476
+    },
+    {
+      "epoch": 0.6595636732623034,
+      "eval_acc": 0.6741899982667059,
+      "eval_loss": 1.3614418506622314,
+      "eval_runtime": 69.5533,
+      "eval_samples_per_second": 91.584,
+      "eval_steps_per_second": 22.903,
+      "step": 26000
+    },
+    {
+      "acc": 0.696841,
+      "epoch": 0.6596905124302385,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.007450178616356e-06,
+      "loss": 1.34596815,
+      "memory(GiB)": 113.67,
+      "step": 26005,
+      "train_speed(iter/s)": 1.315017
+    },
+    {
+      "acc": 0.6830945,
+      "epoch": 0.6598173515981736,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.00661238768027e-06,
+      "loss": 1.37134085,
+      "memory(GiB)": 113.67,
+      "step": 26010,
+      "train_speed(iter/s)": 1.315045
+    },
+    {
+      "acc": 0.68978939,
+      "epoch": 0.6599441907661086,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.005774464499947e-06,
+      "loss": 1.3921978,
+      "memory(GiB)": 113.67,
+      "step": 26015,
+      "train_speed(iter/s)": 1.315075
+    },
+    {
+      "acc": 0.66860304,
+      "epoch": 0.6600710299340437,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.004936409112243e-06,
+      "loss": 1.42660828,
+      "memory(GiB)": 113.67,
+      "step": 26020,
+      "train_speed(iter/s)": 1.315103
+    },
+    {
+      "acc": 0.68835778,
+      "epoch": 0.6601978691019786,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.004098221554018e-06,
+      "loss": 1.3976634,
+      "memory(GiB)": 113.67,
+      "step": 26025,
+      "train_speed(iter/s)": 1.315132
+    },
+    {
+      "acc": 0.68974342,
+      "epoch": 0.6603247082699137,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.003259901862143e-06,
+      "loss": 1.36892262,
+      "memory(GiB)": 113.67,
+      "step": 26030,
+      "train_speed(iter/s)": 1.315161
+    },
+    {
+      "acc": 0.67698345,
+      "epoch": 0.6604515474378488,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.002421450073488e-06,
+      "loss": 1.36021309,
+      "memory(GiB)": 113.67,
+      "step": 26035,
+      "train_speed(iter/s)": 1.315189
+    },
+    {
+      "acc": 0.68674631,
+      "epoch": 0.6605783866057838,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.001582866224932e-06,
+      "loss": 1.33094521,
+      "memory(GiB)": 113.67,
+      "step": 26040,
+      "train_speed(iter/s)": 1.315219
+    },
+    {
+      "acc": 0.69529672,
+      "epoch": 0.6607052257737189,
+      "grad_norm": 3.25,
+      "learning_rate": 8.000744150353362e-06,
+      "loss": 1.41824741,
+      "memory(GiB)": 113.67,
+      "step": 26045,
+      "train_speed(iter/s)": 1.315248
+    },
+    {
+      "acc": 0.68832216,
+      "epoch": 0.660832064941654,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.999905302495667e-06,
+      "loss": 1.41106873,
+      "memory(GiB)": 113.67,
+      "step": 26050,
+      "train_speed(iter/s)": 1.315277
+    },
+    {
+      "acc": 0.68943291,
+      "epoch": 0.660958904109589,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.999066322688743e-06,
+      "loss": 1.38082743,
+      "memory(GiB)": 113.67,
+      "step": 26055,
+      "train_speed(iter/s)": 1.315308
+    },
+    {
+      "acc": 0.68466482,
+      "epoch": 0.6610857432775241,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.998227210969491e-06,
+      "loss": 1.45499792,
+      "memory(GiB)": 113.67,
+      "step": 26060,
+      "train_speed(iter/s)": 1.315332
+    },
+    {
+      "acc": 0.7062922,
+      "epoch": 0.6612125824454591,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.997387967374821e-06,
+      "loss": 1.34366379,
+      "memory(GiB)": 113.67,
+      "step": 26065,
+      "train_speed(iter/s)": 1.315361
+    },
+    {
+      "acc": 0.67884626,
+      "epoch": 0.6613394216133942,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.996548591941647e-06,
+      "loss": 1.47227459,
+      "memory(GiB)": 113.67,
+      "step": 26070,
+      "train_speed(iter/s)": 1.315391
+    },
+    {
+      "acc": 0.6871747,
+      "epoch": 0.6614662607813293,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.995709084706884e-06,
+      "loss": 1.35373011,
+      "memory(GiB)": 113.67,
+      "step": 26075,
+      "train_speed(iter/s)": 1.31542
+    },
+    {
+      "acc": 0.67762632,
+      "epoch": 0.6615930999492643,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.994869445707463e-06,
+      "loss": 1.41493769,
+      "memory(GiB)": 113.67,
+      "step": 26080,
+      "train_speed(iter/s)": 1.31545
+    },
+    {
+      "acc": 0.6827219,
+      "epoch": 0.6617199391171994,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.994029674980313e-06,
+      "loss": 1.38303385,
+      "memory(GiB)": 113.67,
+      "step": 26085,
+      "train_speed(iter/s)": 1.315479
+    },
+    {
+      "acc": 0.69229593,
+      "epoch": 0.6618467782851345,
+      "grad_norm": 2.875,
+      "learning_rate": 7.99318977256237e-06,
+      "loss": 1.37019958,
+      "memory(GiB)": 113.67,
+      "step": 26090,
+      "train_speed(iter/s)": 1.315508
+    },
+    {
+      "acc": 0.67888861,
+      "epoch": 0.6619736174530695,
+      "grad_norm": 2.390625,
+      "learning_rate": 7.992349738490576e-06,
+      "loss": 1.42495308,
+      "memory(GiB)": 113.67,
+      "step": 26095,
+      "train_speed(iter/s)": 1.315538
+    },
+    {
+      "acc": 0.70137787,
+      "epoch": 0.6621004566210046,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.991509572801883e-06,
+      "loss": 1.37298498,
+      "memory(GiB)": 113.67,
+      "step": 26100,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.68458881,
+      "epoch": 0.6622272957889396,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.990669275533241e-06,
+      "loss": 1.42746172,
+      "memory(GiB)": 113.67,
+      "step": 26105,
+      "train_speed(iter/s)": 1.315593
+    },
+    {
+      "acc": 0.67284746,
+      "epoch": 0.6623541349568747,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.989828846721613e-06,
+      "loss": 1.4524931,
+      "memory(GiB)": 113.67,
+      "step": 26110,
+      "train_speed(iter/s)": 1.315623
+    },
+    {
+      "acc": 0.68181686,
+      "epoch": 0.6624809741248098,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.98898828640396e-06,
+      "loss": 1.37787571,
+      "memory(GiB)": 113.67,
+      "step": 26115,
+      "train_speed(iter/s)": 1.315652
+    },
+    {
+      "acc": 0.68469405,
+      "epoch": 0.6626078132927448,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.988147594617262e-06,
+      "loss": 1.38800898,
+      "memory(GiB)": 113.67,
+      "step": 26120,
+      "train_speed(iter/s)": 1.315681
+    },
+    {
+      "acc": 0.68452382,
+      "epoch": 0.6627346524606799,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.987306771398489e-06,
+      "loss": 1.35509043,
+      "memory(GiB)": 113.67,
+      "step": 26125,
+      "train_speed(iter/s)": 1.315711
+    },
+    {
+      "acc": 0.69562106,
+      "epoch": 0.662861491628615,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.986465816784628e-06,
+      "loss": 1.401931,
+      "memory(GiB)": 113.67,
+      "step": 26130,
+      "train_speed(iter/s)": 1.31574
+    },
+    {
+      "acc": 0.68099012,
+      "epoch": 0.66298833079655,
+      "grad_norm": 3.25,
+      "learning_rate": 7.985624730812667e-06,
+      "loss": 1.40378551,
+      "memory(GiB)": 113.67,
+      "step": 26135,
+      "train_speed(iter/s)": 1.315769
+    },
+    {
+      "acc": 0.67209034,
+      "epoch": 0.663115169964485,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.984783513519601e-06,
+      "loss": 1.4227356,
+      "memory(GiB)": 113.67,
+      "step": 26140,
+      "train_speed(iter/s)": 1.315798
+    },
+    {
+      "acc": 0.6773653,
+      "epoch": 0.66324200913242,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.98394216494243e-06,
+      "loss": 1.48292675,
+      "memory(GiB)": 113.67,
+      "step": 26145,
+      "train_speed(iter/s)": 1.315827
+    },
+    {
+      "acc": 0.68382888,
+      "epoch": 0.6633688483003551,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.983100685118157e-06,
+      "loss": 1.42710171,
+      "memory(GiB)": 113.67,
+      "step": 26150,
+      "train_speed(iter/s)": 1.315854
+    },
+    {
+      "acc": 0.69662828,
+      "epoch": 0.6634956874682902,
+      "grad_norm": 2.75,
+      "learning_rate": 7.9822590740838e-06,
+      "loss": 1.30553846,
+      "memory(GiB)": 113.67,
+      "step": 26155,
+      "train_speed(iter/s)": 1.315884
+    },
+    {
+      "acc": 0.687287,
+      "epoch": 0.6636225266362252,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.981417331876373e-06,
+      "loss": 1.35112858,
+      "memory(GiB)": 113.67,
+      "step": 26160,
+      "train_speed(iter/s)": 1.315902
+    },
+    {
+      "acc": 0.68301601,
+      "epoch": 0.6637493658041603,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.980575458532901e-06,
+      "loss": 1.43239937,
+      "memory(GiB)": 113.67,
+      "step": 26165,
+      "train_speed(iter/s)": 1.315931
+    },
+    {
+      "acc": 0.6875443,
+      "epoch": 0.6638762049720954,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.979733454090415e-06,
+      "loss": 1.37955389,
+      "memory(GiB)": 113.67,
+      "step": 26170,
+      "train_speed(iter/s)": 1.315961
+    },
+    {
+      "acc": 0.67217298,
+      "epoch": 0.6640030441400304,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.978891318585947e-06,
+      "loss": 1.37970314,
+      "memory(GiB)": 113.67,
+      "step": 26175,
+      "train_speed(iter/s)": 1.315985
+    },
+    {
+      "acc": 0.6848341,
+      "epoch": 0.6641298833079655,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.978049052056537e-06,
+      "loss": 1.41987152,
+      "memory(GiB)": 113.67,
+      "step": 26180,
+      "train_speed(iter/s)": 1.316014
+    },
+    {
+      "acc": 0.68664188,
+      "epoch": 0.6642567224759005,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.977206654539235e-06,
+      "loss": 1.38368635,
+      "memory(GiB)": 113.67,
+      "step": 26185,
+      "train_speed(iter/s)": 1.316044
+    },
+    {
+      "acc": 0.67402773,
+      "epoch": 0.6643835616438356,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.976364126071092e-06,
+      "loss": 1.44510889,
+      "memory(GiB)": 113.67,
+      "step": 26190,
+      "train_speed(iter/s)": 1.316074
+    },
+    {
+      "acc": 0.66462374,
+      "epoch": 0.6645104008117707,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.975521466689166e-06,
+      "loss": 1.48895016,
+      "memory(GiB)": 113.67,
+      "step": 26195,
+      "train_speed(iter/s)": 1.316104
+    },
+    {
+      "acc": 0.68305712,
+      "epoch": 0.6646372399797057,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.974678676430523e-06,
+      "loss": 1.38333607,
+      "memory(GiB)": 113.67,
+      "step": 26200,
+      "train_speed(iter/s)": 1.316132
+    },
+    {
+      "acc": 0.6946197,
+      "epoch": 0.6647640791476408,
+      "grad_norm": 2.875,
+      "learning_rate": 7.97383575533223e-06,
+      "loss": 1.34419632,
+      "memory(GiB)": 113.67,
+      "step": 26205,
+      "train_speed(iter/s)": 1.316162
+    },
+    {
+      "acc": 0.67160387,
+      "epoch": 0.6648909183155759,
+      "grad_norm": 3.0,
+      "learning_rate": 7.972992703431362e-06,
+      "loss": 1.40588655,
+      "memory(GiB)": 113.67,
+      "step": 26210,
+      "train_speed(iter/s)": 1.316187
+    },
+    {
+      "acc": 0.68348694,
+      "epoch": 0.6650177574835109,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.972149520765e-06,
+      "loss": 1.40807562,
+      "memory(GiB)": 113.67,
+      "step": 26215,
+      "train_speed(iter/s)": 1.316217
+    },
+    {
+      "acc": 0.67543283,
+      "epoch": 0.665144596651446,
+      "grad_norm": 3.125,
+      "learning_rate": 7.971306207370236e-06,
+      "loss": 1.4011672,
+      "memory(GiB)": 113.67,
+      "step": 26220,
+      "train_speed(iter/s)": 1.316246
+    },
+    {
+      "acc": 0.68627768,
+      "epoch": 0.665271435819381,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.970462763284157e-06,
+      "loss": 1.3728611,
+      "memory(GiB)": 113.67,
+      "step": 26225,
+      "train_speed(iter/s)": 1.316275
+    },
+    {
+      "acc": 0.67705274,
+      "epoch": 0.6653982749873161,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.969619188543865e-06,
+      "loss": 1.45404072,
+      "memory(GiB)": 113.67,
+      "step": 26230,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.68444386,
+      "epoch": 0.6655251141552512,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.968775483186462e-06,
+      "loss": 1.3730957,
+      "memory(GiB)": 113.67,
+      "step": 26235,
+      "train_speed(iter/s)": 1.316332
+    },
+    {
+      "acc": 0.70140514,
+      "epoch": 0.6656519533231862,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.967931647249058e-06,
+      "loss": 1.33220215,
+      "memory(GiB)": 113.67,
+      "step": 26240,
+      "train_speed(iter/s)": 1.316361
+    },
+    {
+      "acc": 0.67349949,
+      "epoch": 0.6657787924911213,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.967087680768768e-06,
+      "loss": 1.39562426,
+      "memory(GiB)": 113.67,
+      "step": 26245,
+      "train_speed(iter/s)": 1.31639
+    },
+    {
+      "acc": 0.69007549,
+      "epoch": 0.6659056316590564,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.966243583782718e-06,
+      "loss": 1.41656427,
+      "memory(GiB)": 113.67,
+      "step": 26250,
+      "train_speed(iter/s)": 1.316419
+    },
+    {
+      "acc": 0.67969537,
+      "epoch": 0.6660324708269914,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.96539935632803e-06,
+      "loss": 1.43409424,
+      "memory(GiB)": 113.67,
+      "step": 26255,
+      "train_speed(iter/s)": 1.316448
+    },
+    {
+      "acc": 0.6837234,
+      "epoch": 0.6661593099949265,
+      "grad_norm": 3.125,
+      "learning_rate": 7.964554998441839e-06,
+      "loss": 1.39237547,
+      "memory(GiB)": 113.67,
+      "step": 26260,
+      "train_speed(iter/s)": 1.316477
+    },
+    {
+      "acc": 0.70365829,
+      "epoch": 0.6662861491628614,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.963710510161282e-06,
+      "loss": 1.33791466,
+      "memory(GiB)": 113.67,
+      "step": 26265,
+      "train_speed(iter/s)": 1.316505
+    },
+    {
+      "acc": 0.69514709,
+      "epoch": 0.6664129883307965,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.962865891523508e-06,
+      "loss": 1.33223991,
+      "memory(GiB)": 113.67,
+      "step": 26270,
+      "train_speed(iter/s)": 1.316532
+    },
+    {
+      "acc": 0.67939181,
+      "epoch": 0.6665398274987316,
+      "grad_norm": 3.25,
+      "learning_rate": 7.96202114256566e-06,
+      "loss": 1.48785686,
+      "memory(GiB)": 113.67,
+      "step": 26275,
+      "train_speed(iter/s)": 1.316561
+    },
+    {
+      "acc": 0.68629441,
+      "epoch": 0.6666666666666666,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.961176263324902e-06,
+      "loss": 1.37805176,
+      "memory(GiB)": 113.67,
+      "step": 26280,
+      "train_speed(iter/s)": 1.316589
+    },
+    {
+      "acc": 0.67798214,
+      "epoch": 0.6667935058346017,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.960331253838387e-06,
+      "loss": 1.38253536,
+      "memory(GiB)": 113.67,
+      "step": 26285,
+      "train_speed(iter/s)": 1.31661
+    },
+    {
+      "acc": 0.67579184,
+      "epoch": 0.6669203450025368,
+      "grad_norm": 2.625,
+      "learning_rate": 7.95948611414329e-06,
+      "loss": 1.37392826,
+      "memory(GiB)": 113.67,
+      "step": 26290,
+      "train_speed(iter/s)": 1.316639
+    },
+    {
+      "acc": 0.68503923,
+      "epoch": 0.6670471841704718,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.958640844276776e-06,
+      "loss": 1.43692417,
+      "memory(GiB)": 113.67,
+      "step": 26295,
+      "train_speed(iter/s)": 1.316667
+    },
+    {
+      "acc": 0.66945763,
+      "epoch": 0.6671740233384069,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.957795444276033e-06,
+      "loss": 1.44575968,
+      "memory(GiB)": 113.67,
+      "step": 26300,
+      "train_speed(iter/s)": 1.316695
+    },
+    {
+      "acc": 0.68460436,
+      "epoch": 0.6673008625063419,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.956949914178239e-06,
+      "loss": 1.33293314,
+      "memory(GiB)": 113.67,
+      "step": 26305,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.68749304,
+      "epoch": 0.667427701674277,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.956104254020587e-06,
+      "loss": 1.34113503,
+      "memory(GiB)": 113.67,
+      "step": 26310,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.68707075,
+      "epoch": 0.6675545408422121,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.95525846384027e-06,
+      "loss": 1.37161655,
+      "memory(GiB)": 113.67,
+      "step": 26315,
+      "train_speed(iter/s)": 1.316782
+    },
+    {
+      "acc": 0.68891611,
+      "epoch": 0.6676813800101471,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.954412543674493e-06,
+      "loss": 1.35752773,
+      "memory(GiB)": 113.67,
+      "step": 26320,
+      "train_speed(iter/s)": 1.316811
+    },
+    {
+      "acc": 0.68587093,
+      "epoch": 0.6678082191780822,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.95356649356046e-06,
+      "loss": 1.44123669,
+      "memory(GiB)": 113.67,
+      "step": 26325,
+      "train_speed(iter/s)": 1.31684
+    },
+    {
+      "acc": 0.67677402,
+      "epoch": 0.6679350583460173,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.952720313535387e-06,
+      "loss": 1.46146526,
+      "memory(GiB)": 113.67,
+      "step": 26330,
+      "train_speed(iter/s)": 1.316869
+    },
+    {
+      "acc": 0.67896285,
+      "epoch": 0.6680618975139523,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.951874003636492e-06,
+      "loss": 1.44514084,
+      "memory(GiB)": 113.67,
+      "step": 26335,
+      "train_speed(iter/s)": 1.316898
+    },
+    {
+      "acc": 0.68894849,
+      "epoch": 0.6681887366818874,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.951027563901e-06,
+      "loss": 1.30635386,
+      "memory(GiB)": 113.67,
+      "step": 26340,
+      "train_speed(iter/s)": 1.316927
+    },
+    {
+      "acc": 0.67817707,
+      "epoch": 0.6683155758498224,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.950180994366138e-06,
+      "loss": 1.42715282,
+      "memory(GiB)": 113.67,
+      "step": 26345,
+      "train_speed(iter/s)": 1.316956
+    },
+    {
+      "acc": 0.69988427,
+      "epoch": 0.6684424150177575,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.949334295069147e-06,
+      "loss": 1.31553288,
+      "memory(GiB)": 113.67,
+      "step": 26350,
+      "train_speed(iter/s)": 1.316985
+    },
+    {
+      "acc": 0.67631969,
+      "epoch": 0.6685692541856926,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.948487466047263e-06,
+      "loss": 1.45160246,
+      "memory(GiB)": 113.67,
+      "step": 26355,
+      "train_speed(iter/s)": 1.317014
+    },
+    {
+      "acc": 0.6791688,
+      "epoch": 0.6686960933536276,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.947640507337737e-06,
+      "loss": 1.38683319,
+      "memory(GiB)": 113.67,
+      "step": 26360,
+      "train_speed(iter/s)": 1.317044
+    },
+    {
+      "acc": 0.67545023,
+      "epoch": 0.6688229325215627,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.946793418977821e-06,
+      "loss": 1.40724144,
+      "memory(GiB)": 113.67,
+      "step": 26365,
+      "train_speed(iter/s)": 1.317071
+    },
+    {
+      "acc": 0.67605052,
+      "epoch": 0.6689497716894978,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.945946201004775e-06,
+      "loss": 1.44748135,
+      "memory(GiB)": 113.67,
+      "step": 26370,
+      "train_speed(iter/s)": 1.3171
+    },
+    {
+      "acc": 0.67741489,
+      "epoch": 0.6690766108574328,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.945098853455862e-06,
+      "loss": 1.42640514,
+      "memory(GiB)": 113.67,
+      "step": 26375,
+      "train_speed(iter/s)": 1.317129
+    },
+    {
+      "acc": 0.6970562,
+      "epoch": 0.6692034500253679,
+      "grad_norm": 2.484375,
+      "learning_rate": 7.944251376368352e-06,
+      "loss": 1.3431881,
+      "memory(GiB)": 113.67,
+      "step": 26380,
+      "train_speed(iter/s)": 1.317156
+    },
+    {
+      "acc": 0.68004265,
+      "epoch": 0.6693302891933028,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.943403769779523e-06,
+      "loss": 1.46383038,
+      "memory(GiB)": 113.67,
+      "step": 26385,
+      "train_speed(iter/s)": 1.317185
+    },
+    {
+      "acc": 0.68533325,
+      "epoch": 0.6694571283612379,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.942556033726654e-06,
+      "loss": 1.4562808,
+      "memory(GiB)": 113.67,
+      "step": 26390,
+      "train_speed(iter/s)": 1.317214
+    },
+    {
+      "acc": 0.69988232,
+      "epoch": 0.669583967529173,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.941708168247033e-06,
+      "loss": 1.39247055,
+      "memory(GiB)": 113.67,
+      "step": 26395,
+      "train_speed(iter/s)": 1.317243
+    },
+    {
+      "acc": 0.69363174,
+      "epoch": 0.669710806697108,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.940860173377952e-06,
+      "loss": 1.39257107,
+      "memory(GiB)": 113.67,
+      "step": 26400,
+      "train_speed(iter/s)": 1.317271
+    },
+    {
+      "acc": 0.6775269,
+      "epoch": 0.6698376458650431,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.940012049156711e-06,
+      "loss": 1.39736357,
+      "memory(GiB)": 113.67,
+      "step": 26405,
+      "train_speed(iter/s)": 1.317299
+    },
+    {
+      "acc": 0.69727144,
+      "epoch": 0.6699644850329782,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.939163795620614e-06,
+      "loss": 1.38379345,
+      "memory(GiB)": 113.67,
+      "step": 26410,
+      "train_speed(iter/s)": 1.317328
+    },
+    {
+      "acc": 0.68055511,
+      "epoch": 0.6700913242009132,
+      "grad_norm": 3.25,
+      "learning_rate": 7.938315412806971e-06,
+      "loss": 1.44055367,
+      "memory(GiB)": 113.67,
+      "step": 26415,
+      "train_speed(iter/s)": 1.317357
+    },
+    {
+      "acc": 0.69314375,
+      "epoch": 0.6702181633688483,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.937466900753098e-06,
+      "loss": 1.33299761,
+      "memory(GiB)": 113.67,
+      "step": 26420,
+      "train_speed(iter/s)": 1.317385
+    },
+    {
+      "acc": 0.68052959,
+      "epoch": 0.6703450025367833,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.936618259496316e-06,
+      "loss": 1.44390564,
+      "memory(GiB)": 113.67,
+      "step": 26425,
+      "train_speed(iter/s)": 1.317415
+    },
+    {
+      "acc": 0.69096813,
+      "epoch": 0.6704718417047184,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.935769489073952e-06,
+      "loss": 1.3624444,
+      "memory(GiB)": 113.67,
+      "step": 26430,
+      "train_speed(iter/s)": 1.317444
+    },
+    {
+      "acc": 0.69134798,
+      "epoch": 0.6705986808726535,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.934920589523336e-06,
+      "loss": 1.3121026,
+      "memory(GiB)": 113.67,
+      "step": 26435,
+      "train_speed(iter/s)": 1.317471
+    },
+    {
+      "acc": 0.68157358,
+      "epoch": 0.6707255200405885,
+      "grad_norm": 4.5,
+      "learning_rate": 7.934071560881812e-06,
+      "loss": 1.42416544,
+      "memory(GiB)": 113.67,
+      "step": 26440,
+      "train_speed(iter/s)": 1.317499
+    },
+    {
+      "acc": 0.6928071,
+      "epoch": 0.6708523592085236,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.93322240318672e-06,
+      "loss": 1.36629162,
+      "memory(GiB)": 113.67,
+      "step": 26445,
+      "train_speed(iter/s)": 1.317527
+    },
+    {
+      "acc": 0.68291101,
+      "epoch": 0.6709791983764587,
+      "grad_norm": 2.875,
+      "learning_rate": 7.93237311647541e-06,
+      "loss": 1.4469799,
+      "memory(GiB)": 113.67,
+      "step": 26450,
+      "train_speed(iter/s)": 1.317554
+    },
+    {
+      "acc": 0.69895811,
+      "epoch": 0.6711060375443937,
+      "grad_norm": 3.0,
+      "learning_rate": 7.93152370078524e-06,
+      "loss": 1.32403402,
+      "memory(GiB)": 113.67,
+      "step": 26455,
+      "train_speed(iter/s)": 1.317583
+    },
+    {
+      "acc": 0.68653135,
+      "epoch": 0.6712328767123288,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.930674156153569e-06,
+      "loss": 1.35829926,
+      "memory(GiB)": 113.67,
+      "step": 26460,
+      "train_speed(iter/s)": 1.317609
+    },
+    {
+      "acc": 0.68576221,
+      "epoch": 0.6713597158802638,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.929824482617763e-06,
+      "loss": 1.41101456,
+      "memory(GiB)": 113.67,
+      "step": 26465,
+      "train_speed(iter/s)": 1.317634
+    },
+    {
+      "acc": 0.67816682,
+      "epoch": 0.6714865550481989,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.928974680215196e-06,
+      "loss": 1.46626625,
+      "memory(GiB)": 113.67,
+      "step": 26470,
+      "train_speed(iter/s)": 1.31766
+    },
+    {
+      "acc": 0.67400908,
+      "epoch": 0.671613394216134,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.928124748983244e-06,
+      "loss": 1.44566212,
+      "memory(GiB)": 113.67,
+      "step": 26475,
+      "train_speed(iter/s)": 1.317688
+    },
+    {
+      "acc": 0.68282967,
+      "epoch": 0.671740233384069,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.927274688959294e-06,
+      "loss": 1.45645466,
+      "memory(GiB)": 113.67,
+      "step": 26480,
+      "train_speed(iter/s)": 1.317715
+    },
+    {
+      "acc": 0.68426743,
+      "epoch": 0.6718670725520041,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.926424500180734e-06,
+      "loss": 1.43352509,
+      "memory(GiB)": 113.67,
+      "step": 26485,
+      "train_speed(iter/s)": 1.317742
+    },
+    {
+      "acc": 0.71145554,
+      "epoch": 0.6719939117199392,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.92557418268496e-06,
+      "loss": 1.32366791,
+      "memory(GiB)": 113.67,
+      "step": 26490,
+      "train_speed(iter/s)": 1.317769
+    },
+    {
+      "acc": 0.69201798,
+      "epoch": 0.6721207508878742,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.92472373650937e-06,
+      "loss": 1.38339195,
+      "memory(GiB)": 113.67,
+      "step": 26495,
+      "train_speed(iter/s)": 1.317798
+    },
+    {
+      "acc": 0.66192265,
+      "epoch": 0.6722475900558093,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.923873161691373e-06,
+      "loss": 1.50213013,
+      "memory(GiB)": 113.67,
+      "step": 26500,
+      "train_speed(iter/s)": 1.317825
+    },
+    {
+      "acc": 0.70087442,
+      "epoch": 0.6723744292237442,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.923022458268379e-06,
+      "loss": 1.38610058,
+      "memory(GiB)": 113.67,
+      "step": 26505,
+      "train_speed(iter/s)": 1.317852
+    },
+    {
+      "acc": 0.69274383,
+      "epoch": 0.6725012683916793,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.922171626277809e-06,
+      "loss": 1.4007082,
+      "memory(GiB)": 113.67,
+      "step": 26510,
+      "train_speed(iter/s)": 1.317881
+    },
+    {
+      "acc": 0.67694359,
+      "epoch": 0.6726281075596144,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.921320665757081e-06,
+      "loss": 1.41518354,
+      "memory(GiB)": 113.67,
+      "step": 26515,
+      "train_speed(iter/s)": 1.317909
+    },
+    {
+      "acc": 0.6860425,
+      "epoch": 0.6727549467275494,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.920469576743631e-06,
+      "loss": 1.36679564,
+      "memory(GiB)": 113.67,
+      "step": 26520,
+      "train_speed(iter/s)": 1.317937
+    },
+    {
+      "acc": 0.69039679,
+      "epoch": 0.6728817858954845,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.919618359274888e-06,
+      "loss": 1.40632553,
+      "memory(GiB)": 113.67,
+      "step": 26525,
+      "train_speed(iter/s)": 1.317967
+    },
+    {
+      "acc": 0.68882713,
+      "epoch": 0.6730086250634196,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.918767013388295e-06,
+      "loss": 1.41944656,
+      "memory(GiB)": 113.67,
+      "step": 26530,
+      "train_speed(iter/s)": 1.317994
+    },
+    {
+      "acc": 0.67635241,
+      "epoch": 0.6731354642313546,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.917915539121297e-06,
+      "loss": 1.43547049,
+      "memory(GiB)": 113.67,
+      "step": 26535,
+      "train_speed(iter/s)": 1.318022
+    },
+    {
+      "acc": 0.66299639,
+      "epoch": 0.6732623033992897,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.917063936511347e-06,
+      "loss": 1.50489845,
+      "memory(GiB)": 113.67,
+      "step": 26540,
+      "train_speed(iter/s)": 1.31805
+    },
+    {
+      "acc": 0.6822032,
+      "epoch": 0.6733891425672247,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.9162122055959e-06,
+      "loss": 1.42110224,
+      "memory(GiB)": 113.67,
+      "step": 26545,
+      "train_speed(iter/s)": 1.318077
+    },
+    {
+      "acc": 0.68975296,
+      "epoch": 0.6735159817351598,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.91536034641242e-06,
+      "loss": 1.39221497,
+      "memory(GiB)": 113.67,
+      "step": 26550,
+      "train_speed(iter/s)": 1.318105
+    },
+    {
+      "acc": 0.67760096,
+      "epoch": 0.6736428209030949,
+      "grad_norm": 5.3125,
+      "learning_rate": 7.914508358998376e-06,
+      "loss": 1.45903854,
+      "memory(GiB)": 113.67,
+      "step": 26555,
+      "train_speed(iter/s)": 1.318132
+    },
+    {
+      "acc": 0.70683503,
+      "epoch": 0.6737696600710299,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.913656243391243e-06,
+      "loss": 1.33104544,
+      "memory(GiB)": 113.67,
+      "step": 26560,
+      "train_speed(iter/s)": 1.31816
+    },
+    {
+      "acc": 0.67897043,
+      "epoch": 0.673896499238965,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.9128039996285e-06,
+      "loss": 1.39062519,
+      "memory(GiB)": 113.67,
+      "step": 26565,
+      "train_speed(iter/s)": 1.318188
+    },
+    {
+      "acc": 0.67524433,
+      "epoch": 0.6740233384069001,
+      "grad_norm": 3.75,
+      "learning_rate": 7.911951627747633e-06,
+      "loss": 1.4245903,
+      "memory(GiB)": 113.67,
+      "step": 26570,
+      "train_speed(iter/s)": 1.318214
+    },
+    {
+      "acc": 0.67303228,
+      "epoch": 0.6741501775748351,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.91109912778613e-06,
+      "loss": 1.43396893,
+      "memory(GiB)": 113.67,
+      "step": 26575,
+      "train_speed(iter/s)": 1.318241
+    },
+    {
+      "acc": 0.68158779,
+      "epoch": 0.6742770167427702,
+      "grad_norm": 2.75,
+      "learning_rate": 7.910246499781492e-06,
+      "loss": 1.39726295,
+      "memory(GiB)": 113.67,
+      "step": 26580,
+      "train_speed(iter/s)": 1.318265
+    },
+    {
+      "acc": 0.67559266,
+      "epoch": 0.6744038559107052,
+      "grad_norm": 3.0,
+      "learning_rate": 7.90939374377122e-06,
+      "loss": 1.45035582,
+      "memory(GiB)": 113.67,
+      "step": 26585,
+      "train_speed(iter/s)": 1.318291
+    },
+    {
+      "acc": 0.69046159,
+      "epoch": 0.6745306950786403,
+      "grad_norm": 2.75,
+      "learning_rate": 7.908540859792821e-06,
+      "loss": 1.36215706,
+      "memory(GiB)": 113.67,
+      "step": 26590,
+      "train_speed(iter/s)": 1.318318
+    },
+    {
+      "acc": 0.68872709,
+      "epoch": 0.6746575342465754,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.907687847883809e-06,
+      "loss": 1.35515165,
+      "memory(GiB)": 113.67,
+      "step": 26595,
+      "train_speed(iter/s)": 1.318345
+    },
+    {
+      "acc": 0.69109516,
+      "epoch": 0.6747843734145104,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.906834708081703e-06,
+      "loss": 1.39247379,
+      "memory(GiB)": 113.67,
+      "step": 26600,
+      "train_speed(iter/s)": 1.318373
+    },
+    {
+      "acc": 0.68860936,
+      "epoch": 0.6749112125824455,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.90598144042403e-06,
+      "loss": 1.37141037,
+      "memory(GiB)": 113.67,
+      "step": 26605,
+      "train_speed(iter/s)": 1.318397
+    },
+    {
+      "acc": 0.68719416,
+      "epoch": 0.6750380517503806,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.905128044948318e-06,
+      "loss": 1.41348495,
+      "memory(GiB)": 113.67,
+      "step": 26610,
+      "train_speed(iter/s)": 1.318424
+    },
+    {
+      "acc": 0.68065944,
+      "epoch": 0.6751648909183156,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.904274521692104e-06,
+      "loss": 1.44219046,
+      "memory(GiB)": 113.67,
+      "step": 26615,
+      "train_speed(iter/s)": 1.318451
+    },
+    {
+      "acc": 0.68481827,
+      "epoch": 0.6752917300862507,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.90342087069293e-06,
+      "loss": 1.40160666,
+      "memory(GiB)": 113.67,
+      "step": 26620,
+      "train_speed(iter/s)": 1.318479
+    },
+    {
+      "acc": 0.69019413,
+      "epoch": 0.6754185692541856,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.902567091988343e-06,
+      "loss": 1.3583395,
+      "memory(GiB)": 113.67,
+      "step": 26625,
+      "train_speed(iter/s)": 1.318507
+    },
+    {
+      "acc": 0.69170051,
+      "epoch": 0.6755454084221207,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.901713185615898e-06,
+      "loss": 1.37296629,
+      "memory(GiB)": 113.67,
+      "step": 26630,
+      "train_speed(iter/s)": 1.318535
+    },
+    {
+      "acc": 0.68030901,
+      "epoch": 0.6756722475900558,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.90085915161315e-06,
+      "loss": 1.40830517,
+      "memory(GiB)": 113.67,
+      "step": 26635,
+      "train_speed(iter/s)": 1.318561
+    },
+    {
+      "acc": 0.68148365,
+      "epoch": 0.6757990867579908,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.900004990017667e-06,
+      "loss": 1.44384928,
+      "memory(GiB)": 113.67,
+      "step": 26640,
+      "train_speed(iter/s)": 1.318589
+    },
+    {
+      "acc": 0.66850214,
+      "epoch": 0.6759259259259259,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.899150700867014e-06,
+      "loss": 1.466395,
+      "memory(GiB)": 113.67,
+      "step": 26645,
+      "train_speed(iter/s)": 1.318618
+    },
+    {
+      "acc": 0.68088026,
+      "epoch": 0.676052765093861,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.898296284198772e-06,
+      "loss": 1.36240683,
+      "memory(GiB)": 113.67,
+      "step": 26650,
+      "train_speed(iter/s)": 1.318642
+    },
+    {
+      "acc": 0.68641319,
+      "epoch": 0.676179604261796,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.897441740050518e-06,
+      "loss": 1.44848881,
+      "memory(GiB)": 113.67,
+      "step": 26655,
+      "train_speed(iter/s)": 1.31867
+    },
+    {
+      "acc": 0.66335297,
+      "epoch": 0.6763064434297311,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.89658706845984e-06,
+      "loss": 1.4131258,
+      "memory(GiB)": 113.67,
+      "step": 26660,
+      "train_speed(iter/s)": 1.318698
+    },
+    {
+      "acc": 0.67946916,
+      "epoch": 0.6764332825976661,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.89573226946433e-06,
+      "loss": 1.40648193,
+      "memory(GiB)": 113.67,
+      "step": 26665,
+      "train_speed(iter/s)": 1.318725
+    },
+    {
+      "acc": 0.68543243,
+      "epoch": 0.6765601217656012,
+      "grad_norm": 3.84375,
+      "learning_rate": 7.89487734310159e-06,
+      "loss": 1.40964527,
+      "memory(GiB)": 113.67,
+      "step": 26670,
+      "train_speed(iter/s)": 1.318753
+    },
+    {
+      "acc": 0.68608508,
+      "epoch": 0.6766869609335363,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.894022289409216e-06,
+      "loss": 1.40717707,
+      "memory(GiB)": 113.67,
+      "step": 26675,
+      "train_speed(iter/s)": 1.318777
+    },
+    {
+      "acc": 0.67351151,
+      "epoch": 0.6768138001014713,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.893167108424822e-06,
+      "loss": 1.45163193,
+      "memory(GiB)": 113.67,
+      "step": 26680,
+      "train_speed(iter/s)": 1.318805
+    },
+    {
+      "acc": 0.6981884,
+      "epoch": 0.6769406392694064,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.89231180018602e-06,
+      "loss": 1.32999134,
+      "memory(GiB)": 113.67,
+      "step": 26685,
+      "train_speed(iter/s)": 1.318833
+    },
+    {
+      "acc": 0.6869596,
+      "epoch": 0.6770674784373415,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.891456364730434e-06,
+      "loss": 1.41594572,
+      "memory(GiB)": 113.67,
+      "step": 26690,
+      "train_speed(iter/s)": 1.318861
+    },
+    {
+      "acc": 0.68919373,
+      "epoch": 0.6771943176052765,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.890600802095686e-06,
+      "loss": 1.45474854,
+      "memory(GiB)": 113.67,
+      "step": 26695,
+      "train_speed(iter/s)": 1.318888
+    },
+    {
+      "acc": 0.68683338,
+      "epoch": 0.6773211567732116,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.889745112319411e-06,
+      "loss": 1.38557205,
+      "memory(GiB)": 113.67,
+      "step": 26700,
+      "train_speed(iter/s)": 1.318916
+    },
+    {
+      "acc": 0.68487291,
+      "epoch": 0.6774479959411466,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.888889295439244e-06,
+      "loss": 1.40297432,
+      "memory(GiB)": 113.67,
+      "step": 26705,
+      "train_speed(iter/s)": 1.318943
+    },
+    {
+      "acc": 0.68531771,
+      "epoch": 0.6775748351090817,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.888033351492827e-06,
+      "loss": 1.42926197,
+      "memory(GiB)": 113.67,
+      "step": 26710,
+      "train_speed(iter/s)": 1.318969
+    },
+    {
+      "acc": 0.67020717,
+      "epoch": 0.6777016742770168,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.887177280517808e-06,
+      "loss": 1.49316082,
+      "memory(GiB)": 113.67,
+      "step": 26715,
+      "train_speed(iter/s)": 1.318996
+    },
+    {
+      "acc": 0.69120002,
+      "epoch": 0.6778285134449518,
+      "grad_norm": 3.125,
+      "learning_rate": 7.886321082551845e-06,
+      "loss": 1.32387657,
+      "memory(GiB)": 113.67,
+      "step": 26720,
+      "train_speed(iter/s)": 1.319024
+    },
+    {
+      "acc": 0.70096331,
+      "epoch": 0.6779553526128869,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.88546475763259e-06,
+      "loss": 1.33098736,
+      "memory(GiB)": 113.67,
+      "step": 26725,
+      "train_speed(iter/s)": 1.31905
+    },
+    {
+      "acc": 0.69469166,
+      "epoch": 0.678082191780822,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.884608305797716e-06,
+      "loss": 1.34889421,
+      "memory(GiB)": 113.67,
+      "step": 26730,
+      "train_speed(iter/s)": 1.319076
+    },
+    {
+      "acc": 0.67750406,
+      "epoch": 0.678209030948757,
+      "grad_norm": 2.625,
+      "learning_rate": 7.883751727084888e-06,
+      "loss": 1.45119486,
+      "memory(GiB)": 113.67,
+      "step": 26735,
+      "train_speed(iter/s)": 1.319102
+    },
+    {
+      "acc": 0.68540897,
+      "epoch": 0.678335870116692,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.882895021531784e-06,
+      "loss": 1.43862305,
+      "memory(GiB)": 113.67,
+      "step": 26740,
+      "train_speed(iter/s)": 1.31913
+    },
+    {
+      "acc": 0.6941576,
+      "epoch": 0.678462709284627,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.882038189176085e-06,
+      "loss": 1.36235676,
+      "memory(GiB)": 113.67,
+      "step": 26745,
+      "train_speed(iter/s)": 1.319157
+    },
+    {
+      "acc": 0.6907464,
+      "epoch": 0.6785895484525621,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.881181230055481e-06,
+      "loss": 1.38792973,
+      "memory(GiB)": 113.67,
+      "step": 26750,
+      "train_speed(iter/s)": 1.319183
+    },
+    {
+      "acc": 0.69292564,
+      "epoch": 0.6787163876204972,
+      "grad_norm": 6.5,
+      "learning_rate": 7.880324144207663e-06,
+      "loss": 1.39687767,
+      "memory(GiB)": 113.67,
+      "step": 26755,
+      "train_speed(iter/s)": 1.31921
+    },
+    {
+      "acc": 0.69913816,
+      "epoch": 0.6788432267884322,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.879466931670328e-06,
+      "loss": 1.3710001,
+      "memory(GiB)": 113.67,
+      "step": 26760,
+      "train_speed(iter/s)": 1.319237
+    },
+    {
+      "acc": 0.6720396,
+      "epoch": 0.6789700659563673,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.878609592481182e-06,
+      "loss": 1.49979458,
+      "memory(GiB)": 113.67,
+      "step": 26765,
+      "train_speed(iter/s)": 1.319265
+    },
+    {
+      "acc": 0.69392776,
+      "epoch": 0.6790969051243024,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.877752126677933e-06,
+      "loss": 1.40278463,
+      "memory(GiB)": 113.67,
+      "step": 26770,
+      "train_speed(iter/s)": 1.319291
+    },
+    {
+      "acc": 0.70663419,
+      "epoch": 0.6792237442922374,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.876894534298298e-06,
+      "loss": 1.2996603,
+      "memory(GiB)": 113.67,
+      "step": 26775,
+      "train_speed(iter/s)": 1.319318
+    },
+    {
+      "acc": 0.69116802,
+      "epoch": 0.6793505834601725,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.87603681538e-06,
+      "loss": 1.34290504,
+      "memory(GiB)": 113.67,
+      "step": 26780,
+      "train_speed(iter/s)": 1.319346
+    },
+    {
+      "acc": 0.67016921,
+      "epoch": 0.6794774226281075,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.875178969960757e-06,
+      "loss": 1.49748936,
+      "memory(GiB)": 113.67,
+      "step": 26785,
+      "train_speed(iter/s)": 1.319372
+    },
+    {
+      "acc": 0.67745886,
+      "epoch": 0.6796042617960426,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.87432099807831e-06,
+      "loss": 1.43692703,
+      "memory(GiB)": 113.67,
+      "step": 26790,
+      "train_speed(iter/s)": 1.319399
+    },
+    {
+      "acc": 0.68911576,
+      "epoch": 0.6797311009639777,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.87346289977039e-06,
+      "loss": 1.40401783,
+      "memory(GiB)": 113.67,
+      "step": 26795,
+      "train_speed(iter/s)": 1.319425
+    },
+    {
+      "acc": 0.6903017,
+      "epoch": 0.6798579401319127,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.872604675074745e-06,
+      "loss": 1.37405529,
+      "memory(GiB)": 113.67,
+      "step": 26800,
+      "train_speed(iter/s)": 1.319451
+    },
+    {
+      "acc": 0.680972,
+      "epoch": 0.6799847792998478,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.871746324029119e-06,
+      "loss": 1.37658663,
+      "memory(GiB)": 113.67,
+      "step": 26805,
+      "train_speed(iter/s)": 1.319479
+    },
+    {
+      "acc": 0.6896975,
+      "epoch": 0.6801116184677829,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.87088784667127e-06,
+      "loss": 1.36240635,
+      "memory(GiB)": 113.67,
+      "step": 26810,
+      "train_speed(iter/s)": 1.319505
+    },
+    {
+      "acc": 0.66453795,
+      "epoch": 0.6802384576357179,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.870029243038955e-06,
+      "loss": 1.49538202,
+      "memory(GiB)": 113.67,
+      "step": 26815,
+      "train_speed(iter/s)": 1.319533
+    },
+    {
+      "acc": 0.6741065,
+      "epoch": 0.680365296803653,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.869170513169941e-06,
+      "loss": 1.43487978,
+      "memory(GiB)": 113.67,
+      "step": 26820,
+      "train_speed(iter/s)": 1.319559
+    },
+    {
+      "acc": 0.65936403,
+      "epoch": 0.680492135971588,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.868311657101996e-06,
+      "loss": 1.41271477,
+      "memory(GiB)": 113.67,
+      "step": 26825,
+      "train_speed(iter/s)": 1.319585
+    },
+    {
+      "acc": 0.68201742,
+      "epoch": 0.6806189751395231,
+      "grad_norm": 4.125,
+      "learning_rate": 7.8674526748729e-06,
+      "loss": 1.42480898,
+      "memory(GiB)": 113.67,
+      "step": 26830,
+      "train_speed(iter/s)": 1.319612
+    },
+    {
+      "acc": 0.70735397,
+      "epoch": 0.6807458143074582,
+      "grad_norm": 4.25,
+      "learning_rate": 7.866593566520432e-06,
+      "loss": 1.29308634,
+      "memory(GiB)": 113.67,
+      "step": 26835,
+      "train_speed(iter/s)": 1.31964
+    },
+    {
+      "acc": 0.67841253,
+      "epoch": 0.6808726534753932,
+      "grad_norm": 2.875,
+      "learning_rate": 7.865734332082382e-06,
+      "loss": 1.48537006,
+      "memory(GiB)": 113.67,
+      "step": 26840,
+      "train_speed(iter/s)": 1.319667
+    },
+    {
+      "acc": 0.69701123,
+      "epoch": 0.6809994926433283,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.86487497159654e-06,
+      "loss": 1.32390871,
+      "memory(GiB)": 113.67,
+      "step": 26845,
+      "train_speed(iter/s)": 1.319691
+    },
+    {
+      "acc": 0.683285,
+      "epoch": 0.6811263318112634,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.864015485100706e-06,
+      "loss": 1.39941149,
+      "memory(GiB)": 113.67,
+      "step": 26850,
+      "train_speed(iter/s)": 1.319717
+    },
+    {
+      "acc": 0.69207296,
+      "epoch": 0.6812531709791984,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.863155872632685e-06,
+      "loss": 1.32705355,
+      "memory(GiB)": 113.67,
+      "step": 26855,
+      "train_speed(iter/s)": 1.319743
+    },
+    {
+      "acc": 0.68699312,
+      "epoch": 0.6813800101471335,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.862296134230287e-06,
+      "loss": 1.35777435,
+      "memory(GiB)": 113.67,
+      "step": 26860,
+      "train_speed(iter/s)": 1.319771
+    },
+    {
+      "acc": 0.68652506,
+      "epoch": 0.6815068493150684,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.861436269931322e-06,
+      "loss": 1.37596836,
+      "memory(GiB)": 113.67,
+      "step": 26865,
+      "train_speed(iter/s)": 1.319796
+    },
+    {
+      "acc": 0.68752842,
+      "epoch": 0.6816336884830035,
+      "grad_norm": 2.875,
+      "learning_rate": 7.860576279773617e-06,
+      "loss": 1.33280125,
+      "memory(GiB)": 113.67,
+      "step": 26870,
+      "train_speed(iter/s)": 1.319823
+    },
+    {
+      "acc": 0.69249763,
+      "epoch": 0.6817605276509386,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.859716163794995e-06,
+      "loss": 1.36124401,
+      "memory(GiB)": 113.67,
+      "step": 26875,
+      "train_speed(iter/s)": 1.31985
+    },
+    {
+      "acc": 0.68496218,
+      "epoch": 0.6818873668188736,
+      "grad_norm": 2.34375,
+      "learning_rate": 7.858855922033289e-06,
+      "loss": 1.40563393,
+      "memory(GiB)": 113.67,
+      "step": 26880,
+      "train_speed(iter/s)": 1.319875
+    },
+    {
+      "acc": 0.6809298,
+      "epoch": 0.6820142059868087,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.857995554526334e-06,
+      "loss": 1.46469736,
+      "memory(GiB)": 113.67,
+      "step": 26885,
+      "train_speed(iter/s)": 1.319898
+    },
+    {
+      "acc": 0.67645998,
+      "epoch": 0.6821410451547438,
+      "grad_norm": 2.75,
+      "learning_rate": 7.857135061311977e-06,
+      "loss": 1.41992159,
+      "memory(GiB)": 113.67,
+      "step": 26890,
+      "train_speed(iter/s)": 1.319924
+    },
+    {
+      "acc": 0.67981358,
+      "epoch": 0.6822678843226788,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.856274442428062e-06,
+      "loss": 1.44264803,
+      "memory(GiB)": 113.67,
+      "step": 26895,
+      "train_speed(iter/s)": 1.31995
+    },
+    {
+      "acc": 0.68874359,
+      "epoch": 0.6823947234906139,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.855413697912446e-06,
+      "loss": 1.3927248,
+      "memory(GiB)": 113.67,
+      "step": 26900,
+      "train_speed(iter/s)": 1.319978
+    },
+    {
+      "acc": 0.68444014,
+      "epoch": 0.6825215626585489,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.854552827802987e-06,
+      "loss": 1.41515684,
+      "memory(GiB)": 113.67,
+      "step": 26905,
+      "train_speed(iter/s)": 1.320004
+    },
+    {
+      "acc": 0.69356232,
+      "epoch": 0.682648401826484,
+      "grad_norm": 6.625,
+      "learning_rate": 7.853691832137547e-06,
+      "loss": 1.44916821,
+      "memory(GiB)": 113.67,
+      "step": 26910,
+      "train_speed(iter/s)": 1.320032
+    },
+    {
+      "acc": 0.69863968,
+      "epoch": 0.6827752409944191,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.852830710954003e-06,
+      "loss": 1.32863827,
+      "memory(GiB)": 113.67,
+      "step": 26915,
+      "train_speed(iter/s)": 1.320058
+    },
+    {
+      "acc": 0.68238583,
+      "epoch": 0.6829020801623541,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.851969464290226e-06,
+      "loss": 1.41715622,
+      "memory(GiB)": 113.67,
+      "step": 26920,
+      "train_speed(iter/s)": 1.320084
+    },
+    {
+      "acc": 0.69021106,
+      "epoch": 0.6830289193302892,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.851108092184099e-06,
+      "loss": 1.37563572,
+      "memory(GiB)": 113.67,
+      "step": 26925,
+      "train_speed(iter/s)": 1.320112
+    },
+    {
+      "acc": 0.68630323,
+      "epoch": 0.6831557584982243,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.850246594673508e-06,
+      "loss": 1.39042816,
+      "memory(GiB)": 113.67,
+      "step": 26930,
+      "train_speed(iter/s)": 1.320139
+    },
+    {
+      "acc": 0.67108364,
+      "epoch": 0.6832825976661593,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.849384971796346e-06,
+      "loss": 1.48234024,
+      "memory(GiB)": 113.67,
+      "step": 26935,
+      "train_speed(iter/s)": 1.320166
+    },
+    {
+      "acc": 0.68774152,
+      "epoch": 0.6834094368340944,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.848523223590514e-06,
+      "loss": 1.38472881,
+      "memory(GiB)": 113.67,
+      "step": 26940,
+      "train_speed(iter/s)": 1.320192
+    },
+    {
+      "acc": 0.67525325,
+      "epoch": 0.6835362760020294,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.84766135009391e-06,
+      "loss": 1.47230778,
+      "memory(GiB)": 113.67,
+      "step": 26945,
+      "train_speed(iter/s)": 1.319885
+    },
+    {
+      "acc": 0.68116684,
+      "epoch": 0.6836631151699645,
+      "grad_norm": 2.625,
+      "learning_rate": 7.846799351344447e-06,
+      "loss": 1.42406044,
+      "memory(GiB)": 113.67,
+      "step": 26950,
+      "train_speed(iter/s)": 1.31991
+    },
+    {
+      "acc": 0.69999623,
+      "epoch": 0.6837899543378996,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.845937227380038e-06,
+      "loss": 1.38577461,
+      "memory(GiB)": 113.67,
+      "step": 26955,
+      "train_speed(iter/s)": 1.319938
+    },
+    {
+      "acc": 0.67445812,
+      "epoch": 0.6839167935058346,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.845074978238604e-06,
+      "loss": 1.4651227,
+      "memory(GiB)": 113.67,
+      "step": 26960,
+      "train_speed(iter/s)": 1.319963
+    },
+    {
+      "acc": 0.68602133,
+      "epoch": 0.6840436326737697,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.84421260395807e-06,
+      "loss": 1.40910778,
+      "memory(GiB)": 113.67,
+      "step": 26965,
+      "train_speed(iter/s)": 1.319991
+    },
+    {
+      "acc": 0.68945761,
+      "epoch": 0.6841704718417048,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.84335010457637e-06,
+      "loss": 1.40359802,
+      "memory(GiB)": 113.67,
+      "step": 26970,
+      "train_speed(iter/s)": 1.320017
+    },
+    {
+      "acc": 0.70355511,
+      "epoch": 0.6842973110096398,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.842487480131435e-06,
+      "loss": 1.29976196,
+      "memory(GiB)": 113.67,
+      "step": 26975,
+      "train_speed(iter/s)": 1.320044
+    },
+    {
+      "acc": 0.67867374,
+      "epoch": 0.6844241501775749,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.84162473066121e-06,
+      "loss": 1.48192806,
+      "memory(GiB)": 113.67,
+      "step": 26980,
+      "train_speed(iter/s)": 1.320066
+    },
+    {
+      "acc": 0.68763514,
+      "epoch": 0.6845509893455098,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.840761856203642e-06,
+      "loss": 1.2921628,
+      "memory(GiB)": 113.67,
+      "step": 26985,
+      "train_speed(iter/s)": 1.320091
+    },
+    {
+      "acc": 0.67058887,
+      "epoch": 0.6846778285134449,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.839898856796685e-06,
+      "loss": 1.49018917,
+      "memory(GiB)": 113.67,
+      "step": 26990,
+      "train_speed(iter/s)": 1.320116
+    },
+    {
+      "acc": 0.67486672,
+      "epoch": 0.68480466768138,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.839035732478297e-06,
+      "loss": 1.41821661,
+      "memory(GiB)": 113.67,
+      "step": 26995,
+      "train_speed(iter/s)": 1.32014
+    },
+    {
+      "acc": 0.68492212,
+      "epoch": 0.684931506849315,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.838172483286441e-06,
+      "loss": 1.46099834,
+      "memory(GiB)": 113.67,
+      "step": 27000,
+      "train_speed(iter/s)": 1.320166
+    },
+    {
+      "epoch": 0.684931506849315,
+      "eval_acc": 0.674299843168211,
+      "eval_loss": 1.360476016998291,
+      "eval_runtime": 70.6133,
+      "eval_samples_per_second": 90.21,
+      "eval_steps_per_second": 22.559,
+      "step": 27000
+    },
+    {
+      "acc": 0.70208478,
+      "epoch": 0.6850583460172501,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.83730910925909e-06,
+      "loss": 1.34316463,
+      "memory(GiB)": 113.67,
+      "step": 27005,
+      "train_speed(iter/s)": 1.314849
+    },
+    {
+      "acc": 0.68146415,
+      "epoch": 0.6851851851851852,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.836445610434215e-06,
+      "loss": 1.38639641,
+      "memory(GiB)": 113.67,
+      "step": 27010,
+      "train_speed(iter/s)": 1.314877
+    },
+    {
+      "acc": 0.68383417,
+      "epoch": 0.6853120243531202,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.835581986849799e-06,
+      "loss": 1.41086349,
+      "memory(GiB)": 113.67,
+      "step": 27015,
+      "train_speed(iter/s)": 1.314906
+    },
+    {
+      "acc": 0.68411551,
+      "epoch": 0.6854388635210553,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.834718238543827e-06,
+      "loss": 1.3645792,
+      "memory(GiB)": 113.67,
+      "step": 27020,
+      "train_speed(iter/s)": 1.314935
+    },
+    {
+      "acc": 0.66778655,
+      "epoch": 0.6855657026889903,
+      "grad_norm": 2.75,
+      "learning_rate": 7.833854365554289e-06,
+      "loss": 1.44216728,
+      "memory(GiB)": 113.67,
+      "step": 27025,
+      "train_speed(iter/s)": 1.314965
+    },
+    {
+      "acc": 0.67338667,
+      "epoch": 0.6856925418569254,
+      "grad_norm": 2.625,
+      "learning_rate": 7.832990367919186e-06,
+      "loss": 1.45363255,
+      "memory(GiB)": 113.67,
+      "step": 27030,
+      "train_speed(iter/s)": 1.314994
+    },
+    {
+      "acc": 0.69249544,
+      "epoch": 0.6858193810248605,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.832126245676518e-06,
+      "loss": 1.37738237,
+      "memory(GiB)": 113.67,
+      "step": 27035,
+      "train_speed(iter/s)": 1.315011
+    },
+    {
+      "acc": 0.68728075,
+      "epoch": 0.6859462201927955,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.831261998864293e-06,
+      "loss": 1.38587685,
+      "memory(GiB)": 113.67,
+      "step": 27040,
+      "train_speed(iter/s)": 1.315041
+    },
+    {
+      "acc": 0.68888783,
+      "epoch": 0.6860730593607306,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.830397627520526e-06,
+      "loss": 1.44317856,
+      "memory(GiB)": 113.67,
+      "step": 27045,
+      "train_speed(iter/s)": 1.315071
+    },
+    {
+      "acc": 0.67883668,
+      "epoch": 0.6861998985286657,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.82953313168323e-06,
+      "loss": 1.41792431,
+      "memory(GiB)": 113.67,
+      "step": 27050,
+      "train_speed(iter/s)": 1.315101
+    },
+    {
+      "acc": 0.67996159,
+      "epoch": 0.6863267376966007,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.828668511390439e-06,
+      "loss": 1.37045231,
+      "memory(GiB)": 113.67,
+      "step": 27055,
+      "train_speed(iter/s)": 1.31513
+    },
+    {
+      "acc": 0.67630696,
+      "epoch": 0.6864535768645358,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.827803766680176e-06,
+      "loss": 1.44309797,
+      "memory(GiB)": 113.67,
+      "step": 27060,
+      "train_speed(iter/s)": 1.315156
+    },
+    {
+      "acc": 0.68406806,
+      "epoch": 0.6865804160324708,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.826938897590477e-06,
+      "loss": 1.43148823,
+      "memory(GiB)": 113.67,
+      "step": 27065,
+      "train_speed(iter/s)": 1.315185
+    },
+    {
+      "acc": 0.68902478,
+      "epoch": 0.6867072552004059,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.826073904159384e-06,
+      "loss": 1.32233677,
+      "memory(GiB)": 113.67,
+      "step": 27070,
+      "train_speed(iter/s)": 1.315214
+    },
+    {
+      "acc": 0.69113626,
+      "epoch": 0.686834094368341,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.825208786424944e-06,
+      "loss": 1.41261911,
+      "memory(GiB)": 113.67,
+      "step": 27075,
+      "train_speed(iter/s)": 1.314873
+    },
+    {
+      "acc": 0.67246113,
+      "epoch": 0.686960933536276,
+      "grad_norm": 5.15625,
+      "learning_rate": 7.824343544425207e-06,
+      "loss": 1.47670612,
+      "memory(GiB)": 113.67,
+      "step": 27080,
+      "train_speed(iter/s)": 1.314904
+    },
+    {
+      "acc": 0.68221068,
+      "epoch": 0.6870877727042111,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.823478178198234e-06,
+      "loss": 1.39418564,
+      "memory(GiB)": 113.67,
+      "step": 27085,
+      "train_speed(iter/s)": 1.314933
+    },
+    {
+      "acc": 0.69342556,
+      "epoch": 0.6872146118721462,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.822612687782083e-06,
+      "loss": 1.33685207,
+      "memory(GiB)": 113.67,
+      "step": 27090,
+      "train_speed(iter/s)": 1.314962
+    },
+    {
+      "acc": 0.69389429,
+      "epoch": 0.6873414510400812,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.821747073214823e-06,
+      "loss": 1.3466608,
+      "memory(GiB)": 113.67,
+      "step": 27095,
+      "train_speed(iter/s)": 1.314987
+    },
+    {
+      "acc": 0.68565655,
+      "epoch": 0.6874682902080163,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.820881334534529e-06,
+      "loss": 1.34972649,
+      "memory(GiB)": 113.67,
+      "step": 27100,
+      "train_speed(iter/s)": 1.315016
+    },
+    {
+      "acc": 0.70105524,
+      "epoch": 0.6875951293759512,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.820015471779278e-06,
+      "loss": 1.30702896,
+      "memory(GiB)": 113.67,
+      "step": 27105,
+      "train_speed(iter/s)": 1.315044
+    },
+    {
+      "acc": 0.69614258,
+      "epoch": 0.6877219685438863,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.819149484987159e-06,
+      "loss": 1.35945206,
+      "memory(GiB)": 113.67,
+      "step": 27110,
+      "train_speed(iter/s)": 1.315074
+    },
+    {
+      "acc": 0.69240007,
+      "epoch": 0.6878488077118214,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.818283374196259e-06,
+      "loss": 1.34474859,
+      "memory(GiB)": 113.67,
+      "step": 27115,
+      "train_speed(iter/s)": 1.315104
+    },
+    {
+      "acc": 0.69008107,
+      "epoch": 0.6879756468797564,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.817417139444671e-06,
+      "loss": 1.45705547,
+      "memory(GiB)": 113.67,
+      "step": 27120,
+      "train_speed(iter/s)": 1.315135
+    },
+    {
+      "acc": 0.67521367,
+      "epoch": 0.6881024860476915,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.8165507807705e-06,
+      "loss": 1.5087307,
+      "memory(GiB)": 113.67,
+      "step": 27125,
+      "train_speed(iter/s)": 1.315166
+    },
+    {
+      "acc": 0.67750773,
+      "epoch": 0.6882293252156266,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.81568429821185e-06,
+      "loss": 1.45560379,
+      "memory(GiB)": 113.67,
+      "step": 27130,
+      "train_speed(iter/s)": 1.315196
+    },
+    {
+      "acc": 0.68562026,
+      "epoch": 0.6883561643835616,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.814817691806834e-06,
+      "loss": 1.43221207,
+      "memory(GiB)": 113.67,
+      "step": 27135,
+      "train_speed(iter/s)": 1.315225
+    },
+    {
+      "acc": 0.6861968,
+      "epoch": 0.6884830035514967,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.813950961593569e-06,
+      "loss": 1.36927109,
+      "memory(GiB)": 113.67,
+      "step": 27140,
+      "train_speed(iter/s)": 1.315254
+    },
+    {
+      "acc": 0.6824388,
+      "epoch": 0.6886098427194317,
+      "grad_norm": 2.421875,
+      "learning_rate": 7.813084107610175e-06,
+      "loss": 1.40511456,
+      "memory(GiB)": 113.67,
+      "step": 27145,
+      "train_speed(iter/s)": 1.31528
+    },
+    {
+      "acc": 0.69114914,
+      "epoch": 0.6887366818873668,
+      "grad_norm": 2.875,
+      "learning_rate": 7.812217129894785e-06,
+      "loss": 1.28347816,
+      "memory(GiB)": 113.67,
+      "step": 27150,
+      "train_speed(iter/s)": 1.315308
+    },
+    {
+      "acc": 0.69900331,
+      "epoch": 0.6888635210553019,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.811350028485531e-06,
+      "loss": 1.34795961,
+      "memory(GiB)": 113.67,
+      "step": 27155,
+      "train_speed(iter/s)": 1.315335
+    },
+    {
+      "acc": 0.68836241,
+      "epoch": 0.6889903602232369,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.810482803420549e-06,
+      "loss": 1.33470325,
+      "memory(GiB)": 113.67,
+      "step": 27160,
+      "train_speed(iter/s)": 1.31536
+    },
+    {
+      "acc": 0.68412619,
+      "epoch": 0.689117199391172,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.809615454737984e-06,
+      "loss": 1.43586931,
+      "memory(GiB)": 113.67,
+      "step": 27165,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.67814312,
+      "epoch": 0.6892440385591071,
+      "grad_norm": 3.125,
+      "learning_rate": 7.808747982475991e-06,
+      "loss": 1.43199825,
+      "memory(GiB)": 113.67,
+      "step": 27170,
+      "train_speed(iter/s)": 1.315419
+    },
+    {
+      "acc": 0.67901039,
+      "epoch": 0.6893708777270421,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.807880386672718e-06,
+      "loss": 1.41145029,
+      "memory(GiB)": 113.67,
+      "step": 27175,
+      "train_speed(iter/s)": 1.315449
+    },
+    {
+      "acc": 0.66904159,
+      "epoch": 0.6894977168949772,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.807012667366332e-06,
+      "loss": 1.46008053,
+      "memory(GiB)": 113.67,
+      "step": 27180,
+      "train_speed(iter/s)": 1.31512
+    },
+    {
+      "acc": 0.68295555,
+      "epoch": 0.6896245560629122,
+      "grad_norm": 3.9375,
+      "learning_rate": 7.806144824594994e-06,
+      "loss": 1.43671284,
+      "memory(GiB)": 113.67,
+      "step": 27185,
+      "train_speed(iter/s)": 1.315148
+    },
+    {
+      "acc": 0.68813438,
+      "epoch": 0.6897513952308473,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.805276858396879e-06,
+      "loss": 1.39312048,
+      "memory(GiB)": 113.67,
+      "step": 27190,
+      "train_speed(iter/s)": 1.315178
+    },
+    {
+      "acc": 0.68584666,
+      "epoch": 0.6898782343987824,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.804408768810164e-06,
+      "loss": 1.38195562,
+      "memory(GiB)": 113.67,
+      "step": 27195,
+      "train_speed(iter/s)": 1.315209
+    },
+    {
+      "acc": 0.67978492,
+      "epoch": 0.6900050735667174,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.80354055587303e-06,
+      "loss": 1.47284002,
+      "memory(GiB)": 113.67,
+      "step": 27200,
+      "train_speed(iter/s)": 1.315238
+    },
+    {
+      "acc": 0.68082733,
+      "epoch": 0.6901319127346525,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.802672219623665e-06,
+      "loss": 1.38082848,
+      "memory(GiB)": 113.67,
+      "step": 27205,
+      "train_speed(iter/s)": 1.315266
+    },
+    {
+      "acc": 0.70008011,
+      "epoch": 0.6902587519025876,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.801803760100264e-06,
+      "loss": 1.38540916,
+      "memory(GiB)": 113.67,
+      "step": 27210,
+      "train_speed(iter/s)": 1.315296
+    },
+    {
+      "acc": 0.68497849,
+      "epoch": 0.6903855910705226,
+      "grad_norm": 3.96875,
+      "learning_rate": 7.800935177341022e-06,
+      "loss": 1.3621438,
+      "memory(GiB)": 113.67,
+      "step": 27215,
+      "train_speed(iter/s)": 1.315326
+    },
+    {
+      "acc": 0.69843593,
+      "epoch": 0.6905124302384577,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.800066471384149e-06,
+      "loss": 1.39573727,
+      "memory(GiB)": 113.67,
+      "step": 27220,
+      "train_speed(iter/s)": 1.315354
+    },
+    {
+      "acc": 0.69583349,
+      "epoch": 0.6906392694063926,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.799197642267848e-06,
+      "loss": 1.40435343,
+      "memory(GiB)": 113.67,
+      "step": 27225,
+      "train_speed(iter/s)": 1.315383
+    },
+    {
+      "acc": 0.68500781,
+      "epoch": 0.6907661085743277,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.79832869003034e-06,
+      "loss": 1.36655731,
+      "memory(GiB)": 113.67,
+      "step": 27230,
+      "train_speed(iter/s)": 1.315414
+    },
+    {
+      "acc": 0.69661083,
+      "epoch": 0.6908929477422628,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.797459614709842e-06,
+      "loss": 1.39594584,
+      "memory(GiB)": 113.67,
+      "step": 27235,
+      "train_speed(iter/s)": 1.315444
+    },
+    {
+      "acc": 0.6878828,
+      "epoch": 0.6910197869101978,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.796590416344578e-06,
+      "loss": 1.40278797,
+      "memory(GiB)": 113.67,
+      "step": 27240,
+      "train_speed(iter/s)": 1.315473
+    },
+    {
+      "acc": 0.67815237,
+      "epoch": 0.6911466260781329,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.795721094972783e-06,
+      "loss": 1.48127718,
+      "memory(GiB)": 113.67,
+      "step": 27245,
+      "train_speed(iter/s)": 1.3155
+    },
+    {
+      "acc": 0.66533723,
+      "epoch": 0.691273465246068,
+      "grad_norm": 3.5,
+      "learning_rate": 7.794851650632693e-06,
+      "loss": 1.49002409,
+      "memory(GiB)": 113.67,
+      "step": 27250,
+      "train_speed(iter/s)": 1.315531
+    },
+    {
+      "acc": 0.691995,
+      "epoch": 0.691400304414003,
+      "grad_norm": 2.625,
+      "learning_rate": 7.793982083362548e-06,
+      "loss": 1.30484381,
+      "memory(GiB)": 113.67,
+      "step": 27255,
+      "train_speed(iter/s)": 1.315562
+    },
+    {
+      "acc": 0.70029144,
+      "epoch": 0.6915271435819381,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.7931123932006e-06,
+      "loss": 1.36576147,
+      "memory(GiB)": 113.67,
+      "step": 27260,
+      "train_speed(iter/s)": 1.315593
+    },
+    {
+      "acc": 0.67921634,
+      "epoch": 0.6916539827498731,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.792242580185095e-06,
+      "loss": 1.3831749,
+      "memory(GiB)": 113.67,
+      "step": 27265,
+      "train_speed(iter/s)": 1.315623
+    },
+    {
+      "acc": 0.68872709,
+      "epoch": 0.6917808219178082,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.791372644354295e-06,
+      "loss": 1.33748703,
+      "memory(GiB)": 113.67,
+      "step": 27270,
+      "train_speed(iter/s)": 1.315654
+    },
+    {
+      "acc": 0.6851624,
+      "epoch": 0.6919076610857433,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.790502585746464e-06,
+      "loss": 1.42216225,
+      "memory(GiB)": 113.67,
+      "step": 27275,
+      "train_speed(iter/s)": 1.315684
+    },
+    {
+      "acc": 0.69421768,
+      "epoch": 0.6920345002536783,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.789632404399872e-06,
+      "loss": 1.33547859,
+      "memory(GiB)": 113.67,
+      "step": 27280,
+      "train_speed(iter/s)": 1.315715
+    },
+    {
+      "acc": 0.67823405,
+      "epoch": 0.6921613394216134,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.788762100352791e-06,
+      "loss": 1.41549778,
+      "memory(GiB)": 113.67,
+      "step": 27285,
+      "train_speed(iter/s)": 1.315746
+    },
+    {
+      "acc": 0.70169363,
+      "epoch": 0.6922881785895485,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.787891673643501e-06,
+      "loss": 1.38942432,
+      "memory(GiB)": 113.67,
+      "step": 27290,
+      "train_speed(iter/s)": 1.315775
+    },
+    {
+      "acc": 0.6852705,
+      "epoch": 0.6924150177574835,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.78702112431029e-06,
+      "loss": 1.41622238,
+      "memory(GiB)": 113.67,
+      "step": 27295,
+      "train_speed(iter/s)": 1.315801
+    },
+    {
+      "acc": 0.68978567,
+      "epoch": 0.6925418569254186,
+      "grad_norm": 2.875,
+      "learning_rate": 7.786150452391446e-06,
+      "loss": 1.35744591,
+      "memory(GiB)": 113.67,
+      "step": 27300,
+      "train_speed(iter/s)": 1.315831
+    },
+    {
+      "acc": 0.69616528,
+      "epoch": 0.6926686960933536,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.785279657925265e-06,
+      "loss": 1.35828333,
+      "memory(GiB)": 113.67,
+      "step": 27305,
+      "train_speed(iter/s)": 1.315861
+    },
+    {
+      "acc": 0.67863131,
+      "epoch": 0.6927955352612887,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.784408740950051e-06,
+      "loss": 1.42391748,
+      "memory(GiB)": 113.67,
+      "step": 27310,
+      "train_speed(iter/s)": 1.315891
+    },
+    {
+      "acc": 0.70041037,
+      "epoch": 0.6929223744292238,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.783537701504109e-06,
+      "loss": 1.38755322,
+      "memory(GiB)": 113.67,
+      "step": 27315,
+      "train_speed(iter/s)": 1.31592
+    },
+    {
+      "acc": 0.68981714,
+      "epoch": 0.6930492135971588,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.782666539625749e-06,
+      "loss": 1.4521965,
+      "memory(GiB)": 113.67,
+      "step": 27320,
+      "train_speed(iter/s)": 1.315951
+    },
+    {
+      "acc": 0.69706354,
+      "epoch": 0.6931760527650939,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.781795255353293e-06,
+      "loss": 1.33944073,
+      "memory(GiB)": 113.67,
+      "step": 27325,
+      "train_speed(iter/s)": 1.31598
+    },
+    {
+      "acc": 0.69937019,
+      "epoch": 0.693302891933029,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.780923848725061e-06,
+      "loss": 1.34258041,
+      "memory(GiB)": 113.67,
+      "step": 27330,
+      "train_speed(iter/s)": 1.31601
+    },
+    {
+      "acc": 0.67645907,
+      "epoch": 0.693429731100964,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.780052319779382e-06,
+      "loss": 1.39651518,
+      "memory(GiB)": 113.67,
+      "step": 27335,
+      "train_speed(iter/s)": 1.316039
+    },
+    {
+      "acc": 0.67720528,
+      "epoch": 0.693556570268899,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.779180668554591e-06,
+      "loss": 1.45166073,
+      "memory(GiB)": 113.67,
+      "step": 27340,
+      "train_speed(iter/s)": 1.31607
+    },
+    {
+      "acc": 0.69371071,
+      "epoch": 0.693683409436834,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.778308895089024e-06,
+      "loss": 1.36368017,
+      "memory(GiB)": 113.67,
+      "step": 27345,
+      "train_speed(iter/s)": 1.316099
+    },
+    {
+      "acc": 0.70163326,
+      "epoch": 0.6938102486047691,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.77743699942103e-06,
+      "loss": 1.34653311,
+      "memory(GiB)": 113.67,
+      "step": 27350,
+      "train_speed(iter/s)": 1.316129
+    },
+    {
+      "acc": 0.69544134,
+      "epoch": 0.6939370877727042,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.776564981588955e-06,
+      "loss": 1.36343918,
+      "memory(GiB)": 113.67,
+      "step": 27355,
+      "train_speed(iter/s)": 1.316159
+    },
+    {
+      "acc": 0.67890406,
+      "epoch": 0.6940639269406392,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.775692841631154e-06,
+      "loss": 1.48226852,
+      "memory(GiB)": 113.67,
+      "step": 27360,
+      "train_speed(iter/s)": 1.31619
+    },
+    {
+      "acc": 0.68064585,
+      "epoch": 0.6941907661085743,
+      "grad_norm": 2.46875,
+      "learning_rate": 7.774820579585993e-06,
+      "loss": 1.42100563,
+      "memory(GiB)": 113.67,
+      "step": 27365,
+      "train_speed(iter/s)": 1.31622
+    },
+    {
+      "acc": 0.69702067,
+      "epoch": 0.6943176052765094,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.773948195491831e-06,
+      "loss": 1.38180838,
+      "memory(GiB)": 113.67,
+      "step": 27370,
+      "train_speed(iter/s)": 1.31625
+    },
+    {
+      "acc": 0.68011513,
+      "epoch": 0.6944444444444444,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.773075689387044e-06,
+      "loss": 1.45091887,
+      "memory(GiB)": 113.67,
+      "step": 27375,
+      "train_speed(iter/s)": 1.316281
+    },
+    {
+      "acc": 0.70421619,
+      "epoch": 0.6945712836123795,
+      "grad_norm": 4.25,
+      "learning_rate": 7.772203061310008e-06,
+      "loss": 1.34823809,
+      "memory(GiB)": 113.67,
+      "step": 27380,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.67676497,
+      "epoch": 0.6946981227803145,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.771330311299104e-06,
+      "loss": 1.47360125,
+      "memory(GiB)": 113.67,
+      "step": 27385,
+      "train_speed(iter/s)": 1.31634
+    },
+    {
+      "acc": 0.68019776,
+      "epoch": 0.6948249619482496,
+      "grad_norm": 2.5,
+      "learning_rate": 7.770457439392719e-06,
+      "loss": 1.40169182,
+      "memory(GiB)": 113.67,
+      "step": 27390,
+      "train_speed(iter/s)": 1.31637
+    },
+    {
+      "acc": 0.6732429,
+      "epoch": 0.6949518011161847,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.769584445629247e-06,
+      "loss": 1.47364311,
+      "memory(GiB)": 113.67,
+      "step": 27395,
+      "train_speed(iter/s)": 1.316398
+    },
+    {
+      "acc": 0.68089781,
+      "epoch": 0.6950786402841197,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.768711330047087e-06,
+      "loss": 1.38310347,
+      "memory(GiB)": 113.67,
+      "step": 27400,
+      "train_speed(iter/s)": 1.316066
+    },
+    {
+      "acc": 0.68824363,
+      "epoch": 0.6952054794520548,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.767838092684638e-06,
+      "loss": 1.42682705,
+      "memory(GiB)": 113.67,
+      "step": 27405,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.70430212,
+      "epoch": 0.6953323186199899,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.766964733580316e-06,
+      "loss": 1.32226906,
+      "memory(GiB)": 113.67,
+      "step": 27410,
+      "train_speed(iter/s)": 1.31612
+    },
+    {
+      "acc": 0.68973246,
+      "epoch": 0.6954591577879249,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.76609125277253e-06,
+      "loss": 1.39550867,
+      "memory(GiB)": 113.67,
+      "step": 27415,
+      "train_speed(iter/s)": 1.31615
+    },
+    {
+      "acc": 0.6906055,
+      "epoch": 0.69558599695586,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.7652176502997e-06,
+      "loss": 1.35425682,
+      "memory(GiB)": 113.67,
+      "step": 27420,
+      "train_speed(iter/s)": 1.316179
+    },
+    {
+      "acc": 0.6911396,
+      "epoch": 0.695712836123795,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.764343926200254e-06,
+      "loss": 1.3403326,
+      "memory(GiB)": 113.67,
+      "step": 27425,
+      "train_speed(iter/s)": 1.316209
+    },
+    {
+      "acc": 0.68281817,
+      "epoch": 0.6958396752917301,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.763470080512617e-06,
+      "loss": 1.41958246,
+      "memory(GiB)": 113.67,
+      "step": 27430,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.68839464,
+      "epoch": 0.6959665144596652,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.762596113275229e-06,
+      "loss": 1.37729197,
+      "memory(GiB)": 113.67,
+      "step": 27435,
+      "train_speed(iter/s)": 1.316268
+    },
+    {
+      "acc": 0.68475089,
+      "epoch": 0.6960933536276002,
+      "grad_norm": 3.96875,
+      "learning_rate": 7.761722024526533e-06,
+      "loss": 1.4439435,
+      "memory(GiB)": 113.67,
+      "step": 27440,
+      "train_speed(iter/s)": 1.316298
+    },
+    {
+      "acc": 0.68501863,
+      "epoch": 0.6962201927955353,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.760847814304969e-06,
+      "loss": 1.40984821,
+      "memory(GiB)": 113.67,
+      "step": 27445,
+      "train_speed(iter/s)": 1.316325
+    },
+    {
+      "acc": 0.6789031,
+      "epoch": 0.6963470319634704,
+      "grad_norm": 4.375,
+      "learning_rate": 7.759973482648992e-06,
+      "loss": 1.39156427,
+      "memory(GiB)": 113.67,
+      "step": 27450,
+      "train_speed(iter/s)": 1.316353
+    },
+    {
+      "acc": 0.68452573,
+      "epoch": 0.6964738711314054,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.75909902959706e-06,
+      "loss": 1.46257381,
+      "memory(GiB)": 113.67,
+      "step": 27455,
+      "train_speed(iter/s)": 1.316382
+    },
+    {
+      "acc": 0.67274532,
+      "epoch": 0.6966007102993405,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.758224455187632e-06,
+      "loss": 1.45549269,
+      "memory(GiB)": 113.67,
+      "step": 27460,
+      "train_speed(iter/s)": 1.316409
+    },
+    {
+      "acc": 0.6799789,
+      "epoch": 0.6967275494672754,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.75734975945918e-06,
+      "loss": 1.47349281,
+      "memory(GiB)": 113.67,
+      "step": 27465,
+      "train_speed(iter/s)": 1.316431
+    },
+    {
+      "acc": 0.68443208,
+      "epoch": 0.6968543886352105,
+      "grad_norm": 3.9375,
+      "learning_rate": 7.756474942450174e-06,
+      "loss": 1.43744135,
+      "memory(GiB)": 113.67,
+      "step": 27470,
+      "train_speed(iter/s)": 1.316458
+    },
+    {
+      "acc": 0.70245452,
+      "epoch": 0.6969812278031456,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.755600004199094e-06,
+      "loss": 1.33729553,
+      "memory(GiB)": 113.67,
+      "step": 27475,
+      "train_speed(iter/s)": 1.316485
+    },
+    {
+      "acc": 0.69367132,
+      "epoch": 0.6971080669710806,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.754724944744423e-06,
+      "loss": 1.37735901,
+      "memory(GiB)": 113.67,
+      "step": 27480,
+      "train_speed(iter/s)": 1.316512
+    },
+    {
+      "acc": 0.68661513,
+      "epoch": 0.6972349061390157,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.753849764124648e-06,
+      "loss": 1.44239902,
+      "memory(GiB)": 113.67,
+      "step": 27485,
+      "train_speed(iter/s)": 1.316541
+    },
+    {
+      "acc": 0.6795898,
+      "epoch": 0.6973617453069508,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.752974462378268e-06,
+      "loss": 1.41091204,
+      "memory(GiB)": 113.67,
+      "step": 27490,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.68220091,
+      "epoch": 0.6974885844748858,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.752099039543778e-06,
+      "loss": 1.42725811,
+      "memory(GiB)": 113.67,
+      "step": 27495,
+      "train_speed(iter/s)": 1.316598
+    },
+    {
+      "acc": 0.68657184,
+      "epoch": 0.6976154236428209,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.751223495659685e-06,
+      "loss": 1.35470104,
+      "memory(GiB)": 113.67,
+      "step": 27500,
+      "train_speed(iter/s)": 1.316627
+    },
+    {
+      "acc": 0.68589306,
+      "epoch": 0.6977422628107559,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.7503478307645e-06,
+      "loss": 1.38404951,
+      "memory(GiB)": 113.67,
+      "step": 27505,
+      "train_speed(iter/s)": 1.316655
+    },
+    {
+      "acc": 0.69229403,
+      "epoch": 0.697869101978691,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.74947204489674e-06,
+      "loss": 1.36169739,
+      "memory(GiB)": 113.67,
+      "step": 27510,
+      "train_speed(iter/s)": 1.316683
+    },
+    {
+      "acc": 0.67766395,
+      "epoch": 0.6979959411466261,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.748596138094922e-06,
+      "loss": 1.47127914,
+      "memory(GiB)": 113.67,
+      "step": 27515,
+      "train_speed(iter/s)": 1.316711
+    },
+    {
+      "acc": 0.6562438,
+      "epoch": 0.6981227803145611,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.747720110397573e-06,
+      "loss": 1.48608313,
+      "memory(GiB)": 113.67,
+      "step": 27520,
+      "train_speed(iter/s)": 1.316738
+    },
+    {
+      "acc": 0.68536882,
+      "epoch": 0.6982496194824962,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.746843961843226e-06,
+      "loss": 1.43583679,
+      "memory(GiB)": 113.67,
+      "step": 27525,
+      "train_speed(iter/s)": 1.316766
+    },
+    {
+      "acc": 0.701373,
+      "epoch": 0.6983764586504313,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.74596769247042e-06,
+      "loss": 1.35383158,
+      "memory(GiB)": 113.67,
+      "step": 27530,
+      "train_speed(iter/s)": 1.316796
+    },
+    {
+      "acc": 0.68843298,
+      "epoch": 0.6985032978183663,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.745091302317694e-06,
+      "loss": 1.38515539,
+      "memory(GiB)": 113.67,
+      "step": 27535,
+      "train_speed(iter/s)": 1.316825
+    },
+    {
+      "acc": 0.69328885,
+      "epoch": 0.6986301369863014,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.744214791423597e-06,
+      "loss": 1.39139061,
+      "memory(GiB)": 113.67,
+      "step": 27540,
+      "train_speed(iter/s)": 1.316853
+    },
+    {
+      "acc": 0.66802368,
+      "epoch": 0.6987569761542364,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.74333815982668e-06,
+      "loss": 1.44440355,
+      "memory(GiB)": 113.67,
+      "step": 27545,
+      "train_speed(iter/s)": 1.316882
+    },
+    {
+      "acc": 0.68172398,
+      "epoch": 0.6988838153221715,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.742461407565504e-06,
+      "loss": 1.3686491,
+      "memory(GiB)": 113.67,
+      "step": 27550,
+      "train_speed(iter/s)": 1.31691
+    },
+    {
+      "acc": 0.68809366,
+      "epoch": 0.6990106544901066,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.741584534678632e-06,
+      "loss": 1.40915604,
+      "memory(GiB)": 113.67,
+      "step": 27555,
+      "train_speed(iter/s)": 1.316938
+    },
+    {
+      "acc": 0.6965344,
+      "epoch": 0.6991374936580416,
+      "grad_norm": 2.875,
+      "learning_rate": 7.74070754120463e-06,
+      "loss": 1.40982618,
+      "memory(GiB)": 113.67,
+      "step": 27560,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.6888875,
+      "epoch": 0.6992643328259767,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.739830427182073e-06,
+      "loss": 1.42573032,
+      "memory(GiB)": 113.67,
+      "step": 27565,
+      "train_speed(iter/s)": 1.316993
+    },
+    {
+      "acc": 0.68637438,
+      "epoch": 0.6993911719939118,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.738953192649544e-06,
+      "loss": 1.40664406,
+      "memory(GiB)": 113.67,
+      "step": 27570,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.67539234,
+      "epoch": 0.6995180111618468,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.738075837645625e-06,
+      "loss": 1.41859283,
+      "memory(GiB)": 113.67,
+      "step": 27575,
+      "train_speed(iter/s)": 1.31705
+    },
+    {
+      "acc": 0.69269924,
+      "epoch": 0.6996448503297819,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.737198362208904e-06,
+      "loss": 1.38010235,
+      "memory(GiB)": 113.67,
+      "step": 27580,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.68612881,
+      "epoch": 0.6997716894977168,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.736320766377978e-06,
+      "loss": 1.46556129,
+      "memory(GiB)": 113.67,
+      "step": 27585,
+      "train_speed(iter/s)": 1.317103
+    },
+    {
+      "acc": 0.6716094,
+      "epoch": 0.6998985286656519,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.735443050191452e-06,
+      "loss": 1.38440409,
+      "memory(GiB)": 113.67,
+      "step": 27590,
+      "train_speed(iter/s)": 1.317128
+    },
+    {
+      "acc": 0.68134713,
+      "epoch": 0.700025367833587,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.734565213687923e-06,
+      "loss": 1.3729351,
+      "memory(GiB)": 113.67,
+      "step": 27595,
+      "train_speed(iter/s)": 1.317158
+    },
+    {
+      "acc": 0.68417273,
+      "epoch": 0.700152207001522,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.733687256906009e-06,
+      "loss": 1.41817379,
+      "memory(GiB)": 113.67,
+      "step": 27600,
+      "train_speed(iter/s)": 1.317186
+    },
+    {
+      "acc": 0.68811927,
+      "epoch": 0.7002790461694571,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.732809179884324e-06,
+      "loss": 1.38756752,
+      "memory(GiB)": 113.67,
+      "step": 27605,
+      "train_speed(iter/s)": 1.317214
+    },
+    {
+      "acc": 0.68969746,
+      "epoch": 0.7004058853373922,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.73193098266149e-06,
+      "loss": 1.36519365,
+      "memory(GiB)": 113.67,
+      "step": 27610,
+      "train_speed(iter/s)": 1.317242
+    },
+    {
+      "acc": 0.67417812,
+      "epoch": 0.7005327245053272,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.731052665276135e-06,
+      "loss": 1.44345427,
+      "memory(GiB)": 113.67,
+      "step": 27615,
+      "train_speed(iter/s)": 1.317271
+    },
+    {
+      "acc": 0.70380344,
+      "epoch": 0.7006595636732623,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.730174227766892e-06,
+      "loss": 1.34304218,
+      "memory(GiB)": 113.67,
+      "step": 27620,
+      "train_speed(iter/s)": 1.317298
+    },
+    {
+      "acc": 0.69666653,
+      "epoch": 0.7007864028411973,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.729295670172394e-06,
+      "loss": 1.38378124,
+      "memory(GiB)": 113.67,
+      "step": 27625,
+      "train_speed(iter/s)": 1.317326
+    },
+    {
+      "acc": 0.68773742,
+      "epoch": 0.7009132420091324,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.728416992531287e-06,
+      "loss": 1.39607,
+      "memory(GiB)": 113.67,
+      "step": 27630,
+      "train_speed(iter/s)": 1.317353
+    },
+    {
+      "acc": 0.68292499,
+      "epoch": 0.7010400811770675,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.72753819488222e-06,
+      "loss": 1.43549023,
+      "memory(GiB)": 113.67,
+      "step": 27635,
+      "train_speed(iter/s)": 1.317381
+    },
+    {
+      "acc": 0.68508878,
+      "epoch": 0.7011669203450025,
+      "grad_norm": 3.0,
+      "learning_rate": 7.726659277263848e-06,
+      "loss": 1.36854839,
+      "memory(GiB)": 113.67,
+      "step": 27640,
+      "train_speed(iter/s)": 1.317409
+    },
+    {
+      "acc": 0.68659806,
+      "epoch": 0.7012937595129376,
+      "grad_norm": 5.28125,
+      "learning_rate": 7.725780239714824e-06,
+      "loss": 1.4442852,
+      "memory(GiB)": 113.67,
+      "step": 27645,
+      "train_speed(iter/s)": 1.317435
+    },
+    {
+      "acc": 0.68395224,
+      "epoch": 0.7014205986808727,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.724901082273817e-06,
+      "loss": 1.37260056,
+      "memory(GiB)": 113.67,
+      "step": 27650,
+      "train_speed(iter/s)": 1.317461
+    },
+    {
+      "acc": 0.69167633,
+      "epoch": 0.7015474378488077,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.724021804979493e-06,
+      "loss": 1.35505219,
+      "memory(GiB)": 113.67,
+      "step": 27655,
+      "train_speed(iter/s)": 1.317487
+    },
+    {
+      "acc": 0.6905654,
+      "epoch": 0.7016742770167428,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.723142407870532e-06,
+      "loss": 1.39980001,
+      "memory(GiB)": 113.67,
+      "step": 27660,
+      "train_speed(iter/s)": 1.317515
+    },
+    {
+      "acc": 0.7034163,
+      "epoch": 0.7018011161846778,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.722262890985605e-06,
+      "loss": 1.40444803,
+      "memory(GiB)": 113.67,
+      "step": 27665,
+      "train_speed(iter/s)": 1.317543
+    },
+    {
+      "acc": 0.67838058,
+      "epoch": 0.7019279553526129,
+      "grad_norm": 3.0,
+      "learning_rate": 7.721383254363407e-06,
+      "loss": 1.42782164,
+      "memory(GiB)": 113.67,
+      "step": 27670,
+      "train_speed(iter/s)": 1.31757
+    },
+    {
+      "acc": 0.70784187,
+      "epoch": 0.702054794520548,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.72050349804262e-06,
+      "loss": 1.30793266,
+      "memory(GiB)": 113.67,
+      "step": 27675,
+      "train_speed(iter/s)": 1.317598
+    },
+    {
+      "acc": 0.70117679,
+      "epoch": 0.702181633688483,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.719623622061943e-06,
+      "loss": 1.33565235,
+      "memory(GiB)": 113.67,
+      "step": 27680,
+      "train_speed(iter/s)": 1.317625
+    },
+    {
+      "acc": 0.67812605,
+      "epoch": 0.7023084728564181,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.718743626460076e-06,
+      "loss": 1.42488441,
+      "memory(GiB)": 113.67,
+      "step": 27685,
+      "train_speed(iter/s)": 1.317653
+    },
+    {
+      "acc": 0.69983234,
+      "epoch": 0.7024353120243532,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.71786351127573e-06,
+      "loss": 1.33763628,
+      "memory(GiB)": 113.67,
+      "step": 27690,
+      "train_speed(iter/s)": 1.317679
+    },
+    {
+      "acc": 0.68522749,
+      "epoch": 0.7025621511922882,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.71698327654761e-06,
+      "loss": 1.36088371,
+      "memory(GiB)": 113.67,
+      "step": 27695,
+      "train_speed(iter/s)": 1.317706
+    },
+    {
+      "acc": 0.67995195,
+      "epoch": 0.7026889903602233,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.716102922314435e-06,
+      "loss": 1.39688911,
+      "memory(GiB)": 113.67,
+      "step": 27700,
+      "train_speed(iter/s)": 1.317733
+    },
+    {
+      "acc": 0.66598182,
+      "epoch": 0.7028158295281582,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.715222448614926e-06,
+      "loss": 1.49871988,
+      "memory(GiB)": 113.67,
+      "step": 27705,
+      "train_speed(iter/s)": 1.31776
+    },
+    {
+      "acc": 0.68612494,
+      "epoch": 0.7029426686960933,
+      "grad_norm": 3.875,
+      "learning_rate": 7.714341855487812e-06,
+      "loss": 1.41206093,
+      "memory(GiB)": 113.67,
+      "step": 27710,
+      "train_speed(iter/s)": 1.317788
+    },
+    {
+      "acc": 0.67902756,
+      "epoch": 0.7030695078640284,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.713461142971824e-06,
+      "loss": 1.38966484,
+      "memory(GiB)": 113.67,
+      "step": 27715,
+      "train_speed(iter/s)": 1.317815
+    },
+    {
+      "acc": 0.69626694,
+      "epoch": 0.7031963470319634,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.712580311105701e-06,
+      "loss": 1.39449596,
+      "memory(GiB)": 113.67,
+      "step": 27720,
+      "train_speed(iter/s)": 1.317842
+    },
+    {
+      "acc": 0.685115,
+      "epoch": 0.7033231861998985,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.711699359928184e-06,
+      "loss": 1.35194759,
+      "memory(GiB)": 113.67,
+      "step": 27725,
+      "train_speed(iter/s)": 1.31787
+    },
+    {
+      "acc": 0.70073137,
+      "epoch": 0.7034500253678336,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.710818289478024e-06,
+      "loss": 1.35960045,
+      "memory(GiB)": 113.67,
+      "step": 27730,
+      "train_speed(iter/s)": 1.317898
+    },
+    {
+      "acc": 0.69258127,
+      "epoch": 0.7035768645357686,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.709937099793971e-06,
+      "loss": 1.37972374,
+      "memory(GiB)": 113.67,
+      "step": 27735,
+      "train_speed(iter/s)": 1.317927
+    },
+    {
+      "acc": 0.67662883,
+      "epoch": 0.7037037037037037,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.709055790914787e-06,
+      "loss": 1.42584848,
+      "memory(GiB)": 113.67,
+      "step": 27740,
+      "train_speed(iter/s)": 1.317955
+    },
+    {
+      "acc": 0.6830132,
+      "epoch": 0.7038305428716387,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.708174362879234e-06,
+      "loss": 1.38664846,
+      "memory(GiB)": 113.67,
+      "step": 27745,
+      "train_speed(iter/s)": 1.317978
+    },
+    {
+      "acc": 0.67717466,
+      "epoch": 0.7039573820395738,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.70729281572608e-06,
+      "loss": 1.46629658,
+      "memory(GiB)": 113.67,
+      "step": 27750,
+      "train_speed(iter/s)": 1.318007
+    },
+    {
+      "acc": 0.67381039,
+      "epoch": 0.7040842212075089,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.706411149494102e-06,
+      "loss": 1.44653416,
+      "memory(GiB)": 113.67,
+      "step": 27755,
+      "train_speed(iter/s)": 1.318034
+    },
+    {
+      "acc": 0.68937721,
+      "epoch": 0.7042110603754439,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.705529364222079e-06,
+      "loss": 1.4153265,
+      "memory(GiB)": 113.67,
+      "step": 27760,
+      "train_speed(iter/s)": 1.318061
+    },
+    {
+      "acc": 0.68894091,
+      "epoch": 0.704337899543379,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.704647459948793e-06,
+      "loss": 1.38318691,
+      "memory(GiB)": 113.67,
+      "step": 27765,
+      "train_speed(iter/s)": 1.318089
+    },
+    {
+      "acc": 0.6916255,
+      "epoch": 0.7044647387113141,
+      "grad_norm": 2.875,
+      "learning_rate": 7.703765436713038e-06,
+      "loss": 1.38308249,
+      "memory(GiB)": 113.67,
+      "step": 27770,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.70402603,
+      "epoch": 0.7045915778792491,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.702883294553607e-06,
+      "loss": 1.32748394,
+      "memory(GiB)": 113.67,
+      "step": 27775,
+      "train_speed(iter/s)": 1.318145
+    },
+    {
+      "acc": 0.69555645,
+      "epoch": 0.7047184170471842,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.702001033509302e-06,
+      "loss": 1.33570232,
+      "memory(GiB)": 113.67,
+      "step": 27780,
+      "train_speed(iter/s)": 1.318172
+    },
+    {
+      "acc": 0.67679191,
+      "epoch": 0.7048452562151192,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.701118653618927e-06,
+      "loss": 1.41933041,
+      "memory(GiB)": 113.67,
+      "step": 27785,
+      "train_speed(iter/s)": 1.3182
+    },
+    {
+      "acc": 0.67637415,
+      "epoch": 0.7049720953830543,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.700236154921294e-06,
+      "loss": 1.40455265,
+      "memory(GiB)": 113.67,
+      "step": 27790,
+      "train_speed(iter/s)": 1.318227
+    },
+    {
+      "acc": 0.67603498,
+      "epoch": 0.7050989345509894,
+      "grad_norm": 2.75,
+      "learning_rate": 7.699353537455222e-06,
+      "loss": 1.41159344,
+      "memory(GiB)": 113.67,
+      "step": 27795,
+      "train_speed(iter/s)": 1.318256
+    },
+    {
+      "acc": 0.67648301,
+      "epoch": 0.7052257737189244,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.698470801259526e-06,
+      "loss": 1.42712507,
+      "memory(GiB)": 113.67,
+      "step": 27800,
+      "train_speed(iter/s)": 1.318283
+    },
+    {
+      "acc": 0.68914614,
+      "epoch": 0.7053526128868595,
+      "grad_norm": 3.375,
+      "learning_rate": 7.697587946373037e-06,
+      "loss": 1.43538628,
+      "memory(GiB)": 113.67,
+      "step": 27805,
+      "train_speed(iter/s)": 1.31831
+    },
+    {
+      "acc": 0.68238916,
+      "epoch": 0.7054794520547946,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.696704972834589e-06,
+      "loss": 1.36512432,
+      "memory(GiB)": 113.67,
+      "step": 27810,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "acc": 0.68954344,
+      "epoch": 0.7056062912227296,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.695821880683012e-06,
+      "loss": 1.41139202,
+      "memory(GiB)": 113.67,
+      "step": 27815,
+      "train_speed(iter/s)": 1.318365
+    },
+    {
+      "acc": 0.68539515,
+      "epoch": 0.7057331303906647,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.694938669957156e-06,
+      "loss": 1.37493801,
+      "memory(GiB)": 113.67,
+      "step": 27820,
+      "train_speed(iter/s)": 1.318393
+    },
+    {
+      "acc": 0.68933992,
+      "epoch": 0.7058599695585996,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.694055340695862e-06,
+      "loss": 1.37572298,
+      "memory(GiB)": 113.67,
+      "step": 27825,
+      "train_speed(iter/s)": 1.31842
+    },
+    {
+      "acc": 0.69452915,
+      "epoch": 0.7059868087265347,
+      "grad_norm": 4.125,
+      "learning_rate": 7.693171892937991e-06,
+      "loss": 1.39475746,
+      "memory(GiB)": 113.67,
+      "step": 27830,
+      "train_speed(iter/s)": 1.318447
+    },
+    {
+      "acc": 0.68810349,
+      "epoch": 0.7061136478944698,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.692288326722393e-06,
+      "loss": 1.44691944,
+      "memory(GiB)": 113.67,
+      "step": 27835,
+      "train_speed(iter/s)": 1.31847
+    },
+    {
+      "acc": 0.68176169,
+      "epoch": 0.7062404870624048,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.691404642087933e-06,
+      "loss": 1.37605438,
+      "memory(GiB)": 113.67,
+      "step": 27840,
+      "train_speed(iter/s)": 1.318498
+    },
+    {
+      "acc": 0.70087519,
+      "epoch": 0.7063673262303399,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.690520839073484e-06,
+      "loss": 1.36085634,
+      "memory(GiB)": 113.67,
+      "step": 27845,
+      "train_speed(iter/s)": 1.318526
+    },
+    {
+      "acc": 0.68163652,
+      "epoch": 0.706494165398275,
+      "grad_norm": 3.25,
+      "learning_rate": 7.689636917717913e-06,
+      "loss": 1.38648071,
+      "memory(GiB)": 113.67,
+      "step": 27850,
+      "train_speed(iter/s)": 1.318555
+    },
+    {
+      "acc": 0.68778505,
+      "epoch": 0.70662100456621,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.688752878060103e-06,
+      "loss": 1.39658413,
+      "memory(GiB)": 113.67,
+      "step": 27855,
+      "train_speed(iter/s)": 1.318583
+    },
+    {
+      "acc": 0.6841764,
+      "epoch": 0.7067478437341451,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.687868720138939e-06,
+      "loss": 1.36227751,
+      "memory(GiB)": 113.67,
+      "step": 27860,
+      "train_speed(iter/s)": 1.318611
+    },
+    {
+      "acc": 0.68406801,
+      "epoch": 0.7068746829020801,
+      "grad_norm": 3.625,
+      "learning_rate": 7.686984443993304e-06,
+      "loss": 1.43790407,
+      "memory(GiB)": 113.67,
+      "step": 27865,
+      "train_speed(iter/s)": 1.31864
+    },
+    {
+      "acc": 0.67955041,
+      "epoch": 0.7070015220700152,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.6861000496621e-06,
+      "loss": 1.41355801,
+      "memory(GiB)": 113.67,
+      "step": 27870,
+      "train_speed(iter/s)": 1.318668
+    },
+    {
+      "acc": 0.68976464,
+      "epoch": 0.7071283612379503,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.685215537184223e-06,
+      "loss": 1.44155293,
+      "memory(GiB)": 113.67,
+      "step": 27875,
+      "train_speed(iter/s)": 1.318697
+    },
+    {
+      "acc": 0.68805428,
+      "epoch": 0.7072552004058853,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.684330906598577e-06,
+      "loss": 1.45625353,
+      "memory(GiB)": 113.67,
+      "step": 27880,
+      "train_speed(iter/s)": 1.318727
+    },
+    {
+      "acc": 0.69333124,
+      "epoch": 0.7073820395738204,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.683446157944075e-06,
+      "loss": 1.36977482,
+      "memory(GiB)": 113.67,
+      "step": 27885,
+      "train_speed(iter/s)": 1.318754
+    },
+    {
+      "acc": 0.67815008,
+      "epoch": 0.7075088787417555,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.682561291259628e-06,
+      "loss": 1.4715868,
+      "memory(GiB)": 113.67,
+      "step": 27890,
+      "train_speed(iter/s)": 1.318782
+    },
+    {
+      "acc": 0.68808141,
+      "epoch": 0.7076357179096905,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.681676306584159e-06,
+      "loss": 1.3986867,
+      "memory(GiB)": 113.67,
+      "step": 27895,
+      "train_speed(iter/s)": 1.31881
+    },
+    {
+      "acc": 0.68770208,
+      "epoch": 0.7077625570776256,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.680791203956594e-06,
+      "loss": 1.34720068,
+      "memory(GiB)": 113.67,
+      "step": 27900,
+      "train_speed(iter/s)": 1.31884
+    },
+    {
+      "acc": 0.67605457,
+      "epoch": 0.7078893962455606,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.679905983415861e-06,
+      "loss": 1.43189287,
+      "memory(GiB)": 113.67,
+      "step": 27905,
+      "train_speed(iter/s)": 1.318868
+    },
+    {
+      "acc": 0.69639058,
+      "epoch": 0.7080162354134957,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.6790206450009e-06,
+      "loss": 1.3919426,
+      "memory(GiB)": 113.67,
+      "step": 27910,
+      "train_speed(iter/s)": 1.318896
+    },
+    {
+      "acc": 0.67460432,
+      "epoch": 0.7081430745814308,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.678135188750648e-06,
+      "loss": 1.39339657,
+      "memory(GiB)": 113.67,
+      "step": 27915,
+      "train_speed(iter/s)": 1.318923
+    },
+    {
+      "acc": 0.68638415,
+      "epoch": 0.7082699137493658,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.677249614704057e-06,
+      "loss": 1.36996479,
+      "memory(GiB)": 113.67,
+      "step": 27920,
+      "train_speed(iter/s)": 1.318951
+    },
+    {
+      "acc": 0.69808826,
+      "epoch": 0.7083967529173009,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.676363922900073e-06,
+      "loss": 1.36486225,
+      "memory(GiB)": 113.67,
+      "step": 27925,
+      "train_speed(iter/s)": 1.31898
+    },
+    {
+      "acc": 0.67365732,
+      "epoch": 0.708523592085236,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.675478113377653e-06,
+      "loss": 1.46477814,
+      "memory(GiB)": 113.67,
+      "step": 27930,
+      "train_speed(iter/s)": 1.319005
+    },
+    {
+      "acc": 0.68811312,
+      "epoch": 0.708650431253171,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.674592186175762e-06,
+      "loss": 1.38040466,
+      "memory(GiB)": 113.67,
+      "step": 27935,
+      "train_speed(iter/s)": 1.319033
+    },
+    {
+      "acc": 0.6870575,
+      "epoch": 0.708777270421106,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.673706141333365e-06,
+      "loss": 1.2902935,
+      "memory(GiB)": 113.67,
+      "step": 27940,
+      "train_speed(iter/s)": 1.319061
+    },
+    {
+      "acc": 0.68656645,
+      "epoch": 0.708904109589041,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.672819978889435e-06,
+      "loss": 1.42830343,
+      "memory(GiB)": 113.67,
+      "step": 27945,
+      "train_speed(iter/s)": 1.31909
+    },
+    {
+      "acc": 0.6786582,
+      "epoch": 0.7090309487569761,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.67193369888295e-06,
+      "loss": 1.42749596,
+      "memory(GiB)": 113.67,
+      "step": 27950,
+      "train_speed(iter/s)": 1.319117
+    },
+    {
+      "acc": 0.69350233,
+      "epoch": 0.7091577879249112,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.67104730135289e-06,
+      "loss": 1.32671261,
+      "memory(GiB)": 113.67,
+      "step": 27955,
+      "train_speed(iter/s)": 1.319145
+    },
+    {
+      "acc": 0.69419241,
+      "epoch": 0.7092846270928462,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.670160786338246e-06,
+      "loss": 1.41832581,
+      "memory(GiB)": 113.67,
+      "step": 27960,
+      "train_speed(iter/s)": 1.319174
+    },
+    {
+      "acc": 0.68252201,
+      "epoch": 0.7094114662607813,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.669274153878006e-06,
+      "loss": 1.37967472,
+      "memory(GiB)": 113.67,
+      "step": 27965,
+      "train_speed(iter/s)": 1.319203
+    },
+    {
+      "acc": 0.68092136,
+      "epoch": 0.7095383054287164,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.668387404011176e-06,
+      "loss": 1.4443841,
+      "memory(GiB)": 113.67,
+      "step": 27970,
+      "train_speed(iter/s)": 1.319231
+    },
+    {
+      "acc": 0.671628,
+      "epoch": 0.7096651445966514,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.667500536776748e-06,
+      "loss": 1.46755362,
+      "memory(GiB)": 113.67,
+      "step": 27975,
+      "train_speed(iter/s)": 1.31926
+    },
+    {
+      "acc": 0.67784066,
+      "epoch": 0.7097919837645865,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.666613552213742e-06,
+      "loss": 1.4057765,
+      "memory(GiB)": 113.67,
+      "step": 27980,
+      "train_speed(iter/s)": 1.31929
+    },
+    {
+      "acc": 0.69541812,
+      "epoch": 0.7099188229325215,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.665726450361165e-06,
+      "loss": 1.3733057,
+      "memory(GiB)": 113.67,
+      "step": 27985,
+      "train_speed(iter/s)": 1.319319
+    },
+    {
+      "acc": 0.69440575,
+      "epoch": 0.7100456621004566,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.664839231258036e-06,
+      "loss": 1.37803707,
+      "memory(GiB)": 113.67,
+      "step": 27990,
+      "train_speed(iter/s)": 1.319344
+    },
+    {
+      "acc": 0.69219227,
+      "epoch": 0.7101725012683917,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.663951894943383e-06,
+      "loss": 1.43933449,
+      "memory(GiB)": 113.67,
+      "step": 27995,
+      "train_speed(iter/s)": 1.319375
+    },
+    {
+      "acc": 0.67304573,
+      "epoch": 0.7102993404363267,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.66306444145623e-06,
+      "loss": 1.41616402,
+      "memory(GiB)": 113.67,
+      "step": 28000,
+      "train_speed(iter/s)": 1.319403
+    },
+    {
+      "epoch": 0.7102993404363267,
+      "eval_acc": 0.6743658336413585,
+      "eval_loss": 1.360344648361206,
+      "eval_runtime": 69.5146,
+      "eval_samples_per_second": 91.635,
+      "eval_steps_per_second": 22.916,
+      "step": 28000
+    },
+    {
+      "acc": 0.67581434,
+      "epoch": 0.7104261796042618,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.662176870835614e-06,
+      "loss": 1.42849331,
+      "memory(GiB)": 113.67,
+      "step": 28005,
+      "train_speed(iter/s)": 1.314351
+    },
+    {
+      "acc": 0.69331455,
+      "epoch": 0.7105530187721969,
+      "grad_norm": 3.0,
+      "learning_rate": 7.661289183120572e-06,
+      "loss": 1.36896143,
+      "memory(GiB)": 113.67,
+      "step": 28010,
+      "train_speed(iter/s)": 1.314377
+    },
+    {
+      "acc": 0.6867599,
+      "epoch": 0.7106798579401319,
+      "grad_norm": 3.375,
+      "learning_rate": 7.66040137835015e-06,
+      "loss": 1.41566133,
+      "memory(GiB)": 113.67,
+      "step": 28015,
+      "train_speed(iter/s)": 1.314403
+    },
+    {
+      "acc": 0.67013555,
+      "epoch": 0.710806697108067,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.659513456563399e-06,
+      "loss": 1.47929001,
+      "memory(GiB)": 113.67,
+      "step": 28020,
+      "train_speed(iter/s)": 1.31443
+    },
+    {
+      "acc": 0.69632826,
+      "epoch": 0.710933536276002,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.658625417799372e-06,
+      "loss": 1.36408377,
+      "memory(GiB)": 113.67,
+      "step": 28025,
+      "train_speed(iter/s)": 1.314457
+    },
+    {
+      "acc": 0.69015245,
+      "epoch": 0.7110603754439371,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.657737262097128e-06,
+      "loss": 1.37425957,
+      "memory(GiB)": 113.67,
+      "step": 28030,
+      "train_speed(iter/s)": 1.314485
+    },
+    {
+      "acc": 0.688764,
+      "epoch": 0.7111872146118722,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.656848989495733e-06,
+      "loss": 1.38991375,
+      "memory(GiB)": 113.67,
+      "step": 28035,
+      "train_speed(iter/s)": 1.314507
+    },
+    {
+      "acc": 0.69881067,
+      "epoch": 0.7113140537798072,
+      "grad_norm": 3.125,
+      "learning_rate": 7.65596060003426e-06,
+      "loss": 1.41583443,
+      "memory(GiB)": 113.67,
+      "step": 28040,
+      "train_speed(iter/s)": 1.31453
+    },
+    {
+      "acc": 0.67271757,
+      "epoch": 0.7114408929477423,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.655072093751779e-06,
+      "loss": 1.48297291,
+      "memory(GiB)": 113.67,
+      "step": 28045,
+      "train_speed(iter/s)": 1.314557
+    },
+    {
+      "acc": 0.68654151,
+      "epoch": 0.7115677321156774,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.654183470687375e-06,
+      "loss": 1.41374903,
+      "memory(GiB)": 113.67,
+      "step": 28050,
+      "train_speed(iter/s)": 1.314582
+    },
+    {
+      "acc": 0.69688911,
+      "epoch": 0.7116945712836124,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.653294730880131e-06,
+      "loss": 1.3453373,
+      "memory(GiB)": 113.67,
+      "step": 28055,
+      "train_speed(iter/s)": 1.314609
+    },
+    {
+      "acc": 0.69447727,
+      "epoch": 0.7118214104515475,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.65240587436914e-06,
+      "loss": 1.3104847,
+      "memory(GiB)": 113.67,
+      "step": 28060,
+      "train_speed(iter/s)": 1.314637
+    },
+    {
+      "acc": 0.68504353,
+      "epoch": 0.7119482496194824,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.651516901193494e-06,
+      "loss": 1.45516739,
+      "memory(GiB)": 113.67,
+      "step": 28065,
+      "train_speed(iter/s)": 1.314664
+    },
+    {
+      "acc": 0.69475393,
+      "epoch": 0.7120750887874175,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.650627811392298e-06,
+      "loss": 1.34371166,
+      "memory(GiB)": 113.67,
+      "step": 28070,
+      "train_speed(iter/s)": 1.314692
+    },
+    {
+      "acc": 0.66686907,
+      "epoch": 0.7122019279553526,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.649738605004658e-06,
+      "loss": 1.42724619,
+      "memory(GiB)": 113.67,
+      "step": 28075,
+      "train_speed(iter/s)": 1.31472
+    },
+    {
+      "acc": 0.69354696,
+      "epoch": 0.7123287671232876,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.648849282069682e-06,
+      "loss": 1.30783405,
+      "memory(GiB)": 113.67,
+      "step": 28080,
+      "train_speed(iter/s)": 1.314742
+    },
+    {
+      "acc": 0.68666449,
+      "epoch": 0.7124556062912227,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.647959842626489e-06,
+      "loss": 1.43563499,
+      "memory(GiB)": 113.67,
+      "step": 28085,
+      "train_speed(iter/s)": 1.31477
+    },
+    {
+      "acc": 0.67747149,
+      "epoch": 0.7125824454591578,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.6470702867142e-06,
+      "loss": 1.42325935,
+      "memory(GiB)": 113.67,
+      "step": 28090,
+      "train_speed(iter/s)": 1.314796
+    },
+    {
+      "acc": 0.67512412,
+      "epoch": 0.7127092846270928,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.646180614371941e-06,
+      "loss": 1.42437572,
+      "memory(GiB)": 113.67,
+      "step": 28095,
+      "train_speed(iter/s)": 1.314824
+    },
+    {
+      "acc": 0.68749862,
+      "epoch": 0.7128361237950279,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.645290825638845e-06,
+      "loss": 1.42953577,
+      "memory(GiB)": 113.67,
+      "step": 28100,
+      "train_speed(iter/s)": 1.314852
+    },
+    {
+      "acc": 0.68652024,
+      "epoch": 0.7129629629629629,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.644400920554048e-06,
+      "loss": 1.39634314,
+      "memory(GiB)": 113.67,
+      "step": 28105,
+      "train_speed(iter/s)": 1.314879
+    },
+    {
+      "acc": 0.68510942,
+      "epoch": 0.713089802130898,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.64351089915669e-06,
+      "loss": 1.44771357,
+      "memory(GiB)": 113.67,
+      "step": 28110,
+      "train_speed(iter/s)": 1.314906
+    },
+    {
+      "acc": 0.70866365,
+      "epoch": 0.7132166412988331,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.642620761485921e-06,
+      "loss": 1.29843273,
+      "memory(GiB)": 113.67,
+      "step": 28115,
+      "train_speed(iter/s)": 1.314933
+    },
+    {
+      "acc": 0.68651733,
+      "epoch": 0.7133434804667681,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.641730507580896e-06,
+      "loss": 1.38678875,
+      "memory(GiB)": 113.67,
+      "step": 28120,
+      "train_speed(iter/s)": 1.314959
+    },
+    {
+      "acc": 0.67234411,
+      "epoch": 0.7134703196347032,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.640840137480763e-06,
+      "loss": 1.45184317,
+      "memory(GiB)": 113.67,
+      "step": 28125,
+      "train_speed(iter/s)": 1.314987
+    },
+    {
+      "acc": 0.68462,
+      "epoch": 0.7135971588026383,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.639949651224697e-06,
+      "loss": 1.35463982,
+      "memory(GiB)": 113.67,
+      "step": 28130,
+      "train_speed(iter/s)": 1.315014
+    },
+    {
+      "acc": 0.67793689,
+      "epoch": 0.7137239979705733,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.639059048851853e-06,
+      "loss": 1.45797901,
+      "memory(GiB)": 113.67,
+      "step": 28135,
+      "train_speed(iter/s)": 1.315042
+    },
+    {
+      "acc": 0.69242887,
+      "epoch": 0.7138508371385084,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.638168330401412e-06,
+      "loss": 1.34812298,
+      "memory(GiB)": 113.67,
+      "step": 28140,
+      "train_speed(iter/s)": 1.31507
+    },
+    {
+      "acc": 0.68774128,
+      "epoch": 0.7139776763064434,
+      "grad_norm": 3.96875,
+      "learning_rate": 7.637277495912548e-06,
+      "loss": 1.35317574,
+      "memory(GiB)": 113.67,
+      "step": 28145,
+      "train_speed(iter/s)": 1.315097
+    },
+    {
+      "acc": 0.69592953,
+      "epoch": 0.7141045154743785,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.636386545424447e-06,
+      "loss": 1.32545872,
+      "memory(GiB)": 113.67,
+      "step": 28150,
+      "train_speed(iter/s)": 1.315125
+    },
+    {
+      "acc": 0.67896194,
+      "epoch": 0.7142313546423136,
+      "grad_norm": 3.5,
+      "learning_rate": 7.635495478976294e-06,
+      "loss": 1.41255398,
+      "memory(GiB)": 113.67,
+      "step": 28155,
+      "train_speed(iter/s)": 1.315152
+    },
+    {
+      "acc": 0.6790175,
+      "epoch": 0.7143581938102486,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.634604296607285e-06,
+      "loss": 1.47762518,
+      "memory(GiB)": 113.67,
+      "step": 28160,
+      "train_speed(iter/s)": 1.31518
+    },
+    {
+      "acc": 0.69893236,
+      "epoch": 0.7144850329781837,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.633712998356612e-06,
+      "loss": 1.30239496,
+      "memory(GiB)": 113.67,
+      "step": 28165,
+      "train_speed(iter/s)": 1.315197
+    },
+    {
+      "acc": 0.67743244,
+      "epoch": 0.7146118721461188,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.632821584263486e-06,
+      "loss": 1.43080978,
+      "memory(GiB)": 113.67,
+      "step": 28170,
+      "train_speed(iter/s)": 1.315224
+    },
+    {
+      "acc": 0.69802666,
+      "epoch": 0.7147387113140538,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.631930054367112e-06,
+      "loss": 1.37100067,
+      "memory(GiB)": 113.67,
+      "step": 28175,
+      "train_speed(iter/s)": 1.315253
+    },
+    {
+      "acc": 0.69228206,
+      "epoch": 0.7148655504819889,
+      "grad_norm": 2.875,
+      "learning_rate": 7.631038408706703e-06,
+      "loss": 1.37321472,
+      "memory(GiB)": 113.67,
+      "step": 28180,
+      "train_speed(iter/s)": 1.315281
+    },
+    {
+      "acc": 0.67829876,
+      "epoch": 0.7149923896499238,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.630146647321476e-06,
+      "loss": 1.38750486,
+      "memory(GiB)": 113.67,
+      "step": 28185,
+      "train_speed(iter/s)": 1.315307
+    },
+    {
+      "acc": 0.69169884,
+      "epoch": 0.7151192288178589,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.62925477025066e-06,
+      "loss": 1.46211119,
+      "memory(GiB)": 113.67,
+      "step": 28190,
+      "train_speed(iter/s)": 1.315335
+    },
+    {
+      "acc": 0.67972174,
+      "epoch": 0.715246067985794,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.628362777533479e-06,
+      "loss": 1.45106125,
+      "memory(GiB)": 113.67,
+      "step": 28195,
+      "train_speed(iter/s)": 1.315362
+    },
+    {
+      "acc": 0.68465176,
+      "epoch": 0.715372907153729,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.627470669209169e-06,
+      "loss": 1.38244972,
+      "memory(GiB)": 113.67,
+      "step": 28200,
+      "train_speed(iter/s)": 1.31539
+    },
+    {
+      "acc": 0.69406242,
+      "epoch": 0.7154997463216641,
+      "grad_norm": 2.875,
+      "learning_rate": 7.626578445316968e-06,
+      "loss": 1.36740284,
+      "memory(GiB)": 113.67,
+      "step": 28205,
+      "train_speed(iter/s)": 1.315417
+    },
+    {
+      "acc": 0.6850544,
+      "epoch": 0.7156265854895992,
+      "grad_norm": 3.625,
+      "learning_rate": 7.62568610589612e-06,
+      "loss": 1.41809349,
+      "memory(GiB)": 113.67,
+      "step": 28210,
+      "train_speed(iter/s)": 1.315446
+    },
+    {
+      "acc": 0.68600092,
+      "epoch": 0.7157534246575342,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.624793650985873e-06,
+      "loss": 1.40926094,
+      "memory(GiB)": 113.67,
+      "step": 28215,
+      "train_speed(iter/s)": 1.315474
+    },
+    {
+      "acc": 0.68142223,
+      "epoch": 0.7158802638254693,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.6239010806254835e-06,
+      "loss": 1.41519928,
+      "memory(GiB)": 113.67,
+      "step": 28220,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.67732992,
+      "epoch": 0.7160071029934043,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.6230083948542084e-06,
+      "loss": 1.42409592,
+      "memory(GiB)": 113.67,
+      "step": 28225,
+      "train_speed(iter/s)": 1.315529
+    },
+    {
+      "acc": 0.6964417,
+      "epoch": 0.7161339421613394,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.622115593711314e-06,
+      "loss": 1.32776651,
+      "memory(GiB)": 113.67,
+      "step": 28230,
+      "train_speed(iter/s)": 1.315557
+    },
+    {
+      "acc": 0.68633318,
+      "epoch": 0.7162607813292745,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.62122267723607e-06,
+      "loss": 1.37006016,
+      "memory(GiB)": 113.67,
+      "step": 28235,
+      "train_speed(iter/s)": 1.315584
+    },
+    {
+      "acc": 0.69253178,
+      "epoch": 0.7163876204972095,
+      "grad_norm": 2.625,
+      "learning_rate": 7.620329645467748e-06,
+      "loss": 1.30672188,
+      "memory(GiB)": 113.67,
+      "step": 28240,
+      "train_speed(iter/s)": 1.315611
+    },
+    {
+      "acc": 0.67048078,
+      "epoch": 0.7165144596651446,
+      "grad_norm": 3.125,
+      "learning_rate": 7.61943649844563e-06,
+      "loss": 1.51315956,
+      "memory(GiB)": 113.67,
+      "step": 28245,
+      "train_speed(iter/s)": 1.315639
+    },
+    {
+      "acc": 0.6723495,
+      "epoch": 0.7166412988330797,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.618543236209001e-06,
+      "loss": 1.46763611,
+      "memory(GiB)": 113.67,
+      "step": 28250,
+      "train_speed(iter/s)": 1.315666
+    },
+    {
+      "acc": 0.68256607,
+      "epoch": 0.7167681380010147,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.617649858797147e-06,
+      "loss": 1.43828335,
+      "memory(GiB)": 113.67,
+      "step": 28255,
+      "train_speed(iter/s)": 1.315695
+    },
+    {
+      "acc": 0.69337158,
+      "epoch": 0.7168949771689498,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.616756366249367e-06,
+      "loss": 1.40005293,
+      "memory(GiB)": 113.67,
+      "step": 28260,
+      "train_speed(iter/s)": 1.315721
+    },
+    {
+      "acc": 0.67453508,
+      "epoch": 0.7170218163368848,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.6158627586049586e-06,
+      "loss": 1.46861553,
+      "memory(GiB)": 113.67,
+      "step": 28265,
+      "train_speed(iter/s)": 1.315748
+    },
+    {
+      "acc": 0.69516492,
+      "epoch": 0.7171486555048199,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.614969035903228e-06,
+      "loss": 1.31988039,
+      "memory(GiB)": 113.67,
+      "step": 28270,
+      "train_speed(iter/s)": 1.315775
+    },
+    {
+      "acc": 0.6891221,
+      "epoch": 0.717275494672755,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.614075198183482e-06,
+      "loss": 1.34310055,
+      "memory(GiB)": 113.67,
+      "step": 28275,
+      "train_speed(iter/s)": 1.315803
+    },
+    {
+      "acc": 0.68297644,
+      "epoch": 0.71740233384069,
+      "grad_norm": 4.375,
+      "learning_rate": 7.6131812454850406e-06,
+      "loss": 1.38366146,
+      "memory(GiB)": 113.67,
+      "step": 28280,
+      "train_speed(iter/s)": 1.31583
+    },
+    {
+      "acc": 0.68383522,
+      "epoch": 0.7175291730086251,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.612287177847219e-06,
+      "loss": 1.41361332,
+      "memory(GiB)": 113.67,
+      "step": 28285,
+      "train_speed(iter/s)": 1.315856
+    },
+    {
+      "acc": 0.68586845,
+      "epoch": 0.7176560121765602,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.611392995309345e-06,
+      "loss": 1.40696354,
+      "memory(GiB)": 113.67,
+      "step": 28290,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.68923655,
+      "epoch": 0.7177828513444952,
+      "grad_norm": 3.5,
+      "learning_rate": 7.610498697910748e-06,
+      "loss": 1.35444927,
+      "memory(GiB)": 113.67,
+      "step": 28295,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.68788257,
+      "epoch": 0.7179096905124303,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.609604285690762e-06,
+      "loss": 1.3922184,
+      "memory(GiB)": 113.67,
+      "step": 28300,
+      "train_speed(iter/s)": 1.315934
+    },
+    {
+      "acc": 0.70036316,
+      "epoch": 0.7180365296803652,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.608709758688731e-06,
+      "loss": 1.34520645,
+      "memory(GiB)": 113.67,
+      "step": 28305,
+      "train_speed(iter/s)": 1.315961
+    },
+    {
+      "acc": 0.67863655,
+      "epoch": 0.7181633688483003,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.607815116943995e-06,
+      "loss": 1.35047169,
+      "memory(GiB)": 113.67,
+      "step": 28310,
+      "train_speed(iter/s)": 1.315989
+    },
+    {
+      "acc": 0.68938932,
+      "epoch": 0.7182902080162354,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.606920360495908e-06,
+      "loss": 1.34161224,
+      "memory(GiB)": 113.67,
+      "step": 28315,
+      "train_speed(iter/s)": 1.316017
+    },
+    {
+      "acc": 0.67325191,
+      "epoch": 0.7184170471841704,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.6060254893838255e-06,
+      "loss": 1.43722849,
+      "memory(GiB)": 113.67,
+      "step": 28320,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.67615862,
+      "epoch": 0.7185438863521055,
+      "grad_norm": 3.0,
+      "learning_rate": 7.6051305036471065e-06,
+      "loss": 1.41424732,
+      "memory(GiB)": 113.67,
+      "step": 28325,
+      "train_speed(iter/s)": 1.316069
+    },
+    {
+      "acc": 0.70140972,
+      "epoch": 0.7186707255200406,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.604235403325117e-06,
+      "loss": 1.32327232,
+      "memory(GiB)": 113.67,
+      "step": 28330,
+      "train_speed(iter/s)": 1.316095
+    },
+    {
+      "acc": 0.68459873,
+      "epoch": 0.7187975646879756,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.603340188457227e-06,
+      "loss": 1.43587675,
+      "memory(GiB)": 113.67,
+      "step": 28335,
+      "train_speed(iter/s)": 1.316123
+    },
+    {
+      "acc": 0.67744198,
+      "epoch": 0.7189244038559107,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.602444859082814e-06,
+      "loss": 1.42434855,
+      "memory(GiB)": 113.67,
+      "step": 28340,
+      "train_speed(iter/s)": 1.316151
+    },
+    {
+      "acc": 0.68736763,
+      "epoch": 0.7190512430238457,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.601549415241254e-06,
+      "loss": 1.3698225,
+      "memory(GiB)": 113.67,
+      "step": 28345,
+      "train_speed(iter/s)": 1.31618
+    },
+    {
+      "acc": 0.68573527,
+      "epoch": 0.7191780821917808,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.6006538569719375e-06,
+      "loss": 1.36319771,
+      "memory(GiB)": 113.67,
+      "step": 28350,
+      "train_speed(iter/s)": 1.316199
+    },
+    {
+      "acc": 0.67842417,
+      "epoch": 0.7193049213597159,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.599758184314252e-06,
+      "loss": 1.39513941,
+      "memory(GiB)": 113.67,
+      "step": 28355,
+      "train_speed(iter/s)": 1.316225
+    },
+    {
+      "acc": 0.67856579,
+      "epoch": 0.7194317605276509,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.598862397307596e-06,
+      "loss": 1.40068359,
+      "memory(GiB)": 113.67,
+      "step": 28360,
+      "train_speed(iter/s)": 1.316252
+    },
+    {
+      "acc": 0.71759777,
+      "epoch": 0.719558599695586,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.597966495991368e-06,
+      "loss": 1.28721838,
+      "memory(GiB)": 113.67,
+      "step": 28365,
+      "train_speed(iter/s)": 1.316281
+    },
+    {
+      "acc": 0.70039196,
+      "epoch": 0.7196854388635211,
+      "grad_norm": 3.0,
+      "learning_rate": 7.597070480404974e-06,
+      "loss": 1.29254179,
+      "memory(GiB)": 113.67,
+      "step": 28370,
+      "train_speed(iter/s)": 1.316308
+    },
+    {
+      "acc": 0.68530049,
+      "epoch": 0.7198122780314561,
+      "grad_norm": 2.625,
+      "learning_rate": 7.596174350587826e-06,
+      "loss": 1.37398853,
+      "memory(GiB)": 113.67,
+      "step": 28375,
+      "train_speed(iter/s)": 1.316335
+    },
+    {
+      "acc": 0.67389679,
+      "epoch": 0.7199391171993912,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.595278106579339e-06,
+      "loss": 1.40331583,
+      "memory(GiB)": 113.67,
+      "step": 28380,
+      "train_speed(iter/s)": 1.316362
+    },
+    {
+      "acc": 0.69207373,
+      "epoch": 0.7200659563673262,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.594381748418933e-06,
+      "loss": 1.3572464,
+      "memory(GiB)": 113.67,
+      "step": 28385,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.6831358,
+      "epoch": 0.7201927955352613,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.593485276146035e-06,
+      "loss": 1.3651432,
+      "memory(GiB)": 113.67,
+      "step": 28390,
+      "train_speed(iter/s)": 1.316417
+    },
+    {
+      "acc": 0.69980726,
+      "epoch": 0.7203196347031964,
+      "grad_norm": 4.84375,
+      "learning_rate": 7.592588689800077e-06,
+      "loss": 1.36411419,
+      "memory(GiB)": 113.67,
+      "step": 28395,
+      "train_speed(iter/s)": 1.316444
+    },
+    {
+      "acc": 0.67862997,
+      "epoch": 0.7204464738711314,
+      "grad_norm": 3.125,
+      "learning_rate": 7.591691989420491e-06,
+      "loss": 1.45533104,
+      "memory(GiB)": 113.67,
+      "step": 28400,
+      "train_speed(iter/s)": 1.316472
+    },
+    {
+      "acc": 0.69785786,
+      "epoch": 0.7205733130390665,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.590795175046721e-06,
+      "loss": 1.35623646,
+      "memory(GiB)": 113.67,
+      "step": 28405,
+      "train_speed(iter/s)": 1.316498
+    },
+    {
+      "acc": 0.68533697,
+      "epoch": 0.7207001522070016,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.5898982467182125e-06,
+      "loss": 1.35961838,
+      "memory(GiB)": 113.67,
+      "step": 28410,
+      "train_speed(iter/s)": 1.316526
+    },
+    {
+      "acc": 0.67883472,
+      "epoch": 0.7208269913749366,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.589001204474416e-06,
+      "loss": 1.39937706,
+      "memory(GiB)": 113.67,
+      "step": 28415,
+      "train_speed(iter/s)": 1.31655
+    },
+    {
+      "acc": 0.66790133,
+      "epoch": 0.7209538305428717,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.588104048354787e-06,
+      "loss": 1.43167906,
+      "memory(GiB)": 113.67,
+      "step": 28420,
+      "train_speed(iter/s)": 1.316578
+    },
+    {
+      "acc": 0.68442521,
+      "epoch": 0.7210806697108066,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.587206778398788e-06,
+      "loss": 1.37180538,
+      "memory(GiB)": 113.67,
+      "step": 28425,
+      "train_speed(iter/s)": 1.316606
+    },
+    {
+      "acc": 0.6996583,
+      "epoch": 0.7212075088787417,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.586309394645882e-06,
+      "loss": 1.33576231,
+      "memory(GiB)": 113.67,
+      "step": 28430,
+      "train_speed(iter/s)": 1.316634
+    },
+    {
+      "acc": 0.70030632,
+      "epoch": 0.7213343480466768,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.585411897135544e-06,
+      "loss": 1.31117086,
+      "memory(GiB)": 113.67,
+      "step": 28435,
+      "train_speed(iter/s)": 1.316662
+    },
+    {
+      "acc": 0.69222841,
+      "epoch": 0.7214611872146118,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.584514285907245e-06,
+      "loss": 1.30658875,
+      "memory(GiB)": 113.67,
+      "step": 28440,
+      "train_speed(iter/s)": 1.316689
+    },
+    {
+      "acc": 0.69856987,
+      "epoch": 0.7215880263825469,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.583616561000471e-06,
+      "loss": 1.38560123,
+      "memory(GiB)": 113.67,
+      "step": 28445,
+      "train_speed(iter/s)": 1.316716
+    },
+    {
+      "acc": 0.67587957,
+      "epoch": 0.721714865550482,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.582718722454705e-06,
+      "loss": 1.43005924,
+      "memory(GiB)": 113.67,
+      "step": 28450,
+      "train_speed(iter/s)": 1.316741
+    },
+    {
+      "acc": 0.68013434,
+      "epoch": 0.721841704718417,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.581820770309438e-06,
+      "loss": 1.43567686,
+      "memory(GiB)": 113.67,
+      "step": 28455,
+      "train_speed(iter/s)": 1.316769
+    },
+    {
+      "acc": 0.70403047,
+      "epoch": 0.7219685438863521,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.580922704604168e-06,
+      "loss": 1.33306894,
+      "memory(GiB)": 113.67,
+      "step": 28460,
+      "train_speed(iter/s)": 1.316796
+    },
+    {
+      "acc": 0.67367401,
+      "epoch": 0.7220953830542871,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.5800245253783935e-06,
+      "loss": 1.40674343,
+      "memory(GiB)": 113.67,
+      "step": 28465,
+      "train_speed(iter/s)": 1.316824
+    },
+    {
+      "acc": 0.66891413,
+      "epoch": 0.7222222222222222,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.579126232671621e-06,
+      "loss": 1.4170804,
+      "memory(GiB)": 113.67,
+      "step": 28470,
+      "train_speed(iter/s)": 1.316851
+    },
+    {
+      "acc": 0.68614564,
+      "epoch": 0.7223490613901573,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.578227826523361e-06,
+      "loss": 1.43361959,
+      "memory(GiB)": 113.67,
+      "step": 28475,
+      "train_speed(iter/s)": 1.316878
+    },
+    {
+      "acc": 0.70228477,
+      "epoch": 0.7224759005580923,
+      "grad_norm": 2.875,
+      "learning_rate": 7.577329306973132e-06,
+      "loss": 1.36634865,
+      "memory(GiB)": 113.67,
+      "step": 28480,
+      "train_speed(iter/s)": 1.316905
+    },
+    {
+      "acc": 0.69447298,
+      "epoch": 0.7226027397260274,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.576430674060452e-06,
+      "loss": 1.34109726,
+      "memory(GiB)": 113.67,
+      "step": 28485,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.67514496,
+      "epoch": 0.7227295788939625,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.575531927824849e-06,
+      "loss": 1.44992266,
+      "memory(GiB)": 113.67,
+      "step": 28490,
+      "train_speed(iter/s)": 1.31696
+    },
+    {
+      "acc": 0.69377136,
+      "epoch": 0.7228564180618975,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.574633068305852e-06,
+      "loss": 1.43740759,
+      "memory(GiB)": 113.67,
+      "step": 28495,
+      "train_speed(iter/s)": 1.316984
+    },
+    {
+      "acc": 0.68900967,
+      "epoch": 0.7229832572298326,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.5737340955429995e-06,
+      "loss": 1.40089273,
+      "memory(GiB)": 113.67,
+      "step": 28500,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.69681611,
+      "epoch": 0.7231100963977676,
+      "grad_norm": 3.125,
+      "learning_rate": 7.572835009575828e-06,
+      "loss": 1.32014999,
+      "memory(GiB)": 113.67,
+      "step": 28505,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.68623471,
+      "epoch": 0.7232369355657027,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.571935810443886e-06,
+      "loss": 1.38731871,
+      "memory(GiB)": 113.67,
+      "step": 28510,
+      "train_speed(iter/s)": 1.317061
+    },
+    {
+      "acc": 0.69446449,
+      "epoch": 0.7233637747336378,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.571036498186727e-06,
+      "loss": 1.35313473,
+      "memory(GiB)": 113.67,
+      "step": 28515,
+      "train_speed(iter/s)": 1.317087
+    },
+    {
+      "acc": 0.67868071,
+      "epoch": 0.7234906139015728,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.570137072843902e-06,
+      "loss": 1.44731455,
+      "memory(GiB)": 113.67,
+      "step": 28520,
+      "train_speed(iter/s)": 1.317112
+    },
+    {
+      "acc": 0.68061719,
+      "epoch": 0.7236174530695079,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.569237534454974e-06,
+      "loss": 1.41972008,
+      "memory(GiB)": 113.67,
+      "step": 28525,
+      "train_speed(iter/s)": 1.317137
+    },
+    {
+      "acc": 0.67026267,
+      "epoch": 0.723744292237443,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.568337883059509e-06,
+      "loss": 1.45642405,
+      "memory(GiB)": 113.67,
+      "step": 28530,
+      "train_speed(iter/s)": 1.317161
+    },
+    {
+      "acc": 0.68078833,
+      "epoch": 0.723871131405378,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.567438118697077e-06,
+      "loss": 1.42540684,
+      "memory(GiB)": 113.67,
+      "step": 28535,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.68207326,
+      "epoch": 0.723997970573313,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.566538241407253e-06,
+      "loss": 1.35546989,
+      "memory(GiB)": 113.67,
+      "step": 28540,
+      "train_speed(iter/s)": 1.317213
+    },
+    {
+      "acc": 0.67658396,
+      "epoch": 0.724124809741248,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.565638251229617e-06,
+      "loss": 1.40187874,
+      "memory(GiB)": 113.67,
+      "step": 28545,
+      "train_speed(iter/s)": 1.317237
+    },
+    {
+      "acc": 0.69557314,
+      "epoch": 0.7242516489091831,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.5647381482037585e-06,
+      "loss": 1.32550507,
+      "memory(GiB)": 113.67,
+      "step": 28550,
+      "train_speed(iter/s)": 1.317264
+    },
+    {
+      "acc": 0.69411912,
+      "epoch": 0.7243784880771182,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.563837932369264e-06,
+      "loss": 1.40968914,
+      "memory(GiB)": 113.67,
+      "step": 28555,
+      "train_speed(iter/s)": 1.317289
+    },
+    {
+      "acc": 0.6892199,
+      "epoch": 0.7245053272450532,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.562937603765732e-06,
+      "loss": 1.38843155,
+      "memory(GiB)": 113.67,
+      "step": 28560,
+      "train_speed(iter/s)": 1.317313
+    },
+    {
+      "acc": 0.69163227,
+      "epoch": 0.7246321664129883,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.562037162432761e-06,
+      "loss": 1.38459492,
+      "memory(GiB)": 113.67,
+      "step": 28565,
+      "train_speed(iter/s)": 1.317338
+    },
+    {
+      "acc": 0.67930374,
+      "epoch": 0.7247590055809234,
+      "grad_norm": 3.125,
+      "learning_rate": 7.561136608409956e-06,
+      "loss": 1.4007309,
+      "memory(GiB)": 113.67,
+      "step": 28570,
+      "train_speed(iter/s)": 1.317363
+    },
+    {
+      "acc": 0.66813478,
+      "epoch": 0.7248858447488584,
+      "grad_norm": 2.875,
+      "learning_rate": 7.560235941736929e-06,
+      "loss": 1.51427622,
+      "memory(GiB)": 113.67,
+      "step": 28575,
+      "train_speed(iter/s)": 1.317389
+    },
+    {
+      "acc": 0.67870512,
+      "epoch": 0.7250126839167935,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.559335162453294e-06,
+      "loss": 1.4399087,
+      "memory(GiB)": 113.67,
+      "step": 28580,
+      "train_speed(iter/s)": 1.317414
+    },
+    {
+      "acc": 0.69820299,
+      "epoch": 0.7251395230847285,
+      "grad_norm": 4.125,
+      "learning_rate": 7.558434270598672e-06,
+      "loss": 1.38109903,
+      "memory(GiB)": 113.67,
+      "step": 28585,
+      "train_speed(iter/s)": 1.31744
+    },
+    {
+      "acc": 0.68900995,
+      "epoch": 0.7252663622526636,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.5575332662126885e-06,
+      "loss": 1.37718773,
+      "memory(GiB)": 113.67,
+      "step": 28590,
+      "train_speed(iter/s)": 1.317466
+    },
+    {
+      "acc": 0.69226236,
+      "epoch": 0.7253932014205987,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.556632149334975e-06,
+      "loss": 1.34745493,
+      "memory(GiB)": 113.67,
+      "step": 28595,
+      "train_speed(iter/s)": 1.317492
+    },
+    {
+      "acc": 0.6900188,
+      "epoch": 0.7255200405885337,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.555730920005163e-06,
+      "loss": 1.32783546,
+      "memory(GiB)": 113.67,
+      "step": 28600,
+      "train_speed(iter/s)": 1.317517
+    },
+    {
+      "acc": 0.65993366,
+      "epoch": 0.7256468797564688,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.554829578262894e-06,
+      "loss": 1.48267965,
+      "memory(GiB)": 113.67,
+      "step": 28605,
+      "train_speed(iter/s)": 1.317543
+    },
+    {
+      "acc": 0.69072967,
+      "epoch": 0.7257737189244039,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.5539281241478155e-06,
+      "loss": 1.43000965,
+      "memory(GiB)": 113.67,
+      "step": 28610,
+      "train_speed(iter/s)": 1.317569
+    },
+    {
+      "acc": 0.70057449,
+      "epoch": 0.7259005580923389,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.5530265576995756e-06,
+      "loss": 1.33322144,
+      "memory(GiB)": 113.67,
+      "step": 28615,
+      "train_speed(iter/s)": 1.317593
+    },
+    {
+      "acc": 0.6788012,
+      "epoch": 0.726027397260274,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.552124878957829e-06,
+      "loss": 1.44318981,
+      "memory(GiB)": 113.67,
+      "step": 28620,
+      "train_speed(iter/s)": 1.317618
+    },
+    {
+      "acc": 0.68667526,
+      "epoch": 0.726154236428209,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.551223087962234e-06,
+      "loss": 1.38678875,
+      "memory(GiB)": 113.67,
+      "step": 28625,
+      "train_speed(iter/s)": 1.317642
+    },
+    {
+      "acc": 0.703789,
+      "epoch": 0.7262810755961441,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.55032118475246e-06,
+      "loss": 1.31672535,
+      "memory(GiB)": 113.67,
+      "step": 28630,
+      "train_speed(iter/s)": 1.317666
+    },
+    {
+      "acc": 0.68335161,
+      "epoch": 0.7264079147640792,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.549419169368171e-06,
+      "loss": 1.43784084,
+      "memory(GiB)": 113.67,
+      "step": 28635,
+      "train_speed(iter/s)": 1.317693
+    },
+    {
+      "acc": 0.68088732,
+      "epoch": 0.7265347539320142,
+      "grad_norm": 3.375,
+      "learning_rate": 7.548517041849048e-06,
+      "loss": 1.4445612,
+      "memory(GiB)": 113.67,
+      "step": 28640,
+      "train_speed(iter/s)": 1.317718
+    },
+    {
+      "acc": 0.68827729,
+      "epoch": 0.7266615930999493,
+      "grad_norm": 3.0,
+      "learning_rate": 7.547614802234764e-06,
+      "loss": 1.37627439,
+      "memory(GiB)": 113.67,
+      "step": 28645,
+      "train_speed(iter/s)": 1.317743
+    },
+    {
+      "acc": 0.69966068,
+      "epoch": 0.7267884322678844,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.546712450565008e-06,
+      "loss": 1.39351254,
+      "memory(GiB)": 113.67,
+      "step": 28650,
+      "train_speed(iter/s)": 1.317767
+    },
+    {
+      "acc": 0.69907966,
+      "epoch": 0.7269152714358194,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.545809986879469e-06,
+      "loss": 1.30505199,
+      "memory(GiB)": 113.67,
+      "step": 28655,
+      "train_speed(iter/s)": 1.317791
+    },
+    {
+      "acc": 0.68609514,
+      "epoch": 0.7270421106037545,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.5449074112178385e-06,
+      "loss": 1.39254322,
+      "memory(GiB)": 113.67,
+      "step": 28660,
+      "train_speed(iter/s)": 1.317816
+    },
+    {
+      "acc": 0.68440561,
+      "epoch": 0.7271689497716894,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.54400472361982e-06,
+      "loss": 1.4850894,
+      "memory(GiB)": 113.67,
+      "step": 28665,
+      "train_speed(iter/s)": 1.317842
+    },
+    {
+      "acc": 0.67435098,
+      "epoch": 0.7272957889396245,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.543101924125115e-06,
+      "loss": 1.40209522,
+      "memory(GiB)": 113.67,
+      "step": 28670,
+      "train_speed(iter/s)": 1.317868
+    },
+    {
+      "acc": 0.67734103,
+      "epoch": 0.7274226281075596,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.542199012773432e-06,
+      "loss": 1.47199898,
+      "memory(GiB)": 113.67,
+      "step": 28675,
+      "train_speed(iter/s)": 1.317893
+    },
+    {
+      "acc": 0.68469663,
+      "epoch": 0.7275494672754946,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.541295989604488e-06,
+      "loss": 1.39049339,
+      "memory(GiB)": 113.67,
+      "step": 28680,
+      "train_speed(iter/s)": 1.317919
+    },
+    {
+      "acc": 0.69778023,
+      "epoch": 0.7276763064434297,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.540392854657999e-06,
+      "loss": 1.33574228,
+      "memory(GiB)": 113.67,
+      "step": 28685,
+      "train_speed(iter/s)": 1.317945
+    },
+    {
+      "acc": 0.69829216,
+      "epoch": 0.7278031456113648,
+      "grad_norm": 3.890625,
+      "learning_rate": 7.539489607973691e-06,
+      "loss": 1.37390442,
+      "memory(GiB)": 113.67,
+      "step": 28690,
+      "train_speed(iter/s)": 1.317971
+    },
+    {
+      "acc": 0.68066559,
+      "epoch": 0.7279299847792998,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.5385862495912905e-06,
+      "loss": 1.47343369,
+      "memory(GiB)": 113.67,
+      "step": 28695,
+      "train_speed(iter/s)": 1.317997
+    },
+    {
+      "acc": 0.68536911,
+      "epoch": 0.7280568239472349,
+      "grad_norm": 3.625,
+      "learning_rate": 7.537682779550537e-06,
+      "loss": 1.41004982,
+      "memory(GiB)": 113.67,
+      "step": 28700,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.6650156,
+      "epoch": 0.7281836631151699,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.536779197891159e-06,
+      "loss": 1.45099773,
+      "memory(GiB)": 113.67,
+      "step": 28705,
+      "train_speed(iter/s)": 1.31805
+    },
+    {
+      "acc": 0.69912462,
+      "epoch": 0.728310502283105,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.535875504652912e-06,
+      "loss": 1.35122967,
+      "memory(GiB)": 113.67,
+      "step": 28710,
+      "train_speed(iter/s)": 1.318077
+    },
+    {
+      "acc": 0.68571844,
+      "epoch": 0.7284373414510401,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.534971699875534e-06,
+      "loss": 1.42168465,
+      "memory(GiB)": 113.67,
+      "step": 28715,
+      "train_speed(iter/s)": 1.318102
+    },
+    {
+      "acc": 0.68486481,
+      "epoch": 0.7285641806189751,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.534067783598784e-06,
+      "loss": 1.38307209,
+      "memory(GiB)": 113.67,
+      "step": 28720,
+      "train_speed(iter/s)": 1.318128
+    },
+    {
+      "acc": 0.66383319,
+      "epoch": 0.7286910197869102,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.533163755862419e-06,
+      "loss": 1.44309692,
+      "memory(GiB)": 113.67,
+      "step": 28725,
+      "train_speed(iter/s)": 1.318154
+    },
+    {
+      "acc": 0.67743592,
+      "epoch": 0.7288178589548453,
+      "grad_norm": 2.875,
+      "learning_rate": 7.5322596167062035e-06,
+      "loss": 1.40313358,
+      "memory(GiB)": 113.67,
+      "step": 28730,
+      "train_speed(iter/s)": 1.318181
+    },
+    {
+      "acc": 0.69368792,
+      "epoch": 0.7289446981227803,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.5313553661699035e-06,
+      "loss": 1.35654116,
+      "memory(GiB)": 113.67,
+      "step": 28735,
+      "train_speed(iter/s)": 1.318204
+    },
+    {
+      "acc": 0.70737295,
+      "epoch": 0.7290715372907154,
+      "grad_norm": 9.6875,
+      "learning_rate": 7.530451004293292e-06,
+      "loss": 1.3127737,
+      "memory(GiB)": 113.67,
+      "step": 28740,
+      "train_speed(iter/s)": 1.318231
+    },
+    {
+      "acc": 0.68333535,
+      "epoch": 0.7291983764586504,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.5295465311161485e-06,
+      "loss": 1.36792469,
+      "memory(GiB)": 113.67,
+      "step": 28745,
+      "train_speed(iter/s)": 1.318257
+    },
+    {
+      "acc": 0.700139,
+      "epoch": 0.7293252156265855,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.5286419466782546e-06,
+      "loss": 1.31104584,
+      "memory(GiB)": 113.67,
+      "step": 28750,
+      "train_speed(iter/s)": 1.318284
+    },
+    {
+      "acc": 0.69273748,
+      "epoch": 0.7294520547945206,
+      "grad_norm": 5.96875,
+      "learning_rate": 7.527737251019399e-06,
+      "loss": 1.40027428,
+      "memory(GiB)": 113.67,
+      "step": 28755,
+      "train_speed(iter/s)": 1.318311
+    },
+    {
+      "acc": 0.66848941,
+      "epoch": 0.7295788939624556,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.526832444179373e-06,
+      "loss": 1.43807144,
+      "memory(GiB)": 113.67,
+      "step": 28760,
+      "train_speed(iter/s)": 1.318336
+    },
+    {
+      "acc": 0.69375153,
+      "epoch": 0.7297057331303907,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.525927526197974e-06,
+      "loss": 1.40778284,
+      "memory(GiB)": 113.67,
+      "step": 28765,
+      "train_speed(iter/s)": 1.318362
+    },
+    {
+      "acc": 0.67284565,
+      "epoch": 0.7298325722983258,
+      "grad_norm": 2.75,
+      "learning_rate": 7.5250224971150065e-06,
+      "loss": 1.42379665,
+      "memory(GiB)": 113.67,
+      "step": 28770,
+      "train_speed(iter/s)": 1.318388
+    },
+    {
+      "acc": 0.69628639,
+      "epoch": 0.7299594114662608,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.524117356970275e-06,
+      "loss": 1.3666852,
+      "memory(GiB)": 113.67,
+      "step": 28775,
+      "train_speed(iter/s)": 1.318414
+    },
+    {
+      "acc": 0.67899303,
+      "epoch": 0.7300862506341959,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.523212105803594e-06,
+      "loss": 1.47996559,
+      "memory(GiB)": 113.67,
+      "step": 28780,
+      "train_speed(iter/s)": 1.31844
+    },
+    {
+      "acc": 0.68565321,
+      "epoch": 0.7302130898021308,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.522306743654777e-06,
+      "loss": 1.37011929,
+      "memory(GiB)": 113.67,
+      "step": 28785,
+      "train_speed(iter/s)": 1.318466
+    },
+    {
+      "acc": 0.69281812,
+      "epoch": 0.7303399289700659,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.521401270563651e-06,
+      "loss": 1.41748962,
+      "memory(GiB)": 113.67,
+      "step": 28790,
+      "train_speed(iter/s)": 1.318492
+    },
+    {
+      "acc": 0.67408071,
+      "epoch": 0.730466768138001,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.520495686570037e-06,
+      "loss": 1.40314636,
+      "memory(GiB)": 113.67,
+      "step": 28795,
+      "train_speed(iter/s)": 1.318518
+    },
+    {
+      "acc": 0.69906797,
+      "epoch": 0.730593607305936,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.5195899917137716e-06,
+      "loss": 1.34877567,
+      "memory(GiB)": 113.67,
+      "step": 28800,
+      "train_speed(iter/s)": 1.318545
+    },
+    {
+      "acc": 0.67699084,
+      "epoch": 0.7307204464738711,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.518684186034688e-06,
+      "loss": 1.35605583,
+      "memory(GiB)": 113.67,
+      "step": 28805,
+      "train_speed(iter/s)": 1.31857
+    },
+    {
+      "acc": 0.69780374,
+      "epoch": 0.7308472856418062,
+      "grad_norm": 2.375,
+      "learning_rate": 7.51777826957263e-06,
+      "loss": 1.38125992,
+      "memory(GiB)": 113.67,
+      "step": 28810,
+      "train_speed(iter/s)": 1.318597
+    },
+    {
+      "acc": 0.68557034,
+      "epoch": 0.7309741248097412,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.516872242367441e-06,
+      "loss": 1.34511757,
+      "memory(GiB)": 113.67,
+      "step": 28815,
+      "train_speed(iter/s)": 1.318622
+    },
+    {
+      "acc": 0.68833671,
+      "epoch": 0.7311009639776763,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.5159661044589745e-06,
+      "loss": 1.42607784,
+      "memory(GiB)": 113.67,
+      "step": 28820,
+      "train_speed(iter/s)": 1.318647
+    },
+    {
+      "acc": 0.69663696,
+      "epoch": 0.7312278031456113,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.515059855887087e-06,
+      "loss": 1.37647877,
+      "memory(GiB)": 113.67,
+      "step": 28825,
+      "train_speed(iter/s)": 1.318673
+    },
+    {
+      "acc": 0.68173971,
+      "epoch": 0.7313546423135464,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.514153496691636e-06,
+      "loss": 1.31814232,
+      "memory(GiB)": 113.67,
+      "step": 28830,
+      "train_speed(iter/s)": 1.318698
+    },
+    {
+      "acc": 0.69550295,
+      "epoch": 0.7314814814814815,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.513247026912491e-06,
+      "loss": 1.30223598,
+      "memory(GiB)": 113.67,
+      "step": 28835,
+      "train_speed(iter/s)": 1.318721
+    },
+    {
+      "acc": 0.69659014,
+      "epoch": 0.7316083206494165,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.512340446589521e-06,
+      "loss": 1.37910538,
+      "memory(GiB)": 113.67,
+      "step": 28840,
+      "train_speed(iter/s)": 1.318748
+    },
+    {
+      "acc": 0.69476156,
+      "epoch": 0.7317351598173516,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.5114337557625985e-06,
+      "loss": 1.36168871,
+      "memory(GiB)": 113.67,
+      "step": 28845,
+      "train_speed(iter/s)": 1.318775
+    },
+    {
+      "acc": 0.70546942,
+      "epoch": 0.7318619989852867,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.510526954471611e-06,
+      "loss": 1.2782692,
+      "memory(GiB)": 113.67,
+      "step": 28850,
+      "train_speed(iter/s)": 1.318801
+    },
+    {
+      "acc": 0.68139262,
+      "epoch": 0.7319888381532217,
+      "grad_norm": 5.46875,
+      "learning_rate": 7.509620042756436e-06,
+      "loss": 1.44304714,
+      "memory(GiB)": 113.67,
+      "step": 28855,
+      "train_speed(iter/s)": 1.318827
+    },
+    {
+      "acc": 0.67598448,
+      "epoch": 0.7321156773211568,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.508713020656968e-06,
+      "loss": 1.45248432,
+      "memory(GiB)": 113.67,
+      "step": 28860,
+      "train_speed(iter/s)": 1.318852
+    },
+    {
+      "acc": 0.67854075,
+      "epoch": 0.7322425164890918,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.5078058882131e-06,
+      "loss": 1.43077412,
+      "memory(GiB)": 113.67,
+      "step": 28865,
+      "train_speed(iter/s)": 1.318879
+    },
+    {
+      "acc": 0.67435899,
+      "epoch": 0.7323693556570269,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.506898645464733e-06,
+      "loss": 1.4265913,
+      "memory(GiB)": 113.67,
+      "step": 28870,
+      "train_speed(iter/s)": 1.318905
+    },
+    {
+      "acc": 0.69003925,
+      "epoch": 0.732496194824962,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.505991292451772e-06,
+      "loss": 1.3197238,
+      "memory(GiB)": 113.67,
+      "step": 28875,
+      "train_speed(iter/s)": 1.31893
+    },
+    {
+      "acc": 0.67996049,
+      "epoch": 0.732623033992897,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.505083829214125e-06,
+      "loss": 1.44947138,
+      "memory(GiB)": 113.67,
+      "step": 28880,
+      "train_speed(iter/s)": 1.318958
+    },
+    {
+      "acc": 0.68029289,
+      "epoch": 0.7327498731608321,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.5041762557917065e-06,
+      "loss": 1.46398411,
+      "memory(GiB)": 113.67,
+      "step": 28885,
+      "train_speed(iter/s)": 1.318983
+    },
+    {
+      "acc": 0.6775526,
+      "epoch": 0.7328767123287672,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.5032685722244355e-06,
+      "loss": 1.45094948,
+      "memory(GiB)": 113.67,
+      "step": 28890,
+      "train_speed(iter/s)": 1.31901
+    },
+    {
+      "acc": 0.67475185,
+      "epoch": 0.7330035514967022,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.502360778552238e-06,
+      "loss": 1.38349972,
+      "memory(GiB)": 113.67,
+      "step": 28895,
+      "train_speed(iter/s)": 1.319037
+    },
+    {
+      "acc": 0.68088236,
+      "epoch": 0.7331303906646373,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.5014528748150405e-06,
+      "loss": 1.43546429,
+      "memory(GiB)": 113.67,
+      "step": 28900,
+      "train_speed(iter/s)": 1.319062
+    },
+    {
+      "acc": 0.67193298,
+      "epoch": 0.7332572298325722,
+      "grad_norm": 3.625,
+      "learning_rate": 7.5005448610527765e-06,
+      "loss": 1.37752428,
+      "memory(GiB)": 113.67,
+      "step": 28905,
+      "train_speed(iter/s)": 1.319088
+    },
+    {
+      "acc": 0.69992018,
+      "epoch": 0.7333840690005073,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.499636737305386e-06,
+      "loss": 1.33727188,
+      "memory(GiB)": 113.67,
+      "step": 28910,
+      "train_speed(iter/s)": 1.319114
+    },
+    {
+      "acc": 0.70848155,
+      "epoch": 0.7335109081684424,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.498728503612811e-06,
+      "loss": 1.33494959,
+      "memory(GiB)": 113.67,
+      "step": 28915,
+      "train_speed(iter/s)": 1.319139
+    },
+    {
+      "acc": 0.69021215,
+      "epoch": 0.7336377473363774,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.497820160015002e-06,
+      "loss": 1.39464998,
+      "memory(GiB)": 113.67,
+      "step": 28920,
+      "train_speed(iter/s)": 1.319165
+    },
+    {
+      "acc": 0.66823354,
+      "epoch": 0.7337645865043125,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.496911706551908e-06,
+      "loss": 1.42817526,
+      "memory(GiB)": 113.67,
+      "step": 28925,
+      "train_speed(iter/s)": 1.319191
+    },
+    {
+      "acc": 0.69902987,
+      "epoch": 0.7338914256722476,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.496003143263492e-06,
+      "loss": 1.35251598,
+      "memory(GiB)": 113.67,
+      "step": 28930,
+      "train_speed(iter/s)": 1.319217
+    },
+    {
+      "acc": 0.67452641,
+      "epoch": 0.7340182648401826,
+      "grad_norm": 3.0,
+      "learning_rate": 7.495094470189712e-06,
+      "loss": 1.46643085,
+      "memory(GiB)": 113.67,
+      "step": 28935,
+      "train_speed(iter/s)": 1.319243
+    },
+    {
+      "acc": 0.66369438,
+      "epoch": 0.7341451040081177,
+      "grad_norm": 4.0,
+      "learning_rate": 7.4941856873705376e-06,
+      "loss": 1.53658571,
+      "memory(GiB)": 113.67,
+      "step": 28940,
+      "train_speed(iter/s)": 1.31927
+    },
+    {
+      "acc": 0.6818759,
+      "epoch": 0.7342719431760527,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.493276794845941e-06,
+      "loss": 1.41824856,
+      "memory(GiB)": 113.67,
+      "step": 28945,
+      "train_speed(iter/s)": 1.319296
+    },
+    {
+      "acc": 0.69593191,
+      "epoch": 0.7343987823439878,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.4923677926559005e-06,
+      "loss": 1.32643757,
+      "memory(GiB)": 113.67,
+      "step": 28950,
+      "train_speed(iter/s)": 1.319322
+    },
+    {
+      "acc": 0.69291086,
+      "epoch": 0.7345256215119229,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.491458680840396e-06,
+      "loss": 1.38566484,
+      "memory(GiB)": 113.67,
+      "step": 28955,
+      "train_speed(iter/s)": 1.319349
+    },
+    {
+      "acc": 0.67954993,
+      "epoch": 0.7346524606798579,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.490549459439415e-06,
+      "loss": 1.39172611,
+      "memory(GiB)": 113.67,
+      "step": 28960,
+      "train_speed(iter/s)": 1.319373
+    },
+    {
+      "acc": 0.67919707,
+      "epoch": 0.734779299847793,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.48964012849295e-06,
+      "loss": 1.41531048,
+      "memory(GiB)": 113.67,
+      "step": 28965,
+      "train_speed(iter/s)": 1.3194
+    },
+    {
+      "acc": 0.67761765,
+      "epoch": 0.7349061390157281,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.488730688040995e-06,
+      "loss": 1.38138781,
+      "memory(GiB)": 113.67,
+      "step": 28970,
+      "train_speed(iter/s)": 1.319425
+    },
+    {
+      "acc": 0.68288937,
+      "epoch": 0.7350329781836631,
+      "grad_norm": 4.5,
+      "learning_rate": 7.487821138123554e-06,
+      "loss": 1.46855593,
+      "memory(GiB)": 113.67,
+      "step": 28975,
+      "train_speed(iter/s)": 1.319452
+    },
+    {
+      "acc": 0.71339483,
+      "epoch": 0.7351598173515982,
+      "grad_norm": 3.125,
+      "learning_rate": 7.486911478780633e-06,
+      "loss": 1.32284641,
+      "memory(GiB)": 113.67,
+      "step": 28980,
+      "train_speed(iter/s)": 1.319478
+    },
+    {
+      "acc": 0.69268966,
+      "epoch": 0.7352866565195332,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.4860017100522395e-06,
+      "loss": 1.37755232,
+      "memory(GiB)": 113.67,
+      "step": 28985,
+      "train_speed(iter/s)": 1.319505
+    },
+    {
+      "acc": 0.68184075,
+      "epoch": 0.7354134956874683,
+      "grad_norm": 3.75,
+      "learning_rate": 7.485091831978394e-06,
+      "loss": 1.42733479,
+      "memory(GiB)": 113.67,
+      "step": 28990,
+      "train_speed(iter/s)": 1.319531
+    },
+    {
+      "acc": 0.6879189,
+      "epoch": 0.7355403348554034,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.484181844599113e-06,
+      "loss": 1.42465992,
+      "memory(GiB)": 113.67,
+      "step": 28995,
+      "train_speed(iter/s)": 1.319558
+    },
+    {
+      "acc": 0.69562464,
+      "epoch": 0.7356671740233384,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.483271747954425e-06,
+      "loss": 1.32193279,
+      "memory(GiB)": 113.67,
+      "step": 29000,
+      "train_speed(iter/s)": 1.319585
+    },
+    {
+      "epoch": 0.7356671740233384,
+      "eval_acc": 0.6744606427388552,
+      "eval_loss": 1.3594661951065063,
+      "eval_runtime": 69.2185,
+      "eval_samples_per_second": 92.027,
+      "eval_steps_per_second": 23.014,
+      "step": 29000
+    },
+    {
+      "acc": 0.68344121,
+      "epoch": 0.7357940131912735,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.482361542084356e-06,
+      "loss": 1.37553301,
+      "memory(GiB)": 113.67,
+      "step": 29005,
+      "train_speed(iter/s)": 1.314717
+    },
+    {
+      "acc": 0.6926322,
+      "epoch": 0.7359208523592086,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.481451227028946e-06,
+      "loss": 1.32834835,
+      "memory(GiB)": 113.67,
+      "step": 29010,
+      "train_speed(iter/s)": 1.314743
+    },
+    {
+      "acc": 0.70340018,
+      "epoch": 0.7360476915271436,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.4805408028282316e-06,
+      "loss": 1.33744879,
+      "memory(GiB)": 113.67,
+      "step": 29015,
+      "train_speed(iter/s)": 1.31477
+    },
+    {
+      "acc": 0.6957284,
+      "epoch": 0.7361745306950787,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.479630269522257e-06,
+      "loss": 1.36914883,
+      "memory(GiB)": 113.67,
+      "step": 29020,
+      "train_speed(iter/s)": 1.314796
+    },
+    {
+      "acc": 0.6773201,
+      "epoch": 0.7363013698630136,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.478719627151073e-06,
+      "loss": 1.49160156,
+      "memory(GiB)": 113.67,
+      "step": 29025,
+      "train_speed(iter/s)": 1.314821
+    },
+    {
+      "acc": 0.67692947,
+      "epoch": 0.7364282090309487,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.4778088757547325e-06,
+      "loss": 1.41420841,
+      "memory(GiB)": 113.67,
+      "step": 29030,
+      "train_speed(iter/s)": 1.314846
+    },
+    {
+      "acc": 0.68583646,
+      "epoch": 0.7365550481988838,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.476898015373296e-06,
+      "loss": 1.4394824,
+      "memory(GiB)": 113.67,
+      "step": 29035,
+      "train_speed(iter/s)": 1.314868
+    },
+    {
+      "acc": 0.69417763,
+      "epoch": 0.7366818873668188,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.4759870460468256e-06,
+      "loss": 1.34541435,
+      "memory(GiB)": 113.67,
+      "step": 29040,
+      "train_speed(iter/s)": 1.314884
+    },
+    {
+      "acc": 0.68406496,
+      "epoch": 0.7368087265347539,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.475075967815391e-06,
+      "loss": 1.36348619,
+      "memory(GiB)": 113.67,
+      "step": 29045,
+      "train_speed(iter/s)": 1.314909
+    },
+    {
+      "acc": 0.67941918,
+      "epoch": 0.736935565702689,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.474164780719064e-06,
+      "loss": 1.36641445,
+      "memory(GiB)": 113.67,
+      "step": 29050,
+      "train_speed(iter/s)": 1.314933
+    },
+    {
+      "acc": 0.68405266,
+      "epoch": 0.737062404870624,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.473253484797924e-06,
+      "loss": 1.40595932,
+      "memory(GiB)": 113.67,
+      "step": 29055,
+      "train_speed(iter/s)": 1.314958
+    },
+    {
+      "acc": 0.69610062,
+      "epoch": 0.7371892440385591,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.4723420800920545e-06,
+      "loss": 1.34649639,
+      "memory(GiB)": 113.67,
+      "step": 29060,
+      "train_speed(iter/s)": 1.314985
+    },
+    {
+      "acc": 0.6837121,
+      "epoch": 0.7373160832064941,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.47143056664154e-06,
+      "loss": 1.37319393,
+      "memory(GiB)": 113.67,
+      "step": 29065,
+      "train_speed(iter/s)": 1.315009
+    },
+    {
+      "acc": 0.69238014,
+      "epoch": 0.7374429223744292,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.470518944486476e-06,
+      "loss": 1.37196569,
+      "memory(GiB)": 113.67,
+      "step": 29070,
+      "train_speed(iter/s)": 1.315035
+    },
+    {
+      "acc": 0.68507061,
+      "epoch": 0.7375697615423643,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.469607213666958e-06,
+      "loss": 1.43479958,
+      "memory(GiB)": 113.67,
+      "step": 29075,
+      "train_speed(iter/s)": 1.315061
+    },
+    {
+      "acc": 0.68385382,
+      "epoch": 0.7376966007102993,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.468695374223092e-06,
+      "loss": 1.35022583,
+      "memory(GiB)": 113.67,
+      "step": 29080,
+      "train_speed(iter/s)": 1.315085
+    },
+    {
+      "acc": 0.69826031,
+      "epoch": 0.7378234398782344,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.4677834261949765e-06,
+      "loss": 1.39290533,
+      "memory(GiB)": 113.67,
+      "step": 29085,
+      "train_speed(iter/s)": 1.315111
+    },
+    {
+      "acc": 0.7070334,
+      "epoch": 0.7379502790461695,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.466871369622731e-06,
+      "loss": 1.29384766,
+      "memory(GiB)": 113.67,
+      "step": 29090,
+      "train_speed(iter/s)": 1.315137
+    },
+    {
+      "acc": 0.68159962,
+      "epoch": 0.7380771182141045,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.465959204546469e-06,
+      "loss": 1.43120041,
+      "memory(GiB)": 113.67,
+      "step": 29095,
+      "train_speed(iter/s)": 1.315161
+    },
+    {
+      "acc": 0.67979689,
+      "epoch": 0.7382039573820396,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.465046931006311e-06,
+      "loss": 1.40358028,
+      "memory(GiB)": 113.67,
+      "step": 29100,
+      "train_speed(iter/s)": 1.315186
+    },
+    {
+      "acc": 0.69034367,
+      "epoch": 0.7383307965499746,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.464134549042383e-06,
+      "loss": 1.37406559,
+      "memory(GiB)": 113.67,
+      "step": 29105,
+      "train_speed(iter/s)": 1.315212
+    },
+    {
+      "acc": 0.68551335,
+      "epoch": 0.7384576357179097,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.463222058694817e-06,
+      "loss": 1.41822577,
+      "memory(GiB)": 113.67,
+      "step": 29110,
+      "train_speed(iter/s)": 1.315236
+    },
+    {
+      "acc": 0.68650723,
+      "epoch": 0.7385844748858448,
+      "grad_norm": 2.5625,
+      "learning_rate": 7.462309460003747e-06,
+      "loss": 1.41357632,
+      "memory(GiB)": 113.67,
+      "step": 29115,
+      "train_speed(iter/s)": 1.315262
+    },
+    {
+      "acc": 0.6888546,
+      "epoch": 0.7387113140537798,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.461396753009314e-06,
+      "loss": 1.39808455,
+      "memory(GiB)": 113.67,
+      "step": 29120,
+      "train_speed(iter/s)": 1.315287
+    },
+    {
+      "acc": 0.68898687,
+      "epoch": 0.7388381532217149,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.460483937751662e-06,
+      "loss": 1.37181454,
+      "memory(GiB)": 113.67,
+      "step": 29125,
+      "train_speed(iter/s)": 1.315314
+    },
+    {
+      "acc": 0.67817879,
+      "epoch": 0.73896499238965,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.45957101427094e-06,
+      "loss": 1.36829433,
+      "memory(GiB)": 113.67,
+      "step": 29130,
+      "train_speed(iter/s)": 1.31534
+    },
+    {
+      "acc": 0.68026209,
+      "epoch": 0.739091831557585,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.458657982607303e-06,
+      "loss": 1.40098085,
+      "memory(GiB)": 113.67,
+      "step": 29135,
+      "train_speed(iter/s)": 1.315366
+    },
+    {
+      "acc": 0.68334465,
+      "epoch": 0.73921867072552,
+      "grad_norm": 4.25,
+      "learning_rate": 7.457744842800913e-06,
+      "loss": 1.4306819,
+      "memory(GiB)": 113.67,
+      "step": 29140,
+      "train_speed(iter/s)": 1.315391
+    },
+    {
+      "acc": 0.66850319,
+      "epoch": 0.739345509893455,
+      "grad_norm": 3.0,
+      "learning_rate": 7.45683159489193e-06,
+      "loss": 1.4990407,
+      "memory(GiB)": 113.67,
+      "step": 29145,
+      "train_speed(iter/s)": 1.315416
+    },
+    {
+      "acc": 0.68001308,
+      "epoch": 0.7394723490613901,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.455918238920526e-06,
+      "loss": 1.37888165,
+      "memory(GiB)": 113.67,
+      "step": 29150,
+      "train_speed(iter/s)": 1.315442
+    },
+    {
+      "acc": 0.68413777,
+      "epoch": 0.7395991882293252,
+      "grad_norm": 3.0,
+      "learning_rate": 7.455004774926873e-06,
+      "loss": 1.42935238,
+      "memory(GiB)": 113.67,
+      "step": 29155,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.68506966,
+      "epoch": 0.7397260273972602,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.454091202951148e-06,
+      "loss": 1.38786268,
+      "memory(GiB)": 113.67,
+      "step": 29160,
+      "train_speed(iter/s)": 1.315488
+    },
+    {
+      "acc": 0.68217573,
+      "epoch": 0.7398528665651953,
+      "grad_norm": 3.25,
+      "learning_rate": 7.453177523033536e-06,
+      "loss": 1.40639267,
+      "memory(GiB)": 113.67,
+      "step": 29165,
+      "train_speed(iter/s)": 1.315514
+    },
+    {
+      "acc": 0.67213073,
+      "epoch": 0.7399797057331304,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.452263735214223e-06,
+      "loss": 1.41529312,
+      "memory(GiB)": 113.67,
+      "step": 29170,
+      "train_speed(iter/s)": 1.315538
+    },
+    {
+      "acc": 0.68590345,
+      "epoch": 0.7401065449010654,
+      "grad_norm": 4.125,
+      "learning_rate": 7.451349839533404e-06,
+      "loss": 1.41175308,
+      "memory(GiB)": 113.67,
+      "step": 29175,
+      "train_speed(iter/s)": 1.315565
+    },
+    {
+      "acc": 0.68227425,
+      "epoch": 0.7402333840690005,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.450435836031273e-06,
+      "loss": 1.36106167,
+      "memory(GiB)": 113.67,
+      "step": 29180,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.69161329,
+      "epoch": 0.7403602232369355,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.449521724748034e-06,
+      "loss": 1.36177597,
+      "memory(GiB)": 113.67,
+      "step": 29185,
+      "train_speed(iter/s)": 1.315617
+    },
+    {
+      "acc": 0.69283099,
+      "epoch": 0.7404870624048706,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.4486075057238936e-06,
+      "loss": 1.35557184,
+      "memory(GiB)": 113.67,
+      "step": 29190,
+      "train_speed(iter/s)": 1.315643
+    },
+    {
+      "acc": 0.69140072,
+      "epoch": 0.7406139015728057,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.447693178999062e-06,
+      "loss": 1.40948467,
+      "memory(GiB)": 113.67,
+      "step": 29195,
+      "train_speed(iter/s)": 1.315667
+    },
+    {
+      "acc": 0.68103781,
+      "epoch": 0.7407407407407407,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.446778744613759e-06,
+      "loss": 1.4250021,
+      "memory(GiB)": 113.67,
+      "step": 29200,
+      "train_speed(iter/s)": 1.315693
+    },
+    {
+      "acc": 0.6989172,
+      "epoch": 0.7408675799086758,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.445864202608198e-06,
+      "loss": 1.33629246,
+      "memory(GiB)": 113.67,
+      "step": 29205,
+      "train_speed(iter/s)": 1.31572
+    },
+    {
+      "acc": 0.67940388,
+      "epoch": 0.7409944190766109,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.444949553022613e-06,
+      "loss": 1.40772305,
+      "memory(GiB)": 113.67,
+      "step": 29210,
+      "train_speed(iter/s)": 1.315744
+    },
+    {
+      "acc": 0.68349357,
+      "epoch": 0.7411212582445459,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.444034795897229e-06,
+      "loss": 1.35213556,
+      "memory(GiB)": 113.67,
+      "step": 29215,
+      "train_speed(iter/s)": 1.315771
+    },
+    {
+      "acc": 0.7026257,
+      "epoch": 0.741248097412481,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.443119931272285e-06,
+      "loss": 1.33782597,
+      "memory(GiB)": 113.67,
+      "step": 29220,
+      "train_speed(iter/s)": 1.315797
+    },
+    {
+      "acc": 0.69598198,
+      "epoch": 0.741374936580416,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.442204959188016e-06,
+      "loss": 1.3862999,
+      "memory(GiB)": 113.67,
+      "step": 29225,
+      "train_speed(iter/s)": 1.315823
+    },
+    {
+      "acc": 0.67964387,
+      "epoch": 0.7415017757483511,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.4412898796846724e-06,
+      "loss": 1.40623837,
+      "memory(GiB)": 113.67,
+      "step": 29230,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.69984694,
+      "epoch": 0.7416286149162862,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.440374692802497e-06,
+      "loss": 1.35705395,
+      "memory(GiB)": 113.67,
+      "step": 29235,
+      "train_speed(iter/s)": 1.315873
+    },
+    {
+      "acc": 0.69343591,
+      "epoch": 0.7417554540842212,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.439459398581747e-06,
+      "loss": 1.3611124,
+      "memory(GiB)": 113.67,
+      "step": 29240,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.69085388,
+      "epoch": 0.7418822932521563,
+      "grad_norm": 4.125,
+      "learning_rate": 7.438543997062684e-06,
+      "loss": 1.31418447,
+      "memory(GiB)": 113.67,
+      "step": 29245,
+      "train_speed(iter/s)": 1.315925
+    },
+    {
+      "acc": 0.6705617,
+      "epoch": 0.7420091324200914,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.437628488285568e-06,
+      "loss": 1.38684311,
+      "memory(GiB)": 113.67,
+      "step": 29250,
+      "train_speed(iter/s)": 1.315952
+    },
+    {
+      "acc": 0.68171415,
+      "epoch": 0.7421359715880264,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.4367128722906665e-06,
+      "loss": 1.35190783,
+      "memory(GiB)": 113.67,
+      "step": 29255,
+      "train_speed(iter/s)": 1.315978
+    },
+    {
+      "acc": 0.68431368,
+      "epoch": 0.7422628107559615,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.435797149118255e-06,
+      "loss": 1.4326848,
+      "memory(GiB)": 113.67,
+      "step": 29260,
+      "train_speed(iter/s)": 1.316004
+    },
+    {
+      "acc": 0.68918862,
+      "epoch": 0.7423896499238964,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.434881318808609e-06,
+      "loss": 1.41584892,
+      "memory(GiB)": 113.67,
+      "step": 29265,
+      "train_speed(iter/s)": 1.31603
+    },
+    {
+      "acc": 0.68349438,
+      "epoch": 0.7425164890918315,
+      "grad_norm": 2.875,
+      "learning_rate": 7.433965381402013e-06,
+      "loss": 1.38006496,
+      "memory(GiB)": 113.67,
+      "step": 29270,
+      "train_speed(iter/s)": 1.316056
+    },
+    {
+      "acc": 0.68718543,
+      "epoch": 0.7426433282597666,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.4330493369387514e-06,
+      "loss": 1.37009363,
+      "memory(GiB)": 113.67,
+      "step": 29275,
+      "train_speed(iter/s)": 1.316082
+    },
+    {
+      "acc": 0.66569872,
+      "epoch": 0.7427701674277016,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.432133185459117e-06,
+      "loss": 1.39089127,
+      "memory(GiB)": 113.67,
+      "step": 29280,
+      "train_speed(iter/s)": 1.316106
+    },
+    {
+      "acc": 0.68511667,
+      "epoch": 0.7428970065956367,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.431216927003406e-06,
+      "loss": 1.36135788,
+      "memory(GiB)": 113.67,
+      "step": 29285,
+      "train_speed(iter/s)": 1.316132
+    },
+    {
+      "acc": 0.68965468,
+      "epoch": 0.7430238457635718,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.430300561611922e-06,
+      "loss": 1.36497478,
+      "memory(GiB)": 113.67,
+      "step": 29290,
+      "train_speed(iter/s)": 1.316157
+    },
+    {
+      "acc": 0.69089127,
+      "epoch": 0.7431506849315068,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.429384089324967e-06,
+      "loss": 1.35848351,
+      "memory(GiB)": 113.67,
+      "step": 29295,
+      "train_speed(iter/s)": 1.316183
+    },
+    {
+      "acc": 0.68899412,
+      "epoch": 0.7432775240994419,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.428467510182854e-06,
+      "loss": 1.38271694,
+      "memory(GiB)": 113.67,
+      "step": 29300,
+      "train_speed(iter/s)": 1.316209
+    },
+    {
+      "acc": 0.70042624,
+      "epoch": 0.7434043632673769,
+      "grad_norm": 3.25,
+      "learning_rate": 7.427550824225896e-06,
+      "loss": 1.34086409,
+      "memory(GiB)": 113.67,
+      "step": 29305,
+      "train_speed(iter/s)": 1.316235
+    },
+    {
+      "acc": 0.6913291,
+      "epoch": 0.743531202435312,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.426634031494417e-06,
+      "loss": 1.38869305,
+      "memory(GiB)": 113.67,
+      "step": 29310,
+      "train_speed(iter/s)": 1.316261
+    },
+    {
+      "acc": 0.70043559,
+      "epoch": 0.7436580416032471,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.425717132028738e-06,
+      "loss": 1.33368874,
+      "memory(GiB)": 113.67,
+      "step": 29315,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.67769747,
+      "epoch": 0.7437848807711821,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.42480012586919e-06,
+      "loss": 1.45728893,
+      "memory(GiB)": 113.67,
+      "step": 29320,
+      "train_speed(iter/s)": 1.316313
+    },
+    {
+      "acc": 0.69027476,
+      "epoch": 0.7439117199391172,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.423883013056106e-06,
+      "loss": 1.3100563,
+      "memory(GiB)": 113.67,
+      "step": 29325,
+      "train_speed(iter/s)": 1.316337
+    },
+    {
+      "acc": 0.68918915,
+      "epoch": 0.7440385591070523,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.422965793629825e-06,
+      "loss": 1.36622219,
+      "memory(GiB)": 113.67,
+      "step": 29330,
+      "train_speed(iter/s)": 1.316363
+    },
+    {
+      "acc": 0.68509622,
+      "epoch": 0.7441653982749873,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.422048467630691e-06,
+      "loss": 1.41116095,
+      "memory(GiB)": 113.67,
+      "step": 29335,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.69237008,
+      "epoch": 0.7442922374429224,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.421131035099052e-06,
+      "loss": 1.34051666,
+      "memory(GiB)": 113.67,
+      "step": 29340,
+      "train_speed(iter/s)": 1.316415
+    },
+    {
+      "acc": 0.68902488,
+      "epoch": 0.7444190766108574,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.42021349607526e-06,
+      "loss": 1.3692318,
+      "memory(GiB)": 113.67,
+      "step": 29345,
+      "train_speed(iter/s)": 1.316441
+    },
+    {
+      "acc": 0.70778341,
+      "epoch": 0.7445459157787925,
+      "grad_norm": 2.625,
+      "learning_rate": 7.419295850599673e-06,
+      "loss": 1.277526,
+      "memory(GiB)": 113.67,
+      "step": 29350,
+      "train_speed(iter/s)": 1.316467
+    },
+    {
+      "acc": 0.69007497,
+      "epoch": 0.7446727549467276,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.418378098712653e-06,
+      "loss": 1.40271482,
+      "memory(GiB)": 113.67,
+      "step": 29355,
+      "train_speed(iter/s)": 1.316493
+    },
+    {
+      "acc": 0.67135496,
+      "epoch": 0.7447995941146626,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.417460240454568e-06,
+      "loss": 1.52074318,
+      "memory(GiB)": 113.67,
+      "step": 29360,
+      "train_speed(iter/s)": 1.316518
+    },
+    {
+      "acc": 0.68745313,
+      "epoch": 0.7449264332825977,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.4165422758657865e-06,
+      "loss": 1.3521615,
+      "memory(GiB)": 113.67,
+      "step": 29365,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.69749207,
+      "epoch": 0.7450532724505328,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.415624204986689e-06,
+      "loss": 1.3453227,
+      "memory(GiB)": 113.67,
+      "step": 29370,
+      "train_speed(iter/s)": 1.316568
+    },
+    {
+      "acc": 0.68225608,
+      "epoch": 0.7451801116184678,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.4147060278576525e-06,
+      "loss": 1.42632236,
+      "memory(GiB)": 113.67,
+      "step": 29375,
+      "train_speed(iter/s)": 1.316592
+    },
+    {
+      "acc": 0.6873457,
+      "epoch": 0.7453069507864029,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.413787744519064e-06,
+      "loss": 1.3592248,
+      "memory(GiB)": 113.67,
+      "step": 29380,
+      "train_speed(iter/s)": 1.316618
+    },
+    {
+      "acc": 0.67538099,
+      "epoch": 0.7454337899543378,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.412869355011314e-06,
+      "loss": 1.37102394,
+      "memory(GiB)": 113.67,
+      "step": 29385,
+      "train_speed(iter/s)": 1.316643
+    },
+    {
+      "acc": 0.68215218,
+      "epoch": 0.7455606291222729,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.411950859374797e-06,
+      "loss": 1.42633629,
+      "memory(GiB)": 113.67,
+      "step": 29390,
+      "train_speed(iter/s)": 1.316666
+    },
+    {
+      "acc": 0.69899459,
+      "epoch": 0.745687468290208,
+      "grad_norm": 3.375,
+      "learning_rate": 7.411032257649913e-06,
+      "loss": 1.39636021,
+      "memory(GiB)": 113.67,
+      "step": 29395,
+      "train_speed(iter/s)": 1.316691
+    },
+    {
+      "acc": 0.69228029,
+      "epoch": 0.745814307458143,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.410113549877065e-06,
+      "loss": 1.3796628,
+      "memory(GiB)": 113.67,
+      "step": 29400,
+      "train_speed(iter/s)": 1.316715
+    },
+    {
+      "acc": 0.68297176,
+      "epoch": 0.7459411466260781,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.409194736096663e-06,
+      "loss": 1.41380701,
+      "memory(GiB)": 113.67,
+      "step": 29405,
+      "train_speed(iter/s)": 1.31674
+    },
+    {
+      "acc": 0.6949367,
+      "epoch": 0.7460679857940132,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.408275816349121e-06,
+      "loss": 1.37822466,
+      "memory(GiB)": 113.67,
+      "step": 29410,
+      "train_speed(iter/s)": 1.316766
+    },
+    {
+      "acc": 0.69286299,
+      "epoch": 0.7461948249619482,
+      "grad_norm": 2.875,
+      "learning_rate": 7.4073567906748555e-06,
+      "loss": 1.38984795,
+      "memory(GiB)": 113.67,
+      "step": 29415,
+      "train_speed(iter/s)": 1.316791
+    },
+    {
+      "acc": 0.68594437,
+      "epoch": 0.7463216641298833,
+      "grad_norm": 2.875,
+      "learning_rate": 7.406437659114291e-06,
+      "loss": 1.37247581,
+      "memory(GiB)": 113.67,
+      "step": 29420,
+      "train_speed(iter/s)": 1.316815
+    },
+    {
+      "acc": 0.68689871,
+      "epoch": 0.7464485032978183,
+      "grad_norm": 3.0,
+      "learning_rate": 7.405518421707854e-06,
+      "loss": 1.42085323,
+      "memory(GiB)": 113.67,
+      "step": 29425,
+      "train_speed(iter/s)": 1.31684
+    },
+    {
+      "acc": 0.69132199,
+      "epoch": 0.7465753424657534,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.404599078495977e-06,
+      "loss": 1.37203541,
+      "memory(GiB)": 113.67,
+      "step": 29430,
+      "train_speed(iter/s)": 1.316866
+    },
+    {
+      "acc": 0.68351617,
+      "epoch": 0.7467021816336885,
+      "grad_norm": 2.578125,
+      "learning_rate": 7.403679629519096e-06,
+      "loss": 1.41102276,
+      "memory(GiB)": 113.67,
+      "step": 29435,
+      "train_speed(iter/s)": 1.31689
+    },
+    {
+      "acc": 0.67613001,
+      "epoch": 0.7468290208016235,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.402760074817654e-06,
+      "loss": 1.48316545,
+      "memory(GiB)": 113.67,
+      "step": 29440,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.70914054,
+      "epoch": 0.7469558599695586,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.4018404144320955e-06,
+      "loss": 1.35817947,
+      "memory(GiB)": 113.67,
+      "step": 29445,
+      "train_speed(iter/s)": 1.316942
+    },
+    {
+      "acc": 0.69711781,
+      "epoch": 0.7470826991374937,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.4009206484028735e-06,
+      "loss": 1.41702328,
+      "memory(GiB)": 113.67,
+      "step": 29450,
+      "train_speed(iter/s)": 1.316967
+    },
+    {
+      "acc": 0.69144011,
+      "epoch": 0.7472095383054287,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.400000776770441e-06,
+      "loss": 1.34482479,
+      "memory(GiB)": 113.67,
+      "step": 29455,
+      "train_speed(iter/s)": 1.316992
+    },
+    {
+      "acc": 0.67665644,
+      "epoch": 0.7473363774733638,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.39908079957526e-06,
+      "loss": 1.41518259,
+      "memory(GiB)": 113.67,
+      "step": 29460,
+      "train_speed(iter/s)": 1.317016
+    },
+    {
+      "acc": 0.68397636,
+      "epoch": 0.7474632166412988,
+      "grad_norm": 5.03125,
+      "learning_rate": 7.398160716857794e-06,
+      "loss": 1.4098732,
+      "memory(GiB)": 113.67,
+      "step": 29465,
+      "train_speed(iter/s)": 1.317041
+    },
+    {
+      "acc": 0.69236202,
+      "epoch": 0.7475900558092339,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.397240528658513e-06,
+      "loss": 1.36025829,
+      "memory(GiB)": 113.67,
+      "step": 29470,
+      "train_speed(iter/s)": 1.317063
+    },
+    {
+      "acc": 0.69438004,
+      "epoch": 0.747716894977169,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.39632023501789e-06,
+      "loss": 1.39114895,
+      "memory(GiB)": 113.67,
+      "step": 29475,
+      "train_speed(iter/s)": 1.317088
+    },
+    {
+      "acc": 0.67302904,
+      "epoch": 0.747843734145104,
+      "grad_norm": 3.0,
+      "learning_rate": 7.3953998359764036e-06,
+      "loss": 1.43260117,
+      "memory(GiB)": 113.67,
+      "step": 29480,
+      "train_speed(iter/s)": 1.317112
+    },
+    {
+      "acc": 0.68748837,
+      "epoch": 0.7479705733130391,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.394479331574539e-06,
+      "loss": 1.4047699,
+      "memory(GiB)": 113.67,
+      "step": 29485,
+      "train_speed(iter/s)": 1.317137
+    },
+    {
+      "acc": 0.67787943,
+      "epoch": 0.7480974124809742,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.393558721852783e-06,
+      "loss": 1.44613647,
+      "memory(GiB)": 113.67,
+      "step": 29490,
+      "train_speed(iter/s)": 1.317161
+    },
+    {
+      "acc": 0.67111988,
+      "epoch": 0.7482242516489092,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.392638006851627e-06,
+      "loss": 1.41212091,
+      "memory(GiB)": 113.67,
+      "step": 29495,
+      "train_speed(iter/s)": 1.317188
+    },
+    {
+      "acc": 0.69451256,
+      "epoch": 0.7483510908168443,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.391717186611569e-06,
+      "loss": 1.36260033,
+      "memory(GiB)": 113.67,
+      "step": 29500,
+      "train_speed(iter/s)": 1.317214
+    },
+    {
+      "acc": 0.6975276,
+      "epoch": 0.7484779299847792,
+      "grad_norm": 3.5,
+      "learning_rate": 7.39079626117311e-06,
+      "loss": 1.3592453,
+      "memory(GiB)": 113.67,
+      "step": 29505,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.67746735,
+      "epoch": 0.7486047691527143,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.3898752305767595e-06,
+      "loss": 1.43807411,
+      "memory(GiB)": 113.67,
+      "step": 29510,
+      "train_speed(iter/s)": 1.317265
+    },
+    {
+      "acc": 0.68333769,
+      "epoch": 0.7487316083206494,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.3889540948630245e-06,
+      "loss": 1.41468534,
+      "memory(GiB)": 113.67,
+      "step": 29515,
+      "train_speed(iter/s)": 1.317291
+    },
+    {
+      "acc": 0.68921552,
+      "epoch": 0.7488584474885844,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.388032854072424e-06,
+      "loss": 1.36201658,
+      "memory(GiB)": 113.67,
+      "step": 29520,
+      "train_speed(iter/s)": 1.317317
+    },
+    {
+      "acc": 0.68155785,
+      "epoch": 0.7489852866565195,
+      "grad_norm": 3.375,
+      "learning_rate": 7.387111508245476e-06,
+      "loss": 1.43041086,
+      "memory(GiB)": 113.67,
+      "step": 29525,
+      "train_speed(iter/s)": 1.317344
+    },
+    {
+      "acc": 0.69605799,
+      "epoch": 0.7491121258244546,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.386190057422706e-06,
+      "loss": 1.29574575,
+      "memory(GiB)": 113.67,
+      "step": 29530,
+      "train_speed(iter/s)": 1.317372
+    },
+    {
+      "acc": 0.68080726,
+      "epoch": 0.7492389649923896,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.385268501644645e-06,
+      "loss": 1.3607378,
+      "memory(GiB)": 113.67,
+      "step": 29535,
+      "train_speed(iter/s)": 1.317395
+    },
+    {
+      "acc": 0.6926898,
+      "epoch": 0.7493658041603247,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.384346840951824e-06,
+      "loss": 1.37623806,
+      "memory(GiB)": 113.67,
+      "step": 29540,
+      "train_speed(iter/s)": 1.317421
+    },
+    {
+      "acc": 0.6931797,
+      "epoch": 0.7494926433282597,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.383425075384785e-06,
+      "loss": 1.34544373,
+      "memory(GiB)": 113.67,
+      "step": 29545,
+      "train_speed(iter/s)": 1.317447
+    },
+    {
+      "acc": 0.69286375,
+      "epoch": 0.7496194824961948,
+      "grad_norm": 3.125,
+      "learning_rate": 7.382503204984069e-06,
+      "loss": 1.35340586,
+      "memory(GiB)": 113.67,
+      "step": 29550,
+      "train_speed(iter/s)": 1.317472
+    },
+    {
+      "acc": 0.67898035,
+      "epoch": 0.7497463216641299,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.381581229790226e-06,
+      "loss": 1.36653309,
+      "memory(GiB)": 113.67,
+      "step": 29555,
+      "train_speed(iter/s)": 1.317498
+    },
+    {
+      "acc": 0.68941512,
+      "epoch": 0.7498731608320649,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.380659149843806e-06,
+      "loss": 1.3518507,
+      "memory(GiB)": 113.67,
+      "step": 29560,
+      "train_speed(iter/s)": 1.317519
+    },
+    {
+      "acc": 0.68305645,
+      "epoch": 0.75,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.379736965185369e-06,
+      "loss": 1.38835087,
+      "memory(GiB)": 113.67,
+      "step": 29565,
+      "train_speed(iter/s)": 1.317545
+    },
+    {
+      "acc": 0.66702676,
+      "epoch": 0.7501268391679351,
+      "grad_norm": 2.5625,
+      "learning_rate": 7.378814675855475e-06,
+      "loss": 1.44795275,
+      "memory(GiB)": 113.67,
+      "step": 29570,
+      "train_speed(iter/s)": 1.31757
+    },
+    {
+      "acc": 0.69705782,
+      "epoch": 0.7502536783358701,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.37789228189469e-06,
+      "loss": 1.38437614,
+      "memory(GiB)": 113.67,
+      "step": 29575,
+      "train_speed(iter/s)": 1.317596
+    },
+    {
+      "acc": 0.68953352,
+      "epoch": 0.7503805175038052,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.376969783343588e-06,
+      "loss": 1.38084917,
+      "memory(GiB)": 113.67,
+      "step": 29580,
+      "train_speed(iter/s)": 1.317621
+    },
+    {
+      "acc": 0.68891916,
+      "epoch": 0.7505073566717403,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.37604718024274e-06,
+      "loss": 1.34737778,
+      "memory(GiB)": 113.67,
+      "step": 29585,
+      "train_speed(iter/s)": 1.317647
+    },
+    {
+      "acc": 0.69611473,
+      "epoch": 0.7506341958396753,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.375124472632732e-06,
+      "loss": 1.39352837,
+      "memory(GiB)": 113.67,
+      "step": 29590,
+      "train_speed(iter/s)": 1.317673
+    },
+    {
+      "acc": 0.6744741,
+      "epoch": 0.7507610350076104,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.374201660554142e-06,
+      "loss": 1.44014578,
+      "memory(GiB)": 113.67,
+      "step": 29595,
+      "train_speed(iter/s)": 1.317699
+    },
+    {
+      "acc": 0.68530173,
+      "epoch": 0.7508878741755454,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.373278744047565e-06,
+      "loss": 1.41770792,
+      "memory(GiB)": 113.67,
+      "step": 29600,
+      "train_speed(iter/s)": 1.317725
+    },
+    {
+      "acc": 0.67838383,
+      "epoch": 0.7510147133434805,
+      "grad_norm": 3.84375,
+      "learning_rate": 7.372355723153593e-06,
+      "loss": 1.44290276,
+      "memory(GiB)": 113.67,
+      "step": 29605,
+      "train_speed(iter/s)": 1.317751
+    },
+    {
+      "acc": 0.67680655,
+      "epoch": 0.7511415525114156,
+      "grad_norm": 3.375,
+      "learning_rate": 7.371432597912824e-06,
+      "loss": 1.39238262,
+      "memory(GiB)": 113.67,
+      "step": 29610,
+      "train_speed(iter/s)": 1.317776
+    },
+    {
+      "acc": 0.6935482,
+      "epoch": 0.7512683916793506,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.3705093683658616e-06,
+      "loss": 1.39502411,
+      "memory(GiB)": 113.67,
+      "step": 29615,
+      "train_speed(iter/s)": 1.317802
+    },
+    {
+      "acc": 0.67912049,
+      "epoch": 0.7513952308472857,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.369586034553313e-06,
+      "loss": 1.41193628,
+      "memory(GiB)": 113.67,
+      "step": 29620,
+      "train_speed(iter/s)": 1.317828
+    },
+    {
+      "acc": 0.67830486,
+      "epoch": 0.7515220700152208,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.368662596515792e-06,
+      "loss": 1.48146324,
+      "memory(GiB)": 113.67,
+      "step": 29625,
+      "train_speed(iter/s)": 1.317854
+    },
+    {
+      "acc": 0.6784709,
+      "epoch": 0.7516489091831557,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.367739054293914e-06,
+      "loss": 1.42379608,
+      "memory(GiB)": 113.67,
+      "step": 29630,
+      "train_speed(iter/s)": 1.31788
+    },
+    {
+      "acc": 0.68199854,
+      "epoch": 0.7517757483510908,
+      "grad_norm": 2.625,
+      "learning_rate": 7.366815407928302e-06,
+      "loss": 1.40123234,
+      "memory(GiB)": 113.67,
+      "step": 29635,
+      "train_speed(iter/s)": 1.317906
+    },
+    {
+      "acc": 0.69037561,
+      "epoch": 0.7519025875190258,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.365891657459582e-06,
+      "loss": 1.37359676,
+      "memory(GiB)": 113.67,
+      "step": 29640,
+      "train_speed(iter/s)": 1.317932
+    },
+    {
+      "acc": 0.67072992,
+      "epoch": 0.7520294266869609,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.3649678029283825e-06,
+      "loss": 1.44884539,
+      "memory(GiB)": 113.67,
+      "step": 29645,
+      "train_speed(iter/s)": 1.317957
+    },
+    {
+      "acc": 0.67490106,
+      "epoch": 0.752156265854896,
+      "grad_norm": 3.5,
+      "learning_rate": 7.364043844375342e-06,
+      "loss": 1.43180809,
+      "memory(GiB)": 113.67,
+      "step": 29650,
+      "train_speed(iter/s)": 1.317983
+    },
+    {
+      "acc": 0.67544613,
+      "epoch": 0.752283105022831,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.363119781841095e-06,
+      "loss": 1.44222088,
+      "memory(GiB)": 113.67,
+      "step": 29655,
+      "train_speed(iter/s)": 1.318008
+    },
+    {
+      "acc": 0.67691259,
+      "epoch": 0.7524099441907661,
+      "grad_norm": 3.125,
+      "learning_rate": 7.362195615366293e-06,
+      "loss": 1.41298456,
+      "memory(GiB)": 113.67,
+      "step": 29660,
+      "train_speed(iter/s)": 1.318033
+    },
+    {
+      "acc": 0.7050355,
+      "epoch": 0.7525367833587012,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.361271344991579e-06,
+      "loss": 1.34007864,
+      "memory(GiB)": 113.67,
+      "step": 29665,
+      "train_speed(iter/s)": 1.31806
+    },
+    {
+      "acc": 0.68194838,
+      "epoch": 0.7526636225266362,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.36034697075761e-06,
+      "loss": 1.4221303,
+      "memory(GiB)": 113.67,
+      "step": 29670,
+      "train_speed(iter/s)": 1.318085
+    },
+    {
+      "acc": 0.69053354,
+      "epoch": 0.7527904616945713,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.359422492705043e-06,
+      "loss": 1.3940794,
+      "memory(GiB)": 113.67,
+      "step": 29675,
+      "train_speed(iter/s)": 1.318111
+    },
+    {
+      "acc": 0.69269934,
+      "epoch": 0.7529173008625063,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.3584979108745405e-06,
+      "loss": 1.34428778,
+      "memory(GiB)": 113.67,
+      "step": 29680,
+      "train_speed(iter/s)": 1.318136
+    },
+    {
+      "acc": 0.69296265,
+      "epoch": 0.7530441400304414,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.357573225306771e-06,
+      "loss": 1.39691496,
+      "memory(GiB)": 113.67,
+      "step": 29685,
+      "train_speed(iter/s)": 1.318162
+    },
+    {
+      "acc": 0.68575153,
+      "epoch": 0.7531709791983765,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.356648436042404e-06,
+      "loss": 1.41285591,
+      "memory(GiB)": 113.67,
+      "step": 29690,
+      "train_speed(iter/s)": 1.318188
+    },
+    {
+      "acc": 0.67909169,
+      "epoch": 0.7532978183663115,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.355723543122118e-06,
+      "loss": 1.43945694,
+      "memory(GiB)": 113.67,
+      "step": 29695,
+      "train_speed(iter/s)": 1.318213
+    },
+    {
+      "acc": 0.70411401,
+      "epoch": 0.7534246575342466,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.354798546586592e-06,
+      "loss": 1.33346081,
+      "memory(GiB)": 113.67,
+      "step": 29700,
+      "train_speed(iter/s)": 1.318238
+    },
+    {
+      "acc": 0.6933537,
+      "epoch": 0.7535514967021817,
+      "grad_norm": 3.875,
+      "learning_rate": 7.353873446476512e-06,
+      "loss": 1.36717491,
+      "memory(GiB)": 113.67,
+      "step": 29705,
+      "train_speed(iter/s)": 1.318263
+    },
+    {
+      "acc": 0.6887012,
+      "epoch": 0.7536783358701167,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.3529482428325705e-06,
+      "loss": 1.44697123,
+      "memory(GiB)": 113.67,
+      "step": 29710,
+      "train_speed(iter/s)": 1.318282
+    },
+    {
+      "acc": 0.68392873,
+      "epoch": 0.7538051750380518,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.35202293569546e-06,
+      "loss": 1.40586786,
+      "memory(GiB)": 113.67,
+      "step": 29715,
+      "train_speed(iter/s)": 1.318307
+    },
+    {
+      "acc": 0.69117785,
+      "epoch": 0.7539320142059868,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.351097525105878e-06,
+      "loss": 1.35782833,
+      "memory(GiB)": 113.67,
+      "step": 29720,
+      "train_speed(iter/s)": 1.318331
+    },
+    {
+      "acc": 0.68624496,
+      "epoch": 0.7540588533739219,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.35017201110453e-06,
+      "loss": 1.34953117,
+      "memory(GiB)": 113.67,
+      "step": 29725,
+      "train_speed(iter/s)": 1.318356
+    },
+    {
+      "acc": 0.67794895,
+      "epoch": 0.754185692541857,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.349246393732126e-06,
+      "loss": 1.43559589,
+      "memory(GiB)": 113.67,
+      "step": 29730,
+      "train_speed(iter/s)": 1.318381
+    },
+    {
+      "acc": 0.69438705,
+      "epoch": 0.754312531709792,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.3483206730293755e-06,
+      "loss": 1.38801279,
+      "memory(GiB)": 113.67,
+      "step": 29735,
+      "train_speed(iter/s)": 1.318405
+    },
+    {
+      "acc": 0.69366531,
+      "epoch": 0.7544393708777271,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.347394849036998e-06,
+      "loss": 1.38222713,
+      "memory(GiB)": 113.67,
+      "step": 29740,
+      "train_speed(iter/s)": 1.31843
+    },
+    {
+      "acc": 0.69516993,
+      "epoch": 0.7545662100456622,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.346468921795714e-06,
+      "loss": 1.31973248,
+      "memory(GiB)": 113.67,
+      "step": 29745,
+      "train_speed(iter/s)": 1.318456
+    },
+    {
+      "acc": 0.67615623,
+      "epoch": 0.7546930492135971,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.345542891346251e-06,
+      "loss": 1.41632023,
+      "memory(GiB)": 113.67,
+      "step": 29750,
+      "train_speed(iter/s)": 1.31848
+    },
+    {
+      "acc": 0.68372755,
+      "epoch": 0.7548198883815322,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.344616757729341e-06,
+      "loss": 1.38854275,
+      "memory(GiB)": 113.67,
+      "step": 29755,
+      "train_speed(iter/s)": 1.318505
+    },
+    {
+      "acc": 0.69743519,
+      "epoch": 0.7549467275494672,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.343690520985716e-06,
+      "loss": 1.38251057,
+      "memory(GiB)": 113.67,
+      "step": 29760,
+      "train_speed(iter/s)": 1.318531
+    },
+    {
+      "acc": 0.68380504,
+      "epoch": 0.7550735667174023,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.342764181156119e-06,
+      "loss": 1.37232075,
+      "memory(GiB)": 113.67,
+      "step": 29765,
+      "train_speed(iter/s)": 1.318557
+    },
+    {
+      "acc": 0.69190421,
+      "epoch": 0.7552004058853374,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.341837738281293e-06,
+      "loss": 1.37733011,
+      "memory(GiB)": 113.67,
+      "step": 29770,
+      "train_speed(iter/s)": 1.318583
+    },
+    {
+      "acc": 0.691745,
+      "epoch": 0.7553272450532724,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.3409111924019885e-06,
+      "loss": 1.40728817,
+      "memory(GiB)": 113.67,
+      "step": 29775,
+      "train_speed(iter/s)": 1.318608
+    },
+    {
+      "acc": 0.68566594,
+      "epoch": 0.7554540842212075,
+      "grad_norm": 3.84375,
+      "learning_rate": 7.3399845435589574e-06,
+      "loss": 1.41988487,
+      "memory(GiB)": 113.67,
+      "step": 29780,
+      "train_speed(iter/s)": 1.318635
+    },
+    {
+      "acc": 0.70074711,
+      "epoch": 0.7555809233891426,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.33905779179296e-06,
+      "loss": 1.32273655,
+      "memory(GiB)": 113.67,
+      "step": 29785,
+      "train_speed(iter/s)": 1.318661
+    },
+    {
+      "acc": 0.68597736,
+      "epoch": 0.7557077625570776,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.338130937144756e-06,
+      "loss": 1.38933945,
+      "memory(GiB)": 113.67,
+      "step": 29790,
+      "train_speed(iter/s)": 1.318685
+    },
+    {
+      "acc": 0.70023384,
+      "epoch": 0.7558346017250127,
+      "grad_norm": 3.125,
+      "learning_rate": 7.3372039796551156e-06,
+      "loss": 1.33077736,
+      "memory(GiB)": 113.67,
+      "step": 29795,
+      "train_speed(iter/s)": 1.31871
+    },
+    {
+      "acc": 0.6883153,
+      "epoch": 0.7559614408929477,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.33627691936481e-06,
+      "loss": 1.3865449,
+      "memory(GiB)": 113.67,
+      "step": 29800,
+      "train_speed(iter/s)": 1.318736
+    },
+    {
+      "acc": 0.7170186,
+      "epoch": 0.7560882800608828,
+      "grad_norm": 2.46875,
+      "learning_rate": 7.335349756314614e-06,
+      "loss": 1.27324257,
+      "memory(GiB)": 113.67,
+      "step": 29805,
+      "train_speed(iter/s)": 1.31876
+    },
+    {
+      "acc": 0.6758637,
+      "epoch": 0.7562151192288179,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.33442249054531e-06,
+      "loss": 1.45693932,
+      "memory(GiB)": 113.67,
+      "step": 29810,
+      "train_speed(iter/s)": 1.318785
+    },
+    {
+      "acc": 0.67456164,
+      "epoch": 0.7563419583967529,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.33349512209768e-06,
+      "loss": 1.41851416,
+      "memory(GiB)": 113.67,
+      "step": 29815,
+      "train_speed(iter/s)": 1.318811
+    },
+    {
+      "acc": 0.67839279,
+      "epoch": 0.756468797564688,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.332567651012518e-06,
+      "loss": 1.43073044,
+      "memory(GiB)": 113.67,
+      "step": 29820,
+      "train_speed(iter/s)": 1.318835
+    },
+    {
+      "acc": 0.6912477,
+      "epoch": 0.7565956367326231,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.331640077330616e-06,
+      "loss": 1.35610123,
+      "memory(GiB)": 113.67,
+      "step": 29825,
+      "train_speed(iter/s)": 1.31886
+    },
+    {
+      "acc": 0.69766827,
+      "epoch": 0.7567224759005581,
+      "grad_norm": 3.375,
+      "learning_rate": 7.330712401092773e-06,
+      "loss": 1.42419834,
+      "memory(GiB)": 113.67,
+      "step": 29830,
+      "train_speed(iter/s)": 1.318885
+    },
+    {
+      "acc": 0.68341446,
+      "epoch": 0.7568493150684932,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.329784622339794e-06,
+      "loss": 1.37070284,
+      "memory(GiB)": 113.67,
+      "step": 29835,
+      "train_speed(iter/s)": 1.31891
+    },
+    {
+      "acc": 0.67709193,
+      "epoch": 0.7569761542364282,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.328856741112484e-06,
+      "loss": 1.40691481,
+      "memory(GiB)": 113.67,
+      "step": 29840,
+      "train_speed(iter/s)": 1.318933
+    },
+    {
+      "acc": 0.68172226,
+      "epoch": 0.7571029934043633,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.327928757451659e-06,
+      "loss": 1.37753716,
+      "memory(GiB)": 113.67,
+      "step": 29845,
+      "train_speed(iter/s)": 1.318958
+    },
+    {
+      "acc": 0.69720888,
+      "epoch": 0.7572298325722984,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.3270006713981325e-06,
+      "loss": 1.32718868,
+      "memory(GiB)": 113.67,
+      "step": 29850,
+      "train_speed(iter/s)": 1.318982
+    },
+    {
+      "acc": 0.68193417,
+      "epoch": 0.7573566717402334,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.326072482992728e-06,
+      "loss": 1.4194849,
+      "memory(GiB)": 113.67,
+      "step": 29855,
+      "train_speed(iter/s)": 1.319007
+    },
+    {
+      "acc": 0.67768326,
+      "epoch": 0.7574835109081685,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.325144192276269e-06,
+      "loss": 1.44435215,
+      "memory(GiB)": 113.67,
+      "step": 29860,
+      "train_speed(iter/s)": 1.319032
+    },
+    {
+      "acc": 0.68243785,
+      "epoch": 0.7576103500761036,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.324215799289588e-06,
+      "loss": 1.43999338,
+      "memory(GiB)": 113.67,
+      "step": 29865,
+      "train_speed(iter/s)": 1.319056
+    },
+    {
+      "acc": 0.69796243,
+      "epoch": 0.7577371892440385,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.3232873040735194e-06,
+      "loss": 1.33724613,
+      "memory(GiB)": 113.67,
+      "step": 29870,
+      "train_speed(iter/s)": 1.31908
+    },
+    {
+      "acc": 0.68513722,
+      "epoch": 0.7578640284119736,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.322358706668901e-06,
+      "loss": 1.42535467,
+      "memory(GiB)": 113.67,
+      "step": 29875,
+      "train_speed(iter/s)": 1.319105
+    },
+    {
+      "acc": 0.69058323,
+      "epoch": 0.7579908675799086,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.321430007116582e-06,
+      "loss": 1.39241819,
+      "memory(GiB)": 113.67,
+      "step": 29880,
+      "train_speed(iter/s)": 1.31913
+    },
+    {
+      "acc": 0.66696901,
+      "epoch": 0.7581177067478437,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.320501205457403e-06,
+      "loss": 1.45909519,
+      "memory(GiB)": 113.67,
+      "step": 29885,
+      "train_speed(iter/s)": 1.319155
+    },
+    {
+      "acc": 0.68334322,
+      "epoch": 0.7582445459157788,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.319572301732224e-06,
+      "loss": 1.39307775,
+      "memory(GiB)": 113.67,
+      "step": 29890,
+      "train_speed(iter/s)": 1.319181
+    },
+    {
+      "acc": 0.68127656,
+      "epoch": 0.7583713850837138,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.3186432959818956e-06,
+      "loss": 1.41013775,
+      "memory(GiB)": 113.67,
+      "step": 29895,
+      "train_speed(iter/s)": 1.319205
+    },
+    {
+      "acc": 0.69231539,
+      "epoch": 0.7584982242516489,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.317714188247285e-06,
+      "loss": 1.34936638,
+      "memory(GiB)": 113.67,
+      "step": 29900,
+      "train_speed(iter/s)": 1.319231
+    },
+    {
+      "acc": 0.68308849,
+      "epoch": 0.758625063419584,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.316784978569256e-06,
+      "loss": 1.40981197,
+      "memory(GiB)": 113.67,
+      "step": 29905,
+      "train_speed(iter/s)": 1.319257
+    },
+    {
+      "acc": 0.70618639,
+      "epoch": 0.758751902587519,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.31585566698868e-06,
+      "loss": 1.27975407,
+      "memory(GiB)": 113.67,
+      "step": 29910,
+      "train_speed(iter/s)": 1.319279
+    },
+    {
+      "acc": 0.69589777,
+      "epoch": 0.7588787417554541,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.314926253546433e-06,
+      "loss": 1.30058193,
+      "memory(GiB)": 113.67,
+      "step": 29915,
+      "train_speed(iter/s)": 1.319304
+    },
+    {
+      "acc": 0.68715868,
+      "epoch": 0.7590055809233891,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.313996738283393e-06,
+      "loss": 1.39300327,
+      "memory(GiB)": 113.67,
+      "step": 29920,
+      "train_speed(iter/s)": 1.319329
+    },
+    {
+      "acc": 0.68859396,
+      "epoch": 0.7591324200913242,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.3130671212404455e-06,
+      "loss": 1.42553844,
+      "memory(GiB)": 113.67,
+      "step": 29925,
+      "train_speed(iter/s)": 1.319355
+    },
+    {
+      "acc": 0.69314785,
+      "epoch": 0.7592592592592593,
+      "grad_norm": 2.75,
+      "learning_rate": 7.312137402458479e-06,
+      "loss": 1.37578259,
+      "memory(GiB)": 113.67,
+      "step": 29930,
+      "train_speed(iter/s)": 1.319381
+    },
+    {
+      "acc": 0.68673816,
+      "epoch": 0.7593860984271943,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.3112075819783864e-06,
+      "loss": 1.40923252,
+      "memory(GiB)": 113.67,
+      "step": 29935,
+      "train_speed(iter/s)": 1.319406
+    },
+    {
+      "acc": 0.67294431,
+      "epoch": 0.7595129375951294,
+      "grad_norm": 4.53125,
+      "learning_rate": 7.310277659841066e-06,
+      "loss": 1.48410358,
+      "memory(GiB)": 113.67,
+      "step": 29940,
+      "train_speed(iter/s)": 1.319431
+    },
+    {
+      "acc": 0.67710361,
+      "epoch": 0.7596397767630645,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.309347636087418e-06,
+      "loss": 1.41386585,
+      "memory(GiB)": 113.67,
+      "step": 29945,
+      "train_speed(iter/s)": 1.319456
+    },
+    {
+      "acc": 0.66850424,
+      "epoch": 0.7597666159309995,
+      "grad_norm": 2.5,
+      "learning_rate": 7.308417510758353e-06,
+      "loss": 1.46802826,
+      "memory(GiB)": 113.67,
+      "step": 29950,
+      "train_speed(iter/s)": 1.319481
+    },
+    {
+      "acc": 0.67268271,
+      "epoch": 0.7598934550989346,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.307487283894777e-06,
+      "loss": 1.45201502,
+      "memory(GiB)": 113.67,
+      "step": 29955,
+      "train_speed(iter/s)": 1.319507
+    },
+    {
+      "acc": 0.70079947,
+      "epoch": 0.7600202942668696,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.30655695553761e-06,
+      "loss": 1.31060572,
+      "memory(GiB)": 113.67,
+      "step": 29960,
+      "train_speed(iter/s)": 1.319532
+    },
+    {
+      "acc": 0.69750071,
+      "epoch": 0.7601471334348047,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.305626525727769e-06,
+      "loss": 1.36087608,
+      "memory(GiB)": 113.67,
+      "step": 29965,
+      "train_speed(iter/s)": 1.319558
+    },
+    {
+      "acc": 0.69244032,
+      "epoch": 0.7602739726027398,
+      "grad_norm": 2.75,
+      "learning_rate": 7.30469599450618e-06,
+      "loss": 1.45989189,
+      "memory(GiB)": 113.67,
+      "step": 29970,
+      "train_speed(iter/s)": 1.319583
+    },
+    {
+      "acc": 0.68376293,
+      "epoch": 0.7604008117706748,
+      "grad_norm": 2.484375,
+      "learning_rate": 7.30376536191377e-06,
+      "loss": 1.37835026,
+      "memory(GiB)": 113.67,
+      "step": 29975,
+      "train_speed(iter/s)": 1.319607
+    },
+    {
+      "acc": 0.68728657,
+      "epoch": 0.7605276509386099,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.302834627991477e-06,
+      "loss": 1.40944757,
+      "memory(GiB)": 113.67,
+      "step": 29980,
+      "train_speed(iter/s)": 1.319625
+    },
+    {
+      "acc": 0.68679929,
+      "epoch": 0.760654490106545,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.301903792780233e-06,
+      "loss": 1.42045383,
+      "memory(GiB)": 113.67,
+      "step": 29985,
+      "train_speed(iter/s)": 1.319649
+    },
+    {
+      "acc": 0.698664,
+      "epoch": 0.76078132927448,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.300972856320984e-06,
+      "loss": 1.32624426,
+      "memory(GiB)": 113.67,
+      "step": 29990,
+      "train_speed(iter/s)": 1.319674
+    },
+    {
+      "acc": 0.68907251,
+      "epoch": 0.760908168442415,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.3000418186546754e-06,
+      "loss": 1.38318892,
+      "memory(GiB)": 113.67,
+      "step": 29995,
+      "train_speed(iter/s)": 1.319699
+    },
+    {
+      "acc": 0.70145431,
+      "epoch": 0.76103500761035,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.299110679822258e-06,
+      "loss": 1.2880621,
+      "memory(GiB)": 113.67,
+      "step": 30000,
+      "train_speed(iter/s)": 1.319723
+    },
+    {
+      "epoch": 0.76103500761035,
+      "eval_acc": 0.6744481129021817,
+      "eval_loss": 1.359262228012085,
+      "eval_runtime": 69.7128,
+      "eval_samples_per_second": 91.375,
+      "eval_steps_per_second": 22.851,
+      "step": 30000
+    },
+    {
+      "acc": 0.67641735,
+      "epoch": 0.7611618467782851,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.298179439864689e-06,
+      "loss": 1.43534031,
+      "memory(GiB)": 113.67,
+      "step": 30005,
+      "train_speed(iter/s)": 1.314984
+    },
+    {
+      "acc": 0.69810486,
+      "epoch": 0.7612886859462202,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.297248098822926e-06,
+      "loss": 1.35555515,
+      "memory(GiB)": 113.67,
+      "step": 30010,
+      "train_speed(iter/s)": 1.315008
+    },
+    {
+      "acc": 0.68621178,
+      "epoch": 0.7614155251141552,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.296316656737936e-06,
+      "loss": 1.4015996,
+      "memory(GiB)": 113.67,
+      "step": 30015,
+      "train_speed(iter/s)": 1.315033
+    },
+    {
+      "acc": 0.70754123,
+      "epoch": 0.7615423642820903,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.295385113650689e-06,
+      "loss": 1.35690441,
+      "memory(GiB)": 113.67,
+      "step": 30020,
+      "train_speed(iter/s)": 1.315058
+    },
+    {
+      "acc": 0.68446321,
+      "epoch": 0.7616692034500254,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.294453469602154e-06,
+      "loss": 1.37473183,
+      "memory(GiB)": 113.67,
+      "step": 30025,
+      "train_speed(iter/s)": 1.315082
+    },
+    {
+      "acc": 0.70236073,
+      "epoch": 0.7617960426179604,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.293521724633313e-06,
+      "loss": 1.34713039,
+      "memory(GiB)": 113.67,
+      "step": 30030,
+      "train_speed(iter/s)": 1.315107
+    },
+    {
+      "acc": 0.68575668,
+      "epoch": 0.7619228817858955,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.2925898787851455e-06,
+      "loss": 1.38116808,
+      "memory(GiB)": 113.67,
+      "step": 30035,
+      "train_speed(iter/s)": 1.31513
+    },
+    {
+      "acc": 0.69483089,
+      "epoch": 0.7620497209538305,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.2916579320986415e-06,
+      "loss": 1.35119324,
+      "memory(GiB)": 113.67,
+      "step": 30040,
+      "train_speed(iter/s)": 1.315154
+    },
+    {
+      "acc": 0.68575354,
+      "epoch": 0.7621765601217656,
+      "grad_norm": 3.25,
+      "learning_rate": 7.290725884614787e-06,
+      "loss": 1.41885071,
+      "memory(GiB)": 113.67,
+      "step": 30045,
+      "train_speed(iter/s)": 1.315179
+    },
+    {
+      "acc": 0.68503342,
+      "epoch": 0.7623033992897007,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.2897937363745844e-06,
+      "loss": 1.38336697,
+      "memory(GiB)": 113.67,
+      "step": 30050,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.68372426,
+      "epoch": 0.7624302384576357,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.2888614874190276e-06,
+      "loss": 1.41280098,
+      "memory(GiB)": 113.67,
+      "step": 30055,
+      "train_speed(iter/s)": 1.315227
+    },
+    {
+      "acc": 0.68902583,
+      "epoch": 0.7625570776255708,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.287929137789124e-06,
+      "loss": 1.41078978,
+      "memory(GiB)": 113.67,
+      "step": 30060,
+      "train_speed(iter/s)": 1.315251
+    },
+    {
+      "acc": 0.69904242,
+      "epoch": 0.7626839167935059,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.286996687525882e-06,
+      "loss": 1.39792233,
+      "memory(GiB)": 113.67,
+      "step": 30065,
+      "train_speed(iter/s)": 1.315276
+    },
+    {
+      "acc": 0.67936726,
+      "epoch": 0.7628107559614409,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.2860641366703155e-06,
+      "loss": 1.39322643,
+      "memory(GiB)": 113.67,
+      "step": 30070,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.68083167,
+      "epoch": 0.762937595129376,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.285131485263441e-06,
+      "loss": 1.4207037,
+      "memory(GiB)": 113.67,
+      "step": 30075,
+      "train_speed(iter/s)": 1.315324
+    },
+    {
+      "acc": 0.68012185,
+      "epoch": 0.763064434297311,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.2841987333462815e-06,
+      "loss": 1.477913,
+      "memory(GiB)": 113.67,
+      "step": 30080,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.68607588,
+      "epoch": 0.7631912734652461,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.283265880959863e-06,
+      "loss": 1.4026495,
+      "memory(GiB)": 113.67,
+      "step": 30085,
+      "train_speed(iter/s)": 1.315371
+    },
+    {
+      "acc": 0.68016901,
+      "epoch": 0.7633181126331812,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.282332928145219e-06,
+      "loss": 1.38390303,
+      "memory(GiB)": 113.67,
+      "step": 30090,
+      "train_speed(iter/s)": 1.315395
+    },
+    {
+      "acc": 0.68671894,
+      "epoch": 0.7634449518011162,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.281399874943381e-06,
+      "loss": 1.3525898,
+      "memory(GiB)": 113.67,
+      "step": 30095,
+      "train_speed(iter/s)": 1.315419
+    },
+    {
+      "acc": 0.68391895,
+      "epoch": 0.7635717909690513,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.280466721395393e-06,
+      "loss": 1.47045822,
+      "memory(GiB)": 113.67,
+      "step": 30100,
+      "train_speed(iter/s)": 1.315444
+    },
+    {
+      "acc": 0.67367048,
+      "epoch": 0.7636986301369864,
+      "grad_norm": 2.75,
+      "learning_rate": 7.279533467542295e-06,
+      "loss": 1.40211143,
+      "memory(GiB)": 113.67,
+      "step": 30105,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.6785614,
+      "epoch": 0.7638254693049213,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.2786001134251385e-06,
+      "loss": 1.43100433,
+      "memory(GiB)": 113.67,
+      "step": 30110,
+      "train_speed(iter/s)": 1.31549
+    },
+    {
+      "acc": 0.6894804,
+      "epoch": 0.7639523084728564,
+      "grad_norm": 3.375,
+      "learning_rate": 7.2776666590849744e-06,
+      "loss": 1.37026005,
+      "memory(GiB)": 113.67,
+      "step": 30115,
+      "train_speed(iter/s)": 1.315512
+    },
+    {
+      "acc": 0.69595881,
+      "epoch": 0.7640791476407914,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.276733104562863e-06,
+      "loss": 1.36138821,
+      "memory(GiB)": 113.67,
+      "step": 30120,
+      "train_speed(iter/s)": 1.315536
+    },
+    {
+      "acc": 0.68862791,
+      "epoch": 0.7642059868087265,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.275799449899865e-06,
+      "loss": 1.40554504,
+      "memory(GiB)": 113.67,
+      "step": 30125,
+      "train_speed(iter/s)": 1.31556
+    },
+    {
+      "acc": 0.69306192,
+      "epoch": 0.7643328259766616,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.274865695137046e-06,
+      "loss": 1.40636997,
+      "memory(GiB)": 113.67,
+      "step": 30130,
+      "train_speed(iter/s)": 1.315585
+    },
+    {
+      "acc": 0.687358,
+      "epoch": 0.7644596651445966,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.273931840315477e-06,
+      "loss": 1.36586132,
+      "memory(GiB)": 113.67,
+      "step": 30135,
+      "train_speed(iter/s)": 1.31561
+    },
+    {
+      "acc": 0.69770813,
+      "epoch": 0.7645865043125317,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.272997885476234e-06,
+      "loss": 1.38004627,
+      "memory(GiB)": 113.67,
+      "step": 30140,
+      "train_speed(iter/s)": 1.315635
+    },
+    {
+      "acc": 0.67457657,
+      "epoch": 0.7647133434804668,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.272063830660395e-06,
+      "loss": 1.48367147,
+      "memory(GiB)": 113.67,
+      "step": 30145,
+      "train_speed(iter/s)": 1.315659
+    },
+    {
+      "acc": 0.6915987,
+      "epoch": 0.7648401826484018,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.271129675909046e-06,
+      "loss": 1.35797005,
+      "memory(GiB)": 113.67,
+      "step": 30150,
+      "train_speed(iter/s)": 1.315684
+    },
+    {
+      "acc": 0.70263929,
+      "epoch": 0.7649670218163369,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.270195421263271e-06,
+      "loss": 1.30822811,
+      "memory(GiB)": 113.67,
+      "step": 30155,
+      "train_speed(iter/s)": 1.315709
+    },
+    {
+      "acc": 0.69195075,
+      "epoch": 0.7650938609842719,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.269261066764169e-06,
+      "loss": 1.40215206,
+      "memory(GiB)": 113.67,
+      "step": 30160,
+      "train_speed(iter/s)": 1.315733
+    },
+    {
+      "acc": 0.67493525,
+      "epoch": 0.765220700152207,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.268326612452832e-06,
+      "loss": 1.38746786,
+      "memory(GiB)": 113.67,
+      "step": 30165,
+      "train_speed(iter/s)": 1.315759
+    },
+    {
+      "acc": 0.68665209,
+      "epoch": 0.7653475393201421,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.267392058370364e-06,
+      "loss": 1.36762686,
+      "memory(GiB)": 113.67,
+      "step": 30170,
+      "train_speed(iter/s)": 1.315774
+    },
+    {
+      "acc": 0.69249601,
+      "epoch": 0.7654743784880771,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.2664574045578685e-06,
+      "loss": 1.37832632,
+      "memory(GiB)": 113.67,
+      "step": 30175,
+      "train_speed(iter/s)": 1.315799
+    },
+    {
+      "acc": 0.68525286,
+      "epoch": 0.7656012176560122,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.26552265105646e-06,
+      "loss": 1.39479246,
+      "memory(GiB)": 113.67,
+      "step": 30180,
+      "train_speed(iter/s)": 1.315824
+    },
+    {
+      "acc": 0.66983509,
+      "epoch": 0.7657280568239473,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.264587797907248e-06,
+      "loss": 1.40285006,
+      "memory(GiB)": 113.67,
+      "step": 30185,
+      "train_speed(iter/s)": 1.315848
+    },
+    {
+      "acc": 0.68778305,
+      "epoch": 0.7658548959918823,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.263652845151354e-06,
+      "loss": 1.36423254,
+      "memory(GiB)": 113.67,
+      "step": 30190,
+      "train_speed(iter/s)": 1.315873
+    },
+    {
+      "acc": 0.6807303,
+      "epoch": 0.7659817351598174,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.262717792829903e-06,
+      "loss": 1.38847332,
+      "memory(GiB)": 113.67,
+      "step": 30195,
+      "train_speed(iter/s)": 1.315898
+    },
+    {
+      "acc": 0.69056149,
+      "epoch": 0.7661085743277524,
+      "grad_norm": 4.0625,
+      "learning_rate": 7.261782640984021e-06,
+      "loss": 1.37142868,
+      "memory(GiB)": 113.67,
+      "step": 30200,
+      "train_speed(iter/s)": 1.315924
+    },
+    {
+      "acc": 0.66872005,
+      "epoch": 0.7662354134956875,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.26084738965484e-06,
+      "loss": 1.43911476,
+      "memory(GiB)": 113.67,
+      "step": 30205,
+      "train_speed(iter/s)": 1.315948
+    },
+    {
+      "acc": 0.68405061,
+      "epoch": 0.7663622526636226,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.2599120388834964e-06,
+      "loss": 1.43216038,
+      "memory(GiB)": 113.67,
+      "step": 30210,
+      "train_speed(iter/s)": 1.315972
+    },
+    {
+      "acc": 0.70065498,
+      "epoch": 0.7664890918315576,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.258976588711133e-06,
+      "loss": 1.37288361,
+      "memory(GiB)": 113.67,
+      "step": 30215,
+      "train_speed(iter/s)": 1.315996
+    },
+    {
+      "acc": 0.69600725,
+      "epoch": 0.7666159309994927,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.258041039178891e-06,
+      "loss": 1.35124512,
+      "memory(GiB)": 113.67,
+      "step": 30220,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.67030382,
+      "epoch": 0.7667427701674278,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.257105390327925e-06,
+      "loss": 1.4325532,
+      "memory(GiB)": 113.67,
+      "step": 30225,
+      "train_speed(iter/s)": 1.316042
+    },
+    {
+      "acc": 0.68499804,
+      "epoch": 0.7668696093353627,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.256169642199386e-06,
+      "loss": 1.35857868,
+      "memory(GiB)": 113.67,
+      "step": 30230,
+      "train_speed(iter/s)": 1.316065
+    },
+    {
+      "acc": 0.67457972,
+      "epoch": 0.7669964485032978,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.255233794834432e-06,
+      "loss": 1.40308342,
+      "memory(GiB)": 113.67,
+      "step": 30235,
+      "train_speed(iter/s)": 1.316091
+    },
+    {
+      "acc": 0.71001234,
+      "epoch": 0.7671232876712328,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.254297848274229e-06,
+      "loss": 1.31059313,
+      "memory(GiB)": 113.67,
+      "step": 30240,
+      "train_speed(iter/s)": 1.316116
+    },
+    {
+      "acc": 0.6877696,
+      "epoch": 0.7672501268391679,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.25336180255994e-06,
+      "loss": 1.38808985,
+      "memory(GiB)": 113.67,
+      "step": 30245,
+      "train_speed(iter/s)": 1.31614
+    },
+    {
+      "acc": 0.69043789,
+      "epoch": 0.767376966007103,
+      "grad_norm": 2.625,
+      "learning_rate": 7.25242565773274e-06,
+      "loss": 1.45414391,
+      "memory(GiB)": 113.67,
+      "step": 30250,
+      "train_speed(iter/s)": 1.316162
+    },
+    {
+      "acc": 0.6782485,
+      "epoch": 0.767503805175038,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.251489413833801e-06,
+      "loss": 1.40594034,
+      "memory(GiB)": 113.67,
+      "step": 30255,
+      "train_speed(iter/s)": 1.316186
+    },
+    {
+      "acc": 0.68014684,
+      "epoch": 0.7676306443429731,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.250553070904307e-06,
+      "loss": 1.41767273,
+      "memory(GiB)": 113.67,
+      "step": 30260,
+      "train_speed(iter/s)": 1.316211
+    },
+    {
+      "acc": 0.68799729,
+      "epoch": 0.7677574835109082,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.2496166289854404e-06,
+      "loss": 1.35672722,
+      "memory(GiB)": 113.67,
+      "step": 30265,
+      "train_speed(iter/s)": 1.316236
+    },
+    {
+      "acc": 0.69104619,
+      "epoch": 0.7678843226788432,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.24868008811839e-06,
+      "loss": 1.40243216,
+      "memory(GiB)": 113.67,
+      "step": 30270,
+      "train_speed(iter/s)": 1.31626
+    },
+    {
+      "acc": 0.68647485,
+      "epoch": 0.7680111618467783,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.247743448344351e-06,
+      "loss": 1.3925766,
+      "memory(GiB)": 113.67,
+      "step": 30275,
+      "train_speed(iter/s)": 1.316285
+    },
+    {
+      "acc": 0.67325573,
+      "epoch": 0.7681380010147133,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.246806709704519e-06,
+      "loss": 1.47298594,
+      "memory(GiB)": 113.67,
+      "step": 30280,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.6788219,
+      "epoch": 0.7682648401826484,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.245869872240098e-06,
+      "loss": 1.46145535,
+      "memory(GiB)": 113.67,
+      "step": 30285,
+      "train_speed(iter/s)": 1.316333
+    },
+    {
+      "acc": 0.69054637,
+      "epoch": 0.7683916793505835,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.244932935992292e-06,
+      "loss": 1.42510166,
+      "memory(GiB)": 113.67,
+      "step": 30290,
+      "train_speed(iter/s)": 1.316356
+    },
+    {
+      "acc": 0.68782034,
+      "epoch": 0.7685185185185185,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.243995901002312e-06,
+      "loss": 1.43920326,
+      "memory(GiB)": 113.67,
+      "step": 30295,
+      "train_speed(iter/s)": 1.316381
+    },
+    {
+      "acc": 0.68247776,
+      "epoch": 0.7686453576864536,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.243058767311374e-06,
+      "loss": 1.39385929,
+      "memory(GiB)": 113.67,
+      "step": 30300,
+      "train_speed(iter/s)": 1.316407
+    },
+    {
+      "acc": 0.70461597,
+      "epoch": 0.7687721968543887,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.2421215349606955e-06,
+      "loss": 1.3382165,
+      "memory(GiB)": 113.67,
+      "step": 30305,
+      "train_speed(iter/s)": 1.316431
+    },
+    {
+      "acc": 0.70128059,
+      "epoch": 0.7688990360223237,
+      "grad_norm": 3.25,
+      "learning_rate": 7.241184203991505e-06,
+      "loss": 1.28685436,
+      "memory(GiB)": 113.67,
+      "step": 30310,
+      "train_speed(iter/s)": 1.316456
+    },
+    {
+      "acc": 0.67852907,
+      "epoch": 0.7690258751902588,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.240246774445024e-06,
+      "loss": 1.38353853,
+      "memory(GiB)": 113.67,
+      "step": 30315,
+      "train_speed(iter/s)": 1.316481
+    },
+    {
+      "acc": 0.69082146,
+      "epoch": 0.7691527143581938,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.23930924636249e-06,
+      "loss": 1.41621876,
+      "memory(GiB)": 113.67,
+      "step": 30320,
+      "train_speed(iter/s)": 1.316506
+    },
+    {
+      "acc": 0.67680254,
+      "epoch": 0.7692795535261289,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.238371619785134e-06,
+      "loss": 1.38411026,
+      "memory(GiB)": 113.67,
+      "step": 30325,
+      "train_speed(iter/s)": 1.316531
+    },
+    {
+      "acc": 0.68401985,
+      "epoch": 0.769406392694064,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.237433894754205e-06,
+      "loss": 1.41572504,
+      "memory(GiB)": 113.67,
+      "step": 30330,
+      "train_speed(iter/s)": 1.316555
+    },
+    {
+      "acc": 0.68570919,
+      "epoch": 0.769533231861999,
+      "grad_norm": 3.125,
+      "learning_rate": 7.23649607131094e-06,
+      "loss": 1.42964897,
+      "memory(GiB)": 113.67,
+      "step": 30335,
+      "train_speed(iter/s)": 1.316581
+    },
+    {
+      "acc": 0.67758908,
+      "epoch": 0.7696600710299341,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.235558149496595e-06,
+      "loss": 1.39897385,
+      "memory(GiB)": 113.67,
+      "step": 30340,
+      "train_speed(iter/s)": 1.316606
+    },
+    {
+      "acc": 0.69572129,
+      "epoch": 0.7697869101978692,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.23462012935242e-06,
+      "loss": 1.38022728,
+      "memory(GiB)": 113.67,
+      "step": 30345,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.67221985,
+      "epoch": 0.7699137493658041,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.233682010919676e-06,
+      "loss": 1.4199976,
+      "memory(GiB)": 113.67,
+      "step": 30350,
+      "train_speed(iter/s)": 1.316656
+    },
+    {
+      "acc": 0.69942098,
+      "epoch": 0.7700405885337392,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.2327437942396236e-06,
+      "loss": 1.38192921,
+      "memory(GiB)": 113.67,
+      "step": 30355,
+      "train_speed(iter/s)": 1.316681
+    },
+    {
+      "acc": 0.69931097,
+      "epoch": 0.7701674277016742,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.231805479353532e-06,
+      "loss": 1.32770567,
+      "memory(GiB)": 113.67,
+      "step": 30360,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.69544115,
+      "epoch": 0.7702942668696093,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.2308670663026705e-06,
+      "loss": 1.42126331,
+      "memory(GiB)": 113.67,
+      "step": 30365,
+      "train_speed(iter/s)": 1.316729
+    },
+    {
+      "acc": 0.67652836,
+      "epoch": 0.7704211060375444,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.229928555128315e-06,
+      "loss": 1.43443966,
+      "memory(GiB)": 113.67,
+      "step": 30370,
+      "train_speed(iter/s)": 1.316754
+    },
+    {
+      "acc": 0.69262877,
+      "epoch": 0.7705479452054794,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.228989945871745e-06,
+      "loss": 1.41435919,
+      "memory(GiB)": 113.67,
+      "step": 30375,
+      "train_speed(iter/s)": 1.316778
+    },
+    {
+      "acc": 0.68889165,
+      "epoch": 0.7706747843734145,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.2280512385742475e-06,
+      "loss": 1.36180515,
+      "memory(GiB)": 113.67,
+      "step": 30380,
+      "train_speed(iter/s)": 1.316803
+    },
+    {
+      "acc": 0.68874531,
+      "epoch": 0.7708016235413496,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.227112433277107e-06,
+      "loss": 1.37222872,
+      "memory(GiB)": 113.67,
+      "step": 30385,
+      "train_speed(iter/s)": 1.316828
+    },
+    {
+      "acc": 0.6877984,
+      "epoch": 0.7709284627092846,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.2261735300216195e-06,
+      "loss": 1.41464119,
+      "memory(GiB)": 113.67,
+      "step": 30390,
+      "train_speed(iter/s)": 1.316852
+    },
+    {
+      "acc": 0.68156738,
+      "epoch": 0.7710553018772197,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.22523452884908e-06,
+      "loss": 1.42270813,
+      "memory(GiB)": 113.67,
+      "step": 30395,
+      "train_speed(iter/s)": 1.316877
+    },
+    {
+      "acc": 0.67297034,
+      "epoch": 0.7711821410451547,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.224295429800792e-06,
+      "loss": 1.45652618,
+      "memory(GiB)": 113.67,
+      "step": 30400,
+      "train_speed(iter/s)": 1.316901
+    },
+    {
+      "acc": 0.68924665,
+      "epoch": 0.7713089802130898,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.22335623291806e-06,
+      "loss": 1.4219244,
+      "memory(GiB)": 113.67,
+      "step": 30405,
+      "train_speed(iter/s)": 1.316926
+    },
+    {
+      "acc": 0.67738609,
+      "epoch": 0.7714358193810249,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.222416938242194e-06,
+      "loss": 1.43820457,
+      "memory(GiB)": 113.67,
+      "step": 30410,
+      "train_speed(iter/s)": 1.31695
+    },
+    {
+      "acc": 0.70434909,
+      "epoch": 0.7715626585489599,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.221477545814509e-06,
+      "loss": 1.27047071,
+      "memory(GiB)": 113.67,
+      "step": 30415,
+      "train_speed(iter/s)": 1.316975
+    },
+    {
+      "acc": 0.6864954,
+      "epoch": 0.771689497716895,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.220538055676323e-06,
+      "loss": 1.39206257,
+      "memory(GiB)": 113.67,
+      "step": 30420,
+      "train_speed(iter/s)": 1.316997
+    },
+    {
+      "acc": 0.68324876,
+      "epoch": 0.7718163368848301,
+      "grad_norm": 4.9375,
+      "learning_rate": 7.21959846786896e-06,
+      "loss": 1.4000473,
+      "memory(GiB)": 113.67,
+      "step": 30425,
+      "train_speed(iter/s)": 1.31702
+    },
+    {
+      "acc": 0.69610643,
+      "epoch": 0.7719431760527651,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.218658782433746e-06,
+      "loss": 1.36320305,
+      "memory(GiB)": 113.67,
+      "step": 30430,
+      "train_speed(iter/s)": 1.317044
+    },
+    {
+      "acc": 0.69669318,
+      "epoch": 0.7720700152207002,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.217718999412013e-06,
+      "loss": 1.31981726,
+      "memory(GiB)": 113.67,
+      "step": 30435,
+      "train_speed(iter/s)": 1.317068
+    },
+    {
+      "acc": 0.68856225,
+      "epoch": 0.7721968543886352,
+      "grad_norm": 5.75,
+      "learning_rate": 7.216779118845097e-06,
+      "loss": 1.39628458,
+      "memory(GiB)": 113.67,
+      "step": 30440,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.67804375,
+      "epoch": 0.7723236935565703,
+      "grad_norm": 3.5,
+      "learning_rate": 7.215839140774339e-06,
+      "loss": 1.38914566,
+      "memory(GiB)": 113.67,
+      "step": 30445,
+      "train_speed(iter/s)": 1.317115
+    },
+    {
+      "acc": 0.6906167,
+      "epoch": 0.7724505327245054,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.214899065241082e-06,
+      "loss": 1.42288208,
+      "memory(GiB)": 113.67,
+      "step": 30450,
+      "train_speed(iter/s)": 1.31714
+    },
+    {
+      "acc": 0.70115242,
+      "epoch": 0.7725773718924404,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.213958892286674e-06,
+      "loss": 1.39413815,
+      "memory(GiB)": 113.67,
+      "step": 30455,
+      "train_speed(iter/s)": 1.317164
+    },
+    {
+      "acc": 0.6864274,
+      "epoch": 0.7727042110603755,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.213018621952472e-06,
+      "loss": 1.34596367,
+      "memory(GiB)": 113.67,
+      "step": 30460,
+      "train_speed(iter/s)": 1.317188
+    },
+    {
+      "acc": 0.68983059,
+      "epoch": 0.7728310502283106,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.212078254279828e-06,
+      "loss": 1.34375648,
+      "memory(GiB)": 113.67,
+      "step": 30465,
+      "train_speed(iter/s)": 1.317212
+    },
+    {
+      "acc": 0.67261934,
+      "epoch": 0.7729578893962455,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.211137789310109e-06,
+      "loss": 1.45123777,
+      "memory(GiB)": 113.67,
+      "step": 30470,
+      "train_speed(iter/s)": 1.317236
+    },
+    {
+      "acc": 0.70430737,
+      "epoch": 0.7730847285641806,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.2101972270846756e-06,
+      "loss": 1.3013237,
+      "memory(GiB)": 113.67,
+      "step": 30475,
+      "train_speed(iter/s)": 1.31726
+    },
+    {
+      "acc": 0.70264711,
+      "epoch": 0.7732115677321156,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.2092565676449e-06,
+      "loss": 1.29550285,
+      "memory(GiB)": 113.67,
+      "step": 30480,
+      "train_speed(iter/s)": 1.317284
+    },
+    {
+      "acc": 0.70684695,
+      "epoch": 0.7733384069000507,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.208315811032158e-06,
+      "loss": 1.28708429,
+      "memory(GiB)": 113.67,
+      "step": 30485,
+      "train_speed(iter/s)": 1.317309
+    },
+    {
+      "acc": 0.68628988,
+      "epoch": 0.7734652460679858,
+      "grad_norm": 3.375,
+      "learning_rate": 7.207374957287828e-06,
+      "loss": 1.3859149,
+      "memory(GiB)": 113.67,
+      "step": 30490,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.68056984,
+      "epoch": 0.7735920852359208,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.2064340064532914e-06,
+      "loss": 1.39961996,
+      "memory(GiB)": 113.67,
+      "step": 30495,
+      "train_speed(iter/s)": 1.317355
+    },
+    {
+      "acc": 0.66454611,
+      "epoch": 0.7737189244038559,
+      "grad_norm": 3.125,
+      "learning_rate": 7.205492958569936e-06,
+      "loss": 1.44336834,
+      "memory(GiB)": 113.67,
+      "step": 30500,
+      "train_speed(iter/s)": 1.31738
+    },
+    {
+      "acc": 0.6900589,
+      "epoch": 0.773845763571791,
+      "grad_norm": 2.75,
+      "learning_rate": 7.204551813679154e-06,
+      "loss": 1.46152267,
+      "memory(GiB)": 113.67,
+      "step": 30505,
+      "train_speed(iter/s)": 1.317403
+    },
+    {
+      "acc": 0.67794743,
+      "epoch": 0.773972602739726,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.2036105718223405e-06,
+      "loss": 1.42264347,
+      "memory(GiB)": 113.67,
+      "step": 30510,
+      "train_speed(iter/s)": 1.317422
+    },
+    {
+      "acc": 0.69072948,
+      "epoch": 0.7740994419076611,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.202669233040896e-06,
+      "loss": 1.4130805,
+      "memory(GiB)": 113.67,
+      "step": 30515,
+      "train_speed(iter/s)": 1.317446
+    },
+    {
+      "acc": 0.68666182,
+      "epoch": 0.7742262810755961,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.201727797376223e-06,
+      "loss": 1.43134127,
+      "memory(GiB)": 113.67,
+      "step": 30520,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.6899961,
+      "epoch": 0.7743531202435312,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.200786264869732e-06,
+      "loss": 1.4291502,
+      "memory(GiB)": 113.67,
+      "step": 30525,
+      "train_speed(iter/s)": 1.317492
+    },
+    {
+      "acc": 0.67785683,
+      "epoch": 0.7744799594114663,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.199844635562836e-06,
+      "loss": 1.38991222,
+      "memory(GiB)": 113.67,
+      "step": 30530,
+      "train_speed(iter/s)": 1.317515
+    },
+    {
+      "acc": 0.67929306,
+      "epoch": 0.7746067985794013,
+      "grad_norm": 2.875,
+      "learning_rate": 7.19890290949695e-06,
+      "loss": 1.4373044,
+      "memory(GiB)": 113.67,
+      "step": 30535,
+      "train_speed(iter/s)": 1.317539
+    },
+    {
+      "acc": 0.67280674,
+      "epoch": 0.7747336377473364,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.197961086713498e-06,
+      "loss": 1.42560406,
+      "memory(GiB)": 113.67,
+      "step": 30540,
+      "train_speed(iter/s)": 1.317563
+    },
+    {
+      "acc": 0.6803925,
+      "epoch": 0.7748604769152715,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.197019167253904e-06,
+      "loss": 1.39786949,
+      "memory(GiB)": 113.67,
+      "step": 30545,
+      "train_speed(iter/s)": 1.317588
+    },
+    {
+      "acc": 0.69125543,
+      "epoch": 0.7749873160832065,
+      "grad_norm": 2.40625,
+      "learning_rate": 7.196077151159597e-06,
+      "loss": 1.40877953,
+      "memory(GiB)": 113.67,
+      "step": 30550,
+      "train_speed(iter/s)": 1.317611
+    },
+    {
+      "acc": 0.69618201,
+      "epoch": 0.7751141552511416,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.195135038472013e-06,
+      "loss": 1.37797966,
+      "memory(GiB)": 113.67,
+      "step": 30555,
+      "train_speed(iter/s)": 1.317634
+    },
+    {
+      "acc": 0.7037106,
+      "epoch": 0.7752409944190766,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.194192829232589e-06,
+      "loss": 1.37076273,
+      "memory(GiB)": 113.67,
+      "step": 30560,
+      "train_speed(iter/s)": 1.317655
+    },
+    {
+      "acc": 0.68289022,
+      "epoch": 0.7753678335870117,
+      "grad_norm": 3.625,
+      "learning_rate": 7.1932505234827686e-06,
+      "loss": 1.370012,
+      "memory(GiB)": 113.67,
+      "step": 30565,
+      "train_speed(iter/s)": 1.317678
+    },
+    {
+      "acc": 0.67811503,
+      "epoch": 0.7754946727549468,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.192308121263998e-06,
+      "loss": 1.44373436,
+      "memory(GiB)": 113.67,
+      "step": 30570,
+      "train_speed(iter/s)": 1.317702
+    },
+    {
+      "acc": 0.69041891,
+      "epoch": 0.7756215119228818,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.191365622617728e-06,
+      "loss": 1.37959747,
+      "memory(GiB)": 113.67,
+      "step": 30575,
+      "train_speed(iter/s)": 1.317724
+    },
+    {
+      "acc": 0.68141842,
+      "epoch": 0.7757483510908169,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.190423027585414e-06,
+      "loss": 1.44024076,
+      "memory(GiB)": 113.67,
+      "step": 30580,
+      "train_speed(iter/s)": 1.317747
+    },
+    {
+      "acc": 0.68911877,
+      "epoch": 0.775875190258752,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.189480336208516e-06,
+      "loss": 1.40639906,
+      "memory(GiB)": 113.67,
+      "step": 30585,
+      "train_speed(iter/s)": 1.31777
+    },
+    {
+      "acc": 0.6846694,
+      "epoch": 0.776002029426687,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.188537548528498e-06,
+      "loss": 1.39391956,
+      "memory(GiB)": 113.67,
+      "step": 30590,
+      "train_speed(iter/s)": 1.317793
+    },
+    {
+      "acc": 0.67793961,
+      "epoch": 0.776128868594622,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.187594664586826e-06,
+      "loss": 1.40813026,
+      "memory(GiB)": 113.67,
+      "step": 30595,
+      "train_speed(iter/s)": 1.317814
+    },
+    {
+      "acc": 0.68579836,
+      "epoch": 0.776255707762557,
+      "grad_norm": 2.875,
+      "learning_rate": 7.186651684424975e-06,
+      "loss": 1.44583187,
+      "memory(GiB)": 113.67,
+      "step": 30600,
+      "train_speed(iter/s)": 1.317837
+    },
+    {
+      "acc": 0.68004799,
+      "epoch": 0.7763825469304921,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.185708608084418e-06,
+      "loss": 1.38944969,
+      "memory(GiB)": 113.67,
+      "step": 30605,
+      "train_speed(iter/s)": 1.317861
+    },
+    {
+      "acc": 0.69272537,
+      "epoch": 0.7765093860984272,
+      "grad_norm": 2.75,
+      "learning_rate": 7.184765435606642e-06,
+      "loss": 1.37125263,
+      "memory(GiB)": 113.67,
+      "step": 30610,
+      "train_speed(iter/s)": 1.317885
+    },
+    {
+      "acc": 0.6996707,
+      "epoch": 0.7766362252663622,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.183822167033124e-06,
+      "loss": 1.40389719,
+      "memory(GiB)": 113.67,
+      "step": 30615,
+      "train_speed(iter/s)": 1.317908
+    },
+    {
+      "acc": 0.68469076,
+      "epoch": 0.7767630644342973,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.18287880240536e-06,
+      "loss": 1.40211849,
+      "memory(GiB)": 113.67,
+      "step": 30620,
+      "train_speed(iter/s)": 1.317931
+    },
+    {
+      "acc": 0.70554962,
+      "epoch": 0.7768899036022324,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.1819353417648386e-06,
+      "loss": 1.33036575,
+      "memory(GiB)": 113.67,
+      "step": 30625,
+      "train_speed(iter/s)": 1.317954
+    },
+    {
+      "acc": 0.68741922,
+      "epoch": 0.7770167427701674,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.180991785153059e-06,
+      "loss": 1.39756718,
+      "memory(GiB)": 113.67,
+      "step": 30630,
+      "train_speed(iter/s)": 1.317977
+    },
+    {
+      "acc": 0.68957319,
+      "epoch": 0.7771435819381025,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.180048132611524e-06,
+      "loss": 1.36270828,
+      "memory(GiB)": 113.67,
+      "step": 30635,
+      "train_speed(iter/s)": 1.318
+    },
+    {
+      "acc": 0.68148265,
+      "epoch": 0.7772704211060375,
+      "grad_norm": 5.5,
+      "learning_rate": 7.17910438418174e-06,
+      "loss": 1.36367054,
+      "memory(GiB)": 113.67,
+      "step": 30640,
+      "train_speed(iter/s)": 1.318023
+    },
+    {
+      "acc": 0.68685274,
+      "epoch": 0.7773972602739726,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.178160539905214e-06,
+      "loss": 1.4353426,
+      "memory(GiB)": 113.67,
+      "step": 30645,
+      "train_speed(iter/s)": 1.318048
+    },
+    {
+      "acc": 0.69961448,
+      "epoch": 0.7775240994419077,
+      "grad_norm": 3.375,
+      "learning_rate": 7.1772165998234645e-06,
+      "loss": 1.35855875,
+      "memory(GiB)": 113.67,
+      "step": 30650,
+      "train_speed(iter/s)": 1.318068
+    },
+    {
+      "acc": 0.68374157,
+      "epoch": 0.7776509386098427,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.176272563978007e-06,
+      "loss": 1.41722116,
+      "memory(GiB)": 113.67,
+      "step": 30655,
+      "train_speed(iter/s)": 1.31809
+    },
+    {
+      "acc": 0.68447304,
+      "epoch": 0.7777777777777778,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.175328432410367e-06,
+      "loss": 1.40295048,
+      "memory(GiB)": 113.67,
+      "step": 30660,
+      "train_speed(iter/s)": 1.318114
+    },
+    {
+      "acc": 0.66929865,
+      "epoch": 0.7779046169457129,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.17438420516207e-06,
+      "loss": 1.44182653,
+      "memory(GiB)": 113.67,
+      "step": 30665,
+      "train_speed(iter/s)": 1.318136
+    },
+    {
+      "acc": 0.68981028,
+      "epoch": 0.7780314561136479,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.173439882274647e-06,
+      "loss": 1.32518377,
+      "memory(GiB)": 113.67,
+      "step": 30670,
+      "train_speed(iter/s)": 1.31816
+    },
+    {
+      "acc": 0.68247576,
+      "epoch": 0.778158295281583,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.172495463789635e-06,
+      "loss": 1.42057362,
+      "memory(GiB)": 113.67,
+      "step": 30675,
+      "train_speed(iter/s)": 1.318183
+    },
+    {
+      "acc": 0.68726411,
+      "epoch": 0.778285134449518,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.171550949748574e-06,
+      "loss": 1.40568914,
+      "memory(GiB)": 113.67,
+      "step": 30680,
+      "train_speed(iter/s)": 1.318206
+    },
+    {
+      "acc": 0.68728418,
+      "epoch": 0.7784119736174531,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.170606340193003e-06,
+      "loss": 1.38338985,
+      "memory(GiB)": 113.67,
+      "step": 30685,
+      "train_speed(iter/s)": 1.318231
+    },
+    {
+      "acc": 0.68481617,
+      "epoch": 0.7785388127853882,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.1696616351644786e-06,
+      "loss": 1.45908775,
+      "memory(GiB)": 113.67,
+      "step": 30690,
+      "train_speed(iter/s)": 1.318253
+    },
+    {
+      "acc": 0.69441175,
+      "epoch": 0.7786656519533232,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.168716834704546e-06,
+      "loss": 1.40716696,
+      "memory(GiB)": 113.67,
+      "step": 30695,
+      "train_speed(iter/s)": 1.318278
+    },
+    {
+      "acc": 0.68632264,
+      "epoch": 0.7787924911212583,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.167771938854766e-06,
+      "loss": 1.42775421,
+      "memory(GiB)": 113.67,
+      "step": 30700,
+      "train_speed(iter/s)": 1.318301
+    },
+    {
+      "acc": 0.67956052,
+      "epoch": 0.7789193302891934,
+      "grad_norm": 4.46875,
+      "learning_rate": 7.166826947656696e-06,
+      "loss": 1.49266291,
+      "memory(GiB)": 113.67,
+      "step": 30705,
+      "train_speed(iter/s)": 1.318325
+    },
+    {
+      "acc": 0.68232355,
+      "epoch": 0.7790461694571283,
+      "grad_norm": 2.578125,
+      "learning_rate": 7.165881861151904e-06,
+      "loss": 1.40551224,
+      "memory(GiB)": 113.67,
+      "step": 30710,
+      "train_speed(iter/s)": 1.318349
+    },
+    {
+      "acc": 0.69349833,
+      "epoch": 0.7791730086250634,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.164936679381957e-06,
+      "loss": 1.36166153,
+      "memory(GiB)": 113.67,
+      "step": 30715,
+      "train_speed(iter/s)": 1.318372
+    },
+    {
+      "acc": 0.68615475,
+      "epoch": 0.7792998477929984,
+      "grad_norm": 3.125,
+      "learning_rate": 7.16399140238843e-06,
+      "loss": 1.45096455,
+      "memory(GiB)": 113.67,
+      "step": 30720,
+      "train_speed(iter/s)": 1.318395
+    },
+    {
+      "acc": 0.68333549,
+      "epoch": 0.7794266869609335,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.163046030212899e-06,
+      "loss": 1.41760921,
+      "memory(GiB)": 113.67,
+      "step": 30725,
+      "train_speed(iter/s)": 1.318419
+    },
+    {
+      "acc": 0.69652367,
+      "epoch": 0.7795535261288686,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.1621005628969475e-06,
+      "loss": 1.32813358,
+      "memory(GiB)": 113.67,
+      "step": 30730,
+      "train_speed(iter/s)": 1.318442
+    },
+    {
+      "acc": 0.67834215,
+      "epoch": 0.7796803652968036,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.161155000482159e-06,
+      "loss": 1.47561102,
+      "memory(GiB)": 113.67,
+      "step": 30735,
+      "train_speed(iter/s)": 1.318466
+    },
+    {
+      "acc": 0.6665494,
+      "epoch": 0.7798072044647387,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.160209343010125e-06,
+      "loss": 1.45021648,
+      "memory(GiB)": 113.67,
+      "step": 30740,
+      "train_speed(iter/s)": 1.318488
+    },
+    {
+      "acc": 0.67589164,
+      "epoch": 0.7799340436326738,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.1592635905224386e-06,
+      "loss": 1.46041317,
+      "memory(GiB)": 113.67,
+      "step": 30745,
+      "train_speed(iter/s)": 1.318513
+    },
+    {
+      "acc": 0.68086658,
+      "epoch": 0.7800608828006088,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.1583177430606995e-06,
+      "loss": 1.4598567,
+      "memory(GiB)": 113.67,
+      "step": 30750,
+      "train_speed(iter/s)": 1.318537
+    },
+    {
+      "acc": 0.67591548,
+      "epoch": 0.7801877219685439,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.1573718006665095e-06,
+      "loss": 1.46653156,
+      "memory(GiB)": 113.67,
+      "step": 30755,
+      "train_speed(iter/s)": 1.318563
+    },
+    {
+      "acc": 0.67396512,
+      "epoch": 0.7803145611364789,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.156425763381477e-06,
+      "loss": 1.45425491,
+      "memory(GiB)": 113.67,
+      "step": 30760,
+      "train_speed(iter/s)": 1.318587
+    },
+    {
+      "acc": 0.69714117,
+      "epoch": 0.780441400304414,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.155479631247211e-06,
+      "loss": 1.3649106,
+      "memory(GiB)": 113.67,
+      "step": 30765,
+      "train_speed(iter/s)": 1.318611
+    },
+    {
+      "acc": 0.66884599,
+      "epoch": 0.7805682394723491,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.154533404305327e-06,
+      "loss": 1.44875097,
+      "memory(GiB)": 113.67,
+      "step": 30770,
+      "train_speed(iter/s)": 1.318634
+    },
+    {
+      "acc": 0.6973237,
+      "epoch": 0.7806950786402841,
+      "grad_norm": 3.125,
+      "learning_rate": 7.153587082597445e-06,
+      "loss": 1.26670265,
+      "memory(GiB)": 113.67,
+      "step": 30775,
+      "train_speed(iter/s)": 1.318658
+    },
+    {
+      "acc": 0.70294609,
+      "epoch": 0.7808219178082192,
+      "grad_norm": 3.0,
+      "learning_rate": 7.152640666165187e-06,
+      "loss": 1.32789497,
+      "memory(GiB)": 113.67,
+      "step": 30780,
+      "train_speed(iter/s)": 1.318682
+    },
+    {
+      "acc": 0.69850464,
+      "epoch": 0.7809487569761543,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.151694155050184e-06,
+      "loss": 1.34896336,
+      "memory(GiB)": 113.67,
+      "step": 30785,
+      "train_speed(iter/s)": 1.318706
+    },
+    {
+      "acc": 0.68417645,
+      "epoch": 0.7810755961440893,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.150747549294064e-06,
+      "loss": 1.43940029,
+      "memory(GiB)": 113.67,
+      "step": 30790,
+      "train_speed(iter/s)": 1.318728
+    },
+    {
+      "acc": 0.6910058,
+      "epoch": 0.7812024353120244,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.149800848938464e-06,
+      "loss": 1.36186428,
+      "memory(GiB)": 113.67,
+      "step": 30795,
+      "train_speed(iter/s)": 1.318752
+    },
+    {
+      "acc": 0.66682439,
+      "epoch": 0.7813292744799594,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.1488540540250254e-06,
+      "loss": 1.43736544,
+      "memory(GiB)": 113.67,
+      "step": 30800,
+      "train_speed(iter/s)": 1.318776
+    },
+    {
+      "acc": 0.68373523,
+      "epoch": 0.7814561136478945,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.14790716459539e-06,
+      "loss": 1.39208794,
+      "memory(GiB)": 113.67,
+      "step": 30805,
+      "train_speed(iter/s)": 1.318799
+    },
+    {
+      "acc": 0.69190168,
+      "epoch": 0.7815829528158296,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.146960180691209e-06,
+      "loss": 1.33850956,
+      "memory(GiB)": 113.67,
+      "step": 30810,
+      "train_speed(iter/s)": 1.318822
+    },
+    {
+      "acc": 0.67848897,
+      "epoch": 0.7817097919837646,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.146013102354133e-06,
+      "loss": 1.38863192,
+      "memory(GiB)": 113.67,
+      "step": 30815,
+      "train_speed(iter/s)": 1.318846
+    },
+    {
+      "acc": 0.69436989,
+      "epoch": 0.7818366311516997,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.145065929625821e-06,
+      "loss": 1.33943901,
+      "memory(GiB)": 113.67,
+      "step": 30820,
+      "train_speed(iter/s)": 1.31887
+    },
+    {
+      "acc": 0.68788071,
+      "epoch": 0.7819634703196348,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.1441186625479304e-06,
+      "loss": 1.41311741,
+      "memory(GiB)": 113.67,
+      "step": 30825,
+      "train_speed(iter/s)": 1.318893
+    },
+    {
+      "acc": 0.68508358,
+      "epoch": 0.7820903094875697,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.143171301162131e-06,
+      "loss": 1.40311947,
+      "memory(GiB)": 113.67,
+      "step": 30830,
+      "train_speed(iter/s)": 1.318918
+    },
+    {
+      "acc": 0.6598856,
+      "epoch": 0.7822171486555048,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.142223845510086e-06,
+      "loss": 1.49470901,
+      "memory(GiB)": 113.67,
+      "step": 30835,
+      "train_speed(iter/s)": 1.318942
+    },
+    {
+      "acc": 0.67294798,
+      "epoch": 0.7823439878234398,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.1412762956334746e-06,
+      "loss": 1.43070545,
+      "memory(GiB)": 113.67,
+      "step": 30840,
+      "train_speed(iter/s)": 1.318965
+    },
+    {
+      "acc": 0.68307133,
+      "epoch": 0.7824708269913749,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.140328651573969e-06,
+      "loss": 1.38389912,
+      "memory(GiB)": 113.67,
+      "step": 30845,
+      "train_speed(iter/s)": 1.318988
+    },
+    {
+      "acc": 0.68127308,
+      "epoch": 0.78259766615931,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.139380913373255e-06,
+      "loss": 1.40732079,
+      "memory(GiB)": 113.67,
+      "step": 30850,
+      "train_speed(iter/s)": 1.319012
+    },
+    {
+      "acc": 0.67977428,
+      "epoch": 0.782724505327245,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.138433081073017e-06,
+      "loss": 1.43888855,
+      "memory(GiB)": 113.67,
+      "step": 30855,
+      "train_speed(iter/s)": 1.319035
+    },
+    {
+      "acc": 0.69441271,
+      "epoch": 0.7828513444951801,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.137485154714945e-06,
+      "loss": 1.36476498,
+      "memory(GiB)": 113.67,
+      "step": 30860,
+      "train_speed(iter/s)": 1.319059
+    },
+    {
+      "acc": 0.69902143,
+      "epoch": 0.7829781836631152,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.1365371343407304e-06,
+      "loss": 1.37638874,
+      "memory(GiB)": 113.67,
+      "step": 30865,
+      "train_speed(iter/s)": 1.319082
+    },
+    {
+      "acc": 0.7023344,
+      "epoch": 0.7831050228310502,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.135589019992076e-06,
+      "loss": 1.39065113,
+      "memory(GiB)": 113.67,
+      "step": 30870,
+      "train_speed(iter/s)": 1.319105
+    },
+    {
+      "acc": 0.69605904,
+      "epoch": 0.7832318619989853,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.134640811710681e-06,
+      "loss": 1.3820281,
+      "memory(GiB)": 113.67,
+      "step": 30875,
+      "train_speed(iter/s)": 1.31913
+    },
+    {
+      "acc": 0.66999464,
+      "epoch": 0.7833587011669203,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.133692509538253e-06,
+      "loss": 1.38909597,
+      "memory(GiB)": 113.67,
+      "step": 30880,
+      "train_speed(iter/s)": 1.319149
+    },
+    {
+      "acc": 0.68090734,
+      "epoch": 0.7834855403348554,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.132744113516502e-06,
+      "loss": 1.46664896,
+      "memory(GiB)": 113.67,
+      "step": 30885,
+      "train_speed(iter/s)": 1.319174
+    },
+    {
+      "acc": 0.70414748,
+      "epoch": 0.7836123795027905,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.1317956236871436e-06,
+      "loss": 1.3618187,
+      "memory(GiB)": 113.67,
+      "step": 30890,
+      "train_speed(iter/s)": 1.319197
+    },
+    {
+      "acc": 0.69722238,
+      "epoch": 0.7837392186707255,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.130847040091893e-06,
+      "loss": 1.37452869,
+      "memory(GiB)": 113.67,
+      "step": 30895,
+      "train_speed(iter/s)": 1.319221
+    },
+    {
+      "acc": 0.68639212,
+      "epoch": 0.7838660578386606,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.1298983627724795e-06,
+      "loss": 1.37727509,
+      "memory(GiB)": 113.67,
+      "step": 30900,
+      "train_speed(iter/s)": 1.319246
+    },
+    {
+      "acc": 0.66606464,
+      "epoch": 0.7839928970065957,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.128949591770624e-06,
+      "loss": 1.47093716,
+      "memory(GiB)": 113.67,
+      "step": 30905,
+      "train_speed(iter/s)": 1.319271
+    },
+    {
+      "acc": 0.6946167,
+      "epoch": 0.7841197361745307,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.128000727128063e-06,
+      "loss": 1.423526,
+      "memory(GiB)": 113.67,
+      "step": 30910,
+      "train_speed(iter/s)": 1.319294
+    },
+    {
+      "acc": 0.68176885,
+      "epoch": 0.7842465753424658,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.127051768886527e-06,
+      "loss": 1.42360172,
+      "memory(GiB)": 113.67,
+      "step": 30915,
+      "train_speed(iter/s)": 1.319318
+    },
+    {
+      "acc": 0.71038132,
+      "epoch": 0.7843734145104008,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.126102717087758e-06,
+      "loss": 1.33299561,
+      "memory(GiB)": 113.67,
+      "step": 30920,
+      "train_speed(iter/s)": 1.319342
+    },
+    {
+      "acc": 0.67592077,
+      "epoch": 0.7845002536783359,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.1251535717735e-06,
+      "loss": 1.43988733,
+      "memory(GiB)": 113.67,
+      "step": 30925,
+      "train_speed(iter/s)": 1.319365
+    },
+    {
+      "acc": 0.67362642,
+      "epoch": 0.784627092846271,
+      "grad_norm": 3.890625,
+      "learning_rate": 7.1242043329854995e-06,
+      "loss": 1.47295818,
+      "memory(GiB)": 113.67,
+      "step": 30930,
+      "train_speed(iter/s)": 1.319387
+    },
+    {
+      "acc": 0.68607216,
+      "epoch": 0.784753932014206,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.123255000765508e-06,
+      "loss": 1.37948093,
+      "memory(GiB)": 113.67,
+      "step": 30935,
+      "train_speed(iter/s)": 1.31941
+    },
+    {
+      "acc": 0.69030514,
+      "epoch": 0.7848807711821411,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.122305575155283e-06,
+      "loss": 1.38064842,
+      "memory(GiB)": 113.67,
+      "step": 30940,
+      "train_speed(iter/s)": 1.319433
+    },
+    {
+      "acc": 0.68565669,
+      "epoch": 0.7850076103500762,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.121356056196582e-06,
+      "loss": 1.42814598,
+      "memory(GiB)": 113.67,
+      "step": 30945,
+      "train_speed(iter/s)": 1.319457
+    },
+    {
+      "acc": 0.7059041,
+      "epoch": 0.7851344495180111,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.1204064439311715e-06,
+      "loss": 1.36035614,
+      "memory(GiB)": 113.67,
+      "step": 30950,
+      "train_speed(iter/s)": 1.31948
+    },
+    {
+      "acc": 0.69845481,
+      "epoch": 0.7852612886859462,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.119456738400818e-06,
+      "loss": 1.3458025,
+      "memory(GiB)": 113.67,
+      "step": 30955,
+      "train_speed(iter/s)": 1.319503
+    },
+    {
+      "acc": 0.68852487,
+      "epoch": 0.7853881278538812,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.118506939647295e-06,
+      "loss": 1.29845133,
+      "memory(GiB)": 113.67,
+      "step": 30960,
+      "train_speed(iter/s)": 1.319525
+    },
+    {
+      "acc": 0.68915,
+      "epoch": 0.7855149670218163,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.1175570477123776e-06,
+      "loss": 1.41595917,
+      "memory(GiB)": 113.67,
+      "step": 30965,
+      "train_speed(iter/s)": 1.319549
+    },
+    {
+      "acc": 0.68592505,
+      "epoch": 0.7856418061897514,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.116607062637848e-06,
+      "loss": 1.4103755,
+      "memory(GiB)": 113.67,
+      "step": 30970,
+      "train_speed(iter/s)": 1.319572
+    },
+    {
+      "acc": 0.6869215,
+      "epoch": 0.7857686453576864,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.115656984465489e-06,
+      "loss": 1.35868397,
+      "memory(GiB)": 113.67,
+      "step": 30975,
+      "train_speed(iter/s)": 1.319596
+    },
+    {
+      "acc": 0.67588387,
+      "epoch": 0.7858954845256215,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.114706813237091e-06,
+      "loss": 1.4683197,
+      "memory(GiB)": 113.67,
+      "step": 30980,
+      "train_speed(iter/s)": 1.319619
+    },
+    {
+      "acc": 0.68455276,
+      "epoch": 0.7860223236935566,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.1137565489944445e-06,
+      "loss": 1.41183376,
+      "memory(GiB)": 113.67,
+      "step": 30985,
+      "train_speed(iter/s)": 1.319642
+    },
+    {
+      "acc": 0.69275293,
+      "epoch": 0.7861491628614916,
+      "grad_norm": 3.0,
+      "learning_rate": 7.112806191779349e-06,
+      "loss": 1.35336714,
+      "memory(GiB)": 113.67,
+      "step": 30990,
+      "train_speed(iter/s)": 1.319664
+    },
+    {
+      "acc": 0.68134704,
+      "epoch": 0.7862760020294267,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.111855741633603e-06,
+      "loss": 1.44349232,
+      "memory(GiB)": 113.67,
+      "step": 30995,
+      "train_speed(iter/s)": 1.319687
+    },
+    {
+      "acc": 0.69296937,
+      "epoch": 0.7864028411973617,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.1109051985990145e-06,
+      "loss": 1.41114159,
+      "memory(GiB)": 113.67,
+      "step": 31000,
+      "train_speed(iter/s)": 1.319709
+    },
+    {
+      "epoch": 0.7864028411973617,
+      "eval_acc": 0.674625618921724,
+      "eval_loss": 1.3590041399002075,
+      "eval_runtime": 69.6554,
+      "eval_samples_per_second": 91.45,
+      "eval_steps_per_second": 22.87,
+      "step": 31000
+    },
+    {
+      "acc": 0.68043461,
+      "epoch": 0.7865296803652968,
+      "grad_norm": 2.75,
+      "learning_rate": 7.109954562717389e-06,
+      "loss": 1.43070698,
+      "memory(GiB)": 113.67,
+      "step": 31005,
+      "train_speed(iter/s)": 1.315131
+    },
+    {
+      "acc": 0.69663506,
+      "epoch": 0.7866565195332319,
+      "grad_norm": 6.625,
+      "learning_rate": 7.109003834030543e-06,
+      "loss": 1.36200104,
+      "memory(GiB)": 113.67,
+      "step": 31010,
+      "train_speed(iter/s)": 1.315151
+    },
+    {
+      "acc": 0.68788519,
+      "epoch": 0.7867833587011669,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.108053012580291e-06,
+      "loss": 1.39014606,
+      "memory(GiB)": 113.67,
+      "step": 31015,
+      "train_speed(iter/s)": 1.315174
+    },
+    {
+      "acc": 0.68843412,
+      "epoch": 0.786910197869102,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.107102098408457e-06,
+      "loss": 1.37018909,
+      "memory(GiB)": 113.67,
+      "step": 31020,
+      "train_speed(iter/s)": 1.315195
+    },
+    {
+      "acc": 0.68678408,
+      "epoch": 0.7870370370370371,
+      "grad_norm": 3.125,
+      "learning_rate": 7.106151091556865e-06,
+      "loss": 1.40180931,
+      "memory(GiB)": 113.67,
+      "step": 31025,
+      "train_speed(iter/s)": 1.315219
+    },
+    {
+      "acc": 0.68849697,
+      "epoch": 0.7871638762049721,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.105199992067344e-06,
+      "loss": 1.39641075,
+      "memory(GiB)": 113.67,
+      "step": 31030,
+      "train_speed(iter/s)": 1.315244
+    },
+    {
+      "acc": 0.68479586,
+      "epoch": 0.7872907153729072,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.1042487999817275e-06,
+      "loss": 1.38162174,
+      "memory(GiB)": 113.67,
+      "step": 31035,
+      "train_speed(iter/s)": 1.315267
+    },
+    {
+      "acc": 0.69336104,
+      "epoch": 0.7874175545408422,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.103297515341857e-06,
+      "loss": 1.36813917,
+      "memory(GiB)": 113.67,
+      "step": 31040,
+      "train_speed(iter/s)": 1.315291
+    },
+    {
+      "acc": 0.68092241,
+      "epoch": 0.7875443937087773,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.1023461381895685e-06,
+      "loss": 1.39608316,
+      "memory(GiB)": 113.67,
+      "step": 31045,
+      "train_speed(iter/s)": 1.315305
+    },
+    {
+      "acc": 0.6885982,
+      "epoch": 0.7876712328767124,
+      "grad_norm": 3.0,
+      "learning_rate": 7.1013946685667125e-06,
+      "loss": 1.39728889,
+      "memory(GiB)": 113.67,
+      "step": 31050,
+      "train_speed(iter/s)": 1.31533
+    },
+    {
+      "acc": 0.69284687,
+      "epoch": 0.7877980720446474,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.100443106515135e-06,
+      "loss": 1.34026642,
+      "memory(GiB)": 113.67,
+      "step": 31055,
+      "train_speed(iter/s)": 1.315352
+    },
+    {
+      "acc": 0.69983864,
+      "epoch": 0.7879249112125825,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.099491452076693e-06,
+      "loss": 1.36314392,
+      "memory(GiB)": 113.67,
+      "step": 31060,
+      "train_speed(iter/s)": 1.315376
+    },
+    {
+      "acc": 0.68882456,
+      "epoch": 0.7880517503805176,
+      "grad_norm": 4.875,
+      "learning_rate": 7.098539705293242e-06,
+      "loss": 1.39972553,
+      "memory(GiB)": 113.67,
+      "step": 31065,
+      "train_speed(iter/s)": 1.3154
+    },
+    {
+      "acc": 0.67943296,
+      "epoch": 0.7881785895484525,
+      "grad_norm": 3.75,
+      "learning_rate": 7.097587866206647e-06,
+      "loss": 1.39844761,
+      "memory(GiB)": 113.67,
+      "step": 31070,
+      "train_speed(iter/s)": 1.315423
+    },
+    {
+      "acc": 0.70033827,
+      "epoch": 0.7883054287163876,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.096635934858772e-06,
+      "loss": 1.33206682,
+      "memory(GiB)": 113.67,
+      "step": 31075,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.68641849,
+      "epoch": 0.7884322678843226,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.095683911291488e-06,
+      "loss": 1.40010662,
+      "memory(GiB)": 113.67,
+      "step": 31080,
+      "train_speed(iter/s)": 1.315472
+    },
+    {
+      "acc": 0.68221703,
+      "epoch": 0.7885591070522577,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.0947317955466686e-06,
+      "loss": 1.48033648,
+      "memory(GiB)": 113.67,
+      "step": 31085,
+      "train_speed(iter/s)": 1.315497
+    },
+    {
+      "acc": 0.68558092,
+      "epoch": 0.7886859462201928,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.093779587666193e-06,
+      "loss": 1.40264416,
+      "memory(GiB)": 113.67,
+      "step": 31090,
+      "train_speed(iter/s)": 1.315521
+    },
+    {
+      "acc": 0.69614735,
+      "epoch": 0.7888127853881278,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.092827287691943e-06,
+      "loss": 1.33572149,
+      "memory(GiB)": 113.67,
+      "step": 31095,
+      "train_speed(iter/s)": 1.315546
+    },
+    {
+      "acc": 0.68628082,
+      "epoch": 0.7889396245560629,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.091874895665806e-06,
+      "loss": 1.37050848,
+      "memory(GiB)": 113.67,
+      "step": 31100,
+      "train_speed(iter/s)": 1.315568
+    },
+    {
+      "acc": 0.68988338,
+      "epoch": 0.789066463723998,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.09092241162967e-06,
+      "loss": 1.39156094,
+      "memory(GiB)": 113.67,
+      "step": 31105,
+      "train_speed(iter/s)": 1.315592
+    },
+    {
+      "acc": 0.69595222,
+      "epoch": 0.789193302891933,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.089969835625432e-06,
+      "loss": 1.38047009,
+      "memory(GiB)": 113.67,
+      "step": 31110,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.67335629,
+      "epoch": 0.7893201420598681,
+      "grad_norm": 3.25,
+      "learning_rate": 7.089017167694988e-06,
+      "loss": 1.43143702,
+      "memory(GiB)": 113.67,
+      "step": 31115,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.67448721,
+      "epoch": 0.7894469812278031,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.088064407880244e-06,
+      "loss": 1.46394205,
+      "memory(GiB)": 113.67,
+      "step": 31120,
+      "train_speed(iter/s)": 1.315666
+    },
+    {
+      "acc": 0.68645153,
+      "epoch": 0.7895738203957382,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.087111556223103e-06,
+      "loss": 1.38489885,
+      "memory(GiB)": 113.67,
+      "step": 31125,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.68405957,
+      "epoch": 0.7897006595636733,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.08615861276548e-06,
+      "loss": 1.42756271,
+      "memory(GiB)": 113.67,
+      "step": 31130,
+      "train_speed(iter/s)": 1.315714
+    },
+    {
+      "acc": 0.67644796,
+      "epoch": 0.7898274987316083,
+      "grad_norm": 3.125,
+      "learning_rate": 7.085205577549285e-06,
+      "loss": 1.39948215,
+      "memory(GiB)": 113.67,
+      "step": 31135,
+      "train_speed(iter/s)": 1.315739
+    },
+    {
+      "acc": 0.68993301,
+      "epoch": 0.7899543378995434,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.08425245061644e-06,
+      "loss": 1.36151514,
+      "memory(GiB)": 113.67,
+      "step": 31140,
+      "train_speed(iter/s)": 1.315761
+    },
+    {
+      "acc": 0.68212371,
+      "epoch": 0.7900811770674785,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.083299232008867e-06,
+      "loss": 1.43688345,
+      "memory(GiB)": 113.67,
+      "step": 31145,
+      "train_speed(iter/s)": 1.315785
+    },
+    {
+      "acc": 0.68274679,
+      "epoch": 0.7902080162354135,
+      "grad_norm": 3.25,
+      "learning_rate": 7.082345921768492e-06,
+      "loss": 1.36180096,
+      "memory(GiB)": 113.67,
+      "step": 31150,
+      "train_speed(iter/s)": 1.315809
+    },
+    {
+      "acc": 0.68627748,
+      "epoch": 0.7903348554033486,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.0813925199372455e-06,
+      "loss": 1.40747032,
+      "memory(GiB)": 113.67,
+      "step": 31155,
+      "train_speed(iter/s)": 1.315833
+    },
+    {
+      "acc": 0.66936951,
+      "epoch": 0.7904616945712836,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.080439026557065e-06,
+      "loss": 1.40041885,
+      "memory(GiB)": 113.67,
+      "step": 31160,
+      "train_speed(iter/s)": 1.315856
+    },
+    {
+      "acc": 0.70256271,
+      "epoch": 0.7905885337392187,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.079485441669887e-06,
+      "loss": 1.29548349,
+      "memory(GiB)": 113.67,
+      "step": 31165,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.66721573,
+      "epoch": 0.7907153729071538,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.0785317653176534e-06,
+      "loss": 1.4194437,
+      "memory(GiB)": 113.67,
+      "step": 31170,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.66416807,
+      "epoch": 0.7908422120750888,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.077577997542316e-06,
+      "loss": 1.44231853,
+      "memory(GiB)": 113.67,
+      "step": 31175,
+      "train_speed(iter/s)": 1.315928
+    },
+    {
+      "acc": 0.67078185,
+      "epoch": 0.7909690512430239,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.0766241383858195e-06,
+      "loss": 1.51792908,
+      "memory(GiB)": 113.67,
+      "step": 31180,
+      "train_speed(iter/s)": 1.315951
+    },
+    {
+      "acc": 0.68665352,
+      "epoch": 0.791095890410959,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.075670187890123e-06,
+      "loss": 1.39579182,
+      "memory(GiB)": 113.67,
+      "step": 31185,
+      "train_speed(iter/s)": 1.315975
+    },
+    {
+      "acc": 0.68629818,
+      "epoch": 0.791222729578894,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.0747161460971845e-06,
+      "loss": 1.4015729,
+      "memory(GiB)": 113.67,
+      "step": 31190,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.67814188,
+      "epoch": 0.791349568746829,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.073762013048966e-06,
+      "loss": 1.47956696,
+      "memory(GiB)": 113.67,
+      "step": 31195,
+      "train_speed(iter/s)": 1.316024
+    },
+    {
+      "acc": 0.68246517,
+      "epoch": 0.791476407914764,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.072807788787437e-06,
+      "loss": 1.38946056,
+      "memory(GiB)": 113.67,
+      "step": 31200,
+      "train_speed(iter/s)": 1.316048
+    },
+    {
+      "acc": 0.6896215,
+      "epoch": 0.7916032470826991,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.071853473354566e-06,
+      "loss": 1.37667198,
+      "memory(GiB)": 113.67,
+      "step": 31205,
+      "train_speed(iter/s)": 1.316071
+    },
+    {
+      "acc": 0.68802905,
+      "epoch": 0.7917300862506342,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.070899066792329e-06,
+      "loss": 1.3699791,
+      "memory(GiB)": 113.67,
+      "step": 31210,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.69474459,
+      "epoch": 0.7918569254185692,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.069944569142706e-06,
+      "loss": 1.3358799,
+      "memory(GiB)": 113.67,
+      "step": 31215,
+      "train_speed(iter/s)": 1.316119
+    },
+    {
+      "acc": 0.67850389,
+      "epoch": 0.7919837645865043,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.068989980447679e-06,
+      "loss": 1.45598354,
+      "memory(GiB)": 113.67,
+      "step": 31220,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.69088774,
+      "epoch": 0.7921106037544394,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.068035300749237e-06,
+      "loss": 1.3857728,
+      "memory(GiB)": 113.67,
+      "step": 31225,
+      "train_speed(iter/s)": 1.316166
+    },
+    {
+      "acc": 0.6832233,
+      "epoch": 0.7922374429223744,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.067080530089366e-06,
+      "loss": 1.45482225,
+      "memory(GiB)": 113.67,
+      "step": 31230,
+      "train_speed(iter/s)": 1.316187
+    },
+    {
+      "acc": 0.67725897,
+      "epoch": 0.7923642820903095,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.066125668510067e-06,
+      "loss": 1.44137363,
+      "memory(GiB)": 113.67,
+      "step": 31235,
+      "train_speed(iter/s)": 1.316211
+    },
+    {
+      "acc": 0.67773423,
+      "epoch": 0.7924911212582445,
+      "grad_norm": 3.125,
+      "learning_rate": 7.065170716053336e-06,
+      "loss": 1.42514486,
+      "memory(GiB)": 113.67,
+      "step": 31240,
+      "train_speed(iter/s)": 1.316234
+    },
+    {
+      "acc": 0.67930555,
+      "epoch": 0.7926179604261796,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.064215672761175e-06,
+      "loss": 1.43729229,
+      "memory(GiB)": 113.67,
+      "step": 31245,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.69692121,
+      "epoch": 0.7927447995941147,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.063260538675594e-06,
+      "loss": 1.3527544,
+      "memory(GiB)": 113.67,
+      "step": 31250,
+      "train_speed(iter/s)": 1.316281
+    },
+    {
+      "acc": 0.68835144,
+      "epoch": 0.7928716387620497,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.062305313838601e-06,
+      "loss": 1.38249664,
+      "memory(GiB)": 113.67,
+      "step": 31255,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.67915316,
+      "epoch": 0.7929984779299848,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.061349998292215e-06,
+      "loss": 1.47307539,
+      "memory(GiB)": 113.67,
+      "step": 31260,
+      "train_speed(iter/s)": 1.316327
+    },
+    {
+      "acc": 0.6842782,
+      "epoch": 0.7931253170979199,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.060394592078452e-06,
+      "loss": 1.41328354,
+      "memory(GiB)": 113.67,
+      "step": 31265,
+      "train_speed(iter/s)": 1.31635
+    },
+    {
+      "acc": 0.68847218,
+      "epoch": 0.7932521562658549,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.0594390952393365e-06,
+      "loss": 1.40513,
+      "memory(GiB)": 113.67,
+      "step": 31270,
+      "train_speed(iter/s)": 1.316374
+    },
+    {
+      "acc": 0.68257389,
+      "epoch": 0.79337899543379,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.058483507816894e-06,
+      "loss": 1.41561604,
+      "memory(GiB)": 113.67,
+      "step": 31275,
+      "train_speed(iter/s)": 1.316395
+    },
+    {
+      "acc": 0.69650884,
+      "epoch": 0.793505834601725,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.057527829853157e-06,
+      "loss": 1.33401489,
+      "memory(GiB)": 113.67,
+      "step": 31280,
+      "train_speed(iter/s)": 1.316419
+    },
+    {
+      "acc": 0.68003359,
+      "epoch": 0.7936326737696601,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.056572061390159e-06,
+      "loss": 1.42129307,
+      "memory(GiB)": 113.67,
+      "step": 31285,
+      "train_speed(iter/s)": 1.316444
+    },
+    {
+      "acc": 0.67645741,
+      "epoch": 0.7937595129375952,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.055616202469939e-06,
+      "loss": 1.44355907,
+      "memory(GiB)": 113.67,
+      "step": 31290,
+      "train_speed(iter/s)": 1.316468
+    },
+    {
+      "acc": 0.68418431,
+      "epoch": 0.7938863521055302,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.054660253134543e-06,
+      "loss": 1.40591908,
+      "memory(GiB)": 113.67,
+      "step": 31295,
+      "train_speed(iter/s)": 1.316492
+    },
+    {
+      "acc": 0.68629751,
+      "epoch": 0.7940131912734653,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.053704213426015e-06,
+      "loss": 1.4776413,
+      "memory(GiB)": 113.67,
+      "step": 31300,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.67779889,
+      "epoch": 0.7941400304414004,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.052748083386406e-06,
+      "loss": 1.43919554,
+      "memory(GiB)": 113.67,
+      "step": 31305,
+      "train_speed(iter/s)": 1.316539
+    },
+    {
+      "acc": 0.69884691,
+      "epoch": 0.7942668696093353,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.051791863057772e-06,
+      "loss": 1.38471031,
+      "memory(GiB)": 113.67,
+      "step": 31310,
+      "train_speed(iter/s)": 1.316564
+    },
+    {
+      "acc": 0.67984819,
+      "epoch": 0.7943937087772704,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.050835552482171e-06,
+      "loss": 1.40158176,
+      "memory(GiB)": 113.67,
+      "step": 31315,
+      "train_speed(iter/s)": 1.316588
+    },
+    {
+      "acc": 0.68519115,
+      "epoch": 0.7945205479452054,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.049879151701666e-06,
+      "loss": 1.38857002,
+      "memory(GiB)": 113.67,
+      "step": 31320,
+      "train_speed(iter/s)": 1.316614
+    },
+    {
+      "acc": 0.67739134,
+      "epoch": 0.7946473871131405,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.048922660758324e-06,
+      "loss": 1.36883488,
+      "memory(GiB)": 113.67,
+      "step": 31325,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.69356184,
+      "epoch": 0.7947742262810756,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.047966079694215e-06,
+      "loss": 1.3482275,
+      "memory(GiB)": 113.67,
+      "step": 31330,
+      "train_speed(iter/s)": 1.316663
+    },
+    {
+      "acc": 0.69120855,
+      "epoch": 0.7949010654490106,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.047009408551414e-06,
+      "loss": 1.35504608,
+      "memory(GiB)": 113.67,
+      "step": 31335,
+      "train_speed(iter/s)": 1.316688
+    },
+    {
+      "acc": 0.68021908,
+      "epoch": 0.7950279046169457,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.046052647372002e-06,
+      "loss": 1.41444101,
+      "memory(GiB)": 113.67,
+      "step": 31340,
+      "train_speed(iter/s)": 1.316713
+    },
+    {
+      "acc": 0.67622743,
+      "epoch": 0.7951547437848808,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.045095796198057e-06,
+      "loss": 1.41680622,
+      "memory(GiB)": 113.67,
+      "step": 31345,
+      "train_speed(iter/s)": 1.316737
+    },
+    {
+      "acc": 0.67480106,
+      "epoch": 0.7952815829528158,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.044138855071671e-06,
+      "loss": 1.53810921,
+      "memory(GiB)": 113.67,
+      "step": 31350,
+      "train_speed(iter/s)": 1.316762
+    },
+    {
+      "acc": 0.68588848,
+      "epoch": 0.7954084221207509,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.043181824034929e-06,
+      "loss": 1.39064713,
+      "memory(GiB)": 113.67,
+      "step": 31355,
+      "train_speed(iter/s)": 1.316786
+    },
+    {
+      "acc": 0.69820108,
+      "epoch": 0.7955352612886859,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.042224703129929e-06,
+      "loss": 1.35923929,
+      "memory(GiB)": 113.67,
+      "step": 31360,
+      "train_speed(iter/s)": 1.316811
+    },
+    {
+      "acc": 0.68651319,
+      "epoch": 0.795662100456621,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.0412674923987705e-06,
+      "loss": 1.45610847,
+      "memory(GiB)": 113.67,
+      "step": 31365,
+      "train_speed(iter/s)": 1.316836
+    },
+    {
+      "acc": 0.70334105,
+      "epoch": 0.7957889396245561,
+      "grad_norm": 3.9375,
+      "learning_rate": 7.040310191883552e-06,
+      "loss": 1.32554054,
+      "memory(GiB)": 113.67,
+      "step": 31370,
+      "train_speed(iter/s)": 1.31686
+    },
+    {
+      "acc": 0.6877368,
+      "epoch": 0.7959157787924911,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.039352801626383e-06,
+      "loss": 1.39986553,
+      "memory(GiB)": 113.67,
+      "step": 31375,
+      "train_speed(iter/s)": 1.316885
+    },
+    {
+      "acc": 0.70628796,
+      "epoch": 0.7960426179604262,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.0383953216693725e-06,
+      "loss": 1.25752935,
+      "memory(GiB)": 113.67,
+      "step": 31380,
+      "train_speed(iter/s)": 1.31691
+    },
+    {
+      "acc": 0.6951519,
+      "epoch": 0.7961694571283613,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.037437752054635e-06,
+      "loss": 1.39670448,
+      "memory(GiB)": 113.67,
+      "step": 31385,
+      "train_speed(iter/s)": 1.316934
+    },
+    {
+      "acc": 0.6999959,
+      "epoch": 0.7962962962962963,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.036480092824288e-06,
+      "loss": 1.35882988,
+      "memory(GiB)": 113.67,
+      "step": 31390,
+      "train_speed(iter/s)": 1.316958
+    },
+    {
+      "acc": 0.68392448,
+      "epoch": 0.7964231354642314,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.035522344020455e-06,
+      "loss": 1.3990325,
+      "memory(GiB)": 113.67,
+      "step": 31395,
+      "train_speed(iter/s)": 1.316983
+    },
+    {
+      "acc": 0.67945743,
+      "epoch": 0.7965499746321664,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.034564505685262e-06,
+      "loss": 1.36796408,
+      "memory(GiB)": 113.67,
+      "step": 31400,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.67988405,
+      "epoch": 0.7966768138001015,
+      "grad_norm": 3.375,
+      "learning_rate": 7.0336065778608365e-06,
+      "loss": 1.40452433,
+      "memory(GiB)": 113.67,
+      "step": 31405,
+      "train_speed(iter/s)": 1.317031
+    },
+    {
+      "acc": 0.7012064,
+      "epoch": 0.7968036529680366,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.032648560589316e-06,
+      "loss": 1.33380003,
+      "memory(GiB)": 113.67,
+      "step": 31410,
+      "train_speed(iter/s)": 1.317056
+    },
+    {
+      "acc": 0.68973026,
+      "epoch": 0.7969304921359716,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.031690453912835e-06,
+      "loss": 1.32869463,
+      "memory(GiB)": 113.67,
+      "step": 31415,
+      "train_speed(iter/s)": 1.31708
+    },
+    {
+      "acc": 0.68653016,
+      "epoch": 0.7970573313039067,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.030732257873539e-06,
+      "loss": 1.39665298,
+      "memory(GiB)": 113.67,
+      "step": 31420,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.69009972,
+      "epoch": 0.7971841704718418,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.02977397251357e-06,
+      "loss": 1.39229889,
+      "memory(GiB)": 113.67,
+      "step": 31425,
+      "train_speed(iter/s)": 1.317129
+    },
+    {
+      "acc": 0.69692526,
+      "epoch": 0.7973110096397767,
+      "grad_norm": 2.625,
+      "learning_rate": 7.028815597875081e-06,
+      "loss": 1.32438202,
+      "memory(GiB)": 113.67,
+      "step": 31430,
+      "train_speed(iter/s)": 1.317152
+    },
+    {
+      "acc": 0.67096057,
+      "epoch": 0.7974378488077118,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.027857134000223e-06,
+      "loss": 1.46685829,
+      "memory(GiB)": 113.67,
+      "step": 31435,
+      "train_speed(iter/s)": 1.317177
+    },
+    {
+      "acc": 0.70395651,
+      "epoch": 0.7975646879756468,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.026898580931154e-06,
+      "loss": 1.40606527,
+      "memory(GiB)": 113.67,
+      "step": 31440,
+      "train_speed(iter/s)": 1.317202
+    },
+    {
+      "acc": 0.69924121,
+      "epoch": 0.7976915271435819,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.025939938710037e-06,
+      "loss": 1.3507946,
+      "memory(GiB)": 113.67,
+      "step": 31445,
+      "train_speed(iter/s)": 1.317226
+    },
+    {
+      "acc": 0.66562362,
+      "epoch": 0.797818366311517,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.024981207379036e-06,
+      "loss": 1.44765491,
+      "memory(GiB)": 113.67,
+      "step": 31450,
+      "train_speed(iter/s)": 1.317249
+    },
+    {
+      "acc": 0.66879683,
+      "epoch": 0.797945205479452,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.02402238698032e-06,
+      "loss": 1.45992241,
+      "memory(GiB)": 113.67,
+      "step": 31455,
+      "train_speed(iter/s)": 1.317273
+    },
+    {
+      "acc": 0.6758296,
+      "epoch": 0.7980720446473871,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.023063477556064e-06,
+      "loss": 1.44518919,
+      "memory(GiB)": 113.67,
+      "step": 31460,
+      "train_speed(iter/s)": 1.317297
+    },
+    {
+      "acc": 0.69886823,
+      "epoch": 0.7981988838153222,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.0221044791484424e-06,
+      "loss": 1.33759594,
+      "memory(GiB)": 113.67,
+      "step": 31465,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.68123159,
+      "epoch": 0.7983257229832572,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.021145391799639e-06,
+      "loss": 1.37565346,
+      "memory(GiB)": 113.67,
+      "step": 31470,
+      "train_speed(iter/s)": 1.317347
+    },
+    {
+      "acc": 0.68167458,
+      "epoch": 0.7984525621511923,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.020186215551837e-06,
+      "loss": 1.39474354,
+      "memory(GiB)": 113.67,
+      "step": 31475,
+      "train_speed(iter/s)": 1.317371
+    },
+    {
+      "acc": 0.69055367,
+      "epoch": 0.7985794013191273,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.019226950447227e-06,
+      "loss": 1.38507433,
+      "memory(GiB)": 113.67,
+      "step": 31480,
+      "train_speed(iter/s)": 1.317395
+    },
+    {
+      "acc": 0.67843933,
+      "epoch": 0.7987062404870624,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.018267596527998e-06,
+      "loss": 1.45586948,
+      "memory(GiB)": 113.67,
+      "step": 31485,
+      "train_speed(iter/s)": 1.317419
+    },
+    {
+      "acc": 0.70775223,
+      "epoch": 0.7988330796549975,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.017308153836352e-06,
+      "loss": 1.34195662,
+      "memory(GiB)": 113.67,
+      "step": 31490,
+      "train_speed(iter/s)": 1.317444
+    },
+    {
+      "acc": 0.68432112,
+      "epoch": 0.7989599188229325,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.016348622414484e-06,
+      "loss": 1.42133398,
+      "memory(GiB)": 113.67,
+      "step": 31495,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.69068117,
+      "epoch": 0.7990867579908676,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.015389002304604e-06,
+      "loss": 1.42363462,
+      "memory(GiB)": 113.67,
+      "step": 31500,
+      "train_speed(iter/s)": 1.317489
+    },
+    {
+      "acc": 0.7053813,
+      "epoch": 0.7992135971588027,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.014429293548916e-06,
+      "loss": 1.30410681,
+      "memory(GiB)": 113.67,
+      "step": 31505,
+      "train_speed(iter/s)": 1.317514
+    },
+    {
+      "acc": 0.6900681,
+      "epoch": 0.7993404363267377,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.013469496189633e-06,
+      "loss": 1.40586319,
+      "memory(GiB)": 113.67,
+      "step": 31510,
+      "train_speed(iter/s)": 1.317537
+    },
+    {
+      "acc": 0.69549456,
+      "epoch": 0.7994672754946728,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.012509610268974e-06,
+      "loss": 1.29541864,
+      "memory(GiB)": 113.67,
+      "step": 31515,
+      "train_speed(iter/s)": 1.317561
+    },
+    {
+      "acc": 0.68271265,
+      "epoch": 0.7995941146626078,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.011549635829156e-06,
+      "loss": 1.40477161,
+      "memory(GiB)": 113.67,
+      "step": 31520,
+      "train_speed(iter/s)": 1.317585
+    },
+    {
+      "acc": 0.67474675,
+      "epoch": 0.7997209538305429,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.010589572912404e-06,
+      "loss": 1.38849049,
+      "memory(GiB)": 113.67,
+      "step": 31525,
+      "train_speed(iter/s)": 1.317608
+    },
+    {
+      "acc": 0.67798138,
+      "epoch": 0.799847792998478,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.009629421560946e-06,
+      "loss": 1.41080551,
+      "memory(GiB)": 113.67,
+      "step": 31530,
+      "train_speed(iter/s)": 1.317632
+    },
+    {
+      "acc": 0.68279839,
+      "epoch": 0.799974632166413,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.008669181817015e-06,
+      "loss": 1.41313877,
+      "memory(GiB)": 113.67,
+      "step": 31535,
+      "train_speed(iter/s)": 1.317655
+    },
+    {
+      "acc": 0.68139696,
+      "epoch": 0.8001014713343481,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.007708853722844e-06,
+      "loss": 1.38134203,
+      "memory(GiB)": 113.67,
+      "step": 31540,
+      "train_speed(iter/s)": 1.317678
+    },
+    {
+      "acc": 0.67122545,
+      "epoch": 0.8002283105022832,
+      "grad_norm": 3.125,
+      "learning_rate": 7.006748437320674e-06,
+      "loss": 1.52758026,
+      "memory(GiB)": 113.67,
+      "step": 31545,
+      "train_speed(iter/s)": 1.317701
+    },
+    {
+      "acc": 0.67942605,
+      "epoch": 0.8003551496702181,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.005787932652749e-06,
+      "loss": 1.39390821,
+      "memory(GiB)": 113.67,
+      "step": 31550,
+      "train_speed(iter/s)": 1.317724
+    },
+    {
+      "acc": 0.67504721,
+      "epoch": 0.8004819888381532,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.0048273397613145e-06,
+      "loss": 1.47945251,
+      "memory(GiB)": 113.67,
+      "step": 31555,
+      "train_speed(iter/s)": 1.317747
+    },
+    {
+      "acc": 0.68617287,
+      "epoch": 0.8006088280060882,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.003866658688624e-06,
+      "loss": 1.37815571,
+      "memory(GiB)": 113.67,
+      "step": 31560,
+      "train_speed(iter/s)": 1.317771
+    },
+    {
+      "acc": 0.68433743,
+      "epoch": 0.8007356671740233,
+      "grad_norm": 3.25,
+      "learning_rate": 7.0029058894769295e-06,
+      "loss": 1.44606991,
+      "memory(GiB)": 113.67,
+      "step": 31565,
+      "train_speed(iter/s)": 1.317795
+    },
+    {
+      "acc": 0.67949018,
+      "epoch": 0.8008625063419584,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.001945032168493e-06,
+      "loss": 1.40810566,
+      "memory(GiB)": 113.67,
+      "step": 31570,
+      "train_speed(iter/s)": 1.317819
+    },
+    {
+      "acc": 0.67639728,
+      "epoch": 0.8009893455098934,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.000984086805575e-06,
+      "loss": 1.37939205,
+      "memory(GiB)": 113.67,
+      "step": 31575,
+      "train_speed(iter/s)": 1.317843
+    },
+    {
+      "acc": 0.67858419,
+      "epoch": 0.8011161846778285,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.000023053430444e-06,
+      "loss": 1.40057163,
+      "memory(GiB)": 113.67,
+      "step": 31580,
+      "train_speed(iter/s)": 1.317864
+    },
+    {
+      "acc": 0.67361059,
+      "epoch": 0.8012430238457636,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.999061932085369e-06,
+      "loss": 1.45837107,
+      "memory(GiB)": 113.67,
+      "step": 31585,
+      "train_speed(iter/s)": 1.317888
+    },
+    {
+      "acc": 0.6865581,
+      "epoch": 0.8013698630136986,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.9981007228126255e-06,
+      "loss": 1.34369202,
+      "memory(GiB)": 113.67,
+      "step": 31590,
+      "train_speed(iter/s)": 1.317911
+    },
+    {
+      "acc": 0.67382574,
+      "epoch": 0.8014967021816337,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.997139425654491e-06,
+      "loss": 1.44036446,
+      "memory(GiB)": 113.67,
+      "step": 31595,
+      "train_speed(iter/s)": 1.317936
+    },
+    {
+      "acc": 0.7061058,
+      "epoch": 0.8016235413495687,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.996178040653248e-06,
+      "loss": 1.31044388,
+      "memory(GiB)": 113.67,
+      "step": 31600,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.69793448,
+      "epoch": 0.8017503805175038,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.995216567851183e-06,
+      "loss": 1.3537921,
+      "memory(GiB)": 113.67,
+      "step": 31605,
+      "train_speed(iter/s)": 1.317981
+    },
+    {
+      "acc": 0.67073288,
+      "epoch": 0.8018772196854389,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.994255007290585e-06,
+      "loss": 1.42089462,
+      "memory(GiB)": 113.67,
+      "step": 31610,
+      "train_speed(iter/s)": 1.318005
+    },
+    {
+      "acc": 0.69288502,
+      "epoch": 0.8020040588533739,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.993293359013747e-06,
+      "loss": 1.39231739,
+      "memory(GiB)": 113.67,
+      "step": 31615,
+      "train_speed(iter/s)": 1.318029
+    },
+    {
+      "acc": 0.66377878,
+      "epoch": 0.802130898021309,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.992331623062969e-06,
+      "loss": 1.44713898,
+      "memory(GiB)": 113.67,
+      "step": 31620,
+      "train_speed(iter/s)": 1.318052
+    },
+    {
+      "acc": 0.68754845,
+      "epoch": 0.8022577371892441,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.9913697994805505e-06,
+      "loss": 1.39142265,
+      "memory(GiB)": 113.67,
+      "step": 31625,
+      "train_speed(iter/s)": 1.318076
+    },
+    {
+      "acc": 0.68282881,
+      "epoch": 0.8023845763571791,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.990407888308799e-06,
+      "loss": 1.38093367,
+      "memory(GiB)": 113.67,
+      "step": 31630,
+      "train_speed(iter/s)": 1.3181
+    },
+    {
+      "acc": 0.68224669,
+      "epoch": 0.8025114155251142,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.98944588959002e-06,
+      "loss": 1.41543503,
+      "memory(GiB)": 113.67,
+      "step": 31635,
+      "train_speed(iter/s)": 1.318124
+    },
+    {
+      "acc": 0.68516865,
+      "epoch": 0.8026382546930492,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.9884838033665305e-06,
+      "loss": 1.36692543,
+      "memory(GiB)": 113.67,
+      "step": 31640,
+      "train_speed(iter/s)": 1.318148
+    },
+    {
+      "acc": 0.67221913,
+      "epoch": 0.8027650938609843,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.987521629680643e-06,
+      "loss": 1.44452381,
+      "memory(GiB)": 113.67,
+      "step": 31645,
+      "train_speed(iter/s)": 1.318172
+    },
+    {
+      "acc": 0.67647915,
+      "epoch": 0.8028919330289194,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.9865593685746815e-06,
+      "loss": 1.45677586,
+      "memory(GiB)": 113.67,
+      "step": 31650,
+      "train_speed(iter/s)": 1.318193
+    },
+    {
+      "acc": 0.68188486,
+      "epoch": 0.8030187721968544,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.98559702009097e-06,
+      "loss": 1.48915043,
+      "memory(GiB)": 113.67,
+      "step": 31655,
+      "train_speed(iter/s)": 1.318217
+    },
+    {
+      "acc": 0.68909736,
+      "epoch": 0.8031456113647895,
+      "grad_norm": 4.375,
+      "learning_rate": 6.984634584271836e-06,
+      "loss": 1.42841597,
+      "memory(GiB)": 113.67,
+      "step": 31660,
+      "train_speed(iter/s)": 1.318242
+    },
+    {
+      "acc": 0.68497615,
+      "epoch": 0.8032724505327246,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.983672061159612e-06,
+      "loss": 1.3774971,
+      "memory(GiB)": 113.67,
+      "step": 31665,
+      "train_speed(iter/s)": 1.318266
+    },
+    {
+      "acc": 0.69119043,
+      "epoch": 0.8033992897006595,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.982709450796636e-06,
+      "loss": 1.38055944,
+      "memory(GiB)": 113.67,
+      "step": 31670,
+      "train_speed(iter/s)": 1.31829
+    },
+    {
+      "acc": 0.67045698,
+      "epoch": 0.8035261288685946,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.981746753225245e-06,
+      "loss": 1.49323359,
+      "memory(GiB)": 113.67,
+      "step": 31675,
+      "train_speed(iter/s)": 1.318314
+    },
+    {
+      "acc": 0.69283381,
+      "epoch": 0.8036529680365296,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.980783968487783e-06,
+      "loss": 1.32869596,
+      "memory(GiB)": 113.67,
+      "step": 31680,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "acc": 0.67203202,
+      "epoch": 0.8037798072044647,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.9798210966266e-06,
+      "loss": 1.46473999,
+      "memory(GiB)": 113.67,
+      "step": 31685,
+      "train_speed(iter/s)": 1.318361
+    },
+    {
+      "acc": 0.67636328,
+      "epoch": 0.8039066463723998,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.9788581376840455e-06,
+      "loss": 1.37242317,
+      "memory(GiB)": 113.67,
+      "step": 31690,
+      "train_speed(iter/s)": 1.318384
+    },
+    {
+      "acc": 0.68691244,
+      "epoch": 0.8040334855403348,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.977895091702474e-06,
+      "loss": 1.38027763,
+      "memory(GiB)": 113.67,
+      "step": 31695,
+      "train_speed(iter/s)": 1.318409
+    },
+    {
+      "acc": 0.7003675,
+      "epoch": 0.8041603247082699,
+      "grad_norm": 3.0,
+      "learning_rate": 6.976931958724248e-06,
+      "loss": 1.32249241,
+      "memory(GiB)": 113.67,
+      "step": 31700,
+      "train_speed(iter/s)": 1.318433
+    },
+    {
+      "acc": 0.68388367,
+      "epoch": 0.804287163876205,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.975968738791726e-06,
+      "loss": 1.42565269,
+      "memory(GiB)": 113.67,
+      "step": 31705,
+      "train_speed(iter/s)": 1.318457
+    },
+    {
+      "acc": 0.67706594,
+      "epoch": 0.80441400304414,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.9750054319472785e-06,
+      "loss": 1.47504749,
+      "memory(GiB)": 113.67,
+      "step": 31710,
+      "train_speed(iter/s)": 1.318481
+    },
+    {
+      "acc": 0.69562883,
+      "epoch": 0.8045408422120751,
+      "grad_norm": 3.0,
+      "learning_rate": 6.974042038233272e-06,
+      "loss": 1.39222584,
+      "memory(GiB)": 113.67,
+      "step": 31715,
+      "train_speed(iter/s)": 1.318505
+    },
+    {
+      "acc": 0.66647897,
+      "epoch": 0.8046676813800101,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.9730785576920855e-06,
+      "loss": 1.52797909,
+      "memory(GiB)": 113.67,
+      "step": 31720,
+      "train_speed(iter/s)": 1.31853
+    },
+    {
+      "acc": 0.69883261,
+      "epoch": 0.8047945205479452,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.972114990366094e-06,
+      "loss": 1.35067844,
+      "memory(GiB)": 113.67,
+      "step": 31725,
+      "train_speed(iter/s)": 1.318552
+    },
+    {
+      "acc": 0.67348337,
+      "epoch": 0.8049213597158803,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.97115133629768e-06,
+      "loss": 1.36620035,
+      "memory(GiB)": 113.67,
+      "step": 31730,
+      "train_speed(iter/s)": 1.318577
+    },
+    {
+      "acc": 0.69924974,
+      "epoch": 0.8050481988838153,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.970187595529229e-06,
+      "loss": 1.34903107,
+      "memory(GiB)": 113.67,
+      "step": 31735,
+      "train_speed(iter/s)": 1.3186
+    },
+    {
+      "acc": 0.6836844,
+      "epoch": 0.8051750380517504,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.969223768103133e-06,
+      "loss": 1.47650366,
+      "memory(GiB)": 113.67,
+      "step": 31740,
+      "train_speed(iter/s)": 1.318623
+    },
+    {
+      "acc": 0.68356504,
+      "epoch": 0.8053018772196855,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.968259854061783e-06,
+      "loss": 1.38404636,
+      "memory(GiB)": 113.67,
+      "step": 31745,
+      "train_speed(iter/s)": 1.318646
+    },
+    {
+      "acc": 0.68609457,
+      "epoch": 0.8054287163876205,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.967295853447578e-06,
+      "loss": 1.39548874,
+      "memory(GiB)": 113.67,
+      "step": 31750,
+      "train_speed(iter/s)": 1.318669
+    },
+    {
+      "acc": 0.68970695,
+      "epoch": 0.8055555555555556,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.966331766302916e-06,
+      "loss": 1.42209015,
+      "memory(GiB)": 113.67,
+      "step": 31755,
+      "train_speed(iter/s)": 1.318693
+    },
+    {
+      "acc": 0.68731012,
+      "epoch": 0.8056823947234906,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.965367592670206e-06,
+      "loss": 1.36603279,
+      "memory(GiB)": 113.67,
+      "step": 31760,
+      "train_speed(iter/s)": 1.318717
+    },
+    {
+      "acc": 0.69162493,
+      "epoch": 0.8058092338914257,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.964403332591854e-06,
+      "loss": 1.47053318,
+      "memory(GiB)": 113.67,
+      "step": 31765,
+      "train_speed(iter/s)": 1.31874
+    },
+    {
+      "acc": 0.685111,
+      "epoch": 0.8059360730593608,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.963438986110272e-06,
+      "loss": 1.37998371,
+      "memory(GiB)": 113.67,
+      "step": 31770,
+      "train_speed(iter/s)": 1.318764
+    },
+    {
+      "acc": 0.70211973,
+      "epoch": 0.8060629122272958,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.962474553267877e-06,
+      "loss": 1.3886631,
+      "memory(GiB)": 113.67,
+      "step": 31775,
+      "train_speed(iter/s)": 1.318786
+    },
+    {
+      "acc": 0.6682085,
+      "epoch": 0.8061897513952309,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.96151003410709e-06,
+      "loss": 1.45194721,
+      "memory(GiB)": 113.67,
+      "step": 31780,
+      "train_speed(iter/s)": 1.31881
+    },
+    {
+      "acc": 0.68760929,
+      "epoch": 0.806316590563166,
+      "grad_norm": 2.625,
+      "learning_rate": 6.960545428670333e-06,
+      "loss": 1.44377689,
+      "memory(GiB)": 113.67,
+      "step": 31785,
+      "train_speed(iter/s)": 1.318833
+    },
+    {
+      "acc": 0.69167528,
+      "epoch": 0.806443429731101,
+      "grad_norm": 4.25,
+      "learning_rate": 6.959580737000038e-06,
+      "loss": 1.38598042,
+      "memory(GiB)": 113.67,
+      "step": 31790,
+      "train_speed(iter/s)": 1.318857
+    },
+    {
+      "acc": 0.69860058,
+      "epoch": 0.806570268899036,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.95861595913863e-06,
+      "loss": 1.32269859,
+      "memory(GiB)": 113.67,
+      "step": 31795,
+      "train_speed(iter/s)": 1.318881
+    },
+    {
+      "acc": 0.70396881,
+      "epoch": 0.806697108066971,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.95765109512855e-06,
+      "loss": 1.30690069,
+      "memory(GiB)": 113.67,
+      "step": 31800,
+      "train_speed(iter/s)": 1.318905
+    },
+    {
+      "acc": 0.6866652,
+      "epoch": 0.8068239472349061,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.956686145012233e-06,
+      "loss": 1.37984285,
+      "memory(GiB)": 113.67,
+      "step": 31805,
+      "train_speed(iter/s)": 1.318929
+    },
+    {
+      "acc": 0.68078427,
+      "epoch": 0.8069507864028412,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.955721108832124e-06,
+      "loss": 1.39183445,
+      "memory(GiB)": 113.67,
+      "step": 31810,
+      "train_speed(iter/s)": 1.318952
+    },
+    {
+      "acc": 0.69660463,
+      "epoch": 0.8070776255707762,
+      "grad_norm": 3.125,
+      "learning_rate": 6.9547559866306695e-06,
+      "loss": 1.3388854,
+      "memory(GiB)": 113.67,
+      "step": 31815,
+      "train_speed(iter/s)": 1.318974
+    },
+    {
+      "acc": 0.68983417,
+      "epoch": 0.8072044647387113,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.953790778450318e-06,
+      "loss": 1.40320644,
+      "memory(GiB)": 113.67,
+      "step": 31820,
+      "train_speed(iter/s)": 1.318998
+    },
+    {
+      "acc": 0.68564234,
+      "epoch": 0.8073313039066464,
+      "grad_norm": 3.5,
+      "learning_rate": 6.9528254843335254e-06,
+      "loss": 1.40221386,
+      "memory(GiB)": 113.67,
+      "step": 31825,
+      "train_speed(iter/s)": 1.319021
+    },
+    {
+      "acc": 0.69723053,
+      "epoch": 0.8074581430745814,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.95186010432275e-06,
+      "loss": 1.32316504,
+      "memory(GiB)": 113.67,
+      "step": 31830,
+      "train_speed(iter/s)": 1.319044
+    },
+    {
+      "acc": 0.68465161,
+      "epoch": 0.8075849822425165,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.950894638460452e-06,
+      "loss": 1.41753998,
+      "memory(GiB)": 113.67,
+      "step": 31835,
+      "train_speed(iter/s)": 1.319067
+    },
+    {
+      "acc": 0.68471336,
+      "epoch": 0.8077118214104515,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.949929086789098e-06,
+      "loss": 1.41399412,
+      "memory(GiB)": 113.67,
+      "step": 31840,
+      "train_speed(iter/s)": 1.31909
+    },
+    {
+      "acc": 0.67535477,
+      "epoch": 0.8078386605783866,
+      "grad_norm": 3.0,
+      "learning_rate": 6.948963449351156e-06,
+      "loss": 1.46155262,
+      "memory(GiB)": 113.67,
+      "step": 31845,
+      "train_speed(iter/s)": 1.319113
+    },
+    {
+      "acc": 0.67668452,
+      "epoch": 0.8079654997463217,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.947997726189102e-06,
+      "loss": 1.45598831,
+      "memory(GiB)": 113.67,
+      "step": 31850,
+      "train_speed(iter/s)": 1.319136
+    },
+    {
+      "acc": 0.68678761,
+      "epoch": 0.8080923389142567,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.947031917345409e-06,
+      "loss": 1.43695869,
+      "memory(GiB)": 113.67,
+      "step": 31855,
+      "train_speed(iter/s)": 1.319159
+    },
+    {
+      "acc": 0.67943745,
+      "epoch": 0.8082191780821918,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.946066022862561e-06,
+      "loss": 1.44983788,
+      "memory(GiB)": 113.67,
+      "step": 31860,
+      "train_speed(iter/s)": 1.319182
+    },
+    {
+      "acc": 0.67899699,
+      "epoch": 0.8083460172501269,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.945100042783039e-06,
+      "loss": 1.43881454,
+      "memory(GiB)": 113.67,
+      "step": 31865,
+      "train_speed(iter/s)": 1.319201
+    },
+    {
+      "acc": 0.67849874,
+      "epoch": 0.8084728564180619,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.9441339771493345e-06,
+      "loss": 1.36548185,
+      "memory(GiB)": 113.67,
+      "step": 31870,
+      "train_speed(iter/s)": 1.319225
+    },
+    {
+      "acc": 0.69290681,
+      "epoch": 0.808599695585997,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.943167826003937e-06,
+      "loss": 1.32205105,
+      "memory(GiB)": 113.67,
+      "step": 31875,
+      "train_speed(iter/s)": 1.319248
+    },
+    {
+      "acc": 0.69545269,
+      "epoch": 0.808726534753932,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.942201589389344e-06,
+      "loss": 1.3832202,
+      "memory(GiB)": 113.67,
+      "step": 31880,
+      "train_speed(iter/s)": 1.319272
+    },
+    {
+      "acc": 0.69516144,
+      "epoch": 0.8088533739218671,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.9412352673480525e-06,
+      "loss": 1.38491373,
+      "memory(GiB)": 113.67,
+      "step": 31885,
+      "train_speed(iter/s)": 1.319292
+    },
+    {
+      "acc": 0.67028208,
+      "epoch": 0.8089802130898022,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.940268859922566e-06,
+      "loss": 1.43085995,
+      "memory(GiB)": 113.67,
+      "step": 31890,
+      "train_speed(iter/s)": 1.319315
+    },
+    {
+      "acc": 0.69234571,
+      "epoch": 0.8091070522577372,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.939302367155394e-06,
+      "loss": 1.36243334,
+      "memory(GiB)": 113.67,
+      "step": 31895,
+      "train_speed(iter/s)": 1.319338
+    },
+    {
+      "acc": 0.66591978,
+      "epoch": 0.8092338914256723,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.9383357890890454e-06,
+      "loss": 1.53533516,
+      "memory(GiB)": 113.67,
+      "step": 31900,
+      "train_speed(iter/s)": 1.319361
+    },
+    {
+      "acc": 0.67148809,
+      "epoch": 0.8093607305936074,
+      "grad_norm": 4.125,
+      "learning_rate": 6.937369125766033e-06,
+      "loss": 1.43625822,
+      "memory(GiB)": 113.67,
+      "step": 31905,
+      "train_speed(iter/s)": 1.319385
+    },
+    {
+      "acc": 0.68129244,
+      "epoch": 0.8094875697615423,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.936402377228879e-06,
+      "loss": 1.45265198,
+      "memory(GiB)": 113.67,
+      "step": 31910,
+      "train_speed(iter/s)": 1.319409
+    },
+    {
+      "acc": 0.69165144,
+      "epoch": 0.8096144089294774,
+      "grad_norm": 2.75,
+      "learning_rate": 6.9354355435201015e-06,
+      "loss": 1.32135468,
+      "memory(GiB)": 113.67,
+      "step": 31915,
+      "train_speed(iter/s)": 1.319432
+    },
+    {
+      "acc": 0.68236294,
+      "epoch": 0.8097412480974124,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.934468624682229e-06,
+      "loss": 1.4045249,
+      "memory(GiB)": 113.67,
+      "step": 31920,
+      "train_speed(iter/s)": 1.319452
+    },
+    {
+      "acc": 0.69052629,
+      "epoch": 0.8098680872653475,
+      "grad_norm": 3.78125,
+      "learning_rate": 6.933501620757789e-06,
+      "loss": 1.35796471,
+      "memory(GiB)": 113.67,
+      "step": 31925,
+      "train_speed(iter/s)": 1.319475
+    },
+    {
+      "acc": 0.69408588,
+      "epoch": 0.8099949264332826,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.932534531789317e-06,
+      "loss": 1.34583101,
+      "memory(GiB)": 113.67,
+      "step": 31930,
+      "train_speed(iter/s)": 1.319494
+    },
+    {
+      "acc": 0.68470702,
+      "epoch": 0.8101217656012176,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.931567357819344e-06,
+      "loss": 1.38183937,
+      "memory(GiB)": 113.67,
+      "step": 31935,
+      "train_speed(iter/s)": 1.319518
+    },
+    {
+      "acc": 0.67732,
+      "epoch": 0.8102486047691527,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.930600098890419e-06,
+      "loss": 1.39605913,
+      "memory(GiB)": 113.67,
+      "step": 31940,
+      "train_speed(iter/s)": 1.319541
+    },
+    {
+      "acc": 0.67730241,
+      "epoch": 0.8103754439370878,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.929632755045079e-06,
+      "loss": 1.41059818,
+      "memory(GiB)": 113.67,
+      "step": 31945,
+      "train_speed(iter/s)": 1.319564
+    },
+    {
+      "acc": 0.69186506,
+      "epoch": 0.8105022831050228,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.9286653263258765e-06,
+      "loss": 1.41776829,
+      "memory(GiB)": 113.67,
+      "step": 31950,
+      "train_speed(iter/s)": 1.319588
+    },
+    {
+      "acc": 0.67118311,
+      "epoch": 0.8106291222729579,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.927697812775363e-06,
+      "loss": 1.45416431,
+      "memory(GiB)": 113.67,
+      "step": 31955,
+      "train_speed(iter/s)": 1.319611
+    },
+    {
+      "acc": 0.6714323,
+      "epoch": 0.8107559614408929,
+      "grad_norm": 3.625,
+      "learning_rate": 6.926730214436091e-06,
+      "loss": 1.42134838,
+      "memory(GiB)": 113.67,
+      "step": 31960,
+      "train_speed(iter/s)": 1.319634
+    },
+    {
+      "acc": 0.69153552,
+      "epoch": 0.810882800608828,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.925762531350624e-06,
+      "loss": 1.37084284,
+      "memory(GiB)": 113.67,
+      "step": 31965,
+      "train_speed(iter/s)": 1.319658
+    },
+    {
+      "acc": 0.67859869,
+      "epoch": 0.8110096397767631,
+      "grad_norm": 3.953125,
+      "learning_rate": 6.924794763561522e-06,
+      "loss": 1.43393822,
+      "memory(GiB)": 113.67,
+      "step": 31970,
+      "train_speed(iter/s)": 1.319683
+    },
+    {
+      "acc": 0.67671485,
+      "epoch": 0.8111364789446981,
+      "grad_norm": 3.125,
+      "learning_rate": 6.923826911111353e-06,
+      "loss": 1.40415077,
+      "memory(GiB)": 113.67,
+      "step": 31975,
+      "train_speed(iter/s)": 1.319707
+    },
+    {
+      "acc": 0.69179759,
+      "epoch": 0.8112633181126332,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.922858974042688e-06,
+      "loss": 1.36930656,
+      "memory(GiB)": 113.67,
+      "step": 31980,
+      "train_speed(iter/s)": 1.31973
+    },
+    {
+      "acc": 0.70530529,
+      "epoch": 0.8113901572805683,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.921890952398098e-06,
+      "loss": 1.34854193,
+      "memory(GiB)": 113.67,
+      "step": 31985,
+      "train_speed(iter/s)": 1.319753
+    },
+    {
+      "acc": 0.68192501,
+      "epoch": 0.8115169964485033,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.920922846220166e-06,
+      "loss": 1.46663666,
+      "memory(GiB)": 113.67,
+      "step": 31990,
+      "train_speed(iter/s)": 1.319777
+    },
+    {
+      "acc": 0.66888647,
+      "epoch": 0.8116438356164384,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.919954655551469e-06,
+      "loss": 1.39148655,
+      "memory(GiB)": 113.67,
+      "step": 31995,
+      "train_speed(iter/s)": 1.319801
+    },
+    {
+      "acc": 0.70414476,
+      "epoch": 0.8117706747843734,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.918986380434594e-06,
+      "loss": 1.34295101,
+      "memory(GiB)": 113.67,
+      "step": 32000,
+      "train_speed(iter/s)": 1.319822
+    },
+    {
+      "epoch": 0.8117706747843734,
+      "eval_acc": 0.6747321225334495,
+      "eval_loss": 1.3584792613983154,
+      "eval_runtime": 69.2524,
+      "eval_samples_per_second": 91.982,
+      "eval_steps_per_second": 23.003,
+      "step": 32000
+    },
+    {
+      "acc": 0.70074754,
+      "epoch": 0.8118975139523085,
+      "grad_norm": 3.921875,
+      "learning_rate": 6.918018020912132e-06,
+      "loss": 1.42565403,
+      "memory(GiB)": 113.67,
+      "step": 32005,
+      "train_speed(iter/s)": 1.31541
+    },
+    {
+      "acc": 0.69829941,
+      "epoch": 0.8120243531202436,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.917049577026673e-06,
+      "loss": 1.36632462,
+      "memory(GiB)": 113.67,
+      "step": 32010,
+      "train_speed(iter/s)": 1.315432
+    },
+    {
+      "acc": 0.68855906,
+      "epoch": 0.8121511922881786,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.916081048820815e-06,
+      "loss": 1.35778122,
+      "memory(GiB)": 113.67,
+      "step": 32015,
+      "train_speed(iter/s)": 1.315457
+    },
+    {
+      "acc": 0.68316088,
+      "epoch": 0.8122780314561137,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.915112436337157e-06,
+      "loss": 1.41092987,
+      "memory(GiB)": 113.67,
+      "step": 32020,
+      "train_speed(iter/s)": 1.315479
+    },
+    {
+      "acc": 0.68020787,
+      "epoch": 0.8124048706240488,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.914143739618305e-06,
+      "loss": 1.44798269,
+      "memory(GiB)": 113.67,
+      "step": 32025,
+      "train_speed(iter/s)": 1.315502
+    },
+    {
+      "acc": 0.71392021,
+      "epoch": 0.8125317097919837,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.913174958706865e-06,
+      "loss": 1.30150986,
+      "memory(GiB)": 113.67,
+      "step": 32030,
+      "train_speed(iter/s)": 1.315525
+    },
+    {
+      "acc": 0.68470726,
+      "epoch": 0.8126585489599188,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.912206093645448e-06,
+      "loss": 1.43436909,
+      "memory(GiB)": 113.67,
+      "step": 32035,
+      "train_speed(iter/s)": 1.315549
+    },
+    {
+      "acc": 0.69894166,
+      "epoch": 0.8127853881278538,
+      "grad_norm": 4.375,
+      "learning_rate": 6.91123714447667e-06,
+      "loss": 1.37403841,
+      "memory(GiB)": 113.67,
+      "step": 32040,
+      "train_speed(iter/s)": 1.315572
+    },
+    {
+      "acc": 0.68406553,
+      "epoch": 0.8129122272957889,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.910268111243149e-06,
+      "loss": 1.35905132,
+      "memory(GiB)": 113.67,
+      "step": 32045,
+      "train_speed(iter/s)": 1.315595
+    },
+    {
+      "acc": 0.66690412,
+      "epoch": 0.813039066463724,
+      "grad_norm": 3.25,
+      "learning_rate": 6.909298993987508e-06,
+      "loss": 1.52871084,
+      "memory(GiB)": 113.67,
+      "step": 32050,
+      "train_speed(iter/s)": 1.315618
+    },
+    {
+      "acc": 0.68075728,
+      "epoch": 0.813165905631659,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.908329792752373e-06,
+      "loss": 1.38878527,
+      "memory(GiB)": 113.67,
+      "step": 32055,
+      "train_speed(iter/s)": 1.31564
+    },
+    {
+      "acc": 0.69320273,
+      "epoch": 0.8132927447995941,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.907360507580374e-06,
+      "loss": 1.41454945,
+      "memory(GiB)": 113.67,
+      "step": 32060,
+      "train_speed(iter/s)": 1.315664
+    },
+    {
+      "acc": 0.67696476,
+      "epoch": 0.8134195839675292,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.9063911385141425e-06,
+      "loss": 1.41448002,
+      "memory(GiB)": 113.67,
+      "step": 32065,
+      "train_speed(iter/s)": 1.315687
+    },
+    {
+      "acc": 0.67855086,
+      "epoch": 0.8135464231354642,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.9054216855963194e-06,
+      "loss": 1.43078117,
+      "memory(GiB)": 113.67,
+      "step": 32070,
+      "train_speed(iter/s)": 1.315709
+    },
+    {
+      "acc": 0.67680688,
+      "epoch": 0.8136732623033993,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.904452148869541e-06,
+      "loss": 1.41618004,
+      "memory(GiB)": 113.67,
+      "step": 32075,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.68295541,
+      "epoch": 0.8138001014713343,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.903482528376457e-06,
+      "loss": 1.36136227,
+      "memory(GiB)": 113.67,
+      "step": 32080,
+      "train_speed(iter/s)": 1.315752
+    },
+    {
+      "acc": 0.68318548,
+      "epoch": 0.8139269406392694,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.902512824159711e-06,
+      "loss": 1.46858253,
+      "memory(GiB)": 113.67,
+      "step": 32085,
+      "train_speed(iter/s)": 1.315775
+    },
+    {
+      "acc": 0.67840347,
+      "epoch": 0.8140537798072045,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.901543036261957e-06,
+      "loss": 1.42164812,
+      "memory(GiB)": 113.67,
+      "step": 32090,
+      "train_speed(iter/s)": 1.315797
+    },
+    {
+      "acc": 0.67406135,
+      "epoch": 0.8141806189751395,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.900573164725852e-06,
+      "loss": 1.45595531,
+      "memory(GiB)": 113.67,
+      "step": 32095,
+      "train_speed(iter/s)": 1.315821
+    },
+    {
+      "acc": 0.6988276,
+      "epoch": 0.8143074581430746,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.899603209594052e-06,
+      "loss": 1.35743074,
+      "memory(GiB)": 113.67,
+      "step": 32100,
+      "train_speed(iter/s)": 1.315843
+    },
+    {
+      "acc": 0.6881712,
+      "epoch": 0.8144342973110097,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.898633170909224e-06,
+      "loss": 1.39671497,
+      "memory(GiB)": 113.67,
+      "step": 32105,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68190002,
+      "epoch": 0.8145611364789447,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.897663048714031e-06,
+      "loss": 1.38225698,
+      "memory(GiB)": 113.67,
+      "step": 32110,
+      "train_speed(iter/s)": 1.315886
+    },
+    {
+      "acc": 0.6851963,
+      "epoch": 0.8146879756468798,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.896692843051145e-06,
+      "loss": 1.39682493,
+      "memory(GiB)": 113.67,
+      "step": 32115,
+      "train_speed(iter/s)": 1.315903
+    },
+    {
+      "acc": 0.68397875,
+      "epoch": 0.8148148148148148,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.895722553963239e-06,
+      "loss": 1.395644,
+      "memory(GiB)": 113.67,
+      "step": 32120,
+      "train_speed(iter/s)": 1.315927
+    },
+    {
+      "acc": 0.68386388,
+      "epoch": 0.8149416539827499,
+      "grad_norm": 3.875,
+      "learning_rate": 6.8947521814929915e-06,
+      "loss": 1.42767496,
+      "memory(GiB)": 113.67,
+      "step": 32125,
+      "train_speed(iter/s)": 1.31595
+    },
+    {
+      "acc": 0.67375841,
+      "epoch": 0.815068493150685,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.8937817256830834e-06,
+      "loss": 1.52094803,
+      "memory(GiB)": 113.67,
+      "step": 32130,
+      "train_speed(iter/s)": 1.315974
+    },
+    {
+      "acc": 0.68622475,
+      "epoch": 0.81519533231862,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.892811186576199e-06,
+      "loss": 1.41893797,
+      "memory(GiB)": 113.67,
+      "step": 32135,
+      "train_speed(iter/s)": 1.315997
+    },
+    {
+      "acc": 0.68285046,
+      "epoch": 0.8153221714865551,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.8918405642150295e-06,
+      "loss": 1.41105814,
+      "memory(GiB)": 113.67,
+      "step": 32140,
+      "train_speed(iter/s)": 1.31602
+    },
+    {
+      "acc": 0.68357787,
+      "epoch": 0.8154490106544902,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.890869858642264e-06,
+      "loss": 1.38568516,
+      "memory(GiB)": 113.67,
+      "step": 32145,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.68527808,
+      "epoch": 0.8155758498224251,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.889899069900603e-06,
+      "loss": 1.41489897,
+      "memory(GiB)": 113.67,
+      "step": 32150,
+      "train_speed(iter/s)": 1.316066
+    },
+    {
+      "acc": 0.68576803,
+      "epoch": 0.8157026889903602,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.888928198032741e-06,
+      "loss": 1.38424425,
+      "memory(GiB)": 113.67,
+      "step": 32155,
+      "train_speed(iter/s)": 1.316089
+    },
+    {
+      "acc": 0.69361525,
+      "epoch": 0.8158295281582952,
+      "grad_norm": 3.796875,
+      "learning_rate": 6.887957243081384e-06,
+      "loss": 1.37349854,
+      "memory(GiB)": 113.67,
+      "step": 32160,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.68544455,
+      "epoch": 0.8159563673262303,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.886986205089237e-06,
+      "loss": 1.39184885,
+      "memory(GiB)": 113.67,
+      "step": 32165,
+      "train_speed(iter/s)": 1.316135
+    },
+    {
+      "acc": 0.68017173,
+      "epoch": 0.8160832064941654,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.886015084099011e-06,
+      "loss": 1.39425049,
+      "memory(GiB)": 113.67,
+      "step": 32170,
+      "train_speed(iter/s)": 1.316157
+    },
+    {
+      "acc": 0.68876162,
+      "epoch": 0.8162100456621004,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.885043880153424e-06,
+      "loss": 1.40187588,
+      "memory(GiB)": 113.67,
+      "step": 32175,
+      "train_speed(iter/s)": 1.316171
+    },
+    {
+      "acc": 0.69694023,
+      "epoch": 0.8163368848300355,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.88407259329519e-06,
+      "loss": 1.32798195,
+      "memory(GiB)": 113.67,
+      "step": 32180,
+      "train_speed(iter/s)": 1.316195
+    },
+    {
+      "acc": 0.6928071,
+      "epoch": 0.8164637239979706,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.883101223567031e-06,
+      "loss": 1.39223595,
+      "memory(GiB)": 113.67,
+      "step": 32185,
+      "train_speed(iter/s)": 1.316217
+    },
+    {
+      "acc": 0.69114957,
+      "epoch": 0.8165905631659056,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.882129771011674e-06,
+      "loss": 1.34781837,
+      "memory(GiB)": 113.67,
+      "step": 32190,
+      "train_speed(iter/s)": 1.316241
+    },
+    {
+      "acc": 0.69899759,
+      "epoch": 0.8167174023338407,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.881158235671845e-06,
+      "loss": 1.29396057,
+      "memory(GiB)": 113.67,
+      "step": 32195,
+      "train_speed(iter/s)": 1.316264
+    },
+    {
+      "acc": 0.69279256,
+      "epoch": 0.8168442415017757,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.8801866175902785e-06,
+      "loss": 1.36616211,
+      "memory(GiB)": 113.67,
+      "step": 32200,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.68395667,
+      "epoch": 0.8169710806697108,
+      "grad_norm": 2.875,
+      "learning_rate": 6.87921491680971e-06,
+      "loss": 1.45294724,
+      "memory(GiB)": 113.67,
+      "step": 32205,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.68197889,
+      "epoch": 0.8170979198376459,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.878243133372882e-06,
+      "loss": 1.3726305,
+      "memory(GiB)": 113.67,
+      "step": 32210,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.69035196,
+      "epoch": 0.8172247590055809,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.877271267322532e-06,
+      "loss": 1.38823404,
+      "memory(GiB)": 113.67,
+      "step": 32215,
+      "train_speed(iter/s)": 1.316358
+    },
+    {
+      "acc": 0.67411456,
+      "epoch": 0.817351598173516,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.876299318701412e-06,
+      "loss": 1.46186628,
+      "memory(GiB)": 113.67,
+      "step": 32220,
+      "train_speed(iter/s)": 1.316381
+    },
+    {
+      "acc": 0.68276939,
+      "epoch": 0.8174784373414511,
+      "grad_norm": 3.125,
+      "learning_rate": 6.875327287552269e-06,
+      "loss": 1.40958195,
+      "memory(GiB)": 113.67,
+      "step": 32225,
+      "train_speed(iter/s)": 1.316404
+    },
+    {
+      "acc": 0.6806314,
+      "epoch": 0.8176052765093861,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.8743551739178615e-06,
+      "loss": 1.41550369,
+      "memory(GiB)": 113.67,
+      "step": 32230,
+      "train_speed(iter/s)": 1.316428
+    },
+    {
+      "acc": 0.68403196,
+      "epoch": 0.8177321156773212,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.8733829778409425e-06,
+      "loss": 1.44311199,
+      "memory(GiB)": 113.67,
+      "step": 32235,
+      "train_speed(iter/s)": 1.316451
+    },
+    {
+      "acc": 0.70523448,
+      "epoch": 0.8178589548452562,
+      "grad_norm": 3.125,
+      "learning_rate": 6.872410699364278e-06,
+      "loss": 1.24821434,
+      "memory(GiB)": 113.67,
+      "step": 32240,
+      "train_speed(iter/s)": 1.316475
+    },
+    {
+      "acc": 0.68202467,
+      "epoch": 0.8179857940131913,
+      "grad_norm": 2.875,
+      "learning_rate": 6.8714383385306305e-06,
+      "loss": 1.4266223,
+      "memory(GiB)": 113.67,
+      "step": 32245,
+      "train_speed(iter/s)": 1.316499
+    },
+    {
+      "acc": 0.69042959,
+      "epoch": 0.8181126331811264,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.870465895382769e-06,
+      "loss": 1.33817301,
+      "memory(GiB)": 113.67,
+      "step": 32250,
+      "train_speed(iter/s)": 1.316523
+    },
+    {
+      "acc": 0.69956789,
+      "epoch": 0.8182394723490614,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.869493369963468e-06,
+      "loss": 1.3595562,
+      "memory(GiB)": 113.67,
+      "step": 32255,
+      "train_speed(iter/s)": 1.316546
+    },
+    {
+      "acc": 0.69058385,
+      "epoch": 0.8183663115169965,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.8685207623155e-06,
+      "loss": 1.39397249,
+      "memory(GiB)": 113.67,
+      "step": 32260,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.67239952,
+      "epoch": 0.8184931506849316,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.867548072481649e-06,
+      "loss": 1.4669735,
+      "memory(GiB)": 113.67,
+      "step": 32265,
+      "train_speed(iter/s)": 1.316593
+    },
+    {
+      "acc": 0.66921067,
+      "epoch": 0.8186199898528665,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.866575300504695e-06,
+      "loss": 1.39670372,
+      "memory(GiB)": 113.67,
+      "step": 32270,
+      "train_speed(iter/s)": 1.316616
+    },
+    {
+      "acc": 0.69682226,
+      "epoch": 0.8187468290208016,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.865602446427424e-06,
+      "loss": 1.39233284,
+      "memory(GiB)": 113.67,
+      "step": 32275,
+      "train_speed(iter/s)": 1.31664
+    },
+    {
+      "acc": 0.68490515,
+      "epoch": 0.8188736681887366,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.864629510292629e-06,
+      "loss": 1.41486702,
+      "memory(GiB)": 113.67,
+      "step": 32280,
+      "train_speed(iter/s)": 1.316663
+    },
+    {
+      "acc": 0.69915552,
+      "epoch": 0.8190005073566717,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.863656492143103e-06,
+      "loss": 1.35381937,
+      "memory(GiB)": 113.67,
+      "step": 32285,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.66940522,
+      "epoch": 0.8191273465246068,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.862683392021644e-06,
+      "loss": 1.39089413,
+      "memory(GiB)": 113.67,
+      "step": 32290,
+      "train_speed(iter/s)": 1.316709
+    },
+    {
+      "acc": 0.70140953,
+      "epoch": 0.8192541856925418,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.861710209971052e-06,
+      "loss": 1.27351265,
+      "memory(GiB)": 113.67,
+      "step": 32295,
+      "train_speed(iter/s)": 1.31673
+    },
+    {
+      "acc": 0.67354851,
+      "epoch": 0.8193810248604769,
+      "grad_norm": 2.875,
+      "learning_rate": 6.860736946034136e-06,
+      "loss": 1.43978777,
+      "memory(GiB)": 113.67,
+      "step": 32300,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.67493596,
+      "epoch": 0.819507864028412,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.859763600253698e-06,
+      "loss": 1.49934578,
+      "memory(GiB)": 113.67,
+      "step": 32305,
+      "train_speed(iter/s)": 1.316775
+    },
+    {
+      "acc": 0.69005508,
+      "epoch": 0.819634703196347,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.858790172672556e-06,
+      "loss": 1.41727533,
+      "memory(GiB)": 113.67,
+      "step": 32310,
+      "train_speed(iter/s)": 1.316799
+    },
+    {
+      "acc": 0.69540911,
+      "epoch": 0.8197615423642821,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.857816663333523e-06,
+      "loss": 1.36140184,
+      "memory(GiB)": 113.67,
+      "step": 32315,
+      "train_speed(iter/s)": 1.316824
+    },
+    {
+      "acc": 0.70134721,
+      "epoch": 0.8198883815322171,
+      "grad_norm": 4.25,
+      "learning_rate": 6.856843072279418e-06,
+      "loss": 1.29960833,
+      "memory(GiB)": 113.67,
+      "step": 32320,
+      "train_speed(iter/s)": 1.316847
+    },
+    {
+      "acc": 0.68777442,
+      "epoch": 0.8200152207001522,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.855869399553065e-06,
+      "loss": 1.45591116,
+      "memory(GiB)": 113.67,
+      "step": 32325,
+      "train_speed(iter/s)": 1.316869
+    },
+    {
+      "acc": 0.68480968,
+      "epoch": 0.8201420598680873,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.85489564519729e-06,
+      "loss": 1.39311514,
+      "memory(GiB)": 113.67,
+      "step": 32330,
+      "train_speed(iter/s)": 1.316893
+    },
+    {
+      "acc": 0.69918489,
+      "epoch": 0.8202688990360223,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.853921809254922e-06,
+      "loss": 1.37799702,
+      "memory(GiB)": 113.67,
+      "step": 32335,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.67909007,
+      "epoch": 0.8203957382039574,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.852947891768796e-06,
+      "loss": 1.38114967,
+      "memory(GiB)": 113.67,
+      "step": 32340,
+      "train_speed(iter/s)": 1.31694
+    },
+    {
+      "acc": 0.70302491,
+      "epoch": 0.8205225773718925,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.851973892781749e-06,
+      "loss": 1.29244995,
+      "memory(GiB)": 113.67,
+      "step": 32345,
+      "train_speed(iter/s)": 1.316963
+    },
+    {
+      "acc": 0.68546128,
+      "epoch": 0.8206494165398275,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.850999812336623e-06,
+      "loss": 1.38632336,
+      "memory(GiB)": 113.67,
+      "step": 32350,
+      "train_speed(iter/s)": 1.316985
+    },
+    {
+      "acc": 0.68279877,
+      "epoch": 0.8207762557077626,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.850025650476259e-06,
+      "loss": 1.39548397,
+      "memory(GiB)": 113.67,
+      "step": 32355,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.66967826,
+      "epoch": 0.8209030948756976,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.849051407243509e-06,
+      "loss": 1.41053677,
+      "memory(GiB)": 113.67,
+      "step": 32360,
+      "train_speed(iter/s)": 1.317031
+    },
+    {
+      "acc": 0.67432308,
+      "epoch": 0.8210299340436327,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.8480770826812205e-06,
+      "loss": 1.41541786,
+      "memory(GiB)": 113.67,
+      "step": 32365,
+      "train_speed(iter/s)": 1.317053
+    },
+    {
+      "acc": 0.68571606,
+      "epoch": 0.8211567732115678,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.847102676832253e-06,
+      "loss": 1.44316807,
+      "memory(GiB)": 113.67,
+      "step": 32370,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.68076129,
+      "epoch": 0.8212836123795028,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.8461281897394615e-06,
+      "loss": 1.4140007,
+      "memory(GiB)": 113.67,
+      "step": 32375,
+      "train_speed(iter/s)": 1.317101
+    },
+    {
+      "acc": 0.66793718,
+      "epoch": 0.8214104515474379,
+      "grad_norm": 3.25,
+      "learning_rate": 6.845153621445711e-06,
+      "loss": 1.44850464,
+      "memory(GiB)": 113.67,
+      "step": 32380,
+      "train_speed(iter/s)": 1.317124
+    },
+    {
+      "acc": 0.69182911,
+      "epoch": 0.821537290715373,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.844178971993866e-06,
+      "loss": 1.45634613,
+      "memory(GiB)": 113.67,
+      "step": 32385,
+      "train_speed(iter/s)": 1.317148
+    },
+    {
+      "acc": 0.69537287,
+      "epoch": 0.821664129883308,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.843204241426797e-06,
+      "loss": 1.39509058,
+      "memory(GiB)": 113.67,
+      "step": 32390,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.70103674,
+      "epoch": 0.821790969051243,
+      "grad_norm": 2.453125,
+      "learning_rate": 6.842229429787375e-06,
+      "loss": 1.40258379,
+      "memory(GiB)": 113.67,
+      "step": 32395,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.68812075,
+      "epoch": 0.821917808219178,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.841254537118477e-06,
+      "loss": 1.32725611,
+      "memory(GiB)": 113.67,
+      "step": 32400,
+      "train_speed(iter/s)": 1.317217
+    },
+    {
+      "acc": 0.67010994,
+      "epoch": 0.8220446473871131,
+      "grad_norm": 3.25,
+      "learning_rate": 6.840279563462985e-06,
+      "loss": 1.4754775,
+      "memory(GiB)": 113.67,
+      "step": 32405,
+      "train_speed(iter/s)": 1.317241
+    },
+    {
+      "acc": 0.69154215,
+      "epoch": 0.8221714865550482,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.839304508863781e-06,
+      "loss": 1.39977579,
+      "memory(GiB)": 113.67,
+      "step": 32410,
+      "train_speed(iter/s)": 1.317264
+    },
+    {
+      "acc": 0.69198771,
+      "epoch": 0.8222983257229832,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.838329373363753e-06,
+      "loss": 1.42561188,
+      "memory(GiB)": 113.67,
+      "step": 32415,
+      "train_speed(iter/s)": 1.317287
+    },
+    {
+      "acc": 0.68734283,
+      "epoch": 0.8224251648909183,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.8373541570057924e-06,
+      "loss": 1.40522308,
+      "memory(GiB)": 113.67,
+      "step": 32420,
+      "train_speed(iter/s)": 1.31731
+    },
+    {
+      "acc": 0.6927845,
+      "epoch": 0.8225520040588534,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.836378859832791e-06,
+      "loss": 1.41036091,
+      "memory(GiB)": 113.67,
+      "step": 32425,
+      "train_speed(iter/s)": 1.317334
+    },
+    {
+      "acc": 0.68510618,
+      "epoch": 0.8226788432267884,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.83540348188765e-06,
+      "loss": 1.372999,
+      "memory(GiB)": 113.67,
+      "step": 32430,
+      "train_speed(iter/s)": 1.317358
+    },
+    {
+      "acc": 0.69010162,
+      "epoch": 0.8228056823947235,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.834428023213268e-06,
+      "loss": 1.36363468,
+      "memory(GiB)": 113.67,
+      "step": 32435,
+      "train_speed(iter/s)": 1.317382
+    },
+    {
+      "acc": 0.68114896,
+      "epoch": 0.8229325215626585,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.833452483852554e-06,
+      "loss": 1.37159061,
+      "memory(GiB)": 113.67,
+      "step": 32440,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.69549384,
+      "epoch": 0.8230593607305936,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.832476863848411e-06,
+      "loss": 1.36365738,
+      "memory(GiB)": 113.67,
+      "step": 32445,
+      "train_speed(iter/s)": 1.317427
+    },
+    {
+      "acc": 0.68154516,
+      "epoch": 0.8231861998985287,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.831501163243756e-06,
+      "loss": 1.45780144,
+      "memory(GiB)": 113.67,
+      "step": 32450,
+      "train_speed(iter/s)": 1.31745
+    },
+    {
+      "acc": 0.68032832,
+      "epoch": 0.8233130390664637,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.830525382081501e-06,
+      "loss": 1.41479082,
+      "memory(GiB)": 113.67,
+      "step": 32455,
+      "train_speed(iter/s)": 1.317473
+    },
+    {
+      "acc": 0.68954859,
+      "epoch": 0.8234398782343988,
+      "grad_norm": 2.59375,
+      "learning_rate": 6.829549520404568e-06,
+      "loss": 1.36553478,
+      "memory(GiB)": 113.67,
+      "step": 32460,
+      "train_speed(iter/s)": 1.317496
+    },
+    {
+      "acc": 0.6956027,
+      "epoch": 0.8235667174023339,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.828573578255879e-06,
+      "loss": 1.35696297,
+      "memory(GiB)": 113.67,
+      "step": 32465,
+      "train_speed(iter/s)": 1.317516
+    },
+    {
+      "acc": 0.68252087,
+      "epoch": 0.8236935565702689,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.82759755567836e-06,
+      "loss": 1.39794102,
+      "memory(GiB)": 113.67,
+      "step": 32470,
+      "train_speed(iter/s)": 1.317539
+    },
+    {
+      "acc": 0.68279486,
+      "epoch": 0.823820395738204,
+      "grad_norm": 3.0,
+      "learning_rate": 6.826621452714941e-06,
+      "loss": 1.44284611,
+      "memory(GiB)": 113.67,
+      "step": 32475,
+      "train_speed(iter/s)": 1.317563
+    },
+    {
+      "acc": 0.67953534,
+      "epoch": 0.823947234906139,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.825645269408556e-06,
+      "loss": 1.42178898,
+      "memory(GiB)": 113.67,
+      "step": 32480,
+      "train_speed(iter/s)": 1.317585
+    },
+    {
+      "acc": 0.68580332,
+      "epoch": 0.8240740740740741,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.82466900580214e-06,
+      "loss": 1.37224569,
+      "memory(GiB)": 113.67,
+      "step": 32485,
+      "train_speed(iter/s)": 1.317607
+    },
+    {
+      "acc": 0.6735074,
+      "epoch": 0.8242009132420092,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.823692661938634e-06,
+      "loss": 1.46871958,
+      "memory(GiB)": 113.67,
+      "step": 32490,
+      "train_speed(iter/s)": 1.317631
+    },
+    {
+      "acc": 0.68704176,
+      "epoch": 0.8243277524099442,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.822716237860984e-06,
+      "loss": 1.41832647,
+      "memory(GiB)": 113.67,
+      "step": 32495,
+      "train_speed(iter/s)": 1.317653
+    },
+    {
+      "acc": 0.69226627,
+      "epoch": 0.8244545915778793,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.821739733612135e-06,
+      "loss": 1.35371208,
+      "memory(GiB)": 113.67,
+      "step": 32500,
+      "train_speed(iter/s)": 1.317676
+    },
+    {
+      "acc": 0.68898506,
+      "epoch": 0.8245814307458144,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.820763149235039e-06,
+      "loss": 1.46788301,
+      "memory(GiB)": 113.67,
+      "step": 32505,
+      "train_speed(iter/s)": 1.317699
+    },
+    {
+      "acc": 0.68636012,
+      "epoch": 0.8247082699137493,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.819786484772652e-06,
+      "loss": 1.35938005,
+      "memory(GiB)": 113.67,
+      "step": 32510,
+      "train_speed(iter/s)": 1.317722
+    },
+    {
+      "acc": 0.65850163,
+      "epoch": 0.8248351090816844,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.8188097402679275e-06,
+      "loss": 1.50540762,
+      "memory(GiB)": 113.67,
+      "step": 32515,
+      "train_speed(iter/s)": 1.317743
+    },
+    {
+      "acc": 0.69772243,
+      "epoch": 0.8249619482496194,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.817832915763833e-06,
+      "loss": 1.32795563,
+      "memory(GiB)": 113.67,
+      "step": 32520,
+      "train_speed(iter/s)": 1.317765
+    },
+    {
+      "acc": 0.67888632,
+      "epoch": 0.8250887874175545,
+      "grad_norm": 3.125,
+      "learning_rate": 6.81685601130333e-06,
+      "loss": 1.40335598,
+      "memory(GiB)": 113.67,
+      "step": 32525,
+      "train_speed(iter/s)": 1.317786
+    },
+    {
+      "acc": 0.70466571,
+      "epoch": 0.8252156265854896,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.8158790269293885e-06,
+      "loss": 1.3404356,
+      "memory(GiB)": 113.67,
+      "step": 32530,
+      "train_speed(iter/s)": 1.317809
+    },
+    {
+      "acc": 0.70097256,
+      "epoch": 0.8253424657534246,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.8149019626849785e-06,
+      "loss": 1.3092597,
+      "memory(GiB)": 113.67,
+      "step": 32535,
+      "train_speed(iter/s)": 1.317832
+    },
+    {
+      "acc": 0.67788982,
+      "epoch": 0.8254693049213597,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.813924818613079e-06,
+      "loss": 1.42242584,
+      "memory(GiB)": 113.67,
+      "step": 32540,
+      "train_speed(iter/s)": 1.317853
+    },
+    {
+      "acc": 0.69527874,
+      "epoch": 0.8255961440892948,
+      "grad_norm": 3.375,
+      "learning_rate": 6.812947594756667e-06,
+      "loss": 1.41039276,
+      "memory(GiB)": 113.67,
+      "step": 32545,
+      "train_speed(iter/s)": 1.317875
+    },
+    {
+      "acc": 0.70336823,
+      "epoch": 0.8257229832572298,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.811970291158725e-06,
+      "loss": 1.37286282,
+      "memory(GiB)": 113.67,
+      "step": 32550,
+      "train_speed(iter/s)": 1.317897
+    },
+    {
+      "acc": 0.69121342,
+      "epoch": 0.8258498224251649,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.810992907862239e-06,
+      "loss": 1.38498087,
+      "memory(GiB)": 113.67,
+      "step": 32555,
+      "train_speed(iter/s)": 1.31792
+    },
+    {
+      "acc": 0.67363143,
+      "epoch": 0.8259766615930999,
+      "grad_norm": 3.625,
+      "learning_rate": 6.810015444910202e-06,
+      "loss": 1.38467741,
+      "memory(GiB)": 113.67,
+      "step": 32560,
+      "train_speed(iter/s)": 1.317941
+    },
+    {
+      "acc": 0.6931386,
+      "epoch": 0.826103500761035,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.809037902345603e-06,
+      "loss": 1.44979553,
+      "memory(GiB)": 113.67,
+      "step": 32565,
+      "train_speed(iter/s)": 1.317964
+    },
+    {
+      "acc": 0.68568521,
+      "epoch": 0.8262303399289701,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.808060280211439e-06,
+      "loss": 1.39219227,
+      "memory(GiB)": 113.67,
+      "step": 32570,
+      "train_speed(iter/s)": 1.317985
+    },
+    {
+      "acc": 0.69876938,
+      "epoch": 0.8263571790969051,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.807082578550713e-06,
+      "loss": 1.40190344,
+      "memory(GiB)": 113.67,
+      "step": 32575,
+      "train_speed(iter/s)": 1.318008
+    },
+    {
+      "acc": 0.6788023,
+      "epoch": 0.8264840182648402,
+      "grad_norm": 3.125,
+      "learning_rate": 6.806104797406428e-06,
+      "loss": 1.3794281,
+      "memory(GiB)": 113.67,
+      "step": 32580,
+      "train_speed(iter/s)": 1.318029
+    },
+    {
+      "acc": 0.67475538,
+      "epoch": 0.8266108574327753,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.805126936821588e-06,
+      "loss": 1.42720299,
+      "memory(GiB)": 113.67,
+      "step": 32585,
+      "train_speed(iter/s)": 1.31805
+    },
+    {
+      "acc": 0.68390265,
+      "epoch": 0.8267376966007103,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.804148996839208e-06,
+      "loss": 1.40651264,
+      "memory(GiB)": 113.67,
+      "step": 32590,
+      "train_speed(iter/s)": 1.318073
+    },
+    {
+      "acc": 0.68543863,
+      "epoch": 0.8268645357686454,
+      "grad_norm": 3.0,
+      "learning_rate": 6.803170977502298e-06,
+      "loss": 1.41892538,
+      "memory(GiB)": 113.67,
+      "step": 32595,
+      "train_speed(iter/s)": 1.318095
+    },
+    {
+      "acc": 0.69647522,
+      "epoch": 0.8269913749365804,
+      "grad_norm": 5.0625,
+      "learning_rate": 6.802192878853879e-06,
+      "loss": 1.39158401,
+      "memory(GiB)": 113.67,
+      "step": 32600,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.68500433,
+      "epoch": 0.8271182141045155,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.801214700936972e-06,
+      "loss": 1.34594746,
+      "memory(GiB)": 113.67,
+      "step": 32605,
+      "train_speed(iter/s)": 1.318139
+    },
+    {
+      "acc": 0.68936682,
+      "epoch": 0.8272450532724506,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.8002364437946e-06,
+      "loss": 1.37257814,
+      "memory(GiB)": 113.67,
+      "step": 32610,
+      "train_speed(iter/s)": 1.31816
+    },
+    {
+      "acc": 0.68675737,
+      "epoch": 0.8273718924403856,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.799258107469792e-06,
+      "loss": 1.43001213,
+      "memory(GiB)": 113.67,
+      "step": 32615,
+      "train_speed(iter/s)": 1.318183
+    },
+    {
+      "acc": 0.68549643,
+      "epoch": 0.8274987316083207,
+      "grad_norm": 2.4375,
+      "learning_rate": 6.798279692005578e-06,
+      "loss": 1.34479313,
+      "memory(GiB)": 113.67,
+      "step": 32620,
+      "train_speed(iter/s)": 1.318202
+    },
+    {
+      "acc": 0.69594364,
+      "epoch": 0.8276255707762558,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.7973011974449965e-06,
+      "loss": 1.34537392,
+      "memory(GiB)": 113.67,
+      "step": 32625,
+      "train_speed(iter/s)": 1.318226
+    },
+    {
+      "acc": 0.69570646,
+      "epoch": 0.8277524099441907,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.796322623831082e-06,
+      "loss": 1.34916716,
+      "memory(GiB)": 113.67,
+      "step": 32630,
+      "train_speed(iter/s)": 1.318249
+    },
+    {
+      "acc": 0.68422432,
+      "epoch": 0.8278792491121258,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.795343971206879e-06,
+      "loss": 1.36709976,
+      "memory(GiB)": 113.67,
+      "step": 32635,
+      "train_speed(iter/s)": 1.318273
+    },
+    {
+      "acc": 0.69319563,
+      "epoch": 0.8280060882800608,
+      "grad_norm": 6.0625,
+      "learning_rate": 6.794365239615433e-06,
+      "loss": 1.37155104,
+      "memory(GiB)": 113.67,
+      "step": 32640,
+      "train_speed(iter/s)": 1.318295
+    },
+    {
+      "acc": 0.69204664,
+      "epoch": 0.8281329274479959,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.793386429099792e-06,
+      "loss": 1.40751114,
+      "memory(GiB)": 113.67,
+      "step": 32645,
+      "train_speed(iter/s)": 1.318316
+    },
+    {
+      "acc": 0.6920166,
+      "epoch": 0.828259766615931,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.79240753970301e-06,
+      "loss": 1.36507416,
+      "memory(GiB)": 113.67,
+      "step": 32650,
+      "train_speed(iter/s)": 1.31834
+    },
+    {
+      "acc": 0.67823319,
+      "epoch": 0.828386605783866,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.791428571468139e-06,
+      "loss": 1.44316111,
+      "memory(GiB)": 113.67,
+      "step": 32655,
+      "train_speed(iter/s)": 1.318361
+    },
+    {
+      "acc": 0.67511234,
+      "epoch": 0.8285134449518011,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.7904495244382454e-06,
+      "loss": 1.44703426,
+      "memory(GiB)": 113.67,
+      "step": 32660,
+      "train_speed(iter/s)": 1.318383
+    },
+    {
+      "acc": 0.67904878,
+      "epoch": 0.8286402841197362,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.789470398656385e-06,
+      "loss": 1.45582533,
+      "memory(GiB)": 113.67,
+      "step": 32665,
+      "train_speed(iter/s)": 1.318405
+    },
+    {
+      "acc": 0.68943052,
+      "epoch": 0.8287671232876712,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.788491194165629e-06,
+      "loss": 1.39267159,
+      "memory(GiB)": 113.67,
+      "step": 32670,
+      "train_speed(iter/s)": 1.318428
+    },
+    {
+      "acc": 0.70759459,
+      "epoch": 0.8288939624556063,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.787511911009044e-06,
+      "loss": 1.30704908,
+      "memory(GiB)": 113.67,
+      "step": 32675,
+      "train_speed(iter/s)": 1.31845
+    },
+    {
+      "acc": 0.67647095,
+      "epoch": 0.8290208016235413,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.786532549229704e-06,
+      "loss": 1.44321833,
+      "memory(GiB)": 113.67,
+      "step": 32680,
+      "train_speed(iter/s)": 1.318473
+    },
+    {
+      "acc": 0.68973093,
+      "epoch": 0.8291476407914764,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.785553108870686e-06,
+      "loss": 1.34867668,
+      "memory(GiB)": 113.67,
+      "step": 32685,
+      "train_speed(iter/s)": 1.318495
+    },
+    {
+      "acc": 0.67324657,
+      "epoch": 0.8292744799594115,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.784573589975072e-06,
+      "loss": 1.41995792,
+      "memory(GiB)": 113.67,
+      "step": 32690,
+      "train_speed(iter/s)": 1.318518
+    },
+    {
+      "acc": 0.67282095,
+      "epoch": 0.8294013191273465,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.783593992585943e-06,
+      "loss": 1.44704161,
+      "memory(GiB)": 113.67,
+      "step": 32695,
+      "train_speed(iter/s)": 1.318541
+    },
+    {
+      "acc": 0.67791286,
+      "epoch": 0.8295281582952816,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.7826143167463876e-06,
+      "loss": 1.43709173,
+      "memory(GiB)": 113.67,
+      "step": 32700,
+      "train_speed(iter/s)": 1.318564
+    },
+    {
+      "acc": 0.68665571,
+      "epoch": 0.8296549974632167,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.781634562499495e-06,
+      "loss": 1.43010578,
+      "memory(GiB)": 113.67,
+      "step": 32705,
+      "train_speed(iter/s)": 1.318586
+    },
+    {
+      "acc": 0.6947711,
+      "epoch": 0.8297818366311517,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.780654729888361e-06,
+      "loss": 1.40354156,
+      "memory(GiB)": 113.67,
+      "step": 32710,
+      "train_speed(iter/s)": 1.31861
+    },
+    {
+      "acc": 0.69070473,
+      "epoch": 0.8299086757990868,
+      "grad_norm": 5.96875,
+      "learning_rate": 6.779674818956081e-06,
+      "loss": 1.3988512,
+      "memory(GiB)": 113.67,
+      "step": 32715,
+      "train_speed(iter/s)": 1.318634
+    },
+    {
+      "acc": 0.67815952,
+      "epoch": 0.8300355149670218,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.778694829745756e-06,
+      "loss": 1.42460356,
+      "memory(GiB)": 113.67,
+      "step": 32720,
+      "train_speed(iter/s)": 1.318656
+    },
+    {
+      "acc": 0.67427979,
+      "epoch": 0.8301623541349569,
+      "grad_norm": 3.0,
+      "learning_rate": 6.777714762300492e-06,
+      "loss": 1.3962368,
+      "memory(GiB)": 113.67,
+      "step": 32725,
+      "train_speed(iter/s)": 1.318679
+    },
+    {
+      "acc": 0.70224514,
+      "epoch": 0.830289193302892,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.776734616663397e-06,
+      "loss": 1.3414875,
+      "memory(GiB)": 113.67,
+      "step": 32730,
+      "train_speed(iter/s)": 1.318702
+    },
+    {
+      "acc": 0.70473804,
+      "epoch": 0.830416032470827,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.77575439287758e-06,
+      "loss": 1.278895,
+      "memory(GiB)": 113.67,
+      "step": 32735,
+      "train_speed(iter/s)": 1.318724
+    },
+    {
+      "acc": 0.68700275,
+      "epoch": 0.8305428716387621,
+      "grad_norm": 3.125,
+      "learning_rate": 6.774774090986157e-06,
+      "loss": 1.37760077,
+      "memory(GiB)": 113.67,
+      "step": 32740,
+      "train_speed(iter/s)": 1.318746
+    },
+    {
+      "acc": 0.70025473,
+      "epoch": 0.8306697108066972,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.773793711032244e-06,
+      "loss": 1.38969822,
+      "memory(GiB)": 113.67,
+      "step": 32745,
+      "train_speed(iter/s)": 1.318768
+    },
+    {
+      "acc": 0.67951236,
+      "epoch": 0.8307965499746321,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.772813253058965e-06,
+      "loss": 1.42694283,
+      "memory(GiB)": 113.67,
+      "step": 32750,
+      "train_speed(iter/s)": 1.31879
+    },
+    {
+      "acc": 0.69045486,
+      "epoch": 0.8309233891425672,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.771832717109444e-06,
+      "loss": 1.38167953,
+      "memory(GiB)": 113.67,
+      "step": 32755,
+      "train_speed(iter/s)": 1.318813
+    },
+    {
+      "acc": 0.67643061,
+      "epoch": 0.8310502283105022,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.77085210322681e-06,
+      "loss": 1.43129873,
+      "memory(GiB)": 113.67,
+      "step": 32760,
+      "train_speed(iter/s)": 1.318835
+    },
+    {
+      "acc": 0.67537184,
+      "epoch": 0.8311770674784373,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.769871411454195e-06,
+      "loss": 1.39566631,
+      "memory(GiB)": 113.67,
+      "step": 32765,
+      "train_speed(iter/s)": 1.318857
+    },
+    {
+      "acc": 0.66654449,
+      "epoch": 0.8313039066463724,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.768890641834732e-06,
+      "loss": 1.45438042,
+      "memory(GiB)": 113.67,
+      "step": 32770,
+      "train_speed(iter/s)": 1.318879
+    },
+    {
+      "acc": 0.68634911,
+      "epoch": 0.8314307458143074,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.767909794411562e-06,
+      "loss": 1.39179916,
+      "memory(GiB)": 113.67,
+      "step": 32775,
+      "train_speed(iter/s)": 1.318901
+    },
+    {
+      "acc": 0.68938866,
+      "epoch": 0.8315575849822425,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.7669288692278256e-06,
+      "loss": 1.40316696,
+      "memory(GiB)": 113.67,
+      "step": 32780,
+      "train_speed(iter/s)": 1.318923
+    },
+    {
+      "acc": 0.68204966,
+      "epoch": 0.8316844241501776,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.76594786632667e-06,
+      "loss": 1.32839813,
+      "memory(GiB)": 113.67,
+      "step": 32785,
+      "train_speed(iter/s)": 1.318941
+    },
+    {
+      "acc": 0.6844039,
+      "epoch": 0.8318112633181126,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.764966785751242e-06,
+      "loss": 1.42370586,
+      "memory(GiB)": 113.67,
+      "step": 32790,
+      "train_speed(iter/s)": 1.318963
+    },
+    {
+      "acc": 0.6845602,
+      "epoch": 0.8319381024860477,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.763985627544693e-06,
+      "loss": 1.40373774,
+      "memory(GiB)": 113.67,
+      "step": 32795,
+      "train_speed(iter/s)": 1.318984
+    },
+    {
+      "acc": 0.6734973,
+      "epoch": 0.8320649416539827,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.763004391750183e-06,
+      "loss": 1.40654459,
+      "memory(GiB)": 113.67,
+      "step": 32800,
+      "train_speed(iter/s)": 1.319006
+    },
+    {
+      "acc": 0.68703203,
+      "epoch": 0.8321917808219178,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.762023078410867e-06,
+      "loss": 1.42803869,
+      "memory(GiB)": 113.67,
+      "step": 32805,
+      "train_speed(iter/s)": 1.319027
+    },
+    {
+      "acc": 0.68867941,
+      "epoch": 0.8323186199898529,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.7610416875699095e-06,
+      "loss": 1.36753912,
+      "memory(GiB)": 113.67,
+      "step": 32810,
+      "train_speed(iter/s)": 1.319049
+    },
+    {
+      "acc": 0.70272789,
+      "epoch": 0.8324454591577879,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.760060219270476e-06,
+      "loss": 1.3669961,
+      "memory(GiB)": 113.67,
+      "step": 32815,
+      "train_speed(iter/s)": 1.319071
+    },
+    {
+      "acc": 0.69513044,
+      "epoch": 0.832572298325723,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.759078673555736e-06,
+      "loss": 1.40654898,
+      "memory(GiB)": 113.67,
+      "step": 32820,
+      "train_speed(iter/s)": 1.319093
+    },
+    {
+      "acc": 0.68283434,
+      "epoch": 0.8326991374936581,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.758097050468862e-06,
+      "loss": 1.43042717,
+      "memory(GiB)": 113.67,
+      "step": 32825,
+      "train_speed(iter/s)": 1.319114
+    },
+    {
+      "acc": 0.68900561,
+      "epoch": 0.8328259766615931,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.757115350053032e-06,
+      "loss": 1.43876133,
+      "memory(GiB)": 113.67,
+      "step": 32830,
+      "train_speed(iter/s)": 1.319136
+    },
+    {
+      "acc": 0.68106093,
+      "epoch": 0.8329528158295282,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.756133572351422e-06,
+      "loss": 1.4040761,
+      "memory(GiB)": 113.67,
+      "step": 32835,
+      "train_speed(iter/s)": 1.319158
+    },
+    {
+      "acc": 0.69078369,
+      "epoch": 0.8330796549974632,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.755151717407218e-06,
+      "loss": 1.34673443,
+      "memory(GiB)": 113.67,
+      "step": 32840,
+      "train_speed(iter/s)": 1.319181
+    },
+    {
+      "acc": 0.68349066,
+      "epoch": 0.8332064941653983,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.754169785263605e-06,
+      "loss": 1.4308506,
+      "memory(GiB)": 113.67,
+      "step": 32845,
+      "train_speed(iter/s)": 1.319204
+    },
+    {
+      "acc": 0.69063654,
+      "epoch": 0.8333333333333334,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.753187775963773e-06,
+      "loss": 1.3847702,
+      "memory(GiB)": 113.67,
+      "step": 32850,
+      "train_speed(iter/s)": 1.319227
+    },
+    {
+      "acc": 0.69655566,
+      "epoch": 0.8334601725012684,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.752205689550915e-06,
+      "loss": 1.31247005,
+      "memory(GiB)": 113.67,
+      "step": 32855,
+      "train_speed(iter/s)": 1.319249
+    },
+    {
+      "acc": 0.6890677,
+      "epoch": 0.8335870116692035,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.751223526068228e-06,
+      "loss": 1.38942509,
+      "memory(GiB)": 113.67,
+      "step": 32860,
+      "train_speed(iter/s)": 1.319271
+    },
+    {
+      "acc": 0.67430639,
+      "epoch": 0.8337138508371386,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.75024128555891e-06,
+      "loss": 1.43687611,
+      "memory(GiB)": 113.67,
+      "step": 32865,
+      "train_speed(iter/s)": 1.319295
+    },
+    {
+      "acc": 0.68539448,
+      "epoch": 0.8338406900050735,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.7492589680661695e-06,
+      "loss": 1.39880686,
+      "memory(GiB)": 113.67,
+      "step": 32870,
+      "train_speed(iter/s)": 1.319317
+    },
+    {
+      "acc": 0.680936,
+      "epoch": 0.8339675291730086,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.748276573633207e-06,
+      "loss": 1.45492706,
+      "memory(GiB)": 113.67,
+      "step": 32875,
+      "train_speed(iter/s)": 1.319335
+    },
+    {
+      "acc": 0.68470201,
+      "epoch": 0.8340943683409436,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.747294102303237e-06,
+      "loss": 1.38976479,
+      "memory(GiB)": 113.67,
+      "step": 32880,
+      "train_speed(iter/s)": 1.31936
+    },
+    {
+      "acc": 0.70105138,
+      "epoch": 0.8342212075088787,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.746311554119469e-06,
+      "loss": 1.35234356,
+      "memory(GiB)": 113.67,
+      "step": 32885,
+      "train_speed(iter/s)": 1.319382
+    },
+    {
+      "acc": 0.69868469,
+      "epoch": 0.8343480466768138,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.745328929125125e-06,
+      "loss": 1.32553425,
+      "memory(GiB)": 113.67,
+      "step": 32890,
+      "train_speed(iter/s)": 1.319404
+    },
+    {
+      "acc": 0.6858161,
+      "epoch": 0.8344748858447488,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.7443462273634195e-06,
+      "loss": 1.33957615,
+      "memory(GiB)": 113.67,
+      "step": 32895,
+      "train_speed(iter/s)": 1.319426
+    },
+    {
+      "acc": 0.68738999,
+      "epoch": 0.8346017250126839,
+      "grad_norm": 3.375,
+      "learning_rate": 6.74336344887758e-06,
+      "loss": 1.35743065,
+      "memory(GiB)": 113.67,
+      "step": 32900,
+      "train_speed(iter/s)": 1.319449
+    },
+    {
+      "acc": 0.68869972,
+      "epoch": 0.834728564180619,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.742380593710834e-06,
+      "loss": 1.38445463,
+      "memory(GiB)": 113.67,
+      "step": 32905,
+      "train_speed(iter/s)": 1.31947
+    },
+    {
+      "acc": 0.67530565,
+      "epoch": 0.834855403348554,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.7413976619064085e-06,
+      "loss": 1.45120382,
+      "memory(GiB)": 113.67,
+      "step": 32910,
+      "train_speed(iter/s)": 1.319492
+    },
+    {
+      "acc": 0.69405937,
+      "epoch": 0.8349822425164891,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.74041465350754e-06,
+      "loss": 1.32388229,
+      "memory(GiB)": 113.67,
+      "step": 32915,
+      "train_speed(iter/s)": 1.31951
+    },
+    {
+      "acc": 0.70195522,
+      "epoch": 0.8351090816844241,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.739431568557464e-06,
+      "loss": 1.28183603,
+      "memory(GiB)": 113.67,
+      "step": 32920,
+      "train_speed(iter/s)": 1.319532
+    },
+    {
+      "acc": 0.66269102,
+      "epoch": 0.8352359208523592,
+      "grad_norm": 3.0,
+      "learning_rate": 6.738448407099423e-06,
+      "loss": 1.47978954,
+      "memory(GiB)": 113.67,
+      "step": 32925,
+      "train_speed(iter/s)": 1.319554
+    },
+    {
+      "acc": 0.68324571,
+      "epoch": 0.8353627600202943,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.737465169176658e-06,
+      "loss": 1.35281324,
+      "memory(GiB)": 113.67,
+      "step": 32930,
+      "train_speed(iter/s)": 1.319576
+    },
+    {
+      "acc": 0.68030386,
+      "epoch": 0.8354895991882293,
+      "grad_norm": 4.96875,
+      "learning_rate": 6.736481854832418e-06,
+      "loss": 1.43431387,
+      "memory(GiB)": 113.67,
+      "step": 32935,
+      "train_speed(iter/s)": 1.319598
+    },
+    {
+      "acc": 0.68476686,
+      "epoch": 0.8356164383561644,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.735498464109953e-06,
+      "loss": 1.38413334,
+      "memory(GiB)": 113.67,
+      "step": 32940,
+      "train_speed(iter/s)": 1.31962
+    },
+    {
+      "acc": 0.6907793,
+      "epoch": 0.8357432775240995,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.734514997052517e-06,
+      "loss": 1.33990536,
+      "memory(GiB)": 113.67,
+      "step": 32945,
+      "train_speed(iter/s)": 1.319641
+    },
+    {
+      "acc": 0.69838247,
+      "epoch": 0.8358701166920345,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.733531453703368e-06,
+      "loss": 1.37822189,
+      "memory(GiB)": 113.67,
+      "step": 32950,
+      "train_speed(iter/s)": 1.319664
+    },
+    {
+      "acc": 0.68370171,
+      "epoch": 0.8359969558599696,
+      "grad_norm": 3.78125,
+      "learning_rate": 6.732547834105765e-06,
+      "loss": 1.32229795,
+      "memory(GiB)": 113.67,
+      "step": 32955,
+      "train_speed(iter/s)": 1.319687
+    },
+    {
+      "acc": 0.69682951,
+      "epoch": 0.8361237950279046,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.731564138302975e-06,
+      "loss": 1.38036861,
+      "memory(GiB)": 113.67,
+      "step": 32960,
+      "train_speed(iter/s)": 1.31971
+    },
+    {
+      "acc": 0.69406295,
+      "epoch": 0.8362506341958397,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.730580366338261e-06,
+      "loss": 1.39586573,
+      "memory(GiB)": 113.67,
+      "step": 32965,
+      "train_speed(iter/s)": 1.319733
+    },
+    {
+      "acc": 0.70383749,
+      "epoch": 0.8363774733637748,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.729596518254897e-06,
+      "loss": 1.33905144,
+      "memory(GiB)": 113.67,
+      "step": 32970,
+      "train_speed(iter/s)": 1.319755
+    },
+    {
+      "acc": 0.68990531,
+      "epoch": 0.8365043125317098,
+      "grad_norm": 4.0,
+      "learning_rate": 6.728612594096155e-06,
+      "loss": 1.37604122,
+      "memory(GiB)": 113.67,
+      "step": 32975,
+      "train_speed(iter/s)": 1.319778
+    },
+    {
+      "acc": 0.67673635,
+      "epoch": 0.8366311516996449,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.727628593905315e-06,
+      "loss": 1.41754742,
+      "memory(GiB)": 113.67,
+      "step": 32980,
+      "train_speed(iter/s)": 1.3198
+    },
+    {
+      "acc": 0.69559097,
+      "epoch": 0.83675799086758,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.726644517725655e-06,
+      "loss": 1.38017311,
+      "memory(GiB)": 113.67,
+      "step": 32985,
+      "train_speed(iter/s)": 1.319823
+    },
+    {
+      "acc": 0.66410713,
+      "epoch": 0.836884830035515,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.725660365600462e-06,
+      "loss": 1.46914005,
+      "memory(GiB)": 113.67,
+      "step": 32990,
+      "train_speed(iter/s)": 1.319845
+    },
+    {
+      "acc": 0.67321563,
+      "epoch": 0.83701166920345,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.724676137573021e-06,
+      "loss": 1.4280592,
+      "memory(GiB)": 113.67,
+      "step": 32995,
+      "train_speed(iter/s)": 1.319868
+    },
+    {
+      "acc": 0.66853724,
+      "epoch": 0.837138508371385,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.723691833686622e-06,
+      "loss": 1.39590912,
+      "memory(GiB)": 113.67,
+      "step": 33000,
+      "train_speed(iter/s)": 1.31989
+    },
+    {
+      "epoch": 0.837138508371385,
+      "eval_acc": 0.6747220986641106,
+      "eval_loss": 1.3579063415527344,
+      "eval_runtime": 70.1488,
+      "eval_samples_per_second": 90.807,
+      "eval_steps_per_second": 22.709,
+      "step": 33000
+    },
+    {
+      "acc": 0.68380809,
+      "epoch": 0.8372653475393201,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.722707453984561e-06,
+      "loss": 1.41967402,
+      "memory(GiB)": 113.67,
+      "step": 33005,
+      "train_speed(iter/s)": 1.315557
+    },
+    {
+      "acc": 0.69844022,
+      "epoch": 0.8373921867072552,
+      "grad_norm": 3.625,
+      "learning_rate": 6.721722998510135e-06,
+      "loss": 1.43632746,
+      "memory(GiB)": 113.67,
+      "step": 33010,
+      "train_speed(iter/s)": 1.31558
+    },
+    {
+      "acc": 0.69078722,
+      "epoch": 0.8375190258751902,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.720738467306644e-06,
+      "loss": 1.3922184,
+      "memory(GiB)": 113.67,
+      "step": 33015,
+      "train_speed(iter/s)": 1.315603
+    },
+    {
+      "acc": 0.69086285,
+      "epoch": 0.8376458650431253,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.719753860417394e-06,
+      "loss": 1.43351536,
+      "memory(GiB)": 113.67,
+      "step": 33020,
+      "train_speed(iter/s)": 1.315625
+    },
+    {
+      "acc": 0.67188368,
+      "epoch": 0.8377727042110604,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.718769177885689e-06,
+      "loss": 1.45353127,
+      "memory(GiB)": 113.67,
+      "step": 33025,
+      "train_speed(iter/s)": 1.315647
+    },
+    {
+      "acc": 0.70860224,
+      "epoch": 0.8378995433789954,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.717784419754845e-06,
+      "loss": 1.28283834,
+      "memory(GiB)": 113.67,
+      "step": 33030,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.68926878,
+      "epoch": 0.8380263825469305,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.71679958606817e-06,
+      "loss": 1.34154634,
+      "memory(GiB)": 113.67,
+      "step": 33035,
+      "train_speed(iter/s)": 1.315694
+    },
+    {
+      "acc": 0.66721191,
+      "epoch": 0.8381532217148655,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.715814676868985e-06,
+      "loss": 1.46121826,
+      "memory(GiB)": 113.67,
+      "step": 33040,
+      "train_speed(iter/s)": 1.315717
+    },
+    {
+      "acc": 0.67121792,
+      "epoch": 0.8382800608828006,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.714829692200611e-06,
+      "loss": 1.45914078,
+      "memory(GiB)": 113.67,
+      "step": 33045,
+      "train_speed(iter/s)": 1.315741
+    },
+    {
+      "acc": 0.68404503,
+      "epoch": 0.8384069000507357,
+      "grad_norm": 4.375,
+      "learning_rate": 6.71384463210637e-06,
+      "loss": 1.43567781,
+      "memory(GiB)": 113.67,
+      "step": 33050,
+      "train_speed(iter/s)": 1.315753
+    },
+    {
+      "acc": 0.68798113,
+      "epoch": 0.8385337392186707,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.7128594966295904e-06,
+      "loss": 1.36005592,
+      "memory(GiB)": 113.67,
+      "step": 33055,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.69546132,
+      "epoch": 0.8386605783866058,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.711874285813602e-06,
+      "loss": 1.34917402,
+      "memory(GiB)": 113.67,
+      "step": 33060,
+      "train_speed(iter/s)": 1.315799
+    },
+    {
+      "acc": 0.69661527,
+      "epoch": 0.8387874175545409,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.710888999701741e-06,
+      "loss": 1.38958416,
+      "memory(GiB)": 113.67,
+      "step": 33065,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.70106487,
+      "epoch": 0.8389142567224759,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.7099036383373425e-06,
+      "loss": 1.3322834,
+      "memory(GiB)": 113.67,
+      "step": 33070,
+      "train_speed(iter/s)": 1.315841
+    },
+    {
+      "acc": 0.68602266,
+      "epoch": 0.839041095890411,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.708918201763748e-06,
+      "loss": 1.39475307,
+      "memory(GiB)": 113.67,
+      "step": 33075,
+      "train_speed(iter/s)": 1.315864
+    },
+    {
+      "acc": 0.6885561,
+      "epoch": 0.839167935058346,
+      "grad_norm": 2.875,
+      "learning_rate": 6.707932690024302e-06,
+      "loss": 1.41125793,
+      "memory(GiB)": 113.67,
+      "step": 33080,
+      "train_speed(iter/s)": 1.315886
+    },
+    {
+      "acc": 0.67884302,
+      "epoch": 0.8392947742262811,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.706947103162348e-06,
+      "loss": 1.37758808,
+      "memory(GiB)": 113.67,
+      "step": 33085,
+      "train_speed(iter/s)": 1.315909
+    },
+    {
+      "acc": 0.68759913,
+      "epoch": 0.8394216133942162,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.7059614412212425e-06,
+      "loss": 1.41036539,
+      "memory(GiB)": 113.67,
+      "step": 33090,
+      "train_speed(iter/s)": 1.315932
+    },
+    {
+      "acc": 0.6938652,
+      "epoch": 0.8395484525621512,
+      "grad_norm": 2.875,
+      "learning_rate": 6.704975704244334e-06,
+      "loss": 1.38486452,
+      "memory(GiB)": 113.67,
+      "step": 33095,
+      "train_speed(iter/s)": 1.315955
+    },
+    {
+      "acc": 0.67925553,
+      "epoch": 0.8396752917300863,
+      "grad_norm": 4.0,
+      "learning_rate": 6.703989892274985e-06,
+      "loss": 1.45301666,
+      "memory(GiB)": 113.67,
+      "step": 33100,
+      "train_speed(iter/s)": 1.315977
+    },
+    {
+      "acc": 0.6889502,
+      "epoch": 0.8398021308980214,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.703004005356549e-06,
+      "loss": 1.32914677,
+      "memory(GiB)": 113.67,
+      "step": 33105,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.66648612,
+      "epoch": 0.8399289700659563,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.7020180435323965e-06,
+      "loss": 1.44542274,
+      "memory(GiB)": 113.67,
+      "step": 33110,
+      "train_speed(iter/s)": 1.316024
+    },
+    {
+      "acc": 0.67513714,
+      "epoch": 0.8400558092338914,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.701032006845889e-06,
+      "loss": 1.38838844,
+      "memory(GiB)": 113.67,
+      "step": 33115,
+      "train_speed(iter/s)": 1.316045
+    },
+    {
+      "acc": 0.69698057,
+      "epoch": 0.8401826484018264,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.700045895340401e-06,
+      "loss": 1.32894707,
+      "memory(GiB)": 113.67,
+      "step": 33120,
+      "train_speed(iter/s)": 1.316068
+    },
+    {
+      "acc": 0.69020395,
+      "epoch": 0.8403094875697615,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.699059709059304e-06,
+      "loss": 1.41436996,
+      "memory(GiB)": 113.67,
+      "step": 33125,
+      "train_speed(iter/s)": 1.316091
+    },
+    {
+      "acc": 0.70214376,
+      "epoch": 0.8404363267376966,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.698073448045975e-06,
+      "loss": 1.38041391,
+      "memory(GiB)": 113.67,
+      "step": 33130,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.6949306,
+      "epoch": 0.8405631659056316,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.697087112343795e-06,
+      "loss": 1.32121391,
+      "memory(GiB)": 113.67,
+      "step": 33135,
+      "train_speed(iter/s)": 1.316136
+    },
+    {
+      "acc": 0.67245789,
+      "epoch": 0.8406900050735667,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.696100701996146e-06,
+      "loss": 1.46129427,
+      "memory(GiB)": 113.67,
+      "step": 33140,
+      "train_speed(iter/s)": 1.316157
+    },
+    {
+      "acc": 0.69047413,
+      "epoch": 0.8408168442415018,
+      "grad_norm": 3.5,
+      "learning_rate": 6.6951142170464164e-06,
+      "loss": 1.35578184,
+      "memory(GiB)": 113.67,
+      "step": 33145,
+      "train_speed(iter/s)": 1.316181
+    },
+    {
+      "acc": 0.68579998,
+      "epoch": 0.8409436834094368,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.694127657537995e-06,
+      "loss": 1.43433838,
+      "memory(GiB)": 113.67,
+      "step": 33150,
+      "train_speed(iter/s)": 1.316204
+    },
+    {
+      "acc": 0.67843361,
+      "epoch": 0.8410705225773719,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.693141023514276e-06,
+      "loss": 1.38856773,
+      "memory(GiB)": 113.67,
+      "step": 33155,
+      "train_speed(iter/s)": 1.316227
+    },
+    {
+      "acc": 0.67636385,
+      "epoch": 0.8411973617453069,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.6921543150186555e-06,
+      "loss": 1.42762947,
+      "memory(GiB)": 113.67,
+      "step": 33160,
+      "train_speed(iter/s)": 1.31625
+    },
+    {
+      "acc": 0.69766769,
+      "epoch": 0.841324200913242,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.691167532094531e-06,
+      "loss": 1.3656518,
+      "memory(GiB)": 113.67,
+      "step": 33165,
+      "train_speed(iter/s)": 1.316273
+    },
+    {
+      "acc": 0.69181399,
+      "epoch": 0.8414510400811771,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.690180674785311e-06,
+      "loss": 1.4568346,
+      "memory(GiB)": 113.67,
+      "step": 33170,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.67114935,
+      "epoch": 0.8415778792491121,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.689193743134397e-06,
+      "loss": 1.42633343,
+      "memory(GiB)": 113.67,
+      "step": 33175,
+      "train_speed(iter/s)": 1.316321
+    },
+    {
+      "acc": 0.67980251,
+      "epoch": 0.8417047184170472,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.688206737185201e-06,
+      "loss": 1.40527086,
+      "memory(GiB)": 113.67,
+      "step": 33180,
+      "train_speed(iter/s)": 1.316344
+    },
+    {
+      "acc": 0.6773675,
+      "epoch": 0.8418315575849823,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.687219656981135e-06,
+      "loss": 1.38424749,
+      "memory(GiB)": 113.67,
+      "step": 33185,
+      "train_speed(iter/s)": 1.316366
+    },
+    {
+      "acc": 0.69788499,
+      "epoch": 0.8419583967529173,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.686232502565616e-06,
+      "loss": 1.39115829,
+      "memory(GiB)": 113.67,
+      "step": 33190,
+      "train_speed(iter/s)": 1.31639
+    },
+    {
+      "acc": 0.68758221,
+      "epoch": 0.8420852359208524,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.685245273982063e-06,
+      "loss": 1.41607132,
+      "memory(GiB)": 113.67,
+      "step": 33195,
+      "train_speed(iter/s)": 1.316414
+    },
+    {
+      "acc": 0.67074213,
+      "epoch": 0.8422120750887874,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.684257971273899e-06,
+      "loss": 1.51562366,
+      "memory(GiB)": 113.67,
+      "step": 33200,
+      "train_speed(iter/s)": 1.316437
+    },
+    {
+      "acc": 0.69482884,
+      "epoch": 0.8423389142567225,
+      "grad_norm": 2.875,
+      "learning_rate": 6.68327059448455e-06,
+      "loss": 1.3566988,
+      "memory(GiB)": 113.67,
+      "step": 33205,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.69332824,
+      "epoch": 0.8424657534246576,
+      "grad_norm": 4.125,
+      "learning_rate": 6.682283143657444e-06,
+      "loss": 1.34346132,
+      "memory(GiB)": 113.67,
+      "step": 33210,
+      "train_speed(iter/s)": 1.316483
+    },
+    {
+      "acc": 0.68254781,
+      "epoch": 0.8425925925925926,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.681295618836015e-06,
+      "loss": 1.44954739,
+      "memory(GiB)": 113.67,
+      "step": 33215,
+      "train_speed(iter/s)": 1.316506
+    },
+    {
+      "acc": 0.68037539,
+      "epoch": 0.8427194317605277,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.680308020063699e-06,
+      "loss": 1.39097614,
+      "memory(GiB)": 113.67,
+      "step": 33220,
+      "train_speed(iter/s)": 1.316529
+    },
+    {
+      "acc": 0.68540068,
+      "epoch": 0.8428462709284628,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.679320347383933e-06,
+      "loss": 1.41793938,
+      "memory(GiB)": 113.67,
+      "step": 33225,
+      "train_speed(iter/s)": 1.316552
+    },
+    {
+      "acc": 0.69364519,
+      "epoch": 0.8429731100963977,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.678332600840161e-06,
+      "loss": 1.41888027,
+      "memory(GiB)": 113.67,
+      "step": 33230,
+      "train_speed(iter/s)": 1.316574
+    },
+    {
+      "acc": 0.71006846,
+      "epoch": 0.8430999492643328,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.677344780475827e-06,
+      "loss": 1.32304993,
+      "memory(GiB)": 113.67,
+      "step": 33235,
+      "train_speed(iter/s)": 1.316596
+    },
+    {
+      "acc": 0.69806318,
+      "epoch": 0.8432267884322678,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.676356886334383e-06,
+      "loss": 1.36641483,
+      "memory(GiB)": 113.67,
+      "step": 33240,
+      "train_speed(iter/s)": 1.316618
+    },
+    {
+      "acc": 0.68254538,
+      "epoch": 0.8433536276002029,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.675368918459276e-06,
+      "loss": 1.44172497,
+      "memory(GiB)": 113.67,
+      "step": 33245,
+      "train_speed(iter/s)": 1.316641
+    },
+    {
+      "acc": 0.70550365,
+      "epoch": 0.843480466768138,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.674380876893967e-06,
+      "loss": 1.3101696,
+      "memory(GiB)": 113.67,
+      "step": 33250,
+      "train_speed(iter/s)": 1.316662
+    },
+    {
+      "acc": 0.69031067,
+      "epoch": 0.843607305936073,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.673392761681908e-06,
+      "loss": 1.36513367,
+      "memory(GiB)": 113.67,
+      "step": 33255,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.70043468,
+      "epoch": 0.8437341451040081,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.672404572866566e-06,
+      "loss": 1.34704523,
+      "memory(GiB)": 113.67,
+      "step": 33260,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.69440446,
+      "epoch": 0.8438609842719432,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.671416310491406e-06,
+      "loss": 1.3203557,
+      "memory(GiB)": 113.67,
+      "step": 33265,
+      "train_speed(iter/s)": 1.316726
+    },
+    {
+      "acc": 0.67793059,
+      "epoch": 0.8439878234398782,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.670427974599891e-06,
+      "loss": 1.47841759,
+      "memory(GiB)": 113.67,
+      "step": 33270,
+      "train_speed(iter/s)": 1.316748
+    },
+    {
+      "acc": 0.68896041,
+      "epoch": 0.8441146626078133,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.669439565235498e-06,
+      "loss": 1.42864866,
+      "memory(GiB)": 113.67,
+      "step": 33275,
+      "train_speed(iter/s)": 1.31677
+    },
+    {
+      "acc": 0.68700476,
+      "epoch": 0.8442415017757483,
+      "grad_norm": 3.25,
+      "learning_rate": 6.668451082441698e-06,
+      "loss": 1.39655561,
+      "memory(GiB)": 113.67,
+      "step": 33280,
+      "train_speed(iter/s)": 1.316792
+    },
+    {
+      "acc": 0.66994081,
+      "epoch": 0.8443683409436834,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.667462526261972e-06,
+      "loss": 1.42008829,
+      "memory(GiB)": 113.67,
+      "step": 33285,
+      "train_speed(iter/s)": 1.316813
+    },
+    {
+      "acc": 0.69869804,
+      "epoch": 0.8444951801116185,
+      "grad_norm": 2.53125,
+      "learning_rate": 6.666473896739798e-06,
+      "loss": 1.36755619,
+      "memory(GiB)": 113.67,
+      "step": 33290,
+      "train_speed(iter/s)": 1.316835
+    },
+    {
+      "acc": 0.6888885,
+      "epoch": 0.8446220192795535,
+      "grad_norm": 3.375,
+      "learning_rate": 6.665485193918663e-06,
+      "loss": 1.38754225,
+      "memory(GiB)": 113.67,
+      "step": 33295,
+      "train_speed(iter/s)": 1.316856
+    },
+    {
+      "acc": 0.67282872,
+      "epoch": 0.8447488584474886,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.664496417842053e-06,
+      "loss": 1.41575651,
+      "memory(GiB)": 113.67,
+      "step": 33300,
+      "train_speed(iter/s)": 1.316879
+    },
+    {
+      "acc": 0.69294281,
+      "epoch": 0.8448756976154237,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.6635075685534566e-06,
+      "loss": 1.36747589,
+      "memory(GiB)": 113.67,
+      "step": 33305,
+      "train_speed(iter/s)": 1.3169
+    },
+    {
+      "acc": 0.67993169,
+      "epoch": 0.8450025367833587,
+      "grad_norm": 3.796875,
+      "learning_rate": 6.662518646096374e-06,
+      "loss": 1.42338371,
+      "memory(GiB)": 113.67,
+      "step": 33310,
+      "train_speed(iter/s)": 1.316922
+    },
+    {
+      "acc": 0.68269825,
+      "epoch": 0.8451293759512938,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.661529650514296e-06,
+      "loss": 1.3958025,
+      "memory(GiB)": 113.67,
+      "step": 33315,
+      "train_speed(iter/s)": 1.316944
+    },
+    {
+      "acc": 0.67385702,
+      "epoch": 0.8452562151192288,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.6605405818507274e-06,
+      "loss": 1.4340477,
+      "memory(GiB)": 113.67,
+      "step": 33320,
+      "train_speed(iter/s)": 1.316964
+    },
+    {
+      "acc": 0.68310137,
+      "epoch": 0.8453830542871639,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.659551440149169e-06,
+      "loss": 1.37827835,
+      "memory(GiB)": 113.67,
+      "step": 33325,
+      "train_speed(iter/s)": 1.316985
+    },
+    {
+      "acc": 0.6944355,
+      "epoch": 0.845509893455099,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.65856222545313e-06,
+      "loss": 1.35344458,
+      "memory(GiB)": 113.67,
+      "step": 33330,
+      "train_speed(iter/s)": 1.317008
+    },
+    {
+      "acc": 0.69016647,
+      "epoch": 0.845636732623034,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.657572937806118e-06,
+      "loss": 1.40471849,
+      "memory(GiB)": 113.67,
+      "step": 33335,
+      "train_speed(iter/s)": 1.317025
+    },
+    {
+      "acc": 0.69494419,
+      "epoch": 0.8457635717909691,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.656583577251649e-06,
+      "loss": 1.30879402,
+      "memory(GiB)": 113.67,
+      "step": 33340,
+      "train_speed(iter/s)": 1.317047
+    },
+    {
+      "acc": 0.6687305,
+      "epoch": 0.8458904109589042,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.655594143833237e-06,
+      "loss": 1.44198208,
+      "memory(GiB)": 113.67,
+      "step": 33345,
+      "train_speed(iter/s)": 1.31707
+    },
+    {
+      "acc": 0.68720627,
+      "epoch": 0.8460172501268391,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.654604637594404e-06,
+      "loss": 1.37540836,
+      "memory(GiB)": 113.67,
+      "step": 33350,
+      "train_speed(iter/s)": 1.317092
+    },
+    {
+      "acc": 0.66918821,
+      "epoch": 0.8461440892947742,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.653615058578672e-06,
+      "loss": 1.46652441,
+      "memory(GiB)": 113.67,
+      "step": 33355,
+      "train_speed(iter/s)": 1.317115
+    },
+    {
+      "acc": 0.70367789,
+      "epoch": 0.8462709284627092,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.652625406829566e-06,
+      "loss": 1.36139927,
+      "memory(GiB)": 113.67,
+      "step": 33360,
+      "train_speed(iter/s)": 1.317138
+    },
+    {
+      "acc": 0.68934422,
+      "epoch": 0.8463977676306443,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.651635682390616e-06,
+      "loss": 1.35308619,
+      "memory(GiB)": 113.67,
+      "step": 33365,
+      "train_speed(iter/s)": 1.31716
+    },
+    {
+      "acc": 0.69941578,
+      "epoch": 0.8465246067985794,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.650645885305356e-06,
+      "loss": 1.35932579,
+      "memory(GiB)": 113.67,
+      "step": 33370,
+      "train_speed(iter/s)": 1.317182
+    },
+    {
+      "acc": 0.67627015,
+      "epoch": 0.8466514459665144,
+      "grad_norm": 4.59375,
+      "learning_rate": 6.649656015617319e-06,
+      "loss": 1.47242489,
+      "memory(GiB)": 113.67,
+      "step": 33375,
+      "train_speed(iter/s)": 1.317204
+    },
+    {
+      "acc": 0.6794157,
+      "epoch": 0.8467782851344495,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.648666073370046e-06,
+      "loss": 1.38536911,
+      "memory(GiB)": 113.67,
+      "step": 33380,
+      "train_speed(iter/s)": 1.317221
+    },
+    {
+      "acc": 0.68503647,
+      "epoch": 0.8469051243023846,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.647676058607076e-06,
+      "loss": 1.42425861,
+      "memory(GiB)": 113.67,
+      "step": 33385,
+      "train_speed(iter/s)": 1.317244
+    },
+    {
+      "acc": 0.68375607,
+      "epoch": 0.8470319634703196,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.64668597137196e-06,
+      "loss": 1.43436623,
+      "memory(GiB)": 113.67,
+      "step": 33390,
+      "train_speed(iter/s)": 1.317266
+    },
+    {
+      "acc": 0.69443736,
+      "epoch": 0.8471588026382547,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.645695811708241e-06,
+      "loss": 1.38283787,
+      "memory(GiB)": 113.67,
+      "step": 33395,
+      "train_speed(iter/s)": 1.317288
+    },
+    {
+      "acc": 0.69017758,
+      "epoch": 0.8472856418061897,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.644705579659474e-06,
+      "loss": 1.38235216,
+      "memory(GiB)": 113.67,
+      "step": 33400,
+      "train_speed(iter/s)": 1.31731
+    },
+    {
+      "acc": 0.68311129,
+      "epoch": 0.8474124809741248,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.643715275269212e-06,
+      "loss": 1.42162151,
+      "memory(GiB)": 113.67,
+      "step": 33405,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.69236078,
+      "epoch": 0.8475393201420599,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.642724898581013e-06,
+      "loss": 1.40363331,
+      "memory(GiB)": 113.67,
+      "step": 33410,
+      "train_speed(iter/s)": 1.317356
+    },
+    {
+      "acc": 0.6917398,
+      "epoch": 0.8476661593099949,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.6417344496384394e-06,
+      "loss": 1.37182817,
+      "memory(GiB)": 113.67,
+      "step": 33415,
+      "train_speed(iter/s)": 1.317376
+    },
+    {
+      "acc": 0.65855756,
+      "epoch": 0.84779299847793,
+      "grad_norm": 2.46875,
+      "learning_rate": 6.640743928485054e-06,
+      "loss": 1.45416641,
+      "memory(GiB)": 113.67,
+      "step": 33420,
+      "train_speed(iter/s)": 1.317398
+    },
+    {
+      "acc": 0.69853621,
+      "epoch": 0.8479198376458651,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.639753335164426e-06,
+      "loss": 1.43021889,
+      "memory(GiB)": 113.67,
+      "step": 33425,
+      "train_speed(iter/s)": 1.31742
+    },
+    {
+      "acc": 0.68372493,
+      "epoch": 0.8480466768138001,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.638762669720126e-06,
+      "loss": 1.4105463,
+      "memory(GiB)": 113.67,
+      "step": 33430,
+      "train_speed(iter/s)": 1.317442
+    },
+    {
+      "acc": 0.70585446,
+      "epoch": 0.8481735159817352,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.637771932195726e-06,
+      "loss": 1.32917004,
+      "memory(GiB)": 113.67,
+      "step": 33435,
+      "train_speed(iter/s)": 1.317461
+    },
+    {
+      "acc": 0.70267658,
+      "epoch": 0.8483003551496702,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.636781122634804e-06,
+      "loss": 1.3732585,
+      "memory(GiB)": 113.67,
+      "step": 33440,
+      "train_speed(iter/s)": 1.317483
+    },
+    {
+      "acc": 0.70992994,
+      "epoch": 0.8484271943176053,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.635790241080941e-06,
+      "loss": 1.24016857,
+      "memory(GiB)": 113.67,
+      "step": 33445,
+      "train_speed(iter/s)": 1.317505
+    },
+    {
+      "acc": 0.68538642,
+      "epoch": 0.8485540334855404,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.634799287577721e-06,
+      "loss": 1.36731129,
+      "memory(GiB)": 113.67,
+      "step": 33450,
+      "train_speed(iter/s)": 1.317525
+    },
+    {
+      "acc": 0.69374261,
+      "epoch": 0.8486808726534754,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.6338082621687286e-06,
+      "loss": 1.31589394,
+      "memory(GiB)": 113.67,
+      "step": 33455,
+      "train_speed(iter/s)": 1.317548
+    },
+    {
+      "acc": 0.67923956,
+      "epoch": 0.8488077118214105,
+      "grad_norm": 4.75,
+      "learning_rate": 6.6328171648975545e-06,
+      "loss": 1.46347647,
+      "memory(GiB)": 113.67,
+      "step": 33460,
+      "train_speed(iter/s)": 1.317569
+    },
+    {
+      "acc": 0.69649115,
+      "epoch": 0.8489345509893456,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.63182599580779e-06,
+      "loss": 1.33195896,
+      "memory(GiB)": 113.67,
+      "step": 33465,
+      "train_speed(iter/s)": 1.317591
+    },
+    {
+      "acc": 0.68964834,
+      "epoch": 0.8490613901572805,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.630834754943036e-06,
+      "loss": 1.37888756,
+      "memory(GiB)": 113.67,
+      "step": 33470,
+      "train_speed(iter/s)": 1.317611
+    },
+    {
+      "acc": 0.68813653,
+      "epoch": 0.8491882293252156,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.629843442346886e-06,
+      "loss": 1.39841852,
+      "memory(GiB)": 113.67,
+      "step": 33475,
+      "train_speed(iter/s)": 1.317633
+    },
+    {
+      "acc": 0.69206367,
+      "epoch": 0.8493150684931506,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.628852058062944e-06,
+      "loss": 1.42176208,
+      "memory(GiB)": 113.67,
+      "step": 33480,
+      "train_speed(iter/s)": 1.317653
+    },
+    {
+      "acc": 0.66601267,
+      "epoch": 0.8494419076610857,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.627860602134818e-06,
+      "loss": 1.5537652,
+      "memory(GiB)": 113.67,
+      "step": 33485,
+      "train_speed(iter/s)": 1.317675
+    },
+    {
+      "acc": 0.67301378,
+      "epoch": 0.8495687468290208,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.626869074606113e-06,
+      "loss": 1.44278612,
+      "memory(GiB)": 113.67,
+      "step": 33490,
+      "train_speed(iter/s)": 1.317698
+    },
+    {
+      "acc": 0.66668386,
+      "epoch": 0.8496955859969558,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.625877475520445e-06,
+      "loss": 1.46128197,
+      "memory(GiB)": 113.67,
+      "step": 33495,
+      "train_speed(iter/s)": 1.317719
+    },
+    {
+      "acc": 0.69687982,
+      "epoch": 0.8498224251648909,
+      "grad_norm": 3.375,
+      "learning_rate": 6.624885804921425e-06,
+      "loss": 1.2894578,
+      "memory(GiB)": 113.67,
+      "step": 33500,
+      "train_speed(iter/s)": 1.31774
+    },
+    {
+      "acc": 0.695468,
+      "epoch": 0.849949264332826,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.623894062852673e-06,
+      "loss": 1.32105131,
+      "memory(GiB)": 113.67,
+      "step": 33505,
+      "train_speed(iter/s)": 1.317761
+    },
+    {
+      "acc": 0.70534005,
+      "epoch": 0.850076103500761,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.62290224935781e-06,
+      "loss": 1.30999746,
+      "memory(GiB)": 113.67,
+      "step": 33510,
+      "train_speed(iter/s)": 1.317779
+    },
+    {
+      "acc": 0.69381924,
+      "epoch": 0.8502029426686961,
+      "grad_norm": 2.53125,
+      "learning_rate": 6.621910364480461e-06,
+      "loss": 1.35806971,
+      "memory(GiB)": 113.67,
+      "step": 33515,
+      "train_speed(iter/s)": 1.317801
+    },
+    {
+      "acc": 0.68444395,
+      "epoch": 0.8503297818366311,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.620918408264252e-06,
+      "loss": 1.45502529,
+      "memory(GiB)": 113.67,
+      "step": 33520,
+      "train_speed(iter/s)": 1.317822
+    },
+    {
+      "acc": 0.67043037,
+      "epoch": 0.8504566210045662,
+      "grad_norm": 2.875,
+      "learning_rate": 6.6199263807528136e-06,
+      "loss": 1.46844807,
+      "memory(GiB)": 113.67,
+      "step": 33525,
+      "train_speed(iter/s)": 1.317844
+    },
+    {
+      "acc": 0.68508968,
+      "epoch": 0.8505834601725013,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.618934281989783e-06,
+      "loss": 1.42585049,
+      "memory(GiB)": 113.67,
+      "step": 33530,
+      "train_speed(iter/s)": 1.317866
+    },
+    {
+      "acc": 0.69250841,
+      "epoch": 0.8507102993404363,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.6179421120187915e-06,
+      "loss": 1.34730234,
+      "memory(GiB)": 113.67,
+      "step": 33535,
+      "train_speed(iter/s)": 1.317887
+    },
+    {
+      "acc": 0.70455666,
+      "epoch": 0.8508371385083714,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.616949870883486e-06,
+      "loss": 1.30405359,
+      "memory(GiB)": 113.67,
+      "step": 33540,
+      "train_speed(iter/s)": 1.317909
+    },
+    {
+      "acc": 0.68222904,
+      "epoch": 0.8509639776763065,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.615957558627503e-06,
+      "loss": 1.44008865,
+      "memory(GiB)": 113.67,
+      "step": 33545,
+      "train_speed(iter/s)": 1.317931
+    },
+    {
+      "acc": 0.68779216,
+      "epoch": 0.8510908168442415,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.6149651752944945e-06,
+      "loss": 1.37893,
+      "memory(GiB)": 113.67,
+      "step": 33550,
+      "train_speed(iter/s)": 1.317953
+    },
+    {
+      "acc": 0.67922015,
+      "epoch": 0.8512176560121766,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.613972720928105e-06,
+      "loss": 1.40477629,
+      "memory(GiB)": 113.67,
+      "step": 33555,
+      "train_speed(iter/s)": 1.317976
+    },
+    {
+      "acc": 0.68851318,
+      "epoch": 0.8513444951801116,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.61298019557199e-06,
+      "loss": 1.3402977,
+      "memory(GiB)": 113.67,
+      "step": 33560,
+      "train_speed(iter/s)": 1.317998
+    },
+    {
+      "acc": 0.70700836,
+      "epoch": 0.8514713343480467,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.6119875992698045e-06,
+      "loss": 1.31955299,
+      "memory(GiB)": 113.67,
+      "step": 33565,
+      "train_speed(iter/s)": 1.31802
+    },
+    {
+      "acc": 0.6950552,
+      "epoch": 0.8515981735159818,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.610994932065207e-06,
+      "loss": 1.34138412,
+      "memory(GiB)": 113.67,
+      "step": 33570,
+      "train_speed(iter/s)": 1.318042
+    },
+    {
+      "acc": 0.68641901,
+      "epoch": 0.8517250126839168,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.610002194001861e-06,
+      "loss": 1.40349092,
+      "memory(GiB)": 113.67,
+      "step": 33575,
+      "train_speed(iter/s)": 1.318064
+    },
+    {
+      "acc": 0.69533968,
+      "epoch": 0.8518518518518519,
+      "grad_norm": 3.75,
+      "learning_rate": 6.609009385123429e-06,
+      "loss": 1.30710764,
+      "memory(GiB)": 113.67,
+      "step": 33580,
+      "train_speed(iter/s)": 1.318086
+    },
+    {
+      "acc": 0.69199343,
+      "epoch": 0.851978691019787,
+      "grad_norm": 3.375,
+      "learning_rate": 6.608016505473582e-06,
+      "loss": 1.36615086,
+      "memory(GiB)": 113.67,
+      "step": 33585,
+      "train_speed(iter/s)": 1.318107
+    },
+    {
+      "acc": 0.68851228,
+      "epoch": 0.852105530187722,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.60702355509599e-06,
+      "loss": 1.37148705,
+      "memory(GiB)": 113.67,
+      "step": 33590,
+      "train_speed(iter/s)": 1.318128
+    },
+    {
+      "acc": 0.69482303,
+      "epoch": 0.852232369355657,
+      "grad_norm": 2.875,
+      "learning_rate": 6.606030534034326e-06,
+      "loss": 1.39972591,
+      "memory(GiB)": 113.67,
+      "step": 33595,
+      "train_speed(iter/s)": 1.318149
+    },
+    {
+      "acc": 0.67846565,
+      "epoch": 0.852359208523592,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.6050374423322685e-06,
+      "loss": 1.46027145,
+      "memory(GiB)": 113.67,
+      "step": 33600,
+      "train_speed(iter/s)": 1.318171
+    },
+    {
+      "acc": 0.66712456,
+      "epoch": 0.8524860476915271,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.604044280033498e-06,
+      "loss": 1.45851154,
+      "memory(GiB)": 113.67,
+      "step": 33605,
+      "train_speed(iter/s)": 1.318192
+    },
+    {
+      "acc": 0.67386665,
+      "epoch": 0.8526128868594622,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.6030510471817e-06,
+      "loss": 1.45681562,
+      "memory(GiB)": 113.67,
+      "step": 33610,
+      "train_speed(iter/s)": 1.318214
+    },
+    {
+      "acc": 0.66900148,
+      "epoch": 0.8527397260273972,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.602057743820558e-06,
+      "loss": 1.46226025,
+      "memory(GiB)": 113.67,
+      "step": 33615,
+      "train_speed(iter/s)": 1.318235
+    },
+    {
+      "acc": 0.70387363,
+      "epoch": 0.8528665651953323,
+      "grad_norm": 4.90625,
+      "learning_rate": 6.601064369993766e-06,
+      "loss": 1.31819,
+      "memory(GiB)": 113.67,
+      "step": 33620,
+      "train_speed(iter/s)": 1.318256
+    },
+    {
+      "acc": 0.68135109,
+      "epoch": 0.8529934043632674,
+      "grad_norm": 3.125,
+      "learning_rate": 6.600070925745012e-06,
+      "loss": 1.3825983,
+      "memory(GiB)": 113.67,
+      "step": 33625,
+      "train_speed(iter/s)": 1.318277
+    },
+    {
+      "acc": 0.68172779,
+      "epoch": 0.8531202435312024,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.599077411117998e-06,
+      "loss": 1.42836094,
+      "memory(GiB)": 113.67,
+      "step": 33630,
+      "train_speed(iter/s)": 1.318298
+    },
+    {
+      "acc": 0.70079331,
+      "epoch": 0.8532470826991375,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.598083826156418e-06,
+      "loss": 1.32534218,
+      "memory(GiB)": 113.67,
+      "step": 33635,
+      "train_speed(iter/s)": 1.318319
+    },
+    {
+      "acc": 0.69253244,
+      "epoch": 0.8533739218670725,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.597090170903977e-06,
+      "loss": 1.41480942,
+      "memory(GiB)": 113.67,
+      "step": 33640,
+      "train_speed(iter/s)": 1.318341
+    },
+    {
+      "acc": 0.68935528,
+      "epoch": 0.8535007610350076,
+      "grad_norm": 4.125,
+      "learning_rate": 6.596096445404381e-06,
+      "loss": 1.38641081,
+      "memory(GiB)": 113.67,
+      "step": 33645,
+      "train_speed(iter/s)": 1.318362
+    },
+    {
+      "acc": 0.68066759,
+      "epoch": 0.8536276002029427,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.595102649701336e-06,
+      "loss": 1.38188467,
+      "memory(GiB)": 113.67,
+      "step": 33650,
+      "train_speed(iter/s)": 1.318382
+    },
+    {
+      "acc": 0.68805289,
+      "epoch": 0.8537544393708777,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.5941087838385545e-06,
+      "loss": 1.39367313,
+      "memory(GiB)": 113.67,
+      "step": 33655,
+      "train_speed(iter/s)": 1.318403
+    },
+    {
+      "acc": 0.6975296,
+      "epoch": 0.8538812785388128,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.593114847859752e-06,
+      "loss": 1.3479763,
+      "memory(GiB)": 113.67,
+      "step": 33660,
+      "train_speed(iter/s)": 1.318425
+    },
+    {
+      "acc": 0.68266897,
+      "epoch": 0.8540081177067479,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.592120841808646e-06,
+      "loss": 1.43045435,
+      "memory(GiB)": 113.67,
+      "step": 33665,
+      "train_speed(iter/s)": 1.318446
+    },
+    {
+      "acc": 0.69067602,
+      "epoch": 0.8541349568746829,
+      "grad_norm": 2.875,
+      "learning_rate": 6.5911267657289564e-06,
+      "loss": 1.39345627,
+      "memory(GiB)": 113.67,
+      "step": 33670,
+      "train_speed(iter/s)": 1.318466
+    },
+    {
+      "acc": 0.68212123,
+      "epoch": 0.854261796042618,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.590132619664408e-06,
+      "loss": 1.40467405,
+      "memory(GiB)": 113.67,
+      "step": 33675,
+      "train_speed(iter/s)": 1.318486
+    },
+    {
+      "acc": 0.70274587,
+      "epoch": 0.854388635210553,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.589138403658728e-06,
+      "loss": 1.34945459,
+      "memory(GiB)": 113.67,
+      "step": 33680,
+      "train_speed(iter/s)": 1.318507
+    },
+    {
+      "acc": 0.67062244,
+      "epoch": 0.8545154743784881,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.588144117755645e-06,
+      "loss": 1.45124741,
+      "memory(GiB)": 113.67,
+      "step": 33685,
+      "train_speed(iter/s)": 1.318527
+    },
+    {
+      "acc": 0.6920619,
+      "epoch": 0.8546423135464232,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.5871497619988945e-06,
+      "loss": 1.36744452,
+      "memory(GiB)": 113.67,
+      "step": 33690,
+      "train_speed(iter/s)": 1.318548
+    },
+    {
+      "acc": 0.671803,
+      "epoch": 0.8547691527143582,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.586155336432211e-06,
+      "loss": 1.45957069,
+      "memory(GiB)": 113.67,
+      "step": 33695,
+      "train_speed(iter/s)": 1.318569
+    },
+    {
+      "acc": 0.69947824,
+      "epoch": 0.8548959918822933,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.585160841099333e-06,
+      "loss": 1.32483854,
+      "memory(GiB)": 113.67,
+      "step": 33700,
+      "train_speed(iter/s)": 1.31859
+    },
+    {
+      "acc": 0.69785585,
+      "epoch": 0.8550228310502284,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.584166276044005e-06,
+      "loss": 1.38394985,
+      "memory(GiB)": 113.67,
+      "step": 33705,
+      "train_speed(iter/s)": 1.318612
+    },
+    {
+      "acc": 0.67595959,
+      "epoch": 0.8551496702181633,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.583171641309971e-06,
+      "loss": 1.46400127,
+      "memory(GiB)": 113.67,
+      "step": 33710,
+      "train_speed(iter/s)": 1.318632
+    },
+    {
+      "acc": 0.68731422,
+      "epoch": 0.8552765093860984,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.58217693694098e-06,
+      "loss": 1.40760164,
+      "memory(GiB)": 113.67,
+      "step": 33715,
+      "train_speed(iter/s)": 1.318653
+    },
+    {
+      "acc": 0.68703852,
+      "epoch": 0.8554033485540334,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.581182162980784e-06,
+      "loss": 1.36934605,
+      "memory(GiB)": 113.67,
+      "step": 33720,
+      "train_speed(iter/s)": 1.318385
+    },
+    {
+      "acc": 0.66985111,
+      "epoch": 0.8555301877219685,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.580187319473137e-06,
+      "loss": 1.43524742,
+      "memory(GiB)": 113.67,
+      "step": 33725,
+      "train_speed(iter/s)": 1.318406
+    },
+    {
+      "acc": 0.66653337,
+      "epoch": 0.8556570268899036,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.579192406461796e-06,
+      "loss": 1.43646555,
+      "memory(GiB)": 113.67,
+      "step": 33730,
+      "train_speed(iter/s)": 1.318426
+    },
+    {
+      "acc": 0.68715858,
+      "epoch": 0.8557838660578386,
+      "grad_norm": 3.625,
+      "learning_rate": 6.5781974239905225e-06,
+      "loss": 1.45376244,
+      "memory(GiB)": 113.67,
+      "step": 33735,
+      "train_speed(iter/s)": 1.318449
+    },
+    {
+      "acc": 0.67447186,
+      "epoch": 0.8559107052257737,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.57720237210308e-06,
+      "loss": 1.42250977,
+      "memory(GiB)": 113.67,
+      "step": 33740,
+      "train_speed(iter/s)": 1.318471
+    },
+    {
+      "acc": 0.68226089,
+      "epoch": 0.8560375443937088,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.576207250843235e-06,
+      "loss": 1.37399502,
+      "memory(GiB)": 113.67,
+      "step": 33745,
+      "train_speed(iter/s)": 1.318493
+    },
+    {
+      "acc": 0.70173817,
+      "epoch": 0.8561643835616438,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.575212060254759e-06,
+      "loss": 1.3585,
+      "memory(GiB)": 113.67,
+      "step": 33750,
+      "train_speed(iter/s)": 1.318515
+    },
+    {
+      "acc": 0.68667297,
+      "epoch": 0.8562912227295789,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.574216800381424e-06,
+      "loss": 1.44109383,
+      "memory(GiB)": 113.67,
+      "step": 33755,
+      "train_speed(iter/s)": 1.318538
+    },
+    {
+      "acc": 0.69595194,
+      "epoch": 0.8564180618975139,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.573221471267005e-06,
+      "loss": 1.35648966,
+      "memory(GiB)": 113.67,
+      "step": 33760,
+      "train_speed(iter/s)": 1.318557
+    },
+    {
+      "acc": 0.68667555,
+      "epoch": 0.856544901065449,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.572226072955281e-06,
+      "loss": 1.38190775,
+      "memory(GiB)": 113.67,
+      "step": 33765,
+      "train_speed(iter/s)": 1.318577
+    },
+    {
+      "acc": 0.70567889,
+      "epoch": 0.8566717402333841,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.571230605490036e-06,
+      "loss": 1.31653423,
+      "memory(GiB)": 113.67,
+      "step": 33770,
+      "train_speed(iter/s)": 1.318599
+    },
+    {
+      "acc": 0.7045166,
+      "epoch": 0.8567985794013191,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.570235068915053e-06,
+      "loss": 1.33371105,
+      "memory(GiB)": 113.67,
+      "step": 33775,
+      "train_speed(iter/s)": 1.31862
+    },
+    {
+      "acc": 0.6903882,
+      "epoch": 0.8569254185692542,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.569239463274122e-06,
+      "loss": 1.42346287,
+      "memory(GiB)": 113.67,
+      "step": 33780,
+      "train_speed(iter/s)": 1.318641
+    },
+    {
+      "acc": 0.67272425,
+      "epoch": 0.8570522577371893,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.568243788611033e-06,
+      "loss": 1.38568516,
+      "memory(GiB)": 113.67,
+      "step": 33785,
+      "train_speed(iter/s)": 1.318662
+    },
+    {
+      "acc": 0.69029512,
+      "epoch": 0.8571790969051243,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.56724804496958e-06,
+      "loss": 1.36461029,
+      "memory(GiB)": 113.67,
+      "step": 33790,
+      "train_speed(iter/s)": 1.318684
+    },
+    {
+      "acc": 0.68245697,
+      "epoch": 0.8573059360730594,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.566252232393561e-06,
+      "loss": 1.40890112,
+      "memory(GiB)": 113.67,
+      "step": 33795,
+      "train_speed(iter/s)": 1.318705
+    },
+    {
+      "acc": 0.69585562,
+      "epoch": 0.8574327752409944,
+      "grad_norm": 3.0,
+      "learning_rate": 6.565256350926777e-06,
+      "loss": 1.3654829,
+      "memory(GiB)": 113.67,
+      "step": 33800,
+      "train_speed(iter/s)": 1.318726
+    },
+    {
+      "acc": 0.6972434,
+      "epoch": 0.8575596144089295,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.5642604006130286e-06,
+      "loss": 1.32557001,
+      "memory(GiB)": 113.67,
+      "step": 33805,
+      "train_speed(iter/s)": 1.318748
+    },
+    {
+      "acc": 0.68045092,
+      "epoch": 0.8576864535768646,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.563264381496124e-06,
+      "loss": 1.48230743,
+      "memory(GiB)": 113.67,
+      "step": 33810,
+      "train_speed(iter/s)": 1.318768
+    },
+    {
+      "acc": 0.68167143,
+      "epoch": 0.8578132927447996,
+      "grad_norm": 6.0625,
+      "learning_rate": 6.562268293619872e-06,
+      "loss": 1.48268185,
+      "memory(GiB)": 113.67,
+      "step": 33815,
+      "train_speed(iter/s)": 1.31879
+    },
+    {
+      "acc": 0.68787975,
+      "epoch": 0.8579401319127347,
+      "grad_norm": 4.3125,
+      "learning_rate": 6.561272137028089e-06,
+      "loss": 1.39132118,
+      "memory(GiB)": 113.67,
+      "step": 33820,
+      "train_speed(iter/s)": 1.318812
+    },
+    {
+      "acc": 0.67824345,
+      "epoch": 0.8580669710806698,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.560275911764582e-06,
+      "loss": 1.38870583,
+      "memory(GiB)": 113.67,
+      "step": 33825,
+      "train_speed(iter/s)": 1.318831
+    },
+    {
+      "acc": 0.69565377,
+      "epoch": 0.8581938102486047,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.5592796178731776e-06,
+      "loss": 1.37074795,
+      "memory(GiB)": 113.67,
+      "step": 33830,
+      "train_speed(iter/s)": 1.318852
+    },
+    {
+      "acc": 0.66935625,
+      "epoch": 0.8583206494165398,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.5582832553976924e-06,
+      "loss": 1.39789381,
+      "memory(GiB)": 113.67,
+      "step": 33835,
+      "train_speed(iter/s)": 1.318873
+    },
+    {
+      "acc": 0.69511204,
+      "epoch": 0.8584474885844748,
+      "grad_norm": 2.59375,
+      "learning_rate": 6.557286824381955e-06,
+      "loss": 1.3687748,
+      "memory(GiB)": 113.67,
+      "step": 33840,
+      "train_speed(iter/s)": 1.318895
+    },
+    {
+      "acc": 0.691781,
+      "epoch": 0.8585743277524099,
+      "grad_norm": 3.953125,
+      "learning_rate": 6.556290324869786e-06,
+      "loss": 1.39062576,
+      "memory(GiB)": 113.67,
+      "step": 33845,
+      "train_speed(iter/s)": 1.318918
+    },
+    {
+      "acc": 0.70019331,
+      "epoch": 0.858701166920345,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.555293756905024e-06,
+      "loss": 1.36594591,
+      "memory(GiB)": 113.67,
+      "step": 33850,
+      "train_speed(iter/s)": 1.318939
+    },
+    {
+      "acc": 0.69259624,
+      "epoch": 0.85882800608828,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.554297120531497e-06,
+      "loss": 1.30500889,
+      "memory(GiB)": 113.67,
+      "step": 33855,
+      "train_speed(iter/s)": 1.31896
+    },
+    {
+      "acc": 0.67591057,
+      "epoch": 0.8589548452562151,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.553300415793042e-06,
+      "loss": 1.41865368,
+      "memory(GiB)": 113.67,
+      "step": 33860,
+      "train_speed(iter/s)": 1.318981
+    },
+    {
+      "acc": 0.67653017,
+      "epoch": 0.8590816844241502,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.552303642733502e-06,
+      "loss": 1.41645975,
+      "memory(GiB)": 113.67,
+      "step": 33865,
+      "train_speed(iter/s)": 1.319003
+    },
+    {
+      "acc": 0.68939381,
+      "epoch": 0.8592085235920852,
+      "grad_norm": 2.875,
+      "learning_rate": 6.551306801396715e-06,
+      "loss": 1.40631275,
+      "memory(GiB)": 113.67,
+      "step": 33870,
+      "train_speed(iter/s)": 1.319025
+    },
+    {
+      "acc": 0.68432446,
+      "epoch": 0.8593353627600203,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.550309891826531e-06,
+      "loss": 1.33798609,
+      "memory(GiB)": 113.67,
+      "step": 33875,
+      "train_speed(iter/s)": 1.319047
+    },
+    {
+      "acc": 0.68554344,
+      "epoch": 0.8594622019279553,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.5493129140667955e-06,
+      "loss": 1.35376177,
+      "memory(GiB)": 113.67,
+      "step": 33880,
+      "train_speed(iter/s)": 1.319068
+    },
+    {
+      "acc": 0.67273517,
+      "epoch": 0.8595890410958904,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.54831586816136e-06,
+      "loss": 1.43050346,
+      "memory(GiB)": 113.67,
+      "step": 33885,
+      "train_speed(iter/s)": 1.319091
+    },
+    {
+      "acc": 0.68605747,
+      "epoch": 0.8597158802638255,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.54731875415408e-06,
+      "loss": 1.38459492,
+      "memory(GiB)": 113.67,
+      "step": 33890,
+      "train_speed(iter/s)": 1.319112
+    },
+    {
+      "acc": 0.69519567,
+      "epoch": 0.8598427194317605,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.546321572088814e-06,
+      "loss": 1.36830883,
+      "memory(GiB)": 113.67,
+      "step": 33895,
+      "train_speed(iter/s)": 1.319134
+    },
+    {
+      "acc": 0.68534479,
+      "epoch": 0.8599695585996956,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.545324322009421e-06,
+      "loss": 1.42701826,
+      "memory(GiB)": 113.67,
+      "step": 33900,
+      "train_speed(iter/s)": 1.319156
+    },
+    {
+      "acc": 0.70193429,
+      "epoch": 0.8600963977676307,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.544327003959765e-06,
+      "loss": 1.35567694,
+      "memory(GiB)": 113.67,
+      "step": 33905,
+      "train_speed(iter/s)": 1.319178
+    },
+    {
+      "acc": 0.68918924,
+      "epoch": 0.8602232369355657,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.543329617983713e-06,
+      "loss": 1.38016453,
+      "memory(GiB)": 113.67,
+      "step": 33910,
+      "train_speed(iter/s)": 1.3192
+    },
+    {
+      "acc": 0.69553957,
+      "epoch": 0.8603500761035008,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.5423321641251316e-06,
+      "loss": 1.41416054,
+      "memory(GiB)": 113.67,
+      "step": 33915,
+      "train_speed(iter/s)": 1.319222
+    },
+    {
+      "acc": 0.68131065,
+      "epoch": 0.8604769152714358,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.541334642427898e-06,
+      "loss": 1.42980642,
+      "memory(GiB)": 113.67,
+      "step": 33920,
+      "train_speed(iter/s)": 1.319244
+    },
+    {
+      "acc": 0.68967004,
+      "epoch": 0.8606037544393709,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.540337052935884e-06,
+      "loss": 1.36514263,
+      "memory(GiB)": 113.67,
+      "step": 33925,
+      "train_speed(iter/s)": 1.319267
+    },
+    {
+      "acc": 0.6802268,
+      "epoch": 0.860730593607306,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.53933939569297e-06,
+      "loss": 1.40514889,
+      "memory(GiB)": 113.67,
+      "step": 33930,
+      "train_speed(iter/s)": 1.31929
+    },
+    {
+      "acc": 0.68405924,
+      "epoch": 0.860857432775241,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.538341670743037e-06,
+      "loss": 1.40963974,
+      "memory(GiB)": 113.67,
+      "step": 33935,
+      "train_speed(iter/s)": 1.319311
+    },
+    {
+      "acc": 0.69655185,
+      "epoch": 0.8609842719431761,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.537343878129969e-06,
+      "loss": 1.36760769,
+      "memory(GiB)": 113.67,
+      "step": 33940,
+      "train_speed(iter/s)": 1.319332
+    },
+    {
+      "acc": 0.68483372,
+      "epoch": 0.8611111111111112,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.5363460178976524e-06,
+      "loss": 1.40889463,
+      "memory(GiB)": 113.67,
+      "step": 33945,
+      "train_speed(iter/s)": 1.319352
+    },
+    {
+      "acc": 0.67621994,
+      "epoch": 0.8612379502790461,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.53534809008998e-06,
+      "loss": 1.43445759,
+      "memory(GiB)": 113.67,
+      "step": 33950,
+      "train_speed(iter/s)": 1.319373
+    },
+    {
+      "acc": 0.68206787,
+      "epoch": 0.8613647894469812,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.534350094750843e-06,
+      "loss": 1.40438776,
+      "memory(GiB)": 113.67,
+      "step": 33955,
+      "train_speed(iter/s)": 1.319395
+    },
+    {
+      "acc": 0.67890024,
+      "epoch": 0.8614916286149162,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.5333520319241385e-06,
+      "loss": 1.42473688,
+      "memory(GiB)": 113.67,
+      "step": 33960,
+      "train_speed(iter/s)": 1.319417
+    },
+    {
+      "acc": 0.67585135,
+      "epoch": 0.8616184677828513,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.532353901653765e-06,
+      "loss": 1.45640144,
+      "memory(GiB)": 113.67,
+      "step": 33965,
+      "train_speed(iter/s)": 1.319439
+    },
+    {
+      "acc": 0.67947998,
+      "epoch": 0.8617453069507864,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.531355703983627e-06,
+      "loss": 1.44008951,
+      "memory(GiB)": 113.67,
+      "step": 33970,
+      "train_speed(iter/s)": 1.31946
+    },
+    {
+      "acc": 0.69209685,
+      "epoch": 0.8618721461187214,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.530357438957626e-06,
+      "loss": 1.41877594,
+      "memory(GiB)": 113.67,
+      "step": 33975,
+      "train_speed(iter/s)": 1.319481
+    },
+    {
+      "acc": 0.69155817,
+      "epoch": 0.8619989852866565,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.529359106619675e-06,
+      "loss": 1.38415203,
+      "memory(GiB)": 113.67,
+      "step": 33980,
+      "train_speed(iter/s)": 1.319503
+    },
+    {
+      "acc": 0.67925138,
+      "epoch": 0.8621258244545916,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.528360707013681e-06,
+      "loss": 1.39387445,
+      "memory(GiB)": 113.67,
+      "step": 33985,
+      "train_speed(iter/s)": 1.319523
+    },
+    {
+      "acc": 0.69885187,
+      "epoch": 0.8622526636225266,
+      "grad_norm": 4.0,
+      "learning_rate": 6.52736224018356e-06,
+      "loss": 1.44503031,
+      "memory(GiB)": 113.67,
+      "step": 33990,
+      "train_speed(iter/s)": 1.319273
+    },
+    {
+      "acc": 0.68732672,
+      "epoch": 0.8623795027904617,
+      "grad_norm": 2.625,
+      "learning_rate": 6.526363706173227e-06,
+      "loss": 1.43756304,
+      "memory(GiB)": 113.67,
+      "step": 33995,
+      "train_speed(iter/s)": 1.319293
+    },
+    {
+      "acc": 0.70526009,
+      "epoch": 0.8625063419583967,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.525365105026605e-06,
+      "loss": 1.26859083,
+      "memory(GiB)": 113.67,
+      "step": 34000,
+      "train_speed(iter/s)": 1.319315
+    },
+    {
+      "epoch": 0.8625063419583967,
+      "eval_acc": 0.6747500819660149,
+      "eval_loss": 1.3576042652130127,
+      "eval_runtime": 69.4546,
+      "eval_samples_per_second": 91.715,
+      "eval_steps_per_second": 22.936,
+      "step": 34000
+    },
+    {
+      "acc": 0.70893817,
+      "epoch": 0.8626331811263318,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.524366436787615e-06,
+      "loss": 1.2856863,
+      "memory(GiB)": 113.67,
+      "step": 34005,
+      "train_speed(iter/s)": 1.315151
+    },
+    {
+      "acc": 0.68873148,
+      "epoch": 0.8627600202942669,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.523367701500183e-06,
+      "loss": 1.40052662,
+      "memory(GiB)": 113.67,
+      "step": 34010,
+      "train_speed(iter/s)": 1.31517
+    },
+    {
+      "acc": 0.68232741,
+      "epoch": 0.8628868594622019,
+      "grad_norm": 2.75,
+      "learning_rate": 6.5223688992082375e-06,
+      "loss": 1.37242317,
+      "memory(GiB)": 113.67,
+      "step": 34015,
+      "train_speed(iter/s)": 1.315191
+    },
+    {
+      "acc": 0.67714968,
+      "epoch": 0.863013698630137,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.521370029955713e-06,
+      "loss": 1.44579382,
+      "memory(GiB)": 113.67,
+      "step": 34020,
+      "train_speed(iter/s)": 1.315212
+    },
+    {
+      "acc": 0.67359843,
+      "epoch": 0.8631405377980721,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.520371093786541e-06,
+      "loss": 1.47718811,
+      "memory(GiB)": 113.67,
+      "step": 34025,
+      "train_speed(iter/s)": 1.315232
+    },
+    {
+      "acc": 0.70524702,
+      "epoch": 0.8632673769660071,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.51937209074466e-06,
+      "loss": 1.34327164,
+      "memory(GiB)": 113.67,
+      "step": 34030,
+      "train_speed(iter/s)": 1.315252
+    },
+    {
+      "acc": 0.68519506,
+      "epoch": 0.8633942161339422,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.51837302087401e-06,
+      "loss": 1.36024408,
+      "memory(GiB)": 113.67,
+      "step": 34035,
+      "train_speed(iter/s)": 1.315274
+    },
+    {
+      "acc": 0.68180375,
+      "epoch": 0.8635210553018772,
+      "grad_norm": 4.0,
+      "learning_rate": 6.517373884218539e-06,
+      "loss": 1.4634531,
+      "memory(GiB)": 113.67,
+      "step": 34040,
+      "train_speed(iter/s)": 1.315294
+    },
+    {
+      "acc": 0.68065929,
+      "epoch": 0.8636478944698123,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.5163746808221865e-06,
+      "loss": 1.42795448,
+      "memory(GiB)": 113.67,
+      "step": 34045,
+      "train_speed(iter/s)": 1.315315
+    },
+    {
+      "acc": 0.66823578,
+      "epoch": 0.8637747336377474,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.515375410728907e-06,
+      "loss": 1.38740196,
+      "memory(GiB)": 113.67,
+      "step": 34050,
+      "train_speed(iter/s)": 1.315337
+    },
+    {
+      "acc": 0.68427258,
+      "epoch": 0.8639015728056824,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.51437607398265e-06,
+      "loss": 1.40944042,
+      "memory(GiB)": 113.67,
+      "step": 34055,
+      "train_speed(iter/s)": 1.31506
+    },
+    {
+      "acc": 0.69804072,
+      "epoch": 0.8640284119736175,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.513376670627374e-06,
+      "loss": 1.33278427,
+      "memory(GiB)": 113.67,
+      "step": 34060,
+      "train_speed(iter/s)": 1.315081
+    },
+    {
+      "acc": 0.68094349,
+      "epoch": 0.8641552511415526,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.512377200707033e-06,
+      "loss": 1.42361612,
+      "memory(GiB)": 113.67,
+      "step": 34065,
+      "train_speed(iter/s)": 1.315104
+    },
+    {
+      "acc": 0.69392462,
+      "epoch": 0.8642820903094875,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.511377664265591e-06,
+      "loss": 1.37095785,
+      "memory(GiB)": 113.67,
+      "step": 34070,
+      "train_speed(iter/s)": 1.315126
+    },
+    {
+      "acc": 0.69613686,
+      "epoch": 0.8644089294774226,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.510378061347013e-06,
+      "loss": 1.40060787,
+      "memory(GiB)": 113.67,
+      "step": 34075,
+      "train_speed(iter/s)": 1.315147
+    },
+    {
+      "acc": 0.67792692,
+      "epoch": 0.8645357686453576,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.509378391995264e-06,
+      "loss": 1.45437841,
+      "memory(GiB)": 113.67,
+      "step": 34080,
+      "train_speed(iter/s)": 1.315169
+    },
+    {
+      "acc": 0.68229027,
+      "epoch": 0.8646626078132927,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.508378656254314e-06,
+      "loss": 1.47568607,
+      "memory(GiB)": 113.67,
+      "step": 34085,
+      "train_speed(iter/s)": 1.31519
+    },
+    {
+      "acc": 0.69964371,
+      "epoch": 0.8647894469812278,
+      "grad_norm": 3.0,
+      "learning_rate": 6.507378854168136e-06,
+      "loss": 1.34660702,
+      "memory(GiB)": 113.67,
+      "step": 34090,
+      "train_speed(iter/s)": 1.314913
+    },
+    {
+      "acc": 0.67719188,
+      "epoch": 0.8649162861491628,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.506378985780707e-06,
+      "loss": 1.39375515,
+      "memory(GiB)": 113.67,
+      "step": 34095,
+      "train_speed(iter/s)": 1.314935
+    },
+    {
+      "acc": 0.69345546,
+      "epoch": 0.8650431253170979,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.505379051136004e-06,
+      "loss": 1.37555618,
+      "memory(GiB)": 113.67,
+      "step": 34100,
+      "train_speed(iter/s)": 1.314957
+    },
+    {
+      "acc": 0.69212999,
+      "epoch": 0.865169964485033,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.504379050278009e-06,
+      "loss": 1.36267223,
+      "memory(GiB)": 113.67,
+      "step": 34105,
+      "train_speed(iter/s)": 1.314979
+    },
+    {
+      "acc": 0.67593961,
+      "epoch": 0.865296803652968,
+      "grad_norm": 3.25,
+      "learning_rate": 6.503378983250707e-06,
+      "loss": 1.49922829,
+      "memory(GiB)": 113.67,
+      "step": 34110,
+      "train_speed(iter/s)": 1.315001
+    },
+    {
+      "acc": 0.69011054,
+      "epoch": 0.8654236428209031,
+      "grad_norm": 3.984375,
+      "learning_rate": 6.5023788500980855e-06,
+      "loss": 1.38353882,
+      "memory(GiB)": 113.67,
+      "step": 34115,
+      "train_speed(iter/s)": 1.315023
+    },
+    {
+      "acc": 0.68404541,
+      "epoch": 0.8655504819888381,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.501378650864135e-06,
+      "loss": 1.41552811,
+      "memory(GiB)": 113.67,
+      "step": 34120,
+      "train_speed(iter/s)": 1.315045
+    },
+    {
+      "acc": 0.68323793,
+      "epoch": 0.8656773211567732,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.500378385592847e-06,
+      "loss": 1.40733862,
+      "memory(GiB)": 113.67,
+      "step": 34125,
+      "train_speed(iter/s)": 1.315068
+    },
+    {
+      "acc": 0.69325552,
+      "epoch": 0.8658041603247083,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.49937805432822e-06,
+      "loss": 1.37901506,
+      "memory(GiB)": 113.67,
+      "step": 34130,
+      "train_speed(iter/s)": 1.315089
+    },
+    {
+      "acc": 0.69067373,
+      "epoch": 0.8659309994926433,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.498377657114251e-06,
+      "loss": 1.3316246,
+      "memory(GiB)": 113.67,
+      "step": 34135,
+      "train_speed(iter/s)": 1.31511
+    },
+    {
+      "acc": 0.69644327,
+      "epoch": 0.8660578386605784,
+      "grad_norm": 3.125,
+      "learning_rate": 6.497377193994944e-06,
+      "loss": 1.34297142,
+      "memory(GiB)": 113.67,
+      "step": 34140,
+      "train_speed(iter/s)": 1.315132
+    },
+    {
+      "acc": 0.69474325,
+      "epoch": 0.8661846778285135,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.496376665014301e-06,
+      "loss": 1.38648329,
+      "memory(GiB)": 113.67,
+      "step": 34145,
+      "train_speed(iter/s)": 1.315155
+    },
+    {
+      "acc": 0.67692728,
+      "epoch": 0.8663115169964485,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.4953760702163325e-06,
+      "loss": 1.44983273,
+      "memory(GiB)": 113.67,
+      "step": 34150,
+      "train_speed(iter/s)": 1.315176
+    },
+    {
+      "acc": 0.6887351,
+      "epoch": 0.8664383561643836,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.494375409645049e-06,
+      "loss": 1.37403412,
+      "memory(GiB)": 113.67,
+      "step": 34155,
+      "train_speed(iter/s)": 1.315197
+    },
+    {
+      "acc": 0.68447561,
+      "epoch": 0.8665651953323186,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.493374683344462e-06,
+      "loss": 1.36452465,
+      "memory(GiB)": 113.67,
+      "step": 34160,
+      "train_speed(iter/s)": 1.315215
+    },
+    {
+      "acc": 0.69061146,
+      "epoch": 0.8666920345002537,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.492373891358589e-06,
+      "loss": 1.33251095,
+      "memory(GiB)": 113.67,
+      "step": 34165,
+      "train_speed(iter/s)": 1.315237
+    },
+    {
+      "acc": 0.68599854,
+      "epoch": 0.8668188736681888,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.4913730337314495e-06,
+      "loss": 1.40074787,
+      "memory(GiB)": 113.67,
+      "step": 34170,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.6827312,
+      "epoch": 0.8669457128361238,
+      "grad_norm": 2.59375,
+      "learning_rate": 6.490372110507066e-06,
+      "loss": 1.42557449,
+      "memory(GiB)": 113.67,
+      "step": 34175,
+      "train_speed(iter/s)": 1.315272
+    },
+    {
+      "acc": 0.69300718,
+      "epoch": 0.8670725520040589,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.489371121729462e-06,
+      "loss": 1.35646877,
+      "memory(GiB)": 113.67,
+      "step": 34180,
+      "train_speed(iter/s)": 1.315292
+    },
+    {
+      "acc": 0.68408122,
+      "epoch": 0.867199391171994,
+      "grad_norm": 3.375,
+      "learning_rate": 6.4883700674426666e-06,
+      "loss": 1.35548534,
+      "memory(GiB)": 113.67,
+      "step": 34185,
+      "train_speed(iter/s)": 1.315313
+    },
+    {
+      "acc": 0.68721189,
+      "epoch": 0.867326230339929,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.4873689476907105e-06,
+      "loss": 1.3829174,
+      "memory(GiB)": 113.67,
+      "step": 34190,
+      "train_speed(iter/s)": 1.315334
+    },
+    {
+      "acc": 0.68992391,
+      "epoch": 0.867453069507864,
+      "grad_norm": 3.75,
+      "learning_rate": 6.486367762517628e-06,
+      "loss": 1.41705532,
+      "memory(GiB)": 113.67,
+      "step": 34195,
+      "train_speed(iter/s)": 1.315356
+    },
+    {
+      "acc": 0.67922029,
+      "epoch": 0.867579908675799,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.4853665119674556e-06,
+      "loss": 1.44086933,
+      "memory(GiB)": 113.67,
+      "step": 34200,
+      "train_speed(iter/s)": 1.315378
+    },
+    {
+      "acc": 0.67580361,
+      "epoch": 0.8677067478437341,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.484365196084231e-06,
+      "loss": 1.35721645,
+      "memory(GiB)": 113.67,
+      "step": 34205,
+      "train_speed(iter/s)": 1.3154
+    },
+    {
+      "acc": 0.69274979,
+      "epoch": 0.8678335870116692,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.4833638149119985e-06,
+      "loss": 1.40775566,
+      "memory(GiB)": 113.67,
+      "step": 34210,
+      "train_speed(iter/s)": 1.315422
+    },
+    {
+      "acc": 0.69043036,
+      "epoch": 0.8679604261796042,
+      "grad_norm": 3.375,
+      "learning_rate": 6.4823623684948034e-06,
+      "loss": 1.42418118,
+      "memory(GiB)": 113.67,
+      "step": 34215,
+      "train_speed(iter/s)": 1.315445
+    },
+    {
+      "acc": 0.70025067,
+      "epoch": 0.8680872653475393,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.4813608568766924e-06,
+      "loss": 1.4127243,
+      "memory(GiB)": 113.67,
+      "step": 34220,
+      "train_speed(iter/s)": 1.315468
+    },
+    {
+      "acc": 0.67965145,
+      "epoch": 0.8682141045154744,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.480359280101717e-06,
+      "loss": 1.40252237,
+      "memory(GiB)": 113.67,
+      "step": 34225,
+      "train_speed(iter/s)": 1.31549
+    },
+    {
+      "acc": 0.6915338,
+      "epoch": 0.8683409436834094,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.479357638213931e-06,
+      "loss": 1.33462925,
+      "memory(GiB)": 113.67,
+      "step": 34230,
+      "train_speed(iter/s)": 1.315513
+    },
+    {
+      "acc": 0.69159489,
+      "epoch": 0.8684677828513445,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.478355931257392e-06,
+      "loss": 1.37519951,
+      "memory(GiB)": 113.67,
+      "step": 34235,
+      "train_speed(iter/s)": 1.315535
+    },
+    {
+      "acc": 0.70052948,
+      "epoch": 0.8685946220192795,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.477354159276158e-06,
+      "loss": 1.36497917,
+      "memory(GiB)": 113.67,
+      "step": 34240,
+      "train_speed(iter/s)": 1.315558
+    },
+    {
+      "acc": 0.67654448,
+      "epoch": 0.8687214611872146,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.476352322314292e-06,
+      "loss": 1.40501785,
+      "memory(GiB)": 113.67,
+      "step": 34245,
+      "train_speed(iter/s)": 1.31558
+    },
+    {
+      "acc": 0.6827466,
+      "epoch": 0.8688483003551497,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.47535042041586e-06,
+      "loss": 1.44719563,
+      "memory(GiB)": 113.67,
+      "step": 34250,
+      "train_speed(iter/s)": 1.315603
+    },
+    {
+      "acc": 0.65994983,
+      "epoch": 0.8689751395230847,
+      "grad_norm": 2.875,
+      "learning_rate": 6.474348453624929e-06,
+      "loss": 1.50043325,
+      "memory(GiB)": 113.67,
+      "step": 34255,
+      "train_speed(iter/s)": 1.315626
+    },
+    {
+      "acc": 0.69176812,
+      "epoch": 0.8691019786910198,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.473346421985571e-06,
+      "loss": 1.40467567,
+      "memory(GiB)": 113.67,
+      "step": 34260,
+      "train_speed(iter/s)": 1.315648
+    },
+    {
+      "acc": 0.68214445,
+      "epoch": 0.8692288178589549,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.472344325541859e-06,
+      "loss": 1.44166145,
+      "memory(GiB)": 113.67,
+      "step": 34265,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.68949108,
+      "epoch": 0.8693556570268899,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.4713421643378715e-06,
+      "loss": 1.43404274,
+      "memory(GiB)": 113.67,
+      "step": 34270,
+      "train_speed(iter/s)": 1.315694
+    },
+    {
+      "acc": 0.67855763,
+      "epoch": 0.869482496194825,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.470339938417685e-06,
+      "loss": 1.36959896,
+      "memory(GiB)": 113.67,
+      "step": 34275,
+      "train_speed(iter/s)": 1.315716
+    },
+    {
+      "acc": 0.70431213,
+      "epoch": 0.86960933536276,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.469337647825384e-06,
+      "loss": 1.34436092,
+      "memory(GiB)": 113.67,
+      "step": 34280,
+      "train_speed(iter/s)": 1.315739
+    },
+    {
+      "acc": 0.6815268,
+      "epoch": 0.8697361745306951,
+      "grad_norm": 5.5625,
+      "learning_rate": 6.468335292605053e-06,
+      "loss": 1.39791632,
+      "memory(GiB)": 113.67,
+      "step": 34285,
+      "train_speed(iter/s)": 1.315763
+    },
+    {
+      "acc": 0.69511356,
+      "epoch": 0.8698630136986302,
+      "grad_norm": 3.125,
+      "learning_rate": 6.467332872800779e-06,
+      "loss": 1.31505671,
+      "memory(GiB)": 113.67,
+      "step": 34290,
+      "train_speed(iter/s)": 1.315786
+    },
+    {
+      "acc": 0.68928428,
+      "epoch": 0.8699898528665652,
+      "grad_norm": 3.0,
+      "learning_rate": 6.466330388456655e-06,
+      "loss": 1.40397873,
+      "memory(GiB)": 113.67,
+      "step": 34295,
+      "train_speed(iter/s)": 1.315808
+    },
+    {
+      "acc": 0.67898836,
+      "epoch": 0.8701166920345003,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.465327839616774e-06,
+      "loss": 1.43317585,
+      "memory(GiB)": 113.67,
+      "step": 34300,
+      "train_speed(iter/s)": 1.315831
+    },
+    {
+      "acc": 0.69394913,
+      "epoch": 0.8702435312024354,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.464325226325232e-06,
+      "loss": 1.3828845,
+      "memory(GiB)": 113.67,
+      "step": 34305,
+      "train_speed(iter/s)": 1.315852
+    },
+    {
+      "acc": 0.67111626,
+      "epoch": 0.8703703703703703,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.46332254862613e-06,
+      "loss": 1.4575119,
+      "memory(GiB)": 113.67,
+      "step": 34310,
+      "train_speed(iter/s)": 1.315875
+    },
+    {
+      "acc": 0.68843136,
+      "epoch": 0.8704972095383054,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.462319806563568e-06,
+      "loss": 1.39201241,
+      "memory(GiB)": 113.67,
+      "step": 34315,
+      "train_speed(iter/s)": 1.315897
+    },
+    {
+      "acc": 0.68019433,
+      "epoch": 0.8706240487062404,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.461317000181653e-06,
+      "loss": 1.35604534,
+      "memory(GiB)": 113.67,
+      "step": 34320,
+      "train_speed(iter/s)": 1.315919
+    },
+    {
+      "acc": 0.68506284,
+      "epoch": 0.8707508878741755,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.460314129524491e-06,
+      "loss": 1.42105961,
+      "memory(GiB)": 113.67,
+      "step": 34325,
+      "train_speed(iter/s)": 1.315942
+    },
+    {
+      "acc": 0.6852272,
+      "epoch": 0.8708777270421106,
+      "grad_norm": 2.875,
+      "learning_rate": 6.4593111946361945e-06,
+      "loss": 1.40901814,
+      "memory(GiB)": 113.67,
+      "step": 34330,
+      "train_speed(iter/s)": 1.315964
+    },
+    {
+      "acc": 0.67142906,
+      "epoch": 0.8710045662100456,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.458308195560874e-06,
+      "loss": 1.38481655,
+      "memory(GiB)": 113.67,
+      "step": 34335,
+      "train_speed(iter/s)": 1.315987
+    },
+    {
+      "acc": 0.67751222,
+      "epoch": 0.8711314053779807,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.4573051323426515e-06,
+      "loss": 1.46171341,
+      "memory(GiB)": 113.67,
+      "step": 34340,
+      "train_speed(iter/s)": 1.31601
+    },
+    {
+      "acc": 0.68854747,
+      "epoch": 0.8712582445459158,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.456302005025641e-06,
+      "loss": 1.44087791,
+      "memory(GiB)": 113.67,
+      "step": 34345,
+      "train_speed(iter/s)": 1.316032
+    },
+    {
+      "acc": 0.66575928,
+      "epoch": 0.8713850837138508,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.4552988136539675e-06,
+      "loss": 1.45378742,
+      "memory(GiB)": 113.67,
+      "step": 34350,
+      "train_speed(iter/s)": 1.316055
+    },
+    {
+      "acc": 0.69793348,
+      "epoch": 0.8715119228817859,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.454295558271752e-06,
+      "loss": 1.35560169,
+      "memory(GiB)": 113.67,
+      "step": 34355,
+      "train_speed(iter/s)": 1.316077
+    },
+    {
+      "acc": 0.69374599,
+      "epoch": 0.8716387620497209,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.4532922389231275e-06,
+      "loss": 1.33680382,
+      "memory(GiB)": 113.67,
+      "step": 34360,
+      "train_speed(iter/s)": 1.3161
+    },
+    {
+      "acc": 0.69431949,
+      "epoch": 0.871765601217656,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.452288855652222e-06,
+      "loss": 1.35956821,
+      "memory(GiB)": 113.67,
+      "step": 34365,
+      "train_speed(iter/s)": 1.316122
+    },
+    {
+      "acc": 0.69423304,
+      "epoch": 0.8718924403855911,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.451285408503167e-06,
+      "loss": 1.33934193,
+      "memory(GiB)": 113.67,
+      "step": 34370,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.67530394,
+      "epoch": 0.8720192795535261,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.450281897520102e-06,
+      "loss": 1.41982489,
+      "memory(GiB)": 113.67,
+      "step": 34375,
+      "train_speed(iter/s)": 1.316163
+    },
+    {
+      "acc": 0.70021782,
+      "epoch": 0.8721461187214612,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.449278322747164e-06,
+      "loss": 1.36967278,
+      "memory(GiB)": 113.67,
+      "step": 34380,
+      "train_speed(iter/s)": 1.316185
+    },
+    {
+      "acc": 0.69694448,
+      "epoch": 0.8722729578893963,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.448274684228494e-06,
+      "loss": 1.39572897,
+      "memory(GiB)": 113.67,
+      "step": 34385,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.71164684,
+      "epoch": 0.8723997970573313,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.447270982008237e-06,
+      "loss": 1.30635109,
+      "memory(GiB)": 113.67,
+      "step": 34390,
+      "train_speed(iter/s)": 1.316229
+    },
+    {
+      "acc": 0.68004165,
+      "epoch": 0.8725266362252664,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.446267216130541e-06,
+      "loss": 1.42654829,
+      "memory(GiB)": 113.67,
+      "step": 34395,
+      "train_speed(iter/s)": 1.316252
+    },
+    {
+      "acc": 0.6889502,
+      "epoch": 0.8726534753932014,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.4452633866395555e-06,
+      "loss": 1.38545008,
+      "memory(GiB)": 113.67,
+      "step": 34400,
+      "train_speed(iter/s)": 1.316275
+    },
+    {
+      "acc": 0.69505897,
+      "epoch": 0.8727803145611365,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.444259493579433e-06,
+      "loss": 1.36428156,
+      "memory(GiB)": 113.67,
+      "step": 34405,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.69224353,
+      "epoch": 0.8729071537290716,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.443255536994331e-06,
+      "loss": 1.37495403,
+      "memory(GiB)": 113.67,
+      "step": 34410,
+      "train_speed(iter/s)": 1.31632
+    },
+    {
+      "acc": 0.69771414,
+      "epoch": 0.8730339928970066,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.442251516928406e-06,
+      "loss": 1.39175091,
+      "memory(GiB)": 113.67,
+      "step": 34415,
+      "train_speed(iter/s)": 1.316342
+    },
+    {
+      "acc": 0.69722199,
+      "epoch": 0.8731608320649417,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.441247433425821e-06,
+      "loss": 1.32058678,
+      "memory(GiB)": 113.67,
+      "step": 34420,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.6806838,
+      "epoch": 0.8732876712328768,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.4402432865307384e-06,
+      "loss": 1.43164873,
+      "memory(GiB)": 113.67,
+      "step": 34425,
+      "train_speed(iter/s)": 1.316387
+    },
+    {
+      "acc": 0.67633982,
+      "epoch": 0.8734145104008117,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.439239076287327e-06,
+      "loss": 1.44456968,
+      "memory(GiB)": 113.67,
+      "step": 34430,
+      "train_speed(iter/s)": 1.31641
+    },
+    {
+      "acc": 0.68725781,
+      "epoch": 0.8735413495687468,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.438234802739753e-06,
+      "loss": 1.38644714,
+      "memory(GiB)": 113.67,
+      "step": 34435,
+      "train_speed(iter/s)": 1.316433
+    },
+    {
+      "acc": 0.69221015,
+      "epoch": 0.8736681887366818,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.4372304659321935e-06,
+      "loss": 1.40749197,
+      "memory(GiB)": 113.67,
+      "step": 34440,
+      "train_speed(iter/s)": 1.316452
+    },
+    {
+      "acc": 0.68386393,
+      "epoch": 0.8737950279046169,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.43622606590882e-06,
+      "loss": 1.40839596,
+      "memory(GiB)": 113.67,
+      "step": 34445,
+      "train_speed(iter/s)": 1.316475
+    },
+    {
+      "acc": 0.67592387,
+      "epoch": 0.873921867072552,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.4352216027138125e-06,
+      "loss": 1.34474144,
+      "memory(GiB)": 113.67,
+      "step": 34450,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.67760034,
+      "epoch": 0.874048706240487,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.434217076391351e-06,
+      "loss": 1.41576691,
+      "memory(GiB)": 113.67,
+      "step": 34455,
+      "train_speed(iter/s)": 1.316518
+    },
+    {
+      "acc": 0.68423696,
+      "epoch": 0.8741755454084221,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.433212486985618e-06,
+      "loss": 1.383918,
+      "memory(GiB)": 113.67,
+      "step": 34460,
+      "train_speed(iter/s)": 1.316541
+    },
+    {
+      "acc": 0.68599072,
+      "epoch": 0.8743023845763572,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.432207834540802e-06,
+      "loss": 1.39841347,
+      "memory(GiB)": 113.67,
+      "step": 34465,
+      "train_speed(iter/s)": 1.316564
+    },
+    {
+      "acc": 0.6993185,
+      "epoch": 0.8744292237442922,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.431203119101093e-06,
+      "loss": 1.35994768,
+      "memory(GiB)": 113.67,
+      "step": 34470,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.67700624,
+      "epoch": 0.8745560629122273,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.430198340710677e-06,
+      "loss": 1.38205042,
+      "memory(GiB)": 113.67,
+      "step": 34475,
+      "train_speed(iter/s)": 1.316609
+    },
+    {
+      "acc": 0.70076494,
+      "epoch": 0.8746829020801623,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.4291934994137566e-06,
+      "loss": 1.35434694,
+      "memory(GiB)": 113.67,
+      "step": 34480,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.69368267,
+      "epoch": 0.8748097412480974,
+      "grad_norm": 3.625,
+      "learning_rate": 6.428188595254521e-06,
+      "loss": 1.38731403,
+      "memory(GiB)": 113.67,
+      "step": 34485,
+      "train_speed(iter/s)": 1.316653
+    },
+    {
+      "acc": 0.6931263,
+      "epoch": 0.8749365804160325,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.427183628277178e-06,
+      "loss": 1.40705414,
+      "memory(GiB)": 113.67,
+      "step": 34490,
+      "train_speed(iter/s)": 1.316675
+    },
+    {
+      "acc": 0.68329773,
+      "epoch": 0.8750634195839675,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.426178598525925e-06,
+      "loss": 1.42933998,
+      "memory(GiB)": 113.67,
+      "step": 34495,
+      "train_speed(iter/s)": 1.316698
+    },
+    {
+      "acc": 0.68663068,
+      "epoch": 0.8751902587519026,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.4251735060449725e-06,
+      "loss": 1.45366354,
+      "memory(GiB)": 113.67,
+      "step": 34500,
+      "train_speed(iter/s)": 1.316721
+    },
+    {
+      "acc": 0.66902308,
+      "epoch": 0.8753170979198377,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.424168350878524e-06,
+      "loss": 1.41066637,
+      "memory(GiB)": 113.67,
+      "step": 34505,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.70026245,
+      "epoch": 0.8754439370877727,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.423163133070792e-06,
+      "loss": 1.33276587,
+      "memory(GiB)": 113.67,
+      "step": 34510,
+      "train_speed(iter/s)": 1.316765
+    },
+    {
+      "acc": 0.69413919,
+      "epoch": 0.8755707762557078,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.422157852665993e-06,
+      "loss": 1.46157541,
+      "memory(GiB)": 113.67,
+      "step": 34515,
+      "train_speed(iter/s)": 1.316786
+    },
+    {
+      "acc": 0.69383159,
+      "epoch": 0.8756976154236428,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.421152509708342e-06,
+      "loss": 1.31797838,
+      "memory(GiB)": 113.67,
+      "step": 34520,
+      "train_speed(iter/s)": 1.316807
+    },
+    {
+      "acc": 0.69136114,
+      "epoch": 0.8758244545915779,
+      "grad_norm": 3.75,
+      "learning_rate": 6.4201471042420595e-06,
+      "loss": 1.36989479,
+      "memory(GiB)": 113.67,
+      "step": 34525,
+      "train_speed(iter/s)": 1.316829
+    },
+    {
+      "acc": 0.67995725,
+      "epoch": 0.875951293759513,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.419141636311366e-06,
+      "loss": 1.36891861,
+      "memory(GiB)": 113.67,
+      "step": 34530,
+      "train_speed(iter/s)": 1.316851
+    },
+    {
+      "acc": 0.6873487,
+      "epoch": 0.876078132927448,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.4181361059604875e-06,
+      "loss": 1.37862082,
+      "memory(GiB)": 113.67,
+      "step": 34535,
+      "train_speed(iter/s)": 1.316872
+    },
+    {
+      "acc": 0.67155352,
+      "epoch": 0.8762049720953831,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.4171305132336515e-06,
+      "loss": 1.42530708,
+      "memory(GiB)": 113.67,
+      "step": 34540,
+      "train_speed(iter/s)": 1.316894
+    },
+    {
+      "acc": 0.68056631,
+      "epoch": 0.8763318112633182,
+      "grad_norm": 2.75,
+      "learning_rate": 6.416124858175088e-06,
+      "loss": 1.36853237,
+      "memory(GiB)": 113.67,
+      "step": 34545,
+      "train_speed(iter/s)": 1.316915
+    },
+    {
+      "acc": 0.68383651,
+      "epoch": 0.8764586504312532,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.415119140829031e-06,
+      "loss": 1.4848175,
+      "memory(GiB)": 113.67,
+      "step": 34550,
+      "train_speed(iter/s)": 1.316937
+    },
+    {
+      "acc": 0.68852134,
+      "epoch": 0.8765854895991883,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.414113361239715e-06,
+      "loss": 1.36769848,
+      "memory(GiB)": 113.67,
+      "step": 34555,
+      "train_speed(iter/s)": 1.316959
+    },
+    {
+      "acc": 0.7023241,
+      "epoch": 0.8767123287671232,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.4131075194513825e-06,
+      "loss": 1.27995338,
+      "memory(GiB)": 113.67,
+      "step": 34560,
+      "train_speed(iter/s)": 1.31698
+    },
+    {
+      "acc": 0.70024509,
+      "epoch": 0.8768391679350583,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.41210161550827e-06,
+      "loss": 1.21517649,
+      "memory(GiB)": 113.67,
+      "step": 34565,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.68842101,
+      "epoch": 0.8769660071029934,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.411095649454626e-06,
+      "loss": 1.38100033,
+      "memory(GiB)": 113.67,
+      "step": 34570,
+      "train_speed(iter/s)": 1.317022
+    },
+    {
+      "acc": 0.6818223,
+      "epoch": 0.8770928462709284,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.410089621334693e-06,
+      "loss": 1.42983217,
+      "memory(GiB)": 113.67,
+      "step": 34575,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.68868494,
+      "epoch": 0.8772196854388635,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.4090835311927236e-06,
+      "loss": 1.33182068,
+      "memory(GiB)": 113.67,
+      "step": 34580,
+      "train_speed(iter/s)": 1.317063
+    },
+    {
+      "acc": 0.70277729,
+      "epoch": 0.8773465246067986,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.40807737907297e-06,
+      "loss": 1.38031702,
+      "memory(GiB)": 113.67,
+      "step": 34585,
+      "train_speed(iter/s)": 1.317085
+    },
+    {
+      "acc": 0.68446088,
+      "epoch": 0.8774733637747336,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.407071165019686e-06,
+      "loss": 1.41548767,
+      "memory(GiB)": 113.67,
+      "step": 34590,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.69518275,
+      "epoch": 0.8776002029426687,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.40606488907713e-06,
+      "loss": 1.36524677,
+      "memory(GiB)": 113.67,
+      "step": 34595,
+      "train_speed(iter/s)": 1.317126
+    },
+    {
+      "acc": 0.67341614,
+      "epoch": 0.8777270421106037,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.4050585512895624e-06,
+      "loss": 1.43582954,
+      "memory(GiB)": 113.67,
+      "step": 34600,
+      "train_speed(iter/s)": 1.317148
+    },
+    {
+      "acc": 0.69409566,
+      "epoch": 0.8778538812785388,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.4040521517012475e-06,
+      "loss": 1.33142414,
+      "memory(GiB)": 113.67,
+      "step": 34605,
+      "train_speed(iter/s)": 1.317169
+    },
+    {
+      "acc": 0.67640719,
+      "epoch": 0.8779807204464739,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.40304569035645e-06,
+      "loss": 1.42510834,
+      "memory(GiB)": 113.67,
+      "step": 34610,
+      "train_speed(iter/s)": 1.317188
+    },
+    {
+      "acc": 0.68452244,
+      "epoch": 0.8781075596144089,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.402039167299439e-06,
+      "loss": 1.40387497,
+      "memory(GiB)": 113.67,
+      "step": 34615,
+      "train_speed(iter/s)": 1.31721
+    },
+    {
+      "acc": 0.6769114,
+      "epoch": 0.878234398782344,
+      "grad_norm": 3.0,
+      "learning_rate": 6.401032582574485e-06,
+      "loss": 1.42469254,
+      "memory(GiB)": 113.67,
+      "step": 34620,
+      "train_speed(iter/s)": 1.31723
+    },
+    {
+      "acc": 0.6845974,
+      "epoch": 0.8783612379502791,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.400025936225862e-06,
+      "loss": 1.38127918,
+      "memory(GiB)": 113.67,
+      "step": 34625,
+      "train_speed(iter/s)": 1.317251
+    },
+    {
+      "acc": 0.66991262,
+      "epoch": 0.8784880771182141,
+      "grad_norm": 2.875,
+      "learning_rate": 6.399019228297851e-06,
+      "loss": 1.44828587,
+      "memory(GiB)": 113.67,
+      "step": 34630,
+      "train_speed(iter/s)": 1.317272
+    },
+    {
+      "acc": 0.6895422,
+      "epoch": 0.8786149162861492,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.398012458834724e-06,
+      "loss": 1.35437393,
+      "memory(GiB)": 113.67,
+      "step": 34635,
+      "train_speed(iter/s)": 1.317293
+    },
+    {
+      "acc": 0.67807541,
+      "epoch": 0.8787417554540842,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.397005627880771e-06,
+      "loss": 1.37481976,
+      "memory(GiB)": 113.67,
+      "step": 34640,
+      "train_speed(iter/s)": 1.317314
+    },
+    {
+      "acc": 0.68569589,
+      "epoch": 0.8788685946220193,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.395998735480271e-06,
+      "loss": 1.42707958,
+      "memory(GiB)": 113.67,
+      "step": 34645,
+      "train_speed(iter/s)": 1.317334
+    },
+    {
+      "acc": 0.68019032,
+      "epoch": 0.8789954337899544,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.394991781677516e-06,
+      "loss": 1.34514332,
+      "memory(GiB)": 113.67,
+      "step": 34650,
+      "train_speed(iter/s)": 1.317355
+    },
+    {
+      "acc": 0.68038206,
+      "epoch": 0.8791222729578894,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.393984766516792e-06,
+      "loss": 1.43390636,
+      "memory(GiB)": 113.67,
+      "step": 34655,
+      "train_speed(iter/s)": 1.317376
+    },
+    {
+      "acc": 0.67387304,
+      "epoch": 0.8792491121258245,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.392977690042395e-06,
+      "loss": 1.52381878,
+      "memory(GiB)": 113.67,
+      "step": 34660,
+      "train_speed(iter/s)": 1.317397
+    },
+    {
+      "acc": 0.6974843,
+      "epoch": 0.8793759512937596,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.3919705522986205e-06,
+      "loss": 1.36162806,
+      "memory(GiB)": 113.67,
+      "step": 34665,
+      "train_speed(iter/s)": 1.317417
+    },
+    {
+      "acc": 0.70729179,
+      "epoch": 0.8795027904616946,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.390963353329767e-06,
+      "loss": 1.32206459,
+      "memory(GiB)": 113.67,
+      "step": 34670,
+      "train_speed(iter/s)": 1.317438
+    },
+    {
+      "acc": 0.66907091,
+      "epoch": 0.8796296296296297,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.389956093180134e-06,
+      "loss": 1.4528616,
+      "memory(GiB)": 113.67,
+      "step": 34675,
+      "train_speed(iter/s)": 1.317458
+    },
+    {
+      "acc": 0.70076671,
+      "epoch": 0.8797564687975646,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.388948771894025e-06,
+      "loss": 1.36337414,
+      "memory(GiB)": 113.67,
+      "step": 34680,
+      "train_speed(iter/s)": 1.31748
+    },
+    {
+      "acc": 0.69688711,
+      "epoch": 0.8798833079654997,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.38794138951575e-06,
+      "loss": 1.37021427,
+      "memory(GiB)": 113.67,
+      "step": 34685,
+      "train_speed(iter/s)": 1.3175
+    },
+    {
+      "acc": 0.69901495,
+      "epoch": 0.8800101471334348,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.386933946089615e-06,
+      "loss": 1.38428383,
+      "memory(GiB)": 113.67,
+      "step": 34690,
+      "train_speed(iter/s)": 1.317521
+    },
+    {
+      "acc": 0.69490862,
+      "epoch": 0.8801369863013698,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.385926441659933e-06,
+      "loss": 1.39039345,
+      "memory(GiB)": 113.67,
+      "step": 34695,
+      "train_speed(iter/s)": 1.317542
+    },
+    {
+      "acc": 0.67740412,
+      "epoch": 0.8802638254693049,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.38491887627102e-06,
+      "loss": 1.43636751,
+      "memory(GiB)": 113.67,
+      "step": 34700,
+      "train_speed(iter/s)": 1.317558
+    },
+    {
+      "acc": 0.68394394,
+      "epoch": 0.88039066463724,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.383911249967188e-06,
+      "loss": 1.38846226,
+      "memory(GiB)": 113.67,
+      "step": 34705,
+      "train_speed(iter/s)": 1.31758
+    },
+    {
+      "acc": 0.6817028,
+      "epoch": 0.880517503805175,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.382903562792764e-06,
+      "loss": 1.40123396,
+      "memory(GiB)": 113.67,
+      "step": 34710,
+      "train_speed(iter/s)": 1.317599
+    },
+    {
+      "acc": 0.67240133,
+      "epoch": 0.8806443429731101,
+      "grad_norm": 3.5,
+      "learning_rate": 6.381895814792065e-06,
+      "loss": 1.4527668,
+      "memory(GiB)": 113.67,
+      "step": 34715,
+      "train_speed(iter/s)": 1.317619
+    },
+    {
+      "acc": 0.67974319,
+      "epoch": 0.8807711821410451,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.38088800600942e-06,
+      "loss": 1.39264555,
+      "memory(GiB)": 113.67,
+      "step": 34720,
+      "train_speed(iter/s)": 1.31764
+    },
+    {
+      "acc": 0.69267244,
+      "epoch": 0.8808980213089802,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.3798801364891535e-06,
+      "loss": 1.41883106,
+      "memory(GiB)": 113.67,
+      "step": 34725,
+      "train_speed(iter/s)": 1.317661
+    },
+    {
+      "acc": 0.67426405,
+      "epoch": 0.8810248604769153,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.378872206275599e-06,
+      "loss": 1.38320513,
+      "memory(GiB)": 113.67,
+      "step": 34730,
+      "train_speed(iter/s)": 1.31768
+    },
+    {
+      "acc": 0.69321556,
+      "epoch": 0.8811516996448503,
+      "grad_norm": 3.125,
+      "learning_rate": 6.377864215413088e-06,
+      "loss": 1.3409708,
+      "memory(GiB)": 113.67,
+      "step": 34735,
+      "train_speed(iter/s)": 1.3177
+    },
+    {
+      "acc": 0.69363337,
+      "epoch": 0.8812785388127854,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.376856163945957e-06,
+      "loss": 1.42416773,
+      "memory(GiB)": 113.67,
+      "step": 34740,
+      "train_speed(iter/s)": 1.31772
+    },
+    {
+      "acc": 0.68477182,
+      "epoch": 0.8814053779807205,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.375848051918546e-06,
+      "loss": 1.39356384,
+      "memory(GiB)": 113.67,
+      "step": 34745,
+      "train_speed(iter/s)": 1.317739
+    },
+    {
+      "acc": 0.68872404,
+      "epoch": 0.8815322171486555,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.374839879375194e-06,
+      "loss": 1.35173178,
+      "memory(GiB)": 113.67,
+      "step": 34750,
+      "train_speed(iter/s)": 1.317759
+    },
+    {
+      "acc": 0.67898293,
+      "epoch": 0.8816590563165906,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.373831646360245e-06,
+      "loss": 1.36238823,
+      "memory(GiB)": 113.67,
+      "step": 34755,
+      "train_speed(iter/s)": 1.31778
+    },
+    {
+      "acc": 0.68368635,
+      "epoch": 0.8817858954845256,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.372823352918048e-06,
+      "loss": 1.39463711,
+      "memory(GiB)": 113.67,
+      "step": 34760,
+      "train_speed(iter/s)": 1.317801
+    },
+    {
+      "acc": 0.68188229,
+      "epoch": 0.8819127346524607,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.371814999092951e-06,
+      "loss": 1.37535915,
+      "memory(GiB)": 113.67,
+      "step": 34765,
+      "train_speed(iter/s)": 1.31782
+    },
+    {
+      "acc": 0.68375859,
+      "epoch": 0.8820395738203958,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.370806584929305e-06,
+      "loss": 1.44454718,
+      "memory(GiB)": 113.67,
+      "step": 34770,
+      "train_speed(iter/s)": 1.31784
+    },
+    {
+      "acc": 0.68854475,
+      "epoch": 0.8821664129883308,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.369798110471463e-06,
+      "loss": 1.39910164,
+      "memory(GiB)": 113.67,
+      "step": 34775,
+      "train_speed(iter/s)": 1.31786
+    },
+    {
+      "acc": 0.68772182,
+      "epoch": 0.8822932521562659,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.368789575763787e-06,
+      "loss": 1.3475399,
+      "memory(GiB)": 113.67,
+      "step": 34780,
+      "train_speed(iter/s)": 1.317881
+    },
+    {
+      "acc": 0.68469396,
+      "epoch": 0.882420091324201,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.367780980850633e-06,
+      "loss": 1.3606287,
+      "memory(GiB)": 113.67,
+      "step": 34785,
+      "train_speed(iter/s)": 1.317899
+    },
+    {
+      "acc": 0.68875141,
+      "epoch": 0.882546930492136,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.366772325776367e-06,
+      "loss": 1.42583809,
+      "memory(GiB)": 113.67,
+      "step": 34790,
+      "train_speed(iter/s)": 1.31792
+    },
+    {
+      "acc": 0.68440428,
+      "epoch": 0.882673769660071,
+      "grad_norm": 3.125,
+      "learning_rate": 6.365763610585349e-06,
+      "loss": 1.34012337,
+      "memory(GiB)": 113.67,
+      "step": 34795,
+      "train_speed(iter/s)": 1.317941
+    },
+    {
+      "acc": 0.70198603,
+      "epoch": 0.882800608828006,
+      "grad_norm": 3.625,
+      "learning_rate": 6.3647548353219515e-06,
+      "loss": 1.38183098,
+      "memory(GiB)": 113.67,
+      "step": 34800,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.69478917,
+      "epoch": 0.8829274479959411,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.363746000030543e-06,
+      "loss": 1.40460873,
+      "memory(GiB)": 113.67,
+      "step": 34805,
+      "train_speed(iter/s)": 1.317981
+    },
+    {
+      "acc": 0.68392811,
+      "epoch": 0.8830542871638762,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.362737104755497e-06,
+      "loss": 1.40432472,
+      "memory(GiB)": 113.67,
+      "step": 34810,
+      "train_speed(iter/s)": 1.318002
+    },
+    {
+      "acc": 0.69302144,
+      "epoch": 0.8831811263318112,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.361728149541188e-06,
+      "loss": 1.4025198,
+      "memory(GiB)": 113.67,
+      "step": 34815,
+      "train_speed(iter/s)": 1.318022
+    },
+    {
+      "acc": 0.68893318,
+      "epoch": 0.8833079654997463,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.360719134431995e-06,
+      "loss": 1.41094637,
+      "memory(GiB)": 113.67,
+      "step": 34820,
+      "train_speed(iter/s)": 1.318044
+    },
+    {
+      "acc": 0.70056915,
+      "epoch": 0.8834348046676814,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.359710059472299e-06,
+      "loss": 1.37751846,
+      "memory(GiB)": 113.67,
+      "step": 34825,
+      "train_speed(iter/s)": 1.318064
+    },
+    {
+      "acc": 0.68900933,
+      "epoch": 0.8835616438356164,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.358700924706486e-06,
+      "loss": 1.38692226,
+      "memory(GiB)": 113.67,
+      "step": 34830,
+      "train_speed(iter/s)": 1.318085
+    },
+    {
+      "acc": 0.69241266,
+      "epoch": 0.8836884830035515,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.357691730178939e-06,
+      "loss": 1.40150042,
+      "memory(GiB)": 113.67,
+      "step": 34835,
+      "train_speed(iter/s)": 1.318106
+    },
+    {
+      "acc": 0.67306385,
+      "epoch": 0.8838153221714865,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.356682475934048e-06,
+      "loss": 1.46477261,
+      "memory(GiB)": 113.67,
+      "step": 34840,
+      "train_speed(iter/s)": 1.318127
+    },
+    {
+      "acc": 0.68404894,
+      "epoch": 0.8839421613394216,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.3556731620162036e-06,
+      "loss": 1.38949776,
+      "memory(GiB)": 113.67,
+      "step": 34845,
+      "train_speed(iter/s)": 1.318148
+    },
+    {
+      "acc": 0.70094738,
+      "epoch": 0.8840690005073567,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.354663788469803e-06,
+      "loss": 1.35821657,
+      "memory(GiB)": 113.67,
+      "step": 34850,
+      "train_speed(iter/s)": 1.318168
+    },
+    {
+      "acc": 0.68340969,
+      "epoch": 0.8841958396752917,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.353654355339238e-06,
+      "loss": 1.42499638,
+      "memory(GiB)": 113.67,
+      "step": 34855,
+      "train_speed(iter/s)": 1.318189
+    },
+    {
+      "acc": 0.69596176,
+      "epoch": 0.8843226788432268,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.352644862668914e-06,
+      "loss": 1.36527576,
+      "memory(GiB)": 113.67,
+      "step": 34860,
+      "train_speed(iter/s)": 1.31821
+    },
+    {
+      "acc": 0.700804,
+      "epoch": 0.8844495180111619,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.351635310503228e-06,
+      "loss": 1.35908833,
+      "memory(GiB)": 113.67,
+      "step": 34865,
+      "train_speed(iter/s)": 1.31823
+    },
+    {
+      "acc": 0.69329367,
+      "epoch": 0.8845763571790969,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.3506256988865865e-06,
+      "loss": 1.37777262,
+      "memory(GiB)": 113.67,
+      "step": 34870,
+      "train_speed(iter/s)": 1.318251
+    },
+    {
+      "acc": 0.70780978,
+      "epoch": 0.884703196347032,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.349616027863397e-06,
+      "loss": 1.28652372,
+      "memory(GiB)": 113.67,
+      "step": 34875,
+      "train_speed(iter/s)": 1.318272
+    },
+    {
+      "acc": 0.68395085,
+      "epoch": 0.884830035514967,
+      "grad_norm": 2.875,
+      "learning_rate": 6.34860629747807e-06,
+      "loss": 1.40028553,
+      "memory(GiB)": 113.67,
+      "step": 34880,
+      "train_speed(iter/s)": 1.318293
+    },
+    {
+      "acc": 0.67643719,
+      "epoch": 0.8849568746829021,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.347596507775016e-06,
+      "loss": 1.39289398,
+      "memory(GiB)": 113.67,
+      "step": 34885,
+      "train_speed(iter/s)": 1.318314
+    },
+    {
+      "acc": 0.69285297,
+      "epoch": 0.8850837138508372,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.3465866587986505e-06,
+      "loss": 1.31000862,
+      "memory(GiB)": 113.67,
+      "step": 34890,
+      "train_speed(iter/s)": 1.318333
+    },
+    {
+      "acc": 0.70440569,
+      "epoch": 0.8852105530187722,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.345576750593392e-06,
+      "loss": 1.34079361,
+      "memory(GiB)": 113.67,
+      "step": 34895,
+      "train_speed(iter/s)": 1.318352
+    },
+    {
+      "acc": 0.66838031,
+      "epoch": 0.8853373921867073,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.34456678320366e-06,
+      "loss": 1.46124096,
+      "memory(GiB)": 113.67,
+      "step": 34900,
+      "train_speed(iter/s)": 1.318373
+    },
+    {
+      "acc": 0.68000302,
+      "epoch": 0.8854642313546424,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.343556756673879e-06,
+      "loss": 1.44185753,
+      "memory(GiB)": 113.67,
+      "step": 34905,
+      "train_speed(iter/s)": 1.318394
+    },
+    {
+      "acc": 0.68355989,
+      "epoch": 0.8855910705225774,
+      "grad_norm": 2.59375,
+      "learning_rate": 6.3425466710484726e-06,
+      "loss": 1.38334742,
+      "memory(GiB)": 113.67,
+      "step": 34910,
+      "train_speed(iter/s)": 1.318415
+    },
+    {
+      "acc": 0.68149295,
+      "epoch": 0.8857179096905125,
+      "grad_norm": 3.25,
+      "learning_rate": 6.3415365263718686e-06,
+      "loss": 1.439394,
+      "memory(GiB)": 113.67,
+      "step": 34915,
+      "train_speed(iter/s)": 1.318434
+    },
+    {
+      "acc": 0.68039789,
+      "epoch": 0.8858447488584474,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.340526322688501e-06,
+      "loss": 1.3728982,
+      "memory(GiB)": 113.67,
+      "step": 34920,
+      "train_speed(iter/s)": 1.318455
+    },
+    {
+      "acc": 0.68775916,
+      "epoch": 0.8859715880263825,
+      "grad_norm": 3.375,
+      "learning_rate": 6.339516060042798e-06,
+      "loss": 1.42021675,
+      "memory(GiB)": 113.67,
+      "step": 34925,
+      "train_speed(iter/s)": 1.318475
+    },
+    {
+      "acc": 0.67272568,
+      "epoch": 0.8860984271943176,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.3385057384792e-06,
+      "loss": 1.4326951,
+      "memory(GiB)": 113.67,
+      "step": 34930,
+      "train_speed(iter/s)": 1.318495
+    },
+    {
+      "acc": 0.66983652,
+      "epoch": 0.8862252663622526,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.337495358042143e-06,
+      "loss": 1.43800545,
+      "memory(GiB)": 113.67,
+      "step": 34935,
+      "train_speed(iter/s)": 1.318515
+    },
+    {
+      "acc": 0.6853683,
+      "epoch": 0.8863521055301877,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.336484918776069e-06,
+      "loss": 1.42331867,
+      "memory(GiB)": 113.67,
+      "step": 34940,
+      "train_speed(iter/s)": 1.318536
+    },
+    {
+      "acc": 0.69639702,
+      "epoch": 0.8864789446981228,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.335474420725421e-06,
+      "loss": 1.38142509,
+      "memory(GiB)": 113.67,
+      "step": 34945,
+      "train_speed(iter/s)": 1.318556
+    },
+    {
+      "acc": 0.67251558,
+      "epoch": 0.8866057838660578,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.334463863934646e-06,
+      "loss": 1.43056755,
+      "memory(GiB)": 113.67,
+      "step": 34950,
+      "train_speed(iter/s)": 1.318578
+    },
+    {
+      "acc": 0.68310585,
+      "epoch": 0.8867326230339929,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.333453248448192e-06,
+      "loss": 1.35675526,
+      "memory(GiB)": 113.67,
+      "step": 34955,
+      "train_speed(iter/s)": 1.318598
+    },
+    {
+      "acc": 0.67221432,
+      "epoch": 0.8868594622019279,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.33244257431051e-06,
+      "loss": 1.5295639,
+      "memory(GiB)": 113.67,
+      "step": 34960,
+      "train_speed(iter/s)": 1.318619
+    },
+    {
+      "acc": 0.67971306,
+      "epoch": 0.886986301369863,
+      "grad_norm": 2.875,
+      "learning_rate": 6.331431841566056e-06,
+      "loss": 1.42841606,
+      "memory(GiB)": 113.67,
+      "step": 34965,
+      "train_speed(iter/s)": 1.31864
+    },
+    {
+      "acc": 0.69410524,
+      "epoch": 0.8871131405377981,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.330421050259283e-06,
+      "loss": 1.40052547,
+      "memory(GiB)": 113.67,
+      "step": 34970,
+      "train_speed(iter/s)": 1.318662
+    },
+    {
+      "acc": 0.69142132,
+      "epoch": 0.8872399797057331,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.329410200434655e-06,
+      "loss": 1.34964628,
+      "memory(GiB)": 113.67,
+      "step": 34975,
+      "train_speed(iter/s)": 1.318682
+    },
+    {
+      "acc": 0.681814,
+      "epoch": 0.8873668188736682,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.328399292136629e-06,
+      "loss": 1.43471613,
+      "memory(GiB)": 113.67,
+      "step": 34980,
+      "train_speed(iter/s)": 1.318702
+    },
+    {
+      "acc": 0.68623743,
+      "epoch": 0.8874936580416033,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.327388325409672e-06,
+      "loss": 1.43655291,
+      "memory(GiB)": 113.67,
+      "step": 34985,
+      "train_speed(iter/s)": 1.318723
+    },
+    {
+      "acc": 0.67141514,
+      "epoch": 0.8876204972095383,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.326377300298251e-06,
+      "loss": 1.42496996,
+      "memory(GiB)": 113.67,
+      "step": 34990,
+      "train_speed(iter/s)": 1.318744
+    },
+    {
+      "acc": 0.69527736,
+      "epoch": 0.8877473363774734,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.325366216846832e-06,
+      "loss": 1.37230072,
+      "memory(GiB)": 113.67,
+      "step": 34995,
+      "train_speed(iter/s)": 1.318764
+    },
+    {
+      "acc": 0.70066357,
+      "epoch": 0.8878741755454084,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.324355075099893e-06,
+      "loss": 1.43615332,
+      "memory(GiB)": 113.67,
+      "step": 35000,
+      "train_speed(iter/s)": 1.318785
+    },
+    {
+      "epoch": 0.8878741755454084,
+      "eval_acc": 0.6748160724391624,
+      "eval_loss": 1.3577944040298462,
+      "eval_runtime": 69.2235,
+      "eval_samples_per_second": 92.021,
+      "eval_steps_per_second": 23.012,
+      "step": 35000
+    },
+    {
+      "acc": 0.68538094,
+      "epoch": 0.8880010147133435,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.3233438751019016e-06,
+      "loss": 1.41558743,
+      "memory(GiB)": 113.67,
+      "step": 35005,
+      "train_speed(iter/s)": 1.314746
+    },
+    {
+      "acc": 0.69535913,
+      "epoch": 0.8881278538812786,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.322332616897341e-06,
+      "loss": 1.34956846,
+      "memory(GiB)": 113.67,
+      "step": 35010,
+      "train_speed(iter/s)": 1.314768
+    },
+    {
+      "acc": 0.68401718,
+      "epoch": 0.8882546930492136,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.321321300530685e-06,
+      "loss": 1.31306372,
+      "memory(GiB)": 113.67,
+      "step": 35015,
+      "train_speed(iter/s)": 1.314792
+    },
+    {
+      "acc": 0.70806408,
+      "epoch": 0.8883815322171487,
+      "grad_norm": 3.75,
+      "learning_rate": 6.320309926046421e-06,
+      "loss": 1.36048985,
+      "memory(GiB)": 113.67,
+      "step": 35020,
+      "train_speed(iter/s)": 1.314814
+    },
+    {
+      "acc": 0.67075548,
+      "epoch": 0.8885083713850838,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.319298493489032e-06,
+      "loss": 1.47395592,
+      "memory(GiB)": 113.67,
+      "step": 35025,
+      "train_speed(iter/s)": 1.314837
+    },
+    {
+      "acc": 0.6787303,
+      "epoch": 0.8886352105530188,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.318287002903004e-06,
+      "loss": 1.3815155,
+      "memory(GiB)": 113.67,
+      "step": 35030,
+      "train_speed(iter/s)": 1.314859
+    },
+    {
+      "acc": 0.67098894,
+      "epoch": 0.8887620497209539,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.317275454332829e-06,
+      "loss": 1.424333,
+      "memory(GiB)": 113.67,
+      "step": 35035,
+      "train_speed(iter/s)": 1.314882
+    },
+    {
+      "acc": 0.6857234,
+      "epoch": 0.8888888888888888,
+      "grad_norm": 4.96875,
+      "learning_rate": 6.3162638478229965e-06,
+      "loss": 1.42110987,
+      "memory(GiB)": 113.67,
+      "step": 35040,
+      "train_speed(iter/s)": 1.314904
+    },
+    {
+      "acc": 0.682796,
+      "epoch": 0.8890157280568239,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.315252183418005e-06,
+      "loss": 1.39080105,
+      "memory(GiB)": 113.67,
+      "step": 35045,
+      "train_speed(iter/s)": 1.314926
+    },
+    {
+      "acc": 0.67477794,
+      "epoch": 0.889142567224759,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.31424046116235e-06,
+      "loss": 1.50131111,
+      "memory(GiB)": 113.67,
+      "step": 35050,
+      "train_speed(iter/s)": 1.31494
+    },
+    {
+      "acc": 0.67725859,
+      "epoch": 0.889269406392694,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.313228681100532e-06,
+      "loss": 1.41309814,
+      "memory(GiB)": 113.67,
+      "step": 35055,
+      "train_speed(iter/s)": 1.314961
+    },
+    {
+      "acc": 0.7005528,
+      "epoch": 0.8893962455606291,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.312216843277052e-06,
+      "loss": 1.36824665,
+      "memory(GiB)": 113.67,
+      "step": 35060,
+      "train_speed(iter/s)": 1.314984
+    },
+    {
+      "acc": 0.65574293,
+      "epoch": 0.8895230847285642,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.3112049477364165e-06,
+      "loss": 1.52109222,
+      "memory(GiB)": 113.67,
+      "step": 35065,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.68689198,
+      "epoch": 0.8896499238964992,
+      "grad_norm": 3.75,
+      "learning_rate": 6.310192994523137e-06,
+      "loss": 1.42557392,
+      "memory(GiB)": 113.67,
+      "step": 35070,
+      "train_speed(iter/s)": 1.315029
+    },
+    {
+      "acc": 0.67947888,
+      "epoch": 0.8897767630644343,
+      "grad_norm": 2.875,
+      "learning_rate": 6.309180983681716e-06,
+      "loss": 1.42054911,
+      "memory(GiB)": 113.67,
+      "step": 35075,
+      "train_speed(iter/s)": 1.31505
+    },
+    {
+      "acc": 0.67723789,
+      "epoch": 0.8899036022323693,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.308168915256671e-06,
+      "loss": 1.45486288,
+      "memory(GiB)": 113.67,
+      "step": 35080,
+      "train_speed(iter/s)": 1.315071
+    },
+    {
+      "acc": 0.68086119,
+      "epoch": 0.8900304414003044,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.307156789292518e-06,
+      "loss": 1.45126696,
+      "memory(GiB)": 113.67,
+      "step": 35085,
+      "train_speed(iter/s)": 1.315093
+    },
+    {
+      "acc": 0.69270358,
+      "epoch": 0.8901572805682395,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.306144605833773e-06,
+      "loss": 1.3801445,
+      "memory(GiB)": 113.67,
+      "step": 35090,
+      "train_speed(iter/s)": 1.315115
+    },
+    {
+      "acc": 0.67178669,
+      "epoch": 0.8902841197361745,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.305132364924955e-06,
+      "loss": 1.40880833,
+      "memory(GiB)": 113.67,
+      "step": 35095,
+      "train_speed(iter/s)": 1.315137
+    },
+    {
+      "acc": 0.67872963,
+      "epoch": 0.8904109589041096,
+      "grad_norm": 3.625,
+      "learning_rate": 6.3041200666105905e-06,
+      "loss": 1.37831383,
+      "memory(GiB)": 113.67,
+      "step": 35100,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.68451848,
+      "epoch": 0.8905377980720447,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.303107710935202e-06,
+      "loss": 1.40921345,
+      "memory(GiB)": 113.67,
+      "step": 35105,
+      "train_speed(iter/s)": 1.31518
+    },
+    {
+      "acc": 0.66582627,
+      "epoch": 0.8906646372399797,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.302095297943319e-06,
+      "loss": 1.38381357,
+      "memory(GiB)": 113.67,
+      "step": 35110,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.6757473,
+      "epoch": 0.8907914764079148,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.301082827679472e-06,
+      "loss": 1.41087074,
+      "memory(GiB)": 113.67,
+      "step": 35115,
+      "train_speed(iter/s)": 1.315223
+    },
+    {
+      "acc": 0.7030921,
+      "epoch": 0.8909183155758498,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.300070300188192e-06,
+      "loss": 1.35085239,
+      "memory(GiB)": 113.67,
+      "step": 35120,
+      "train_speed(iter/s)": 1.315245
+    },
+    {
+      "acc": 0.69411645,
+      "epoch": 0.8910451547437849,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.2990577155140164e-06,
+      "loss": 1.40712547,
+      "memory(GiB)": 113.67,
+      "step": 35125,
+      "train_speed(iter/s)": 1.315267
+    },
+    {
+      "acc": 0.68764753,
+      "epoch": 0.89117199391172,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.298045073701483e-06,
+      "loss": 1.37800217,
+      "memory(GiB)": 113.67,
+      "step": 35130,
+      "train_speed(iter/s)": 1.315289
+    },
+    {
+      "acc": 0.68638248,
+      "epoch": 0.891298833079655,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.29703237479513e-06,
+      "loss": 1.37028465,
+      "memory(GiB)": 113.67,
+      "step": 35135,
+      "train_speed(iter/s)": 1.315312
+    },
+    {
+      "acc": 0.68162632,
+      "epoch": 0.8914256722475901,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.296019618839505e-06,
+      "loss": 1.43304462,
+      "memory(GiB)": 113.67,
+      "step": 35140,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.70085936,
+      "epoch": 0.8915525114155252,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.295006805879149e-06,
+      "loss": 1.34768152,
+      "memory(GiB)": 113.67,
+      "step": 35145,
+      "train_speed(iter/s)": 1.315356
+    },
+    {
+      "acc": 0.68489432,
+      "epoch": 0.8916793505834602,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.293993935958613e-06,
+      "loss": 1.35110655,
+      "memory(GiB)": 113.67,
+      "step": 35150,
+      "train_speed(iter/s)": 1.315379
+    },
+    {
+      "acc": 0.69087954,
+      "epoch": 0.8918061897513953,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.292981009122445e-06,
+      "loss": 1.36649256,
+      "memory(GiB)": 113.67,
+      "step": 35155,
+      "train_speed(iter/s)": 1.315401
+    },
+    {
+      "acc": 0.68245997,
+      "epoch": 0.8919330289193302,
+      "grad_norm": 3.5,
+      "learning_rate": 6.291968025415202e-06,
+      "loss": 1.42170277,
+      "memory(GiB)": 113.67,
+      "step": 35160,
+      "train_speed(iter/s)": 1.315422
+    },
+    {
+      "acc": 0.68501816,
+      "epoch": 0.8920598680872653,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.290954984881434e-06,
+      "loss": 1.37417011,
+      "memory(GiB)": 113.67,
+      "step": 35165,
+      "train_speed(iter/s)": 1.315444
+    },
+    {
+      "acc": 0.68759766,
+      "epoch": 0.8921867072552004,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.289941887565703e-06,
+      "loss": 1.42582598,
+      "memory(GiB)": 113.67,
+      "step": 35170,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.70108109,
+      "epoch": 0.8923135464231354,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.288928733512569e-06,
+      "loss": 1.36183691,
+      "memory(GiB)": 113.67,
+      "step": 35175,
+      "train_speed(iter/s)": 1.315491
+    },
+    {
+      "acc": 0.69545059,
+      "epoch": 0.8924403855910705,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.287915522766596e-06,
+      "loss": 1.30015068,
+      "memory(GiB)": 113.67,
+      "step": 35180,
+      "train_speed(iter/s)": 1.315511
+    },
+    {
+      "acc": 0.68397856,
+      "epoch": 0.8925672247590056,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.2869022553723465e-06,
+      "loss": 1.36835718,
+      "memory(GiB)": 113.67,
+      "step": 35185,
+      "train_speed(iter/s)": 1.315534
+    },
+    {
+      "acc": 0.6944313,
+      "epoch": 0.8926940639269406,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.285888931374391e-06,
+      "loss": 1.36185894,
+      "memory(GiB)": 113.67,
+      "step": 35190,
+      "train_speed(iter/s)": 1.315556
+    },
+    {
+      "acc": 0.68508139,
+      "epoch": 0.8928209030948757,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.284875550817299e-06,
+      "loss": 1.34005938,
+      "memory(GiB)": 113.67,
+      "step": 35195,
+      "train_speed(iter/s)": 1.315579
+    },
+    {
+      "acc": 0.68270373,
+      "epoch": 0.8929477422628107,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.2838621137456425e-06,
+      "loss": 1.40332184,
+      "memory(GiB)": 113.67,
+      "step": 35200,
+      "train_speed(iter/s)": 1.315601
+    },
+    {
+      "acc": 0.68260689,
+      "epoch": 0.8930745814307458,
+      "grad_norm": 2.484375,
+      "learning_rate": 6.282848620203999e-06,
+      "loss": 1.37947779,
+      "memory(GiB)": 113.67,
+      "step": 35205,
+      "train_speed(iter/s)": 1.315623
+    },
+    {
+      "acc": 0.68902636,
+      "epoch": 0.8932014205986809,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.2818350702369466e-06,
+      "loss": 1.38054399,
+      "memory(GiB)": 113.67,
+      "step": 35210,
+      "train_speed(iter/s)": 1.315645
+    },
+    {
+      "acc": 0.69046555,
+      "epoch": 0.8933282597666159,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.280821463889063e-06,
+      "loss": 1.40551376,
+      "memory(GiB)": 113.67,
+      "step": 35215,
+      "train_speed(iter/s)": 1.315668
+    },
+    {
+      "acc": 0.69584675,
+      "epoch": 0.893455098934551,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.279807801204936e-06,
+      "loss": 1.38722458,
+      "memory(GiB)": 113.67,
+      "step": 35220,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.70346727,
+      "epoch": 0.8935819381024861,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.278794082229145e-06,
+      "loss": 1.36866121,
+      "memory(GiB)": 113.67,
+      "step": 35225,
+      "train_speed(iter/s)": 1.315712
+    },
+    {
+      "acc": 0.69253278,
+      "epoch": 0.8937087772704211,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.2777803070062825e-06,
+      "loss": 1.39229631,
+      "memory(GiB)": 113.67,
+      "step": 35230,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.68904142,
+      "epoch": 0.8938356164383562,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.276766475580935e-06,
+      "loss": 1.35242815,
+      "memory(GiB)": 113.67,
+      "step": 35235,
+      "train_speed(iter/s)": 1.315756
+    },
+    {
+      "acc": 0.66891193,
+      "epoch": 0.8939624556062912,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.2757525879977e-06,
+      "loss": 1.49567871,
+      "memory(GiB)": 113.67,
+      "step": 35240,
+      "train_speed(iter/s)": 1.315779
+    },
+    {
+      "acc": 0.6868021,
+      "epoch": 0.8940892947742263,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.27473864430117e-06,
+      "loss": 1.43852997,
+      "memory(GiB)": 113.67,
+      "step": 35245,
+      "train_speed(iter/s)": 1.3158
+    },
+    {
+      "acc": 0.6724659,
+      "epoch": 0.8942161339421614,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.273724644535942e-06,
+      "loss": 1.43040876,
+      "memory(GiB)": 113.67,
+      "step": 35250,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.69988565,
+      "epoch": 0.8943429731100964,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.272710588746619e-06,
+      "loss": 1.34654655,
+      "memory(GiB)": 113.67,
+      "step": 35255,
+      "train_speed(iter/s)": 1.315841
+    },
+    {
+      "acc": 0.67999926,
+      "epoch": 0.8944698122780315,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.271696476977801e-06,
+      "loss": 1.44577303,
+      "memory(GiB)": 113.67,
+      "step": 35260,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68977175,
+      "epoch": 0.8945966514459666,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.270682309274094e-06,
+      "loss": 1.33527784,
+      "memory(GiB)": 113.67,
+      "step": 35265,
+      "train_speed(iter/s)": 1.315885
+    },
+    {
+      "acc": 0.68898287,
+      "epoch": 0.8947234906139016,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.269668085680106e-06,
+      "loss": 1.36939917,
+      "memory(GiB)": 113.67,
+      "step": 35270,
+      "train_speed(iter/s)": 1.315908
+    },
+    {
+      "acc": 0.67123671,
+      "epoch": 0.8948503297818367,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.268653806240448e-06,
+      "loss": 1.46448545,
+      "memory(GiB)": 113.67,
+      "step": 35275,
+      "train_speed(iter/s)": 1.315931
+    },
+    {
+      "acc": 0.69067178,
+      "epoch": 0.8949771689497716,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.26763947099973e-06,
+      "loss": 1.37932472,
+      "memory(GiB)": 113.67,
+      "step": 35280,
+      "train_speed(iter/s)": 1.315954
+    },
+    {
+      "acc": 0.67773113,
+      "epoch": 0.8951040081177067,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.266625080002569e-06,
+      "loss": 1.47427206,
+      "memory(GiB)": 113.67,
+      "step": 35285,
+      "train_speed(iter/s)": 1.315977
+    },
+    {
+      "acc": 0.67383838,
+      "epoch": 0.8952308472856418,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.265610633293582e-06,
+      "loss": 1.43732786,
+      "memory(GiB)": 113.67,
+      "step": 35290,
+      "train_speed(iter/s)": 1.315999
+    },
+    {
+      "acc": 0.68138099,
+      "epoch": 0.8953576864535768,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.264596130917389e-06,
+      "loss": 1.38199482,
+      "memory(GiB)": 113.67,
+      "step": 35295,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.6978374,
+      "epoch": 0.8954845256215119,
+      "grad_norm": 3.625,
+      "learning_rate": 6.2635815729186124e-06,
+      "loss": 1.31381693,
+      "memory(GiB)": 113.67,
+      "step": 35300,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.67909632,
+      "epoch": 0.895611364789447,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.2625669593418744e-06,
+      "loss": 1.4165575,
+      "memory(GiB)": 113.67,
+      "step": 35305,
+      "train_speed(iter/s)": 1.316065
+    },
+    {
+      "acc": 0.67956371,
+      "epoch": 0.895738203957382,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.261552290231807e-06,
+      "loss": 1.38910379,
+      "memory(GiB)": 113.67,
+      "step": 35310,
+      "train_speed(iter/s)": 1.316087
+    },
+    {
+      "acc": 0.68772631,
+      "epoch": 0.8958650431253171,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.260537565633037e-06,
+      "loss": 1.3231432,
+      "memory(GiB)": 113.67,
+      "step": 35315,
+      "train_speed(iter/s)": 1.316109
+    },
+    {
+      "acc": 0.68605886,
+      "epoch": 0.8959918822932521,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.259522785590197e-06,
+      "loss": 1.40650902,
+      "memory(GiB)": 113.67,
+      "step": 35320,
+      "train_speed(iter/s)": 1.316131
+    },
+    {
+      "acc": 0.67518406,
+      "epoch": 0.8961187214611872,
+      "grad_norm": 3.984375,
+      "learning_rate": 6.2585079501479205e-06,
+      "loss": 1.4463232,
+      "memory(GiB)": 113.67,
+      "step": 35325,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.68756652,
+      "epoch": 0.8962455606291223,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.257493059350848e-06,
+      "loss": 1.35918875,
+      "memory(GiB)": 113.67,
+      "step": 35330,
+      "train_speed(iter/s)": 1.316177
+    },
+    {
+      "acc": 0.68905106,
+      "epoch": 0.8963723997970573,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.256478113243613e-06,
+      "loss": 1.4118412,
+      "memory(GiB)": 113.67,
+      "step": 35335,
+      "train_speed(iter/s)": 1.316199
+    },
+    {
+      "acc": 0.69374189,
+      "epoch": 0.8964992389649924,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.255463111870864e-06,
+      "loss": 1.36041126,
+      "memory(GiB)": 113.67,
+      "step": 35340,
+      "train_speed(iter/s)": 1.316221
+    },
+    {
+      "acc": 0.68631916,
+      "epoch": 0.8966260781329275,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.25444805527724e-06,
+      "loss": 1.40413074,
+      "memory(GiB)": 113.67,
+      "step": 35345,
+      "train_speed(iter/s)": 1.316242
+    },
+    {
+      "acc": 0.6746047,
+      "epoch": 0.8967529173008625,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.253432943507391e-06,
+      "loss": 1.47726469,
+      "memory(GiB)": 113.67,
+      "step": 35350,
+      "train_speed(iter/s)": 1.316264
+    },
+    {
+      "acc": 0.68850932,
+      "epoch": 0.8968797564687976,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.252417776605964e-06,
+      "loss": 1.35325127,
+      "memory(GiB)": 113.67,
+      "step": 35355,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.70016737,
+      "epoch": 0.8970065956367326,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.251402554617613e-06,
+      "loss": 1.33723993,
+      "memory(GiB)": 113.67,
+      "step": 35360,
+      "train_speed(iter/s)": 1.316308
+    },
+    {
+      "acc": 0.66826439,
+      "epoch": 0.8971334348046677,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.2503872775869886e-06,
+      "loss": 1.48769102,
+      "memory(GiB)": 113.67,
+      "step": 35365,
+      "train_speed(iter/s)": 1.316328
+    },
+    {
+      "acc": 0.70493999,
+      "epoch": 0.8972602739726028,
+      "grad_norm": 3.0,
+      "learning_rate": 6.249371945558751e-06,
+      "loss": 1.35983543,
+      "memory(GiB)": 113.67,
+      "step": 35370,
+      "train_speed(iter/s)": 1.31635
+    },
+    {
+      "acc": 0.69305067,
+      "epoch": 0.8973871131405378,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.248356558577555e-06,
+      "loss": 1.35573311,
+      "memory(GiB)": 113.67,
+      "step": 35375,
+      "train_speed(iter/s)": 1.316372
+    },
+    {
+      "acc": 0.6936738,
+      "epoch": 0.8975139523084729,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.247341116688067e-06,
+      "loss": 1.42604666,
+      "memory(GiB)": 113.67,
+      "step": 35380,
+      "train_speed(iter/s)": 1.316395
+    },
+    {
+      "acc": 0.69239092,
+      "epoch": 0.897640791476408,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.246325619934945e-06,
+      "loss": 1.34312944,
+      "memory(GiB)": 113.67,
+      "step": 35385,
+      "train_speed(iter/s)": 1.316417
+    },
+    {
+      "acc": 0.68749504,
+      "epoch": 0.897767630644343,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.245310068362859e-06,
+      "loss": 1.3438508,
+      "memory(GiB)": 113.67,
+      "step": 35390,
+      "train_speed(iter/s)": 1.316441
+    },
+    {
+      "acc": 0.68910522,
+      "epoch": 0.897894469812278,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.244294462016476e-06,
+      "loss": 1.40560284,
+      "memory(GiB)": 113.67,
+      "step": 35395,
+      "train_speed(iter/s)": 1.316463
+    },
+    {
+      "acc": 0.70430279,
+      "epoch": 0.898021308980213,
+      "grad_norm": 3.625,
+      "learning_rate": 6.243278800940468e-06,
+      "loss": 1.31053782,
+      "memory(GiB)": 113.67,
+      "step": 35400,
+      "train_speed(iter/s)": 1.316485
+    },
+    {
+      "acc": 0.68754387,
+      "epoch": 0.8981481481481481,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.242263085179506e-06,
+      "loss": 1.39373817,
+      "memory(GiB)": 113.67,
+      "step": 35405,
+      "train_speed(iter/s)": 1.316508
+    },
+    {
+      "acc": 0.68517284,
+      "epoch": 0.8982749873160832,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.241247314778269e-06,
+      "loss": 1.40625286,
+      "memory(GiB)": 113.67,
+      "step": 35410,
+      "train_speed(iter/s)": 1.31653
+    },
+    {
+      "acc": 0.68299775,
+      "epoch": 0.8984018264840182,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.240231489781432e-06,
+      "loss": 1.47221203,
+      "memory(GiB)": 113.67,
+      "step": 35415,
+      "train_speed(iter/s)": 1.316552
+    },
+    {
+      "acc": 0.67230234,
+      "epoch": 0.8985286656519533,
+      "grad_norm": 3.25,
+      "learning_rate": 6.239215610233678e-06,
+      "loss": 1.4076416,
+      "memory(GiB)": 113.67,
+      "step": 35420,
+      "train_speed(iter/s)": 1.316574
+    },
+    {
+      "acc": 0.69698291,
+      "epoch": 0.8986555048198884,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.238199676179688e-06,
+      "loss": 1.35124283,
+      "memory(GiB)": 113.67,
+      "step": 35425,
+      "train_speed(iter/s)": 1.316596
+    },
+    {
+      "acc": 0.69635949,
+      "epoch": 0.8987823439878234,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.2371836876641475e-06,
+      "loss": 1.33392687,
+      "memory(GiB)": 113.67,
+      "step": 35430,
+      "train_speed(iter/s)": 1.316617
+    },
+    {
+      "acc": 0.68764682,
+      "epoch": 0.8989091831557585,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.236167644731745e-06,
+      "loss": 1.40393448,
+      "memory(GiB)": 113.67,
+      "step": 35435,
+      "train_speed(iter/s)": 1.316639
+    },
+    {
+      "acc": 0.69152884,
+      "epoch": 0.8990360223236935,
+      "grad_norm": 3.125,
+      "learning_rate": 6.235151547427172e-06,
+      "loss": 1.38559895,
+      "memory(GiB)": 113.67,
+      "step": 35440,
+      "train_speed(iter/s)": 1.31666
+    },
+    {
+      "acc": 0.68744617,
+      "epoch": 0.8991628614916286,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.2341353957951165e-06,
+      "loss": 1.41929827,
+      "memory(GiB)": 113.67,
+      "step": 35445,
+      "train_speed(iter/s)": 1.316681
+    },
+    {
+      "acc": 0.66641817,
+      "epoch": 0.8992897006595637,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.233119189880279e-06,
+      "loss": 1.42959976,
+      "memory(GiB)": 113.67,
+      "step": 35450,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.67414103,
+      "epoch": 0.8994165398274987,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.232102929727353e-06,
+      "loss": 1.43102732,
+      "memory(GiB)": 113.67,
+      "step": 35455,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.6774313,
+      "epoch": 0.8995433789954338,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.231086615381039e-06,
+      "loss": 1.46317768,
+      "memory(GiB)": 113.67,
+      "step": 35460,
+      "train_speed(iter/s)": 1.316744
+    },
+    {
+      "acc": 0.69807482,
+      "epoch": 0.8996702181633689,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.2300702468860385e-06,
+      "loss": 1.38530197,
+      "memory(GiB)": 113.67,
+      "step": 35465,
+      "train_speed(iter/s)": 1.316766
+    },
+    {
+      "acc": 0.69194279,
+      "epoch": 0.8997970573313039,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.229053824287058e-06,
+      "loss": 1.33833752,
+      "memory(GiB)": 113.67,
+      "step": 35470,
+      "train_speed(iter/s)": 1.316787
+    },
+    {
+      "acc": 0.67183485,
+      "epoch": 0.899923896499239,
+      "grad_norm": 3.5,
+      "learning_rate": 6.228037347628803e-06,
+      "loss": 1.41479893,
+      "memory(GiB)": 113.67,
+      "step": 35475,
+      "train_speed(iter/s)": 1.316808
+    },
+    {
+      "acc": 0.69972134,
+      "epoch": 0.900050735667174,
+      "grad_norm": 2.59375,
+      "learning_rate": 6.227020816955982e-06,
+      "loss": 1.32115002,
+      "memory(GiB)": 113.67,
+      "step": 35480,
+      "train_speed(iter/s)": 1.316829
+    },
+    {
+      "acc": 0.6894691,
+      "epoch": 0.9001775748351091,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.226004232313308e-06,
+      "loss": 1.38217793,
+      "memory(GiB)": 113.67,
+      "step": 35485,
+      "train_speed(iter/s)": 1.31685
+    },
+    {
+      "acc": 0.67778668,
+      "epoch": 0.9003044140030442,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.224987593745493e-06,
+      "loss": 1.40290012,
+      "memory(GiB)": 113.67,
+      "step": 35490,
+      "train_speed(iter/s)": 1.316871
+    },
+    {
+      "acc": 0.67889652,
+      "epoch": 0.9004312531709792,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.223970901297255e-06,
+      "loss": 1.36564159,
+      "memory(GiB)": 113.67,
+      "step": 35495,
+      "train_speed(iter/s)": 1.316893
+    },
+    {
+      "acc": 0.68539357,
+      "epoch": 0.9005580923389143,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.222954155013312e-06,
+      "loss": 1.3866888,
+      "memory(GiB)": 113.67,
+      "step": 35500,
+      "train_speed(iter/s)": 1.316913
+    },
+    {
+      "acc": 0.69784369,
+      "epoch": 0.9006849315068494,
+      "grad_norm": 3.0,
+      "learning_rate": 6.221937354938386e-06,
+      "loss": 1.3616538,
+      "memory(GiB)": 113.67,
+      "step": 35505,
+      "train_speed(iter/s)": 1.316935
+    },
+    {
+      "acc": 0.69218178,
+      "epoch": 0.9008117706747844,
+      "grad_norm": 3.0,
+      "learning_rate": 6.2209205011171995e-06,
+      "loss": 1.39557114,
+      "memory(GiB)": 113.67,
+      "step": 35510,
+      "train_speed(iter/s)": 1.316954
+    },
+    {
+      "acc": 0.66791019,
+      "epoch": 0.9009386098427195,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.219903593594476e-06,
+      "loss": 1.40216484,
+      "memory(GiB)": 113.67,
+      "step": 35515,
+      "train_speed(iter/s)": 1.316976
+    },
+    {
+      "acc": 0.69611206,
+      "epoch": 0.9010654490106544,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.218886632414949e-06,
+      "loss": 1.33343687,
+      "memory(GiB)": 113.67,
+      "step": 35520,
+      "train_speed(iter/s)": 1.316997
+    },
+    {
+      "acc": 0.68131094,
+      "epoch": 0.9011922881785895,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.217869617623343e-06,
+      "loss": 1.41282768,
+      "memory(GiB)": 113.67,
+      "step": 35525,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.69843087,
+      "epoch": 0.9013191273465246,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.216852549264396e-06,
+      "loss": 1.37774506,
+      "memory(GiB)": 113.67,
+      "step": 35530,
+      "train_speed(iter/s)": 1.317039
+    },
+    {
+      "acc": 0.69814653,
+      "epoch": 0.9014459665144596,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.215835427382842e-06,
+      "loss": 1.3677309,
+      "memory(GiB)": 113.67,
+      "step": 35535,
+      "train_speed(iter/s)": 1.317057
+    },
+    {
+      "acc": 0.69004579,
+      "epoch": 0.9015728056823947,
+      "grad_norm": 2.875,
+      "learning_rate": 6.214818252023415e-06,
+      "loss": 1.398911,
+      "memory(GiB)": 113.67,
+      "step": 35540,
+      "train_speed(iter/s)": 1.317078
+    },
+    {
+      "acc": 0.68692112,
+      "epoch": 0.9016996448503298,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.2138010232308585e-06,
+      "loss": 1.33070354,
+      "memory(GiB)": 113.67,
+      "step": 35545,
+      "train_speed(iter/s)": 1.317096
+    },
+    {
+      "acc": 0.69863276,
+      "epoch": 0.9018264840182648,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.212783741049915e-06,
+      "loss": 1.34468565,
+      "memory(GiB)": 113.67,
+      "step": 35550,
+      "train_speed(iter/s)": 1.317117
+    },
+    {
+      "acc": 0.69141226,
+      "epoch": 0.9019533231861999,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.211766405525326e-06,
+      "loss": 1.35698595,
+      "memory(GiB)": 113.67,
+      "step": 35555,
+      "train_speed(iter/s)": 1.317138
+    },
+    {
+      "acc": 0.68892822,
+      "epoch": 0.9020801623541349,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.210749016701842e-06,
+      "loss": 1.41168404,
+      "memory(GiB)": 113.67,
+      "step": 35560,
+      "train_speed(iter/s)": 1.317159
+    },
+    {
+      "acc": 0.67491932,
+      "epoch": 0.90220700152207,
+      "grad_norm": 3.5,
+      "learning_rate": 6.2097315746242095e-06,
+      "loss": 1.43924561,
+      "memory(GiB)": 113.67,
+      "step": 35565,
+      "train_speed(iter/s)": 1.317178
+    },
+    {
+      "acc": 0.66956654,
+      "epoch": 0.9023338406900051,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.208714079337181e-06,
+      "loss": 1.47608376,
+      "memory(GiB)": 113.67,
+      "step": 35570,
+      "train_speed(iter/s)": 1.317199
+    },
+    {
+      "acc": 0.67865915,
+      "epoch": 0.9024606798579401,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.207696530885511e-06,
+      "loss": 1.41816025,
+      "memory(GiB)": 113.67,
+      "step": 35575,
+      "train_speed(iter/s)": 1.31722
+    },
+    {
+      "acc": 0.68618302,
+      "epoch": 0.9025875190258752,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.2066789293139565e-06,
+      "loss": 1.39813223,
+      "memory(GiB)": 113.67,
+      "step": 35580,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.6708571,
+      "epoch": 0.9027143581938103,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.2056612746672736e-06,
+      "loss": 1.46067324,
+      "memory(GiB)": 113.67,
+      "step": 35585,
+      "train_speed(iter/s)": 1.31726
+    },
+    {
+      "acc": 0.69142532,
+      "epoch": 0.9028411973617453,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.204643566990227e-06,
+      "loss": 1.41506739,
+      "memory(GiB)": 113.67,
+      "step": 35590,
+      "train_speed(iter/s)": 1.317281
+    },
+    {
+      "acc": 0.68467813,
+      "epoch": 0.9029680365296804,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.2036258063275764e-06,
+      "loss": 1.37068901,
+      "memory(GiB)": 113.67,
+      "step": 35595,
+      "train_speed(iter/s)": 1.317302
+    },
+    {
+      "acc": 0.69087896,
+      "epoch": 0.9030948756976154,
+      "grad_norm": 3.875,
+      "learning_rate": 6.20260799272409e-06,
+      "loss": 1.38991852,
+      "memory(GiB)": 113.67,
+      "step": 35600,
+      "train_speed(iter/s)": 1.317324
+    },
+    {
+      "acc": 0.69361033,
+      "epoch": 0.9032217148655505,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.201590126224534e-06,
+      "loss": 1.39915905,
+      "memory(GiB)": 113.67,
+      "step": 35605,
+      "train_speed(iter/s)": 1.317345
+    },
+    {
+      "acc": 0.69606104,
+      "epoch": 0.9033485540334856,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.20057220687368e-06,
+      "loss": 1.31142273,
+      "memory(GiB)": 113.67,
+      "step": 35610,
+      "train_speed(iter/s)": 1.317365
+    },
+    {
+      "acc": 0.68417902,
+      "epoch": 0.9034753932014206,
+      "grad_norm": 2.875,
+      "learning_rate": 6.199554234716301e-06,
+      "loss": 1.37486115,
+      "memory(GiB)": 113.67,
+      "step": 35615,
+      "train_speed(iter/s)": 1.317386
+    },
+    {
+      "acc": 0.68613248,
+      "epoch": 0.9036022323693557,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.19853620979717e-06,
+      "loss": 1.36004925,
+      "memory(GiB)": 113.67,
+      "step": 35620,
+      "train_speed(iter/s)": 1.317407
+    },
+    {
+      "acc": 0.69213676,
+      "epoch": 0.9037290715372908,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.1975181321610655e-06,
+      "loss": 1.37238188,
+      "memory(GiB)": 113.67,
+      "step": 35625,
+      "train_speed(iter/s)": 1.317427
+    },
+    {
+      "acc": 0.67485323,
+      "epoch": 0.9038559107052258,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.1965000018527676e-06,
+      "loss": 1.43108864,
+      "memory(GiB)": 113.67,
+      "step": 35630,
+      "train_speed(iter/s)": 1.317448
+    },
+    {
+      "acc": 0.67843313,
+      "epoch": 0.9039827498731609,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.195481818917057e-06,
+      "loss": 1.40063066,
+      "memory(GiB)": 113.67,
+      "step": 35635,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.67861938,
+      "epoch": 0.9041095890410958,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.194463583398719e-06,
+      "loss": 1.40363579,
+      "memory(GiB)": 113.67,
+      "step": 35640,
+      "train_speed(iter/s)": 1.317488
+    },
+    {
+      "acc": 0.67936926,
+      "epoch": 0.9042364282090309,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.193445295342538e-06,
+      "loss": 1.36473484,
+      "memory(GiB)": 113.67,
+      "step": 35645,
+      "train_speed(iter/s)": 1.317509
+    },
+    {
+      "acc": 0.66824636,
+      "epoch": 0.904363267376966,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.192426954793308e-06,
+      "loss": 1.43387184,
+      "memory(GiB)": 113.67,
+      "step": 35650,
+      "train_speed(iter/s)": 1.317529
+    },
+    {
+      "acc": 0.6841311,
+      "epoch": 0.904490106544901,
+      "grad_norm": 3.125,
+      "learning_rate": 6.1914085617958135e-06,
+      "loss": 1.39905901,
+      "memory(GiB)": 113.67,
+      "step": 35655,
+      "train_speed(iter/s)": 1.317549
+    },
+    {
+      "acc": 0.69372358,
+      "epoch": 0.9046169457128361,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.190390116394853e-06,
+      "loss": 1.37770481,
+      "memory(GiB)": 113.67,
+      "step": 35660,
+      "train_speed(iter/s)": 1.317568
+    },
+    {
+      "acc": 0.66836338,
+      "epoch": 0.9047437848807712,
+      "grad_norm": 5.15625,
+      "learning_rate": 6.189371618635219e-06,
+      "loss": 1.43461151,
+      "memory(GiB)": 113.67,
+      "step": 35665,
+      "train_speed(iter/s)": 1.317589
+    },
+    {
+      "acc": 0.67612371,
+      "epoch": 0.9048706240487062,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.188353068561714e-06,
+      "loss": 1.387537,
+      "memory(GiB)": 113.67,
+      "step": 35670,
+      "train_speed(iter/s)": 1.317609
+    },
+    {
+      "acc": 0.69414463,
+      "epoch": 0.9049974632166413,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.187334466219133e-06,
+      "loss": 1.34484453,
+      "memory(GiB)": 113.67,
+      "step": 35675,
+      "train_speed(iter/s)": 1.31763
+    },
+    {
+      "acc": 0.69175959,
+      "epoch": 0.9051243023845763,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.18631581165228e-06,
+      "loss": 1.31954947,
+      "memory(GiB)": 113.67,
+      "step": 35680,
+      "train_speed(iter/s)": 1.31765
+    },
+    {
+      "acc": 0.6765749,
+      "epoch": 0.9052511415525114,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.185297104905963e-06,
+      "loss": 1.40323277,
+      "memory(GiB)": 113.67,
+      "step": 35685,
+      "train_speed(iter/s)": 1.317669
+    },
+    {
+      "acc": 0.68332105,
+      "epoch": 0.9053779807204465,
+      "grad_norm": 3.25,
+      "learning_rate": 6.184278346024988e-06,
+      "loss": 1.43148479,
+      "memory(GiB)": 113.67,
+      "step": 35690,
+      "train_speed(iter/s)": 1.317691
+    },
+    {
+      "acc": 0.68808527,
+      "epoch": 0.9055048198883815,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.183259535054163e-06,
+      "loss": 1.3672534,
+      "memory(GiB)": 113.67,
+      "step": 35695,
+      "train_speed(iter/s)": 1.317711
+    },
+    {
+      "acc": 0.67055869,
+      "epoch": 0.9056316590563166,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.1822406720383e-06,
+      "loss": 1.39340219,
+      "memory(GiB)": 113.67,
+      "step": 35700,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "acc": 0.68389297,
+      "epoch": 0.9057584982242517,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.181221757022215e-06,
+      "loss": 1.39834652,
+      "memory(GiB)": 113.67,
+      "step": 35705,
+      "train_speed(iter/s)": 1.31775
+    },
+    {
+      "acc": 0.6789917,
+      "epoch": 0.9058853373921867,
+      "grad_norm": 3.5,
+      "learning_rate": 6.180202790050724e-06,
+      "loss": 1.37872858,
+      "memory(GiB)": 113.67,
+      "step": 35710,
+      "train_speed(iter/s)": 1.317771
+    },
+    {
+      "acc": 0.68560963,
+      "epoch": 0.9060121765601218,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.179183771168643e-06,
+      "loss": 1.42231522,
+      "memory(GiB)": 113.67,
+      "step": 35715,
+      "train_speed(iter/s)": 1.317791
+    },
+    {
+      "acc": 0.67990346,
+      "epoch": 0.9061390157280568,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.1781647004207965e-06,
+      "loss": 1.42386627,
+      "memory(GiB)": 113.67,
+      "step": 35720,
+      "train_speed(iter/s)": 1.317811
+    },
+    {
+      "acc": 0.67262259,
+      "epoch": 0.9062658548959919,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.177145577852005e-06,
+      "loss": 1.45550365,
+      "memory(GiB)": 113.67,
+      "step": 35725,
+      "train_speed(iter/s)": 1.317831
+    },
+    {
+      "acc": 0.67436008,
+      "epoch": 0.906392694063927,
+      "grad_norm": 3.375,
+      "learning_rate": 6.176126403507097e-06,
+      "loss": 1.42676783,
+      "memory(GiB)": 113.67,
+      "step": 35730,
+      "train_speed(iter/s)": 1.317851
+    },
+    {
+      "acc": 0.69393239,
+      "epoch": 0.906519533231862,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.175107177430897e-06,
+      "loss": 1.36754971,
+      "memory(GiB)": 113.67,
+      "step": 35735,
+      "train_speed(iter/s)": 1.317872
+    },
+    {
+      "acc": 0.68596821,
+      "epoch": 0.9066463723997971,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.17408789966824e-06,
+      "loss": 1.34781284,
+      "memory(GiB)": 113.67,
+      "step": 35740,
+      "train_speed(iter/s)": 1.317893
+    },
+    {
+      "acc": 0.68834763,
+      "epoch": 0.9067732115677322,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.173068570263951e-06,
+      "loss": 1.37924662,
+      "memory(GiB)": 113.67,
+      "step": 35745,
+      "train_speed(iter/s)": 1.317912
+    },
+    {
+      "acc": 0.68101625,
+      "epoch": 0.9069000507356672,
+      "grad_norm": 3.25,
+      "learning_rate": 6.172049189262872e-06,
+      "loss": 1.4198616,
+      "memory(GiB)": 113.67,
+      "step": 35750,
+      "train_speed(iter/s)": 1.317934
+    },
+    {
+      "acc": 0.6954361,
+      "epoch": 0.9070268899036023,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.1710297567098354e-06,
+      "loss": 1.40313606,
+      "memory(GiB)": 113.67,
+      "step": 35755,
+      "train_speed(iter/s)": 1.317955
+    },
+    {
+      "acc": 0.6806735,
+      "epoch": 0.9071537290715372,
+      "grad_norm": 3.875,
+      "learning_rate": 6.170010272649682e-06,
+      "loss": 1.41537294,
+      "memory(GiB)": 113.67,
+      "step": 35760,
+      "train_speed(iter/s)": 1.317975
+    },
+    {
+      "acc": 0.67437801,
+      "epoch": 0.9072805682394723,
+      "grad_norm": 4.0,
+      "learning_rate": 6.168990737127254e-06,
+      "loss": 1.43641243,
+      "memory(GiB)": 113.67,
+      "step": 35765,
+      "train_speed(iter/s)": 1.317995
+    },
+    {
+      "acc": 0.69131737,
+      "epoch": 0.9074074074074074,
+      "grad_norm": 3.5,
+      "learning_rate": 6.167971150187394e-06,
+      "loss": 1.37719479,
+      "memory(GiB)": 113.67,
+      "step": 35770,
+      "train_speed(iter/s)": 1.318015
+    },
+    {
+      "acc": 0.67502146,
+      "epoch": 0.9075342465753424,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.166951511874948e-06,
+      "loss": 1.44962864,
+      "memory(GiB)": 113.67,
+      "step": 35775,
+      "train_speed(iter/s)": 1.318035
+    },
+    {
+      "acc": 0.68808317,
+      "epoch": 0.9076610857432775,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.165931822234764e-06,
+      "loss": 1.37900209,
+      "memory(GiB)": 113.67,
+      "step": 35780,
+      "train_speed(iter/s)": 1.318056
+    },
+    {
+      "acc": 0.68730316,
+      "epoch": 0.9077879249112126,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.164912081311694e-06,
+      "loss": 1.46251011,
+      "memory(GiB)": 113.67,
+      "step": 35785,
+      "train_speed(iter/s)": 1.318077
+    },
+    {
+      "acc": 0.68969193,
+      "epoch": 0.9079147640791476,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.163892289150588e-06,
+      "loss": 1.3475913,
+      "memory(GiB)": 113.67,
+      "step": 35790,
+      "train_speed(iter/s)": 1.318097
+    },
+    {
+      "acc": 0.69597006,
+      "epoch": 0.9080416032470827,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.162872445796303e-06,
+      "loss": 1.37441788,
+      "memory(GiB)": 113.67,
+      "step": 35795,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.67600379,
+      "epoch": 0.9081684424150177,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.161852551293697e-06,
+      "loss": 1.43759012,
+      "memory(GiB)": 113.67,
+      "step": 35800,
+      "train_speed(iter/s)": 1.318138
+    },
+    {
+      "acc": 0.68533578,
+      "epoch": 0.9082952815829528,
+      "grad_norm": 3.625,
+      "learning_rate": 6.160832605687628e-06,
+      "loss": 1.43646536,
+      "memory(GiB)": 113.67,
+      "step": 35805,
+      "train_speed(iter/s)": 1.318159
+    },
+    {
+      "acc": 0.68464565,
+      "epoch": 0.9084221207508879,
+      "grad_norm": 3.375,
+      "learning_rate": 6.159812609022961e-06,
+      "loss": 1.38927374,
+      "memory(GiB)": 113.67,
+      "step": 35810,
+      "train_speed(iter/s)": 1.31818
+    },
+    {
+      "acc": 0.68140736,
+      "epoch": 0.9085489599188229,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.158792561344553e-06,
+      "loss": 1.43976784,
+      "memory(GiB)": 113.67,
+      "step": 35815,
+      "train_speed(iter/s)": 1.3182
+    },
+    {
+      "acc": 0.68367128,
+      "epoch": 0.908675799086758,
+      "grad_norm": 3.5,
+      "learning_rate": 6.157772462697277e-06,
+      "loss": 1.41832981,
+      "memory(GiB)": 113.67,
+      "step": 35820,
+      "train_speed(iter/s)": 1.31822
+    },
+    {
+      "acc": 0.69547482,
+      "epoch": 0.9088026382546931,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.156752313125998e-06,
+      "loss": 1.43612537,
+      "memory(GiB)": 113.67,
+      "step": 35825,
+      "train_speed(iter/s)": 1.318241
+    },
+    {
+      "acc": 0.69185762,
+      "epoch": 0.9089294774226281,
+      "grad_norm": 3.5,
+      "learning_rate": 6.155732112675587e-06,
+      "loss": 1.38338156,
+      "memory(GiB)": 113.67,
+      "step": 35830,
+      "train_speed(iter/s)": 1.318262
+    },
+    {
+      "acc": 0.70496244,
+      "epoch": 0.9090563165905632,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.154711861390919e-06,
+      "loss": 1.29207201,
+      "memory(GiB)": 113.67,
+      "step": 35835,
+      "train_speed(iter/s)": 1.318283
+    },
+    {
+      "acc": 0.69210072,
+      "epoch": 0.9091831557584982,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.153691559316868e-06,
+      "loss": 1.40174408,
+      "memory(GiB)": 113.67,
+      "step": 35840,
+      "train_speed(iter/s)": 1.318304
+    },
+    {
+      "acc": 0.66956077,
+      "epoch": 0.9093099949264333,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.152671206498311e-06,
+      "loss": 1.50035572,
+      "memory(GiB)": 113.67,
+      "step": 35845,
+      "train_speed(iter/s)": 1.318325
+    },
+    {
+      "acc": 0.70214782,
+      "epoch": 0.9094368340943684,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.151650802980128e-06,
+      "loss": 1.34741287,
+      "memory(GiB)": 113.67,
+      "step": 35850,
+      "train_speed(iter/s)": 1.318345
+    },
+    {
+      "acc": 0.67623324,
+      "epoch": 0.9095636732623034,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.150630348807201e-06,
+      "loss": 1.4592783,
+      "memory(GiB)": 113.67,
+      "step": 35855,
+      "train_speed(iter/s)": 1.318366
+    },
+    {
+      "acc": 0.66908741,
+      "epoch": 0.9096905124302385,
+      "grad_norm": 4.125,
+      "learning_rate": 6.149609844024413e-06,
+      "loss": 1.46331978,
+      "memory(GiB)": 113.67,
+      "step": 35860,
+      "train_speed(iter/s)": 1.318388
+    },
+    {
+      "acc": 0.68261857,
+      "epoch": 0.9098173515981736,
+      "grad_norm": 2.75,
+      "learning_rate": 6.148589288676652e-06,
+      "loss": 1.45802803,
+      "memory(GiB)": 113.67,
+      "step": 35865,
+      "train_speed(iter/s)": 1.318409
+    },
+    {
+      "acc": 0.69322453,
+      "epoch": 0.9099441907661086,
+      "grad_norm": 4.0,
+      "learning_rate": 6.147568682808808e-06,
+      "loss": 1.380618,
+      "memory(GiB)": 113.67,
+      "step": 35870,
+      "train_speed(iter/s)": 1.318431
+    },
+    {
+      "acc": 0.68944364,
+      "epoch": 0.9100710299340437,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.146548026465766e-06,
+      "loss": 1.42818003,
+      "memory(GiB)": 113.67,
+      "step": 35875,
+      "train_speed(iter/s)": 1.318451
+    },
+    {
+      "acc": 0.68543587,
+      "epoch": 0.9101978691019786,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.145527319692427e-06,
+      "loss": 1.33861637,
+      "memory(GiB)": 113.67,
+      "step": 35880,
+      "train_speed(iter/s)": 1.318472
+    },
+    {
+      "acc": 0.69064226,
+      "epoch": 0.9103247082699137,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.144506562533678e-06,
+      "loss": 1.35932178,
+      "memory(GiB)": 113.67,
+      "step": 35885,
+      "train_speed(iter/s)": 1.318494
+    },
+    {
+      "acc": 0.68776579,
+      "epoch": 0.9104515474378488,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.143485755034425e-06,
+      "loss": 1.37156773,
+      "memory(GiB)": 113.67,
+      "step": 35890,
+      "train_speed(iter/s)": 1.318515
+    },
+    {
+      "acc": 0.6781702,
+      "epoch": 0.9105783866057838,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.14246489723956e-06,
+      "loss": 1.39239063,
+      "memory(GiB)": 113.67,
+      "step": 35895,
+      "train_speed(iter/s)": 1.318536
+    },
+    {
+      "acc": 0.68817596,
+      "epoch": 0.9107052257737189,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.141443989193988e-06,
+      "loss": 1.39268131,
+      "memory(GiB)": 113.67,
+      "step": 35900,
+      "train_speed(iter/s)": 1.318558
+    },
+    {
+      "acc": 0.67412472,
+      "epoch": 0.910832064941654,
+      "grad_norm": 3.375,
+      "learning_rate": 6.140423030942615e-06,
+      "loss": 1.5044179,
+      "memory(GiB)": 113.67,
+      "step": 35905,
+      "train_speed(iter/s)": 1.318578
+    },
+    {
+      "acc": 0.68604355,
+      "epoch": 0.910958904109589,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.139402022530344e-06,
+      "loss": 1.39914074,
+      "memory(GiB)": 113.67,
+      "step": 35910,
+      "train_speed(iter/s)": 1.3186
+    },
+    {
+      "acc": 0.67934189,
+      "epoch": 0.9110857432775241,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.138380964002087e-06,
+      "loss": 1.41534538,
+      "memory(GiB)": 113.67,
+      "step": 35915,
+      "train_speed(iter/s)": 1.318621
+    },
+    {
+      "acc": 0.69707341,
+      "epoch": 0.9112125824454591,
+      "grad_norm": 4.90625,
+      "learning_rate": 6.13735985540275e-06,
+      "loss": 1.34046936,
+      "memory(GiB)": 113.67,
+      "step": 35920,
+      "train_speed(iter/s)": 1.318642
+    },
+    {
+      "acc": 0.69121771,
+      "epoch": 0.9113394216133942,
+      "grad_norm": 3.90625,
+      "learning_rate": 6.13633869677725e-06,
+      "loss": 1.35165691,
+      "memory(GiB)": 123.79,
+      "step": 35925,
+      "train_speed(iter/s)": 1.318657
+    },
+    {
+      "acc": 0.67635593,
+      "epoch": 0.9114662607813293,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.1353174881705e-06,
+      "loss": 1.41793108,
+      "memory(GiB)": 123.79,
+      "step": 35930,
+      "train_speed(iter/s)": 1.318676
+    },
+    {
+      "acc": 0.67078986,
+      "epoch": 0.9115930999492643,
+      "grad_norm": 3.125,
+      "learning_rate": 6.134296229627419e-06,
+      "loss": 1.47979374,
+      "memory(GiB)": 123.79,
+      "step": 35935,
+      "train_speed(iter/s)": 1.318697
+    },
+    {
+      "acc": 0.68794427,
+      "epoch": 0.9117199391171994,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.1332749211929255e-06,
+      "loss": 1.38859501,
+      "memory(GiB)": 123.79,
+      "step": 35940,
+      "train_speed(iter/s)": 1.318717
+    },
+    {
+      "acc": 0.70735588,
+      "epoch": 0.9118467782851345,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.132253562911941e-06,
+      "loss": 1.33588333,
+      "memory(GiB)": 123.79,
+      "step": 35945,
+      "train_speed(iter/s)": 1.318737
+    },
+    {
+      "acc": 0.69470692,
+      "epoch": 0.9119736174530695,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.1312321548293895e-06,
+      "loss": 1.35676098,
+      "memory(GiB)": 123.79,
+      "step": 35950,
+      "train_speed(iter/s)": 1.318759
+    },
+    {
+      "acc": 0.68593464,
+      "epoch": 0.9121004566210046,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.130210696990197e-06,
+      "loss": 1.48057594,
+      "memory(GiB)": 123.79,
+      "step": 35955,
+      "train_speed(iter/s)": 1.318779
+    },
+    {
+      "acc": 0.68610506,
+      "epoch": 0.9122272957889396,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.129189189439293e-06,
+      "loss": 1.44336834,
+      "memory(GiB)": 123.79,
+      "step": 35960,
+      "train_speed(iter/s)": 1.318801
+    },
+    {
+      "acc": 0.68392096,
+      "epoch": 0.9123541349568747,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.128167632221605e-06,
+      "loss": 1.42755508,
+      "memory(GiB)": 123.79,
+      "step": 35965,
+      "train_speed(iter/s)": 1.318822
+    },
+    {
+      "acc": 0.69089537,
+      "epoch": 0.9124809741248098,
+      "grad_norm": 3.0,
+      "learning_rate": 6.127146025382069e-06,
+      "loss": 1.3448534,
+      "memory(GiB)": 123.79,
+      "step": 35970,
+      "train_speed(iter/s)": 1.31884
+    },
+    {
+      "acc": 0.67804599,
+      "epoch": 0.9126078132927448,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.126124368965619e-06,
+      "loss": 1.48159618,
+      "memory(GiB)": 123.79,
+      "step": 35975,
+      "train_speed(iter/s)": 1.318862
+    },
+    {
+      "acc": 0.65978174,
+      "epoch": 0.9127346524606799,
+      "grad_norm": 3.125,
+      "learning_rate": 6.125102663017191e-06,
+      "loss": 1.50882568,
+      "memory(GiB)": 123.79,
+      "step": 35980,
+      "train_speed(iter/s)": 1.318882
+    },
+    {
+      "acc": 0.69080172,
+      "epoch": 0.912861491628615,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.124080907581724e-06,
+      "loss": 1.37198591,
+      "memory(GiB)": 123.79,
+      "step": 35985,
+      "train_speed(iter/s)": 1.318902
+    },
+    {
+      "acc": 0.69472065,
+      "epoch": 0.91298833079655,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.1230591027041605e-06,
+      "loss": 1.38442764,
+      "memory(GiB)": 123.79,
+      "step": 35990,
+      "train_speed(iter/s)": 1.318922
+    },
+    {
+      "acc": 0.69603376,
+      "epoch": 0.913115169964485,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.1220372484294444e-06,
+      "loss": 1.33867226,
+      "memory(GiB)": 123.79,
+      "step": 35995,
+      "train_speed(iter/s)": 1.318942
+    },
+    {
+      "acc": 0.67958903,
+      "epoch": 0.91324200913242,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.12101534480252e-06,
+      "loss": 1.41058445,
+      "memory(GiB)": 123.79,
+      "step": 36000,
+      "train_speed(iter/s)": 1.318962
+    },
+    {
+      "epoch": 0.91324200913242,
+      "eval_acc": 0.6748256786472788,
+      "eval_loss": 1.3574117422103882,
+      "eval_runtime": 69.5979,
+      "eval_samples_per_second": 91.526,
+      "eval_steps_per_second": 22.889,
+      "step": 36000
+    },
+    {
+      "acc": 0.68375988,
+      "epoch": 0.9133688483003551,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.119993391868335e-06,
+      "loss": 1.38504019,
+      "memory(GiB)": 123.79,
+      "step": 36005,
+      "train_speed(iter/s)": 1.315026
+    },
+    {
+      "acc": 0.70208731,
+      "epoch": 0.9134956874682902,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.118971389671842e-06,
+      "loss": 1.38037663,
+      "memory(GiB)": 123.79,
+      "step": 36010,
+      "train_speed(iter/s)": 1.315047
+    },
+    {
+      "acc": 0.67623973,
+      "epoch": 0.9136225266362252,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.117949338257989e-06,
+      "loss": 1.42386932,
+      "memory(GiB)": 123.79,
+      "step": 36015,
+      "train_speed(iter/s)": 1.315069
+    },
+    {
+      "acc": 0.69760494,
+      "epoch": 0.9137493658041603,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.116927237671735e-06,
+      "loss": 1.36895771,
+      "memory(GiB)": 123.79,
+      "step": 36020,
+      "train_speed(iter/s)": 1.31509
+    },
+    {
+      "acc": 0.68319993,
+      "epoch": 0.9138762049720954,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.115905087958032e-06,
+      "loss": 1.36566124,
+      "memory(GiB)": 123.79,
+      "step": 36025,
+      "train_speed(iter/s)": 1.315112
+    },
+    {
+      "acc": 0.68433409,
+      "epoch": 0.9140030441400304,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.114882889161844e-06,
+      "loss": 1.43192673,
+      "memory(GiB)": 123.79,
+      "step": 36030,
+      "train_speed(iter/s)": 1.315133
+    },
+    {
+      "acc": 0.69437437,
+      "epoch": 0.9141298833079655,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.113860641328127e-06,
+      "loss": 1.3852499,
+      "memory(GiB)": 123.79,
+      "step": 36035,
+      "train_speed(iter/s)": 1.315155
+    },
+    {
+      "acc": 0.68791943,
+      "epoch": 0.9142567224759005,
+      "grad_norm": 3.125,
+      "learning_rate": 6.112838344501846e-06,
+      "loss": 1.36131315,
+      "memory(GiB)": 123.79,
+      "step": 36040,
+      "train_speed(iter/s)": 1.315176
+    },
+    {
+      "acc": 0.67688303,
+      "epoch": 0.9143835616438356,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.111815998727966e-06,
+      "loss": 1.45198517,
+      "memory(GiB)": 123.79,
+      "step": 36045,
+      "train_speed(iter/s)": 1.315198
+    },
+    {
+      "acc": 0.68918438,
+      "epoch": 0.9145104008117707,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.110793604051455e-06,
+      "loss": 1.36557703,
+      "memory(GiB)": 123.79,
+      "step": 36050,
+      "train_speed(iter/s)": 1.315218
+    },
+    {
+      "acc": 0.68776307,
+      "epoch": 0.9146372399797057,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.109771160517283e-06,
+      "loss": 1.40030155,
+      "memory(GiB)": 123.79,
+      "step": 36055,
+      "train_speed(iter/s)": 1.315238
+    },
+    {
+      "acc": 0.68884602,
+      "epoch": 0.9147640791476408,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.108748668170419e-06,
+      "loss": 1.41369286,
+      "memory(GiB)": 123.79,
+      "step": 36060,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.69480686,
+      "epoch": 0.9148909183155759,
+      "grad_norm": 2.75,
+      "learning_rate": 6.1077261270558385e-06,
+      "loss": 1.39180841,
+      "memory(GiB)": 123.79,
+      "step": 36065,
+      "train_speed(iter/s)": 1.315279
+    },
+    {
+      "acc": 0.68520336,
+      "epoch": 0.9150177574835109,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.106703537218518e-06,
+      "loss": 1.41459827,
+      "memory(GiB)": 123.79,
+      "step": 36070,
+      "train_speed(iter/s)": 1.315299
+    },
+    {
+      "acc": 0.68162661,
+      "epoch": 0.915144596651446,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.105680898703434e-06,
+      "loss": 1.41338253,
+      "memory(GiB)": 123.79,
+      "step": 36075,
+      "train_speed(iter/s)": 1.31532
+    },
+    {
+      "acc": 0.67308478,
+      "epoch": 0.915271435819381,
+      "grad_norm": 2.53125,
+      "learning_rate": 6.104658211555568e-06,
+      "loss": 1.45389233,
+      "memory(GiB)": 123.79,
+      "step": 36080,
+      "train_speed(iter/s)": 1.31534
+    },
+    {
+      "acc": 0.68202715,
+      "epoch": 0.9153982749873161,
+      "grad_norm": 3.90625,
+      "learning_rate": 6.103635475819902e-06,
+      "loss": 1.41406307,
+      "memory(GiB)": 123.79,
+      "step": 36085,
+      "train_speed(iter/s)": 1.315361
+    },
+    {
+      "acc": 0.68719025,
+      "epoch": 0.9155251141552512,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.102612691541422e-06,
+      "loss": 1.40083628,
+      "memory(GiB)": 123.79,
+      "step": 36090,
+      "train_speed(iter/s)": 1.315383
+    },
+    {
+      "acc": 0.66669436,
+      "epoch": 0.9156519533231862,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.10158985876511e-06,
+      "loss": 1.38383884,
+      "memory(GiB)": 123.79,
+      "step": 36095,
+      "train_speed(iter/s)": 1.315405
+    },
+    {
+      "acc": 0.68878546,
+      "epoch": 0.9157787924911213,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.10056697753596e-06,
+      "loss": 1.35853577,
+      "memory(GiB)": 123.79,
+      "step": 36100,
+      "train_speed(iter/s)": 1.315426
+    },
+    {
+      "acc": 0.6799448,
+      "epoch": 0.9159056316590564,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.0995440478989595e-06,
+      "loss": 1.38385258,
+      "memory(GiB)": 123.79,
+      "step": 36105,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.71058059,
+      "epoch": 0.9160324708269914,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.098521069899104e-06,
+      "loss": 1.24411955,
+      "memory(GiB)": 123.79,
+      "step": 36110,
+      "train_speed(iter/s)": 1.315469
+    },
+    {
+      "acc": 0.68125153,
+      "epoch": 0.9161593099949265,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.097498043581385e-06,
+      "loss": 1.4407093,
+      "memory(GiB)": 123.79,
+      "step": 36115,
+      "train_speed(iter/s)": 1.315491
+    },
+    {
+      "acc": 0.69505749,
+      "epoch": 0.9162861491628614,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.096474968990804e-06,
+      "loss": 1.37043171,
+      "memory(GiB)": 123.79,
+      "step": 36120,
+      "train_speed(iter/s)": 1.315511
+    },
+    {
+      "acc": 0.68688874,
+      "epoch": 0.9164129883307965,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.095451846172358e-06,
+      "loss": 1.40146255,
+      "memory(GiB)": 123.79,
+      "step": 36125,
+      "train_speed(iter/s)": 1.315533
+    },
+    {
+      "acc": 0.67696686,
+      "epoch": 0.9165398274987316,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.094428675171049e-06,
+      "loss": 1.4301487,
+      "memory(GiB)": 123.79,
+      "step": 36130,
+      "train_speed(iter/s)": 1.315555
+    },
+    {
+      "acc": 0.68055391,
+      "epoch": 0.9166666666666666,
+      "grad_norm": 3.375,
+      "learning_rate": 6.09340545603188e-06,
+      "loss": 1.4169136,
+      "memory(GiB)": 123.79,
+      "step": 36135,
+      "train_speed(iter/s)": 1.315577
+    },
+    {
+      "acc": 0.70132437,
+      "epoch": 0.9167935058346017,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.092382188799858e-06,
+      "loss": 1.36613312,
+      "memory(GiB)": 123.79,
+      "step": 36140,
+      "train_speed(iter/s)": 1.315598
+    },
+    {
+      "acc": 0.69523449,
+      "epoch": 0.9169203450025368,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.09135887351999e-06,
+      "loss": 1.38027954,
+      "memory(GiB)": 123.79,
+      "step": 36145,
+      "train_speed(iter/s)": 1.31562
+    },
+    {
+      "acc": 0.68132973,
+      "epoch": 0.9170471841704718,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.090335510237286e-06,
+      "loss": 1.3994154,
+      "memory(GiB)": 123.79,
+      "step": 36150,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.69147825,
+      "epoch": 0.9171740233384069,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.089312098996758e-06,
+      "loss": 1.34599094,
+      "memory(GiB)": 123.79,
+      "step": 36155,
+      "train_speed(iter/s)": 1.31566
+    },
+    {
+      "acc": 0.690903,
+      "epoch": 0.9173008625063419,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.088288639843422e-06,
+      "loss": 1.39337225,
+      "memory(GiB)": 123.79,
+      "step": 36160,
+      "train_speed(iter/s)": 1.315682
+    },
+    {
+      "acc": 0.69706068,
+      "epoch": 0.917427701674277,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.08726513282229e-06,
+      "loss": 1.38266764,
+      "memory(GiB)": 123.79,
+      "step": 36165,
+      "train_speed(iter/s)": 1.315702
+    },
+    {
+      "acc": 0.68970022,
+      "epoch": 0.9175545408422121,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.0862415779783855e-06,
+      "loss": 1.39471817,
+      "memory(GiB)": 123.79,
+      "step": 36170,
+      "train_speed(iter/s)": 1.315723
+    },
+    {
+      "acc": 0.67542658,
+      "epoch": 0.9176813800101471,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.085217975356726e-06,
+      "loss": 1.42269974,
+      "memory(GiB)": 123.79,
+      "step": 36175,
+      "train_speed(iter/s)": 1.315743
+    },
+    {
+      "acc": 0.67646856,
+      "epoch": 0.9178082191780822,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.084194325002335e-06,
+      "loss": 1.45775776,
+      "memory(GiB)": 123.79,
+      "step": 36180,
+      "train_speed(iter/s)": 1.315756
+    },
+    {
+      "acc": 0.70094781,
+      "epoch": 0.9179350583460173,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.083170626960237e-06,
+      "loss": 1.3574008,
+      "memory(GiB)": 123.79,
+      "step": 36185,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.69378228,
+      "epoch": 0.9180618975139523,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.082146881275458e-06,
+      "loss": 1.37792244,
+      "memory(GiB)": 123.79,
+      "step": 36190,
+      "train_speed(iter/s)": 1.315798
+    },
+    {
+      "acc": 0.68365016,
+      "epoch": 0.9181887366818874,
+      "grad_norm": 3.125,
+      "learning_rate": 6.081123087993028e-06,
+      "loss": 1.38209877,
+      "memory(GiB)": 123.79,
+      "step": 36195,
+      "train_speed(iter/s)": 1.315819
+    },
+    {
+      "acc": 0.66647863,
+      "epoch": 0.9183155758498224,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.0800992471579775e-06,
+      "loss": 1.43038559,
+      "memory(GiB)": 123.79,
+      "step": 36200,
+      "train_speed(iter/s)": 1.315841
+    },
+    {
+      "acc": 0.69025221,
+      "epoch": 0.9184424150177575,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.079075358815341e-06,
+      "loss": 1.35020266,
+      "memory(GiB)": 123.79,
+      "step": 36205,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68222618,
+      "epoch": 0.9185692541856926,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.078051423010152e-06,
+      "loss": 1.42538328,
+      "memory(GiB)": 123.79,
+      "step": 36210,
+      "train_speed(iter/s)": 1.315884
+    },
+    {
+      "acc": 0.68151331,
+      "epoch": 0.9186960933536276,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.077027439787448e-06,
+      "loss": 1.46144733,
+      "memory(GiB)": 123.79,
+      "step": 36215,
+      "train_speed(iter/s)": 1.315906
+    },
+    {
+      "acc": 0.68943739,
+      "epoch": 0.9188229325215627,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.076003409192268e-06,
+      "loss": 1.42865057,
+      "memory(GiB)": 123.79,
+      "step": 36220,
+      "train_speed(iter/s)": 1.315927
+    },
+    {
+      "acc": 0.68478003,
+      "epoch": 0.9189497716894978,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.074979331269656e-06,
+      "loss": 1.40817909,
+      "memory(GiB)": 123.79,
+      "step": 36225,
+      "train_speed(iter/s)": 1.315949
+    },
+    {
+      "acc": 0.68284793,
+      "epoch": 0.9190766108574328,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.0739552060646525e-06,
+      "loss": 1.4448103,
+      "memory(GiB)": 123.79,
+      "step": 36230,
+      "train_speed(iter/s)": 1.315971
+    },
+    {
+      "acc": 0.68858681,
+      "epoch": 0.9192034500253679,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.0729310336223025e-06,
+      "loss": 1.36528835,
+      "memory(GiB)": 123.79,
+      "step": 36235,
+      "train_speed(iter/s)": 1.315992
+    },
+    {
+      "acc": 0.69185047,
+      "epoch": 0.9193302891933028,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.071906813987658e-06,
+      "loss": 1.35972109,
+      "memory(GiB)": 123.79,
+      "step": 36240,
+      "train_speed(iter/s)": 1.316013
+    },
+    {
+      "acc": 0.69316168,
+      "epoch": 0.9194571283612379,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.070882547205764e-06,
+      "loss": 1.41000223,
+      "memory(GiB)": 123.79,
+      "step": 36245,
+      "train_speed(iter/s)": 1.316034
+    },
+    {
+      "acc": 0.69443817,
+      "epoch": 0.919583967529173,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.069858233321677e-06,
+      "loss": 1.32051563,
+      "memory(GiB)": 123.79,
+      "step": 36250,
+      "train_speed(iter/s)": 1.316054
+    },
+    {
+      "acc": 0.69305882,
+      "epoch": 0.919710806697108,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.068833872380445e-06,
+      "loss": 1.35154257,
+      "memory(GiB)": 123.79,
+      "step": 36255,
+      "train_speed(iter/s)": 1.316076
+    },
+    {
+      "acc": 0.67056036,
+      "epoch": 0.9198376458650431,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.067809464427129e-06,
+      "loss": 1.44846668,
+      "memory(GiB)": 123.79,
+      "step": 36260,
+      "train_speed(iter/s)": 1.316097
+    },
+    {
+      "acc": 0.69426374,
+      "epoch": 0.9199644850329782,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.066785009506786e-06,
+      "loss": 1.37085972,
+      "memory(GiB)": 123.79,
+      "step": 36265,
+      "train_speed(iter/s)": 1.316119
+    },
+    {
+      "acc": 0.69019418,
+      "epoch": 0.9200913242009132,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.065760507664474e-06,
+      "loss": 1.39219036,
+      "memory(GiB)": 123.79,
+      "step": 36270,
+      "train_speed(iter/s)": 1.31614
+    },
+    {
+      "acc": 0.6940474,
+      "epoch": 0.9202181633688483,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.064735958945258e-06,
+      "loss": 1.39349766,
+      "memory(GiB)": 123.79,
+      "step": 36275,
+      "train_speed(iter/s)": 1.316162
+    },
+    {
+      "acc": 0.69574432,
+      "epoch": 0.9203450025367833,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.0637113633942006e-06,
+      "loss": 1.41798325,
+      "memory(GiB)": 123.79,
+      "step": 36280,
+      "train_speed(iter/s)": 1.316182
+    },
+    {
+      "acc": 0.6710629,
+      "epoch": 0.9204718417047184,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.0626867210563675e-06,
+      "loss": 1.50531301,
+      "memory(GiB)": 123.79,
+      "step": 36285,
+      "train_speed(iter/s)": 1.316204
+    },
+    {
+      "acc": 0.6989584,
+      "epoch": 0.9205986808726535,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.061662031976828e-06,
+      "loss": 1.36036472,
+      "memory(GiB)": 123.79,
+      "step": 36290,
+      "train_speed(iter/s)": 1.316225
+    },
+    {
+      "acc": 0.68653364,
+      "epoch": 0.9207255200405885,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.0606372962006534e-06,
+      "loss": 1.35276318,
+      "memory(GiB)": 123.79,
+      "step": 36295,
+      "train_speed(iter/s)": 1.316247
+    },
+    {
+      "acc": 0.67875996,
+      "epoch": 0.9208523592085236,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.0596125137729145e-06,
+      "loss": 1.41682138,
+      "memory(GiB)": 123.79,
+      "step": 36300,
+      "train_speed(iter/s)": 1.316264
+    },
+    {
+      "acc": 0.6841825,
+      "epoch": 0.9209791983764587,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.058587684738685e-06,
+      "loss": 1.44076424,
+      "memory(GiB)": 123.79,
+      "step": 36305,
+      "train_speed(iter/s)": 1.316285
+    },
+    {
+      "acc": 0.69283462,
+      "epoch": 0.9211060375443937,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.057562809143045e-06,
+      "loss": 1.35826473,
+      "memory(GiB)": 123.79,
+      "step": 36310,
+      "train_speed(iter/s)": 1.316306
+    },
+    {
+      "acc": 0.67957554,
+      "epoch": 0.9212328767123288,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.056537887031069e-06,
+      "loss": 1.41057224,
+      "memory(GiB)": 123.79,
+      "step": 36315,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.69297729,
+      "epoch": 0.9213597158802638,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.055512918447841e-06,
+      "loss": 1.38862019,
+      "memory(GiB)": 123.79,
+      "step": 36320,
+      "train_speed(iter/s)": 1.316346
+    },
+    {
+      "acc": 0.67919817,
+      "epoch": 0.9214865550481989,
+      "grad_norm": 2.75,
+      "learning_rate": 6.054487903438442e-06,
+      "loss": 1.37897005,
+      "memory(GiB)": 123.79,
+      "step": 36325,
+      "train_speed(iter/s)": 1.316367
+    },
+    {
+      "acc": 0.68720913,
+      "epoch": 0.921613394216134,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.0534628420479576e-06,
+      "loss": 1.41675987,
+      "memory(GiB)": 123.79,
+      "step": 36330,
+      "train_speed(iter/s)": 1.316388
+    },
+    {
+      "acc": 0.6873353,
+      "epoch": 0.921740233384069,
+      "grad_norm": 3.0,
+      "learning_rate": 6.0524377343214724e-06,
+      "loss": 1.44921751,
+      "memory(GiB)": 123.79,
+      "step": 36335,
+      "train_speed(iter/s)": 1.316408
+    },
+    {
+      "acc": 0.67574968,
+      "epoch": 0.9218670725520041,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.051412580304079e-06,
+      "loss": 1.42584915,
+      "memory(GiB)": 123.79,
+      "step": 36340,
+      "train_speed(iter/s)": 1.316428
+    },
+    {
+      "acc": 0.6791297,
+      "epoch": 0.9219939117199392,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.050387380040864e-06,
+      "loss": 1.44616375,
+      "memory(GiB)": 123.79,
+      "step": 36345,
+      "train_speed(iter/s)": 1.316448
+    },
+    {
+      "acc": 0.67421722,
+      "epoch": 0.9221207508878742,
+      "grad_norm": 3.0,
+      "learning_rate": 6.049362133576924e-06,
+      "loss": 1.43652887,
+      "memory(GiB)": 123.79,
+      "step": 36350,
+      "train_speed(iter/s)": 1.316468
+    },
+    {
+      "acc": 0.707899,
+      "epoch": 0.9222475900558093,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.048336840957351e-06,
+      "loss": 1.35199337,
+      "memory(GiB)": 123.79,
+      "step": 36355,
+      "train_speed(iter/s)": 1.316489
+    },
+    {
+      "acc": 0.69245882,
+      "epoch": 0.9223744292237442,
+      "grad_norm": 2.703125,
+      "learning_rate": 6.047311502227245e-06,
+      "loss": 1.38497219,
+      "memory(GiB)": 123.79,
+      "step": 36360,
+      "train_speed(iter/s)": 1.31651
+    },
+    {
+      "acc": 0.68087459,
+      "epoch": 0.9225012683916793,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.046286117431703e-06,
+      "loss": 1.40824537,
+      "memory(GiB)": 123.79,
+      "step": 36365,
+      "train_speed(iter/s)": 1.316529
+    },
+    {
+      "acc": 0.68656821,
+      "epoch": 0.9226281075596144,
+      "grad_norm": 2.390625,
+      "learning_rate": 6.0452606866158246e-06,
+      "loss": 1.4240057,
+      "memory(GiB)": 123.79,
+      "step": 36370,
+      "train_speed(iter/s)": 1.316549
+    },
+    {
+      "acc": 0.69283314,
+      "epoch": 0.9227549467275494,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.044235209824716e-06,
+      "loss": 1.33994188,
+      "memory(GiB)": 123.79,
+      "step": 36375,
+      "train_speed(iter/s)": 1.316569
+    },
+    {
+      "acc": 0.67382755,
+      "epoch": 0.9228817858954845,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.04320968710348e-06,
+      "loss": 1.50345583,
+      "memory(GiB)": 123.79,
+      "step": 36380,
+      "train_speed(iter/s)": 1.31659
+    },
+    {
+      "acc": 0.66832056,
+      "epoch": 0.9230086250634196,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.042184118497223e-06,
+      "loss": 1.45647936,
+      "memory(GiB)": 123.79,
+      "step": 36385,
+      "train_speed(iter/s)": 1.31661
+    },
+    {
+      "acc": 0.69360189,
+      "epoch": 0.9231354642313546,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.0411585040510576e-06,
+      "loss": 1.38068171,
+      "memory(GiB)": 123.79,
+      "step": 36390,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.69660749,
+      "epoch": 0.9232623033992897,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.040132843810091e-06,
+      "loss": 1.36526575,
+      "memory(GiB)": 123.79,
+      "step": 36395,
+      "train_speed(iter/s)": 1.31665
+    },
+    {
+      "acc": 0.71048951,
+      "epoch": 0.9233891425672247,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.03910713781944e-06,
+      "loss": 1.28712883,
+      "memory(GiB)": 123.79,
+      "step": 36400,
+      "train_speed(iter/s)": 1.316669
+    },
+    {
+      "acc": 0.6769659,
+      "epoch": 0.9235159817351598,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.038081386124216e-06,
+      "loss": 1.41603317,
+      "memory(GiB)": 123.79,
+      "step": 36405,
+      "train_speed(iter/s)": 1.31669
+    },
+    {
+      "acc": 0.69245434,
+      "epoch": 0.9236428209030949,
+      "grad_norm": 3.125,
+      "learning_rate": 6.037055588769539e-06,
+      "loss": 1.35684376,
+      "memory(GiB)": 123.79,
+      "step": 36410,
+      "train_speed(iter/s)": 1.31671
+    },
+    {
+      "acc": 0.69039536,
+      "epoch": 0.9237696600710299,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.036029745800527e-06,
+      "loss": 1.34510393,
+      "memory(GiB)": 123.79,
+      "step": 36415,
+      "train_speed(iter/s)": 1.316729
+    },
+    {
+      "acc": 0.68766918,
+      "epoch": 0.923896499238965,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.0350038572623e-06,
+      "loss": 1.46024961,
+      "memory(GiB)": 123.79,
+      "step": 36420,
+      "train_speed(iter/s)": 1.316749
+    },
+    {
+      "acc": 0.67676167,
+      "epoch": 0.9240233384069001,
+      "grad_norm": 2.875,
+      "learning_rate": 6.033977923199984e-06,
+      "loss": 1.408004,
+      "memory(GiB)": 123.79,
+      "step": 36425,
+      "train_speed(iter/s)": 1.316769
+    },
+    {
+      "acc": 0.70337038,
+      "epoch": 0.9241501775748351,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.032951943658702e-06,
+      "loss": 1.32189159,
+      "memory(GiB)": 123.79,
+      "step": 36430,
+      "train_speed(iter/s)": 1.316789
+    },
+    {
+      "acc": 0.67905889,
+      "epoch": 0.9242770167427702,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.031925918683582e-06,
+      "loss": 1.40127258,
+      "memory(GiB)": 123.79,
+      "step": 36435,
+      "train_speed(iter/s)": 1.316807
+    },
+    {
+      "acc": 0.67705631,
+      "epoch": 0.9244038559107052,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.030899848319754e-06,
+      "loss": 1.46253967,
+      "memory(GiB)": 123.79,
+      "step": 36440,
+      "train_speed(iter/s)": 1.316827
+    },
+    {
+      "acc": 0.67523909,
+      "epoch": 0.9245306950786403,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.029873732612346e-06,
+      "loss": 1.42919331,
+      "memory(GiB)": 123.79,
+      "step": 36445,
+      "train_speed(iter/s)": 1.316845
+    },
+    {
+      "acc": 0.68192148,
+      "epoch": 0.9246575342465754,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.028847571606493e-06,
+      "loss": 1.46302977,
+      "memory(GiB)": 123.79,
+      "step": 36450,
+      "train_speed(iter/s)": 1.316864
+    },
+    {
+      "acc": 0.67659783,
+      "epoch": 0.9247843734145104,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.0278213653473305e-06,
+      "loss": 1.39486752,
+      "memory(GiB)": 123.79,
+      "step": 36455,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.68534994,
+      "epoch": 0.9249112125824455,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.026795113879998e-06,
+      "loss": 1.40946226,
+      "memory(GiB)": 123.79,
+      "step": 36460,
+      "train_speed(iter/s)": 1.316904
+    },
+    {
+      "acc": 0.68326206,
+      "epoch": 0.9250380517503806,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.025768817249629e-06,
+      "loss": 1.43190737,
+      "memory(GiB)": 123.79,
+      "step": 36465,
+      "train_speed(iter/s)": 1.316925
+    },
+    {
+      "acc": 0.6639061,
+      "epoch": 0.9251648909183156,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.024742475501369e-06,
+      "loss": 1.48257217,
+      "memory(GiB)": 123.79,
+      "step": 36470,
+      "train_speed(iter/s)": 1.316946
+    },
+    {
+      "acc": 0.68441529,
+      "epoch": 0.9252917300862507,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.023716088680359e-06,
+      "loss": 1.39664984,
+      "memory(GiB)": 123.79,
+      "step": 36475,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.68642664,
+      "epoch": 0.9254185692541856,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.022689656831746e-06,
+      "loss": 1.31643333,
+      "memory(GiB)": 123.79,
+      "step": 36480,
+      "train_speed(iter/s)": 1.316986
+    },
+    {
+      "acc": 0.68786783,
+      "epoch": 0.9255454084221207,
+      "grad_norm": 4.5,
+      "learning_rate": 6.021663180000675e-06,
+      "loss": 1.35379858,
+      "memory(GiB)": 123.79,
+      "step": 36485,
+      "train_speed(iter/s)": 1.317006
+    },
+    {
+      "acc": 0.68203893,
+      "epoch": 0.9256722475900558,
+      "grad_norm": 3.125,
+      "learning_rate": 6.020636658232297e-06,
+      "loss": 1.40203609,
+      "memory(GiB)": 123.79,
+      "step": 36490,
+      "train_speed(iter/s)": 1.317026
+    },
+    {
+      "acc": 0.68243828,
+      "epoch": 0.9257990867579908,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.019610091571762e-06,
+      "loss": 1.43577881,
+      "memory(GiB)": 123.79,
+      "step": 36495,
+      "train_speed(iter/s)": 1.317047
+    },
+    {
+      "acc": 0.67897921,
+      "epoch": 0.9259259259259259,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.018583480064222e-06,
+      "loss": 1.45136642,
+      "memory(GiB)": 123.79,
+      "step": 36500,
+      "train_speed(iter/s)": 1.317066
+    },
+    {
+      "acc": 0.68626032,
+      "epoch": 0.926052765093861,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.017556823754833e-06,
+      "loss": 1.2879508,
+      "memory(GiB)": 123.79,
+      "step": 36505,
+      "train_speed(iter/s)": 1.317087
+    },
+    {
+      "acc": 0.68812304,
+      "epoch": 0.926179604261796,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.016530122688753e-06,
+      "loss": 1.38455944,
+      "memory(GiB)": 123.79,
+      "step": 36510,
+      "train_speed(iter/s)": 1.317109
+    },
+    {
+      "acc": 0.69949455,
+      "epoch": 0.9263064434297311,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.015503376911138e-06,
+      "loss": 1.37122526,
+      "memory(GiB)": 123.79,
+      "step": 36515,
+      "train_speed(iter/s)": 1.317129
+    },
+    {
+      "acc": 0.68711739,
+      "epoch": 0.9264332825976661,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.0144765864671515e-06,
+      "loss": 1.38626499,
+      "memory(GiB)": 123.79,
+      "step": 36520,
+      "train_speed(iter/s)": 1.317149
+    },
+    {
+      "acc": 0.66060905,
+      "epoch": 0.9265601217656012,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.013449751401954e-06,
+      "loss": 1.45503473,
+      "memory(GiB)": 123.79,
+      "step": 36525,
+      "train_speed(iter/s)": 1.317169
+    },
+    {
+      "acc": 0.6931375,
+      "epoch": 0.9266869609335363,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.012422871760715e-06,
+      "loss": 1.36574936,
+      "memory(GiB)": 123.79,
+      "step": 36530,
+      "train_speed(iter/s)": 1.317189
+    },
+    {
+      "acc": 0.68710108,
+      "epoch": 0.9268138001014713,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.011395947588594e-06,
+      "loss": 1.42780781,
+      "memory(GiB)": 123.79,
+      "step": 36535,
+      "train_speed(iter/s)": 1.317208
+    },
+    {
+      "acc": 0.67331796,
+      "epoch": 0.9269406392694064,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.010368978930767e-06,
+      "loss": 1.40006561,
+      "memory(GiB)": 123.79,
+      "step": 36540,
+      "train_speed(iter/s)": 1.317228
+    },
+    {
+      "acc": 0.68327751,
+      "epoch": 0.9270674784373415,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.0093419658323995e-06,
+      "loss": 1.44965839,
+      "memory(GiB)": 123.79,
+      "step": 36545,
+      "train_speed(iter/s)": 1.317248
+    },
+    {
+      "acc": 0.67673244,
+      "epoch": 0.9271943176052765,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.0083149083386675e-06,
+      "loss": 1.49763184,
+      "memory(GiB)": 123.79,
+      "step": 36550,
+      "train_speed(iter/s)": 1.317268
+    },
+    {
+      "acc": 0.67450805,
+      "epoch": 0.9273211567732116,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.007287806494742e-06,
+      "loss": 1.43258553,
+      "memory(GiB)": 123.79,
+      "step": 36555,
+      "train_speed(iter/s)": 1.317288
+    },
+    {
+      "acc": 0.69247313,
+      "epoch": 0.9274479959411466,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.006260660345802e-06,
+      "loss": 1.38830881,
+      "memory(GiB)": 123.79,
+      "step": 36560,
+      "train_speed(iter/s)": 1.317307
+    },
+    {
+      "acc": 0.68851051,
+      "epoch": 0.9275748351090817,
+      "grad_norm": 2.75,
+      "learning_rate": 6.005233469937027e-06,
+      "loss": 1.39126749,
+      "memory(GiB)": 123.79,
+      "step": 36565,
+      "train_speed(iter/s)": 1.317328
+    },
+    {
+      "acc": 0.70367603,
+      "epoch": 0.9277016742770168,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.004206235313594e-06,
+      "loss": 1.33231859,
+      "memory(GiB)": 123.79,
+      "step": 36570,
+      "train_speed(iter/s)": 1.317347
+    },
+    {
+      "acc": 0.69035954,
+      "epoch": 0.9278285134449518,
+      "grad_norm": 2.875,
+      "learning_rate": 6.003178956520688e-06,
+      "loss": 1.37743101,
+      "memory(GiB)": 123.79,
+      "step": 36575,
+      "train_speed(iter/s)": 1.317367
+    },
+    {
+      "acc": 0.6864604,
+      "epoch": 0.9279553526128869,
+      "grad_norm": 2.421875,
+      "learning_rate": 6.002151633603493e-06,
+      "loss": 1.41311684,
+      "memory(GiB)": 123.79,
+      "step": 36580,
+      "train_speed(iter/s)": 1.317386
+    },
+    {
+      "acc": 0.68290854,
+      "epoch": 0.928082191780822,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.0011242666071945e-06,
+      "loss": 1.40669041,
+      "memory(GiB)": 123.79,
+      "step": 36585,
+      "train_speed(iter/s)": 1.317407
+    },
+    {
+      "acc": 0.6825644,
+      "epoch": 0.928209030948757,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.000096855576982e-06,
+      "loss": 1.3607584,
+      "memory(GiB)": 123.79,
+      "step": 36590,
+      "train_speed(iter/s)": 1.317425
+    },
+    {
+      "acc": 0.69153042,
+      "epoch": 0.928335870116692,
+      "grad_norm": 2.875,
+      "learning_rate": 5.999069400558044e-06,
+      "loss": 1.41368656,
+      "memory(GiB)": 123.79,
+      "step": 36595,
+      "train_speed(iter/s)": 1.317442
+    },
+    {
+      "acc": 0.68743811,
+      "epoch": 0.928462709284627,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.998041901595573e-06,
+      "loss": 1.35957108,
+      "memory(GiB)": 123.79,
+      "step": 36600,
+      "train_speed(iter/s)": 1.317462
+    },
+    {
+      "acc": 0.66534939,
+      "epoch": 0.9285895484525621,
+      "grad_norm": 3.125,
+      "learning_rate": 5.997014358734763e-06,
+      "loss": 1.42663231,
+      "memory(GiB)": 123.79,
+      "step": 36605,
+      "train_speed(iter/s)": 1.317481
+    },
+    {
+      "acc": 0.69133134,
+      "epoch": 0.9287163876204972,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.995986772020811e-06,
+      "loss": 1.33558197,
+      "memory(GiB)": 123.79,
+      "step": 36610,
+      "train_speed(iter/s)": 1.3175
+    },
+    {
+      "acc": 0.68809505,
+      "epoch": 0.9288432267884322,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.994959141498913e-06,
+      "loss": 1.39793282,
+      "memory(GiB)": 123.79,
+      "step": 36615,
+      "train_speed(iter/s)": 1.317519
+    },
+    {
+      "acc": 0.68777857,
+      "epoch": 0.9289700659563673,
+      "grad_norm": 2.625,
+      "learning_rate": 5.993931467214272e-06,
+      "loss": 1.41132202,
+      "memory(GiB)": 123.79,
+      "step": 36620,
+      "train_speed(iter/s)": 1.317539
+    },
+    {
+      "acc": 0.68217716,
+      "epoch": 0.9290969051243024,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.992903749212084e-06,
+      "loss": 1.40680695,
+      "memory(GiB)": 123.79,
+      "step": 36625,
+      "train_speed(iter/s)": 1.317559
+    },
+    {
+      "acc": 0.68101063,
+      "epoch": 0.9292237442922374,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.991875987537559e-06,
+      "loss": 1.42006245,
+      "memory(GiB)": 123.79,
+      "step": 36630,
+      "train_speed(iter/s)": 1.317578
+    },
+    {
+      "acc": 0.6865521,
+      "epoch": 0.9293505834601725,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.990848182235898e-06,
+      "loss": 1.40695038,
+      "memory(GiB)": 123.79,
+      "step": 36635,
+      "train_speed(iter/s)": 1.317598
+    },
+    {
+      "acc": 0.67613745,
+      "epoch": 0.9294774226281075,
+      "grad_norm": 3.859375,
+      "learning_rate": 5.98982033335231e-06,
+      "loss": 1.44887257,
+      "memory(GiB)": 123.79,
+      "step": 36640,
+      "train_speed(iter/s)": 1.317617
+    },
+    {
+      "acc": 0.69524555,
+      "epoch": 0.9296042617960426,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.988792440932006e-06,
+      "loss": 1.29166756,
+      "memory(GiB)": 123.79,
+      "step": 36645,
+      "train_speed(iter/s)": 1.317637
+    },
+    {
+      "acc": 0.69756627,
+      "epoch": 0.9297311009639777,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.987764505020195e-06,
+      "loss": 1.31303005,
+      "memory(GiB)": 123.79,
+      "step": 36650,
+      "train_speed(iter/s)": 1.317657
+    },
+    {
+      "acc": 0.67609005,
+      "epoch": 0.9298579401319127,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.986736525662091e-06,
+      "loss": 1.40097551,
+      "memory(GiB)": 123.79,
+      "step": 36655,
+      "train_speed(iter/s)": 1.317677
+    },
+    {
+      "acc": 0.68703375,
+      "epoch": 0.9299847792998478,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.985708502902909e-06,
+      "loss": 1.36302958,
+      "memory(GiB)": 123.79,
+      "step": 36660,
+      "train_speed(iter/s)": 1.317695
+    },
+    {
+      "acc": 0.6899157,
+      "epoch": 0.9301116184677829,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.984680436787867e-06,
+      "loss": 1.40452366,
+      "memory(GiB)": 123.79,
+      "step": 36665,
+      "train_speed(iter/s)": 1.317715
+    },
+    {
+      "acc": 0.67594767,
+      "epoch": 0.9302384576357179,
+      "grad_norm": 2.875,
+      "learning_rate": 5.983652327362182e-06,
+      "loss": 1.41634865,
+      "memory(GiB)": 123.79,
+      "step": 36670,
+      "train_speed(iter/s)": 1.317734
+    },
+    {
+      "acc": 0.68837576,
+      "epoch": 0.930365296803653,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.982624174671077e-06,
+      "loss": 1.43169632,
+      "memory(GiB)": 123.79,
+      "step": 36675,
+      "train_speed(iter/s)": 1.317754
+    },
+    {
+      "acc": 0.69324923,
+      "epoch": 0.930492135971588,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.981595978759773e-06,
+      "loss": 1.39170589,
+      "memory(GiB)": 123.79,
+      "step": 36680,
+      "train_speed(iter/s)": 1.317775
+    },
+    {
+      "acc": 0.67770538,
+      "epoch": 0.9306189751395231,
+      "grad_norm": 2.671875,
+      "learning_rate": 5.980567739673495e-06,
+      "loss": 1.44647751,
+      "memory(GiB)": 123.79,
+      "step": 36685,
+      "train_speed(iter/s)": 1.317795
+    },
+    {
+      "acc": 0.6948874,
+      "epoch": 0.9307458143074582,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.979539457457472e-06,
+      "loss": 1.39604492,
+      "memory(GiB)": 123.79,
+      "step": 36690,
+      "train_speed(iter/s)": 1.317816
+    },
+    {
+      "acc": 0.69127893,
+      "epoch": 0.9308726534753932,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.978511132156928e-06,
+      "loss": 1.40929375,
+      "memory(GiB)": 123.79,
+      "step": 36695,
+      "train_speed(iter/s)": 1.317836
+    },
+    {
+      "acc": 0.67626801,
+      "epoch": 0.9309994926433283,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.9774827638170965e-06,
+      "loss": 1.44409161,
+      "memory(GiB)": 123.79,
+      "step": 36700,
+      "train_speed(iter/s)": 1.317856
+    },
+    {
+      "acc": 0.69170337,
+      "epoch": 0.9311263318112634,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.9764543524832085e-06,
+      "loss": 1.34578428,
+      "memory(GiB)": 123.79,
+      "step": 36705,
+      "train_speed(iter/s)": 1.317876
+    },
+    {
+      "acc": 0.70245171,
+      "epoch": 0.9312531709791984,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.975425898200499e-06,
+      "loss": 1.37763948,
+      "memory(GiB)": 123.79,
+      "step": 36710,
+      "train_speed(iter/s)": 1.317896
+    },
+    {
+      "acc": 0.67931681,
+      "epoch": 0.9313800101471335,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.974397401014202e-06,
+      "loss": 1.44432631,
+      "memory(GiB)": 123.79,
+      "step": 36715,
+      "train_speed(iter/s)": 1.317916
+    },
+    {
+      "acc": 0.6812314,
+      "epoch": 0.9315068493150684,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.973368860969559e-06,
+      "loss": 1.42879877,
+      "memory(GiB)": 123.79,
+      "step": 36720,
+      "train_speed(iter/s)": 1.317936
+    },
+    {
+      "acc": 0.68567004,
+      "epoch": 0.9316336884830035,
+      "grad_norm": 3.703125,
+      "learning_rate": 5.972340278111808e-06,
+      "loss": 1.41651154,
+      "memory(GiB)": 123.79,
+      "step": 36725,
+      "train_speed(iter/s)": 1.317954
+    },
+    {
+      "acc": 0.68841724,
+      "epoch": 0.9317605276509386,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.9713116524861895e-06,
+      "loss": 1.31323652,
+      "memory(GiB)": 123.79,
+      "step": 36730,
+      "train_speed(iter/s)": 1.317973
+    },
+    {
+      "acc": 0.68991017,
+      "epoch": 0.9318873668188736,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.970282984137947e-06,
+      "loss": 1.42917385,
+      "memory(GiB)": 123.79,
+      "step": 36735,
+      "train_speed(iter/s)": 1.317989
+    },
+    {
+      "acc": 0.7071547,
+      "epoch": 0.9320142059868087,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.969254273112328e-06,
+      "loss": 1.3400034,
+      "memory(GiB)": 123.79,
+      "step": 36740,
+      "train_speed(iter/s)": 1.318009
+    },
+    {
+      "acc": 0.68486328,
+      "epoch": 0.9321410451547438,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.968225519454577e-06,
+      "loss": 1.43065748,
+      "memory(GiB)": 123.79,
+      "step": 36745,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.69507751,
+      "epoch": 0.9322678843226788,
+      "grad_norm": 3.25,
+      "learning_rate": 5.967196723209947e-06,
+      "loss": 1.36570053,
+      "memory(GiB)": 123.79,
+      "step": 36750,
+      "train_speed(iter/s)": 1.318044
+    },
+    {
+      "acc": 0.67489243,
+      "epoch": 0.9323947234906139,
+      "grad_norm": 3.5,
+      "learning_rate": 5.966167884423686e-06,
+      "loss": 1.48059464,
+      "memory(GiB)": 123.79,
+      "step": 36755,
+      "train_speed(iter/s)": 1.318063
+    },
+    {
+      "acc": 0.70791192,
+      "epoch": 0.9325215626585489,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.965139003141048e-06,
+      "loss": 1.30423994,
+      "memory(GiB)": 123.79,
+      "step": 36760,
+      "train_speed(iter/s)": 1.318084
+    },
+    {
+      "acc": 0.68336554,
+      "epoch": 0.932648401826484,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.964110079407287e-06,
+      "loss": 1.37228813,
+      "memory(GiB)": 123.79,
+      "step": 36765,
+      "train_speed(iter/s)": 1.318104
+    },
+    {
+      "acc": 0.68233619,
+      "epoch": 0.9327752409944191,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.9630811132676625e-06,
+      "loss": 1.4515666,
+      "memory(GiB)": 123.79,
+      "step": 36770,
+      "train_speed(iter/s)": 1.318124
+    },
+    {
+      "acc": 0.687358,
+      "epoch": 0.9329020801623541,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.962052104767427e-06,
+      "loss": 1.43827972,
+      "memory(GiB)": 123.79,
+      "step": 36775,
+      "train_speed(iter/s)": 1.318141
+    },
+    {
+      "acc": 0.69611549,
+      "epoch": 0.9330289193302892,
+      "grad_norm": 3.75,
+      "learning_rate": 5.961023053951848e-06,
+      "loss": 1.3893981,
+      "memory(GiB)": 123.79,
+      "step": 36780,
+      "train_speed(iter/s)": 1.31816
+    },
+    {
+      "acc": 0.69037728,
+      "epoch": 0.9331557584982243,
+      "grad_norm": 3.375,
+      "learning_rate": 5.9599939608661825e-06,
+      "loss": 1.4446044,
+      "memory(GiB)": 123.79,
+      "step": 36785,
+      "train_speed(iter/s)": 1.318178
+    },
+    {
+      "acc": 0.68350306,
+      "epoch": 0.9332825976661593,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.9589648255556975e-06,
+      "loss": 1.43527088,
+      "memory(GiB)": 123.79,
+      "step": 36790,
+      "train_speed(iter/s)": 1.318197
+    },
+    {
+      "acc": 0.68581572,
+      "epoch": 0.9334094368340944,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.957935648065658e-06,
+      "loss": 1.37167902,
+      "memory(GiB)": 123.79,
+      "step": 36795,
+      "train_speed(iter/s)": 1.318218
+    },
+    {
+      "acc": 0.68458948,
+      "epoch": 0.9335362760020294,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.956906428441331e-06,
+      "loss": 1.3909399,
+      "memory(GiB)": 123.79,
+      "step": 36800,
+      "train_speed(iter/s)": 1.318239
+    },
+    {
+      "acc": 0.68405027,
+      "epoch": 0.9336631151699645,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.955877166727988e-06,
+      "loss": 1.41275635,
+      "memory(GiB)": 123.79,
+      "step": 36805,
+      "train_speed(iter/s)": 1.318259
+    },
+    {
+      "acc": 0.68674393,
+      "epoch": 0.9337899543378996,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.954847862970898e-06,
+      "loss": 1.38463478,
+      "memory(GiB)": 123.79,
+      "step": 36810,
+      "train_speed(iter/s)": 1.31828
+    },
+    {
+      "acc": 0.67705431,
+      "epoch": 0.9339167935058346,
+      "grad_norm": 2.625,
+      "learning_rate": 5.953818517215338e-06,
+      "loss": 1.45176468,
+      "memory(GiB)": 123.79,
+      "step": 36815,
+      "train_speed(iter/s)": 1.318301
+    },
+    {
+      "acc": 0.69115229,
+      "epoch": 0.9340436326737697,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.95278912950658e-06,
+      "loss": 1.39590759,
+      "memory(GiB)": 123.79,
+      "step": 36820,
+      "train_speed(iter/s)": 1.318321
+    },
+    {
+      "acc": 0.70703135,
+      "epoch": 0.9341704718417048,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.9517596998899e-06,
+      "loss": 1.30549536,
+      "memory(GiB)": 123.79,
+      "step": 36825,
+      "train_speed(iter/s)": 1.318342
+    },
+    {
+      "acc": 0.69810109,
+      "epoch": 0.9342973110096398,
+      "grad_norm": 3.25,
+      "learning_rate": 5.9507302284105836e-06,
+      "loss": 1.36235352,
+      "memory(GiB)": 123.79,
+      "step": 36830,
+      "train_speed(iter/s)": 1.318361
+    },
+    {
+      "acc": 0.69490857,
+      "epoch": 0.9344241501775749,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.949700715113904e-06,
+      "loss": 1.40266781,
+      "memory(GiB)": 123.79,
+      "step": 36835,
+      "train_speed(iter/s)": 1.318382
+    },
+    {
+      "acc": 0.68624873,
+      "epoch": 0.9345509893455098,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.9486711600451484e-06,
+      "loss": 1.36912422,
+      "memory(GiB)": 123.79,
+      "step": 36840,
+      "train_speed(iter/s)": 1.318402
+    },
+    {
+      "acc": 0.6917706,
+      "epoch": 0.9346778285134449,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.9476415632495974e-06,
+      "loss": 1.39415474,
+      "memory(GiB)": 123.79,
+      "step": 36845,
+      "train_speed(iter/s)": 1.318423
+    },
+    {
+      "acc": 0.67632475,
+      "epoch": 0.93480466768138,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.946611924772542e-06,
+      "loss": 1.40250587,
+      "memory(GiB)": 123.79,
+      "step": 36850,
+      "train_speed(iter/s)": 1.318443
+    },
+    {
+      "acc": 0.69609575,
+      "epoch": 0.934931506849315,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.945582244659267e-06,
+      "loss": 1.36077147,
+      "memory(GiB)": 123.79,
+      "step": 36855,
+      "train_speed(iter/s)": 1.318463
+    },
+    {
+      "acc": 0.69040942,
+      "epoch": 0.9350583460172501,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.944552522955063e-06,
+      "loss": 1.41363354,
+      "memory(GiB)": 123.79,
+      "step": 36860,
+      "train_speed(iter/s)": 1.318485
+    },
+    {
+      "acc": 0.7047677,
+      "epoch": 0.9351851851851852,
+      "grad_norm": 2.46875,
+      "learning_rate": 5.943522759705221e-06,
+      "loss": 1.35777655,
+      "memory(GiB)": 123.79,
+      "step": 36865,
+      "train_speed(iter/s)": 1.318505
+    },
+    {
+      "acc": 0.67603159,
+      "epoch": 0.9353120243531202,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.942492954955037e-06,
+      "loss": 1.45803947,
+      "memory(GiB)": 123.79,
+      "step": 36870,
+      "train_speed(iter/s)": 1.318527
+    },
+    {
+      "acc": 0.6783433,
+      "epoch": 0.9354388635210553,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.941463108749804e-06,
+      "loss": 1.44300747,
+      "memory(GiB)": 123.79,
+      "step": 36875,
+      "train_speed(iter/s)": 1.318548
+    },
+    {
+      "acc": 0.68462887,
+      "epoch": 0.9355657026889903,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.940433221134821e-06,
+      "loss": 1.40803909,
+      "memory(GiB)": 123.79,
+      "step": 36880,
+      "train_speed(iter/s)": 1.318569
+    },
+    {
+      "acc": 0.69451895,
+      "epoch": 0.9356925418569254,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.9394032921553856e-06,
+      "loss": 1.41175346,
+      "memory(GiB)": 123.79,
+      "step": 36885,
+      "train_speed(iter/s)": 1.318589
+    },
+    {
+      "acc": 0.68293324,
+      "epoch": 0.9358193810248605,
+      "grad_norm": 3.5,
+      "learning_rate": 5.9383733218568e-06,
+      "loss": 1.44731388,
+      "memory(GiB)": 123.79,
+      "step": 36890,
+      "train_speed(iter/s)": 1.31861
+    },
+    {
+      "acc": 0.68460226,
+      "epoch": 0.9359462201927955,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.937343310284365e-06,
+      "loss": 1.43567486,
+      "memory(GiB)": 123.79,
+      "step": 36895,
+      "train_speed(iter/s)": 1.318632
+    },
+    {
+      "acc": 0.68424492,
+      "epoch": 0.9360730593607306,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.936313257483387e-06,
+      "loss": 1.38530006,
+      "memory(GiB)": 123.79,
+      "step": 36900,
+      "train_speed(iter/s)": 1.318653
+    },
+    {
+      "acc": 0.68103204,
+      "epoch": 0.9361998985286657,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.935283163499171e-06,
+      "loss": 1.4432909,
+      "memory(GiB)": 123.79,
+      "step": 36905,
+      "train_speed(iter/s)": 1.318674
+    },
+    {
+      "acc": 0.6730607,
+      "epoch": 0.9363267376966007,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.9342530283770274e-06,
+      "loss": 1.43960238,
+      "memory(GiB)": 123.79,
+      "step": 36910,
+      "train_speed(iter/s)": 1.318694
+    },
+    {
+      "acc": 0.69758282,
+      "epoch": 0.9364535768645358,
+      "grad_norm": 2.546875,
+      "learning_rate": 5.9332228521622615e-06,
+      "loss": 1.29135399,
+      "memory(GiB)": 123.79,
+      "step": 36915,
+      "train_speed(iter/s)": 1.318714
+    },
+    {
+      "acc": 0.70007606,
+      "epoch": 0.9365804160324708,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.93219263490019e-06,
+      "loss": 1.40289488,
+      "memory(GiB)": 123.79,
+      "step": 36920,
+      "train_speed(iter/s)": 1.318735
+    },
+    {
+      "acc": 0.6930418,
+      "epoch": 0.9367072552004059,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.931162376636123e-06,
+      "loss": 1.46011848,
+      "memory(GiB)": 123.79,
+      "step": 36925,
+      "train_speed(iter/s)": 1.318754
+    },
+    {
+      "acc": 0.68778586,
+      "epoch": 0.936834094368341,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.93013207741538e-06,
+      "loss": 1.35796995,
+      "memory(GiB)": 123.79,
+      "step": 36930,
+      "train_speed(iter/s)": 1.318775
+    },
+    {
+      "acc": 0.69858589,
+      "epoch": 0.936960933536276,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.929101737283274e-06,
+      "loss": 1.34122429,
+      "memory(GiB)": 123.79,
+      "step": 36935,
+      "train_speed(iter/s)": 1.318796
+    },
+    {
+      "acc": 0.6731782,
+      "epoch": 0.9370877727042111,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.928071356285126e-06,
+      "loss": 1.38488111,
+      "memory(GiB)": 123.79,
+      "step": 36940,
+      "train_speed(iter/s)": 1.318817
+    },
+    {
+      "acc": 0.70338373,
+      "epoch": 0.9372146118721462,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.927040934466255e-06,
+      "loss": 1.34964027,
+      "memory(GiB)": 123.79,
+      "step": 36945,
+      "train_speed(iter/s)": 1.318838
+    },
+    {
+      "acc": 0.69995165,
+      "epoch": 0.9373414510400812,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.926010471871986e-06,
+      "loss": 1.3561223,
+      "memory(GiB)": 123.79,
+      "step": 36950,
+      "train_speed(iter/s)": 1.318859
+    },
+    {
+      "acc": 0.69651699,
+      "epoch": 0.9374682902080163,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.924979968547642e-06,
+      "loss": 1.32209539,
+      "memory(GiB)": 123.79,
+      "step": 36955,
+      "train_speed(iter/s)": 1.31888
+    },
+    {
+      "acc": 0.66502748,
+      "epoch": 0.9375951293759512,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.9239494245385485e-06,
+      "loss": 1.50459023,
+      "memory(GiB)": 123.79,
+      "step": 36960,
+      "train_speed(iter/s)": 1.318901
+    },
+    {
+      "acc": 0.69413552,
+      "epoch": 0.9377219685438863,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.9229188398900325e-06,
+      "loss": 1.41543999,
+      "memory(GiB)": 123.79,
+      "step": 36965,
+      "train_speed(iter/s)": 1.318922
+    },
+    {
+      "acc": 0.67297468,
+      "epoch": 0.9378488077118214,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.921888214647429e-06,
+      "loss": 1.45201197,
+      "memory(GiB)": 123.79,
+      "step": 36970,
+      "train_speed(iter/s)": 1.318942
+    },
+    {
+      "acc": 0.6873383,
+      "epoch": 0.9379756468797564,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.920857548856064e-06,
+      "loss": 1.36569777,
+      "memory(GiB)": 123.79,
+      "step": 36975,
+      "train_speed(iter/s)": 1.318963
+    },
+    {
+      "acc": 0.67968059,
+      "epoch": 0.9381024860476915,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.919826842561274e-06,
+      "loss": 1.38920383,
+      "memory(GiB)": 123.79,
+      "step": 36980,
+      "train_speed(iter/s)": 1.318984
+    },
+    {
+      "acc": 0.68382797,
+      "epoch": 0.9382293252156266,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.91879609580839e-06,
+      "loss": 1.43278046,
+      "memory(GiB)": 123.79,
+      "step": 36985,
+      "train_speed(iter/s)": 1.319004
+    },
+    {
+      "acc": 0.67751307,
+      "epoch": 0.9383561643835616,
+      "grad_norm": 4.375,
+      "learning_rate": 5.917765308642754e-06,
+      "loss": 1.46014423,
+      "memory(GiB)": 123.79,
+      "step": 36990,
+      "train_speed(iter/s)": 1.319025
+    },
+    {
+      "acc": 0.68507648,
+      "epoch": 0.9384830035514967,
+      "grad_norm": 4.78125,
+      "learning_rate": 5.9167344811097014e-06,
+      "loss": 1.4548955,
+      "memory(GiB)": 123.79,
+      "step": 36995,
+      "train_speed(iter/s)": 1.319046
+    },
+    {
+      "acc": 0.69449043,
+      "epoch": 0.9386098427194317,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.9157036132545735e-06,
+      "loss": 1.33846302,
+      "memory(GiB)": 123.79,
+      "step": 37000,
+      "train_speed(iter/s)": 1.319067
+    },
+    {
+      "epoch": 0.9386098427194317,
+      "eval_acc": 0.674868280091969,
+      "eval_loss": 1.3572280406951904,
+      "eval_runtime": 69.296,
+      "eval_samples_per_second": 91.925,
+      "eval_steps_per_second": 22.988,
+      "step": 37000
+    },
+    {
+      "acc": 0.69752517,
+      "epoch": 0.9387366818873668,
+      "grad_norm": 3.875,
+      "learning_rate": 5.914672705122713e-06,
+      "loss": 1.39682541,
+      "memory(GiB)": 123.79,
+      "step": 37005,
+      "train_speed(iter/s)": 1.315245
+    },
+    {
+      "acc": 0.68128109,
+      "epoch": 0.9388635210553019,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.9136417567594615e-06,
+      "loss": 1.38434458,
+      "memory(GiB)": 123.79,
+      "step": 37010,
+      "train_speed(iter/s)": 1.315265
+    },
+    {
+      "acc": 0.68871279,
+      "epoch": 0.9389903602232369,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.9126107682101675e-06,
+      "loss": 1.43167648,
+      "memory(GiB)": 123.79,
+      "step": 37015,
+      "train_speed(iter/s)": 1.315287
+    },
+    {
+      "acc": 0.6638793,
+      "epoch": 0.939117199391172,
+      "grad_norm": 2.328125,
+      "learning_rate": 5.911579739520178e-06,
+      "loss": 1.42589207,
+      "memory(GiB)": 123.79,
+      "step": 37020,
+      "train_speed(iter/s)": 1.315307
+    },
+    {
+      "acc": 0.67444944,
+      "epoch": 0.9392440385591071,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.91054867073484e-06,
+      "loss": 1.43461094,
+      "memory(GiB)": 123.79,
+      "step": 37025,
+      "train_speed(iter/s)": 1.315327
+    },
+    {
+      "acc": 0.67173038,
+      "epoch": 0.9393708777270421,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.909517561899508e-06,
+      "loss": 1.44078751,
+      "memory(GiB)": 123.79,
+      "step": 37030,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.65186548,
+      "epoch": 0.9394977168949772,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.908486413059532e-06,
+      "loss": 1.44570751,
+      "memory(GiB)": 123.79,
+      "step": 37035,
+      "train_speed(iter/s)": 1.315369
+    },
+    {
+      "acc": 0.68192453,
+      "epoch": 0.9396245560629122,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.907455224260268e-06,
+      "loss": 1.39854259,
+      "memory(GiB)": 123.79,
+      "step": 37040,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.67956204,
+      "epoch": 0.9397513952308473,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.9064239955470704e-06,
+      "loss": 1.40723305,
+      "memory(GiB)": 123.79,
+      "step": 37045,
+      "train_speed(iter/s)": 1.31541
+    },
+    {
+      "acc": 0.659548,
+      "epoch": 0.9398782343987824,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.9053927269653e-06,
+      "loss": 1.49778538,
+      "memory(GiB)": 123.79,
+      "step": 37050,
+      "train_speed(iter/s)": 1.315431
+    },
+    {
+      "acc": 0.69435682,
+      "epoch": 0.9400050735667174,
+      "grad_norm": 3.125,
+      "learning_rate": 5.904361418560314e-06,
+      "loss": 1.37466507,
+      "memory(GiB)": 123.79,
+      "step": 37055,
+      "train_speed(iter/s)": 1.315444
+    },
+    {
+      "acc": 0.69586782,
+      "epoch": 0.9401319127346525,
+      "grad_norm": 3.703125,
+      "learning_rate": 5.903330070377477e-06,
+      "loss": 1.41261406,
+      "memory(GiB)": 123.79,
+      "step": 37060,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.69350214,
+      "epoch": 0.9402587519025876,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.902298682462147e-06,
+      "loss": 1.30958376,
+      "memory(GiB)": 123.79,
+      "step": 37065,
+      "train_speed(iter/s)": 1.315485
+    },
+    {
+      "acc": 0.69621124,
+      "epoch": 0.9403855910705226,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.901267254859695e-06,
+      "loss": 1.38569794,
+      "memory(GiB)": 123.79,
+      "step": 37070,
+      "train_speed(iter/s)": 1.315505
+    },
+    {
+      "acc": 0.67851849,
+      "epoch": 0.9405124302384577,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.900235787615485e-06,
+      "loss": 1.47394304,
+      "memory(GiB)": 123.79,
+      "step": 37075,
+      "train_speed(iter/s)": 1.315526
+    },
+    {
+      "acc": 0.68956919,
+      "epoch": 0.9406392694063926,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.8992042807748866e-06,
+      "loss": 1.3819068,
+      "memory(GiB)": 123.79,
+      "step": 37080,
+      "train_speed(iter/s)": 1.315547
+    },
+    {
+      "acc": 0.68416872,
+      "epoch": 0.9407661085743277,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.898172734383267e-06,
+      "loss": 1.42304382,
+      "memory(GiB)": 123.79,
+      "step": 37085,
+      "train_speed(iter/s)": 1.315568
+    },
+    {
+      "acc": 0.67436199,
+      "epoch": 0.9408929477422628,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.897141148486003e-06,
+      "loss": 1.45963612,
+      "memory(GiB)": 123.79,
+      "step": 37090,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.69305983,
+      "epoch": 0.9410197869101978,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.8961095231284645e-06,
+      "loss": 1.35458298,
+      "memory(GiB)": 123.79,
+      "step": 37095,
+      "train_speed(iter/s)": 1.315608
+    },
+    {
+      "acc": 0.6806601,
+      "epoch": 0.9411466260781329,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.895077858356029e-06,
+      "loss": 1.40429535,
+      "memory(GiB)": 123.79,
+      "step": 37100,
+      "train_speed(iter/s)": 1.315629
+    },
+    {
+      "acc": 0.68260193,
+      "epoch": 0.941273465246068,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.8940461542140725e-06,
+      "loss": 1.38190308,
+      "memory(GiB)": 123.79,
+      "step": 37105,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.70031099,
+      "epoch": 0.941400304414003,
+      "grad_norm": 4.25,
+      "learning_rate": 5.893014410747975e-06,
+      "loss": 1.3497344,
+      "memory(GiB)": 123.79,
+      "step": 37110,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.69482207,
+      "epoch": 0.9415271435819381,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.891982628003114e-06,
+      "loss": 1.38212738,
+      "memory(GiB)": 123.79,
+      "step": 37115,
+      "train_speed(iter/s)": 1.315692
+    },
+    {
+      "acc": 0.68122358,
+      "epoch": 0.9416539827498731,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.890950806024879e-06,
+      "loss": 1.39410238,
+      "memory(GiB)": 123.79,
+      "step": 37120,
+      "train_speed(iter/s)": 1.315713
+    },
+    {
+      "acc": 0.68970308,
+      "epoch": 0.9417808219178082,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.889918944858647e-06,
+      "loss": 1.42260752,
+      "memory(GiB)": 123.79,
+      "step": 37125,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.69887352,
+      "epoch": 0.9419076610857433,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.888887044549808e-06,
+      "loss": 1.34509144,
+      "memory(GiB)": 123.79,
+      "step": 37130,
+      "train_speed(iter/s)": 1.315755
+    },
+    {
+      "acc": 0.6748693,
+      "epoch": 0.9420345002536783,
+      "grad_norm": 3.125,
+      "learning_rate": 5.887855105143746e-06,
+      "loss": 1.40342197,
+      "memory(GiB)": 123.79,
+      "step": 37135,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.68755007,
+      "epoch": 0.9421613394216134,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.886823126685855e-06,
+      "loss": 1.36548576,
+      "memory(GiB)": 123.79,
+      "step": 37140,
+      "train_speed(iter/s)": 1.315797
+    },
+    {
+      "acc": 0.68484039,
+      "epoch": 0.9422881785895485,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.8857911092215214e-06,
+      "loss": 1.45654783,
+      "memory(GiB)": 123.79,
+      "step": 37145,
+      "train_speed(iter/s)": 1.315819
+    },
+    {
+      "acc": 0.67900248,
+      "epoch": 0.9424150177574835,
+      "grad_norm": 3.0,
+      "learning_rate": 5.884759052796142e-06,
+      "loss": 1.46269112,
+      "memory(GiB)": 123.79,
+      "step": 37150,
+      "train_speed(iter/s)": 1.31584
+    },
+    {
+      "acc": 0.68602777,
+      "epoch": 0.9425418569254186,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.883726957455108e-06,
+      "loss": 1.39906807,
+      "memory(GiB)": 123.79,
+      "step": 37155,
+      "train_speed(iter/s)": 1.315861
+    },
+    {
+      "acc": 0.67911787,
+      "epoch": 0.9426686960933536,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.8826948232438176e-06,
+      "loss": 1.41652336,
+      "memory(GiB)": 123.79,
+      "step": 37160,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.69276967,
+      "epoch": 0.9427955352612887,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.881662650207667e-06,
+      "loss": 1.36155815,
+      "memory(GiB)": 123.79,
+      "step": 37165,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.70288105,
+      "epoch": 0.9429223744292238,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.880630438392057e-06,
+      "loss": 1.3575531,
+      "memory(GiB)": 123.79,
+      "step": 37170,
+      "train_speed(iter/s)": 1.315926
+    },
+    {
+      "acc": 0.6923728,
+      "epoch": 0.9430492135971588,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.879598187842389e-06,
+      "loss": 1.35164089,
+      "memory(GiB)": 123.79,
+      "step": 37175,
+      "train_speed(iter/s)": 1.315947
+    },
+    {
+      "acc": 0.69735003,
+      "epoch": 0.9431760527650939,
+      "grad_norm": 5.4375,
+      "learning_rate": 5.878565898604066e-06,
+      "loss": 1.40482578,
+      "memory(GiB)": 123.79,
+      "step": 37180,
+      "train_speed(iter/s)": 1.315968
+    },
+    {
+      "acc": 0.69735298,
+      "epoch": 0.943302891933029,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.87753357072249e-06,
+      "loss": 1.31006765,
+      "memory(GiB)": 123.79,
+      "step": 37185,
+      "train_speed(iter/s)": 1.315987
+    },
+    {
+      "acc": 0.69428692,
+      "epoch": 0.943429731100964,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.876501204243072e-06,
+      "loss": 1.38061028,
+      "memory(GiB)": 123.79,
+      "step": 37190,
+      "train_speed(iter/s)": 1.316008
+    },
+    {
+      "acc": 0.69963503,
+      "epoch": 0.943556570268899,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.875468799211217e-06,
+      "loss": 1.34880219,
+      "memory(GiB)": 123.79,
+      "step": 37195,
+      "train_speed(iter/s)": 1.316029
+    },
+    {
+      "acc": 0.68292885,
+      "epoch": 0.943683409436834,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.874436355672337e-06,
+      "loss": 1.33872843,
+      "memory(GiB)": 123.79,
+      "step": 37200,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.68404498,
+      "epoch": 0.9438102486047691,
+      "grad_norm": 4.375,
+      "learning_rate": 5.873403873671839e-06,
+      "loss": 1.41408119,
+      "memory(GiB)": 123.79,
+      "step": 37205,
+      "train_speed(iter/s)": 1.316072
+    },
+    {
+      "acc": 0.68770876,
+      "epoch": 0.9439370877727042,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.872371353255142e-06,
+      "loss": 1.42262135,
+      "memory(GiB)": 123.79,
+      "step": 37210,
+      "train_speed(iter/s)": 1.316093
+    },
+    {
+      "acc": 0.69606957,
+      "epoch": 0.9440639269406392,
+      "grad_norm": 3.375,
+      "learning_rate": 5.871338794467656e-06,
+      "loss": 1.38161469,
+      "memory(GiB)": 123.79,
+      "step": 37215,
+      "train_speed(iter/s)": 1.316114
+    },
+    {
+      "acc": 0.67975569,
+      "epoch": 0.9441907661085743,
+      "grad_norm": 3.0,
+      "learning_rate": 5.8703061973548e-06,
+      "loss": 1.37598476,
+      "memory(GiB)": 123.79,
+      "step": 37220,
+      "train_speed(iter/s)": 1.316135
+    },
+    {
+      "acc": 0.68895764,
+      "epoch": 0.9443176052765094,
+      "grad_norm": 2.421875,
+      "learning_rate": 5.869273561961992e-06,
+      "loss": 1.3793355,
+      "memory(GiB)": 123.79,
+      "step": 37225,
+      "train_speed(iter/s)": 1.316156
+    },
+    {
+      "acc": 0.69125128,
+      "epoch": 0.9444444444444444,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.8682408883346535e-06,
+      "loss": 1.4017952,
+      "memory(GiB)": 123.79,
+      "step": 37230,
+      "train_speed(iter/s)": 1.316176
+    },
+    {
+      "acc": 0.69225883,
+      "epoch": 0.9445712836123795,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.867208176518202e-06,
+      "loss": 1.36517277,
+      "memory(GiB)": 123.79,
+      "step": 37235,
+      "train_speed(iter/s)": 1.316197
+    },
+    {
+      "acc": 0.69379807,
+      "epoch": 0.9446981227803145,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.866175426558064e-06,
+      "loss": 1.35373077,
+      "memory(GiB)": 123.79,
+      "step": 37240,
+      "train_speed(iter/s)": 1.316215
+    },
+    {
+      "acc": 0.68542705,
+      "epoch": 0.9448249619482496,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.865142638499664e-06,
+      "loss": 1.45035315,
+      "memory(GiB)": 123.79,
+      "step": 37245,
+      "train_speed(iter/s)": 1.316235
+    },
+    {
+      "acc": 0.6681797,
+      "epoch": 0.9449518011161847,
+      "grad_norm": 3.125,
+      "learning_rate": 5.864109812388426e-06,
+      "loss": 1.52961941,
+      "memory(GiB)": 123.79,
+      "step": 37250,
+      "train_speed(iter/s)": 1.316256
+    },
+    {
+      "acc": 0.67980318,
+      "epoch": 0.9450786402841197,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.863076948269782e-06,
+      "loss": 1.38385859,
+      "memory(GiB)": 123.79,
+      "step": 37255,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.69081521,
+      "epoch": 0.9452054794520548,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.862044046189162e-06,
+      "loss": 1.40432091,
+      "memory(GiB)": 123.79,
+      "step": 37260,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.69936814,
+      "epoch": 0.9453323186199899,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.8610111061919924e-06,
+      "loss": 1.39102573,
+      "memory(GiB)": 123.79,
+      "step": 37265,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.67718911,
+      "epoch": 0.9454591577879249,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.859978128323713e-06,
+      "loss": 1.40576315,
+      "memory(GiB)": 123.79,
+      "step": 37270,
+      "train_speed(iter/s)": 1.316336
+    },
+    {
+      "acc": 0.69196134,
+      "epoch": 0.94558599695586,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.858945112629755e-06,
+      "loss": 1.37837677,
+      "memory(GiB)": 123.79,
+      "step": 37275,
+      "train_speed(iter/s)": 1.316355
+    },
+    {
+      "acc": 0.69408593,
+      "epoch": 0.945712836123795,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.857912059155557e-06,
+      "loss": 1.38230448,
+      "memory(GiB)": 123.79,
+      "step": 37280,
+      "train_speed(iter/s)": 1.316374
+    },
+    {
+      "acc": 0.70130873,
+      "epoch": 0.9458396752917301,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.856878967946555e-06,
+      "loss": 1.34562111,
+      "memory(GiB)": 123.79,
+      "step": 37285,
+      "train_speed(iter/s)": 1.316394
+    },
+    {
+      "acc": 0.68106551,
+      "epoch": 0.9459665144596652,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.855845839048191e-06,
+      "loss": 1.41055508,
+      "memory(GiB)": 123.79,
+      "step": 37290,
+      "train_speed(iter/s)": 1.316413
+    },
+    {
+      "acc": 0.68750019,
+      "epoch": 0.9460933536276002,
+      "grad_norm": 3.0,
+      "learning_rate": 5.854812672505906e-06,
+      "loss": 1.46424217,
+      "memory(GiB)": 123.79,
+      "step": 37295,
+      "train_speed(iter/s)": 1.316433
+    },
+    {
+      "acc": 0.69106331,
+      "epoch": 0.9462201927955353,
+      "grad_norm": 3.375,
+      "learning_rate": 5.853779468365144e-06,
+      "loss": 1.3487175,
+      "memory(GiB)": 123.79,
+      "step": 37300,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.68163533,
+      "epoch": 0.9463470319634704,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.852746226671348e-06,
+      "loss": 1.4600976,
+      "memory(GiB)": 123.79,
+      "step": 37305,
+      "train_speed(iter/s)": 1.316472
+    },
+    {
+      "acc": 0.68320622,
+      "epoch": 0.9464738711314054,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.851712947469966e-06,
+      "loss": 1.40190773,
+      "memory(GiB)": 123.79,
+      "step": 37310,
+      "train_speed(iter/s)": 1.316493
+    },
+    {
+      "acc": 0.67794237,
+      "epoch": 0.9466007102993405,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.850679630806446e-06,
+      "loss": 1.44427137,
+      "memory(GiB)": 123.79,
+      "step": 37315,
+      "train_speed(iter/s)": 1.316509
+    },
+    {
+      "acc": 0.69341707,
+      "epoch": 0.9467275494672754,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.849646276726237e-06,
+      "loss": 1.40909176,
+      "memory(GiB)": 123.79,
+      "step": 37320,
+      "train_speed(iter/s)": 1.316528
+    },
+    {
+      "acc": 0.68388119,
+      "epoch": 0.9468543886352105,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.848612885274792e-06,
+      "loss": 1.43773994,
+      "memory(GiB)": 123.79,
+      "step": 37325,
+      "train_speed(iter/s)": 1.316544
+    },
+    {
+      "acc": 0.67464156,
+      "epoch": 0.9469812278031456,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.847579456497564e-06,
+      "loss": 1.45758963,
+      "memory(GiB)": 123.79,
+      "step": 37330,
+      "train_speed(iter/s)": 1.316563
+    },
+    {
+      "acc": 0.68607259,
+      "epoch": 0.9471080669710806,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.8465459904400065e-06,
+      "loss": 1.34098177,
+      "memory(GiB)": 123.79,
+      "step": 37335,
+      "train_speed(iter/s)": 1.316582
+    },
+    {
+      "acc": 0.69474697,
+      "epoch": 0.9472349061390157,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.845512487147579e-06,
+      "loss": 1.40921354,
+      "memory(GiB)": 123.79,
+      "step": 37340,
+      "train_speed(iter/s)": 1.316602
+    },
+    {
+      "acc": 0.68666101,
+      "epoch": 0.9473617453069508,
+      "grad_norm": 3.5,
+      "learning_rate": 5.844478946665733e-06,
+      "loss": 1.39299297,
+      "memory(GiB)": 123.79,
+      "step": 37345,
+      "train_speed(iter/s)": 1.316621
+    },
+    {
+      "acc": 0.68826833,
+      "epoch": 0.9474885844748858,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.843445369039937e-06,
+      "loss": 1.45222263,
+      "memory(GiB)": 123.79,
+      "step": 37350,
+      "train_speed(iter/s)": 1.31664
+    },
+    {
+      "acc": 0.70514908,
+      "epoch": 0.9476154236428209,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.842411754315645e-06,
+      "loss": 1.30940437,
+      "memory(GiB)": 123.79,
+      "step": 37355,
+      "train_speed(iter/s)": 1.31666
+    },
+    {
+      "acc": 0.68109345,
+      "epoch": 0.9477422628107559,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.841378102538324e-06,
+      "loss": 1.4642417,
+      "memory(GiB)": 123.79,
+      "step": 37360,
+      "train_speed(iter/s)": 1.316679
+    },
+    {
+      "acc": 0.69463801,
+      "epoch": 0.947869101978691,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.840344413753438e-06,
+      "loss": 1.35907822,
+      "memory(GiB)": 123.79,
+      "step": 37365,
+      "train_speed(iter/s)": 1.316698
+    },
+    {
+      "acc": 0.67992096,
+      "epoch": 0.9479959411466261,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.8393106880064535e-06,
+      "loss": 1.38604412,
+      "memory(GiB)": 123.79,
+      "step": 37370,
+      "train_speed(iter/s)": 1.316718
+    },
+    {
+      "acc": 0.68163424,
+      "epoch": 0.9481227803145611,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.838276925342836e-06,
+      "loss": 1.36222801,
+      "memory(GiB)": 123.79,
+      "step": 37375,
+      "train_speed(iter/s)": 1.316737
+    },
+    {
+      "acc": 0.6866992,
+      "epoch": 0.9482496194824962,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.837243125808058e-06,
+      "loss": 1.36050243,
+      "memory(GiB)": 123.79,
+      "step": 37380,
+      "train_speed(iter/s)": 1.316756
+    },
+    {
+      "acc": 0.67065778,
+      "epoch": 0.9483764586504313,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.8362092894475886e-06,
+      "loss": 1.47752199,
+      "memory(GiB)": 123.79,
+      "step": 37385,
+      "train_speed(iter/s)": 1.316775
+    },
+    {
+      "acc": 0.68647079,
+      "epoch": 0.9485032978183663,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.835175416306901e-06,
+      "loss": 1.41084995,
+      "memory(GiB)": 123.79,
+      "step": 37390,
+      "train_speed(iter/s)": 1.316795
+    },
+    {
+      "acc": 0.69186354,
+      "epoch": 0.9486301369863014,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.83414150643147e-06,
+      "loss": 1.32847099,
+      "memory(GiB)": 123.79,
+      "step": 37395,
+      "train_speed(iter/s)": 1.316814
+    },
+    {
+      "acc": 0.66915474,
+      "epoch": 0.9487569761542364,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.833107559866772e-06,
+      "loss": 1.32493839,
+      "memory(GiB)": 123.79,
+      "step": 37400,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.68919201,
+      "epoch": 0.9488838153221715,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.832073576658282e-06,
+      "loss": 1.39776506,
+      "memory(GiB)": 123.79,
+      "step": 37405,
+      "train_speed(iter/s)": 1.316852
+    },
+    {
+      "acc": 0.69064126,
+      "epoch": 0.9490106544901066,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.831039556851485e-06,
+      "loss": 1.38163929,
+      "memory(GiB)": 123.79,
+      "step": 37410,
+      "train_speed(iter/s)": 1.316872
+    },
+    {
+      "acc": 0.678794,
+      "epoch": 0.9491374936580416,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.8300055004918535e-06,
+      "loss": 1.39741802,
+      "memory(GiB)": 123.79,
+      "step": 37415,
+      "train_speed(iter/s)": 1.31689
+    },
+    {
+      "acc": 0.68004994,
+      "epoch": 0.9492643328259767,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.828971407624877e-06,
+      "loss": 1.43640223,
+      "memory(GiB)": 123.79,
+      "step": 37420,
+      "train_speed(iter/s)": 1.316908
+    },
+    {
+      "acc": 0.691892,
+      "epoch": 0.9493911719939118,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.827937278296037e-06,
+      "loss": 1.38608608,
+      "memory(GiB)": 123.79,
+      "step": 37425,
+      "train_speed(iter/s)": 1.316926
+    },
+    {
+      "acc": 0.70088806,
+      "epoch": 0.9495180111618468,
+      "grad_norm": 3.5,
+      "learning_rate": 5.826903112550819e-06,
+      "loss": 1.36950893,
+      "memory(GiB)": 123.79,
+      "step": 37430,
+      "train_speed(iter/s)": 1.316945
+    },
+    {
+      "acc": 0.6818202,
+      "epoch": 0.9496448503297819,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.825868910434708e-06,
+      "loss": 1.37430162,
+      "memory(GiB)": 123.79,
+      "step": 37435,
+      "train_speed(iter/s)": 1.316964
+    },
+    {
+      "acc": 0.70550165,
+      "epoch": 0.9497716894977168,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.824834671993197e-06,
+      "loss": 1.33347054,
+      "memory(GiB)": 123.79,
+      "step": 37440,
+      "train_speed(iter/s)": 1.316983
+    },
+    {
+      "acc": 0.6782649,
+      "epoch": 0.9498985286656519,
+      "grad_norm": 3.5,
+      "learning_rate": 5.823800397271774e-06,
+      "loss": 1.41959314,
+      "memory(GiB)": 123.79,
+      "step": 37445,
+      "train_speed(iter/s)": 1.317002
+    },
+    {
+      "acc": 0.68500934,
+      "epoch": 0.950025367833587,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.822766086315932e-06,
+      "loss": 1.41947517,
+      "memory(GiB)": 123.79,
+      "step": 37450,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.68034515,
+      "epoch": 0.950152207001522,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.821731739171164e-06,
+      "loss": 1.4258112,
+      "memory(GiB)": 123.79,
+      "step": 37455,
+      "train_speed(iter/s)": 1.317041
+    },
+    {
+      "acc": 0.69477806,
+      "epoch": 0.9502790461694571,
+      "grad_norm": 3.125,
+      "learning_rate": 5.820697355882965e-06,
+      "loss": 1.45753517,
+      "memory(GiB)": 123.79,
+      "step": 37460,
+      "train_speed(iter/s)": 1.31706
+    },
+    {
+      "acc": 0.69000473,
+      "epoch": 0.9504058853373922,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.819662936496833e-06,
+      "loss": 1.42072868,
+      "memory(GiB)": 123.79,
+      "step": 37465,
+      "train_speed(iter/s)": 1.317076
+    },
+    {
+      "acc": 0.68470535,
+      "epoch": 0.9505327245053272,
+      "grad_norm": 2.875,
+      "learning_rate": 5.818628481058265e-06,
+      "loss": 1.33554287,
+      "memory(GiB)": 123.79,
+      "step": 37470,
+      "train_speed(iter/s)": 1.317096
+    },
+    {
+      "acc": 0.67734885,
+      "epoch": 0.9506595636732623,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.81759398961276e-06,
+      "loss": 1.42016926,
+      "memory(GiB)": 123.79,
+      "step": 37475,
+      "train_speed(iter/s)": 1.317116
+    },
+    {
+      "acc": 0.68629284,
+      "epoch": 0.9507864028411973,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.816559462205824e-06,
+      "loss": 1.38050413,
+      "memory(GiB)": 123.79,
+      "step": 37480,
+      "train_speed(iter/s)": 1.317135
+    },
+    {
+      "acc": 0.69433002,
+      "epoch": 0.9509132420091324,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.815524898882954e-06,
+      "loss": 1.37260618,
+      "memory(GiB)": 123.79,
+      "step": 37485,
+      "train_speed(iter/s)": 1.317154
+    },
+    {
+      "acc": 0.69972348,
+      "epoch": 0.9510400811770675,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.8144902996896615e-06,
+      "loss": 1.38151302,
+      "memory(GiB)": 123.79,
+      "step": 37490,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.67539539,
+      "epoch": 0.9511669203450025,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.813455664671446e-06,
+      "loss": 1.44779224,
+      "memory(GiB)": 123.79,
+      "step": 37495,
+      "train_speed(iter/s)": 1.317192
+    },
+    {
+      "acc": 0.70342884,
+      "epoch": 0.9512937595129376,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.812420993873819e-06,
+      "loss": 1.27513123,
+      "memory(GiB)": 123.79,
+      "step": 37500,
+      "train_speed(iter/s)": 1.317211
+    },
+    {
+      "acc": 0.68550134,
+      "epoch": 0.9514205986808727,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.81138628734229e-06,
+      "loss": 1.38187304,
+      "memory(GiB)": 123.79,
+      "step": 37505,
+      "train_speed(iter/s)": 1.317231
+    },
+    {
+      "acc": 0.69276009,
+      "epoch": 0.9515474378488077,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.81035154512237e-06,
+      "loss": 1.34894505,
+      "memory(GiB)": 123.79,
+      "step": 37510,
+      "train_speed(iter/s)": 1.31725
+    },
+    {
+      "acc": 0.67210784,
+      "epoch": 0.9516742770167428,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.809316767259571e-06,
+      "loss": 1.44364433,
+      "memory(GiB)": 123.79,
+      "step": 37515,
+      "train_speed(iter/s)": 1.31727
+    },
+    {
+      "acc": 0.67364659,
+      "epoch": 0.9518011161846778,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.808281953799408e-06,
+      "loss": 1.42000237,
+      "memory(GiB)": 123.79,
+      "step": 37520,
+      "train_speed(iter/s)": 1.317288
+    },
+    {
+      "acc": 0.66830573,
+      "epoch": 0.9519279553526129,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.807247104787395e-06,
+      "loss": 1.41286201,
+      "memory(GiB)": 123.79,
+      "step": 37525,
+      "train_speed(iter/s)": 1.317308
+    },
+    {
+      "acc": 0.70063176,
+      "epoch": 0.952054794520548,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.806212220269049e-06,
+      "loss": 1.3794384,
+      "memory(GiB)": 123.79,
+      "step": 37530,
+      "train_speed(iter/s)": 1.317328
+    },
+    {
+      "acc": 0.67139611,
+      "epoch": 0.952181633688483,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.805177300289891e-06,
+      "loss": 1.48791218,
+      "memory(GiB)": 123.79,
+      "step": 37535,
+      "train_speed(iter/s)": 1.317344
+    },
+    {
+      "acc": 0.68262043,
+      "epoch": 0.9523084728564181,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.804142344895441e-06,
+      "loss": 1.4099678,
+      "memory(GiB)": 123.79,
+      "step": 37540,
+      "train_speed(iter/s)": 1.317364
+    },
+    {
+      "acc": 0.6922884,
+      "epoch": 0.9524353120243532,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.803107354131221e-06,
+      "loss": 1.36200733,
+      "memory(GiB)": 123.79,
+      "step": 37545,
+      "train_speed(iter/s)": 1.317383
+    },
+    {
+      "acc": 0.6852469,
+      "epoch": 0.9525621511922882,
+      "grad_norm": 3.25,
+      "learning_rate": 5.802072328042753e-06,
+      "loss": 1.4370595,
+      "memory(GiB)": 123.79,
+      "step": 37550,
+      "train_speed(iter/s)": 1.317404
+    },
+    {
+      "acc": 0.67714186,
+      "epoch": 0.9526889903602233,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.8010372666755625e-06,
+      "loss": 1.43240261,
+      "memory(GiB)": 123.79,
+      "step": 37555,
+      "train_speed(iter/s)": 1.31742
+    },
+    {
+      "acc": 0.6846694,
+      "epoch": 0.9528158295281582,
+      "grad_norm": 3.5,
+      "learning_rate": 5.800002170075179e-06,
+      "loss": 1.43098946,
+      "memory(GiB)": 123.79,
+      "step": 37560,
+      "train_speed(iter/s)": 1.317439
+    },
+    {
+      "acc": 0.71274462,
+      "epoch": 0.9529426686960933,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.798967038287125e-06,
+      "loss": 1.34449577,
+      "memory(GiB)": 123.79,
+      "step": 37565,
+      "train_speed(iter/s)": 1.317458
+    },
+    {
+      "acc": 0.69893684,
+      "epoch": 0.9530695078640284,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.797931871356936e-06,
+      "loss": 1.36005144,
+      "memory(GiB)": 123.79,
+      "step": 37570,
+      "train_speed(iter/s)": 1.317478
+    },
+    {
+      "acc": 0.68665953,
+      "epoch": 0.9531963470319634,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.796896669330139e-06,
+      "loss": 1.38883944,
+      "memory(GiB)": 123.79,
+      "step": 37575,
+      "train_speed(iter/s)": 1.317498
+    },
+    {
+      "acc": 0.66915436,
+      "epoch": 0.9533231861998985,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.79586143225227e-06,
+      "loss": 1.47462234,
+      "memory(GiB)": 123.79,
+      "step": 37580,
+      "train_speed(iter/s)": 1.317516
+    },
+    {
+      "acc": 0.67312107,
+      "epoch": 0.9534500253678336,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.79482616016886e-06,
+      "loss": 1.4526866,
+      "memory(GiB)": 123.79,
+      "step": 37585,
+      "train_speed(iter/s)": 1.317536
+    },
+    {
+      "acc": 0.68318048,
+      "epoch": 0.9535768645357686,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.793790853125449e-06,
+      "loss": 1.41836052,
+      "memory(GiB)": 123.79,
+      "step": 37590,
+      "train_speed(iter/s)": 1.317555
+    },
+    {
+      "acc": 0.68226972,
+      "epoch": 0.9537037037037037,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.792755511167572e-06,
+      "loss": 1.43210506,
+      "memory(GiB)": 123.79,
+      "step": 37595,
+      "train_speed(iter/s)": 1.317575
+    },
+    {
+      "acc": 0.68326063,
+      "epoch": 0.9538305428716387,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.7917201343407685e-06,
+      "loss": 1.3890379,
+      "memory(GiB)": 123.79,
+      "step": 37600,
+      "train_speed(iter/s)": 1.317595
+    },
+    {
+      "acc": 0.68325281,
+      "epoch": 0.9539573820395738,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.790684722690577e-06,
+      "loss": 1.47044296,
+      "memory(GiB)": 123.79,
+      "step": 37605,
+      "train_speed(iter/s)": 1.317614
+    },
+    {
+      "acc": 0.68231258,
+      "epoch": 0.9540842212075089,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.789649276262542e-06,
+      "loss": 1.4282733,
+      "memory(GiB)": 123.79,
+      "step": 37610,
+      "train_speed(iter/s)": 1.317632
+    },
+    {
+      "acc": 0.68817306,
+      "epoch": 0.9542110603754439,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.788613795102207e-06,
+      "loss": 1.39334831,
+      "memory(GiB)": 123.79,
+      "step": 37615,
+      "train_speed(iter/s)": 1.317651
+    },
+    {
+      "acc": 0.68446636,
+      "epoch": 0.954337899543379,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.787578279255116e-06,
+      "loss": 1.34620113,
+      "memory(GiB)": 123.79,
+      "step": 37620,
+      "train_speed(iter/s)": 1.31767
+    },
+    {
+      "acc": 0.70149641,
+      "epoch": 0.9544647387113141,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.786542728766815e-06,
+      "loss": 1.34999599,
+      "memory(GiB)": 123.79,
+      "step": 37625,
+      "train_speed(iter/s)": 1.317689
+    },
+    {
+      "acc": 0.68370867,
+      "epoch": 0.9545915778792491,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.785507143682856e-06,
+      "loss": 1.38726215,
+      "memory(GiB)": 123.79,
+      "step": 37630,
+      "train_speed(iter/s)": 1.317708
+    },
+    {
+      "acc": 0.66158352,
+      "epoch": 0.9547184170471842,
+      "grad_norm": 3.0,
+      "learning_rate": 5.784471524048782e-06,
+      "loss": 1.4274168,
+      "memory(GiB)": 123.79,
+      "step": 37635,
+      "train_speed(iter/s)": 1.317726
+    },
+    {
+      "acc": 0.68909397,
+      "epoch": 0.9548452562151192,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.783435869910151e-06,
+      "loss": 1.40981102,
+      "memory(GiB)": 123.79,
+      "step": 37640,
+      "train_speed(iter/s)": 1.317745
+    },
+    {
+      "acc": 0.68193464,
+      "epoch": 0.9549720953830543,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.782400181312511e-06,
+      "loss": 1.4229476,
+      "memory(GiB)": 123.79,
+      "step": 37645,
+      "train_speed(iter/s)": 1.317764
+    },
+    {
+      "acc": 0.69306698,
+      "epoch": 0.9550989345509894,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.781364458301419e-06,
+      "loss": 1.35960665,
+      "memory(GiB)": 123.79,
+      "step": 37650,
+      "train_speed(iter/s)": 1.31778
+    },
+    {
+      "acc": 0.69023409,
+      "epoch": 0.9552257737189244,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.780328700922427e-06,
+      "loss": 1.39576168,
+      "memory(GiB)": 123.79,
+      "step": 37655,
+      "train_speed(iter/s)": 1.317798
+    },
+    {
+      "acc": 0.70254951,
+      "epoch": 0.9553526128868595,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.779292909221097e-06,
+      "loss": 1.37471466,
+      "memory(GiB)": 123.79,
+      "step": 37660,
+      "train_speed(iter/s)": 1.317816
+    },
+    {
+      "acc": 0.69330683,
+      "epoch": 0.9554794520547946,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.778257083242986e-06,
+      "loss": 1.38806038,
+      "memory(GiB)": 123.79,
+      "step": 37665,
+      "train_speed(iter/s)": 1.317836
+    },
+    {
+      "acc": 0.68231697,
+      "epoch": 0.9556062912227296,
+      "grad_norm": 2.671875,
+      "learning_rate": 5.777221223033653e-06,
+      "loss": 1.40274601,
+      "memory(GiB)": 123.79,
+      "step": 37670,
+      "train_speed(iter/s)": 1.317854
+    },
+    {
+      "acc": 0.67840071,
+      "epoch": 0.9557331303906647,
+      "grad_norm": 2.625,
+      "learning_rate": 5.77618532863866e-06,
+      "loss": 1.40840302,
+      "memory(GiB)": 123.79,
+      "step": 37675,
+      "train_speed(iter/s)": 1.317872
+    },
+    {
+      "acc": 0.70861716,
+      "epoch": 0.9558599695585996,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.775149400103572e-06,
+      "loss": 1.28258724,
+      "memory(GiB)": 123.79,
+      "step": 37680,
+      "train_speed(iter/s)": 1.31789
+    },
+    {
+      "acc": 0.68206997,
+      "epoch": 0.9559868087265347,
+      "grad_norm": 3.25,
+      "learning_rate": 5.774113437473953e-06,
+      "loss": 1.40039959,
+      "memory(GiB)": 123.79,
+      "step": 37685,
+      "train_speed(iter/s)": 1.317909
+    },
+    {
+      "acc": 0.684902,
+      "epoch": 0.9561136478944698,
+      "grad_norm": 2.671875,
+      "learning_rate": 5.7730774407953675e-06,
+      "loss": 1.35597963,
+      "memory(GiB)": 123.79,
+      "step": 37690,
+      "train_speed(iter/s)": 1.317928
+    },
+    {
+      "acc": 0.6948101,
+      "epoch": 0.9562404870624048,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.772041410113384e-06,
+      "loss": 1.3399066,
+      "memory(GiB)": 123.79,
+      "step": 37695,
+      "train_speed(iter/s)": 1.317947
+    },
+    {
+      "acc": 0.67958002,
+      "epoch": 0.9563673262303399,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.771005345473575e-06,
+      "loss": 1.4162035,
+      "memory(GiB)": 123.79,
+      "step": 37700,
+      "train_speed(iter/s)": 1.317966
+    },
+    {
+      "acc": 0.69359627,
+      "epoch": 0.956494165398275,
+      "grad_norm": 2.75,
+      "learning_rate": 5.769969246921505e-06,
+      "loss": 1.34419098,
+      "memory(GiB)": 123.79,
+      "step": 37705,
+      "train_speed(iter/s)": 1.317985
+    },
+    {
+      "acc": 0.6839241,
+      "epoch": 0.95662100456621,
+      "grad_norm": 2.875,
+      "learning_rate": 5.768933114502753e-06,
+      "loss": 1.40703964,
+      "memory(GiB)": 123.79,
+      "step": 37710,
+      "train_speed(iter/s)": 1.318005
+    },
+    {
+      "acc": 0.69977093,
+      "epoch": 0.9567478437341451,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.7678969482628875e-06,
+      "loss": 1.34183493,
+      "memory(GiB)": 123.79,
+      "step": 37715,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.69323244,
+      "epoch": 0.9568746829020801,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.766860748247488e-06,
+      "loss": 1.32766647,
+      "memory(GiB)": 123.79,
+      "step": 37720,
+      "train_speed(iter/s)": 1.318043
+    },
+    {
+      "acc": 0.67831984,
+      "epoch": 0.9570015220700152,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.765824514502126e-06,
+      "loss": 1.42224064,
+      "memory(GiB)": 123.79,
+      "step": 37725,
+      "train_speed(iter/s)": 1.318062
+    },
+    {
+      "acc": 0.70548182,
+      "epoch": 0.9571283612379503,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.7647882470723846e-06,
+      "loss": 1.3623806,
+      "memory(GiB)": 123.79,
+      "step": 37730,
+      "train_speed(iter/s)": 1.318082
+    },
+    {
+      "acc": 0.70166001,
+      "epoch": 0.9572552004058853,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.763751946003842e-06,
+      "loss": 1.31687241,
+      "memory(GiB)": 123.79,
+      "step": 37735,
+      "train_speed(iter/s)": 1.3181
+    },
+    {
+      "acc": 0.68227634,
+      "epoch": 0.9573820395738204,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.7627156113420775e-06,
+      "loss": 1.35268478,
+      "memory(GiB)": 123.79,
+      "step": 37740,
+      "train_speed(iter/s)": 1.318119
+    },
+    {
+      "acc": 0.68370509,
+      "epoch": 0.9575088787417555,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.761679243132677e-06,
+      "loss": 1.36940842,
+      "memory(GiB)": 123.79,
+      "step": 37745,
+      "train_speed(iter/s)": 1.318138
+    },
+    {
+      "acc": 0.66688061,
+      "epoch": 0.9576357179096905,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.760642841421222e-06,
+      "loss": 1.40169344,
+      "memory(GiB)": 123.79,
+      "step": 37750,
+      "train_speed(iter/s)": 1.318157
+    },
+    {
+      "acc": 0.69682608,
+      "epoch": 0.9577625570776256,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.759606406253299e-06,
+      "loss": 1.34856844,
+      "memory(GiB)": 123.79,
+      "step": 37755,
+      "train_speed(iter/s)": 1.318175
+    },
+    {
+      "acc": 0.69211893,
+      "epoch": 0.9578893962455606,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.758569937674494e-06,
+      "loss": 1.33109922,
+      "memory(GiB)": 123.79,
+      "step": 37760,
+      "train_speed(iter/s)": 1.318193
+    },
+    {
+      "acc": 0.69315071,
+      "epoch": 0.9580162354134957,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.7575334357303954e-06,
+      "loss": 1.35805149,
+      "memory(GiB)": 123.79,
+      "step": 37765,
+      "train_speed(iter/s)": 1.318213
+    },
+    {
+      "acc": 0.68642998,
+      "epoch": 0.9581430745814308,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.756496900466596e-06,
+      "loss": 1.37482891,
+      "memory(GiB)": 123.79,
+      "step": 37770,
+      "train_speed(iter/s)": 1.318231
+    },
+    {
+      "acc": 0.68631206,
+      "epoch": 0.9582699137493658,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.755460331928684e-06,
+      "loss": 1.36946983,
+      "memory(GiB)": 123.79,
+      "step": 37775,
+      "train_speed(iter/s)": 1.31825
+    },
+    {
+      "acc": 0.68061862,
+      "epoch": 0.9583967529173009,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.754423730162257e-06,
+      "loss": 1.44919949,
+      "memory(GiB)": 123.79,
+      "step": 37780,
+      "train_speed(iter/s)": 1.31827
+    },
+    {
+      "acc": 0.6854425,
+      "epoch": 0.958523592085236,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.753387095212901e-06,
+      "loss": 1.40368023,
+      "memory(GiB)": 123.79,
+      "step": 37785,
+      "train_speed(iter/s)": 1.318288
+    },
+    {
+      "acc": 0.68747268,
+      "epoch": 0.958650431253171,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.752350427126221e-06,
+      "loss": 1.42385559,
+      "memory(GiB)": 123.79,
+      "step": 37790,
+      "train_speed(iter/s)": 1.318308
+    },
+    {
+      "acc": 0.68884354,
+      "epoch": 0.958777270421106,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.751313725947808e-06,
+      "loss": 1.33805342,
+      "memory(GiB)": 123.79,
+      "step": 37795,
+      "train_speed(iter/s)": 1.318326
+    },
+    {
+      "acc": 0.69364805,
+      "epoch": 0.958904109589041,
+      "grad_norm": 4.125,
+      "learning_rate": 5.7502769917232635e-06,
+      "loss": 1.36912737,
+      "memory(GiB)": 123.79,
+      "step": 37800,
+      "train_speed(iter/s)": 1.318345
+    },
+    {
+      "acc": 0.6802536,
+      "epoch": 0.9590309487569761,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.7492402244981885e-06,
+      "loss": 1.44862976,
+      "memory(GiB)": 123.79,
+      "step": 37805,
+      "train_speed(iter/s)": 1.318365
+    },
+    {
+      "acc": 0.69102073,
+      "epoch": 0.9591577879249112,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.748203424318182e-06,
+      "loss": 1.39663906,
+      "memory(GiB)": 123.79,
+      "step": 37810,
+      "train_speed(iter/s)": 1.318384
+    },
+    {
+      "acc": 0.7076695,
+      "epoch": 0.9592846270928462,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.747166591228849e-06,
+      "loss": 1.32299519,
+      "memory(GiB)": 123.79,
+      "step": 37815,
+      "train_speed(iter/s)": 1.318403
+    },
+    {
+      "acc": 0.69155831,
+      "epoch": 0.9594114662607813,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.746129725275793e-06,
+      "loss": 1.39248428,
+      "memory(GiB)": 123.79,
+      "step": 37820,
+      "train_speed(iter/s)": 1.318421
+    },
+    {
+      "acc": 0.68737564,
+      "epoch": 0.9595383054287164,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.74509282650462e-06,
+      "loss": 1.41284885,
+      "memory(GiB)": 123.79,
+      "step": 37825,
+      "train_speed(iter/s)": 1.318439
+    },
+    {
+      "acc": 0.67827559,
+      "epoch": 0.9596651445966514,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.744055894960938e-06,
+      "loss": 1.46372395,
+      "memory(GiB)": 123.79,
+      "step": 37830,
+      "train_speed(iter/s)": 1.318458
+    },
+    {
+      "acc": 0.67172556,
+      "epoch": 0.9597919837645865,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.743018930690357e-06,
+      "loss": 1.48761139,
+      "memory(GiB)": 123.79,
+      "step": 37835,
+      "train_speed(iter/s)": 1.318476
+    },
+    {
+      "acc": 0.68376651,
+      "epoch": 0.9599188229325215,
+      "grad_norm": 5.4375,
+      "learning_rate": 5.7419819337384855e-06,
+      "loss": 1.39914837,
+      "memory(GiB)": 123.79,
+      "step": 37840,
+      "train_speed(iter/s)": 1.318495
+    },
+    {
+      "acc": 0.67478752,
+      "epoch": 0.9600456621004566,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.740944904150934e-06,
+      "loss": 1.37875938,
+      "memory(GiB)": 123.79,
+      "step": 37845,
+      "train_speed(iter/s)": 1.318513
+    },
+    {
+      "acc": 0.68782978,
+      "epoch": 0.9601725012683917,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.739907841973321e-06,
+      "loss": 1.35511885,
+      "memory(GiB)": 123.79,
+      "step": 37850,
+      "train_speed(iter/s)": 1.318531
+    },
+    {
+      "acc": 0.68360291,
+      "epoch": 0.9602993404363267,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.738870747251255e-06,
+      "loss": 1.41675797,
+      "memory(GiB)": 123.79,
+      "step": 37855,
+      "train_speed(iter/s)": 1.318549
+    },
+    {
+      "acc": 0.68415956,
+      "epoch": 0.9604261796042618,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.737833620030357e-06,
+      "loss": 1.41554489,
+      "memory(GiB)": 123.79,
+      "step": 37860,
+      "train_speed(iter/s)": 1.318569
+    },
+    {
+      "acc": 0.71476417,
+      "epoch": 0.9605530187721969,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.7367964603562385e-06,
+      "loss": 1.31012802,
+      "memory(GiB)": 123.79,
+      "step": 37865,
+      "train_speed(iter/s)": 1.318588
+    },
+    {
+      "acc": 0.7006391,
+      "epoch": 0.9606798579401319,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.7357592682745245e-06,
+      "loss": 1.38447609,
+      "memory(GiB)": 123.79,
+      "step": 37870,
+      "train_speed(iter/s)": 1.318607
+    },
+    {
+      "acc": 0.70072961,
+      "epoch": 0.960806697108067,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.734722043830833e-06,
+      "loss": 1.30635424,
+      "memory(GiB)": 123.79,
+      "step": 37875,
+      "train_speed(iter/s)": 1.318624
+    },
+    {
+      "acc": 0.68650842,
+      "epoch": 0.960933536276002,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.7336847870707855e-06,
+      "loss": 1.41033363,
+      "memory(GiB)": 123.79,
+      "step": 37880,
+      "train_speed(iter/s)": 1.318644
+    },
+    {
+      "acc": 0.68552074,
+      "epoch": 0.9610603754439371,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.732647498040006e-06,
+      "loss": 1.41489229,
+      "memory(GiB)": 123.79,
+      "step": 37885,
+      "train_speed(iter/s)": 1.318663
+    },
+    {
+      "acc": 0.68431726,
+      "epoch": 0.9611872146118722,
+      "grad_norm": 5.09375,
+      "learning_rate": 5.731610176784118e-06,
+      "loss": 1.38405285,
+      "memory(GiB)": 123.79,
+      "step": 37890,
+      "train_speed(iter/s)": 1.318681
+    },
+    {
+      "acc": 0.68537722,
+      "epoch": 0.9613140537798072,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.730572823348748e-06,
+      "loss": 1.35050974,
+      "memory(GiB)": 123.79,
+      "step": 37895,
+      "train_speed(iter/s)": 1.318701
+    },
+    {
+      "acc": 0.68858538,
+      "epoch": 0.9614408929477423,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.729535437779523e-06,
+      "loss": 1.38770332,
+      "memory(GiB)": 123.79,
+      "step": 37900,
+      "train_speed(iter/s)": 1.318719
+    },
+    {
+      "acc": 0.66953464,
+      "epoch": 0.9615677321156774,
+      "grad_norm": 2.875,
+      "learning_rate": 5.728498020122073e-06,
+      "loss": 1.42866745,
+      "memory(GiB)": 123.79,
+      "step": 37905,
+      "train_speed(iter/s)": 1.318739
+    },
+    {
+      "acc": 0.68997264,
+      "epoch": 0.9616945712836124,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.727460570422028e-06,
+      "loss": 1.33794022,
+      "memory(GiB)": 123.79,
+      "step": 37910,
+      "train_speed(iter/s)": 1.318756
+    },
+    {
+      "acc": 0.69878511,
+      "epoch": 0.9618214104515475,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.726423088725017e-06,
+      "loss": 1.38791504,
+      "memory(GiB)": 123.79,
+      "step": 37915,
+      "train_speed(iter/s)": 1.318774
+    },
+    {
+      "acc": 0.67802391,
+      "epoch": 0.9619482496194824,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.725385575076677e-06,
+      "loss": 1.38081093,
+      "memory(GiB)": 123.79,
+      "step": 37920,
+      "train_speed(iter/s)": 1.318793
+    },
+    {
+      "acc": 0.6746439,
+      "epoch": 0.9620750887874175,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.7243480295226405e-06,
+      "loss": 1.43080368,
+      "memory(GiB)": 123.79,
+      "step": 37925,
+      "train_speed(iter/s)": 1.318812
+    },
+    {
+      "acc": 0.68891087,
+      "epoch": 0.9622019279553526,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.723310452108545e-06,
+      "loss": 1.39130783,
+      "memory(GiB)": 123.79,
+      "step": 37930,
+      "train_speed(iter/s)": 1.318831
+    },
+    {
+      "acc": 0.67256174,
+      "epoch": 0.9623287671232876,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.722272842880023e-06,
+      "loss": 1.47897263,
+      "memory(GiB)": 123.79,
+      "step": 37935,
+      "train_speed(iter/s)": 1.31885
+    },
+    {
+      "acc": 0.68769255,
+      "epoch": 0.9624556062912227,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.7212352018827215e-06,
+      "loss": 1.39348135,
+      "memory(GiB)": 123.79,
+      "step": 37940,
+      "train_speed(iter/s)": 1.318869
+    },
+    {
+      "acc": 0.6958725,
+      "epoch": 0.9625824454591578,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.720197529162272e-06,
+      "loss": 1.36697569,
+      "memory(GiB)": 123.79,
+      "step": 37945,
+      "train_speed(iter/s)": 1.318889
+    },
+    {
+      "acc": 0.69540272,
+      "epoch": 0.9627092846270928,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.719159824764321e-06,
+      "loss": 1.31064224,
+      "memory(GiB)": 123.79,
+      "step": 37950,
+      "train_speed(iter/s)": 1.318908
+    },
+    {
+      "acc": 0.69051619,
+      "epoch": 0.9628361237950279,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.71812208873451e-06,
+      "loss": 1.43259335,
+      "memory(GiB)": 123.79,
+      "step": 37955,
+      "train_speed(iter/s)": 1.318927
+    },
+    {
+      "acc": 0.68763647,
+      "epoch": 0.9629629629629629,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.717084321118482e-06,
+      "loss": 1.36658401,
+      "memory(GiB)": 123.79,
+      "step": 37960,
+      "train_speed(iter/s)": 1.318946
+    },
+    {
+      "acc": 0.68789992,
+      "epoch": 0.963089802130898,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.716046521961887e-06,
+      "loss": 1.4446703,
+      "memory(GiB)": 123.79,
+      "step": 37965,
+      "train_speed(iter/s)": 1.318965
+    },
+    {
+      "acc": 0.69640756,
+      "epoch": 0.9632166412988331,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.715008691310366e-06,
+      "loss": 1.39436398,
+      "memory(GiB)": 123.79,
+      "step": 37970,
+      "train_speed(iter/s)": 1.318985
+    },
+    {
+      "acc": 0.67793102,
+      "epoch": 0.9633434804667681,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.713970829209573e-06,
+      "loss": 1.4848032,
+      "memory(GiB)": 123.79,
+      "step": 37975,
+      "train_speed(iter/s)": 1.319004
+    },
+    {
+      "acc": 0.6856245,
+      "epoch": 0.9634703196347032,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.712932935705153e-06,
+      "loss": 1.45799561,
+      "memory(GiB)": 123.79,
+      "step": 37980,
+      "train_speed(iter/s)": 1.319021
+    },
+    {
+      "acc": 0.66341114,
+      "epoch": 0.9635971588026383,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.711895010842762e-06,
+      "loss": 1.49984751,
+      "memory(GiB)": 123.79,
+      "step": 37985,
+      "train_speed(iter/s)": 1.319041
+    },
+    {
+      "acc": 0.69058919,
+      "epoch": 0.9637239979705733,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.710857054668048e-06,
+      "loss": 1.41312084,
+      "memory(GiB)": 123.79,
+      "step": 37990,
+      "train_speed(iter/s)": 1.31906
+    },
+    {
+      "acc": 0.69853263,
+      "epoch": 0.9638508371385084,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.7098190672266675e-06,
+      "loss": 1.38687496,
+      "memory(GiB)": 123.79,
+      "step": 37995,
+      "train_speed(iter/s)": 1.319079
+    },
+    {
+      "acc": 0.68450422,
+      "epoch": 0.9639776763064434,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.708781048564276e-06,
+      "loss": 1.3648015,
+      "memory(GiB)": 123.79,
+      "step": 38000,
+      "train_speed(iter/s)": 1.319097
+    },
+    {
+      "epoch": 0.9639776763064434,
+      "eval_acc": 0.674912552181549,
+      "eval_loss": 1.357008695602417,
+      "eval_runtime": 69.5968,
+      "eval_samples_per_second": 91.527,
+      "eval_steps_per_second": 22.889,
+      "step": 38000
+    },
+    {
+      "acc": 0.69691005,
+      "epoch": 0.9641045154743785,
+      "grad_norm": 4.75,
+      "learning_rate": 5.707742998726527e-06,
+      "loss": 1.35002766,
+      "memory(GiB)": 123.79,
+      "step": 38005,
+      "train_speed(iter/s)": 1.315363
+    },
+    {
+      "acc": 0.68650899,
+      "epoch": 0.9642313546423136,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.706704917759085e-06,
+      "loss": 1.34669266,
+      "memory(GiB)": 123.79,
+      "step": 38010,
+      "train_speed(iter/s)": 1.315381
+    },
+    {
+      "acc": 0.70135469,
+      "epoch": 0.9643581938102486,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.705666805707603e-06,
+      "loss": 1.333957,
+      "memory(GiB)": 123.79,
+      "step": 38015,
+      "train_speed(iter/s)": 1.3154
+    },
+    {
+      "acc": 0.68076763,
+      "epoch": 0.9644850329781837,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.704628662617744e-06,
+      "loss": 1.37159824,
+      "memory(GiB)": 123.79,
+      "step": 38020,
+      "train_speed(iter/s)": 1.315419
+    },
+    {
+      "acc": 0.6973793,
+      "epoch": 0.9646118721461188,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.703590488535171e-06,
+      "loss": 1.3827879,
+      "memory(GiB)": 123.79,
+      "step": 38025,
+      "train_speed(iter/s)": 1.315438
+    },
+    {
+      "acc": 0.69102273,
+      "epoch": 0.9647387113140538,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.702552283505548e-06,
+      "loss": 1.38422689,
+      "memory(GiB)": 123.79,
+      "step": 38030,
+      "train_speed(iter/s)": 1.315457
+    },
+    {
+      "acc": 0.70198932,
+      "epoch": 0.9648655504819889,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.7015140475745376e-06,
+      "loss": 1.33846264,
+      "memory(GiB)": 123.79,
+      "step": 38035,
+      "train_speed(iter/s)": 1.315476
+    },
+    {
+      "acc": 0.69061422,
+      "epoch": 0.9649923896499238,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.700475780787809e-06,
+      "loss": 1.35471907,
+      "memory(GiB)": 123.79,
+      "step": 38040,
+      "train_speed(iter/s)": 1.315495
+    },
+    {
+      "acc": 0.68636427,
+      "epoch": 0.9651192288178589,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.699437483191027e-06,
+      "loss": 1.33107748,
+      "memory(GiB)": 123.79,
+      "step": 38045,
+      "train_speed(iter/s)": 1.315513
+    },
+    {
+      "acc": 0.691471,
+      "epoch": 0.965246067985794,
+      "grad_norm": 2.875,
+      "learning_rate": 5.6983991548298615e-06,
+      "loss": 1.3902523,
+      "memory(GiB)": 123.79,
+      "step": 38050,
+      "train_speed(iter/s)": 1.315531
+    },
+    {
+      "acc": 0.69879375,
+      "epoch": 0.965372907153729,
+      "grad_norm": 4.28125,
+      "learning_rate": 5.697360795749983e-06,
+      "loss": 1.37198782,
+      "memory(GiB)": 123.79,
+      "step": 38055,
+      "train_speed(iter/s)": 1.315551
+    },
+    {
+      "acc": 0.68062024,
+      "epoch": 0.9654997463216641,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.696322405997064e-06,
+      "loss": 1.43291206,
+      "memory(GiB)": 123.79,
+      "step": 38060,
+      "train_speed(iter/s)": 1.31557
+    },
+    {
+      "acc": 0.68917809,
+      "epoch": 0.9656265854895992,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.695283985616775e-06,
+      "loss": 1.42502937,
+      "memory(GiB)": 123.79,
+      "step": 38065,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.67924023,
+      "epoch": 0.9657534246575342,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.694245534654795e-06,
+      "loss": 1.43618345,
+      "memory(GiB)": 123.79,
+      "step": 38070,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.69266005,
+      "epoch": 0.9658802638254693,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.693207053156794e-06,
+      "loss": 1.25542431,
+      "memory(GiB)": 123.79,
+      "step": 38075,
+      "train_speed(iter/s)": 1.315627
+    },
+    {
+      "acc": 0.68885455,
+      "epoch": 0.9660071029934043,
+      "grad_norm": 2.5,
+      "learning_rate": 5.692168541168455e-06,
+      "loss": 1.37991295,
+      "memory(GiB)": 123.79,
+      "step": 38080,
+      "train_speed(iter/s)": 1.315645
+    },
+    {
+      "acc": 0.69453893,
+      "epoch": 0.9661339421613394,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.691129998735449e-06,
+      "loss": 1.30517435,
+      "memory(GiB)": 123.79,
+      "step": 38085,
+      "train_speed(iter/s)": 1.315664
+    },
+    {
+      "acc": 0.68950291,
+      "epoch": 0.9662607813292745,
+      "grad_norm": 3.125,
+      "learning_rate": 5.690091425903464e-06,
+      "loss": 1.43639584,
+      "memory(GiB)": 123.79,
+      "step": 38090,
+      "train_speed(iter/s)": 1.315678
+    },
+    {
+      "acc": 0.68453751,
+      "epoch": 0.9663876204972095,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.689052822718175e-06,
+      "loss": 1.39250183,
+      "memory(GiB)": 123.79,
+      "step": 38095,
+      "train_speed(iter/s)": 1.315698
+    },
+    {
+      "acc": 0.69728985,
+      "epoch": 0.9665144596651446,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.688014189225266e-06,
+      "loss": 1.32559624,
+      "memory(GiB)": 123.79,
+      "step": 38100,
+      "train_speed(iter/s)": 1.315716
+    },
+    {
+      "acc": 0.68363929,
+      "epoch": 0.9666412988330797,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.686975525470423e-06,
+      "loss": 1.42960415,
+      "memory(GiB)": 123.79,
+      "step": 38105,
+      "train_speed(iter/s)": 1.315735
+    },
+    {
+      "acc": 0.68892064,
+      "epoch": 0.9667681380010147,
+      "grad_norm": 4.25,
+      "learning_rate": 5.685936831499328e-06,
+      "loss": 1.41594429,
+      "memory(GiB)": 123.79,
+      "step": 38110,
+      "train_speed(iter/s)": 1.315751
+    },
+    {
+      "acc": 0.68595352,
+      "epoch": 0.9668949771689498,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.684898107357669e-06,
+      "loss": 1.39635773,
+      "memory(GiB)": 123.79,
+      "step": 38115,
+      "train_speed(iter/s)": 1.315767
+    },
+    {
+      "acc": 0.67912374,
+      "epoch": 0.9670218163368848,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.683859353091133e-06,
+      "loss": 1.47352324,
+      "memory(GiB)": 123.79,
+      "step": 38120,
+      "train_speed(iter/s)": 1.315787
+    },
+    {
+      "acc": 0.67709832,
+      "epoch": 0.9671486555048199,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.6828205687454094e-06,
+      "loss": 1.42289038,
+      "memory(GiB)": 123.79,
+      "step": 38125,
+      "train_speed(iter/s)": 1.315805
+    },
+    {
+      "acc": 0.67491517,
+      "epoch": 0.967275494672755,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.68178175436619e-06,
+      "loss": 1.45437708,
+      "memory(GiB)": 123.79,
+      "step": 38130,
+      "train_speed(iter/s)": 1.315825
+    },
+    {
+      "acc": 0.6873383,
+      "epoch": 0.96740233384069,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.680742909999163e-06,
+      "loss": 1.41616354,
+      "memory(GiB)": 123.79,
+      "step": 38135,
+      "train_speed(iter/s)": 1.315843
+    },
+    {
+      "acc": 0.667623,
+      "epoch": 0.9675291730086251,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.679704035690026e-06,
+      "loss": 1.50121689,
+      "memory(GiB)": 123.79,
+      "step": 38140,
+      "train_speed(iter/s)": 1.315861
+    },
+    {
+      "acc": 0.69102058,
+      "epoch": 0.9676560121765602,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.6786651314844675e-06,
+      "loss": 1.31794882,
+      "memory(GiB)": 123.79,
+      "step": 38145,
+      "train_speed(iter/s)": 1.315879
+    },
+    {
+      "acc": 0.68291273,
+      "epoch": 0.9677828513444952,
+      "grad_norm": 3.125,
+      "learning_rate": 5.67762619742819e-06,
+      "loss": 1.40261402,
+      "memory(GiB)": 123.79,
+      "step": 38150,
+      "train_speed(iter/s)": 1.315898
+    },
+    {
+      "acc": 0.67617378,
+      "epoch": 0.9679096905124303,
+      "grad_norm": 3.25,
+      "learning_rate": 5.676587233566885e-06,
+      "loss": 1.40001945,
+      "memory(GiB)": 123.79,
+      "step": 38155,
+      "train_speed(iter/s)": 1.315916
+    },
+    {
+      "acc": 0.68456125,
+      "epoch": 0.9680365296803652,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.675548239946254e-06,
+      "loss": 1.38343983,
+      "memory(GiB)": 123.79,
+      "step": 38160,
+      "train_speed(iter/s)": 1.315936
+    },
+    {
+      "acc": 0.66841164,
+      "epoch": 0.9681633688483003,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.674509216611993e-06,
+      "loss": 1.48532028,
+      "memory(GiB)": 123.79,
+      "step": 38165,
+      "train_speed(iter/s)": 1.315954
+    },
+    {
+      "acc": 0.68209677,
+      "epoch": 0.9682902080162354,
+      "grad_norm": 3.125,
+      "learning_rate": 5.673470163609806e-06,
+      "loss": 1.37249889,
+      "memory(GiB)": 123.79,
+      "step": 38170,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.69995408,
+      "epoch": 0.9684170471841704,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.672431080985395e-06,
+      "loss": 1.31943226,
+      "memory(GiB)": 123.79,
+      "step": 38175,
+      "train_speed(iter/s)": 1.315992
+    },
+    {
+      "acc": 0.68689623,
+      "epoch": 0.9685438863521055,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.671391968784464e-06,
+      "loss": 1.3850421,
+      "memory(GiB)": 123.79,
+      "step": 38180,
+      "train_speed(iter/s)": 1.31601
+    },
+    {
+      "acc": 0.69642344,
+      "epoch": 0.9686707255200406,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.670352827052715e-06,
+      "loss": 1.39369555,
+      "memory(GiB)": 123.79,
+      "step": 38185,
+      "train_speed(iter/s)": 1.316022
+    },
+    {
+      "acc": 0.69961977,
+      "epoch": 0.9687975646879756,
+      "grad_norm": 3.0,
+      "learning_rate": 5.6693136558358565e-06,
+      "loss": 1.399716,
+      "memory(GiB)": 123.79,
+      "step": 38190,
+      "train_speed(iter/s)": 1.316042
+    },
+    {
+      "acc": 0.679391,
+      "epoch": 0.9689244038559107,
+      "grad_norm": 3.125,
+      "learning_rate": 5.668274455179595e-06,
+      "loss": 1.42855015,
+      "memory(GiB)": 123.79,
+      "step": 38195,
+      "train_speed(iter/s)": 1.31606
+    },
+    {
+      "acc": 0.66441774,
+      "epoch": 0.9690512430238457,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.667235225129639e-06,
+      "loss": 1.46188107,
+      "memory(GiB)": 123.79,
+      "step": 38200,
+      "train_speed(iter/s)": 1.316078
+    },
+    {
+      "acc": 0.67833018,
+      "epoch": 0.9691780821917808,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.6661959657317e-06,
+      "loss": 1.39263563,
+      "memory(GiB)": 123.79,
+      "step": 38205,
+      "train_speed(iter/s)": 1.316098
+    },
+    {
+      "acc": 0.70486183,
+      "epoch": 0.9693049213597159,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.665156677031487e-06,
+      "loss": 1.35202084,
+      "memory(GiB)": 123.79,
+      "step": 38210,
+      "train_speed(iter/s)": 1.316117
+    },
+    {
+      "acc": 0.68262148,
+      "epoch": 0.9694317605276509,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.664117359074712e-06,
+      "loss": 1.41352901,
+      "memory(GiB)": 123.79,
+      "step": 38215,
+      "train_speed(iter/s)": 1.316136
+    },
+    {
+      "acc": 0.68317528,
+      "epoch": 0.969558599695586,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.6630780119070935e-06,
+      "loss": 1.46090794,
+      "memory(GiB)": 123.79,
+      "step": 38220,
+      "train_speed(iter/s)": 1.316155
+    },
+    {
+      "acc": 0.68600011,
+      "epoch": 0.9696854388635211,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.6620386355743415e-06,
+      "loss": 1.43450251,
+      "memory(GiB)": 123.79,
+      "step": 38225,
+      "train_speed(iter/s)": 1.316174
+    },
+    {
+      "acc": 0.68818283,
+      "epoch": 0.9698122780314561,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.660999230122177e-06,
+      "loss": 1.35185204,
+      "memory(GiB)": 123.79,
+      "step": 38230,
+      "train_speed(iter/s)": 1.316193
+    },
+    {
+      "acc": 0.67872896,
+      "epoch": 0.9699391171993912,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.659959795596313e-06,
+      "loss": 1.37694054,
+      "memory(GiB)": 123.79,
+      "step": 38235,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.67816334,
+      "epoch": 0.9700659563673262,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.65892033204247e-06,
+      "loss": 1.42758789,
+      "memory(GiB)": 123.79,
+      "step": 38240,
+      "train_speed(iter/s)": 1.316232
+    },
+    {
+      "acc": 0.70221791,
+      "epoch": 0.9701927955352613,
+      "grad_norm": 2.75,
+      "learning_rate": 5.657880839506371e-06,
+      "loss": 1.31611786,
+      "memory(GiB)": 123.79,
+      "step": 38245,
+      "train_speed(iter/s)": 1.316251
+    },
+    {
+      "acc": 0.69133835,
+      "epoch": 0.9703196347031964,
+      "grad_norm": 3.96875,
+      "learning_rate": 5.656841318033735e-06,
+      "loss": 1.36791897,
+      "memory(GiB)": 123.79,
+      "step": 38250,
+      "train_speed(iter/s)": 1.31627
+    },
+    {
+      "acc": 0.68568487,
+      "epoch": 0.9704464738711314,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.6558017676702846e-06,
+      "loss": 1.35911741,
+      "memory(GiB)": 123.79,
+      "step": 38255,
+      "train_speed(iter/s)": 1.316288
+    },
+    {
+      "acc": 0.6833107,
+      "epoch": 0.9705733130390665,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.654762188461744e-06,
+      "loss": 1.42673836,
+      "memory(GiB)": 123.79,
+      "step": 38260,
+      "train_speed(iter/s)": 1.316307
+    },
+    {
+      "acc": 0.67424774,
+      "epoch": 0.9707001522070016,
+      "grad_norm": 3.25,
+      "learning_rate": 5.653722580453841e-06,
+      "loss": 1.43464737,
+      "memory(GiB)": 123.79,
+      "step": 38265,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.67099786,
+      "epoch": 0.9708269913749366,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.652682943692299e-06,
+      "loss": 1.46004696,
+      "memory(GiB)": 123.79,
+      "step": 38270,
+      "train_speed(iter/s)": 1.316345
+    },
+    {
+      "acc": 0.69279914,
+      "epoch": 0.9709538305428717,
+      "grad_norm": 4.0625,
+      "learning_rate": 5.651643278222847e-06,
+      "loss": 1.40465717,
+      "memory(GiB)": 123.79,
+      "step": 38275,
+      "train_speed(iter/s)": 1.316364
+    },
+    {
+      "acc": 0.6997251,
+      "epoch": 0.9710806697108066,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.6506035840912145e-06,
+      "loss": 1.34196663,
+      "memory(GiB)": 123.79,
+      "step": 38280,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.69591985,
+      "epoch": 0.9712075088787417,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.649563861343131e-06,
+      "loss": 1.38004227,
+      "memory(GiB)": 123.79,
+      "step": 38285,
+      "train_speed(iter/s)": 1.316402
+    },
+    {
+      "acc": 0.66599641,
+      "epoch": 0.9713343480466768,
+      "grad_norm": 3.75,
+      "learning_rate": 5.648524110024331e-06,
+      "loss": 1.41896782,
+      "memory(GiB)": 123.79,
+      "step": 38290,
+      "train_speed(iter/s)": 1.31642
+    },
+    {
+      "acc": 0.69507666,
+      "epoch": 0.9714611872146118,
+      "grad_norm": 3.125,
+      "learning_rate": 5.647484330180542e-06,
+      "loss": 1.38544836,
+      "memory(GiB)": 123.79,
+      "step": 38295,
+      "train_speed(iter/s)": 1.316439
+    },
+    {
+      "acc": 0.6865396,
+      "epoch": 0.9715880263825469,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.646444521857504e-06,
+      "loss": 1.42073822,
+      "memory(GiB)": 123.79,
+      "step": 38300,
+      "train_speed(iter/s)": 1.316458
+    },
+    {
+      "acc": 0.70534964,
+      "epoch": 0.971714865550482,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.645404685100948e-06,
+      "loss": 1.2931778,
+      "memory(GiB)": 123.79,
+      "step": 38305,
+      "train_speed(iter/s)": 1.316476
+    },
+    {
+      "acc": 0.68969488,
+      "epoch": 0.971841704718417,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.644364819956613e-06,
+      "loss": 1.42131767,
+      "memory(GiB)": 123.79,
+      "step": 38310,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.68567047,
+      "epoch": 0.9719685438863521,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.643324926470236e-06,
+      "loss": 1.38579769,
+      "memory(GiB)": 123.79,
+      "step": 38315,
+      "train_speed(iter/s)": 1.316514
+    },
+    {
+      "acc": 0.69419174,
+      "epoch": 0.9720953830542871,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.642285004687557e-06,
+      "loss": 1.3900219,
+      "memory(GiB)": 123.79,
+      "step": 38320,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.69791379,
+      "epoch": 0.9722222222222222,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.6412450546543165e-06,
+      "loss": 1.38981762,
+      "memory(GiB)": 123.79,
+      "step": 38325,
+      "train_speed(iter/s)": 1.316554
+    },
+    {
+      "acc": 0.67460895,
+      "epoch": 0.9723490613901573,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.640205076416254e-06,
+      "loss": 1.49096317,
+      "memory(GiB)": 123.79,
+      "step": 38330,
+      "train_speed(iter/s)": 1.316573
+    },
+    {
+      "acc": 0.68244252,
+      "epoch": 0.9724759005580923,
+      "grad_norm": 4.25,
+      "learning_rate": 5.639165070019116e-06,
+      "loss": 1.42724609,
+      "memory(GiB)": 123.79,
+      "step": 38335,
+      "train_speed(iter/s)": 1.316592
+    },
+    {
+      "acc": 0.68410912,
+      "epoch": 0.9726027397260274,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.638125035508642e-06,
+      "loss": 1.3943387,
+      "memory(GiB)": 123.79,
+      "step": 38340,
+      "train_speed(iter/s)": 1.316611
+    },
+    {
+      "acc": 0.69354239,
+      "epoch": 0.9727295788939625,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.6370849729305825e-06,
+      "loss": 1.3860014,
+      "memory(GiB)": 123.79,
+      "step": 38345,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.67244711,
+      "epoch": 0.9728564180618975,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.63604488233068e-06,
+      "loss": 1.44541492,
+      "memory(GiB)": 123.79,
+      "step": 38350,
+      "train_speed(iter/s)": 1.31665
+    },
+    {
+      "acc": 0.68834352,
+      "epoch": 0.9729832572298326,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.635004763754683e-06,
+      "loss": 1.37132931,
+      "memory(GiB)": 123.79,
+      "step": 38355,
+      "train_speed(iter/s)": 1.316667
+    },
+    {
+      "acc": 0.68121858,
+      "epoch": 0.9731100963977676,
+      "grad_norm": 3.5,
+      "learning_rate": 5.633964617248345e-06,
+      "loss": 1.42766647,
+      "memory(GiB)": 123.79,
+      "step": 38360,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.69871893,
+      "epoch": 0.9732369355657027,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.6329244428574085e-06,
+      "loss": 1.33653297,
+      "memory(GiB)": 123.79,
+      "step": 38365,
+      "train_speed(iter/s)": 1.316704
+    },
+    {
+      "acc": 0.68612661,
+      "epoch": 0.9733637747336378,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.631884240627632e-06,
+      "loss": 1.3988863,
+      "memory(GiB)": 123.79,
+      "step": 38370,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.68724828,
+      "epoch": 0.9734906139015728,
+      "grad_norm": 2.5625,
+      "learning_rate": 5.6308440106047634e-06,
+      "loss": 1.41923485,
+      "memory(GiB)": 123.79,
+      "step": 38375,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.69279618,
+      "epoch": 0.9736174530695079,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.62980375283456e-06,
+      "loss": 1.3792717,
+      "memory(GiB)": 123.79,
+      "step": 38380,
+      "train_speed(iter/s)": 1.316763
+    },
+    {
+      "acc": 0.68654151,
+      "epoch": 0.973744292237443,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.628763467362775e-06,
+      "loss": 1.39320202,
+      "memory(GiB)": 123.79,
+      "step": 38385,
+      "train_speed(iter/s)": 1.316781
+    },
+    {
+      "acc": 0.67618923,
+      "epoch": 0.973871131405378,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.627723154235165e-06,
+      "loss": 1.40657578,
+      "memory(GiB)": 123.79,
+      "step": 38390,
+      "train_speed(iter/s)": 1.316801
+    },
+    {
+      "acc": 0.68247576,
+      "epoch": 0.973997970573313,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.62668281349749e-06,
+      "loss": 1.44778156,
+      "memory(GiB)": 123.79,
+      "step": 38395,
+      "train_speed(iter/s)": 1.31682
+    },
+    {
+      "acc": 0.67405128,
+      "epoch": 0.974124809741248,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.625642445195505e-06,
+      "loss": 1.47116947,
+      "memory(GiB)": 123.79,
+      "step": 38400,
+      "train_speed(iter/s)": 1.31684
+    },
+    {
+      "acc": 0.69952774,
+      "epoch": 0.9742516489091831,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.6246020493749735e-06,
+      "loss": 1.39353342,
+      "memory(GiB)": 123.79,
+      "step": 38405,
+      "train_speed(iter/s)": 1.316859
+    },
+    {
+      "acc": 0.69132638,
+      "epoch": 0.9743784880771182,
+      "grad_norm": 3.625,
+      "learning_rate": 5.623561626081654e-06,
+      "loss": 1.40961132,
+      "memory(GiB)": 123.79,
+      "step": 38410,
+      "train_speed(iter/s)": 1.316878
+    },
+    {
+      "acc": 0.69264717,
+      "epoch": 0.9745053272450532,
+      "grad_norm": 4.0,
+      "learning_rate": 5.622521175361311e-06,
+      "loss": 1.37767792,
+      "memory(GiB)": 123.79,
+      "step": 38415,
+      "train_speed(iter/s)": 1.316897
+    },
+    {
+      "acc": 0.68899608,
+      "epoch": 0.9746321664129883,
+      "grad_norm": 3.125,
+      "learning_rate": 5.621480697259707e-06,
+      "loss": 1.46775341,
+      "memory(GiB)": 123.79,
+      "step": 38420,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.66937714,
+      "epoch": 0.9747590055809234,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.620440191822607e-06,
+      "loss": 1.41191425,
+      "memory(GiB)": 123.79,
+      "step": 38425,
+      "train_speed(iter/s)": 1.316935
+    },
+    {
+      "acc": 0.66830807,
+      "epoch": 0.9748858447488584,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.619399659095778e-06,
+      "loss": 1.42972307,
+      "memory(GiB)": 123.79,
+      "step": 38430,
+      "train_speed(iter/s)": 1.316954
+    },
+    {
+      "acc": 0.68282642,
+      "epoch": 0.9750126839167935,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.618359099124985e-06,
+      "loss": 1.40371284,
+      "memory(GiB)": 123.79,
+      "step": 38435,
+      "train_speed(iter/s)": 1.316974
+    },
+    {
+      "acc": 0.6856986,
+      "epoch": 0.9751395230847285,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.617318511956001e-06,
+      "loss": 1.41881094,
+      "memory(GiB)": 123.79,
+      "step": 38440,
+      "train_speed(iter/s)": 1.316993
+    },
+    {
+      "acc": 0.67629681,
+      "epoch": 0.9752663622526636,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.61627789763459e-06,
+      "loss": 1.43718357,
+      "memory(GiB)": 123.79,
+      "step": 38445,
+      "train_speed(iter/s)": 1.31701
+    },
+    {
+      "acc": 0.69419403,
+      "epoch": 0.9753932014205987,
+      "grad_norm": 3.0,
+      "learning_rate": 5.6152372562065275e-06,
+      "loss": 1.35526772,
+      "memory(GiB)": 123.79,
+      "step": 38450,
+      "train_speed(iter/s)": 1.317029
+    },
+    {
+      "acc": 0.67836285,
+      "epoch": 0.9755200405885337,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.614196587717581e-06,
+      "loss": 1.39905319,
+      "memory(GiB)": 123.79,
+      "step": 38455,
+      "train_speed(iter/s)": 1.317048
+    },
+    {
+      "acc": 0.70282526,
+      "epoch": 0.9756468797564688,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.613155892213529e-06,
+      "loss": 1.34762783,
+      "memory(GiB)": 123.79,
+      "step": 38460,
+      "train_speed(iter/s)": 1.317067
+    },
+    {
+      "acc": 0.69718151,
+      "epoch": 0.9757737189244039,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.612115169740142e-06,
+      "loss": 1.34407158,
+      "memory(GiB)": 123.79,
+      "step": 38465,
+      "train_speed(iter/s)": 1.317083
+    },
+    {
+      "acc": 0.70368681,
+      "epoch": 0.9759005580923389,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.611074420343197e-06,
+      "loss": 1.37988405,
+      "memory(GiB)": 123.79,
+      "step": 38470,
+      "train_speed(iter/s)": 1.317102
+    },
+    {
+      "acc": 0.69907055,
+      "epoch": 0.976027397260274,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.610033644068471e-06,
+      "loss": 1.37368765,
+      "memory(GiB)": 123.79,
+      "step": 38475,
+      "train_speed(iter/s)": 1.317121
+    },
+    {
+      "acc": 0.69847636,
+      "epoch": 0.976154236428209,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.608992840961742e-06,
+      "loss": 1.34670496,
+      "memory(GiB)": 123.79,
+      "step": 38480,
+      "train_speed(iter/s)": 1.317139
+    },
+    {
+      "acc": 0.6843626,
+      "epoch": 0.9762810755961441,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.6079520110687876e-06,
+      "loss": 1.36999731,
+      "memory(GiB)": 123.79,
+      "step": 38485,
+      "train_speed(iter/s)": 1.317158
+    },
+    {
+      "acc": 0.69161034,
+      "epoch": 0.9764079147640792,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.606911154435392e-06,
+      "loss": 1.35322418,
+      "memory(GiB)": 123.79,
+      "step": 38490,
+      "train_speed(iter/s)": 1.317176
+    },
+    {
+      "acc": 0.68450031,
+      "epoch": 0.9765347539320142,
+      "grad_norm": 4.125,
+      "learning_rate": 5.605870271107332e-06,
+      "loss": 1.40223885,
+      "memory(GiB)": 123.79,
+      "step": 38495,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.68311725,
+      "epoch": 0.9766615930999493,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.6048293611303925e-06,
+      "loss": 1.40983391,
+      "memory(GiB)": 123.79,
+      "step": 38500,
+      "train_speed(iter/s)": 1.317213
+    },
+    {
+      "acc": 0.70258236,
+      "epoch": 0.9767884322678844,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.603788424550357e-06,
+      "loss": 1.37514277,
+      "memory(GiB)": 123.79,
+      "step": 38505,
+      "train_speed(iter/s)": 1.317231
+    },
+    {
+      "acc": 0.68718839,
+      "epoch": 0.9769152714358194,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.602747461413014e-06,
+      "loss": 1.37475042,
+      "memory(GiB)": 123.79,
+      "step": 38510,
+      "train_speed(iter/s)": 1.31725
+    },
+    {
+      "acc": 0.69529572,
+      "epoch": 0.9770421106037545,
+      "grad_norm": 4.28125,
+      "learning_rate": 5.6017064717641435e-06,
+      "loss": 1.31141396,
+      "memory(GiB)": 123.79,
+      "step": 38515,
+      "train_speed(iter/s)": 1.317265
+    },
+    {
+      "acc": 0.68607759,
+      "epoch": 0.9771689497716894,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.600665455649538e-06,
+      "loss": 1.41628523,
+      "memory(GiB)": 123.79,
+      "step": 38520,
+      "train_speed(iter/s)": 1.317284
+    },
+    {
+      "acc": 0.69189115,
+      "epoch": 0.9772957889396245,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.599624413114981e-06,
+      "loss": 1.3814312,
+      "memory(GiB)": 123.79,
+      "step": 38525,
+      "train_speed(iter/s)": 1.317303
+    },
+    {
+      "acc": 0.69095988,
+      "epoch": 0.9774226281075596,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.5985833442062676e-06,
+      "loss": 1.41157494,
+      "memory(GiB)": 123.79,
+      "step": 38530,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.69745646,
+      "epoch": 0.9775494672754946,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.597542248969185e-06,
+      "loss": 1.32369652,
+      "memory(GiB)": 123.79,
+      "step": 38535,
+      "train_speed(iter/s)": 1.317341
+    },
+    {
+      "acc": 0.67173901,
+      "epoch": 0.9776763064434297,
+      "grad_norm": 4.0,
+      "learning_rate": 5.596501127449527e-06,
+      "loss": 1.4112771,
+      "memory(GiB)": 123.79,
+      "step": 38540,
+      "train_speed(iter/s)": 1.317359
+    },
+    {
+      "acc": 0.68808818,
+      "epoch": 0.9778031456113648,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.595459979693086e-06,
+      "loss": 1.36863937,
+      "memory(GiB)": 123.79,
+      "step": 38545,
+      "train_speed(iter/s)": 1.317378
+    },
+    {
+      "acc": 0.69044375,
+      "epoch": 0.9779299847792998,
+      "grad_norm": 3.25,
+      "learning_rate": 5.594418805745657e-06,
+      "loss": 1.37370129,
+      "memory(GiB)": 123.79,
+      "step": 38550,
+      "train_speed(iter/s)": 1.317398
+    },
+    {
+      "acc": 0.67919664,
+      "epoch": 0.9780568239472349,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.593377605653035e-06,
+      "loss": 1.37235165,
+      "memory(GiB)": 123.79,
+      "step": 38555,
+      "train_speed(iter/s)": 1.317417
+    },
+    {
+      "acc": 0.69089122,
+      "epoch": 0.9781836631151699,
+      "grad_norm": 3.0,
+      "learning_rate": 5.592336379461018e-06,
+      "loss": 1.36287594,
+      "memory(GiB)": 123.79,
+      "step": 38560,
+      "train_speed(iter/s)": 1.317437
+    },
+    {
+      "acc": 0.68136201,
+      "epoch": 0.978310502283105,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.5912951272154004e-06,
+      "loss": 1.3483923,
+      "memory(GiB)": 123.79,
+      "step": 38565,
+      "train_speed(iter/s)": 1.317455
+    },
+    {
+      "acc": 0.68182478,
+      "epoch": 0.9784373414510401,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.590253848961984e-06,
+      "loss": 1.4499753,
+      "memory(GiB)": 123.79,
+      "step": 38570,
+      "train_speed(iter/s)": 1.317474
+    },
+    {
+      "acc": 0.6751018,
+      "epoch": 0.9785641806189751,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.589212544746566e-06,
+      "loss": 1.41323261,
+      "memory(GiB)": 123.79,
+      "step": 38575,
+      "train_speed(iter/s)": 1.317492
+    },
+    {
+      "acc": 0.69208264,
+      "epoch": 0.9786910197869102,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.588171214614953e-06,
+      "loss": 1.44380083,
+      "memory(GiB)": 123.79,
+      "step": 38580,
+      "train_speed(iter/s)": 1.317511
+    },
+    {
+      "acc": 0.68463449,
+      "epoch": 0.9788178589548453,
+      "grad_norm": 3.0,
+      "learning_rate": 5.587129858612941e-06,
+      "loss": 1.37291994,
+      "memory(GiB)": 123.79,
+      "step": 38585,
+      "train_speed(iter/s)": 1.31753
+    },
+    {
+      "acc": 0.67984028,
+      "epoch": 0.9789446981227803,
+      "grad_norm": 2.546875,
+      "learning_rate": 5.586088476786339e-06,
+      "loss": 1.41537533,
+      "memory(GiB)": 123.79,
+      "step": 38590,
+      "train_speed(iter/s)": 1.317549
+    },
+    {
+      "acc": 0.67414379,
+      "epoch": 0.9790715372907154,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.585047069180947e-06,
+      "loss": 1.45839558,
+      "memory(GiB)": 123.79,
+      "step": 38595,
+      "train_speed(iter/s)": 1.317567
+    },
+    {
+      "acc": 0.68567395,
+      "epoch": 0.9791983764586504,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.5840056358425755e-06,
+      "loss": 1.34984531,
+      "memory(GiB)": 123.79,
+      "step": 38600,
+      "train_speed(iter/s)": 1.317586
+    },
+    {
+      "acc": 0.67173781,
+      "epoch": 0.9793252156265855,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.582964176817025e-06,
+      "loss": 1.46231041,
+      "memory(GiB)": 123.79,
+      "step": 38605,
+      "train_speed(iter/s)": 1.317602
+    },
+    {
+      "acc": 0.69506707,
+      "epoch": 0.9794520547945206,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.58192269215011e-06,
+      "loss": 1.33809357,
+      "memory(GiB)": 123.79,
+      "step": 38610,
+      "train_speed(iter/s)": 1.31762
+    },
+    {
+      "acc": 0.69799848,
+      "epoch": 0.9795788939624556,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.580881181887636e-06,
+      "loss": 1.40070105,
+      "memory(GiB)": 123.79,
+      "step": 38615,
+      "train_speed(iter/s)": 1.317639
+    },
+    {
+      "acc": 0.6852519,
+      "epoch": 0.9797057331303907,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.579839646075414e-06,
+      "loss": 1.38276205,
+      "memory(GiB)": 123.79,
+      "step": 38620,
+      "train_speed(iter/s)": 1.317657
+    },
+    {
+      "acc": 0.67856808,
+      "epoch": 0.9798325722983258,
+      "grad_norm": 2.46875,
+      "learning_rate": 5.578798084759257e-06,
+      "loss": 1.4342947,
+      "memory(GiB)": 123.79,
+      "step": 38625,
+      "train_speed(iter/s)": 1.317676
+    },
+    {
+      "acc": 0.69299908,
+      "epoch": 0.9799594114662608,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.577756497984975e-06,
+      "loss": 1.32234802,
+      "memory(GiB)": 123.79,
+      "step": 38630,
+      "train_speed(iter/s)": 1.317693
+    },
+    {
+      "acc": 0.68411264,
+      "epoch": 0.9800862506341959,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.576714885798382e-06,
+      "loss": 1.3454524,
+      "memory(GiB)": 123.79,
+      "step": 38635,
+      "train_speed(iter/s)": 1.317712
+    },
+    {
+      "acc": 0.70173607,
+      "epoch": 0.9802130898021308,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.575673248245295e-06,
+      "loss": 1.35616789,
+      "memory(GiB)": 123.79,
+      "step": 38640,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "acc": 0.67577524,
+      "epoch": 0.9803399289700659,
+      "grad_norm": 4.5,
+      "learning_rate": 5.574631585371527e-06,
+      "loss": 1.4900773,
+      "memory(GiB)": 123.79,
+      "step": 38645,
+      "train_speed(iter/s)": 1.31775
+    },
+    {
+      "acc": 0.67671232,
+      "epoch": 0.980466768138001,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.573589897222897e-06,
+      "loss": 1.37697124,
+      "memory(GiB)": 123.79,
+      "step": 38650,
+      "train_speed(iter/s)": 1.317769
+    },
+    {
+      "acc": 0.69385605,
+      "epoch": 0.980593607305936,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.572548183845222e-06,
+      "loss": 1.33913479,
+      "memory(GiB)": 123.79,
+      "step": 38655,
+      "train_speed(iter/s)": 1.317788
+    },
+    {
+      "acc": 0.67724423,
+      "epoch": 0.9807204464738711,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.571506445284322e-06,
+      "loss": 1.41000872,
+      "memory(GiB)": 123.79,
+      "step": 38660,
+      "train_speed(iter/s)": 1.317807
+    },
+    {
+      "acc": 0.70007782,
+      "epoch": 0.9808472856418062,
+      "grad_norm": 2.390625,
+      "learning_rate": 5.570464681586017e-06,
+      "loss": 1.36947947,
+      "memory(GiB)": 123.79,
+      "step": 38665,
+      "train_speed(iter/s)": 1.317823
+    },
+    {
+      "acc": 0.67768288,
+      "epoch": 0.9809741248097412,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.569422892796129e-06,
+      "loss": 1.39097252,
+      "memory(GiB)": 123.79,
+      "step": 38670,
+      "train_speed(iter/s)": 1.317842
+    },
+    {
+      "acc": 0.67390509,
+      "epoch": 0.9811009639776763,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.568381078960479e-06,
+      "loss": 1.52952309,
+      "memory(GiB)": 123.79,
+      "step": 38675,
+      "train_speed(iter/s)": 1.317862
+    },
+    {
+      "acc": 0.69424124,
+      "epoch": 0.9812278031456113,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.567339240124892e-06,
+      "loss": 1.31350946,
+      "memory(GiB)": 123.79,
+      "step": 38680,
+      "train_speed(iter/s)": 1.317881
+    },
+    {
+      "acc": 0.69141994,
+      "epoch": 0.9813546423135464,
+      "grad_norm": 2.875,
+      "learning_rate": 5.5662973763351915e-06,
+      "loss": 1.41080017,
+      "memory(GiB)": 123.79,
+      "step": 38685,
+      "train_speed(iter/s)": 1.3179
+    },
+    {
+      "acc": 0.67816381,
+      "epoch": 0.9814814814814815,
+      "grad_norm": 3.96875,
+      "learning_rate": 5.565255487637204e-06,
+      "loss": 1.41748524,
+      "memory(GiB)": 123.79,
+      "step": 38690,
+      "train_speed(iter/s)": 1.317921
+    },
+    {
+      "acc": 0.68098941,
+      "epoch": 0.9816083206494165,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.564213574076757e-06,
+      "loss": 1.38967943,
+      "memory(GiB)": 123.79,
+      "step": 38695,
+      "train_speed(iter/s)": 1.31794
+    },
+    {
+      "acc": 0.67328019,
+      "epoch": 0.9817351598173516,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.563171635699678e-06,
+      "loss": 1.42953434,
+      "memory(GiB)": 123.79,
+      "step": 38700,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.69998522,
+      "epoch": 0.9818619989852867,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.562129672551796e-06,
+      "loss": 1.34255848,
+      "memory(GiB)": 123.79,
+      "step": 38705,
+      "train_speed(iter/s)": 1.31798
+    },
+    {
+      "acc": 0.67285538,
+      "epoch": 0.9819888381532217,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.561087684678941e-06,
+      "loss": 1.4163516,
+      "memory(GiB)": 123.79,
+      "step": 38710,
+      "train_speed(iter/s)": 1.318
+    },
+    {
+      "acc": 0.67783971,
+      "epoch": 0.9821156773211568,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.560045672126945e-06,
+      "loss": 1.41357517,
+      "memory(GiB)": 123.79,
+      "step": 38715,
+      "train_speed(iter/s)": 1.318019
+    },
+    {
+      "acc": 0.68542452,
+      "epoch": 0.9822425164890918,
+      "grad_norm": 3.125,
+      "learning_rate": 5.55900363494164e-06,
+      "loss": 1.40150414,
+      "memory(GiB)": 123.79,
+      "step": 38720,
+      "train_speed(iter/s)": 1.318039
+    },
+    {
+      "acc": 0.69568257,
+      "epoch": 0.9823693556570269,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.557961573168857e-06,
+      "loss": 1.41347151,
+      "memory(GiB)": 123.79,
+      "step": 38725,
+      "train_speed(iter/s)": 1.318058
+    },
+    {
+      "acc": 0.68835554,
+      "epoch": 0.982496194824962,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.5569194868544376e-06,
+      "loss": 1.37911034,
+      "memory(GiB)": 123.79,
+      "step": 38730,
+      "train_speed(iter/s)": 1.318078
+    },
+    {
+      "acc": 0.68319693,
+      "epoch": 0.982623033992897,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.555877376044209e-06,
+      "loss": 1.34975872,
+      "memory(GiB)": 123.79,
+      "step": 38735,
+      "train_speed(iter/s)": 1.318097
+    },
+    {
+      "acc": 0.70638084,
+      "epoch": 0.9827498731608321,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.554835240784013e-06,
+      "loss": 1.37446108,
+      "memory(GiB)": 123.79,
+      "step": 38740,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.68643217,
+      "epoch": 0.9828767123287672,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.553793081119685e-06,
+      "loss": 1.44026966,
+      "memory(GiB)": 123.79,
+      "step": 38745,
+      "train_speed(iter/s)": 1.318136
+    },
+    {
+      "acc": 0.68059225,
+      "epoch": 0.9830035514967022,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.552750897097065e-06,
+      "loss": 1.39628048,
+      "memory(GiB)": 123.79,
+      "step": 38750,
+      "train_speed(iter/s)": 1.318154
+    },
+    {
+      "acc": 0.67837868,
+      "epoch": 0.9831303906646373,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.551708688761993e-06,
+      "loss": 1.41306648,
+      "memory(GiB)": 123.79,
+      "step": 38755,
+      "train_speed(iter/s)": 1.318173
+    },
+    {
+      "acc": 0.68966703,
+      "epoch": 0.9832572298325722,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.550666456160311e-06,
+      "loss": 1.36933832,
+      "memory(GiB)": 123.79,
+      "step": 38760,
+      "train_speed(iter/s)": 1.318192
+    },
+    {
+      "acc": 0.68373671,
+      "epoch": 0.9833840690005073,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.549624199337857e-06,
+      "loss": 1.40151587,
+      "memory(GiB)": 123.79,
+      "step": 38765,
+      "train_speed(iter/s)": 1.318212
+    },
+    {
+      "acc": 0.67994394,
+      "epoch": 0.9835109081684424,
+      "grad_norm": 4.0,
+      "learning_rate": 5.548581918340479e-06,
+      "loss": 1.42944098,
+      "memory(GiB)": 123.79,
+      "step": 38770,
+      "train_speed(iter/s)": 1.318232
+    },
+    {
+      "acc": 0.69728122,
+      "epoch": 0.9836377473363774,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.547539613214019e-06,
+      "loss": 1.36426191,
+      "memory(GiB)": 123.79,
+      "step": 38775,
+      "train_speed(iter/s)": 1.31825
+    },
+    {
+      "acc": 0.68516617,
+      "epoch": 0.9837645865043125,
+      "grad_norm": 3.5,
+      "learning_rate": 5.546497284004321e-06,
+      "loss": 1.41016846,
+      "memory(GiB)": 123.79,
+      "step": 38780,
+      "train_speed(iter/s)": 1.31827
+    },
+    {
+      "acc": 0.69009199,
+      "epoch": 0.9838914256722476,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.545454930757233e-06,
+      "loss": 1.38792667,
+      "memory(GiB)": 123.79,
+      "step": 38785,
+      "train_speed(iter/s)": 1.318289
+    },
+    {
+      "acc": 0.70018291,
+      "epoch": 0.9840182648401826,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.544412553518602e-06,
+      "loss": 1.30349445,
+      "memory(GiB)": 123.79,
+      "step": 38790,
+      "train_speed(iter/s)": 1.318309
+    },
+    {
+      "acc": 0.69033861,
+      "epoch": 0.9841451040081177,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.543370152334275e-06,
+      "loss": 1.36598148,
+      "memory(GiB)": 123.79,
+      "step": 38795,
+      "train_speed(iter/s)": 1.318328
+    },
+    {
+      "acc": 0.69000359,
+      "epoch": 0.9842719431760527,
+      "grad_norm": 2.75,
+      "learning_rate": 5.542327727250105e-06,
+      "loss": 1.36232901,
+      "memory(GiB)": 123.79,
+      "step": 38800,
+      "train_speed(iter/s)": 1.318345
+    },
+    {
+      "acc": 0.67691216,
+      "epoch": 0.9843987823439878,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.5412852783119385e-06,
+      "loss": 1.42165928,
+      "memory(GiB)": 123.79,
+      "step": 38805,
+      "train_speed(iter/s)": 1.318364
+    },
+    {
+      "acc": 0.68680835,
+      "epoch": 0.9845256215119229,
+      "grad_norm": 2.625,
+      "learning_rate": 5.54024280556563e-06,
+      "loss": 1.38170528,
+      "memory(GiB)": 123.79,
+      "step": 38810,
+      "train_speed(iter/s)": 1.318384
+    },
+    {
+      "acc": 0.68748226,
+      "epoch": 0.9846524606798579,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.53920030905703e-06,
+      "loss": 1.46366358,
+      "memory(GiB)": 123.79,
+      "step": 38815,
+      "train_speed(iter/s)": 1.318403
+    },
+    {
+      "acc": 0.6781539,
+      "epoch": 0.984779299847793,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.538157788831993e-06,
+      "loss": 1.3670043,
+      "memory(GiB)": 123.79,
+      "step": 38820,
+      "train_speed(iter/s)": 1.318422
+    },
+    {
+      "acc": 0.683039,
+      "epoch": 0.9849061390157281,
+      "grad_norm": 6.375,
+      "learning_rate": 5.537115244936374e-06,
+      "loss": 1.4183754,
+      "memory(GiB)": 123.79,
+      "step": 38825,
+      "train_speed(iter/s)": 1.318442
+    },
+    {
+      "acc": 0.671486,
+      "epoch": 0.9850329781836631,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.536072677416029e-06,
+      "loss": 1.47557774,
+      "memory(GiB)": 123.79,
+      "step": 38830,
+      "train_speed(iter/s)": 1.318462
+    },
+    {
+      "acc": 0.69487028,
+      "epoch": 0.9851598173515982,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.535030086316814e-06,
+      "loss": 1.35275536,
+      "memory(GiB)": 123.79,
+      "step": 38835,
+      "train_speed(iter/s)": 1.318481
+    },
+    {
+      "acc": 0.69585609,
+      "epoch": 0.9852866565195332,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.533987471684586e-06,
+      "loss": 1.33342533,
+      "memory(GiB)": 123.79,
+      "step": 38840,
+      "train_speed(iter/s)": 1.318499
+    },
+    {
+      "acc": 0.68337326,
+      "epoch": 0.9854134956874683,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.532944833565207e-06,
+      "loss": 1.37561741,
+      "memory(GiB)": 123.79,
+      "step": 38845,
+      "train_speed(iter/s)": 1.318518
+    },
+    {
+      "acc": 0.67423048,
+      "epoch": 0.9855403348554034,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.531902172004533e-06,
+      "loss": 1.39807539,
+      "memory(GiB)": 123.79,
+      "step": 38850,
+      "train_speed(iter/s)": 1.318537
+    },
+    {
+      "acc": 0.68757381,
+      "epoch": 0.9856671740233384,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.530859487048427e-06,
+      "loss": 1.42044115,
+      "memory(GiB)": 123.79,
+      "step": 38855,
+      "train_speed(iter/s)": 1.318556
+    },
+    {
+      "acc": 0.666924,
+      "epoch": 0.9857940131912735,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.529816778742752e-06,
+      "loss": 1.49954996,
+      "memory(GiB)": 123.79,
+      "step": 38860,
+      "train_speed(iter/s)": 1.318575
+    },
+    {
+      "acc": 0.70819879,
+      "epoch": 0.9859208523592086,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.528774047133369e-06,
+      "loss": 1.3025259,
+      "memory(GiB)": 123.79,
+      "step": 38865,
+      "train_speed(iter/s)": 1.318595
+    },
+    {
+      "acc": 0.69097576,
+      "epoch": 0.9860476915271436,
+      "grad_norm": 2.75,
+      "learning_rate": 5.527731292266142e-06,
+      "loss": 1.38745871,
+      "memory(GiB)": 123.79,
+      "step": 38870,
+      "train_speed(iter/s)": 1.318612
+    },
+    {
+      "acc": 0.69389315,
+      "epoch": 0.9861745306950787,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.5266885141869355e-06,
+      "loss": 1.40966606,
+      "memory(GiB)": 123.79,
+      "step": 38875,
+      "train_speed(iter/s)": 1.318632
+    },
+    {
+      "acc": 0.68742638,
+      "epoch": 0.9863013698630136,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.5256457129416185e-06,
+      "loss": 1.40668621,
+      "memory(GiB)": 123.79,
+      "step": 38880,
+      "train_speed(iter/s)": 1.318653
+    },
+    {
+      "acc": 0.70185795,
+      "epoch": 0.9864282090309487,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.524602888576055e-06,
+      "loss": 1.34954758,
+      "memory(GiB)": 123.79,
+      "step": 38885,
+      "train_speed(iter/s)": 1.318674
+    },
+    {
+      "acc": 0.66611519,
+      "epoch": 0.9865550481988838,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.523560041136116e-06,
+      "loss": 1.4354516,
+      "memory(GiB)": 123.79,
+      "step": 38890,
+      "train_speed(iter/s)": 1.318693
+    },
+    {
+      "acc": 0.66284528,
+      "epoch": 0.9866818873668188,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.522517170667667e-06,
+      "loss": 1.50507545,
+      "memory(GiB)": 123.79,
+      "step": 38895,
+      "train_speed(iter/s)": 1.318714
+    },
+    {
+      "acc": 0.69381976,
+      "epoch": 0.9868087265347539,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.5214742772165806e-06,
+      "loss": 1.34677525,
+      "memory(GiB)": 123.79,
+      "step": 38900,
+      "train_speed(iter/s)": 1.318735
+    },
+    {
+      "acc": 0.67433367,
+      "epoch": 0.986935565702689,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.520431360828728e-06,
+      "loss": 1.46483269,
+      "memory(GiB)": 123.79,
+      "step": 38905,
+      "train_speed(iter/s)": 1.318756
+    },
+    {
+      "acc": 0.67032976,
+      "epoch": 0.987062404870624,
+      "grad_norm": 3.5,
+      "learning_rate": 5.51938842154998e-06,
+      "loss": 1.46888638,
+      "memory(GiB)": 123.79,
+      "step": 38910,
+      "train_speed(iter/s)": 1.318776
+    },
+    {
+      "acc": 0.69553185,
+      "epoch": 0.9871892440385591,
+      "grad_norm": 4.125,
+      "learning_rate": 5.51834545942621e-06,
+      "loss": 1.34654846,
+      "memory(GiB)": 123.79,
+      "step": 38915,
+      "train_speed(iter/s)": 1.318797
+    },
+    {
+      "acc": 0.68432841,
+      "epoch": 0.9873160832064941,
+      "grad_norm": 4.125,
+      "learning_rate": 5.5173024745032925e-06,
+      "loss": 1.36979141,
+      "memory(GiB)": 123.79,
+      "step": 38920,
+      "train_speed(iter/s)": 1.318817
+    },
+    {
+      "acc": 0.67918015,
+      "epoch": 0.9874429223744292,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.516259466827103e-06,
+      "loss": 1.44674816,
+      "memory(GiB)": 123.79,
+      "step": 38925,
+      "train_speed(iter/s)": 1.318838
+    },
+    {
+      "acc": 0.68611097,
+      "epoch": 0.9875697615423643,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.515216436443517e-06,
+      "loss": 1.39919071,
+      "memory(GiB)": 123.79,
+      "step": 38930,
+      "train_speed(iter/s)": 1.318859
+    },
+    {
+      "acc": 0.68279095,
+      "epoch": 0.9876966007102993,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.514173383398412e-06,
+      "loss": 1.43317242,
+      "memory(GiB)": 123.79,
+      "step": 38935,
+      "train_speed(iter/s)": 1.318879
+    },
+    {
+      "acc": 0.68212819,
+      "epoch": 0.9878234398782344,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.513130307737666e-06,
+      "loss": 1.41781454,
+      "memory(GiB)": 123.79,
+      "step": 38940,
+      "train_speed(iter/s)": 1.318899
+    },
+    {
+      "acc": 0.67673712,
+      "epoch": 0.9879502790461695,
+      "grad_norm": 2.625,
+      "learning_rate": 5.512087209507157e-06,
+      "loss": 1.45814333,
+      "memory(GiB)": 123.79,
+      "step": 38945,
+      "train_speed(iter/s)": 1.31892
+    },
+    {
+      "acc": 0.68739328,
+      "epoch": 0.9880771182141045,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.5110440887527684e-06,
+      "loss": 1.31755629,
+      "memory(GiB)": 123.79,
+      "step": 38950,
+      "train_speed(iter/s)": 1.31894
+    },
+    {
+      "acc": 0.69284506,
+      "epoch": 0.9882039573820396,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.510000945520377e-06,
+      "loss": 1.35481119,
+      "memory(GiB)": 123.79,
+      "step": 38955,
+      "train_speed(iter/s)": 1.318957
+    },
+    {
+      "acc": 0.68517976,
+      "epoch": 0.9883307965499746,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.508957779855869e-06,
+      "loss": 1.42332363,
+      "memory(GiB)": 123.79,
+      "step": 38960,
+      "train_speed(iter/s)": 1.318976
+    },
+    {
+      "acc": 0.68608899,
+      "epoch": 0.9884576357179097,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.507914591805124e-06,
+      "loss": 1.42329273,
+      "memory(GiB)": 123.79,
+      "step": 38965,
+      "train_speed(iter/s)": 1.318997
+    },
+    {
+      "acc": 0.69393454,
+      "epoch": 0.9885844748858448,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.506871381414027e-06,
+      "loss": 1.33735752,
+      "memory(GiB)": 123.79,
+      "step": 38970,
+      "train_speed(iter/s)": 1.319017
+    },
+    {
+      "acc": 0.68499441,
+      "epoch": 0.9887113140537798,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.505828148728465e-06,
+      "loss": 1.41139107,
+      "memory(GiB)": 123.79,
+      "step": 38975,
+      "train_speed(iter/s)": 1.319037
+    },
+    {
+      "acc": 0.69022341,
+      "epoch": 0.9888381532217149,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.5047848937943225e-06,
+      "loss": 1.38698387,
+      "memory(GiB)": 123.79,
+      "step": 38980,
+      "train_speed(iter/s)": 1.319055
+    },
+    {
+      "acc": 0.68917947,
+      "epoch": 0.98896499238965,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.503741616657486e-06,
+      "loss": 1.36863213,
+      "memory(GiB)": 123.79,
+      "step": 38985,
+      "train_speed(iter/s)": 1.319073
+    },
+    {
+      "acc": 0.68727674,
+      "epoch": 0.989091831557585,
+      "grad_norm": 4.25,
+      "learning_rate": 5.502698317363846e-06,
+      "loss": 1.43573771,
+      "memory(GiB)": 123.79,
+      "step": 38990,
+      "train_speed(iter/s)": 1.319092
+    },
+    {
+      "acc": 0.68300524,
+      "epoch": 0.98921867072552,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.501654995959288e-06,
+      "loss": 1.42089615,
+      "memory(GiB)": 123.79,
+      "step": 38995,
+      "train_speed(iter/s)": 1.319108
+    },
+    {
+      "acc": 0.67447357,
+      "epoch": 0.989345509893455,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.5006116524897034e-06,
+      "loss": 1.41937695,
+      "memory(GiB)": 123.79,
+      "step": 39000,
+      "train_speed(iter/s)": 1.319127
+    },
+    {
+      "epoch": 0.989345509893455,
+      "eval_acc": 0.6749351058875614,
+      "eval_loss": 1.3568341732025146,
+      "eval_runtime": 69.2713,
+      "eval_samples_per_second": 91.957,
+      "eval_steps_per_second": 22.997,
+      "step": 39000
+    },
+    {
+      "acc": 0.68490853,
+      "epoch": 0.9894723490613901,
+      "grad_norm": 3.5,
+      "learning_rate": 5.499568287000984e-06,
+      "loss": 1.43501825,
+      "memory(GiB)": 123.79,
+      "step": 39005,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.68964562,
+      "epoch": 0.9895991882293252,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.49852489953902e-06,
+      "loss": 1.39588737,
+      "memory(GiB)": 123.79,
+      "step": 39010,
+      "train_speed(iter/s)": 1.315524
+    },
+    {
+      "acc": 0.69556952,
+      "epoch": 0.9897260273972602,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.497481490149705e-06,
+      "loss": 1.37354851,
+      "memory(GiB)": 123.79,
+      "step": 39015,
+      "train_speed(iter/s)": 1.315542
+    },
+    {
+      "acc": 0.68085008,
+      "epoch": 0.9898528665651953,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.496438058878936e-06,
+      "loss": 1.42083559,
+      "memory(GiB)": 123.79,
+      "step": 39020,
+      "train_speed(iter/s)": 1.315562
+    },
+    {
+      "acc": 0.67774491,
+      "epoch": 0.9899797057331304,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.4953946057726005e-06,
+      "loss": 1.45507822,
+      "memory(GiB)": 123.79,
+      "step": 39025,
+      "train_speed(iter/s)": 1.315582
+    },
+    {
+      "acc": 0.69048424,
+      "epoch": 0.9901065449010654,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.494351130876602e-06,
+      "loss": 1.37635822,
+      "memory(GiB)": 123.79,
+      "step": 39030,
+      "train_speed(iter/s)": 1.315601
+    },
+    {
+      "acc": 0.69990358,
+      "epoch": 0.9902333840690005,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.493307634236831e-06,
+      "loss": 1.35525579,
+      "memory(GiB)": 123.79,
+      "step": 39035,
+      "train_speed(iter/s)": 1.315621
+    },
+    {
+      "acc": 0.6814436,
+      "epoch": 0.9903602232369355,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.492264115899189e-06,
+      "loss": 1.39971275,
+      "memory(GiB)": 123.79,
+      "step": 39040,
+      "train_speed(iter/s)": 1.31564
+    },
+    {
+      "acc": 0.67934647,
+      "epoch": 0.9904870624048706,
+      "grad_norm": 5.5625,
+      "learning_rate": 5.491220575909573e-06,
+      "loss": 1.4050499,
+      "memory(GiB)": 123.79,
+      "step": 39045,
+      "train_speed(iter/s)": 1.315659
+    },
+    {
+      "acc": 0.68781424,
+      "epoch": 0.9906139015728057,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.4901770143138835e-06,
+      "loss": 1.43856869,
+      "memory(GiB)": 123.79,
+      "step": 39050,
+      "train_speed(iter/s)": 1.315679
+    },
+    {
+      "acc": 0.68962517,
+      "epoch": 0.9907407407407407,
+      "grad_norm": 2.453125,
+      "learning_rate": 5.48913343115802e-06,
+      "loss": 1.37858467,
+      "memory(GiB)": 123.79,
+      "step": 39055,
+      "train_speed(iter/s)": 1.315698
+    },
+    {
+      "acc": 0.68012228,
+      "epoch": 0.9908675799086758,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.488089826487884e-06,
+      "loss": 1.36877804,
+      "memory(GiB)": 123.79,
+      "step": 39060,
+      "train_speed(iter/s)": 1.31571
+    },
+    {
+      "acc": 0.69662256,
+      "epoch": 0.9909944190766109,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.48704620034938e-06,
+      "loss": 1.35638618,
+      "memory(GiB)": 123.79,
+      "step": 39065,
+      "train_speed(iter/s)": 1.315728
+    },
+    {
+      "acc": 0.69747415,
+      "epoch": 0.9911212582445459,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.486002552788408e-06,
+      "loss": 1.38124838,
+      "memory(GiB)": 123.79,
+      "step": 39070,
+      "train_speed(iter/s)": 1.315749
+    },
+    {
+      "acc": 0.67800312,
+      "epoch": 0.991248097412481,
+      "grad_norm": 3.5,
+      "learning_rate": 5.4849588838508734e-06,
+      "loss": 1.38521996,
+      "memory(GiB)": 123.79,
+      "step": 39075,
+      "train_speed(iter/s)": 1.315768
+    },
+    {
+      "acc": 0.69953928,
+      "epoch": 0.991374936580416,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.483915193582684e-06,
+      "loss": 1.37240353,
+      "memory(GiB)": 123.79,
+      "step": 39080,
+      "train_speed(iter/s)": 1.315786
+    },
+    {
+      "acc": 0.67906585,
+      "epoch": 0.9915017757483511,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.482871482029742e-06,
+      "loss": 1.35831919,
+      "memory(GiB)": 123.79,
+      "step": 39085,
+      "train_speed(iter/s)": 1.315804
+    },
+    {
+      "acc": 0.69482298,
+      "epoch": 0.9916286149162862,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.4818277492379565e-06,
+      "loss": 1.40450916,
+      "memory(GiB)": 123.79,
+      "step": 39090,
+      "train_speed(iter/s)": 1.315825
+    },
+    {
+      "acc": 0.69024568,
+      "epoch": 0.9917554540842212,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.480783995253236e-06,
+      "loss": 1.40165634,
+      "memory(GiB)": 123.79,
+      "step": 39095,
+      "train_speed(iter/s)": 1.315845
+    },
+    {
+      "acc": 0.69939909,
+      "epoch": 0.9918822932521563,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.47974022012149e-06,
+      "loss": 1.35167027,
+      "memory(GiB)": 123.79,
+      "step": 39100,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.69382901,
+      "epoch": 0.9920091324200914,
+      "grad_norm": 3.5,
+      "learning_rate": 5.478696423888624e-06,
+      "loss": 1.34443092,
+      "memory(GiB)": 123.79,
+      "step": 39105,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.68594971,
+      "epoch": 0.9921359715880264,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.477652606600555e-06,
+      "loss": 1.3729579,
+      "memory(GiB)": 123.79,
+      "step": 39110,
+      "train_speed(iter/s)": 1.315903
+    },
+    {
+      "acc": 0.69666138,
+      "epoch": 0.9922628107559615,
+      "grad_norm": 3.25,
+      "learning_rate": 5.47660876830319e-06,
+      "loss": 1.33560581,
+      "memory(GiB)": 123.79,
+      "step": 39115,
+      "train_speed(iter/s)": 1.315921
+    },
+    {
+      "acc": 0.69726906,
+      "epoch": 0.9923896499238964,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.475564909042444e-06,
+      "loss": 1.30480423,
+      "memory(GiB)": 123.79,
+      "step": 39120,
+      "train_speed(iter/s)": 1.31594
+    },
+    {
+      "acc": 0.68045535,
+      "epoch": 0.9925164890918315,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.4745210288642306e-06,
+      "loss": 1.3920188,
+      "memory(GiB)": 123.79,
+      "step": 39125,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.67379389,
+      "epoch": 0.9926433282597666,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.473477127814464e-06,
+      "loss": 1.38305531,
+      "memory(GiB)": 123.79,
+      "step": 39130,
+      "train_speed(iter/s)": 1.315979
+    },
+    {
+      "acc": 0.66830196,
+      "epoch": 0.9927701674277016,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.472433205939058e-06,
+      "loss": 1.50924816,
+      "memory(GiB)": 123.79,
+      "step": 39135,
+      "train_speed(iter/s)": 1.315999
+    },
+    {
+      "acc": 0.70264521,
+      "epoch": 0.9928970065956367,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.471389263283932e-06,
+      "loss": 1.35003147,
+      "memory(GiB)": 123.79,
+      "step": 39140,
+      "train_speed(iter/s)": 1.316019
+    },
+    {
+      "acc": 0.70449252,
+      "epoch": 0.9930238457635718,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.4703452998950005e-06,
+      "loss": 1.3105588,
+      "memory(GiB)": 123.79,
+      "step": 39145,
+      "train_speed(iter/s)": 1.316038
+    },
+    {
+      "acc": 0.6852397,
+      "epoch": 0.9931506849315068,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.469301315818183e-06,
+      "loss": 1.43338261,
+      "memory(GiB)": 123.79,
+      "step": 39150,
+      "train_speed(iter/s)": 1.316057
+    },
+    {
+      "acc": 0.68991947,
+      "epoch": 0.9932775240994419,
+      "grad_norm": 2.75,
+      "learning_rate": 5.468257311099399e-06,
+      "loss": 1.27565947,
+      "memory(GiB)": 123.79,
+      "step": 39155,
+      "train_speed(iter/s)": 1.316077
+    },
+    {
+      "acc": 0.71052427,
+      "epoch": 0.9934043632673769,
+      "grad_norm": 3.75,
+      "learning_rate": 5.467213285784567e-06,
+      "loss": 1.33140278,
+      "memory(GiB)": 123.79,
+      "step": 39160,
+      "train_speed(iter/s)": 1.316096
+    },
+    {
+      "acc": 0.68670979,
+      "epoch": 0.993531202435312,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.466169239919608e-06,
+      "loss": 1.45268612,
+      "memory(GiB)": 123.79,
+      "step": 39165,
+      "train_speed(iter/s)": 1.316115
+    },
+    {
+      "acc": 0.68065143,
+      "epoch": 0.9936580416032471,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.465125173550446e-06,
+      "loss": 1.44554787,
+      "memory(GiB)": 123.79,
+      "step": 39170,
+      "train_speed(iter/s)": 1.316133
+    },
+    {
+      "acc": 0.68982992,
+      "epoch": 0.9937848807711821,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.464081086723001e-06,
+      "loss": 1.41747971,
+      "memory(GiB)": 123.79,
+      "step": 39175,
+      "train_speed(iter/s)": 1.316153
+    },
+    {
+      "acc": 0.67857428,
+      "epoch": 0.9939117199391172,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.4630369794832006e-06,
+      "loss": 1.43591747,
+      "memory(GiB)": 123.79,
+      "step": 39180,
+      "train_speed(iter/s)": 1.316173
+    },
+    {
+      "acc": 0.69379854,
+      "epoch": 0.9940385591070523,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.461992851876963e-06,
+      "loss": 1.41649027,
+      "memory(GiB)": 123.79,
+      "step": 39185,
+      "train_speed(iter/s)": 1.316189
+    },
+    {
+      "acc": 0.68752069,
+      "epoch": 0.9941653982749873,
+      "grad_norm": 4.125,
+      "learning_rate": 5.460948703950218e-06,
+      "loss": 1.47125759,
+      "memory(GiB)": 123.79,
+      "step": 39190,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.67724094,
+      "epoch": 0.9942922374429224,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.459904535748892e-06,
+      "loss": 1.46503201,
+      "memory(GiB)": 123.79,
+      "step": 39195,
+      "train_speed(iter/s)": 1.316228
+    },
+    {
+      "acc": 0.71004758,
+      "epoch": 0.9944190766108574,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.458860347318912e-06,
+      "loss": 1.29629021,
+      "memory(GiB)": 123.79,
+      "step": 39200,
+      "train_speed(iter/s)": 1.316247
+    },
+    {
+      "acc": 0.69321041,
+      "epoch": 0.9945459157787925,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.457816138706203e-06,
+      "loss": 1.45453815,
+      "memory(GiB)": 123.79,
+      "step": 39205,
+      "train_speed(iter/s)": 1.316267
+    },
+    {
+      "acc": 0.6989481,
+      "epoch": 0.9946727549467276,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.456771909956697e-06,
+      "loss": 1.39139128,
+      "memory(GiB)": 123.79,
+      "step": 39210,
+      "train_speed(iter/s)": 1.316286
+    },
+    {
+      "acc": 0.69691505,
+      "epoch": 0.9947995941146626,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.455727661116324e-06,
+      "loss": 1.37917366,
+      "memory(GiB)": 123.79,
+      "step": 39215,
+      "train_speed(iter/s)": 1.316305
+    },
+    {
+      "acc": 0.69432282,
+      "epoch": 0.9949264332825977,
+      "grad_norm": 7.71875,
+      "learning_rate": 5.454683392231014e-06,
+      "loss": 1.41838541,
+      "memory(GiB)": 123.79,
+      "step": 39220,
+      "train_speed(iter/s)": 1.316325
+    },
+    {
+      "acc": 0.6935174,
+      "epoch": 0.9950532724505328,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.453639103346697e-06,
+      "loss": 1.36184273,
+      "memory(GiB)": 123.79,
+      "step": 39225,
+      "train_speed(iter/s)": 1.316344
+    },
+    {
+      "acc": 0.68579698,
+      "epoch": 0.9951801116184678,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.452594794509307e-06,
+      "loss": 1.3372056,
+      "memory(GiB)": 123.79,
+      "step": 39230,
+      "train_speed(iter/s)": 1.316364
+    },
+    {
+      "acc": 0.69323969,
+      "epoch": 0.9953069507864029,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.4515504657647765e-06,
+      "loss": 1.42372389,
+      "memory(GiB)": 123.79,
+      "step": 39235,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.68268766,
+      "epoch": 0.9954337899543378,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.450506117159044e-06,
+      "loss": 1.46206198,
+      "memory(GiB)": 123.79,
+      "step": 39240,
+      "train_speed(iter/s)": 1.316403
+    },
+    {
+      "acc": 0.68525429,
+      "epoch": 0.9955606291222729,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.449461748738037e-06,
+      "loss": 1.43083515,
+      "memory(GiB)": 123.79,
+      "step": 39245,
+      "train_speed(iter/s)": 1.316422
+    },
+    {
+      "acc": 0.70146265,
+      "epoch": 0.995687468290208,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.448417360547699e-06,
+      "loss": 1.30406837,
+      "memory(GiB)": 123.79,
+      "step": 39250,
+      "train_speed(iter/s)": 1.31644
+    },
+    {
+      "acc": 0.69873371,
+      "epoch": 0.995814307458143,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.44737295263396e-06,
+      "loss": 1.32116165,
+      "memory(GiB)": 123.79,
+      "step": 39255,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.67807751,
+      "epoch": 0.9959411466260781,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.446328525042764e-06,
+      "loss": 1.46262512,
+      "memory(GiB)": 123.79,
+      "step": 39260,
+      "train_speed(iter/s)": 1.316477
+    },
+    {
+      "acc": 0.67917271,
+      "epoch": 0.9960679857940132,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.4452840778200456e-06,
+      "loss": 1.45545216,
+      "memory(GiB)": 123.79,
+      "step": 39265,
+      "train_speed(iter/s)": 1.316496
+    },
+    {
+      "acc": 0.6812026,
+      "epoch": 0.9961948249619482,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.444239611011746e-06,
+      "loss": 1.38926449,
+      "memory(GiB)": 123.79,
+      "step": 39270,
+      "train_speed(iter/s)": 1.316514
+    },
+    {
+      "acc": 0.681042,
+      "epoch": 0.9963216641298833,
+      "grad_norm": 2.671875,
+      "learning_rate": 5.443195124663804e-06,
+      "loss": 1.42513447,
+      "memory(GiB)": 123.79,
+      "step": 39275,
+      "train_speed(iter/s)": 1.316532
+    },
+    {
+      "acc": 0.68281116,
+      "epoch": 0.9964485032978183,
+      "grad_norm": 3.75,
+      "learning_rate": 5.442150618822162e-06,
+      "loss": 1.41844282,
+      "memory(GiB)": 123.79,
+      "step": 39280,
+      "train_speed(iter/s)": 1.316551
+    },
+    {
+      "acc": 0.69515405,
+      "epoch": 0.9965753424657534,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.441106093532762e-06,
+      "loss": 1.43543053,
+      "memory(GiB)": 123.79,
+      "step": 39285,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.69834795,
+      "epoch": 0.9967021816336885,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.440061548841546e-06,
+      "loss": 1.36860847,
+      "memory(GiB)": 123.79,
+      "step": 39290,
+      "train_speed(iter/s)": 1.316589
+    },
+    {
+      "acc": 0.69379616,
+      "epoch": 0.9968290208016235,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.43901698479446e-06,
+      "loss": 1.40147877,
+      "memory(GiB)": 123.79,
+      "step": 39295,
+      "train_speed(iter/s)": 1.316608
+    },
+    {
+      "acc": 0.6789464,
+      "epoch": 0.9969558599695586,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.4379724014374455e-06,
+      "loss": 1.43141918,
+      "memory(GiB)": 123.79,
+      "step": 39300,
+      "train_speed(iter/s)": 1.316628
+    },
+    {
+      "acc": 0.6681241,
+      "epoch": 0.9970826991374937,
+      "grad_norm": 2.625,
+      "learning_rate": 5.436927798816448e-06,
+      "loss": 1.43567295,
+      "memory(GiB)": 123.79,
+      "step": 39305,
+      "train_speed(iter/s)": 1.316646
+    },
+    {
+      "acc": 0.70633059,
+      "epoch": 0.9972095383054287,
+      "grad_norm": 5.625,
+      "learning_rate": 5.4358831769774174e-06,
+      "loss": 1.31346521,
+      "memory(GiB)": 123.79,
+      "step": 39310,
+      "train_speed(iter/s)": 1.316665
+    },
+    {
+      "acc": 0.68907228,
+      "epoch": 0.9973363774733638,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.434838535966298e-06,
+      "loss": 1.43180838,
+      "memory(GiB)": 123.79,
+      "step": 39315,
+      "train_speed(iter/s)": 1.316684
+    },
+    {
+      "acc": 0.67403293,
+      "epoch": 0.9974632166412988,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.43379387582904e-06,
+      "loss": 1.39573364,
+      "memory(GiB)": 123.79,
+      "step": 39320,
+      "train_speed(iter/s)": 1.316704
+    },
+    {
+      "acc": 0.68983135,
+      "epoch": 0.9975900558092339,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.432749196611587e-06,
+      "loss": 1.36199045,
+      "memory(GiB)": 123.79,
+      "step": 39325,
+      "train_speed(iter/s)": 1.316723
+    },
+    {
+      "acc": 0.6851903,
+      "epoch": 0.997716894977169,
+      "grad_norm": 3.125,
+      "learning_rate": 5.431704498359896e-06,
+      "loss": 1.36537571,
+      "memory(GiB)": 123.79,
+      "step": 39330,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.69444399,
+      "epoch": 0.997843734145104,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.43065978111991e-06,
+      "loss": 1.38934097,
+      "memory(GiB)": 123.79,
+      "step": 39335,
+      "train_speed(iter/s)": 1.316761
+    },
+    {
+      "acc": 0.68624964,
+      "epoch": 0.9979705733130391,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.429615044937586e-06,
+      "loss": 1.37298536,
+      "memory(GiB)": 123.79,
+      "step": 39340,
+      "train_speed(iter/s)": 1.316781
+    },
+    {
+      "acc": 0.68142376,
+      "epoch": 0.9980974124809742,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.4285702898588754e-06,
+      "loss": 1.45214577,
+      "memory(GiB)": 123.79,
+      "step": 39345,
+      "train_speed(iter/s)": 1.3168
+    },
+    {
+      "acc": 0.68821974,
+      "epoch": 0.9982242516489092,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.427525515929729e-06,
+      "loss": 1.37411966,
+      "memory(GiB)": 123.79,
+      "step": 39350,
+      "train_speed(iter/s)": 1.316819
+    },
+    {
+      "acc": 0.70207453,
+      "epoch": 0.9983510908168443,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.426480723196102e-06,
+      "loss": 1.30172405,
+      "memory(GiB)": 123.79,
+      "step": 39355,
+      "train_speed(iter/s)": 1.316839
+    },
+    {
+      "acc": 0.68965087,
+      "epoch": 0.9984779299847792,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.425435911703948e-06,
+      "loss": 1.38429203,
+      "memory(GiB)": 123.79,
+      "step": 39360,
+      "train_speed(iter/s)": 1.316856
+    },
+    {
+      "acc": 0.68453608,
+      "epoch": 0.9986047691527143,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.424391081499223e-06,
+      "loss": 1.40196257,
+      "memory(GiB)": 123.79,
+      "step": 39365,
+      "train_speed(iter/s)": 1.316876
+    },
+    {
+      "acc": 0.68593454,
+      "epoch": 0.9987316083206494,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.423346232627884e-06,
+      "loss": 1.39465332,
+      "memory(GiB)": 123.79,
+      "step": 39370,
+      "train_speed(iter/s)": 1.316895
+    },
+    {
+      "acc": 0.67786765,
+      "epoch": 0.9988584474885844,
+      "grad_norm": 3.75,
+      "learning_rate": 5.422301365135887e-06,
+      "loss": 1.4438406,
+      "memory(GiB)": 123.79,
+      "step": 39375,
+      "train_speed(iter/s)": 1.316912
+    },
+    {
+      "acc": 0.67514257,
+      "epoch": 0.9989852866565195,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.421256479069191e-06,
+      "loss": 1.42497129,
+      "memory(GiB)": 123.79,
+      "step": 39380,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.69954915,
+      "epoch": 0.9991121258244546,
+      "grad_norm": 3.375,
+      "learning_rate": 5.420211574473754e-06,
+      "loss": 1.32207718,
+      "memory(GiB)": 123.79,
+      "step": 39385,
+      "train_speed(iter/s)": 1.316952
+    },
+    {
+      "acc": 0.68957138,
+      "epoch": 0.9992389649923896,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.419166651395536e-06,
+      "loss": 1.380233,
+      "memory(GiB)": 123.79,
+      "step": 39390,
+      "train_speed(iter/s)": 1.316971
+    },
+    {
+      "acc": 0.7015944,
+      "epoch": 0.9993658041603247,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.418121709880497e-06,
+      "loss": 1.34586754,
+      "memory(GiB)": 123.79,
+      "step": 39395,
+      "train_speed(iter/s)": 1.31699
+    },
+    {
+      "acc": 0.68557234,
+      "epoch": 0.9994926433282597,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.4170767499746e-06,
+      "loss": 1.37385893,
+      "memory(GiB)": 123.79,
+      "step": 39400,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.68342791,
+      "epoch": 0.9996194824961948,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.416031771723803e-06,
+      "loss": 1.42451105,
+      "memory(GiB)": 123.79,
+      "step": 39405,
+      "train_speed(iter/s)": 1.317029
+    },
+    {
+      "acc": 0.69478769,
+      "epoch": 0.9997463216641299,
+      "grad_norm": 3.890625,
+      "learning_rate": 5.414986775174073e-06,
+      "loss": 1.35866356,
+      "memory(GiB)": 123.79,
+      "step": 39410,
+      "train_speed(iter/s)": 1.317048
+    },
+    {
+      "acc": 0.69848895,
+      "epoch": 0.9998731608320649,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.41394176037137e-06,
+      "loss": 1.35277996,
+      "memory(GiB)": 123.79,
+      "step": 39415,
+      "train_speed(iter/s)": 1.317067
+    },
+    {
+      "acc": 0.67717419,
+      "epoch": 1.0,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.412896727361663e-06,
+      "loss": 1.4732089,
+      "memory(GiB)": 123.79,
+      "step": 39420,
+      "train_speed(iter/s)": 1.317054
+    },
+    {
+      "acc": 0.69368315,
+      "epoch": 1.000126839167935,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.411851676190912e-06,
+      "loss": 1.37227974,
+      "memory(GiB)": 123.79,
+      "step": 39425,
+      "train_speed(iter/s)": 1.317054
+    },
+    {
+      "acc": 0.68234239,
+      "epoch": 1.0002536783358702,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.4108066069050864e-06,
+      "loss": 1.41736851,
+      "memory(GiB)": 123.79,
+      "step": 39430,
+      "train_speed(iter/s)": 1.317072
+    },
+    {
+      "acc": 0.66503525,
+      "epoch": 1.0003805175038052,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.409761519550153e-06,
+      "loss": 1.43278503,
+      "memory(GiB)": 123.79,
+      "step": 39435,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.69783616,
+      "epoch": 1.0005073566717402,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.408716414172077e-06,
+      "loss": 1.3454525,
+      "memory(GiB)": 123.79,
+      "step": 39440,
+      "train_speed(iter/s)": 1.317108
+    },
+    {
+      "acc": 0.69810114,
+      "epoch": 1.0006341958396754,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.407671290816829e-06,
+      "loss": 1.37866592,
+      "memory(GiB)": 123.79,
+      "step": 39445,
+      "train_speed(iter/s)": 1.317127
+    },
+    {
+      "acc": 0.6623704,
+      "epoch": 1.0007610350076104,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.406626149530378e-06,
+      "loss": 1.50022812,
+      "memory(GiB)": 123.79,
+      "step": 39450,
+      "train_speed(iter/s)": 1.317145
+    },
+    {
+      "acc": 0.69952092,
+      "epoch": 1.0008878741755454,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.405580990358692e-06,
+      "loss": 1.31767349,
+      "memory(GiB)": 123.79,
+      "step": 39455,
+      "train_speed(iter/s)": 1.317164
+    },
+    {
+      "acc": 0.69271793,
+      "epoch": 1.0010147133434804,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.404535813347746e-06,
+      "loss": 1.36808348,
+      "memory(GiB)": 123.79,
+      "step": 39460,
+      "train_speed(iter/s)": 1.317183
+    },
+    {
+      "acc": 0.69211392,
+      "epoch": 1.0011415525114156,
+      "grad_norm": 4.4375,
+      "learning_rate": 5.403490618543505e-06,
+      "loss": 1.42602024,
+      "memory(GiB)": 123.79,
+      "step": 39465,
+      "train_speed(iter/s)": 1.317202
+    },
+    {
+      "acc": 0.69862709,
+      "epoch": 1.0012683916793506,
+      "grad_norm": 3.0,
+      "learning_rate": 5.40244540599195e-06,
+      "loss": 1.3604723,
+      "memory(GiB)": 123.79,
+      "step": 39470,
+      "train_speed(iter/s)": 1.317221
+    },
+    {
+      "acc": 0.68596458,
+      "epoch": 1.0013952308472855,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.401400175739045e-06,
+      "loss": 1.43681297,
+      "memory(GiB)": 123.79,
+      "step": 39475,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.67743073,
+      "epoch": 1.0015220700152208,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.400354927830769e-06,
+      "loss": 1.40911913,
+      "memory(GiB)": 123.79,
+      "step": 39480,
+      "train_speed(iter/s)": 1.317259
+    },
+    {
+      "acc": 0.69587727,
+      "epoch": 1.0016489091831557,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.399309662313097e-06,
+      "loss": 1.3579464,
+      "memory(GiB)": 123.79,
+      "step": 39485,
+      "train_speed(iter/s)": 1.317278
+    },
+    {
+      "acc": 0.68159256,
+      "epoch": 1.0017757483510907,
+      "grad_norm": 3.125,
+      "learning_rate": 5.3982643792320024e-06,
+      "loss": 1.39621754,
+      "memory(GiB)": 123.79,
+      "step": 39490,
+      "train_speed(iter/s)": 1.317295
+    },
+    {
+      "acc": 0.68458252,
+      "epoch": 1.001902587519026,
+      "grad_norm": 2.75,
+      "learning_rate": 5.397219078633462e-06,
+      "loss": 1.37658434,
+      "memory(GiB)": 123.79,
+      "step": 39495,
+      "train_speed(iter/s)": 1.317312
+    },
+    {
+      "acc": 0.68769717,
+      "epoch": 1.002029426686961,
+      "grad_norm": 3.75,
+      "learning_rate": 5.3961737605634546e-06,
+      "loss": 1.41844559,
+      "memory(GiB)": 123.79,
+      "step": 39500,
+      "train_speed(iter/s)": 1.31733
+    },
+    {
+      "acc": 0.69708586,
+      "epoch": 1.002156265854896,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.395128425067954e-06,
+      "loss": 1.37539644,
+      "memory(GiB)": 123.79,
+      "step": 39505,
+      "train_speed(iter/s)": 1.317347
+    },
+    {
+      "acc": 0.70254121,
+      "epoch": 1.0022831050228311,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.394083072192944e-06,
+      "loss": 1.32537155,
+      "memory(GiB)": 123.79,
+      "step": 39510,
+      "train_speed(iter/s)": 1.317366
+    },
+    {
+      "acc": 0.69805222,
+      "epoch": 1.0024099441907661,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.393037701984399e-06,
+      "loss": 1.30691223,
+      "memory(GiB)": 123.79,
+      "step": 39515,
+      "train_speed(iter/s)": 1.317383
+    },
+    {
+      "acc": 0.68130646,
+      "epoch": 1.0025367833587011,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.391992314488303e-06,
+      "loss": 1.3681675,
+      "memory(GiB)": 123.79,
+      "step": 39520,
+      "train_speed(iter/s)": 1.317402
+    },
+    {
+      "acc": 0.68218374,
+      "epoch": 1.0026636225266363,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.3909469097506314e-06,
+      "loss": 1.37114277,
+      "memory(GiB)": 123.79,
+      "step": 39525,
+      "train_speed(iter/s)": 1.31742
+    },
+    {
+      "acc": 0.66949816,
+      "epoch": 1.0027904616945713,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.389901487817373e-06,
+      "loss": 1.41762953,
+      "memory(GiB)": 123.79,
+      "step": 39530,
+      "train_speed(iter/s)": 1.317438
+    },
+    {
+      "acc": 0.7020669,
+      "epoch": 1.0029173008625063,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.388856048734505e-06,
+      "loss": 1.3792696,
+      "memory(GiB)": 123.79,
+      "step": 39535,
+      "train_speed(iter/s)": 1.317457
+    },
+    {
+      "acc": 0.6776125,
+      "epoch": 1.0030441400304415,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.3878105925480115e-06,
+      "loss": 1.47417698,
+      "memory(GiB)": 123.79,
+      "step": 39540,
+      "train_speed(iter/s)": 1.317475
+    },
+    {
+      "acc": 0.68474913,
+      "epoch": 1.0031709791983765,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.3867651193038765e-06,
+      "loss": 1.39078426,
+      "memory(GiB)": 123.79,
+      "step": 39545,
+      "train_speed(iter/s)": 1.317494
+    },
+    {
+      "acc": 0.68636532,
+      "epoch": 1.0032978183663115,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.385719629048086e-06,
+      "loss": 1.37544527,
+      "memory(GiB)": 123.79,
+      "step": 39550,
+      "train_speed(iter/s)": 1.317512
+    },
+    {
+      "acc": 0.67383018,
+      "epoch": 1.0034246575342465,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.384674121826622e-06,
+      "loss": 1.42656765,
+      "memory(GiB)": 123.79,
+      "step": 39555,
+      "train_speed(iter/s)": 1.317532
+    },
+    {
+      "acc": 0.68708415,
+      "epoch": 1.0035514967021817,
+      "grad_norm": 3.5,
+      "learning_rate": 5.383628597685474e-06,
+      "loss": 1.38360672,
+      "memory(GiB)": 123.79,
+      "step": 39560,
+      "train_speed(iter/s)": 1.317551
+    },
+    {
+      "acc": 0.70953994,
+      "epoch": 1.0036783358701167,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.382583056670627e-06,
+      "loss": 1.31590328,
+      "memory(GiB)": 123.79,
+      "step": 39565,
+      "train_speed(iter/s)": 1.317571
+    },
+    {
+      "acc": 0.68418207,
+      "epoch": 1.0038051750380517,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.38153749882807e-06,
+      "loss": 1.41487122,
+      "memory(GiB)": 123.79,
+      "step": 39570,
+      "train_speed(iter/s)": 1.31759
+    },
+    {
+      "acc": 0.68583717,
+      "epoch": 1.0039320142059869,
+      "grad_norm": 3.375,
+      "learning_rate": 5.38049192420379e-06,
+      "loss": 1.38047123,
+      "memory(GiB)": 123.79,
+      "step": 39575,
+      "train_speed(iter/s)": 1.31761
+    },
+    {
+      "acc": 0.68636007,
+      "epoch": 1.0040588533739219,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.3794463328437766e-06,
+      "loss": 1.38852158,
+      "memory(GiB)": 123.79,
+      "step": 39580,
+      "train_speed(iter/s)": 1.31763
+    },
+    {
+      "acc": 0.69522004,
+      "epoch": 1.0041856925418569,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.3784007247940185e-06,
+      "loss": 1.32494297,
+      "memory(GiB)": 123.79,
+      "step": 39585,
+      "train_speed(iter/s)": 1.317648
+    },
+    {
+      "acc": 0.68404322,
+      "epoch": 1.004312531709792,
+      "grad_norm": 4.75,
+      "learning_rate": 5.377355100100508e-06,
+      "loss": 1.32664642,
+      "memory(GiB)": 123.79,
+      "step": 39590,
+      "train_speed(iter/s)": 1.317668
+    },
+    {
+      "acc": 0.70473781,
+      "epoch": 1.004439370877727,
+      "grad_norm": 3.625,
+      "learning_rate": 5.376309458809235e-06,
+      "loss": 1.32524796,
+      "memory(GiB)": 123.79,
+      "step": 39595,
+      "train_speed(iter/s)": 1.317688
+    },
+    {
+      "acc": 0.67277393,
+      "epoch": 1.004566210045662,
+      "grad_norm": 3.375,
+      "learning_rate": 5.375263800966192e-06,
+      "loss": 1.40313787,
+      "memory(GiB)": 123.79,
+      "step": 39600,
+      "train_speed(iter/s)": 1.317707
+    },
+    {
+      "acc": 0.68057528,
+      "epoch": 1.0046930492135973,
+      "grad_norm": 3.90625,
+      "learning_rate": 5.374218126617371e-06,
+      "loss": 1.37067232,
+      "memory(GiB)": 123.79,
+      "step": 39605,
+      "train_speed(iter/s)": 1.317727
+    },
+    {
+      "acc": 0.68670273,
+      "epoch": 1.0048198883815322,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.373172435808768e-06,
+      "loss": 1.41163721,
+      "memory(GiB)": 123.79,
+      "step": 39610,
+      "train_speed(iter/s)": 1.317746
+    },
+    {
+      "acc": 0.68248396,
+      "epoch": 1.0049467275494672,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.372126728586372e-06,
+      "loss": 1.42729044,
+      "memory(GiB)": 123.79,
+      "step": 39615,
+      "train_speed(iter/s)": 1.317765
+    },
+    {
+      "acc": 0.68330202,
+      "epoch": 1.0050735667174022,
+      "grad_norm": 3.0,
+      "learning_rate": 5.371081004996184e-06,
+      "loss": 1.37844372,
+      "memory(GiB)": 123.79,
+      "step": 39620,
+      "train_speed(iter/s)": 1.317784
+    },
+    {
+      "acc": 0.681142,
+      "epoch": 1.0052004058853374,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.370035265084195e-06,
+      "loss": 1.39417152,
+      "memory(GiB)": 123.79,
+      "step": 39625,
+      "train_speed(iter/s)": 1.317804
+    },
+    {
+      "acc": 0.69228926,
+      "epoch": 1.0053272450532724,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.3689895088964025e-06,
+      "loss": 1.40499563,
+      "memory(GiB)": 123.79,
+      "step": 39630,
+      "train_speed(iter/s)": 1.317823
+    },
+    {
+      "acc": 0.67921481,
+      "epoch": 1.0054540842212074,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.367943736478806e-06,
+      "loss": 1.41826324,
+      "memory(GiB)": 123.79,
+      "step": 39635,
+      "train_speed(iter/s)": 1.317841
+    },
+    {
+      "acc": 0.67239118,
+      "epoch": 1.0055809233891426,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.3668979478774e-06,
+      "loss": 1.46662769,
+      "memory(GiB)": 123.79,
+      "step": 39640,
+      "train_speed(iter/s)": 1.31786
+    },
+    {
+      "acc": 0.69771261,
+      "epoch": 1.0057077625570776,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.3658521431381836e-06,
+      "loss": 1.3662322,
+      "memory(GiB)": 123.79,
+      "step": 39645,
+      "train_speed(iter/s)": 1.317878
+    },
+    {
+      "acc": 0.68688402,
+      "epoch": 1.0058346017250126,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.364806322307158e-06,
+      "loss": 1.38137255,
+      "memory(GiB)": 123.79,
+      "step": 39650,
+      "train_speed(iter/s)": 1.317897
+    },
+    {
+      "acc": 0.68110285,
+      "epoch": 1.0059614408929478,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.363760485430321e-06,
+      "loss": 1.4291976,
+      "memory(GiB)": 123.79,
+      "step": 39655,
+      "train_speed(iter/s)": 1.317916
+    },
+    {
+      "acc": 0.7040452,
+      "epoch": 1.0060882800608828,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.3627146325536725e-06,
+      "loss": 1.39108448,
+      "memory(GiB)": 123.79,
+      "step": 39660,
+      "train_speed(iter/s)": 1.317935
+    },
+    {
+      "acc": 0.70749302,
+      "epoch": 1.0062151192288178,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.361668763723216e-06,
+      "loss": 1.25321522,
+      "memory(GiB)": 123.79,
+      "step": 39665,
+      "train_speed(iter/s)": 1.317953
+    },
+    {
+      "acc": 0.686203,
+      "epoch": 1.006341958396753,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.360622878984954e-06,
+      "loss": 1.35593967,
+      "memory(GiB)": 123.79,
+      "step": 39670,
+      "train_speed(iter/s)": 1.317971
+    },
+    {
+      "acc": 0.6841898,
+      "epoch": 1.006468797564688,
+      "grad_norm": 3.125,
+      "learning_rate": 5.359576978384885e-06,
+      "loss": 1.3938921,
+      "memory(GiB)": 123.79,
+      "step": 39675,
+      "train_speed(iter/s)": 1.317989
+    },
+    {
+      "acc": 0.68829718,
+      "epoch": 1.006595636732623,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.358531061969018e-06,
+      "loss": 1.37073898,
+      "memory(GiB)": 123.79,
+      "step": 39680,
+      "train_speed(iter/s)": 1.318009
+    },
+    {
+      "acc": 0.67991395,
+      "epoch": 1.0067224759005582,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.357485129783351e-06,
+      "loss": 1.4642231,
+      "memory(GiB)": 123.79,
+      "step": 39685,
+      "train_speed(iter/s)": 1.318026
+    },
+    {
+      "acc": 0.68139377,
+      "epoch": 1.0068493150684932,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.356439181873895e-06,
+      "loss": 1.39669971,
+      "memory(GiB)": 123.79,
+      "step": 39690,
+      "train_speed(iter/s)": 1.318045
+    },
+    {
+      "acc": 0.70035653,
+      "epoch": 1.0069761542364282,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.35539321828665e-06,
+      "loss": 1.32493744,
+      "memory(GiB)": 123.79,
+      "step": 39695,
+      "train_speed(iter/s)": 1.318064
+    },
+    {
+      "acc": 0.68632412,
+      "epoch": 1.0071029934043634,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.354347239067625e-06,
+      "loss": 1.3826704,
+      "memory(GiB)": 123.79,
+      "step": 39700,
+      "train_speed(iter/s)": 1.318083
+    },
+    {
+      "acc": 0.69257417,
+      "epoch": 1.0072298325722984,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.3533012442628275e-06,
+      "loss": 1.32326717,
+      "memory(GiB)": 123.79,
+      "step": 39705,
+      "train_speed(iter/s)": 1.318101
+    },
+    {
+      "acc": 0.6991611,
+      "epoch": 1.0073566717402334,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.3522552339182635e-06,
+      "loss": 1.29164982,
+      "memory(GiB)": 123.79,
+      "step": 39710,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.68821697,
+      "epoch": 1.0074835109081683,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.351209208079941e-06,
+      "loss": 1.40398998,
+      "memory(GiB)": 123.79,
+      "step": 39715,
+      "train_speed(iter/s)": 1.318135
+    },
+    {
+      "acc": 0.69508095,
+      "epoch": 1.0076103500761036,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.35016316679387e-06,
+      "loss": 1.39173679,
+      "memory(GiB)": 123.79,
+      "step": 39720,
+      "train_speed(iter/s)": 1.318154
+    },
+    {
+      "acc": 0.70074596,
+      "epoch": 1.0077371892440385,
+      "grad_norm": 3.875,
+      "learning_rate": 5.349117110106059e-06,
+      "loss": 1.29010715,
+      "memory(GiB)": 123.79,
+      "step": 39725,
+      "train_speed(iter/s)": 1.318174
+    },
+    {
+      "acc": 0.6921741,
+      "epoch": 1.0078640284119735,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.34807103806252e-06,
+      "loss": 1.37788153,
+      "memory(GiB)": 123.79,
+      "step": 39730,
+      "train_speed(iter/s)": 1.318193
+    },
+    {
+      "acc": 0.6806313,
+      "epoch": 1.0079908675799087,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.347024950709262e-06,
+      "loss": 1.39078474,
+      "memory(GiB)": 123.79,
+      "step": 39735,
+      "train_speed(iter/s)": 1.318211
+    },
+    {
+      "acc": 0.67520924,
+      "epoch": 1.0081177067478437,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.345978848092297e-06,
+      "loss": 1.44406395,
+      "memory(GiB)": 123.79,
+      "step": 39740,
+      "train_speed(iter/s)": 1.318229
+    },
+    {
+      "acc": 0.69119711,
+      "epoch": 1.0082445459157787,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.344932730257637e-06,
+      "loss": 1.38152332,
+      "memory(GiB)": 123.79,
+      "step": 39745,
+      "train_speed(iter/s)": 1.318247
+    },
+    {
+      "acc": 0.686127,
+      "epoch": 1.008371385083714,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.343886597251298e-06,
+      "loss": 1.38848152,
+      "memory(GiB)": 123.79,
+      "step": 39750,
+      "train_speed(iter/s)": 1.318265
+    },
+    {
+      "acc": 0.69275446,
+      "epoch": 1.008498224251649,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.342840449119287e-06,
+      "loss": 1.41497478,
+      "memory(GiB)": 123.79,
+      "step": 39755,
+      "train_speed(iter/s)": 1.318284
+    },
+    {
+      "acc": 0.69269013,
+      "epoch": 1.008625063419584,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.341794285907627e-06,
+      "loss": 1.39422016,
+      "memory(GiB)": 123.79,
+      "step": 39760,
+      "train_speed(iter/s)": 1.318302
+    },
+    {
+      "acc": 0.67844458,
+      "epoch": 1.0087519025875191,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.340748107662324e-06,
+      "loss": 1.45389442,
+      "memory(GiB)": 123.79,
+      "step": 39765,
+      "train_speed(iter/s)": 1.318321
+    },
+    {
+      "acc": 0.68420835,
+      "epoch": 1.0088787417554541,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.339701914429402e-06,
+      "loss": 1.42164669,
+      "memory(GiB)": 123.79,
+      "step": 39770,
+      "train_speed(iter/s)": 1.31834
+    },
+    {
+      "acc": 0.67897949,
+      "epoch": 1.009005580923389,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.338655706254871e-06,
+      "loss": 1.50231361,
+      "memory(GiB)": 123.79,
+      "step": 39775,
+      "train_speed(iter/s)": 1.31836
+    },
+    {
+      "acc": 0.67595329,
+      "epoch": 1.009132420091324,
+      "grad_norm": 2.5625,
+      "learning_rate": 5.33760948318475e-06,
+      "loss": 1.40755215,
+      "memory(GiB)": 123.79,
+      "step": 39780,
+      "train_speed(iter/s)": 1.318378
+    },
+    {
+      "acc": 0.69285707,
+      "epoch": 1.0092592592592593,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.336563245265056e-06,
+      "loss": 1.42123585,
+      "memory(GiB)": 123.79,
+      "step": 39785,
+      "train_speed(iter/s)": 1.318397
+    },
+    {
+      "acc": 0.67762051,
+      "epoch": 1.0093860984271943,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.3355169925418095e-06,
+      "loss": 1.46956139,
+      "memory(GiB)": 123.79,
+      "step": 39790,
+      "train_speed(iter/s)": 1.318416
+    },
+    {
+      "acc": 0.68200645,
+      "epoch": 1.0095129375951293,
+      "grad_norm": 2.75,
+      "learning_rate": 5.334470725061027e-06,
+      "loss": 1.41321602,
+      "memory(GiB)": 123.79,
+      "step": 39795,
+      "train_speed(iter/s)": 1.318434
+    },
+    {
+      "acc": 0.69340687,
+      "epoch": 1.0096397767630645,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.333424442868729e-06,
+      "loss": 1.43382902,
+      "memory(GiB)": 123.79,
+      "step": 39800,
+      "train_speed(iter/s)": 1.318453
+    },
+    {
+      "acc": 0.68250265,
+      "epoch": 1.0097666159309995,
+      "grad_norm": 3.625,
+      "learning_rate": 5.3323781460109345e-06,
+      "loss": 1.45288448,
+      "memory(GiB)": 123.79,
+      "step": 39805,
+      "train_speed(iter/s)": 1.318472
+    },
+    {
+      "acc": 0.69256802,
+      "epoch": 1.0098934550989345,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.3313318345336665e-06,
+      "loss": 1.38281574,
+      "memory(GiB)": 123.79,
+      "step": 39810,
+      "train_speed(iter/s)": 1.318491
+    },
+    {
+      "acc": 0.70444841,
+      "epoch": 1.0100202942668697,
+      "grad_norm": 3.5,
+      "learning_rate": 5.330285508482944e-06,
+      "loss": 1.36813908,
+      "memory(GiB)": 123.79,
+      "step": 39815,
+      "train_speed(iter/s)": 1.318509
+    },
+    {
+      "acc": 0.6872417,
+      "epoch": 1.0101471334348047,
+      "grad_norm": 5.0625,
+      "learning_rate": 5.3292391679047905e-06,
+      "loss": 1.38960476,
+      "memory(GiB)": 123.79,
+      "step": 39820,
+      "train_speed(iter/s)": 1.318528
+    },
+    {
+      "acc": 0.68280497,
+      "epoch": 1.0102739726027397,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.328192812845228e-06,
+      "loss": 1.43255072,
+      "memory(GiB)": 123.79,
+      "step": 39825,
+      "train_speed(iter/s)": 1.318546
+    },
+    {
+      "acc": 0.68831902,
+      "epoch": 1.0104008117706749,
+      "grad_norm": 3.96875,
+      "learning_rate": 5.3271464433502805e-06,
+      "loss": 1.35324793,
+      "memory(GiB)": 123.79,
+      "step": 39830,
+      "train_speed(iter/s)": 1.318565
+    },
+    {
+      "acc": 0.68393459,
+      "epoch": 1.0105276509386099,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.3261000594659715e-06,
+      "loss": 1.45954323,
+      "memory(GiB)": 123.79,
+      "step": 39835,
+      "train_speed(iter/s)": 1.318583
+    },
+    {
+      "acc": 0.6715951,
+      "epoch": 1.0106544901065448,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.3250536612383275e-06,
+      "loss": 1.48155947,
+      "memory(GiB)": 123.79,
+      "step": 39840,
+      "train_speed(iter/s)": 1.318601
+    },
+    {
+      "acc": 0.69521241,
+      "epoch": 1.01078132927448,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.32400724871337e-06,
+      "loss": 1.32716789,
+      "memory(GiB)": 123.79,
+      "step": 39845,
+      "train_speed(iter/s)": 1.31862
+    },
+    {
+      "acc": 0.69465055,
+      "epoch": 1.010908168442415,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.322960821937129e-06,
+      "loss": 1.36022806,
+      "memory(GiB)": 123.79,
+      "step": 39850,
+      "train_speed(iter/s)": 1.318642
+    },
+    {
+      "acc": 0.67148142,
+      "epoch": 1.01103500761035,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.321914380955628e-06,
+      "loss": 1.45008564,
+      "memory(GiB)": 123.79,
+      "step": 39855,
+      "train_speed(iter/s)": 1.318661
+    },
+    {
+      "acc": 0.69458008,
+      "epoch": 1.0111618467782852,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.320867925814896e-06,
+      "loss": 1.37869263,
+      "memory(GiB)": 123.79,
+      "step": 39860,
+      "train_speed(iter/s)": 1.318681
+    },
+    {
+      "acc": 0.69212408,
+      "epoch": 1.0112886859462202,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.31982145656096e-06,
+      "loss": 1.39726505,
+      "memory(GiB)": 123.79,
+      "step": 39865,
+      "train_speed(iter/s)": 1.318701
+    },
+    {
+      "acc": 0.68076038,
+      "epoch": 1.0114155251141552,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.318774973239849e-06,
+      "loss": 1.345051,
+      "memory(GiB)": 123.79,
+      "step": 39870,
+      "train_speed(iter/s)": 1.318719
+    },
+    {
+      "acc": 0.6858202,
+      "epoch": 1.0115423642820902,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.31772847589759e-06,
+      "loss": 1.41058407,
+      "memory(GiB)": 123.79,
+      "step": 39875,
+      "train_speed(iter/s)": 1.318737
+    },
+    {
+      "acc": 0.6969974,
+      "epoch": 1.0116692034500254,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.316681964580215e-06,
+      "loss": 1.36325932,
+      "memory(GiB)": 123.79,
+      "step": 39880,
+      "train_speed(iter/s)": 1.318756
+    },
+    {
+      "acc": 0.68981957,
+      "epoch": 1.0117960426179604,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.315635439333753e-06,
+      "loss": 1.39392395,
+      "memory(GiB)": 123.79,
+      "step": 39885,
+      "train_speed(iter/s)": 1.318771
+    },
+    {
+      "acc": 0.67873321,
+      "epoch": 1.0119228817858954,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.314588900204235e-06,
+      "loss": 1.40418472,
+      "memory(GiB)": 123.79,
+      "step": 39890,
+      "train_speed(iter/s)": 1.318789
+    },
+    {
+      "acc": 0.68271761,
+      "epoch": 1.0120497209538306,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.313542347237692e-06,
+      "loss": 1.44117231,
+      "memory(GiB)": 123.79,
+      "step": 39895,
+      "train_speed(iter/s)": 1.318807
+    },
+    {
+      "acc": 0.67205906,
+      "epoch": 1.0121765601217656,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.312495780480159e-06,
+      "loss": 1.41751556,
+      "memory(GiB)": 123.79,
+      "step": 39900,
+      "train_speed(iter/s)": 1.318825
+    },
+    {
+      "acc": 0.69137673,
+      "epoch": 1.0123033992897006,
+      "grad_norm": 3.0,
+      "learning_rate": 5.311449199977664e-06,
+      "loss": 1.39275446,
+      "memory(GiB)": 123.79,
+      "step": 39905,
+      "train_speed(iter/s)": 1.318844
+    },
+    {
+      "acc": 0.69893003,
+      "epoch": 1.0124302384576358,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.310402605776245e-06,
+      "loss": 1.35447292,
+      "memory(GiB)": 123.79,
+      "step": 39910,
+      "train_speed(iter/s)": 1.318861
+    },
+    {
+      "acc": 0.68962584,
+      "epoch": 1.0125570776255708,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.309355997921931e-06,
+      "loss": 1.37687893,
+      "memory(GiB)": 123.79,
+      "step": 39915,
+      "train_speed(iter/s)": 1.31888
+    },
+    {
+      "acc": 0.70192137,
+      "epoch": 1.0126839167935058,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.308309376460761e-06,
+      "loss": 1.39544487,
+      "memory(GiB)": 123.79,
+      "step": 39920,
+      "train_speed(iter/s)": 1.318898
+    },
+    {
+      "acc": 0.70122757,
+      "epoch": 1.012810755961441,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.307262741438767e-06,
+      "loss": 1.33195782,
+      "memory(GiB)": 123.79,
+      "step": 39925,
+      "train_speed(iter/s)": 1.318916
+    },
+    {
+      "acc": 0.69792118,
+      "epoch": 1.012937595129376,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.3062160929019855e-06,
+      "loss": 1.38867912,
+      "memory(GiB)": 123.79,
+      "step": 39930,
+      "train_speed(iter/s)": 1.318936
+    },
+    {
+      "acc": 0.69845943,
+      "epoch": 1.013064434297311,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.305169430896454e-06,
+      "loss": 1.33994579,
+      "memory(GiB)": 123.79,
+      "step": 39935,
+      "train_speed(iter/s)": 1.318955
+    },
+    {
+      "acc": 0.68758974,
+      "epoch": 1.013191273465246,
+      "grad_norm": 2.5625,
+      "learning_rate": 5.304122755468209e-06,
+      "loss": 1.31226778,
+      "memory(GiB)": 123.79,
+      "step": 39940,
+      "train_speed(iter/s)": 1.318973
+    },
+    {
+      "acc": 0.69208941,
+      "epoch": 1.0133181126331812,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.303076066663286e-06,
+      "loss": 1.29714088,
+      "memory(GiB)": 123.79,
+      "step": 39945,
+      "train_speed(iter/s)": 1.318993
+    },
+    {
+      "acc": 0.67696743,
+      "epoch": 1.0134449518011162,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.302029364527726e-06,
+      "loss": 1.41363287,
+      "memory(GiB)": 123.79,
+      "step": 39950,
+      "train_speed(iter/s)": 1.319012
+    },
+    {
+      "acc": 0.67918639,
+      "epoch": 1.0135717909690511,
+      "grad_norm": 3.5,
+      "learning_rate": 5.3009826491075645e-06,
+      "loss": 1.46551323,
+      "memory(GiB)": 123.79,
+      "step": 39955,
+      "train_speed(iter/s)": 1.319031
+    },
+    {
+      "acc": 0.68849187,
+      "epoch": 1.0136986301369864,
+      "grad_norm": 4.4375,
+      "learning_rate": 5.299935920448843e-06,
+      "loss": 1.33952532,
+      "memory(GiB)": 123.79,
+      "step": 39960,
+      "train_speed(iter/s)": 1.31905
+    },
+    {
+      "acc": 0.69071493,
+      "epoch": 1.0138254693049213,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.298889178597599e-06,
+      "loss": 1.36563663,
+      "memory(GiB)": 123.79,
+      "step": 39965,
+      "train_speed(iter/s)": 1.319069
+    },
+    {
+      "acc": 0.6777895,
+      "epoch": 1.0139523084728563,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.297842423599877e-06,
+      "loss": 1.39653149,
+      "memory(GiB)": 123.79,
+      "step": 39970,
+      "train_speed(iter/s)": 1.319089
+    },
+    {
+      "acc": 0.69221454,
+      "epoch": 1.0140791476407915,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.296795655501714e-06,
+      "loss": 1.42941971,
+      "memory(GiB)": 123.79,
+      "step": 39975,
+      "train_speed(iter/s)": 1.319109
+    },
+    {
+      "acc": 0.682302,
+      "epoch": 1.0142059868087265,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.295748874349155e-06,
+      "loss": 1.43743649,
+      "memory(GiB)": 123.79,
+      "step": 39980,
+      "train_speed(iter/s)": 1.319128
+    },
+    {
+      "acc": 0.68350196,
+      "epoch": 1.0143328259766615,
+      "grad_norm": 3.125,
+      "learning_rate": 5.294702080188236e-06,
+      "loss": 1.35843019,
+      "memory(GiB)": 123.79,
+      "step": 39985,
+      "train_speed(iter/s)": 1.319148
+    },
+    {
+      "acc": 0.68375239,
+      "epoch": 1.0144596651445967,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.293655273065008e-06,
+      "loss": 1.40181332,
+      "memory(GiB)": 123.79,
+      "step": 39990,
+      "train_speed(iter/s)": 1.319166
+    },
+    {
+      "acc": 0.70034294,
+      "epoch": 1.0145865043125317,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.2926084530255076e-06,
+      "loss": 1.36825962,
+      "memory(GiB)": 123.79,
+      "step": 39995,
+      "train_speed(iter/s)": 1.319185
+    },
+    {
+      "acc": 0.68538475,
+      "epoch": 1.0147133434804667,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.291561620115781e-06,
+      "loss": 1.37764235,
+      "memory(GiB)": 123.79,
+      "step": 40000,
+      "train_speed(iter/s)": 1.319206
+    },
+    {
+      "epoch": 1.0147133434804667,
+      "eval_acc": 0.6749359412100063,
+      "eval_loss": 1.3566780090332031,
+      "eval_runtime": 69.3132,
+      "eval_samples_per_second": 91.902,
+      "eval_steps_per_second": 22.983,
+      "step": 40000
+    },
+    {
+      "acc": 0.68577514,
+      "epoch": 1.014840182648402,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.290514774381874e-06,
+      "loss": 1.44029226,
+      "memory(GiB)": 123.79,
+      "step": 40005,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.69018898,
+      "epoch": 1.014967021816337,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.289467915869829e-06,
+      "loss": 1.38287888,
+      "memory(GiB)": 123.79,
+      "step": 40010,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.69527311,
+      "epoch": 1.015093860984272,
+      "grad_norm": 3.625,
+      "learning_rate": 5.288421044625694e-06,
+      "loss": 1.38699942,
+      "memory(GiB)": 123.79,
+      "step": 40015,
+      "train_speed(iter/s)": 1.31571
+    },
+    {
+      "acc": 0.68901997,
+      "epoch": 1.0152207001522071,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.287374160695513e-06,
+      "loss": 1.39893484,
+      "memory(GiB)": 123.79,
+      "step": 40020,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.68518386,
+      "epoch": 1.015347539320142,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.286327264125332e-06,
+      "loss": 1.46309395,
+      "memory(GiB)": 123.79,
+      "step": 40025,
+      "train_speed(iter/s)": 1.315751
+    },
+    {
+      "acc": 0.68118553,
+      "epoch": 1.015474378488077,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.285280354961202e-06,
+      "loss": 1.44921455,
+      "memory(GiB)": 123.79,
+      "step": 40030,
+      "train_speed(iter/s)": 1.315771
+    },
+    {
+      "acc": 0.68407006,
+      "epoch": 1.015601217656012,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.284233433249167e-06,
+      "loss": 1.31314297,
+      "memory(GiB)": 123.79,
+      "step": 40035,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.6939908,
+      "epoch": 1.0157280568239473,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.283186499035276e-06,
+      "loss": 1.34378748,
+      "memory(GiB)": 123.79,
+      "step": 40040,
+      "train_speed(iter/s)": 1.315812
+    },
+    {
+      "acc": 0.69148202,
+      "epoch": 1.0158548959918823,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.2821395523655795e-06,
+      "loss": 1.33001728,
+      "memory(GiB)": 123.79,
+      "step": 40045,
+      "train_speed(iter/s)": 1.315832
+    },
+    {
+      "acc": 0.70178843,
+      "epoch": 1.0159817351598173,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.281092593286127e-06,
+      "loss": 1.33119488,
+      "memory(GiB)": 123.79,
+      "step": 40050,
+      "train_speed(iter/s)": 1.315849
+    },
+    {
+      "acc": 0.6828763,
+      "epoch": 1.0161085743277525,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.280045621842964e-06,
+      "loss": 1.37732067,
+      "memory(GiB)": 123.79,
+      "step": 40055,
+      "train_speed(iter/s)": 1.31587
+    },
+    {
+      "acc": 0.66566057,
+      "epoch": 1.0162354134956875,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.278998638082148e-06,
+      "loss": 1.41148796,
+      "memory(GiB)": 123.79,
+      "step": 40060,
+      "train_speed(iter/s)": 1.315889
+    },
+    {
+      "acc": 0.68759656,
+      "epoch": 1.0163622526636225,
+      "grad_norm": 2.75,
+      "learning_rate": 5.277951642049722e-06,
+      "loss": 1.41410494,
+      "memory(GiB)": 123.79,
+      "step": 40065,
+      "train_speed(iter/s)": 1.315908
+    },
+    {
+      "acc": 0.68794422,
+      "epoch": 1.0164890918315577,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.276904633791745e-06,
+      "loss": 1.36186256,
+      "memory(GiB)": 123.79,
+      "step": 40070,
+      "train_speed(iter/s)": 1.315928
+    },
+    {
+      "acc": 0.68777785,
+      "epoch": 1.0166159309994927,
+      "grad_norm": 3.125,
+      "learning_rate": 5.275857613354265e-06,
+      "loss": 1.3443923,
+      "memory(GiB)": 123.79,
+      "step": 40075,
+      "train_speed(iter/s)": 1.315942
+    },
+    {
+      "acc": 0.69741025,
+      "epoch": 1.0167427701674276,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.274810580783335e-06,
+      "loss": 1.36396275,
+      "memory(GiB)": 123.79,
+      "step": 40080,
+      "train_speed(iter/s)": 1.315962
+    },
+    {
+      "acc": 0.70367279,
+      "epoch": 1.0168696093353629,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.2737635361250094e-06,
+      "loss": 1.34869461,
+      "memory(GiB)": 123.79,
+      "step": 40085,
+      "train_speed(iter/s)": 1.315982
+    },
+    {
+      "acc": 0.68433657,
+      "epoch": 1.0169964485032978,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.2727164794253415e-06,
+      "loss": 1.38051195,
+      "memory(GiB)": 123.79,
+      "step": 40090,
+      "train_speed(iter/s)": 1.316002
+    },
+    {
+      "acc": 0.68065457,
+      "epoch": 1.0171232876712328,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.271669410730384e-06,
+      "loss": 1.3798708,
+      "memory(GiB)": 123.79,
+      "step": 40095,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.69727912,
+      "epoch": 1.0172501268391678,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.270622330086194e-06,
+      "loss": 1.41236572,
+      "memory(GiB)": 123.79,
+      "step": 40100,
+      "train_speed(iter/s)": 1.316041
+    },
+    {
+      "acc": 0.67525015,
+      "epoch": 1.017376966007103,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.269575237538827e-06,
+      "loss": 1.37095089,
+      "memory(GiB)": 123.79,
+      "step": 40105,
+      "train_speed(iter/s)": 1.31606
+    },
+    {
+      "acc": 0.67505302,
+      "epoch": 1.017503805175038,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.268528133134335e-06,
+      "loss": 1.43305521,
+      "memory(GiB)": 123.79,
+      "step": 40110,
+      "train_speed(iter/s)": 1.316079
+    },
+    {
+      "acc": 0.6940043,
+      "epoch": 1.017630644342973,
+      "grad_norm": 3.0,
+      "learning_rate": 5.267481016918776e-06,
+      "loss": 1.35057411,
+      "memory(GiB)": 123.79,
+      "step": 40115,
+      "train_speed(iter/s)": 1.316099
+    },
+    {
+      "acc": 0.6908711,
+      "epoch": 1.0177574835109082,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.266433888938212e-06,
+      "loss": 1.38233643,
+      "memory(GiB)": 123.79,
+      "step": 40120,
+      "train_speed(iter/s)": 1.316119
+    },
+    {
+      "acc": 0.69931879,
+      "epoch": 1.0178843226788432,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.265386749238691e-06,
+      "loss": 1.33824959,
+      "memory(GiB)": 123.79,
+      "step": 40125,
+      "train_speed(iter/s)": 1.316139
+    },
+    {
+      "acc": 0.68166623,
+      "epoch": 1.0180111618467782,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.26433959786628e-06,
+      "loss": 1.40574389,
+      "memory(GiB)": 123.79,
+      "step": 40130,
+      "train_speed(iter/s)": 1.316158
+    },
+    {
+      "acc": 0.67866449,
+      "epoch": 1.0181380010147134,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.263292434867031e-06,
+      "loss": 1.39365158,
+      "memory(GiB)": 123.79,
+      "step": 40135,
+      "train_speed(iter/s)": 1.316178
+    },
+    {
+      "acc": 0.70113678,
+      "epoch": 1.0182648401826484,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.262245260287006e-06,
+      "loss": 1.29192085,
+      "memory(GiB)": 123.79,
+      "step": 40140,
+      "train_speed(iter/s)": 1.316198
+    },
+    {
+      "acc": 0.67700663,
+      "epoch": 1.0183916793505834,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.261198074172262e-06,
+      "loss": 1.46667109,
+      "memory(GiB)": 123.79,
+      "step": 40145,
+      "train_speed(iter/s)": 1.316218
+    },
+    {
+      "acc": 0.69193993,
+      "epoch": 1.0185185185185186,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.260150876568862e-06,
+      "loss": 1.39304161,
+      "memory(GiB)": 123.79,
+      "step": 40150,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.68848295,
+      "epoch": 1.0186453576864536,
+      "grad_norm": 2.546875,
+      "learning_rate": 5.259103667522866e-06,
+      "loss": 1.38703394,
+      "memory(GiB)": 123.79,
+      "step": 40155,
+      "train_speed(iter/s)": 1.316259
+    },
+    {
+      "acc": 0.68380418,
+      "epoch": 1.0187721968543886,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.258056447080333e-06,
+      "loss": 1.4285078,
+      "memory(GiB)": 123.79,
+      "step": 40160,
+      "train_speed(iter/s)": 1.316277
+    },
+    {
+      "acc": 0.68513303,
+      "epoch": 1.0188990360223238,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.257009215287325e-06,
+      "loss": 1.43948326,
+      "memory(GiB)": 123.79,
+      "step": 40165,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.68852892,
+      "epoch": 1.0190258751902588,
+      "grad_norm": 3.0,
+      "learning_rate": 5.255961972189905e-06,
+      "loss": 1.39924383,
+      "memory(GiB)": 123.79,
+      "step": 40170,
+      "train_speed(iter/s)": 1.316317
+    },
+    {
+      "acc": 0.67891407,
+      "epoch": 1.0191527143581938,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.254914717834133e-06,
+      "loss": 1.41096897,
+      "memory(GiB)": 123.79,
+      "step": 40175,
+      "train_speed(iter/s)": 1.316336
+    },
+    {
+      "acc": 0.68286781,
+      "epoch": 1.019279553526129,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.253867452266075e-06,
+      "loss": 1.35759811,
+      "memory(GiB)": 123.79,
+      "step": 40180,
+      "train_speed(iter/s)": 1.316357
+    },
+    {
+      "acc": 0.70406952,
+      "epoch": 1.019406392694064,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.252820175531792e-06,
+      "loss": 1.34873247,
+      "memory(GiB)": 123.79,
+      "step": 40185,
+      "train_speed(iter/s)": 1.316377
+    },
+    {
+      "acc": 0.68199358,
+      "epoch": 1.019533231861999,
+      "grad_norm": 4.15625,
+      "learning_rate": 5.25177288767735e-06,
+      "loss": 1.4042366,
+      "memory(GiB)": 123.79,
+      "step": 40190,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.69709024,
+      "epoch": 1.019660071029934,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.250725588748811e-06,
+      "loss": 1.36417751,
+      "memory(GiB)": 123.79,
+      "step": 40195,
+      "train_speed(iter/s)": 1.316409
+    },
+    {
+      "acc": 0.67304745,
+      "epoch": 1.0197869101978692,
+      "grad_norm": 3.0,
+      "learning_rate": 5.249678278792243e-06,
+      "loss": 1.48304892,
+      "memory(GiB)": 123.79,
+      "step": 40200,
+      "train_speed(iter/s)": 1.316427
+    },
+    {
+      "acc": 0.70143876,
+      "epoch": 1.0199137493658041,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.248630957853708e-06,
+      "loss": 1.32979889,
+      "memory(GiB)": 123.79,
+      "step": 40205,
+      "train_speed(iter/s)": 1.316446
+    },
+    {
+      "acc": 0.69010134,
+      "epoch": 1.0200405885337391,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.247583625979276e-06,
+      "loss": 1.33473549,
+      "memory(GiB)": 123.79,
+      "step": 40210,
+      "train_speed(iter/s)": 1.316466
+    },
+    {
+      "acc": 0.67300825,
+      "epoch": 1.0201674277016743,
+      "grad_norm": 3.5,
+      "learning_rate": 5.246536283215007e-06,
+      "loss": 1.39045801,
+      "memory(GiB)": 123.79,
+      "step": 40215,
+      "train_speed(iter/s)": 1.316486
+    },
+    {
+      "acc": 0.68492723,
+      "epoch": 1.0202942668696093,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.245488929606974e-06,
+      "loss": 1.41106968,
+      "memory(GiB)": 123.79,
+      "step": 40220,
+      "train_speed(iter/s)": 1.316505
+    },
+    {
+      "acc": 0.69516511,
+      "epoch": 1.0204211060375443,
+      "grad_norm": 2.5,
+      "learning_rate": 5.244441565201241e-06,
+      "loss": 1.40358448,
+      "memory(GiB)": 123.79,
+      "step": 40225,
+      "train_speed(iter/s)": 1.316524
+    },
+    {
+      "acc": 0.68608451,
+      "epoch": 1.0205479452054795,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.243394190043877e-06,
+      "loss": 1.31609116,
+      "memory(GiB)": 123.79,
+      "step": 40230,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.68521748,
+      "epoch": 1.0206747843734145,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.242346804180949e-06,
+      "loss": 1.37659492,
+      "memory(GiB)": 123.79,
+      "step": 40235,
+      "train_speed(iter/s)": 1.316563
+    },
+    {
+      "acc": 0.68271551,
+      "epoch": 1.0208016235413495,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.241299407658528e-06,
+      "loss": 1.41307926,
+      "memory(GiB)": 123.79,
+      "step": 40240,
+      "train_speed(iter/s)": 1.316582
+    },
+    {
+      "acc": 0.69893932,
+      "epoch": 1.0209284627092847,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.240252000522681e-06,
+      "loss": 1.35526962,
+      "memory(GiB)": 123.79,
+      "step": 40245,
+      "train_speed(iter/s)": 1.3166
+    },
+    {
+      "acc": 0.67578459,
+      "epoch": 1.0210553018772197,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.239204582819479e-06,
+      "loss": 1.44591389,
+      "memory(GiB)": 123.79,
+      "step": 40250,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "acc": 0.67509232,
+      "epoch": 1.0211821410451547,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.238157154594989e-06,
+      "loss": 1.39451046,
+      "memory(GiB)": 123.79,
+      "step": 40255,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.70620489,
+      "epoch": 1.0213089802130897,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.237109715895287e-06,
+      "loss": 1.28904762,
+      "memory(GiB)": 123.79,
+      "step": 40260,
+      "train_speed(iter/s)": 1.316658
+    },
+    {
+      "acc": 0.69000769,
+      "epoch": 1.021435819381025,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.2360622667664385e-06,
+      "loss": 1.37303562,
+      "memory(GiB)": 123.79,
+      "step": 40265,
+      "train_speed(iter/s)": 1.316677
+    },
+    {
+      "acc": 0.69612942,
+      "epoch": 1.02156265854896,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.235014807254521e-06,
+      "loss": 1.35164747,
+      "memory(GiB)": 123.79,
+      "step": 40270,
+      "train_speed(iter/s)": 1.316696
+    },
+    {
+      "acc": 0.67858968,
+      "epoch": 1.0216894977168949,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.233967337405599e-06,
+      "loss": 1.3874752,
+      "memory(GiB)": 123.79,
+      "step": 40275,
+      "train_speed(iter/s)": 1.316716
+    },
+    {
+      "acc": 0.6944725,
+      "epoch": 1.02181633688483,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.232919857265752e-06,
+      "loss": 1.38685827,
+      "memory(GiB)": 123.79,
+      "step": 40280,
+      "train_speed(iter/s)": 1.316735
+    },
+    {
+      "acc": 0.6850749,
+      "epoch": 1.021943176052765,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.231872366881048e-06,
+      "loss": 1.38453255,
+      "memory(GiB)": 123.79,
+      "step": 40285,
+      "train_speed(iter/s)": 1.316754
+    },
+    {
+      "acc": 0.69623747,
+      "epoch": 1.0220700152207,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.230824866297563e-06,
+      "loss": 1.37452049,
+      "memory(GiB)": 123.79,
+      "step": 40290,
+      "train_speed(iter/s)": 1.316773
+    },
+    {
+      "acc": 0.69129076,
+      "epoch": 1.0221968543886353,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.229777355561368e-06,
+      "loss": 1.37406092,
+      "memory(GiB)": 123.79,
+      "step": 40295,
+      "train_speed(iter/s)": 1.316793
+    },
+    {
+      "acc": 0.6940659,
+      "epoch": 1.0223236935565703,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.2287298347185415e-06,
+      "loss": 1.38541889,
+      "memory(GiB)": 123.79,
+      "step": 40300,
+      "train_speed(iter/s)": 1.316812
+    },
+    {
+      "acc": 0.67019887,
+      "epoch": 1.0224505327245053,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.227682303815155e-06,
+      "loss": 1.41684284,
+      "memory(GiB)": 123.79,
+      "step": 40305,
+      "train_speed(iter/s)": 1.316828
+    },
+    {
+      "acc": 0.68294325,
+      "epoch": 1.0225773718924405,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.226634762897284e-06,
+      "loss": 1.42331371,
+      "memory(GiB)": 123.79,
+      "step": 40310,
+      "train_speed(iter/s)": 1.316848
+    },
+    {
+      "acc": 0.68886013,
+      "epoch": 1.0227042110603755,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.225587212011004e-06,
+      "loss": 1.32194557,
+      "memory(GiB)": 123.79,
+      "step": 40315,
+      "train_speed(iter/s)": 1.316866
+    },
+    {
+      "acc": 0.68977103,
+      "epoch": 1.0228310502283104,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.224539651202391e-06,
+      "loss": 1.35695076,
+      "memory(GiB)": 123.79,
+      "step": 40320,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.68489652,
+      "epoch": 1.0229578893962457,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.223492080517523e-06,
+      "loss": 1.40345345,
+      "memory(GiB)": 123.79,
+      "step": 40325,
+      "train_speed(iter/s)": 1.316671
+    },
+    {
+      "acc": 0.67955875,
+      "epoch": 1.0230847285641806,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.2224445000024744e-06,
+      "loss": 1.43767853,
+      "memory(GiB)": 123.79,
+      "step": 40330,
+      "train_speed(iter/s)": 1.316689
+    },
+    {
+      "acc": 0.67792845,
+      "epoch": 1.0232115677321156,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.221396909703322e-06,
+      "loss": 1.3687685,
+      "memory(GiB)": 123.79,
+      "step": 40335,
+      "train_speed(iter/s)": 1.316707
+    },
+    {
+      "acc": 0.69791164,
+      "epoch": 1.0233384069000508,
+      "grad_norm": 3.25,
+      "learning_rate": 5.220349309666148e-06,
+      "loss": 1.31246634,
+      "memory(GiB)": 123.79,
+      "step": 40340,
+      "train_speed(iter/s)": 1.316725
+    },
+    {
+      "acc": 0.6816783,
+      "epoch": 1.0234652460679858,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.2193016999370265e-06,
+      "loss": 1.39326591,
+      "memory(GiB)": 123.79,
+      "step": 40345,
+      "train_speed(iter/s)": 1.316744
+    },
+    {
+      "acc": 0.68662348,
+      "epoch": 1.0235920852359208,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.218254080562038e-06,
+      "loss": 1.35142784,
+      "memory(GiB)": 123.79,
+      "step": 40350,
+      "train_speed(iter/s)": 1.316762
+    },
+    {
+      "acc": 0.71264105,
+      "epoch": 1.0237189244038558,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.2172064515872585e-06,
+      "loss": 1.28696585,
+      "memory(GiB)": 123.79,
+      "step": 40355,
+      "train_speed(iter/s)": 1.31678
+    },
+    {
+      "acc": 0.6835391,
+      "epoch": 1.023845763571791,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.21615881305877e-06,
+      "loss": 1.41985178,
+      "memory(GiB)": 123.79,
+      "step": 40360,
+      "train_speed(iter/s)": 1.316799
+    },
+    {
+      "acc": 0.68578625,
+      "epoch": 1.023972602739726,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.215111165022653e-06,
+      "loss": 1.3620863,
+      "memory(GiB)": 123.79,
+      "step": 40365,
+      "train_speed(iter/s)": 1.316816
+    },
+    {
+      "acc": 0.70411458,
+      "epoch": 1.024099441907661,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.2140635075249856e-06,
+      "loss": 1.30191402,
+      "memory(GiB)": 123.79,
+      "step": 40370,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.69066157,
+      "epoch": 1.0242262810755962,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.213015840611851e-06,
+      "loss": 1.35351276,
+      "memory(GiB)": 123.79,
+      "step": 40375,
+      "train_speed(iter/s)": 1.316851
+    },
+    {
+      "acc": 0.68125448,
+      "epoch": 1.0243531202435312,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.211968164329328e-06,
+      "loss": 1.42443657,
+      "memory(GiB)": 123.79,
+      "step": 40380,
+      "train_speed(iter/s)": 1.316868
+    },
+    {
+      "acc": 0.68918872,
+      "epoch": 1.0244799594114662,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.210920478723497e-06,
+      "loss": 1.40292625,
+      "memory(GiB)": 123.79,
+      "step": 40385,
+      "train_speed(iter/s)": 1.316886
+    },
+    {
+      "acc": 0.67561846,
+      "epoch": 1.0246067985794014,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.209872783840443e-06,
+      "loss": 1.44732447,
+      "memory(GiB)": 123.79,
+      "step": 40390,
+      "train_speed(iter/s)": 1.316904
+    },
+    {
+      "acc": 0.68918648,
+      "epoch": 1.0247336377473364,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.208825079726248e-06,
+      "loss": 1.40436144,
+      "memory(GiB)": 123.79,
+      "step": 40395,
+      "train_speed(iter/s)": 1.316922
+    },
+    {
+      "acc": 0.68909068,
+      "epoch": 1.0248604769152714,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.207777366426992e-06,
+      "loss": 1.38765249,
+      "memory(GiB)": 123.79,
+      "step": 40400,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.688835,
+      "epoch": 1.0249873160832066,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.206729643988759e-06,
+      "loss": 1.37937698,
+      "memory(GiB)": 123.79,
+      "step": 40405,
+      "train_speed(iter/s)": 1.316958
+    },
+    {
+      "acc": 0.69358296,
+      "epoch": 1.0251141552511416,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.205681912457635e-06,
+      "loss": 1.36215973,
+      "memory(GiB)": 123.79,
+      "step": 40410,
+      "train_speed(iter/s)": 1.316975
+    },
+    {
+      "acc": 0.67833252,
+      "epoch": 1.0252409944190766,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.204634171879701e-06,
+      "loss": 1.47580662,
+      "memory(GiB)": 123.79,
+      "step": 40415,
+      "train_speed(iter/s)": 1.316993
+    },
+    {
+      "acc": 0.68991442,
+      "epoch": 1.0253678335870116,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.2035864223010445e-06,
+      "loss": 1.29170341,
+      "memory(GiB)": 123.79,
+      "step": 40420,
+      "train_speed(iter/s)": 1.317012
+    },
+    {
+      "acc": 0.68057804,
+      "epoch": 1.0254946727549468,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.202538663767746e-06,
+      "loss": 1.41810055,
+      "memory(GiB)": 123.79,
+      "step": 40425,
+      "train_speed(iter/s)": 1.31703
+    },
+    {
+      "acc": 0.68494482,
+      "epoch": 1.0256215119228818,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.201490896325895e-06,
+      "loss": 1.42313232,
+      "memory(GiB)": 123.79,
+      "step": 40430,
+      "train_speed(iter/s)": 1.317049
+    },
+    {
+      "acc": 0.69171019,
+      "epoch": 1.0257483510908167,
+      "grad_norm": 3.125,
+      "learning_rate": 5.200443120021572e-06,
+      "loss": 1.37278233,
+      "memory(GiB)": 123.79,
+      "step": 40435,
+      "train_speed(iter/s)": 1.317068
+    },
+    {
+      "acc": 0.69614305,
+      "epoch": 1.025875190258752,
+      "grad_norm": 3.125,
+      "learning_rate": 5.199395334900868e-06,
+      "loss": 1.3317503,
+      "memory(GiB)": 123.79,
+      "step": 40440,
+      "train_speed(iter/s)": 1.317086
+    },
+    {
+      "acc": 0.68483162,
+      "epoch": 1.026002029426687,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.198347541009866e-06,
+      "loss": 1.39999847,
+      "memory(GiB)": 123.79,
+      "step": 40445,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.68538413,
+      "epoch": 1.026128868594622,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.197299738394654e-06,
+      "loss": 1.44867935,
+      "memory(GiB)": 123.79,
+      "step": 40450,
+      "train_speed(iter/s)": 1.317121
+    },
+    {
+      "acc": 0.68297243,
+      "epoch": 1.0262557077625571,
+      "grad_norm": 2.875,
+      "learning_rate": 5.196251927101318e-06,
+      "loss": 1.3995142,
+      "memory(GiB)": 123.79,
+      "step": 40455,
+      "train_speed(iter/s)": 1.317139
+    },
+    {
+      "acc": 0.69814382,
+      "epoch": 1.0263825469304921,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.195204107175946e-06,
+      "loss": 1.2981699,
+      "memory(GiB)": 123.79,
+      "step": 40460,
+      "train_speed(iter/s)": 1.317157
+    },
+    {
+      "acc": 0.68700213,
+      "epoch": 1.0265093860984271,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.194156278664627e-06,
+      "loss": 1.42594128,
+      "memory(GiB)": 123.79,
+      "step": 40465,
+      "train_speed(iter/s)": 1.317175
+    },
+    {
+      "acc": 0.69041924,
+      "epoch": 1.0266362252663623,
+      "grad_norm": 2.75,
+      "learning_rate": 5.1931084416134466e-06,
+      "loss": 1.37417555,
+      "memory(GiB)": 123.79,
+      "step": 40470,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.68360624,
+      "epoch": 1.0267630644342973,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.192060596068496e-06,
+      "loss": 1.41494331,
+      "memory(GiB)": 123.79,
+      "step": 40475,
+      "train_speed(iter/s)": 1.317212
+    },
+    {
+      "acc": 0.68752308,
+      "epoch": 1.0268899036022323,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.191012742075863e-06,
+      "loss": 1.37100258,
+      "memory(GiB)": 123.79,
+      "step": 40480,
+      "train_speed(iter/s)": 1.31723
+    },
+    {
+      "acc": 0.69514041,
+      "epoch": 1.0270167427701675,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.189964879681635e-06,
+      "loss": 1.39465857,
+      "memory(GiB)": 123.79,
+      "step": 40485,
+      "train_speed(iter/s)": 1.317246
+    },
+    {
+      "acc": 0.6853302,
+      "epoch": 1.0271435819381025,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.188917008931905e-06,
+      "loss": 1.40698652,
+      "memory(GiB)": 123.79,
+      "step": 40490,
+      "train_speed(iter/s)": 1.317264
+    },
+    {
+      "acc": 0.68460865,
+      "epoch": 1.0272704211060375,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.18786912987276e-06,
+      "loss": 1.38819914,
+      "memory(GiB)": 123.79,
+      "step": 40495,
+      "train_speed(iter/s)": 1.317282
+    },
+    {
+      "acc": 0.69373927,
+      "epoch": 1.0273972602739727,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.186821242550294e-06,
+      "loss": 1.3613348,
+      "memory(GiB)": 123.79,
+      "step": 40500,
+      "train_speed(iter/s)": 1.3173
+    },
+    {
+      "acc": 0.67677951,
+      "epoch": 1.0275240994419077,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.185773347010594e-06,
+      "loss": 1.42722015,
+      "memory(GiB)": 123.79,
+      "step": 40505,
+      "train_speed(iter/s)": 1.317318
+    },
+    {
+      "acc": 0.6894906,
+      "epoch": 1.0276509386098427,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.184725443299753e-06,
+      "loss": 1.37712049,
+      "memory(GiB)": 123.79,
+      "step": 40510,
+      "train_speed(iter/s)": 1.317336
+    },
+    {
+      "acc": 0.70161519,
+      "epoch": 1.0277777777777777,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.183677531463863e-06,
+      "loss": 1.33234549,
+      "memory(GiB)": 123.79,
+      "step": 40515,
+      "train_speed(iter/s)": 1.317354
+    },
+    {
+      "acc": 0.6810564,
+      "epoch": 1.027904616945713,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.182629611549015e-06,
+      "loss": 1.44944649,
+      "memory(GiB)": 123.79,
+      "step": 40520,
+      "train_speed(iter/s)": 1.31737
+    },
+    {
+      "acc": 0.69201365,
+      "epoch": 1.0280314561136479,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.181581683601301e-06,
+      "loss": 1.33941879,
+      "memory(GiB)": 123.79,
+      "step": 40525,
+      "train_speed(iter/s)": 1.317388
+    },
+    {
+      "acc": 0.69300461,
+      "epoch": 1.0281582952815829,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.1805337476668135e-06,
+      "loss": 1.4962863,
+      "memory(GiB)": 123.79,
+      "step": 40530,
+      "train_speed(iter/s)": 1.317406
+    },
+    {
+      "acc": 0.68893003,
+      "epoch": 1.028285134449518,
+      "grad_norm": 2.390625,
+      "learning_rate": 5.179485803791646e-06,
+      "loss": 1.41852417,
+      "memory(GiB)": 123.79,
+      "step": 40535,
+      "train_speed(iter/s)": 1.317421
+    },
+    {
+      "acc": 0.68807096,
+      "epoch": 1.028411973617453,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.178437852021892e-06,
+      "loss": 1.45615835,
+      "memory(GiB)": 123.79,
+      "step": 40540,
+      "train_speed(iter/s)": 1.31744
+    },
+    {
+      "acc": 0.69886265,
+      "epoch": 1.028538812785388,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.177389892403645e-06,
+      "loss": 1.34360399,
+      "memory(GiB)": 123.79,
+      "step": 40545,
+      "train_speed(iter/s)": 1.317459
+    },
+    {
+      "acc": 0.68709221,
+      "epoch": 1.0286656519533233,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.176341924982997e-06,
+      "loss": 1.37269897,
+      "memory(GiB)": 123.79,
+      "step": 40550,
+      "train_speed(iter/s)": 1.317477
+    },
+    {
+      "acc": 0.69407206,
+      "epoch": 1.0287924911212583,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.1752939498060435e-06,
+      "loss": 1.35582943,
+      "memory(GiB)": 123.79,
+      "step": 40555,
+      "train_speed(iter/s)": 1.317495
+    },
+    {
+      "acc": 0.68052821,
+      "epoch": 1.0289193302891932,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.174245966918883e-06,
+      "loss": 1.48516941,
+      "memory(GiB)": 123.79,
+      "step": 40560,
+      "train_speed(iter/s)": 1.317514
+    },
+    {
+      "acc": 0.67657757,
+      "epoch": 1.0290461694571285,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.173197976367603e-06,
+      "loss": 1.39728251,
+      "memory(GiB)": 123.79,
+      "step": 40565,
+      "train_speed(iter/s)": 1.317532
+    },
+    {
+      "acc": 0.69162488,
+      "epoch": 1.0291730086250634,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.1721499781983055e-06,
+      "loss": 1.3439827,
+      "memory(GiB)": 123.79,
+      "step": 40570,
+      "train_speed(iter/s)": 1.31755
+    },
+    {
+      "acc": 0.67087202,
+      "epoch": 1.0292998477929984,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.171101972457081e-06,
+      "loss": 1.38827972,
+      "memory(GiB)": 123.79,
+      "step": 40575,
+      "train_speed(iter/s)": 1.317567
+    },
+    {
+      "acc": 0.68365278,
+      "epoch": 1.0294266869609334,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.170053959190029e-06,
+      "loss": 1.3727354,
+      "memory(GiB)": 123.79,
+      "step": 40580,
+      "train_speed(iter/s)": 1.317584
+    },
+    {
+      "acc": 0.6889122,
+      "epoch": 1.0295535261288686,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.169005938443245e-06,
+      "loss": 1.40131073,
+      "memory(GiB)": 123.79,
+      "step": 40585,
+      "train_speed(iter/s)": 1.317602
+    },
+    {
+      "acc": 0.68856101,
+      "epoch": 1.0296803652968036,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.1679579102628245e-06,
+      "loss": 1.39372931,
+      "memory(GiB)": 123.79,
+      "step": 40590,
+      "train_speed(iter/s)": 1.317619
+    },
+    {
+      "acc": 0.68267183,
+      "epoch": 1.0298072044647386,
+      "grad_norm": 2.875,
+      "learning_rate": 5.166909874694866e-06,
+      "loss": 1.38426294,
+      "memory(GiB)": 123.79,
+      "step": 40595,
+      "train_speed(iter/s)": 1.317637
+    },
+    {
+      "acc": 0.67482605,
+      "epoch": 1.0299340436326738,
+      "grad_norm": 3.875,
+      "learning_rate": 5.165861831785465e-06,
+      "loss": 1.46655979,
+      "memory(GiB)": 123.79,
+      "step": 40600,
+      "train_speed(iter/s)": 1.317655
+    },
+    {
+      "acc": 0.67853193,
+      "epoch": 1.0300608828006088,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.164813781580721e-06,
+      "loss": 1.40338345,
+      "memory(GiB)": 123.79,
+      "step": 40605,
+      "train_speed(iter/s)": 1.317672
+    },
+    {
+      "acc": 0.68242054,
+      "epoch": 1.0301877219685438,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.16376572412673e-06,
+      "loss": 1.3710495,
+      "memory(GiB)": 123.79,
+      "step": 40610,
+      "train_speed(iter/s)": 1.31769
+    },
+    {
+      "acc": 0.69804387,
+      "epoch": 1.030314561136479,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.162717659469593e-06,
+      "loss": 1.41035624,
+      "memory(GiB)": 123.79,
+      "step": 40615,
+      "train_speed(iter/s)": 1.317705
+    },
+    {
+      "acc": 0.69950542,
+      "epoch": 1.030441400304414,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.161669587655406e-06,
+      "loss": 1.31480389,
+      "memory(GiB)": 123.79,
+      "step": 40620,
+      "train_speed(iter/s)": 1.317723
+    },
+    {
+      "acc": 0.68728371,
+      "epoch": 1.030568239472349,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.160621508730267e-06,
+      "loss": 1.35415707,
+      "memory(GiB)": 123.79,
+      "step": 40625,
+      "train_speed(iter/s)": 1.317741
+    },
+    {
+      "acc": 0.67785554,
+      "epoch": 1.0306950786402842,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.15957342274028e-06,
+      "loss": 1.41602974,
+      "memory(GiB)": 123.79,
+      "step": 40630,
+      "train_speed(iter/s)": 1.317759
+    },
+    {
+      "acc": 0.6774106,
+      "epoch": 1.0308219178082192,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.158525329731539e-06,
+      "loss": 1.40699902,
+      "memory(GiB)": 123.79,
+      "step": 40635,
+      "train_speed(iter/s)": 1.317778
+    },
+    {
+      "acc": 0.68874273,
+      "epoch": 1.0309487569761542,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.157477229750149e-06,
+      "loss": 1.47040281,
+      "memory(GiB)": 123.79,
+      "step": 40640,
+      "train_speed(iter/s)": 1.317796
+    },
+    {
+      "acc": 0.69379501,
+      "epoch": 1.0310755961440894,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.156429122842204e-06,
+      "loss": 1.3906661,
+      "memory(GiB)": 123.79,
+      "step": 40645,
+      "train_speed(iter/s)": 1.317814
+    },
+    {
+      "acc": 0.68916221,
+      "epoch": 1.0312024353120244,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.15538100905381e-06,
+      "loss": 1.36881752,
+      "memory(GiB)": 123.79,
+      "step": 40650,
+      "train_speed(iter/s)": 1.317831
+    },
+    {
+      "acc": 0.68637762,
+      "epoch": 1.0313292744799594,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.154332888431064e-06,
+      "loss": 1.38661289,
+      "memory(GiB)": 123.79,
+      "step": 40655,
+      "train_speed(iter/s)": 1.317849
+    },
+    {
+      "acc": 0.68040304,
+      "epoch": 1.0314561136478946,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.15328476102007e-06,
+      "loss": 1.39472466,
+      "memory(GiB)": 123.79,
+      "step": 40660,
+      "train_speed(iter/s)": 1.317867
+    },
+    {
+      "acc": 0.67448826,
+      "epoch": 1.0315829528158296,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.1522366268669264e-06,
+      "loss": 1.40569878,
+      "memory(GiB)": 123.79,
+      "step": 40665,
+      "train_speed(iter/s)": 1.317886
+    },
+    {
+      "acc": 0.68636842,
+      "epoch": 1.0317097919837646,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.1511884860177376e-06,
+      "loss": 1.38132057,
+      "memory(GiB)": 123.79,
+      "step": 40670,
+      "train_speed(iter/s)": 1.317904
+    },
+    {
+      "acc": 0.69226665,
+      "epoch": 1.0318366311516995,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.150140338518603e-06,
+      "loss": 1.39013815,
+      "memory(GiB)": 123.79,
+      "step": 40675,
+      "train_speed(iter/s)": 1.317922
+    },
+    {
+      "acc": 0.70380583,
+      "epoch": 1.0319634703196348,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.149092184415627e-06,
+      "loss": 1.32541504,
+      "memory(GiB)": 123.79,
+      "step": 40680,
+      "train_speed(iter/s)": 1.31794
+    },
+    {
+      "acc": 0.7024611,
+      "epoch": 1.0320903094875697,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.148044023754911e-06,
+      "loss": 1.36336765,
+      "memory(GiB)": 123.79,
+      "step": 40685,
+      "train_speed(iter/s)": 1.317958
+    },
+    {
+      "acc": 0.67075109,
+      "epoch": 1.0322171486555047,
+      "grad_norm": 2.625,
+      "learning_rate": 5.146995856582557e-06,
+      "loss": 1.41202421,
+      "memory(GiB)": 123.79,
+      "step": 40690,
+      "train_speed(iter/s)": 1.317975
+    },
+    {
+      "acc": 0.68841105,
+      "epoch": 1.03234398782344,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.14594768294467e-06,
+      "loss": 1.44015121,
+      "memory(GiB)": 123.79,
+      "step": 40695,
+      "train_speed(iter/s)": 1.317994
+    },
+    {
+      "acc": 0.68634377,
+      "epoch": 1.032470826991375,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.1448995028873515e-06,
+      "loss": 1.40880318,
+      "memory(GiB)": 123.79,
+      "step": 40700,
+      "train_speed(iter/s)": 1.318012
+    },
+    {
+      "acc": 0.68805761,
+      "epoch": 1.03259766615931,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.143851316456706e-06,
+      "loss": 1.34544439,
+      "memory(GiB)": 123.79,
+      "step": 40705,
+      "train_speed(iter/s)": 1.318026
+    },
+    {
+      "acc": 0.68272223,
+      "epoch": 1.0327245053272451,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.142803123698838e-06,
+      "loss": 1.42984333,
+      "memory(GiB)": 123.79,
+      "step": 40710,
+      "train_speed(iter/s)": 1.318044
+    },
+    {
+      "acc": 0.6941834,
+      "epoch": 1.0328513444951801,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.14175492465985e-06,
+      "loss": 1.31621466,
+      "memory(GiB)": 123.79,
+      "step": 40715,
+      "train_speed(iter/s)": 1.318062
+    },
+    {
+      "acc": 0.69375582,
+      "epoch": 1.0329781836631151,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.14070671938585e-06,
+      "loss": 1.41126699,
+      "memory(GiB)": 123.79,
+      "step": 40720,
+      "train_speed(iter/s)": 1.31808
+    },
+    {
+      "acc": 0.68689604,
+      "epoch": 1.0331050228310503,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.139658507922937e-06,
+      "loss": 1.38346128,
+      "memory(GiB)": 123.79,
+      "step": 40725,
+      "train_speed(iter/s)": 1.318098
+    },
+    {
+      "acc": 0.67836461,
+      "epoch": 1.0332318619989853,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.138610290317221e-06,
+      "loss": 1.42691431,
+      "memory(GiB)": 123.79,
+      "step": 40730,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.68429041,
+      "epoch": 1.0333587011669203,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.137562066614805e-06,
+      "loss": 1.39947634,
+      "memory(GiB)": 123.79,
+      "step": 40735,
+      "train_speed(iter/s)": 1.318135
+    },
+    {
+      "acc": 0.69670172,
+      "epoch": 1.0334855403348553,
+      "grad_norm": 4.625,
+      "learning_rate": 5.136513836861795e-06,
+      "loss": 1.32590504,
+      "memory(GiB)": 123.79,
+      "step": 40740,
+      "train_speed(iter/s)": 1.318154
+    },
+    {
+      "acc": 0.69553962,
+      "epoch": 1.0336123795027905,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.135465601104298e-06,
+      "loss": 1.44545174,
+      "memory(GiB)": 123.79,
+      "step": 40745,
+      "train_speed(iter/s)": 1.318174
+    },
+    {
+      "acc": 0.71149435,
+      "epoch": 1.0337392186707255,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.134417359388418e-06,
+      "loss": 1.33304024,
+      "memory(GiB)": 123.79,
+      "step": 40750,
+      "train_speed(iter/s)": 1.318193
+    },
+    {
+      "acc": 0.67674351,
+      "epoch": 1.0338660578386605,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.133369111760264e-06,
+      "loss": 1.39510422,
+      "memory(GiB)": 123.79,
+      "step": 40755,
+      "train_speed(iter/s)": 1.31821
+    },
+    {
+      "acc": 0.69851761,
+      "epoch": 1.0339928970065957,
+      "grad_norm": 4.40625,
+      "learning_rate": 5.132320858265939e-06,
+      "loss": 1.36258888,
+      "memory(GiB)": 123.79,
+      "step": 40760,
+      "train_speed(iter/s)": 1.318228
+    },
+    {
+      "acc": 0.69356499,
+      "epoch": 1.0341197361745307,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.131272598951554e-06,
+      "loss": 1.38666677,
+      "memory(GiB)": 123.79,
+      "step": 40765,
+      "train_speed(iter/s)": 1.318247
+    },
+    {
+      "acc": 0.69580054,
+      "epoch": 1.0342465753424657,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.130224333863212e-06,
+      "loss": 1.29245453,
+      "memory(GiB)": 123.79,
+      "step": 40770,
+      "train_speed(iter/s)": 1.318264
+    },
+    {
+      "acc": 0.68662205,
+      "epoch": 1.0343734145104009,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.129176063047022e-06,
+      "loss": 1.39230852,
+      "memory(GiB)": 123.79,
+      "step": 40775,
+      "train_speed(iter/s)": 1.318282
+    },
+    {
+      "acc": 0.67239127,
+      "epoch": 1.0345002536783359,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.128127786549094e-06,
+      "loss": 1.45294437,
+      "memory(GiB)": 123.79,
+      "step": 40780,
+      "train_speed(iter/s)": 1.318301
+    },
+    {
+      "acc": 0.68501987,
+      "epoch": 1.0346270928462709,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.127079504415532e-06,
+      "loss": 1.40775089,
+      "memory(GiB)": 123.79,
+      "step": 40785,
+      "train_speed(iter/s)": 1.318319
+    },
+    {
+      "acc": 0.70624104,
+      "epoch": 1.034753932014206,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.126031216692449e-06,
+      "loss": 1.31061049,
+      "memory(GiB)": 123.79,
+      "step": 40790,
+      "train_speed(iter/s)": 1.318336
+    },
+    {
+      "acc": 0.6872488,
+      "epoch": 1.034880771182141,
+      "grad_norm": 2.546875,
+      "learning_rate": 5.124982923425947e-06,
+      "loss": 1.38824711,
+      "memory(GiB)": 123.79,
+      "step": 40795,
+      "train_speed(iter/s)": 1.318353
+    },
+    {
+      "acc": 0.68361893,
+      "epoch": 1.035007610350076,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.123934624662139e-06,
+      "loss": 1.39306669,
+      "memory(GiB)": 123.79,
+      "step": 40800,
+      "train_speed(iter/s)": 1.318371
+    },
+    {
+      "acc": 0.66937404,
+      "epoch": 1.0351344495180113,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.1228863204471335e-06,
+      "loss": 1.43368721,
+      "memory(GiB)": 123.79,
+      "step": 40805,
+      "train_speed(iter/s)": 1.31839
+    },
+    {
+      "acc": 0.69007034,
+      "epoch": 1.0352612886859462,
+      "grad_norm": 2.4375,
+      "learning_rate": 5.121838010827039e-06,
+      "loss": 1.36311378,
+      "memory(GiB)": 123.79,
+      "step": 40810,
+      "train_speed(iter/s)": 1.318409
+    },
+    {
+      "acc": 0.68344841,
+      "epoch": 1.0353881278538812,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.120789695847965e-06,
+      "loss": 1.39604006,
+      "memory(GiB)": 123.79,
+      "step": 40815,
+      "train_speed(iter/s)": 1.318428
+    },
+    {
+      "acc": 0.68343234,
+      "epoch": 1.0355149670218164,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.119741375556021e-06,
+      "loss": 1.39961557,
+      "memory(GiB)": 123.79,
+      "step": 40820,
+      "train_speed(iter/s)": 1.318446
+    },
+    {
+      "acc": 0.71068249,
+      "epoch": 1.0356418061897514,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.118693049997316e-06,
+      "loss": 1.31492081,
+      "memory(GiB)": 123.79,
+      "step": 40825,
+      "train_speed(iter/s)": 1.318465
+    },
+    {
+      "acc": 0.68205132,
+      "epoch": 1.0357686453576864,
+      "grad_norm": 3.375,
+      "learning_rate": 5.117644719217961e-06,
+      "loss": 1.41586895,
+      "memory(GiB)": 123.79,
+      "step": 40830,
+      "train_speed(iter/s)": 1.318484
+    },
+    {
+      "acc": 0.67894812,
+      "epoch": 1.0358954845256214,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.116596383264066e-06,
+      "loss": 1.38940344,
+      "memory(GiB)": 123.79,
+      "step": 40835,
+      "train_speed(iter/s)": 1.318503
+    },
+    {
+      "acc": 0.6725081,
+      "epoch": 1.0360223236935566,
+      "grad_norm": 4.84375,
+      "learning_rate": 5.115548042181742e-06,
+      "loss": 1.44809284,
+      "memory(GiB)": 123.79,
+      "step": 40840,
+      "train_speed(iter/s)": 1.318521
+    },
+    {
+      "acc": 0.70204029,
+      "epoch": 1.0361491628614916,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.114499696017098e-06,
+      "loss": 1.29127388,
+      "memory(GiB)": 123.79,
+      "step": 40845,
+      "train_speed(iter/s)": 1.318282
+    },
+    {
+      "acc": 0.67786207,
+      "epoch": 1.0362760020294266,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.1134513448162475e-06,
+      "loss": 1.44865007,
+      "memory(GiB)": 123.79,
+      "step": 40850,
+      "train_speed(iter/s)": 1.318301
+    },
+    {
+      "acc": 0.66896205,
+      "epoch": 1.0364028411973618,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.112402988625299e-06,
+      "loss": 1.42542191,
+      "memory(GiB)": 123.79,
+      "step": 40855,
+      "train_speed(iter/s)": 1.31832
+    },
+    {
+      "acc": 0.69915519,
+      "epoch": 1.0365296803652968,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.111354627490367e-06,
+      "loss": 1.35427399,
+      "memory(GiB)": 123.79,
+      "step": 40860,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "acc": 0.67618637,
+      "epoch": 1.0366565195332318,
+      "grad_norm": 3.125,
+      "learning_rate": 5.110306261457559e-06,
+      "loss": 1.43267193,
+      "memory(GiB)": 123.79,
+      "step": 40865,
+      "train_speed(iter/s)": 1.318356
+    },
+    {
+      "acc": 0.67690315,
+      "epoch": 1.036783358701167,
+      "grad_norm": 2.875,
+      "learning_rate": 5.109257890572991e-06,
+      "loss": 1.46325512,
+      "memory(GiB)": 123.79,
+      "step": 40870,
+      "train_speed(iter/s)": 1.318373
+    },
+    {
+      "acc": 0.69372435,
+      "epoch": 1.036910197869102,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.108209514882772e-06,
+      "loss": 1.36774721,
+      "memory(GiB)": 123.79,
+      "step": 40875,
+      "train_speed(iter/s)": 1.318389
+    },
+    {
+      "acc": 0.67417397,
+      "epoch": 1.037037037037037,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.107161134433017e-06,
+      "loss": 1.42189865,
+      "memory(GiB)": 123.79,
+      "step": 40880,
+      "train_speed(iter/s)": 1.318407
+    },
+    {
+      "acc": 0.68491769,
+      "epoch": 1.0371638762049722,
+      "grad_norm": 3.875,
+      "learning_rate": 5.106112749269835e-06,
+      "loss": 1.35195913,
+      "memory(GiB)": 123.79,
+      "step": 40885,
+      "train_speed(iter/s)": 1.318421
+    },
+    {
+      "acc": 0.68767333,
+      "epoch": 1.0372907153729072,
+      "grad_norm": 5.5,
+      "learning_rate": 5.105064359439341e-06,
+      "loss": 1.43937397,
+      "memory(GiB)": 123.79,
+      "step": 40890,
+      "train_speed(iter/s)": 1.318439
+    },
+    {
+      "acc": 0.69224653,
+      "epoch": 1.0374175545408422,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.1040159649876485e-06,
+      "loss": 1.30783882,
+      "memory(GiB)": 123.79,
+      "step": 40895,
+      "train_speed(iter/s)": 1.318457
+    },
+    {
+      "acc": 0.68139973,
+      "epoch": 1.0375443937087772,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.102967565960868e-06,
+      "loss": 1.42717123,
+      "memory(GiB)": 123.79,
+      "step": 40900,
+      "train_speed(iter/s)": 1.318474
+    },
+    {
+      "acc": 0.68287773,
+      "epoch": 1.0376712328767124,
+      "grad_norm": 3.125,
+      "learning_rate": 5.101919162405116e-06,
+      "loss": 1.40660343,
+      "memory(GiB)": 123.79,
+      "step": 40905,
+      "train_speed(iter/s)": 1.31849
+    },
+    {
+      "acc": 0.69685988,
+      "epoch": 1.0377980720446474,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.100870754366503e-06,
+      "loss": 1.35973988,
+      "memory(GiB)": 123.79,
+      "step": 40910,
+      "train_speed(iter/s)": 1.31851
+    },
+    {
+      "acc": 0.70345378,
+      "epoch": 1.0379249112125823,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.099822341891144e-06,
+      "loss": 1.28684378,
+      "memory(GiB)": 123.79,
+      "step": 40915,
+      "train_speed(iter/s)": 1.318528
+    },
+    {
+      "acc": 0.67872577,
+      "epoch": 1.0380517503805176,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.098773925025152e-06,
+      "loss": 1.41483927,
+      "memory(GiB)": 123.79,
+      "step": 40920,
+      "train_speed(iter/s)": 1.318547
+    },
+    {
+      "acc": 0.67099295,
+      "epoch": 1.0381785895484525,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.097725503814643e-06,
+      "loss": 1.46670094,
+      "memory(GiB)": 123.79,
+      "step": 40925,
+      "train_speed(iter/s)": 1.318564
+    },
+    {
+      "acc": 0.69356556,
+      "epoch": 1.0383054287163875,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.09667707830573e-06,
+      "loss": 1.33853016,
+      "memory(GiB)": 123.79,
+      "step": 40930,
+      "train_speed(iter/s)": 1.318583
+    },
+    {
+      "acc": 0.70566168,
+      "epoch": 1.0384322678843227,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.095628648544526e-06,
+      "loss": 1.28496094,
+      "memory(GiB)": 123.79,
+      "step": 40935,
+      "train_speed(iter/s)": 1.3186
+    },
+    {
+      "acc": 0.69076858,
+      "epoch": 1.0385591070522577,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.0945802145771495e-06,
+      "loss": 1.35043468,
+      "memory(GiB)": 123.79,
+      "step": 40940,
+      "train_speed(iter/s)": 1.318619
+    },
+    {
+      "acc": 0.69211092,
+      "epoch": 1.0386859462201927,
+      "grad_norm": 2.875,
+      "learning_rate": 5.093531776449711e-06,
+      "loss": 1.33352776,
+      "memory(GiB)": 123.79,
+      "step": 40945,
+      "train_speed(iter/s)": 1.318636
+    },
+    {
+      "acc": 0.68802242,
+      "epoch": 1.038812785388128,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.092483334208327e-06,
+      "loss": 1.39372091,
+      "memory(GiB)": 123.79,
+      "step": 40950,
+      "train_speed(iter/s)": 1.318655
+    },
+    {
+      "acc": 0.68201504,
+      "epoch": 1.038939624556063,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.091434887899114e-06,
+      "loss": 1.42672529,
+      "memory(GiB)": 123.79,
+      "step": 40955,
+      "train_speed(iter/s)": 1.318673
+    },
+    {
+      "acc": 0.69027066,
+      "epoch": 1.039066463723998,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.0903864375681866e-06,
+      "loss": 1.37977734,
+      "memory(GiB)": 123.79,
+      "step": 40960,
+      "train_speed(iter/s)": 1.318691
+    },
+    {
+      "acc": 0.69407339,
+      "epoch": 1.0391933028919331,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.0893379832616594e-06,
+      "loss": 1.37721004,
+      "memory(GiB)": 123.79,
+      "step": 40965,
+      "train_speed(iter/s)": 1.318708
+    },
+    {
+      "acc": 0.69553928,
+      "epoch": 1.0393201420598681,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.08828952502565e-06,
+      "loss": 1.38429737,
+      "memory(GiB)": 123.79,
+      "step": 40970,
+      "train_speed(iter/s)": 1.318493
+    },
+    {
+      "acc": 0.69409227,
+      "epoch": 1.039446981227803,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.087241062906272e-06,
+      "loss": 1.37967291,
+      "memory(GiB)": 123.79,
+      "step": 40975,
+      "train_speed(iter/s)": 1.318511
+    },
+    {
+      "acc": 0.67463388,
+      "epoch": 1.0395738203957383,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.086192596949643e-06,
+      "loss": 1.40699482,
+      "memory(GiB)": 123.79,
+      "step": 40980,
+      "train_speed(iter/s)": 1.318529
+    },
+    {
+      "acc": 0.69450674,
+      "epoch": 1.0397006595636733,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.085144127201879e-06,
+      "loss": 1.34971609,
+      "memory(GiB)": 123.79,
+      "step": 40985,
+      "train_speed(iter/s)": 1.318548
+    },
+    {
+      "acc": 0.69385929,
+      "epoch": 1.0398274987316083,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.084095653709096e-06,
+      "loss": 1.28998852,
+      "memory(GiB)": 123.79,
+      "step": 40990,
+      "train_speed(iter/s)": 1.318566
+    },
+    {
+      "acc": 0.70451484,
+      "epoch": 1.0399543378995433,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.0830471765174096e-06,
+      "loss": 1.37070751,
+      "memory(GiB)": 123.79,
+      "step": 40995,
+      "train_speed(iter/s)": 1.318583
+    },
+    {
+      "acc": 0.67780714,
+      "epoch": 1.0400811770674785,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.0819986956729395e-06,
+      "loss": 1.41744251,
+      "memory(GiB)": 123.79,
+      "step": 41000,
+      "train_speed(iter/s)": 1.318602
+    },
+    {
+      "epoch": 1.0400811770674785,
+      "eval_acc": 0.6749363588712288,
+      "eval_loss": 1.3566526174545288,
+      "eval_runtime": 69.7378,
+      "eval_samples_per_second": 91.342,
+      "eval_steps_per_second": 22.843,
+      "step": 41000
+    },
+    {
+      "acc": 0.70514808,
+      "epoch": 1.0402080162354135,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.080950211221799e-06,
+      "loss": 1.34332647,
+      "memory(GiB)": 123.79,
+      "step": 41005,
+      "train_speed(iter/s)": 1.31514
+    },
+    {
+      "acc": 0.68024516,
+      "epoch": 1.0403348554033485,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.079901723210109e-06,
+      "loss": 1.36504402,
+      "memory(GiB)": 123.79,
+      "step": 41010,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.68408556,
+      "epoch": 1.0404616945712837,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.078853231683981e-06,
+      "loss": 1.39124784,
+      "memory(GiB)": 123.79,
+      "step": 41015,
+      "train_speed(iter/s)": 1.315176
+    },
+    {
+      "acc": 0.68636293,
+      "epoch": 1.0405885337392187,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.077804736689539e-06,
+      "loss": 1.41176424,
+      "memory(GiB)": 123.79,
+      "step": 41020,
+      "train_speed(iter/s)": 1.315194
+    },
+    {
+      "acc": 0.69627929,
+      "epoch": 1.0407153729071537,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.0767562382728955e-06,
+      "loss": 1.33701925,
+      "memory(GiB)": 123.79,
+      "step": 41025,
+      "train_speed(iter/s)": 1.315213
+    },
+    {
+      "acc": 0.67673292,
+      "epoch": 1.0408422120750889,
+      "grad_norm": 3.703125,
+      "learning_rate": 5.075707736480171e-06,
+      "loss": 1.48249092,
+      "memory(GiB)": 123.79,
+      "step": 41030,
+      "train_speed(iter/s)": 1.315232
+    },
+    {
+      "acc": 0.68493013,
+      "epoch": 1.0409690512430239,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.074659231357482e-06,
+      "loss": 1.40126171,
+      "memory(GiB)": 123.79,
+      "step": 41035,
+      "train_speed(iter/s)": 1.31525
+    },
+    {
+      "acc": 0.68880081,
+      "epoch": 1.0410958904109588,
+      "grad_norm": 2.546875,
+      "learning_rate": 5.073610722950947e-06,
+      "loss": 1.38948994,
+      "memory(GiB)": 123.79,
+      "step": 41040,
+      "train_speed(iter/s)": 1.315269
+    },
+    {
+      "acc": 0.68080816,
+      "epoch": 1.041222729578894,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.072562211306683e-06,
+      "loss": 1.38444405,
+      "memory(GiB)": 123.79,
+      "step": 41045,
+      "train_speed(iter/s)": 1.315287
+    },
+    {
+      "acc": 0.66926022,
+      "epoch": 1.041349568746829,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.071513696470809e-06,
+      "loss": 1.45516319,
+      "memory(GiB)": 123.79,
+      "step": 41050,
+      "train_speed(iter/s)": 1.315305
+    },
+    {
+      "acc": 0.69081736,
+      "epoch": 1.041476407914764,
+      "grad_norm": 3.890625,
+      "learning_rate": 5.070465178489443e-06,
+      "loss": 1.37311935,
+      "memory(GiB)": 123.79,
+      "step": 41055,
+      "train_speed(iter/s)": 1.315324
+    },
+    {
+      "acc": 0.69352598,
+      "epoch": 1.041603247082699,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.069416657408704e-06,
+      "loss": 1.35811749,
+      "memory(GiB)": 123.79,
+      "step": 41060,
+      "train_speed(iter/s)": 1.315343
+    },
+    {
+      "acc": 0.69104557,
+      "epoch": 1.0417300862506342,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.0683681332747105e-06,
+      "loss": 1.43416538,
+      "memory(GiB)": 123.79,
+      "step": 41065,
+      "train_speed(iter/s)": 1.315353
+    },
+    {
+      "acc": 0.69197683,
+      "epoch": 1.0418569254185692,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.067319606133583e-06,
+      "loss": 1.41546049,
+      "memory(GiB)": 123.79,
+      "step": 41070,
+      "train_speed(iter/s)": 1.315372
+    },
+    {
+      "acc": 0.70572753,
+      "epoch": 1.0419837645865042,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.066271076031436e-06,
+      "loss": 1.34759636,
+      "memory(GiB)": 123.79,
+      "step": 41075,
+      "train_speed(iter/s)": 1.31539
+    },
+    {
+      "acc": 0.68357038,
+      "epoch": 1.0421106037544394,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.065222543014394e-06,
+      "loss": 1.396348,
+      "memory(GiB)": 123.79,
+      "step": 41080,
+      "train_speed(iter/s)": 1.315408
+    },
+    {
+      "acc": 0.68423119,
+      "epoch": 1.0422374429223744,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.06417400712857e-06,
+      "loss": 1.42039175,
+      "memory(GiB)": 123.79,
+      "step": 41085,
+      "train_speed(iter/s)": 1.315427
+    },
+    {
+      "acc": 0.69369497,
+      "epoch": 1.0423642820903094,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.0631254684200906e-06,
+      "loss": 1.3267849,
+      "memory(GiB)": 123.79,
+      "step": 41090,
+      "train_speed(iter/s)": 1.315446
+    },
+    {
+      "acc": 0.69978905,
+      "epoch": 1.0424911212582446,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.062076926935068e-06,
+      "loss": 1.36062851,
+      "memory(GiB)": 123.79,
+      "step": 41095,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.69240518,
+      "epoch": 1.0426179604261796,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.061028382719626e-06,
+      "loss": 1.33238955,
+      "memory(GiB)": 123.79,
+      "step": 41100,
+      "train_speed(iter/s)": 1.315483
+    },
+    {
+      "acc": 0.69227304,
+      "epoch": 1.0427447995941146,
+      "grad_norm": 3.375,
+      "learning_rate": 5.0599798358198835e-06,
+      "loss": 1.3462944,
+      "memory(GiB)": 123.79,
+      "step": 41105,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.68592491,
+      "epoch": 1.0428716387620498,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.0589312862819605e-06,
+      "loss": 1.38886061,
+      "memory(GiB)": 123.79,
+      "step": 41110,
+      "train_speed(iter/s)": 1.315518
+    },
+    {
+      "acc": 0.69274826,
+      "epoch": 1.0429984779299848,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.057882734151977e-06,
+      "loss": 1.34329224,
+      "memory(GiB)": 123.79,
+      "step": 41115,
+      "train_speed(iter/s)": 1.315535
+    },
+    {
+      "acc": 0.69203649,
+      "epoch": 1.0431253170979198,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.05683417947605e-06,
+      "loss": 1.39255743,
+      "memory(GiB)": 123.79,
+      "step": 41120,
+      "train_speed(iter/s)": 1.315553
+    },
+    {
+      "acc": 0.68146524,
+      "epoch": 1.043252156265855,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.055785622300303e-06,
+      "loss": 1.39695568,
+      "memory(GiB)": 123.79,
+      "step": 41125,
+      "train_speed(iter/s)": 1.315571
+    },
+    {
+      "acc": 0.68223224,
+      "epoch": 1.04337899543379,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.054737062670857e-06,
+      "loss": 1.41858606,
+      "memory(GiB)": 123.79,
+      "step": 41130,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.68053689,
+      "epoch": 1.043505834601725,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.053688500633828e-06,
+      "loss": 1.36643009,
+      "memory(GiB)": 123.79,
+      "step": 41135,
+      "train_speed(iter/s)": 1.315606
+    },
+    {
+      "acc": 0.69239016,
+      "epoch": 1.0436326737696602,
+      "grad_norm": 3.5,
+      "learning_rate": 5.052639936235341e-06,
+      "loss": 1.35936413,
+      "memory(GiB)": 123.79,
+      "step": 41140,
+      "train_speed(iter/s)": 1.315624
+    },
+    {
+      "acc": 0.68837366,
+      "epoch": 1.0437595129375952,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.051591369521513e-06,
+      "loss": 1.40910912,
+      "memory(GiB)": 123.79,
+      "step": 41145,
+      "train_speed(iter/s)": 1.315642
+    },
+    {
+      "acc": 0.68409595,
+      "epoch": 1.0438863521055302,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.050542800538469e-06,
+      "loss": 1.38424187,
+      "memory(GiB)": 123.79,
+      "step": 41150,
+      "train_speed(iter/s)": 1.31566
+    },
+    {
+      "acc": 0.68805008,
+      "epoch": 1.0440131912734651,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.049494229332324e-06,
+      "loss": 1.44757624,
+      "memory(GiB)": 123.79,
+      "step": 41155,
+      "train_speed(iter/s)": 1.315678
+    },
+    {
+      "acc": 0.68969579,
+      "epoch": 1.0441400304414004,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.048445655949204e-06,
+      "loss": 1.40881004,
+      "memory(GiB)": 123.79,
+      "step": 41160,
+      "train_speed(iter/s)": 1.315696
+    },
+    {
+      "acc": 0.6769968,
+      "epoch": 1.0442668696093353,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.047397080435225e-06,
+      "loss": 1.49435349,
+      "memory(GiB)": 123.79,
+      "step": 41165,
+      "train_speed(iter/s)": 1.315714
+    },
+    {
+      "acc": 0.68909845,
+      "epoch": 1.0443937087772703,
+      "grad_norm": 3.0,
+      "learning_rate": 5.046348502836512e-06,
+      "loss": 1.44308472,
+      "memory(GiB)": 123.79,
+      "step": 41170,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.67928925,
+      "epoch": 1.0445205479452055,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.045299923199186e-06,
+      "loss": 1.4107501,
+      "memory(GiB)": 123.79,
+      "step": 41175,
+      "train_speed(iter/s)": 1.315749
+    },
+    {
+      "acc": 0.68980737,
+      "epoch": 1.0446473871131405,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.044251341569366e-06,
+      "loss": 1.34787931,
+      "memory(GiB)": 123.79,
+      "step": 41180,
+      "train_speed(iter/s)": 1.315766
+    },
+    {
+      "acc": 0.66879759,
+      "epoch": 1.0447742262810755,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.043202757993175e-06,
+      "loss": 1.49200668,
+      "memory(GiB)": 123.79,
+      "step": 41185,
+      "train_speed(iter/s)": 1.315783
+    },
+    {
+      "acc": 0.67335777,
+      "epoch": 1.0449010654490107,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.042154172516734e-06,
+      "loss": 1.4370348,
+      "memory(GiB)": 123.79,
+      "step": 41190,
+      "train_speed(iter/s)": 1.315801
+    },
+    {
+      "acc": 0.68831558,
+      "epoch": 1.0450279046169457,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.041105585186164e-06,
+      "loss": 1.39720306,
+      "memory(GiB)": 123.79,
+      "step": 41195,
+      "train_speed(iter/s)": 1.315819
+    },
+    {
+      "acc": 0.69617519,
+      "epoch": 1.0451547437848807,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.040056996047587e-06,
+      "loss": 1.41748333,
+      "memory(GiB)": 123.79,
+      "step": 41200,
+      "train_speed(iter/s)": 1.315835
+    },
+    {
+      "acc": 0.69395213,
+      "epoch": 1.045281582952816,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.039008405147125e-06,
+      "loss": 1.35388632,
+      "memory(GiB)": 123.79,
+      "step": 41205,
+      "train_speed(iter/s)": 1.315854
+    },
+    {
+      "acc": 0.67363029,
+      "epoch": 1.045408422120751,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.0379598125308984e-06,
+      "loss": 1.42942219,
+      "memory(GiB)": 123.79,
+      "step": 41210,
+      "train_speed(iter/s)": 1.315872
+    },
+    {
+      "acc": 0.67687392,
+      "epoch": 1.045535261288686,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.036911218245029e-06,
+      "loss": 1.46771774,
+      "memory(GiB)": 123.79,
+      "step": 41215,
+      "train_speed(iter/s)": 1.31589
+    },
+    {
+      "acc": 0.68553133,
+      "epoch": 1.045662100456621,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.035862622335641e-06,
+      "loss": 1.40338745,
+      "memory(GiB)": 123.79,
+      "step": 41220,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.700737,
+      "epoch": 1.045788939624556,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.034814024848853e-06,
+      "loss": 1.33531113,
+      "memory(GiB)": 123.79,
+      "step": 41225,
+      "train_speed(iter/s)": 1.315925
+    },
+    {
+      "acc": 0.6917757,
+      "epoch": 1.045915778792491,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.033765425830791e-06,
+      "loss": 1.363375,
+      "memory(GiB)": 123.79,
+      "step": 41230,
+      "train_speed(iter/s)": 1.315944
+    },
+    {
+      "acc": 0.68602986,
+      "epoch": 1.046042617960426,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.032716825327573e-06,
+      "loss": 1.37337742,
+      "memory(GiB)": 123.79,
+      "step": 41235,
+      "train_speed(iter/s)": 1.315962
+    },
+    {
+      "acc": 0.68384724,
+      "epoch": 1.0461694571283613,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.031668223385323e-06,
+      "loss": 1.3656765,
+      "memory(GiB)": 123.79,
+      "step": 41240,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.67985582,
+      "epoch": 1.0462962962962963,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.030619620050163e-06,
+      "loss": 1.41175375,
+      "memory(GiB)": 123.79,
+      "step": 41245,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.70806341,
+      "epoch": 1.0464231354642313,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.029571015368217e-06,
+      "loss": 1.26494532,
+      "memory(GiB)": 123.79,
+      "step": 41250,
+      "train_speed(iter/s)": 1.316018
+    },
+    {
+      "acc": 0.67728372,
+      "epoch": 1.0465499746321665,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.028522409385605e-06,
+      "loss": 1.39388237,
+      "memory(GiB)": 123.79,
+      "step": 41255,
+      "train_speed(iter/s)": 1.316036
+    },
+    {
+      "acc": 0.69215374,
+      "epoch": 1.0466768138001015,
+      "grad_norm": 2.75,
+      "learning_rate": 5.0274738021484495e-06,
+      "loss": 1.32341471,
+      "memory(GiB)": 123.79,
+      "step": 41260,
+      "train_speed(iter/s)": 1.316054
+    },
+    {
+      "acc": 0.67598286,
+      "epoch": 1.0468036529680365,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.026425193702874e-06,
+      "loss": 1.42516232,
+      "memory(GiB)": 123.79,
+      "step": 41265,
+      "train_speed(iter/s)": 1.316073
+    },
+    {
+      "acc": 0.71298876,
+      "epoch": 1.0469304921359717,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.025376584095001e-06,
+      "loss": 1.24404182,
+      "memory(GiB)": 123.79,
+      "step": 41270,
+      "train_speed(iter/s)": 1.316088
+    },
+    {
+      "acc": 0.70167546,
+      "epoch": 1.0470573313039067,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.024327973370951e-06,
+      "loss": 1.29617863,
+      "memory(GiB)": 123.79,
+      "step": 41275,
+      "train_speed(iter/s)": 1.316107
+    },
+    {
+      "acc": 0.67333717,
+      "epoch": 1.0471841704718416,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.02327936157685e-06,
+      "loss": 1.44018707,
+      "memory(GiB)": 123.79,
+      "step": 41280,
+      "train_speed(iter/s)": 1.316125
+    },
+    {
+      "acc": 0.68083954,
+      "epoch": 1.0473110096397769,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.022230748758816e-06,
+      "loss": 1.36068153,
+      "memory(GiB)": 123.79,
+      "step": 41285,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.69344473,
+      "epoch": 1.0474378488077118,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.021182134962978e-06,
+      "loss": 1.40802269,
+      "memory(GiB)": 123.79,
+      "step": 41290,
+      "train_speed(iter/s)": 1.316156
+    },
+    {
+      "acc": 0.69712067,
+      "epoch": 1.0475646879756468,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.020133520235453e-06,
+      "loss": 1.40318708,
+      "memory(GiB)": 123.79,
+      "step": 41295,
+      "train_speed(iter/s)": 1.316172
+    },
+    {
+      "acc": 0.67666006,
+      "epoch": 1.047691527143582,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.019084904622367e-06,
+      "loss": 1.40555782,
+      "memory(GiB)": 123.79,
+      "step": 41300,
+      "train_speed(iter/s)": 1.316189
+    },
+    {
+      "acc": 0.68323712,
+      "epoch": 1.047818366311517,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.01803628816984e-06,
+      "loss": 1.38388138,
+      "memory(GiB)": 123.79,
+      "step": 41305,
+      "train_speed(iter/s)": 1.316207
+    },
+    {
+      "acc": 0.69203529,
+      "epoch": 1.047945205479452,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.016987670923998e-06,
+      "loss": 1.35278635,
+      "memory(GiB)": 123.79,
+      "step": 41310,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.69767528,
+      "epoch": 1.048072044647387,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.0159390529309615e-06,
+      "loss": 1.27844505,
+      "memory(GiB)": 123.79,
+      "step": 41315,
+      "train_speed(iter/s)": 1.316244
+    },
+    {
+      "acc": 0.69085398,
+      "epoch": 1.0481988838153222,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.014890434236854e-06,
+      "loss": 1.34962749,
+      "memory(GiB)": 123.79,
+      "step": 41320,
+      "train_speed(iter/s)": 1.316261
+    },
+    {
+      "acc": 0.67097497,
+      "epoch": 1.0483257229832572,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.0138418148878e-06,
+      "loss": 1.46227808,
+      "memory(GiB)": 123.79,
+      "step": 41325,
+      "train_speed(iter/s)": 1.316279
+    },
+    {
+      "acc": 0.69744859,
+      "epoch": 1.0484525621511922,
+      "grad_norm": 3.90625,
+      "learning_rate": 5.01279319492992e-06,
+      "loss": 1.39355526,
+      "memory(GiB)": 123.79,
+      "step": 41330,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.67433414,
+      "epoch": 1.0485794013191274,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.01174457440934e-06,
+      "loss": 1.39545069,
+      "memory(GiB)": 123.79,
+      "step": 41335,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.7041862,
+      "epoch": 1.0487062404870624,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.010695953372179e-06,
+      "loss": 1.34721928,
+      "memory(GiB)": 123.79,
+      "step": 41340,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.6890255,
+      "epoch": 1.0488330796549974,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.009647331864563e-06,
+      "loss": 1.35846777,
+      "memory(GiB)": 123.79,
+      "step": 41345,
+      "train_speed(iter/s)": 1.316352
+    },
+    {
+      "acc": 0.69091458,
+      "epoch": 1.0489599188229326,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.008598709932615e-06,
+      "loss": 1.40060825,
+      "memory(GiB)": 123.79,
+      "step": 41350,
+      "train_speed(iter/s)": 1.31637
+    },
+    {
+      "acc": 0.67730789,
+      "epoch": 1.0490867579908676,
+      "grad_norm": 2.875,
+      "learning_rate": 5.007550087622456e-06,
+      "loss": 1.43037663,
+      "memory(GiB)": 123.79,
+      "step": 41355,
+      "train_speed(iter/s)": 1.316388
+    },
+    {
+      "acc": 0.70004239,
+      "epoch": 1.0492135971588026,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.0065014649802124e-06,
+      "loss": 1.353442,
+      "memory(GiB)": 123.79,
+      "step": 41360,
+      "train_speed(iter/s)": 1.316406
+    },
+    {
+      "acc": 0.68031111,
+      "epoch": 1.0493404363267378,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.005452842052003e-06,
+      "loss": 1.41038647,
+      "memory(GiB)": 123.79,
+      "step": 41365,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.6927464,
+      "epoch": 1.0494672754946728,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.004404218883955e-06,
+      "loss": 1.40980377,
+      "memory(GiB)": 123.79,
+      "step": 41370,
+      "train_speed(iter/s)": 1.316441
+    },
+    {
+      "acc": 0.70375137,
+      "epoch": 1.0495941146626078,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.0033555955221875e-06,
+      "loss": 1.36949139,
+      "memory(GiB)": 123.79,
+      "step": 41375,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.69025917,
+      "epoch": 1.0497209538305428,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.002306972012829e-06,
+      "loss": 1.37603598,
+      "memory(GiB)": 123.79,
+      "step": 41380,
+      "train_speed(iter/s)": 1.316477
+    },
+    {
+      "acc": 0.68624773,
+      "epoch": 1.049847792998478,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.001258348401998e-06,
+      "loss": 1.3859252,
+      "memory(GiB)": 123.79,
+      "step": 41385,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.68855019,
+      "epoch": 1.049974632166413,
+      "grad_norm": 3.1875,
+      "learning_rate": 5.000209724735819e-06,
+      "loss": 1.41933193,
+      "memory(GiB)": 123.79,
+      "step": 41390,
+      "train_speed(iter/s)": 1.316513
+    },
+    {
+      "acc": 0.68366566,
+      "epoch": 1.050101471334348,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.999161101060416e-06,
+      "loss": 1.44510298,
+      "memory(GiB)": 123.79,
+      "step": 41395,
+      "train_speed(iter/s)": 1.316531
+    },
+    {
+      "acc": 0.67748508,
+      "epoch": 1.0502283105022832,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.99811247742191e-06,
+      "loss": 1.45273857,
+      "memory(GiB)": 123.79,
+      "step": 41400,
+      "train_speed(iter/s)": 1.316549
+    },
+    {
+      "acc": 0.68937173,
+      "epoch": 1.0503551496702181,
+      "grad_norm": 3.25,
+      "learning_rate": 4.9970638538664275e-06,
+      "loss": 1.40820045,
+      "memory(GiB)": 123.79,
+      "step": 41405,
+      "train_speed(iter/s)": 1.316566
+    },
+    {
+      "acc": 0.68267574,
+      "epoch": 1.0504819888381531,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.996015230440091e-06,
+      "loss": 1.37532568,
+      "memory(GiB)": 123.79,
+      "step": 41410,
+      "train_speed(iter/s)": 1.316583
+    },
+    {
+      "acc": 0.70209889,
+      "epoch": 1.0506088280060883,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.99496660718902e-06,
+      "loss": 1.30622692,
+      "memory(GiB)": 123.79,
+      "step": 41415,
+      "train_speed(iter/s)": 1.316601
+    },
+    {
+      "acc": 0.68244028,
+      "epoch": 1.0507356671740233,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.99391798415934e-06,
+      "loss": 1.399296,
+      "memory(GiB)": 123.79,
+      "step": 41420,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "acc": 0.68896866,
+      "epoch": 1.0508625063419583,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.992869361397175e-06,
+      "loss": 1.38459139,
+      "memory(GiB)": 123.79,
+      "step": 41425,
+      "train_speed(iter/s)": 1.316637
+    },
+    {
+      "acc": 0.68339252,
+      "epoch": 1.0509893455098935,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.991820738948649e-06,
+      "loss": 1.35987301,
+      "memory(GiB)": 123.79,
+      "step": 41430,
+      "train_speed(iter/s)": 1.316655
+    },
+    {
+      "acc": 0.67783833,
+      "epoch": 1.0511161846778285,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.9907721168598805e-06,
+      "loss": 1.40904465,
+      "memory(GiB)": 123.79,
+      "step": 41435,
+      "train_speed(iter/s)": 1.316673
+    },
+    {
+      "acc": 0.70122824,
+      "epoch": 1.0512430238457635,
+      "grad_norm": 2.75,
+      "learning_rate": 4.989723495176997e-06,
+      "loss": 1.31216106,
+      "memory(GiB)": 123.79,
+      "step": 41440,
+      "train_speed(iter/s)": 1.31669
+    },
+    {
+      "acc": 0.70187073,
+      "epoch": 1.0513698630136987,
+      "grad_norm": 2.875,
+      "learning_rate": 4.988674873946118e-06,
+      "loss": 1.31239815,
+      "memory(GiB)": 123.79,
+      "step": 41445,
+      "train_speed(iter/s)": 1.316709
+    },
+    {
+      "acc": 0.7062315,
+      "epoch": 1.0514967021816337,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.987626253213373e-06,
+      "loss": 1.35886211,
+      "memory(GiB)": 123.79,
+      "step": 41450,
+      "train_speed(iter/s)": 1.316727
+    },
+    {
+      "acc": 0.69554396,
+      "epoch": 1.0516235413495687,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.986577633024877e-06,
+      "loss": 1.38427429,
+      "memory(GiB)": 123.79,
+      "step": 41455,
+      "train_speed(iter/s)": 1.316745
+    },
+    {
+      "acc": 0.68351064,
+      "epoch": 1.051750380517504,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.985529013426758e-06,
+      "loss": 1.38653727,
+      "memory(GiB)": 123.79,
+      "step": 41460,
+      "train_speed(iter/s)": 1.316764
+    },
+    {
+      "acc": 0.68860884,
+      "epoch": 1.051877219685439,
+      "grad_norm": 3.125,
+      "learning_rate": 4.984480394465136e-06,
+      "loss": 1.35100651,
+      "memory(GiB)": 123.79,
+      "step": 41465,
+      "train_speed(iter/s)": 1.316779
+    },
+    {
+      "acc": 0.68639784,
+      "epoch": 1.052004058853374,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.9834317761861385e-06,
+      "loss": 1.39457779,
+      "memory(GiB)": 123.79,
+      "step": 41470,
+      "train_speed(iter/s)": 1.316797
+    },
+    {
+      "acc": 0.67577572,
+      "epoch": 1.0521308980213089,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.982383158635884e-06,
+      "loss": 1.40155802,
+      "memory(GiB)": 123.79,
+      "step": 41475,
+      "train_speed(iter/s)": 1.316815
+    },
+    {
+      "acc": 0.6784193,
+      "epoch": 1.052257737189244,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.981334541860496e-06,
+      "loss": 1.40158205,
+      "memory(GiB)": 123.79,
+      "step": 41480,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.68410816,
+      "epoch": 1.052384576357179,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.980285925906098e-06,
+      "loss": 1.42386131,
+      "memory(GiB)": 123.79,
+      "step": 41485,
+      "train_speed(iter/s)": 1.31685
+    },
+    {
+      "acc": 0.68598123,
+      "epoch": 1.052511415525114,
+      "grad_norm": 3.375,
+      "learning_rate": 4.9792373108188155e-06,
+      "loss": 1.39221764,
+      "memory(GiB)": 123.79,
+      "step": 41490,
+      "train_speed(iter/s)": 1.316867
+    },
+    {
+      "acc": 0.70062494,
+      "epoch": 1.0526382546930493,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.978188696644767e-06,
+      "loss": 1.35800915,
+      "memory(GiB)": 123.79,
+      "step": 41495,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.70121675,
+      "epoch": 1.0527650938609843,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.977140083430075e-06,
+      "loss": 1.33213711,
+      "memory(GiB)": 123.79,
+      "step": 41500,
+      "train_speed(iter/s)": 1.316901
+    },
+    {
+      "acc": 0.69826088,
+      "epoch": 1.0528919330289193,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.976091471220867e-06,
+      "loss": 1.34615612,
+      "memory(GiB)": 123.79,
+      "step": 41505,
+      "train_speed(iter/s)": 1.316918
+    },
+    {
+      "acc": 0.6785408,
+      "epoch": 1.0530187721968545,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.975042860063263e-06,
+      "loss": 1.40303984,
+      "memory(GiB)": 123.79,
+      "step": 41510,
+      "train_speed(iter/s)": 1.316935
+    },
+    {
+      "acc": 0.68332663,
+      "epoch": 1.0531456113647895,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.973994250003384e-06,
+      "loss": 1.41471786,
+      "memory(GiB)": 123.79,
+      "step": 41515,
+      "train_speed(iter/s)": 1.316952
+    },
+    {
+      "acc": 0.67248783,
+      "epoch": 1.0532724505327244,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.972945641087355e-06,
+      "loss": 1.46463499,
+      "memory(GiB)": 123.79,
+      "step": 41520,
+      "train_speed(iter/s)": 1.316968
+    },
+    {
+      "acc": 0.69341135,
+      "epoch": 1.0533992897006597,
+      "grad_norm": 2.875,
+      "learning_rate": 4.9718970333612955e-06,
+      "loss": 1.37654772,
+      "memory(GiB)": 123.79,
+      "step": 41525,
+      "train_speed(iter/s)": 1.316987
+    },
+    {
+      "acc": 0.6828146,
+      "epoch": 1.0535261288685946,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.970848426871333e-06,
+      "loss": 1.39432793,
+      "memory(GiB)": 123.79,
+      "step": 41530,
+      "train_speed(iter/s)": 1.317004
+    },
+    {
+      "acc": 0.68138566,
+      "epoch": 1.0536529680365296,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.9697998216635854e-06,
+      "loss": 1.42016773,
+      "memory(GiB)": 123.79,
+      "step": 41535,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.6913693,
+      "epoch": 1.0537798072044646,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.9687512177841765e-06,
+      "loss": 1.3525486,
+      "memory(GiB)": 123.79,
+      "step": 41540,
+      "train_speed(iter/s)": 1.317038
+    },
+    {
+      "acc": 0.67398615,
+      "epoch": 1.0539066463723998,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.967702615279227e-06,
+      "loss": 1.42418365,
+      "memory(GiB)": 123.79,
+      "step": 41545,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.67526608,
+      "epoch": 1.0540334855403348,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.966654014194863e-06,
+      "loss": 1.45549946,
+      "memory(GiB)": 123.79,
+      "step": 41550,
+      "train_speed(iter/s)": 1.317072
+    },
+    {
+      "acc": 0.69960136,
+      "epoch": 1.0541603247082698,
+      "grad_norm": 2.75,
+      "learning_rate": 4.965605414577204e-06,
+      "loss": 1.29549618,
+      "memory(GiB)": 123.79,
+      "step": 41555,
+      "train_speed(iter/s)": 1.317088
+    },
+    {
+      "acc": 0.69788408,
+      "epoch": 1.054287163876205,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.964556816472371e-06,
+      "loss": 1.36490536,
+      "memory(GiB)": 123.79,
+      "step": 41560,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.68913798,
+      "epoch": 1.05441400304414,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.9635082199264874e-06,
+      "loss": 1.41030912,
+      "memory(GiB)": 123.79,
+      "step": 41565,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.69621315,
+      "epoch": 1.054540842212075,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.962459624985677e-06,
+      "loss": 1.37980299,
+      "memory(GiB)": 123.79,
+      "step": 41570,
+      "train_speed(iter/s)": 1.317138
+    },
+    {
+      "acc": 0.68037353,
+      "epoch": 1.0546676813800102,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.961411031696059e-06,
+      "loss": 1.42727013,
+      "memory(GiB)": 123.79,
+      "step": 41575,
+      "train_speed(iter/s)": 1.316903
+    },
+    {
+      "acc": 0.69883795,
+      "epoch": 1.0547945205479452,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.960362440103756e-06,
+      "loss": 1.38087215,
+      "memory(GiB)": 123.79,
+      "step": 41580,
+      "train_speed(iter/s)": 1.31692
+    },
+    {
+      "acc": 0.70244579,
+      "epoch": 1.0549213597158802,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.95931385025489e-06,
+      "loss": 1.33525658,
+      "memory(GiB)": 123.79,
+      "step": 41585,
+      "train_speed(iter/s)": 1.316937
+    },
+    {
+      "acc": 0.68837709,
+      "epoch": 1.0550481988838154,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.958265262195584e-06,
+      "loss": 1.38721313,
+      "memory(GiB)": 123.79,
+      "step": 41590,
+      "train_speed(iter/s)": 1.316956
+    },
+    {
+      "acc": 0.70064688,
+      "epoch": 1.0551750380517504,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.957216675971955e-06,
+      "loss": 1.36784983,
+      "memory(GiB)": 123.79,
+      "step": 41595,
+      "train_speed(iter/s)": 1.316973
+    },
+    {
+      "acc": 0.68801208,
+      "epoch": 1.0553018772196854,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.9561680916301295e-06,
+      "loss": 1.45314579,
+      "memory(GiB)": 123.79,
+      "step": 41600,
+      "train_speed(iter/s)": 1.316991
+    },
+    {
+      "acc": 0.68968248,
+      "epoch": 1.0554287163876206,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.955119509216226e-06,
+      "loss": 1.36623268,
+      "memory(GiB)": 123.79,
+      "step": 41605,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.68703928,
+      "epoch": 1.0555555555555556,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.9540709287763685e-06,
+      "loss": 1.41497707,
+      "memory(GiB)": 123.79,
+      "step": 41610,
+      "train_speed(iter/s)": 1.317025
+    },
+    {
+      "acc": 0.68395128,
+      "epoch": 1.0556823947234906,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.953022350356676e-06,
+      "loss": 1.41918392,
+      "memory(GiB)": 123.79,
+      "step": 41615,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.68556142,
+      "epoch": 1.0558092338914258,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.951973774003269e-06,
+      "loss": 1.40372696,
+      "memory(GiB)": 123.79,
+      "step": 41620,
+      "train_speed(iter/s)": 1.317059
+    },
+    {
+      "acc": 0.687603,
+      "epoch": 1.0559360730593608,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.950925199762271e-06,
+      "loss": 1.43110228,
+      "memory(GiB)": 123.79,
+      "step": 41625,
+      "train_speed(iter/s)": 1.317076
+    },
+    {
+      "acc": 0.6855525,
+      "epoch": 1.0560629122272958,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.949876627679803e-06,
+      "loss": 1.40398083,
+      "memory(GiB)": 123.79,
+      "step": 41630,
+      "train_speed(iter/s)": 1.317094
+    },
+    {
+      "acc": 0.69546347,
+      "epoch": 1.0561897513952307,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.948828057801983e-06,
+      "loss": 1.37458477,
+      "memory(GiB)": 123.79,
+      "step": 41635,
+      "train_speed(iter/s)": 1.317112
+    },
+    {
+      "acc": 0.67329016,
+      "epoch": 1.056316590563166,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.947779490174933e-06,
+      "loss": 1.41619816,
+      "memory(GiB)": 123.79,
+      "step": 41640,
+      "train_speed(iter/s)": 1.317129
+    },
+    {
+      "acc": 0.69019446,
+      "epoch": 1.056443429731101,
+      "grad_norm": 3.5,
+      "learning_rate": 4.946730924844775e-06,
+      "loss": 1.41293583,
+      "memory(GiB)": 123.79,
+      "step": 41645,
+      "train_speed(iter/s)": 1.317146
+    },
+    {
+      "acc": 0.68222313,
+      "epoch": 1.056570268899036,
+      "grad_norm": 3.125,
+      "learning_rate": 4.945682361857631e-06,
+      "loss": 1.40958529,
+      "memory(GiB)": 123.79,
+      "step": 41650,
+      "train_speed(iter/s)": 1.317164
+    },
+    {
+      "acc": 0.692452,
+      "epoch": 1.0566971080669711,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.944633801259615e-06,
+      "loss": 1.38079557,
+      "memory(GiB)": 123.79,
+      "step": 41655,
+      "train_speed(iter/s)": 1.317182
+    },
+    {
+      "acc": 0.67812786,
+      "epoch": 1.0568239472349061,
+      "grad_norm": 3.96875,
+      "learning_rate": 4.943585243096854e-06,
+      "loss": 1.44228764,
+      "memory(GiB)": 123.79,
+      "step": 41660,
+      "train_speed(iter/s)": 1.317197
+    },
+    {
+      "acc": 0.7018343,
+      "epoch": 1.0569507864028411,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.942536687415465e-06,
+      "loss": 1.32808867,
+      "memory(GiB)": 123.79,
+      "step": 41665,
+      "train_speed(iter/s)": 1.317215
+    },
+    {
+      "acc": 0.67580128,
+      "epoch": 1.0570776255707763,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.941488134261571e-06,
+      "loss": 1.47694092,
+      "memory(GiB)": 123.79,
+      "step": 41670,
+      "train_speed(iter/s)": 1.317232
+    },
+    {
+      "acc": 0.70036712,
+      "epoch": 1.0572044647387113,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.940439583681288e-06,
+      "loss": 1.28730068,
+      "memory(GiB)": 123.79,
+      "step": 41675,
+      "train_speed(iter/s)": 1.31725
+    },
+    {
+      "acc": 0.69147987,
+      "epoch": 1.0573313039066463,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.939391035720739e-06,
+      "loss": 1.34731007,
+      "memory(GiB)": 123.79,
+      "step": 41680,
+      "train_speed(iter/s)": 1.317267
+    },
+    {
+      "acc": 0.70175724,
+      "epoch": 1.0574581430745815,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.938342490426041e-06,
+      "loss": 1.3103879,
+      "memory(GiB)": 123.79,
+      "step": 41685,
+      "train_speed(iter/s)": 1.317285
+    },
+    {
+      "acc": 0.69005313,
+      "epoch": 1.0575849822425165,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.937293947843318e-06,
+      "loss": 1.39130049,
+      "memory(GiB)": 123.79,
+      "step": 41690,
+      "train_speed(iter/s)": 1.317303
+    },
+    {
+      "acc": 0.69062524,
+      "epoch": 1.0577118214104515,
+      "grad_norm": 2.5,
+      "learning_rate": 4.936245408018687e-06,
+      "loss": 1.3448143,
+      "memory(GiB)": 123.79,
+      "step": 41695,
+      "train_speed(iter/s)": 1.317321
+    },
+    {
+      "acc": 0.68890142,
+      "epoch": 1.0578386605783865,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.935196870998265e-06,
+      "loss": 1.41435604,
+      "memory(GiB)": 123.79,
+      "step": 41700,
+      "train_speed(iter/s)": 1.317336
+    },
+    {
+      "acc": 0.68087692,
+      "epoch": 1.0579654997463217,
+      "grad_norm": 7.34375,
+      "learning_rate": 4.934148336828176e-06,
+      "loss": 1.45629435,
+      "memory(GiB)": 123.79,
+      "step": 41705,
+      "train_speed(iter/s)": 1.317352
+    },
+    {
+      "acc": 0.66222515,
+      "epoch": 1.0580923389142567,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.933099805554538e-06,
+      "loss": 1.49466181,
+      "memory(GiB)": 123.79,
+      "step": 41710,
+      "train_speed(iter/s)": 1.317368
+    },
+    {
+      "acc": 0.67583075,
+      "epoch": 1.0582191780821917,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.932051277223468e-06,
+      "loss": 1.45804167,
+      "memory(GiB)": 123.79,
+      "step": 41715,
+      "train_speed(iter/s)": 1.317385
+    },
+    {
+      "acc": 0.69026966,
+      "epoch": 1.058346017250127,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.931002751881086e-06,
+      "loss": 1.40150023,
+      "memory(GiB)": 123.79,
+      "step": 41720,
+      "train_speed(iter/s)": 1.3174
+    },
+    {
+      "acc": 0.68840265,
+      "epoch": 1.0584728564180619,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.929954229573512e-06,
+      "loss": 1.37382412,
+      "memory(GiB)": 123.79,
+      "step": 41725,
+      "train_speed(iter/s)": 1.317417
+    },
+    {
+      "acc": 0.67969503,
+      "epoch": 1.0585996955859969,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.9289057103468635e-06,
+      "loss": 1.36309023,
+      "memory(GiB)": 123.79,
+      "step": 41730,
+      "train_speed(iter/s)": 1.317434
+    },
+    {
+      "acc": 0.68782682,
+      "epoch": 1.058726534753932,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.927857194247258e-06,
+      "loss": 1.36434269,
+      "memory(GiB)": 123.79,
+      "step": 41735,
+      "train_speed(iter/s)": 1.31745
+    },
+    {
+      "acc": 0.67286181,
+      "epoch": 1.058853373921867,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.926808681320816e-06,
+      "loss": 1.43207245,
+      "memory(GiB)": 123.79,
+      "step": 41740,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.67939434,
+      "epoch": 1.058980213089802,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.925760171613654e-06,
+      "loss": 1.41663208,
+      "memory(GiB)": 123.79,
+      "step": 41745,
+      "train_speed(iter/s)": 1.317486
+    },
+    {
+      "acc": 0.6772119,
+      "epoch": 1.0591070522577373,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.9247116651718925e-06,
+      "loss": 1.40526199,
+      "memory(GiB)": 123.79,
+      "step": 41750,
+      "train_speed(iter/s)": 1.317503
+    },
+    {
+      "acc": 0.68116655,
+      "epoch": 1.0592338914256723,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.9236631620416486e-06,
+      "loss": 1.45487156,
+      "memory(GiB)": 123.79,
+      "step": 41755,
+      "train_speed(iter/s)": 1.317521
+    },
+    {
+      "acc": 0.69504814,
+      "epoch": 1.0593607305936072,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.922614662269038e-06,
+      "loss": 1.39469471,
+      "memory(GiB)": 123.79,
+      "step": 41760,
+      "train_speed(iter/s)": 1.317538
+    },
+    {
+      "acc": 0.69734125,
+      "epoch": 1.0594875697615425,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.9215661659001805e-06,
+      "loss": 1.33895645,
+      "memory(GiB)": 123.79,
+      "step": 41765,
+      "train_speed(iter/s)": 1.317552
+    },
+    {
+      "acc": 0.713416,
+      "epoch": 1.0596144089294774,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.920517672981195e-06,
+      "loss": 1.28497295,
+      "memory(GiB)": 123.79,
+      "step": 41770,
+      "train_speed(iter/s)": 1.317569
+    },
+    {
+      "acc": 0.68837676,
+      "epoch": 1.0597412480974124,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.919469183558195e-06,
+      "loss": 1.39973907,
+      "memory(GiB)": 123.79,
+      "step": 41775,
+      "train_speed(iter/s)": 1.317586
+    },
+    {
+      "acc": 0.69827452,
+      "epoch": 1.0598680872653476,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.9184206976773e-06,
+      "loss": 1.36464996,
+      "memory(GiB)": 123.79,
+      "step": 41780,
+      "train_speed(iter/s)": 1.317603
+    },
+    {
+      "acc": 0.67735014,
+      "epoch": 1.0599949264332826,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.917372215384627e-06,
+      "loss": 1.46106586,
+      "memory(GiB)": 123.79,
+      "step": 41785,
+      "train_speed(iter/s)": 1.317617
+    },
+    {
+      "acc": 0.68770986,
+      "epoch": 1.0601217656012176,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.916323736726295e-06,
+      "loss": 1.3508997,
+      "memory(GiB)": 123.79,
+      "step": 41790,
+      "train_speed(iter/s)": 1.317632
+    },
+    {
+      "acc": 0.68250899,
+      "epoch": 1.0602486047691526,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.9152752617484156e-06,
+      "loss": 1.39964352,
+      "memory(GiB)": 123.79,
+      "step": 41795,
+      "train_speed(iter/s)": 1.317649
+    },
+    {
+      "acc": 0.68214369,
+      "epoch": 1.0603754439370878,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.91422679049711e-06,
+      "loss": 1.40432177,
+      "memory(GiB)": 123.79,
+      "step": 41800,
+      "train_speed(iter/s)": 1.317666
+    },
+    {
+      "acc": 0.69586596,
+      "epoch": 1.0605022831050228,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.913178323018493e-06,
+      "loss": 1.33998299,
+      "memory(GiB)": 123.79,
+      "step": 41805,
+      "train_speed(iter/s)": 1.317682
+    },
+    {
+      "acc": 0.69630899,
+      "epoch": 1.0606291222729578,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.912129859358682e-06,
+      "loss": 1.3801157,
+      "memory(GiB)": 123.79,
+      "step": 41810,
+      "train_speed(iter/s)": 1.3177
+    },
+    {
+      "acc": 0.70026746,
+      "epoch": 1.060755961440893,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.9110813995637905e-06,
+      "loss": 1.3349761,
+      "memory(GiB)": 123.79,
+      "step": 41815,
+      "train_speed(iter/s)": 1.317716
+    },
+    {
+      "acc": 0.69722013,
+      "epoch": 1.060882800608828,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.910032943679936e-06,
+      "loss": 1.35427456,
+      "memory(GiB)": 123.79,
+      "step": 41820,
+      "train_speed(iter/s)": 1.317733
+    },
+    {
+      "acc": 0.69739513,
+      "epoch": 1.061009639776763,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.908984491753234e-06,
+      "loss": 1.35035629,
+      "memory(GiB)": 123.79,
+      "step": 41825,
+      "train_speed(iter/s)": 1.31775
+    },
+    {
+      "acc": 0.68285956,
+      "epoch": 1.0611364789446982,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.907936043829802e-06,
+      "loss": 1.40193977,
+      "memory(GiB)": 123.79,
+      "step": 41830,
+      "train_speed(iter/s)": 1.317768
+    },
+    {
+      "acc": 0.69066358,
+      "epoch": 1.0612633181126332,
+      "grad_norm": 3.0,
+      "learning_rate": 4.906887599955754e-06,
+      "loss": 1.38948393,
+      "memory(GiB)": 123.79,
+      "step": 41835,
+      "train_speed(iter/s)": 1.317785
+    },
+    {
+      "acc": 0.69732084,
+      "epoch": 1.0613901572805682,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.905839160177203e-06,
+      "loss": 1.32657738,
+      "memory(GiB)": 123.79,
+      "step": 41840,
+      "train_speed(iter/s)": 1.317802
+    },
+    {
+      "acc": 0.68642774,
+      "epoch": 1.0615169964485034,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.904790724540267e-06,
+      "loss": 1.35764561,
+      "memory(GiB)": 123.79,
+      "step": 41845,
+      "train_speed(iter/s)": 1.317819
+    },
+    {
+      "acc": 0.69186978,
+      "epoch": 1.0616438356164384,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.903742293091061e-06,
+      "loss": 1.36721153,
+      "memory(GiB)": 123.79,
+      "step": 41850,
+      "train_speed(iter/s)": 1.317836
+    },
+    {
+      "acc": 0.67866392,
+      "epoch": 1.0617706747843734,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.902693865875698e-06,
+      "loss": 1.43110647,
+      "memory(GiB)": 123.79,
+      "step": 41855,
+      "train_speed(iter/s)": 1.317852
+    },
+    {
+      "acc": 0.69401503,
+      "epoch": 1.0618975139523084,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.901645442940293e-06,
+      "loss": 1.37301178,
+      "memory(GiB)": 123.79,
+      "step": 41860,
+      "train_speed(iter/s)": 1.317868
+    },
+    {
+      "acc": 0.69375839,
+      "epoch": 1.0620243531202436,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.900597024330961e-06,
+      "loss": 1.35437231,
+      "memory(GiB)": 123.79,
+      "step": 41865,
+      "train_speed(iter/s)": 1.317885
+    },
+    {
+      "acc": 0.6839777,
+      "epoch": 1.0621511922881786,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.899548610093816e-06,
+      "loss": 1.41094656,
+      "memory(GiB)": 123.79,
+      "step": 41870,
+      "train_speed(iter/s)": 1.317902
+    },
+    {
+      "acc": 0.69750748,
+      "epoch": 1.0622780314561135,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.89850020027497e-06,
+      "loss": 1.37162361,
+      "memory(GiB)": 123.79,
+      "step": 41875,
+      "train_speed(iter/s)": 1.317919
+    },
+    {
+      "acc": 0.69634228,
+      "epoch": 1.0624048706240488,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.89745179492054e-06,
+      "loss": 1.36760254,
+      "memory(GiB)": 123.79,
+      "step": 41880,
+      "train_speed(iter/s)": 1.317935
+    },
+    {
+      "acc": 0.69060712,
+      "epoch": 1.0625317097919837,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.896403394076636e-06,
+      "loss": 1.39444389,
+      "memory(GiB)": 123.79,
+      "step": 41885,
+      "train_speed(iter/s)": 1.317952
+    },
+    {
+      "acc": 0.68275814,
+      "epoch": 1.0626585489599187,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.895354997789377e-06,
+      "loss": 1.4553525,
+      "memory(GiB)": 123.79,
+      "step": 41890,
+      "train_speed(iter/s)": 1.31797
+    },
+    {
+      "acc": 0.68997836,
+      "epoch": 1.062785388127854,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.894306606104869e-06,
+      "loss": 1.37566872,
+      "memory(GiB)": 123.79,
+      "step": 41895,
+      "train_speed(iter/s)": 1.317987
+    },
+    {
+      "acc": 0.69017882,
+      "epoch": 1.062912227295789,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.893258219069229e-06,
+      "loss": 1.36972771,
+      "memory(GiB)": 123.79,
+      "step": 41900,
+      "train_speed(iter/s)": 1.318004
+    },
+    {
+      "acc": 0.69852314,
+      "epoch": 1.063039066463724,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.892209836728569e-06,
+      "loss": 1.33775291,
+      "memory(GiB)": 123.79,
+      "step": 41905,
+      "train_speed(iter/s)": 1.318021
+    },
+    {
+      "acc": 0.68708744,
+      "epoch": 1.0631659056316591,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.891161459129003e-06,
+      "loss": 1.42424908,
+      "memory(GiB)": 123.79,
+      "step": 41910,
+      "train_speed(iter/s)": 1.318039
+    },
+    {
+      "acc": 0.69828997,
+      "epoch": 1.0632927447995941,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.890113086316641e-06,
+      "loss": 1.34873438,
+      "memory(GiB)": 123.79,
+      "step": 41915,
+      "train_speed(iter/s)": 1.318055
+    },
+    {
+      "acc": 0.69058104,
+      "epoch": 1.0634195839675291,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.889064718337595e-06,
+      "loss": 1.36505108,
+      "memory(GiB)": 123.79,
+      "step": 41920,
+      "train_speed(iter/s)": 1.318072
+    },
+    {
+      "acc": 0.69170065,
+      "epoch": 1.0635464231354643,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.888016355237979e-06,
+      "loss": 1.36455708,
+      "memory(GiB)": 123.79,
+      "step": 41925,
+      "train_speed(iter/s)": 1.318089
+    },
+    {
+      "acc": 0.68835793,
+      "epoch": 1.0636732623033993,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.886967997063905e-06,
+      "loss": 1.35614758,
+      "memory(GiB)": 123.79,
+      "step": 41930,
+      "train_speed(iter/s)": 1.318106
+    },
+    {
+      "acc": 0.69174771,
+      "epoch": 1.0638001014713343,
+      "grad_norm": 3.625,
+      "learning_rate": 4.885919643861482e-06,
+      "loss": 1.30944118,
+      "memory(GiB)": 123.79,
+      "step": 41935,
+      "train_speed(iter/s)": 1.318123
+    },
+    {
+      "acc": 0.69716167,
+      "epoch": 1.0639269406392695,
+      "grad_norm": 5.25,
+      "learning_rate": 4.884871295676821e-06,
+      "loss": 1.27372589,
+      "memory(GiB)": 123.79,
+      "step": 41940,
+      "train_speed(iter/s)": 1.318136
+    },
+    {
+      "acc": 0.68881569,
+      "epoch": 1.0640537798072045,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.883822952556036e-06,
+      "loss": 1.37028828,
+      "memory(GiB)": 123.79,
+      "step": 41945,
+      "train_speed(iter/s)": 1.318153
+    },
+    {
+      "acc": 0.68664694,
+      "epoch": 1.0641806189751395,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.882774614545237e-06,
+      "loss": 1.34929962,
+      "memory(GiB)": 123.79,
+      "step": 41950,
+      "train_speed(iter/s)": 1.318171
+    },
+    {
+      "acc": 0.68796382,
+      "epoch": 1.0643074581430745,
+      "grad_norm": 3.5,
+      "learning_rate": 4.881726281690531e-06,
+      "loss": 1.40832615,
+      "memory(GiB)": 123.79,
+      "step": 41955,
+      "train_speed(iter/s)": 1.318187
+    },
+    {
+      "acc": 0.67361078,
+      "epoch": 1.0644342973110097,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.8806779540380335e-06,
+      "loss": 1.43644266,
+      "memory(GiB)": 123.79,
+      "step": 41960,
+      "train_speed(iter/s)": 1.318204
+    },
+    {
+      "acc": 0.69961576,
+      "epoch": 1.0645611364789447,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.879629631633851e-06,
+      "loss": 1.36061707,
+      "memory(GiB)": 123.79,
+      "step": 41965,
+      "train_speed(iter/s)": 1.318221
+    },
+    {
+      "acc": 0.67795,
+      "epoch": 1.0646879756468797,
+      "grad_norm": 3.625,
+      "learning_rate": 4.8785813145240965e-06,
+      "loss": 1.38920288,
+      "memory(GiB)": 123.79,
+      "step": 41970,
+      "train_speed(iter/s)": 1.318237
+    },
+    {
+      "acc": 0.6826746,
+      "epoch": 1.0648148148148149,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.877533002754877e-06,
+      "loss": 1.36768417,
+      "memory(GiB)": 123.79,
+      "step": 41975,
+      "train_speed(iter/s)": 1.318254
+    },
+    {
+      "acc": 0.66969547,
+      "epoch": 1.0649416539827499,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.8764846963723025e-06,
+      "loss": 1.45168247,
+      "memory(GiB)": 123.79,
+      "step": 41980,
+      "train_speed(iter/s)": 1.318271
+    },
+    {
+      "acc": 0.68236341,
+      "epoch": 1.0650684931506849,
+      "grad_norm": 3.5,
+      "learning_rate": 4.875436395422481e-06,
+      "loss": 1.39468498,
+      "memory(GiB)": 123.79,
+      "step": 41985,
+      "train_speed(iter/s)": 1.318287
+    },
+    {
+      "acc": 0.68865366,
+      "epoch": 1.06519533231862,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.874388099951527e-06,
+      "loss": 1.36500902,
+      "memory(GiB)": 123.79,
+      "step": 41990,
+      "train_speed(iter/s)": 1.318304
+    },
+    {
+      "acc": 0.71486712,
+      "epoch": 1.065322171486555,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.873339810005543e-06,
+      "loss": 1.33164835,
+      "memory(GiB)": 123.79,
+      "step": 41995,
+      "train_speed(iter/s)": 1.318321
+    },
+    {
+      "acc": 0.68078265,
+      "epoch": 1.06544901065449,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.872291525630638e-06,
+      "loss": 1.45259886,
+      "memory(GiB)": 123.79,
+      "step": 42000,
+      "train_speed(iter/s)": 1.318338
+    },
+    {
+      "epoch": 1.06544901065449,
+      "eval_acc": 0.6749731130588046,
+      "eval_loss": 1.3566150665283203,
+      "eval_runtime": 69.5118,
+      "eval_samples_per_second": 91.639,
+      "eval_steps_per_second": 22.917,
+      "step": 42000
+    },
+    {
+      "acc": 0.69336996,
+      "epoch": 1.0655758498224253,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.871243246872923e-06,
+      "loss": 1.36313763,
+      "memory(GiB)": 123.79,
+      "step": 42005,
+      "train_speed(iter/s)": 1.314968
+    },
+    {
+      "acc": 0.68941412,
+      "epoch": 1.0657026889903602,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.870194973778506e-06,
+      "loss": 1.37677708,
+      "memory(GiB)": 123.79,
+      "step": 42010,
+      "train_speed(iter/s)": 1.314984
+    },
+    {
+      "acc": 0.67747755,
+      "epoch": 1.0658295281582952,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.869146706393493e-06,
+      "loss": 1.43487244,
+      "memory(GiB)": 123.79,
+      "step": 42015,
+      "train_speed(iter/s)": 1.315002
+    },
+    {
+      "acc": 0.68886299,
+      "epoch": 1.0659563673262302,
+      "grad_norm": 3.375,
+      "learning_rate": 4.868098444763991e-06,
+      "loss": 1.33202934,
+      "memory(GiB)": 123.79,
+      "step": 42020,
+      "train_speed(iter/s)": 1.31502
+    },
+    {
+      "acc": 0.68795166,
+      "epoch": 1.0660832064941654,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.86705018893611e-06,
+      "loss": 1.39436865,
+      "memory(GiB)": 123.79,
+      "step": 42025,
+      "train_speed(iter/s)": 1.315038
+    },
+    {
+      "acc": 0.69692578,
+      "epoch": 1.0662100456621004,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.866001938955955e-06,
+      "loss": 1.35457592,
+      "memory(GiB)": 123.79,
+      "step": 42030,
+      "train_speed(iter/s)": 1.315056
+    },
+    {
+      "acc": 0.68545647,
+      "epoch": 1.0663368848300354,
+      "grad_norm": 2.75,
+      "learning_rate": 4.864953694869632e-06,
+      "loss": 1.38228025,
+      "memory(GiB)": 123.79,
+      "step": 42035,
+      "train_speed(iter/s)": 1.315073
+    },
+    {
+      "acc": 0.69479413,
+      "epoch": 1.0664637239979706,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.863905456723249e-06,
+      "loss": 1.38499203,
+      "memory(GiB)": 123.79,
+      "step": 42040,
+      "train_speed(iter/s)": 1.315091
+    },
+    {
+      "acc": 0.66507301,
+      "epoch": 1.0665905631659056,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.8628572245629105e-06,
+      "loss": 1.47754927,
+      "memory(GiB)": 123.79,
+      "step": 42045,
+      "train_speed(iter/s)": 1.315108
+    },
+    {
+      "acc": 0.68155336,
+      "epoch": 1.0667174023338406,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.861808998434726e-06,
+      "loss": 1.38732662,
+      "memory(GiB)": 123.79,
+      "step": 42050,
+      "train_speed(iter/s)": 1.315125
+    },
+    {
+      "acc": 0.69675846,
+      "epoch": 1.0668442415017758,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.860760778384797e-06,
+      "loss": 1.31606331,
+      "memory(GiB)": 123.79,
+      "step": 42055,
+      "train_speed(iter/s)": 1.315142
+    },
+    {
+      "acc": 0.68654156,
+      "epoch": 1.0669710806697108,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.85971256445923e-06,
+      "loss": 1.33651419,
+      "memory(GiB)": 123.79,
+      "step": 42060,
+      "train_speed(iter/s)": 1.315159
+    },
+    {
+      "acc": 0.68042846,
+      "epoch": 1.0670979198376458,
+      "grad_norm": 3.625,
+      "learning_rate": 4.858664356704131e-06,
+      "loss": 1.42237206,
+      "memory(GiB)": 123.79,
+      "step": 42065,
+      "train_speed(iter/s)": 1.315177
+    },
+    {
+      "acc": 0.71346378,
+      "epoch": 1.067224759005581,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.857616155165606e-06,
+      "loss": 1.23811646,
+      "memory(GiB)": 123.79,
+      "step": 42070,
+      "train_speed(iter/s)": 1.315194
+    },
+    {
+      "acc": 0.66731954,
+      "epoch": 1.067351598173516,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.856567959889758e-06,
+      "loss": 1.46808405,
+      "memory(GiB)": 123.79,
+      "step": 42075,
+      "train_speed(iter/s)": 1.315211
+    },
+    {
+      "acc": 0.69049625,
+      "epoch": 1.067478437341451,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.855519770922691e-06,
+      "loss": 1.40292683,
+      "memory(GiB)": 123.79,
+      "step": 42080,
+      "train_speed(iter/s)": 1.315229
+    },
+    {
+      "acc": 0.6861043,
+      "epoch": 1.0676052765093862,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.8544715883105084e-06,
+      "loss": 1.38639297,
+      "memory(GiB)": 123.79,
+      "step": 42085,
+      "train_speed(iter/s)": 1.315247
+    },
+    {
+      "acc": 0.68816314,
+      "epoch": 1.0677321156773212,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.853423412099318e-06,
+      "loss": 1.39509735,
+      "memory(GiB)": 123.79,
+      "step": 42090,
+      "train_speed(iter/s)": 1.315264
+    },
+    {
+      "acc": 0.67623444,
+      "epoch": 1.0678589548452562,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.852375242335217e-06,
+      "loss": 1.42287922,
+      "memory(GiB)": 123.79,
+      "step": 42095,
+      "train_speed(iter/s)": 1.315282
+    },
+    {
+      "acc": 0.68591385,
+      "epoch": 1.0679857940131914,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.851327079064314e-06,
+      "loss": 1.39017582,
+      "memory(GiB)": 123.79,
+      "step": 42100,
+      "train_speed(iter/s)": 1.315295
+    },
+    {
+      "acc": 0.68025155,
+      "epoch": 1.0681126331811264,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.850278922332708e-06,
+      "loss": 1.47019691,
+      "memory(GiB)": 123.79,
+      "step": 42105,
+      "train_speed(iter/s)": 1.315312
+    },
+    {
+      "acc": 0.69324083,
+      "epoch": 1.0682394723490614,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.849230772186508e-06,
+      "loss": 1.38092709,
+      "memory(GiB)": 123.79,
+      "step": 42110,
+      "train_speed(iter/s)": 1.315329
+    },
+    {
+      "acc": 0.69374704,
+      "epoch": 1.0683663115169963,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.848182628671806e-06,
+      "loss": 1.35213947,
+      "memory(GiB)": 123.79,
+      "step": 42115,
+      "train_speed(iter/s)": 1.315346
+    },
+    {
+      "acc": 0.70327306,
+      "epoch": 1.0684931506849316,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.847134491834713e-06,
+      "loss": 1.34299288,
+      "memory(GiB)": 123.79,
+      "step": 42120,
+      "train_speed(iter/s)": 1.315363
+    },
+    {
+      "acc": 0.69076471,
+      "epoch": 1.0686199898528665,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.846086361721326e-06,
+      "loss": 1.34305172,
+      "memory(GiB)": 123.79,
+      "step": 42125,
+      "train_speed(iter/s)": 1.315381
+    },
+    {
+      "acc": 0.68385506,
+      "epoch": 1.0687468290208015,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.84503823837775e-06,
+      "loss": 1.38433142,
+      "memory(GiB)": 123.79,
+      "step": 42130,
+      "train_speed(iter/s)": 1.315399
+    },
+    {
+      "acc": 0.68804836,
+      "epoch": 1.0688736681887367,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.843990121850083e-06,
+      "loss": 1.34983997,
+      "memory(GiB)": 123.79,
+      "step": 42135,
+      "train_speed(iter/s)": 1.315416
+    },
+    {
+      "acc": 0.68068423,
+      "epoch": 1.0690005073566717,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.842942012184426e-06,
+      "loss": 1.41075182,
+      "memory(GiB)": 123.79,
+      "step": 42140,
+      "train_speed(iter/s)": 1.315431
+    },
+    {
+      "acc": 0.67701235,
+      "epoch": 1.0691273465246067,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.841893909426881e-06,
+      "loss": 1.42131596,
+      "memory(GiB)": 123.79,
+      "step": 42145,
+      "train_speed(iter/s)": 1.315449
+    },
+    {
+      "acc": 0.68228474,
+      "epoch": 1.069254185692542,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.84084581362355e-06,
+      "loss": 1.43077574,
+      "memory(GiB)": 123.79,
+      "step": 42150,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.69408312,
+      "epoch": 1.069381024860477,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.839797724820529e-06,
+      "loss": 1.39281454,
+      "memory(GiB)": 123.79,
+      "step": 42155,
+      "train_speed(iter/s)": 1.315484
+    },
+    {
+      "acc": 0.68279982,
+      "epoch": 1.069507864028412,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.838749643063918e-06,
+      "loss": 1.42620392,
+      "memory(GiB)": 123.79,
+      "step": 42160,
+      "train_speed(iter/s)": 1.315502
+    },
+    {
+      "acc": 0.6829164,
+      "epoch": 1.0696347031963471,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.837701568399819e-06,
+      "loss": 1.42123413,
+      "memory(GiB)": 123.79,
+      "step": 42165,
+      "train_speed(iter/s)": 1.31552
+    },
+    {
+      "acc": 0.67131171,
+      "epoch": 1.0697615423642821,
+      "grad_norm": 2.625,
+      "learning_rate": 4.836653500874331e-06,
+      "loss": 1.41741695,
+      "memory(GiB)": 123.79,
+      "step": 42170,
+      "train_speed(iter/s)": 1.315537
+    },
+    {
+      "acc": 0.68244648,
+      "epoch": 1.069888381532217,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.835605440533549e-06,
+      "loss": 1.43020277,
+      "memory(GiB)": 123.79,
+      "step": 42175,
+      "train_speed(iter/s)": 1.315555
+    },
+    {
+      "acc": 0.69378738,
+      "epoch": 1.070015220700152,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.834557387423575e-06,
+      "loss": 1.30366821,
+      "memory(GiB)": 123.79,
+      "step": 42180,
+      "train_speed(iter/s)": 1.315573
+    },
+    {
+      "acc": 0.68771639,
+      "epoch": 1.0701420598680873,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.833509341590503e-06,
+      "loss": 1.40145826,
+      "memory(GiB)": 123.79,
+      "step": 42185,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.69375887,
+      "epoch": 1.0702688990360223,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.8324613030804374e-06,
+      "loss": 1.31909485,
+      "memory(GiB)": 123.79,
+      "step": 42190,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.67827182,
+      "epoch": 1.0703957382039573,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.83141327193947e-06,
+      "loss": 1.44485817,
+      "memory(GiB)": 123.79,
+      "step": 42195,
+      "train_speed(iter/s)": 1.315617
+    },
+    {
+      "acc": 0.69497409,
+      "epoch": 1.0705225773718925,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.8303652482137e-06,
+      "loss": 1.34619989,
+      "memory(GiB)": 123.79,
+      "step": 42200,
+      "train_speed(iter/s)": 1.315635
+    },
+    {
+      "acc": 0.66477442,
+      "epoch": 1.0706494165398275,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.829317231949222e-06,
+      "loss": 1.45125208,
+      "memory(GiB)": 123.79,
+      "step": 42205,
+      "train_speed(iter/s)": 1.315652
+    },
+    {
+      "acc": 0.69998617,
+      "epoch": 1.0707762557077625,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.828269223192137e-06,
+      "loss": 1.38125458,
+      "memory(GiB)": 123.79,
+      "step": 42210,
+      "train_speed(iter/s)": 1.315669
+    },
+    {
+      "acc": 0.68963428,
+      "epoch": 1.0709030948756977,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.827221221988537e-06,
+      "loss": 1.36191387,
+      "memory(GiB)": 123.79,
+      "step": 42215,
+      "train_speed(iter/s)": 1.315687
+    },
+    {
+      "acc": 0.68884268,
+      "epoch": 1.0710299340436327,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.826173228384518e-06,
+      "loss": 1.36079559,
+      "memory(GiB)": 123.79,
+      "step": 42220,
+      "train_speed(iter/s)": 1.315705
+    },
+    {
+      "acc": 0.68920517,
+      "epoch": 1.0711567732115677,
+      "grad_norm": 2.75,
+      "learning_rate": 4.8251252424261775e-06,
+      "loss": 1.42328272,
+      "memory(GiB)": 123.79,
+      "step": 42225,
+      "train_speed(iter/s)": 1.315722
+    },
+    {
+      "acc": 0.70638714,
+      "epoch": 1.0712836123795029,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.8240772641596105e-06,
+      "loss": 1.29822559,
+      "memory(GiB)": 123.79,
+      "step": 42230,
+      "train_speed(iter/s)": 1.31574
+    },
+    {
+      "acc": 0.6997592,
+      "epoch": 1.0714104515474379,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.82302929363091e-06,
+      "loss": 1.38390923,
+      "memory(GiB)": 123.79,
+      "step": 42235,
+      "train_speed(iter/s)": 1.315757
+    },
+    {
+      "acc": 0.68283935,
+      "epoch": 1.0715372907153728,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.8219813308861705e-06,
+      "loss": 1.40877991,
+      "memory(GiB)": 123.79,
+      "step": 42240,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.70098815,
+      "epoch": 1.071664129883308,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.820933375971487e-06,
+      "loss": 1.30973177,
+      "memory(GiB)": 123.79,
+      "step": 42245,
+      "train_speed(iter/s)": 1.315793
+    },
+    {
+      "acc": 0.68922582,
+      "epoch": 1.071790969051243,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.819885428932955e-06,
+      "loss": 1.38889084,
+      "memory(GiB)": 123.79,
+      "step": 42250,
+      "train_speed(iter/s)": 1.315811
+    },
+    {
+      "acc": 0.69750824,
+      "epoch": 1.071917808219178,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.818837489816664e-06,
+      "loss": 1.34858465,
+      "memory(GiB)": 123.79,
+      "step": 42255,
+      "train_speed(iter/s)": 1.315829
+    },
+    {
+      "acc": 0.68363938,
+      "epoch": 1.0720446473871132,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.81778955866871e-06,
+      "loss": 1.43478146,
+      "memory(GiB)": 123.79,
+      "step": 42260,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.69742594,
+      "epoch": 1.0721714865550482,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.816741635535183e-06,
+      "loss": 1.35909557,
+      "memory(GiB)": 123.79,
+      "step": 42265,
+      "train_speed(iter/s)": 1.315864
+    },
+    {
+      "acc": 0.66986127,
+      "epoch": 1.0722983257229832,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.81569372046218e-06,
+      "loss": 1.48864861,
+      "memory(GiB)": 123.79,
+      "step": 42270,
+      "train_speed(iter/s)": 1.315882
+    },
+    {
+      "acc": 0.67541547,
+      "epoch": 1.0724251648909182,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.814645813495788e-06,
+      "loss": 1.44724855,
+      "memory(GiB)": 123.79,
+      "step": 42275,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.68247781,
+      "epoch": 1.0725520040588534,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.8135979146821e-06,
+      "loss": 1.40128031,
+      "memory(GiB)": 123.79,
+      "step": 42280,
+      "train_speed(iter/s)": 1.315917
+    },
+    {
+      "acc": 0.67842264,
+      "epoch": 1.0726788432267884,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.81255002406721e-06,
+      "loss": 1.38718615,
+      "memory(GiB)": 123.79,
+      "step": 42285,
+      "train_speed(iter/s)": 1.315934
+    },
+    {
+      "acc": 0.70683966,
+      "epoch": 1.0728056823947234,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.811502141697206e-06,
+      "loss": 1.3082181,
+      "memory(GiB)": 123.79,
+      "step": 42290,
+      "train_speed(iter/s)": 1.315952
+    },
+    {
+      "acc": 0.68439131,
+      "epoch": 1.0729325215626586,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.81045426761818e-06,
+      "loss": 1.38064175,
+      "memory(GiB)": 123.79,
+      "step": 42295,
+      "train_speed(iter/s)": 1.31597
+    },
+    {
+      "acc": 0.69093857,
+      "epoch": 1.0730593607305936,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.80940640187622e-06,
+      "loss": 1.4043437,
+      "memory(GiB)": 123.79,
+      "step": 42300,
+      "train_speed(iter/s)": 1.315986
+    },
+    {
+      "acc": 0.69094629,
+      "epoch": 1.0731861998985286,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.808358544517418e-06,
+      "loss": 1.33199253,
+      "memory(GiB)": 123.79,
+      "step": 42305,
+      "train_speed(iter/s)": 1.316005
+    },
+    {
+      "acc": 0.69518533,
+      "epoch": 1.0733130390664638,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.807310695587865e-06,
+      "loss": 1.31435099,
+      "memory(GiB)": 123.79,
+      "step": 42310,
+      "train_speed(iter/s)": 1.316023
+    },
+    {
+      "acc": 0.66695709,
+      "epoch": 1.0734398782343988,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.8062628551336445e-06,
+      "loss": 1.4860446,
+      "memory(GiB)": 123.79,
+      "step": 42315,
+      "train_speed(iter/s)": 1.31604
+    },
+    {
+      "acc": 0.68958349,
+      "epoch": 1.0735667174023338,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.80521502320085e-06,
+      "loss": 1.41283836,
+      "memory(GiB)": 123.79,
+      "step": 42320,
+      "train_speed(iter/s)": 1.316058
+    },
+    {
+      "acc": 0.68110747,
+      "epoch": 1.073693556570269,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.804167199835567e-06,
+      "loss": 1.38596153,
+      "memory(GiB)": 123.79,
+      "step": 42325,
+      "train_speed(iter/s)": 1.316073
+    },
+    {
+      "acc": 0.68990521,
+      "epoch": 1.073820395738204,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.8031193850838894e-06,
+      "loss": 1.33286839,
+      "memory(GiB)": 123.79,
+      "step": 42330,
+      "train_speed(iter/s)": 1.31609
+    },
+    {
+      "acc": 0.68667359,
+      "epoch": 1.073947234906139,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.802071578991896e-06,
+      "loss": 1.41093159,
+      "memory(GiB)": 123.79,
+      "step": 42335,
+      "train_speed(iter/s)": 1.316108
+    },
+    {
+      "acc": 0.68357396,
+      "epoch": 1.074074074074074,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.801023781605679e-06,
+      "loss": 1.40569286,
+      "memory(GiB)": 123.79,
+      "step": 42340,
+      "train_speed(iter/s)": 1.316124
+    },
+    {
+      "acc": 0.68440313,
+      "epoch": 1.0742009132420092,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.799975992971325e-06,
+      "loss": 1.39234409,
+      "memory(GiB)": 123.79,
+      "step": 42345,
+      "train_speed(iter/s)": 1.316141
+    },
+    {
+      "acc": 0.6974195,
+      "epoch": 1.0743277524099442,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.798928213134921e-06,
+      "loss": 1.3413312,
+      "memory(GiB)": 123.79,
+      "step": 42350,
+      "train_speed(iter/s)": 1.316158
+    },
+    {
+      "acc": 0.6946846,
+      "epoch": 1.0744545915778791,
+      "grad_norm": 3.375,
+      "learning_rate": 4.797880442142551e-06,
+      "loss": 1.41558523,
+      "memory(GiB)": 123.79,
+      "step": 42355,
+      "train_speed(iter/s)": 1.316175
+    },
+    {
+      "acc": 0.6839859,
+      "epoch": 1.0745814307458144,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.7968326800403e-06,
+      "loss": 1.35536051,
+      "memory(GiB)": 123.79,
+      "step": 42360,
+      "train_speed(iter/s)": 1.316191
+    },
+    {
+      "acc": 0.67837996,
+      "epoch": 1.0747082699137493,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.795784926874255e-06,
+      "loss": 1.42456064,
+      "memory(GiB)": 123.79,
+      "step": 42365,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.68718743,
+      "epoch": 1.0748351090816843,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.794737182690503e-06,
+      "loss": 1.39976978,
+      "memory(GiB)": 123.79,
+      "step": 42370,
+      "train_speed(iter/s)": 1.316225
+    },
+    {
+      "acc": 0.66365671,
+      "epoch": 1.0749619482496195,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.793689447535126e-06,
+      "loss": 1.37817583,
+      "memory(GiB)": 123.79,
+      "step": 42375,
+      "train_speed(iter/s)": 1.316242
+    },
+    {
+      "acc": 0.70727758,
+      "epoch": 1.0750887874175545,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.792641721454206e-06,
+      "loss": 1.3888155,
+      "memory(GiB)": 123.79,
+      "step": 42380,
+      "train_speed(iter/s)": 1.316259
+    },
+    {
+      "acc": 0.69507017,
+      "epoch": 1.0752156265854895,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.79159400449383e-06,
+      "loss": 1.36644135,
+      "memory(GiB)": 123.79,
+      "step": 42385,
+      "train_speed(iter/s)": 1.316275
+    },
+    {
+      "acc": 0.6870358,
+      "epoch": 1.0753424657534247,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.7905462967000816e-06,
+      "loss": 1.40440369,
+      "memory(GiB)": 123.79,
+      "step": 42390,
+      "train_speed(iter/s)": 1.316292
+    },
+    {
+      "acc": 0.67982659,
+      "epoch": 1.0754693049213597,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.789498598119039e-06,
+      "loss": 1.43694487,
+      "memory(GiB)": 123.79,
+      "step": 42395,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.69262261,
+      "epoch": 1.0755961440892947,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.78845090879679e-06,
+      "loss": 1.37925615,
+      "memory(GiB)": 123.79,
+      "step": 42400,
+      "train_speed(iter/s)": 1.316328
+    },
+    {
+      "acc": 0.68291998,
+      "epoch": 1.07572298325723,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.787403228779413e-06,
+      "loss": 1.38226528,
+      "memory(GiB)": 123.79,
+      "step": 42405,
+      "train_speed(iter/s)": 1.316345
+    },
+    {
+      "acc": 0.68802938,
+      "epoch": 1.075849822425165,
+      "grad_norm": 4.125,
+      "learning_rate": 4.786355558112994e-06,
+      "loss": 1.43257141,
+      "memory(GiB)": 123.79,
+      "step": 42410,
+      "train_speed(iter/s)": 1.316361
+    },
+    {
+      "acc": 0.70172377,
+      "epoch": 1.0759766615931,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.78530789684361e-06,
+      "loss": 1.33823757,
+      "memory(GiB)": 123.79,
+      "step": 42415,
+      "train_speed(iter/s)": 1.316378
+    },
+    {
+      "acc": 0.67627296,
+      "epoch": 1.0761035007610351,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.784260245017343e-06,
+      "loss": 1.44360447,
+      "memory(GiB)": 123.79,
+      "step": 42420,
+      "train_speed(iter/s)": 1.316395
+    },
+    {
+      "acc": 0.68637695,
+      "epoch": 1.07623033992897,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.7832126026802725e-06,
+      "loss": 1.39731121,
+      "memory(GiB)": 123.79,
+      "step": 42425,
+      "train_speed(iter/s)": 1.316412
+    },
+    {
+      "acc": 0.69489985,
+      "epoch": 1.076357179096905,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.782164969878482e-06,
+      "loss": 1.32589788,
+      "memory(GiB)": 123.79,
+      "step": 42430,
+      "train_speed(iter/s)": 1.31643
+    },
+    {
+      "acc": 0.67402563,
+      "epoch": 1.07648401826484,
+      "grad_norm": 2.75,
+      "learning_rate": 4.781117346658047e-06,
+      "loss": 1.41264048,
+      "memory(GiB)": 123.79,
+      "step": 42435,
+      "train_speed(iter/s)": 1.316446
+    },
+    {
+      "acc": 0.6797462,
+      "epoch": 1.0766108574327753,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.780069733065048e-06,
+      "loss": 1.38716679,
+      "memory(GiB)": 123.79,
+      "step": 42440,
+      "train_speed(iter/s)": 1.31646
+    },
+    {
+      "acc": 0.70088654,
+      "epoch": 1.0767376966007103,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.779022129145566e-06,
+      "loss": 1.33729973,
+      "memory(GiB)": 123.79,
+      "step": 42445,
+      "train_speed(iter/s)": 1.316476
+    },
+    {
+      "acc": 0.68463516,
+      "epoch": 1.0768645357686453,
+      "grad_norm": 3.125,
+      "learning_rate": 4.777974534945677e-06,
+      "loss": 1.46228533,
+      "memory(GiB)": 123.79,
+      "step": 42450,
+      "train_speed(iter/s)": 1.316493
+    },
+    {
+      "acc": 0.68049431,
+      "epoch": 1.0769913749365805,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.776926950511457e-06,
+      "loss": 1.46029959,
+      "memory(GiB)": 123.79,
+      "step": 42455,
+      "train_speed(iter/s)": 1.31651
+    },
+    {
+      "acc": 0.66977673,
+      "epoch": 1.0771182141045155,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.775879375888986e-06,
+      "loss": 1.46528234,
+      "memory(GiB)": 123.79,
+      "step": 42460,
+      "train_speed(iter/s)": 1.316526
+    },
+    {
+      "acc": 0.69246454,
+      "epoch": 1.0772450532724505,
+      "grad_norm": 3.0,
+      "learning_rate": 4.774831811124343e-06,
+      "loss": 1.38221388,
+      "memory(GiB)": 123.79,
+      "step": 42465,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.68684196,
+      "epoch": 1.0773718924403857,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.773784256263601e-06,
+      "loss": 1.37614889,
+      "memory(GiB)": 123.79,
+      "step": 42470,
+      "train_speed(iter/s)": 1.316561
+    },
+    {
+      "acc": 0.67734318,
+      "epoch": 1.0774987316083207,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.7727367113528374e-06,
+      "loss": 1.40535355,
+      "memory(GiB)": 123.79,
+      "step": 42475,
+      "train_speed(iter/s)": 1.316577
+    },
+    {
+      "acc": 0.68214269,
+      "epoch": 1.0776255707762556,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.771689176438128e-06,
+      "loss": 1.43583002,
+      "memory(GiB)": 123.79,
+      "step": 42480,
+      "train_speed(iter/s)": 1.316595
+    },
+    {
+      "acc": 0.68837304,
+      "epoch": 1.0777524099441909,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.770641651565546e-06,
+      "loss": 1.3353426,
+      "memory(GiB)": 123.79,
+      "step": 42485,
+      "train_speed(iter/s)": 1.31661
+    },
+    {
+      "acc": 0.69648018,
+      "epoch": 1.0778792491121258,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.769594136781172e-06,
+      "loss": 1.35560322,
+      "memory(GiB)": 123.79,
+      "step": 42490,
+      "train_speed(iter/s)": 1.316627
+    },
+    {
+      "acc": 0.70075865,
+      "epoch": 1.0780060882800608,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.768546632131074e-06,
+      "loss": 1.36621323,
+      "memory(GiB)": 123.79,
+      "step": 42495,
+      "train_speed(iter/s)": 1.316645
+    },
+    {
+      "acc": 0.68458843,
+      "epoch": 1.0781329274479958,
+      "grad_norm": 4.0,
+      "learning_rate": 4.767499137661328e-06,
+      "loss": 1.44493351,
+      "memory(GiB)": 123.79,
+      "step": 42500,
+      "train_speed(iter/s)": 1.31666
+    },
+    {
+      "acc": 0.69516973,
+      "epoch": 1.078259766615931,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.76645165341801e-06,
+      "loss": 1.30206194,
+      "memory(GiB)": 123.79,
+      "step": 42505,
+      "train_speed(iter/s)": 1.316677
+    },
+    {
+      "acc": 0.67547431,
+      "epoch": 1.078386605783866,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.76540417944719e-06,
+      "loss": 1.37526951,
+      "memory(GiB)": 123.79,
+      "step": 42510,
+      "train_speed(iter/s)": 1.316694
+    },
+    {
+      "acc": 0.69062123,
+      "epoch": 1.078513444951801,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.764356715794942e-06,
+      "loss": 1.39958515,
+      "memory(GiB)": 123.79,
+      "step": 42515,
+      "train_speed(iter/s)": 1.316711
+    },
+    {
+      "acc": 0.69079418,
+      "epoch": 1.0786402841197362,
+      "grad_norm": 4.6875,
+      "learning_rate": 4.763309262507336e-06,
+      "loss": 1.32971821,
+      "memory(GiB)": 123.79,
+      "step": 42520,
+      "train_speed(iter/s)": 1.316728
+    },
+    {
+      "acc": 0.6847609,
+      "epoch": 1.0787671232876712,
+      "grad_norm": 3.75,
+      "learning_rate": 4.762261819630447e-06,
+      "loss": 1.39856691,
+      "memory(GiB)": 123.79,
+      "step": 42525,
+      "train_speed(iter/s)": 1.316744
+    },
+    {
+      "acc": 0.69595184,
+      "epoch": 1.0788939624556062,
+      "grad_norm": 3.125,
+      "learning_rate": 4.761214387210345e-06,
+      "loss": 1.33512564,
+      "memory(GiB)": 123.79,
+      "step": 42530,
+      "train_speed(iter/s)": 1.316761
+    },
+    {
+      "acc": 0.7015461,
+      "epoch": 1.0790208016235414,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.760166965293099e-06,
+      "loss": 1.38159447,
+      "memory(GiB)": 123.79,
+      "step": 42535,
+      "train_speed(iter/s)": 1.316778
+    },
+    {
+      "acc": 0.69502969,
+      "epoch": 1.0791476407914764,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.759119553924781e-06,
+      "loss": 1.35403824,
+      "memory(GiB)": 123.79,
+      "step": 42540,
+      "train_speed(iter/s)": 1.316795
+    },
+    {
+      "acc": 0.68224344,
+      "epoch": 1.0792744799594114,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.758072153151461e-06,
+      "loss": 1.35558882,
+      "memory(GiB)": 123.79,
+      "step": 42545,
+      "train_speed(iter/s)": 1.316812
+    },
+    {
+      "acc": 0.68895092,
+      "epoch": 1.0794013191273466,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.757024763019209e-06,
+      "loss": 1.3337842,
+      "memory(GiB)": 123.79,
+      "step": 42550,
+      "train_speed(iter/s)": 1.316828
+    },
+    {
+      "acc": 0.68779244,
+      "epoch": 1.0795281582952816,
+      "grad_norm": 2.75,
+      "learning_rate": 4.755977383574091e-06,
+      "loss": 1.36512375,
+      "memory(GiB)": 123.79,
+      "step": 42555,
+      "train_speed(iter/s)": 1.316845
+    },
+    {
+      "acc": 0.69605393,
+      "epoch": 1.0796549974632166,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.754930014862177e-06,
+      "loss": 1.37745457,
+      "memory(GiB)": 123.79,
+      "step": 42560,
+      "train_speed(iter/s)": 1.316862
+    },
+    {
+      "acc": 0.68945141,
+      "epoch": 1.0797818366311518,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.753882656929535e-06,
+      "loss": 1.45165501,
+      "memory(GiB)": 123.79,
+      "step": 42565,
+      "train_speed(iter/s)": 1.31688
+    },
+    {
+      "acc": 0.68327026,
+      "epoch": 1.0799086757990868,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.752835309822234e-06,
+      "loss": 1.3917881,
+      "memory(GiB)": 123.79,
+      "step": 42570,
+      "train_speed(iter/s)": 1.316898
+    },
+    {
+      "acc": 0.69473333,
+      "epoch": 1.0800355149670218,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.7517879735863385e-06,
+      "loss": 1.36426401,
+      "memory(GiB)": 123.79,
+      "step": 42575,
+      "train_speed(iter/s)": 1.316915
+    },
+    {
+      "acc": 0.67630882,
+      "epoch": 1.080162354134957,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.750740648267916e-06,
+      "loss": 1.42393932,
+      "memory(GiB)": 123.79,
+      "step": 42580,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.68514676,
+      "epoch": 1.080289193302892,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.749693333913033e-06,
+      "loss": 1.39680357,
+      "memory(GiB)": 123.79,
+      "step": 42585,
+      "train_speed(iter/s)": 1.316949
+    },
+    {
+      "acc": 0.67839775,
+      "epoch": 1.080416032470827,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.748646030567755e-06,
+      "loss": 1.40767689,
+      "memory(GiB)": 123.79,
+      "step": 42590,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.6924612,
+      "epoch": 1.080542871638762,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.747598738278147e-06,
+      "loss": 1.39993401,
+      "memory(GiB)": 123.79,
+      "step": 42595,
+      "train_speed(iter/s)": 1.316984
+    },
+    {
+      "acc": 0.69142342,
+      "epoch": 1.0806697108066972,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.746551457090272e-06,
+      "loss": 1.38244314,
+      "memory(GiB)": 123.79,
+      "step": 42600,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.68312206,
+      "epoch": 1.0807965499746321,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.745504187050197e-06,
+      "loss": 1.44825115,
+      "memory(GiB)": 123.79,
+      "step": 42605,
+      "train_speed(iter/s)": 1.317019
+    },
+    {
+      "acc": 0.67951312,
+      "epoch": 1.0809233891425671,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.744456928203985e-06,
+      "loss": 1.41487417,
+      "memory(GiB)": 123.79,
+      "step": 42610,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.68640766,
+      "epoch": 1.0810502283105023,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.743409680597695e-06,
+      "loss": 1.38780117,
+      "memory(GiB)": 123.79,
+      "step": 42615,
+      "train_speed(iter/s)": 1.317052
+    },
+    {
+      "acc": 0.67620101,
+      "epoch": 1.0811770674784373,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.742362444277394e-06,
+      "loss": 1.46063633,
+      "memory(GiB)": 123.79,
+      "step": 42620,
+      "train_speed(iter/s)": 1.317069
+    },
+    {
+      "acc": 0.67970695,
+      "epoch": 1.0813039066463723,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.741315219289142e-06,
+      "loss": 1.42852211,
+      "memory(GiB)": 123.79,
+      "step": 42625,
+      "train_speed(iter/s)": 1.317086
+    },
+    {
+      "acc": 0.69385967,
+      "epoch": 1.0814307458143075,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.740268005679005e-06,
+      "loss": 1.35840206,
+      "memory(GiB)": 123.79,
+      "step": 42630,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.68850541,
+      "epoch": 1.0815575849822425,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.739220803493039e-06,
+      "loss": 1.34447002,
+      "memory(GiB)": 123.79,
+      "step": 42635,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.69620056,
+      "epoch": 1.0816844241501775,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.738173612777306e-06,
+      "loss": 1.41594715,
+      "memory(GiB)": 123.79,
+      "step": 42640,
+      "train_speed(iter/s)": 1.317137
+    },
+    {
+      "acc": 0.6708941,
+      "epoch": 1.0818112633181127,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.737126433577866e-06,
+      "loss": 1.47990789,
+      "memory(GiB)": 123.79,
+      "step": 42645,
+      "train_speed(iter/s)": 1.317154
+    },
+    {
+      "acc": 0.67808456,
+      "epoch": 1.0819381024860477,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.736079265940781e-06,
+      "loss": 1.42484512,
+      "memory(GiB)": 123.79,
+      "step": 42650,
+      "train_speed(iter/s)": 1.317173
+    },
+    {
+      "acc": 0.67629209,
+      "epoch": 1.0820649416539827,
+      "grad_norm": 3.75,
+      "learning_rate": 4.735032109912107e-06,
+      "loss": 1.4311677,
+      "memory(GiB)": 123.79,
+      "step": 42655,
+      "train_speed(iter/s)": 1.31719
+    },
+    {
+      "acc": 0.68039446,
+      "epoch": 1.0821917808219177,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.733984965537903e-06,
+      "loss": 1.37970085,
+      "memory(GiB)": 123.79,
+      "step": 42660,
+      "train_speed(iter/s)": 1.317205
+    },
+    {
+      "acc": 0.70256791,
+      "epoch": 1.082318619989853,
+      "grad_norm": 3.375,
+      "learning_rate": 4.732937832864229e-06,
+      "loss": 1.32368832,
+      "memory(GiB)": 123.79,
+      "step": 42665,
+      "train_speed(iter/s)": 1.317223
+    },
+    {
+      "acc": 0.67144699,
+      "epoch": 1.082445459157788,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.731890711937141e-06,
+      "loss": 1.43388004,
+      "memory(GiB)": 123.79,
+      "step": 42670,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.6919075,
+      "epoch": 1.0825722983257229,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.730843602802696e-06,
+      "loss": 1.37166119,
+      "memory(GiB)": 123.79,
+      "step": 42675,
+      "train_speed(iter/s)": 1.317257
+    },
+    {
+      "acc": 0.68969269,
+      "epoch": 1.082699137493658,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.729796505506951e-06,
+      "loss": 1.37844276,
+      "memory(GiB)": 123.79,
+      "step": 42680,
+      "train_speed(iter/s)": 1.317275
+    },
+    {
+      "acc": 0.67372146,
+      "epoch": 1.082825976661593,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.728749420095964e-06,
+      "loss": 1.42386703,
+      "memory(GiB)": 123.79,
+      "step": 42685,
+      "train_speed(iter/s)": 1.317292
+    },
+    {
+      "acc": 0.68796721,
+      "epoch": 1.082952815829528,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.727702346615788e-06,
+      "loss": 1.43039217,
+      "memory(GiB)": 123.79,
+      "step": 42690,
+      "train_speed(iter/s)": 1.317309
+    },
+    {
+      "acc": 0.69520521,
+      "epoch": 1.0830796549974633,
+      "grad_norm": 2.5,
+      "learning_rate": 4.726655285112477e-06,
+      "loss": 1.35257931,
+      "memory(GiB)": 123.79,
+      "step": 42695,
+      "train_speed(iter/s)": 1.317327
+    },
+    {
+      "acc": 0.68308682,
+      "epoch": 1.0832064941653983,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.725608235632088e-06,
+      "loss": 1.38610973,
+      "memory(GiB)": 123.79,
+      "step": 42700,
+      "train_speed(iter/s)": 1.317345
+    },
+    {
+      "acc": 0.68843455,
+      "epoch": 1.0833333333333333,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.724561198220672e-06,
+      "loss": 1.38057089,
+      "memory(GiB)": 123.79,
+      "step": 42705,
+      "train_speed(iter/s)": 1.317362
+    },
+    {
+      "acc": 0.67582169,
+      "epoch": 1.0834601725012685,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.723514172924287e-06,
+      "loss": 1.43184357,
+      "memory(GiB)": 123.79,
+      "step": 42710,
+      "train_speed(iter/s)": 1.317379
+    },
+    {
+      "acc": 0.67852707,
+      "epoch": 1.0835870116692035,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.7224671597889825e-06,
+      "loss": 1.40793133,
+      "memory(GiB)": 123.79,
+      "step": 42715,
+      "train_speed(iter/s)": 1.317398
+    },
+    {
+      "acc": 0.67525358,
+      "epoch": 1.0837138508371384,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.72142015886081e-06,
+      "loss": 1.44285555,
+      "memory(GiB)": 123.79,
+      "step": 42720,
+      "train_speed(iter/s)": 1.317416
+    },
+    {
+      "acc": 0.70395436,
+      "epoch": 1.0838406900050737,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.720373170185823e-06,
+      "loss": 1.342171,
+      "memory(GiB)": 123.79,
+      "step": 42725,
+      "train_speed(iter/s)": 1.317433
+    },
+    {
+      "acc": 0.68915534,
+      "epoch": 1.0839675291730086,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.719326193810075e-06,
+      "loss": 1.43899832,
+      "memory(GiB)": 123.79,
+      "step": 42730,
+      "train_speed(iter/s)": 1.317451
+    },
+    {
+      "acc": 0.673803,
+      "epoch": 1.0840943683409436,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.718279229779612e-06,
+      "loss": 1.46628084,
+      "memory(GiB)": 123.79,
+      "step": 42735,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.68129282,
+      "epoch": 1.0842212075088788,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.717232278140485e-06,
+      "loss": 1.4617795,
+      "memory(GiB)": 123.79,
+      "step": 42740,
+      "train_speed(iter/s)": 1.317485
+    },
+    {
+      "acc": 0.69888482,
+      "epoch": 1.0843480466768138,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.716185338938746e-06,
+      "loss": 1.32747421,
+      "memory(GiB)": 123.79,
+      "step": 42745,
+      "train_speed(iter/s)": 1.317502
+    },
+    {
+      "acc": 0.67816439,
+      "epoch": 1.0844748858447488,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.7151384122204445e-06,
+      "loss": 1.34731121,
+      "memory(GiB)": 123.79,
+      "step": 42750,
+      "train_speed(iter/s)": 1.317519
+    },
+    {
+      "acc": 0.68319616,
+      "epoch": 1.0846017250126838,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.7140914980316254e-06,
+      "loss": 1.36231422,
+      "memory(GiB)": 123.79,
+      "step": 42755,
+      "train_speed(iter/s)": 1.317536
+    },
+    {
+      "acc": 0.68563704,
+      "epoch": 1.084728564180619,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.713044596418339e-06,
+      "loss": 1.41397028,
+      "memory(GiB)": 123.79,
+      "step": 42760,
+      "train_speed(iter/s)": 1.317554
+    },
+    {
+      "acc": 0.68861275,
+      "epoch": 1.084855403348554,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.711997707426632e-06,
+      "loss": 1.40008221,
+      "memory(GiB)": 123.79,
+      "step": 42765,
+      "train_speed(iter/s)": 1.317571
+    },
+    {
+      "acc": 0.68259416,
+      "epoch": 1.084982242516489,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.710950831102555e-06,
+      "loss": 1.39403925,
+      "memory(GiB)": 123.79,
+      "step": 42770,
+      "train_speed(iter/s)": 1.317588
+    },
+    {
+      "acc": 0.68933749,
+      "epoch": 1.0851090816844242,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.709903967492147e-06,
+      "loss": 1.37057266,
+      "memory(GiB)": 123.79,
+      "step": 42775,
+      "train_speed(iter/s)": 1.317606
+    },
+    {
+      "acc": 0.68106589,
+      "epoch": 1.0852359208523592,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.7088571166414595e-06,
+      "loss": 1.38777847,
+      "memory(GiB)": 123.79,
+      "step": 42780,
+      "train_speed(iter/s)": 1.317624
+    },
+    {
+      "acc": 0.68706665,
+      "epoch": 1.0853627600202942,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.707810278596534e-06,
+      "loss": 1.40479031,
+      "memory(GiB)": 123.79,
+      "step": 42785,
+      "train_speed(iter/s)": 1.317641
+    },
+    {
+      "acc": 0.69547033,
+      "epoch": 1.0854895991882294,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.7067634534034205e-06,
+      "loss": 1.39094429,
+      "memory(GiB)": 123.79,
+      "step": 42790,
+      "train_speed(iter/s)": 1.317658
+    },
+    {
+      "acc": 0.68483281,
+      "epoch": 1.0856164383561644,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.705716641108157e-06,
+      "loss": 1.36829271,
+      "memory(GiB)": 123.79,
+      "step": 42795,
+      "train_speed(iter/s)": 1.317677
+    },
+    {
+      "acc": 0.68141479,
+      "epoch": 1.0857432775240994,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.7046698417567894e-06,
+      "loss": 1.4182476,
+      "memory(GiB)": 123.79,
+      "step": 42800,
+      "train_speed(iter/s)": 1.317694
+    },
+    {
+      "acc": 0.68845444,
+      "epoch": 1.0858701166920346,
+      "grad_norm": 3.375,
+      "learning_rate": 4.7036230553953616e-06,
+      "loss": 1.4054081,
+      "memory(GiB)": 123.79,
+      "step": 42805,
+      "train_speed(iter/s)": 1.317712
+    },
+    {
+      "acc": 0.69734559,
+      "epoch": 1.0859969558599696,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.702576282069916e-06,
+      "loss": 1.31521292,
+      "memory(GiB)": 123.79,
+      "step": 42810,
+      "train_speed(iter/s)": 1.317729
+    },
+    {
+      "acc": 0.67501259,
+      "epoch": 1.0861237950279046,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.701529521826492e-06,
+      "loss": 1.43529577,
+      "memory(GiB)": 123.79,
+      "step": 42815,
+      "train_speed(iter/s)": 1.317747
+    },
+    {
+      "acc": 0.70172606,
+      "epoch": 1.0862506341958396,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.700482774711131e-06,
+      "loss": 1.38360386,
+      "memory(GiB)": 123.79,
+      "step": 42820,
+      "train_speed(iter/s)": 1.317764
+    },
+    {
+      "acc": 0.69844122,
+      "epoch": 1.0863774733637748,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.699436040769877e-06,
+      "loss": 1.30841217,
+      "memory(GiB)": 123.79,
+      "step": 42825,
+      "train_speed(iter/s)": 1.317782
+    },
+    {
+      "acc": 0.69446955,
+      "epoch": 1.0865043125317098,
+      "grad_norm": 3.375,
+      "learning_rate": 4.698389320048768e-06,
+      "loss": 1.38715563,
+      "memory(GiB)": 123.79,
+      "step": 42830,
+      "train_speed(iter/s)": 1.317798
+    },
+    {
+      "acc": 0.67416286,
+      "epoch": 1.0866311516996447,
+      "grad_norm": 3.125,
+      "learning_rate": 4.697342612593841e-06,
+      "loss": 1.4069355,
+      "memory(GiB)": 123.79,
+      "step": 42835,
+      "train_speed(iter/s)": 1.317815
+    },
+    {
+      "acc": 0.68583288,
+      "epoch": 1.08675799086758,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.696295918451139e-06,
+      "loss": 1.36150532,
+      "memory(GiB)": 123.79,
+      "step": 42840,
+      "train_speed(iter/s)": 1.317832
+    },
+    {
+      "acc": 0.69299574,
+      "epoch": 1.086884830035515,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.695249237666697e-06,
+      "loss": 1.36075935,
+      "memory(GiB)": 123.79,
+      "step": 42845,
+      "train_speed(iter/s)": 1.317849
+    },
+    {
+      "acc": 0.7107996,
+      "epoch": 1.08701166920345,
+      "grad_norm": 3.25,
+      "learning_rate": 4.694202570286556e-06,
+      "loss": 1.32748814,
+      "memory(GiB)": 123.79,
+      "step": 42850,
+      "train_speed(iter/s)": 1.317864
+    },
+    {
+      "acc": 0.68339343,
+      "epoch": 1.0871385083713851,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.693155916356751e-06,
+      "loss": 1.4299139,
+      "memory(GiB)": 123.79,
+      "step": 42855,
+      "train_speed(iter/s)": 1.317881
+    },
+    {
+      "acc": 0.66999502,
+      "epoch": 1.0872653475393201,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.692109275923318e-06,
+      "loss": 1.47569389,
+      "memory(GiB)": 123.79,
+      "step": 42860,
+      "train_speed(iter/s)": 1.317898
+    },
+    {
+      "acc": 0.6892971,
+      "epoch": 1.0873921867072551,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.6910626490322925e-06,
+      "loss": 1.43878222,
+      "memory(GiB)": 123.79,
+      "step": 42865,
+      "train_speed(iter/s)": 1.317915
+    },
+    {
+      "acc": 0.69360399,
+      "epoch": 1.0875190258751903,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.690016035729714e-06,
+      "loss": 1.35979347,
+      "memory(GiB)": 123.79,
+      "step": 42870,
+      "train_speed(iter/s)": 1.317932
+    },
+    {
+      "acc": 0.68398943,
+      "epoch": 1.0876458650431253,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.688969436061612e-06,
+      "loss": 1.41372738,
+      "memory(GiB)": 123.79,
+      "step": 42875,
+      "train_speed(iter/s)": 1.31795
+    },
+    {
+      "acc": 0.69025979,
+      "epoch": 1.0877727042110603,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.687922850074022e-06,
+      "loss": 1.38052855,
+      "memory(GiB)": 123.79,
+      "step": 42880,
+      "train_speed(iter/s)": 1.317967
+    },
+    {
+      "acc": 0.68344908,
+      "epoch": 1.0878995433789955,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.686876277812981e-06,
+      "loss": 1.4556097,
+      "memory(GiB)": 123.79,
+      "step": 42885,
+      "train_speed(iter/s)": 1.317983
+    },
+    {
+      "acc": 0.67742691,
+      "epoch": 1.0880263825469305,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.685829719324519e-06,
+      "loss": 1.45342102,
+      "memory(GiB)": 123.79,
+      "step": 42890,
+      "train_speed(iter/s)": 1.318001
+    },
+    {
+      "acc": 0.67971911,
+      "epoch": 1.0881532217148655,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.6847831746546664e-06,
+      "loss": 1.47444782,
+      "memory(GiB)": 123.79,
+      "step": 42895,
+      "train_speed(iter/s)": 1.318017
+    },
+    {
+      "acc": 0.69088893,
+      "epoch": 1.0882800608828007,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.683736643849459e-06,
+      "loss": 1.35790386,
+      "memory(GiB)": 123.79,
+      "step": 42900,
+      "train_speed(iter/s)": 1.318035
+    },
+    {
+      "acc": 0.68901606,
+      "epoch": 1.0884069000507357,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.6826901269549255e-06,
+      "loss": 1.33982983,
+      "memory(GiB)": 123.79,
+      "step": 42905,
+      "train_speed(iter/s)": 1.318049
+    },
+    {
+      "acc": 0.68193445,
+      "epoch": 1.0885337392186707,
+      "grad_norm": 3.96875,
+      "learning_rate": 4.681643624017097e-06,
+      "loss": 1.39890175,
+      "memory(GiB)": 123.79,
+      "step": 42910,
+      "train_speed(iter/s)": 1.318066
+    },
+    {
+      "acc": 0.70666323,
+      "epoch": 1.0886605783866057,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.680597135082002e-06,
+      "loss": 1.29064798,
+      "memory(GiB)": 123.79,
+      "step": 42915,
+      "train_speed(iter/s)": 1.318083
+    },
+    {
+      "acc": 0.71300535,
+      "epoch": 1.088787417554541,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.679550660195673e-06,
+      "loss": 1.33591394,
+      "memory(GiB)": 123.79,
+      "step": 42920,
+      "train_speed(iter/s)": 1.3181
+    },
+    {
+      "acc": 0.68157687,
+      "epoch": 1.0889142567224759,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.6785041994041345e-06,
+      "loss": 1.41393461,
+      "memory(GiB)": 123.79,
+      "step": 42925,
+      "train_speed(iter/s)": 1.318117
+    },
+    {
+      "acc": 0.68506231,
+      "epoch": 1.0890410958904109,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.6774577527534195e-06,
+      "loss": 1.41665659,
+      "memory(GiB)": 123.79,
+      "step": 42930,
+      "train_speed(iter/s)": 1.318135
+    },
+    {
+      "acc": 0.69177814,
+      "epoch": 1.089167935058346,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.676411320289551e-06,
+      "loss": 1.4267252,
+      "memory(GiB)": 123.79,
+      "step": 42935,
+      "train_speed(iter/s)": 1.318152
+    },
+    {
+      "acc": 0.68193598,
+      "epoch": 1.089294774226281,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.675364902058556e-06,
+      "loss": 1.40140858,
+      "memory(GiB)": 123.79,
+      "step": 42940,
+      "train_speed(iter/s)": 1.318169
+    },
+    {
+      "acc": 0.71038647,
+      "epoch": 1.089421613394216,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.674318498106464e-06,
+      "loss": 1.40072784,
+      "memory(GiB)": 123.79,
+      "step": 42945,
+      "train_speed(iter/s)": 1.318187
+    },
+    {
+      "acc": 0.70206661,
+      "epoch": 1.0895484525621513,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.6732721084792985e-06,
+      "loss": 1.37314663,
+      "memory(GiB)": 123.79,
+      "step": 42950,
+      "train_speed(iter/s)": 1.318204
+    },
+    {
+      "acc": 0.67672391,
+      "epoch": 1.0896752917300863,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.672225733223084e-06,
+      "loss": 1.44547405,
+      "memory(GiB)": 123.79,
+      "step": 42955,
+      "train_speed(iter/s)": 1.318222
+    },
+    {
+      "acc": 0.70006781,
+      "epoch": 1.0898021308980212,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.671179372383844e-06,
+      "loss": 1.33367805,
+      "memory(GiB)": 123.79,
+      "step": 42960,
+      "train_speed(iter/s)": 1.318239
+    },
+    {
+      "acc": 0.68583889,
+      "epoch": 1.0899289700659565,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.670133026007604e-06,
+      "loss": 1.43693428,
+      "memory(GiB)": 123.79,
+      "step": 42965,
+      "train_speed(iter/s)": 1.318257
+    },
+    {
+      "acc": 0.68949738,
+      "epoch": 1.0900558092338914,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.669086694140388e-06,
+      "loss": 1.43386517,
+      "memory(GiB)": 123.79,
+      "step": 42970,
+      "train_speed(iter/s)": 1.318274
+    },
+    {
+      "acc": 0.68712525,
+      "epoch": 1.0901826484018264,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.668040376828214e-06,
+      "loss": 1.36964417,
+      "memory(GiB)": 123.79,
+      "step": 42975,
+      "train_speed(iter/s)": 1.318291
+    },
+    {
+      "acc": 0.69967537,
+      "epoch": 1.0903094875697614,
+      "grad_norm": 3.125,
+      "learning_rate": 4.666994074117108e-06,
+      "loss": 1.39120979,
+      "memory(GiB)": 123.79,
+      "step": 42980,
+      "train_speed(iter/s)": 1.318308
+    },
+    {
+      "acc": 0.67899771,
+      "epoch": 1.0904363267376966,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.665947786053088e-06,
+      "loss": 1.44196968,
+      "memory(GiB)": 123.79,
+      "step": 42985,
+      "train_speed(iter/s)": 1.318325
+    },
+    {
+      "acc": 0.67910986,
+      "epoch": 1.0905631659056316,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.664901512682179e-06,
+      "loss": 1.42101927,
+      "memory(GiB)": 123.79,
+      "step": 42990,
+      "train_speed(iter/s)": 1.318342
+    },
+    {
+      "acc": 0.68783197,
+      "epoch": 1.0906900050735666,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.663855254050394e-06,
+      "loss": 1.39759712,
+      "memory(GiB)": 123.79,
+      "step": 42995,
+      "train_speed(iter/s)": 1.318357
+    },
+    {
+      "acc": 0.67899008,
+      "epoch": 1.0908168442415018,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.662809010203757e-06,
+      "loss": 1.44015284,
+      "memory(GiB)": 123.79,
+      "step": 43000,
+      "train_speed(iter/s)": 1.318374
+    },
+    {
+      "epoch": 1.0908168442415018,
+      "eval_acc": 0.6750015140219314,
+      "eval_loss": 1.3567273616790771,
+      "eval_runtime": 69.6849,
+      "eval_samples_per_second": 91.412,
+      "eval_steps_per_second": 22.86,
+      "step": 43000
+    },
+    {
+      "acc": 0.68163443,
+      "epoch": 1.0909436834094368,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.661762781188284e-06,
+      "loss": 1.40870199,
+      "memory(GiB)": 123.79,
+      "step": 43005,
+      "train_speed(iter/s)": 1.315072
+    },
+    {
+      "acc": 0.6935338,
+      "epoch": 1.0910705225773718,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.660716567049997e-06,
+      "loss": 1.38799047,
+      "memory(GiB)": 123.79,
+      "step": 43010,
+      "train_speed(iter/s)": 1.31509
+    },
+    {
+      "acc": 0.68605108,
+      "epoch": 1.091197361745307,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.659670367834908e-06,
+      "loss": 1.39398613,
+      "memory(GiB)": 123.79,
+      "step": 43015,
+      "train_speed(iter/s)": 1.315106
+    },
+    {
+      "acc": 0.68351889,
+      "epoch": 1.091324200913242,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.658624183589035e-06,
+      "loss": 1.38760004,
+      "memory(GiB)": 123.79,
+      "step": 43020,
+      "train_speed(iter/s)": 1.315122
+    },
+    {
+      "acc": 0.7031178,
+      "epoch": 1.091451040081177,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.657578014358395e-06,
+      "loss": 1.35219917,
+      "memory(GiB)": 123.79,
+      "step": 43025,
+      "train_speed(iter/s)": 1.315138
+    },
+    {
+      "acc": 0.67694273,
+      "epoch": 1.0915778792491122,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.656531860189005e-06,
+      "loss": 1.37920551,
+      "memory(GiB)": 123.79,
+      "step": 43030,
+      "train_speed(iter/s)": 1.315155
+    },
+    {
+      "acc": 0.6837676,
+      "epoch": 1.0917047184170472,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.655485721126875e-06,
+      "loss": 1.385886,
+      "memory(GiB)": 123.79,
+      "step": 43035,
+      "train_speed(iter/s)": 1.315172
+    },
+    {
+      "acc": 0.67788887,
+      "epoch": 1.0918315575849822,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.6544395972180214e-06,
+      "loss": 1.44180183,
+      "memory(GiB)": 123.79,
+      "step": 43040,
+      "train_speed(iter/s)": 1.315189
+    },
+    {
+      "acc": 0.67581105,
+      "epoch": 1.0919583967529174,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.653393488508457e-06,
+      "loss": 1.43165169,
+      "memory(GiB)": 123.79,
+      "step": 43045,
+      "train_speed(iter/s)": 1.315206
+    },
+    {
+      "acc": 0.67934752,
+      "epoch": 1.0920852359208524,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.652347395044197e-06,
+      "loss": 1.38625898,
+      "memory(GiB)": 123.79,
+      "step": 43050,
+      "train_speed(iter/s)": 1.315223
+    },
+    {
+      "acc": 0.69005504,
+      "epoch": 1.0922120750887874,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.651301316871247e-06,
+      "loss": 1.41944141,
+      "memory(GiB)": 123.79,
+      "step": 43055,
+      "train_speed(iter/s)": 1.315238
+    },
+    {
+      "acc": 0.68801484,
+      "epoch": 1.0923389142567226,
+      "grad_norm": 2.875,
+      "learning_rate": 4.6502552540356235e-06,
+      "loss": 1.38958111,
+      "memory(GiB)": 123.79,
+      "step": 43060,
+      "train_speed(iter/s)": 1.315257
+    },
+    {
+      "acc": 0.6869771,
+      "epoch": 1.0924657534246576,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.649209206583335e-06,
+      "loss": 1.42057734,
+      "memory(GiB)": 123.79,
+      "step": 43065,
+      "train_speed(iter/s)": 1.315274
+    },
+    {
+      "acc": 0.70641165,
+      "epoch": 1.0925925925925926,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.648163174560393e-06,
+      "loss": 1.36177921,
+      "memory(GiB)": 123.79,
+      "step": 43070,
+      "train_speed(iter/s)": 1.315284
+    },
+    {
+      "acc": 0.69267311,
+      "epoch": 1.0927194317605275,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.647117158012804e-06,
+      "loss": 1.37502003,
+      "memory(GiB)": 123.79,
+      "step": 43075,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.68411245,
+      "epoch": 1.0928462709284628,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.646071156986579e-06,
+      "loss": 1.35955772,
+      "memory(GiB)": 123.79,
+      "step": 43080,
+      "train_speed(iter/s)": 1.31532
+    },
+    {
+      "acc": 0.6745985,
+      "epoch": 1.0929731100963977,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.645025171527723e-06,
+      "loss": 1.42369051,
+      "memory(GiB)": 123.79,
+      "step": 43085,
+      "train_speed(iter/s)": 1.315334
+    },
+    {
+      "acc": 0.67613497,
+      "epoch": 1.0930999492643327,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.643979201682247e-06,
+      "loss": 1.40878468,
+      "memory(GiB)": 123.79,
+      "step": 43090,
+      "train_speed(iter/s)": 1.315351
+    },
+    {
+      "acc": 0.68196006,
+      "epoch": 1.093226788432268,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.642933247496155e-06,
+      "loss": 1.44008293,
+      "memory(GiB)": 123.79,
+      "step": 43095,
+      "train_speed(iter/s)": 1.315368
+    },
+    {
+      "acc": 0.68276339,
+      "epoch": 1.093353627600203,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.641887309015451e-06,
+      "loss": 1.43291302,
+      "memory(GiB)": 123.79,
+      "step": 43100,
+      "train_speed(iter/s)": 1.315385
+    },
+    {
+      "acc": 0.68750734,
+      "epoch": 1.093480466768138,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.640841386286143e-06,
+      "loss": 1.31452026,
+      "memory(GiB)": 123.79,
+      "step": 43105,
+      "train_speed(iter/s)": 1.315401
+    },
+    {
+      "acc": 0.68300862,
+      "epoch": 1.0936073059360731,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.639795479354236e-06,
+      "loss": 1.45585079,
+      "memory(GiB)": 123.79,
+      "step": 43110,
+      "train_speed(iter/s)": 1.315418
+    },
+    {
+      "acc": 0.69914045,
+      "epoch": 1.0937341451040081,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.6387495882657295e-06,
+      "loss": 1.40214853,
+      "memory(GiB)": 123.79,
+      "step": 43115,
+      "train_speed(iter/s)": 1.315435
+    },
+    {
+      "acc": 0.68979836,
+      "epoch": 1.0938609842719431,
+      "grad_norm": 3.0,
+      "learning_rate": 4.63770371306663e-06,
+      "loss": 1.39486141,
+      "memory(GiB)": 123.79,
+      "step": 43120,
+      "train_speed(iter/s)": 1.315452
+    },
+    {
+      "acc": 0.68331642,
+      "epoch": 1.0939878234398783,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.636657853802939e-06,
+      "loss": 1.31884851,
+      "memory(GiB)": 123.79,
+      "step": 43125,
+      "train_speed(iter/s)": 1.31547
+    },
+    {
+      "acc": 0.70043612,
+      "epoch": 1.0941146626078133,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.635612010520659e-06,
+      "loss": 1.41459618,
+      "memory(GiB)": 123.79,
+      "step": 43130,
+      "train_speed(iter/s)": 1.315487
+    },
+    {
+      "acc": 0.68428421,
+      "epoch": 1.0942415017757483,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.6345661832657866e-06,
+      "loss": 1.42910614,
+      "memory(GiB)": 123.79,
+      "step": 43135,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.69174976,
+      "epoch": 1.0943683409436833,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.633520372084327e-06,
+      "loss": 1.3298418,
+      "memory(GiB)": 123.79,
+      "step": 43140,
+      "train_speed(iter/s)": 1.315522
+    },
+    {
+      "acc": 0.682864,
+      "epoch": 1.0944951801116185,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.632474577022276e-06,
+      "loss": 1.45385103,
+      "memory(GiB)": 123.79,
+      "step": 43145,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.70315804,
+      "epoch": 1.0946220192795535,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.631428798125637e-06,
+      "loss": 1.36685944,
+      "memory(GiB)": 123.79,
+      "step": 43150,
+      "train_speed(iter/s)": 1.315556
+    },
+    {
+      "acc": 0.69669476,
+      "epoch": 1.0947488584474885,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.630383035440403e-06,
+      "loss": 1.38966579,
+      "memory(GiB)": 123.79,
+      "step": 43155,
+      "train_speed(iter/s)": 1.315574
+    },
+    {
+      "acc": 0.69824491,
+      "epoch": 1.0948756976154237,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.6293372890125724e-06,
+      "loss": 1.3774972,
+      "memory(GiB)": 123.79,
+      "step": 43160,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.68061304,
+      "epoch": 1.0950025367833587,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.628291558888144e-06,
+      "loss": 1.41233501,
+      "memory(GiB)": 123.79,
+      "step": 43165,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.69193096,
+      "epoch": 1.0951293759512937,
+      "grad_norm": 3.25,
+      "learning_rate": 4.627245845113113e-06,
+      "loss": 1.36922436,
+      "memory(GiB)": 123.79,
+      "step": 43170,
+      "train_speed(iter/s)": 1.315624
+    },
+    {
+      "acc": 0.6842998,
+      "epoch": 1.0952562151192289,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.626200147733474e-06,
+      "loss": 1.383815,
+      "memory(GiB)": 123.79,
+      "step": 43175,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.68937826,
+      "epoch": 1.0953830542871639,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.62515446679522e-06,
+      "loss": 1.45051069,
+      "memory(GiB)": 123.79,
+      "step": 43180,
+      "train_speed(iter/s)": 1.315656
+    },
+    {
+      "acc": 0.70441041,
+      "epoch": 1.0955098934550989,
+      "grad_norm": 2.875,
+      "learning_rate": 4.624108802344347e-06,
+      "loss": 1.29286823,
+      "memory(GiB)": 123.79,
+      "step": 43185,
+      "train_speed(iter/s)": 1.315674
+    },
+    {
+      "acc": 0.68988733,
+      "epoch": 1.095636732623034,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.623063154426848e-06,
+      "loss": 1.37278051,
+      "memory(GiB)": 123.79,
+      "step": 43190,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.6834475,
+      "epoch": 1.095763571790969,
+      "grad_norm": 3.125,
+      "learning_rate": 4.622017523088712e-06,
+      "loss": 1.38256912,
+      "memory(GiB)": 123.79,
+      "step": 43195,
+      "train_speed(iter/s)": 1.315707
+    },
+    {
+      "acc": 0.68739614,
+      "epoch": 1.095890410958904,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.620971908375934e-06,
+      "loss": 1.4098341,
+      "memory(GiB)": 123.79,
+      "step": 43200,
+      "train_speed(iter/s)": 1.315724
+    },
+    {
+      "acc": 0.70007825,
+      "epoch": 1.0960172501268393,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.619926310334503e-06,
+      "loss": 1.33261108,
+      "memory(GiB)": 123.79,
+      "step": 43205,
+      "train_speed(iter/s)": 1.315741
+    },
+    {
+      "acc": 0.67725887,
+      "epoch": 1.0961440892947742,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.618880729010413e-06,
+      "loss": 1.40788937,
+      "memory(GiB)": 123.79,
+      "step": 43210,
+      "train_speed(iter/s)": 1.315759
+    },
+    {
+      "acc": 0.6943862,
+      "epoch": 1.0962709284627092,
+      "grad_norm": 3.125,
+      "learning_rate": 4.617835164449647e-06,
+      "loss": 1.33302364,
+      "memory(GiB)": 123.79,
+      "step": 43215,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.68836164,
+      "epoch": 1.0963977676306444,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.616789616698197e-06,
+      "loss": 1.37497625,
+      "memory(GiB)": 123.79,
+      "step": 43220,
+      "train_speed(iter/s)": 1.315793
+    },
+    {
+      "acc": 0.67531309,
+      "epoch": 1.0965246067985794,
+      "grad_norm": 4.0,
+      "learning_rate": 4.61574408580205e-06,
+      "loss": 1.43255405,
+      "memory(GiB)": 123.79,
+      "step": 43225,
+      "train_speed(iter/s)": 1.315811
+    },
+    {
+      "acc": 0.69852791,
+      "epoch": 1.0966514459665144,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.614698571807196e-06,
+      "loss": 1.32441187,
+      "memory(GiB)": 123.79,
+      "step": 43230,
+      "train_speed(iter/s)": 1.315827
+    },
+    {
+      "acc": 0.68399563,
+      "epoch": 1.0967782851344494,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.6136530747596185e-06,
+      "loss": 1.39903078,
+      "memory(GiB)": 123.79,
+      "step": 43235,
+      "train_speed(iter/s)": 1.315844
+    },
+    {
+      "acc": 0.68683119,
+      "epoch": 1.0969051243023846,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.612607594705301e-06,
+      "loss": 1.44141827,
+      "memory(GiB)": 123.79,
+      "step": 43240,
+      "train_speed(iter/s)": 1.315862
+    },
+    {
+      "acc": 0.69941864,
+      "epoch": 1.0970319634703196,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.611562131690234e-06,
+      "loss": 1.35287113,
+      "memory(GiB)": 123.79,
+      "step": 43245,
+      "train_speed(iter/s)": 1.315879
+    },
+    {
+      "acc": 0.68413754,
+      "epoch": 1.0971588026382546,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.610516685760399e-06,
+      "loss": 1.37313356,
+      "memory(GiB)": 123.79,
+      "step": 43250,
+      "train_speed(iter/s)": 1.315896
+    },
+    {
+      "acc": 0.70680618,
+      "epoch": 1.0972856418061898,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.6094712569617775e-06,
+      "loss": 1.30642681,
+      "memory(GiB)": 123.79,
+      "step": 43255,
+      "train_speed(iter/s)": 1.315912
+    },
+    {
+      "acc": 0.67966733,
+      "epoch": 1.0974124809741248,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.608425845340353e-06,
+      "loss": 1.40945444,
+      "memory(GiB)": 123.79,
+      "step": 43260,
+      "train_speed(iter/s)": 1.315929
+    },
+    {
+      "acc": 0.66709375,
+      "epoch": 1.0975393201420598,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.607380450942109e-06,
+      "loss": 1.45520115,
+      "memory(GiB)": 123.79,
+      "step": 43265,
+      "train_speed(iter/s)": 1.315947
+    },
+    {
+      "acc": 0.68417778,
+      "epoch": 1.097666159309995,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.606335073813028e-06,
+      "loss": 1.42919044,
+      "memory(GiB)": 123.79,
+      "step": 43270,
+      "train_speed(iter/s)": 1.315964
+    },
+    {
+      "acc": 0.67318268,
+      "epoch": 1.09779299847793,
+      "grad_norm": 2.75,
+      "learning_rate": 4.605289713999085e-06,
+      "loss": 1.4465064,
+      "memory(GiB)": 123.79,
+      "step": 43275,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.68536391,
+      "epoch": 1.097919837645865,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.604244371546263e-06,
+      "loss": 1.38008671,
+      "memory(GiB)": 123.79,
+      "step": 43280,
+      "train_speed(iter/s)": 1.315998
+    },
+    {
+      "acc": 0.67268829,
+      "epoch": 1.0980466768138002,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.603199046500539e-06,
+      "loss": 1.44361315,
+      "memory(GiB)": 123.79,
+      "step": 43285,
+      "train_speed(iter/s)": 1.316015
+    },
+    {
+      "acc": 0.6836082,
+      "epoch": 1.0981735159817352,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.602153738907896e-06,
+      "loss": 1.40556259,
+      "memory(GiB)": 123.79,
+      "step": 43290,
+      "train_speed(iter/s)": 1.316032
+    },
+    {
+      "acc": 0.69040613,
+      "epoch": 1.0983003551496702,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.601108448814306e-06,
+      "loss": 1.40061398,
+      "memory(GiB)": 123.79,
+      "step": 43295,
+      "train_speed(iter/s)": 1.316049
+    },
+    {
+      "acc": 0.68537288,
+      "epoch": 1.0984271943176052,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.600063176265749e-06,
+      "loss": 1.31802263,
+      "memory(GiB)": 123.79,
+      "step": 43300,
+      "train_speed(iter/s)": 1.316066
+    },
+    {
+      "acc": 0.69273658,
+      "epoch": 1.0985540334855404,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.599017921308196e-06,
+      "loss": 1.37483139,
+      "memory(GiB)": 123.79,
+      "step": 43305,
+      "train_speed(iter/s)": 1.316083
+    },
+    {
+      "acc": 0.6814846,
+      "epoch": 1.0986808726534754,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.5979726839876285e-06,
+      "loss": 1.39017715,
+      "memory(GiB)": 123.79,
+      "step": 43310,
+      "train_speed(iter/s)": 1.3161
+    },
+    {
+      "acc": 0.67421789,
+      "epoch": 1.0988077118214103,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.596927464350015e-06,
+      "loss": 1.42751007,
+      "memory(GiB)": 123.79,
+      "step": 43315,
+      "train_speed(iter/s)": 1.316117
+    },
+    {
+      "acc": 0.6743165,
+      "epoch": 1.0989345509893456,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.595882262441331e-06,
+      "loss": 1.38852367,
+      "memory(GiB)": 123.79,
+      "step": 43320,
+      "train_speed(iter/s)": 1.316134
+    },
+    {
+      "acc": 0.69160557,
+      "epoch": 1.0990613901572805,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.5948370783075505e-06,
+      "loss": 1.34607601,
+      "memory(GiB)": 123.79,
+      "step": 43325,
+      "train_speed(iter/s)": 1.31615
+    },
+    {
+      "acc": 0.68853517,
+      "epoch": 1.0991882293252155,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.5937919119946445e-06,
+      "loss": 1.41950302,
+      "memory(GiB)": 123.79,
+      "step": 43330,
+      "train_speed(iter/s)": 1.316167
+    },
+    {
+      "acc": 0.6980588,
+      "epoch": 1.0993150684931507,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.592746763548582e-06,
+      "loss": 1.34306173,
+      "memory(GiB)": 123.79,
+      "step": 43335,
+      "train_speed(iter/s)": 1.316184
+    },
+    {
+      "acc": 0.68498187,
+      "epoch": 1.0994419076610857,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.591701633015336e-06,
+      "loss": 1.40605049,
+      "memory(GiB)": 123.79,
+      "step": 43340,
+      "train_speed(iter/s)": 1.316201
+    },
+    {
+      "acc": 0.69063015,
+      "epoch": 1.0995687468290207,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.590656520440876e-06,
+      "loss": 1.41059551,
+      "memory(GiB)": 123.79,
+      "step": 43345,
+      "train_speed(iter/s)": 1.316218
+    },
+    {
+      "acc": 0.70950851,
+      "epoch": 1.099695585996956,
+      "grad_norm": 3.0,
+      "learning_rate": 4.58961142587117e-06,
+      "loss": 1.30228148,
+      "memory(GiB)": 123.79,
+      "step": 43350,
+      "train_speed(iter/s)": 1.316235
+    },
+    {
+      "acc": 0.68488288,
+      "epoch": 1.099822425164891,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.588566349352185e-06,
+      "loss": 1.33752937,
+      "memory(GiB)": 123.79,
+      "step": 43355,
+      "train_speed(iter/s)": 1.316252
+    },
+    {
+      "acc": 0.6895607,
+      "epoch": 1.099949264332826,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.5875212909298885e-06,
+      "loss": 1.35419273,
+      "memory(GiB)": 123.79,
+      "step": 43360,
+      "train_speed(iter/s)": 1.31627
+    },
+    {
+      "acc": 0.69440675,
+      "epoch": 1.1000761035007611,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.586476250650246e-06,
+      "loss": 1.34554195,
+      "memory(GiB)": 123.79,
+      "step": 43365,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.67967911,
+      "epoch": 1.1002029426686961,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.585431228559228e-06,
+      "loss": 1.38663692,
+      "memory(GiB)": 123.79,
+      "step": 43370,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.686204,
+      "epoch": 1.100329781836631,
+      "grad_norm": 2.875,
+      "learning_rate": 4.584386224702792e-06,
+      "loss": 1.36289186,
+      "memory(GiB)": 123.79,
+      "step": 43375,
+      "train_speed(iter/s)": 1.316321
+    },
+    {
+      "acc": 0.68715935,
+      "epoch": 1.1004566210045663,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.583341239126906e-06,
+      "loss": 1.39333801,
+      "memory(GiB)": 123.79,
+      "step": 43380,
+      "train_speed(iter/s)": 1.316338
+    },
+    {
+      "acc": 0.68331513,
+      "epoch": 1.1005834601725013,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.582296271877534e-06,
+      "loss": 1.43719978,
+      "memory(GiB)": 123.79,
+      "step": 43385,
+      "train_speed(iter/s)": 1.316355
+    },
+    {
+      "acc": 0.68492503,
+      "epoch": 1.1007102993404363,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.581251323000636e-06,
+      "loss": 1.41020756,
+      "memory(GiB)": 123.79,
+      "step": 43390,
+      "train_speed(iter/s)": 1.316371
+    },
+    {
+      "acc": 0.66348243,
+      "epoch": 1.1008371385083713,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.580206392542175e-06,
+      "loss": 1.43065252,
+      "memory(GiB)": 123.79,
+      "step": 43395,
+      "train_speed(iter/s)": 1.316387
+    },
+    {
+      "acc": 0.69218359,
+      "epoch": 1.1009639776763065,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.579161480548109e-06,
+      "loss": 1.38117943,
+      "memory(GiB)": 123.79,
+      "step": 43400,
+      "train_speed(iter/s)": 1.316403
+    },
+    {
+      "acc": 0.68576627,
+      "epoch": 1.1010908168442415,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.578116587064402e-06,
+      "loss": 1.41656313,
+      "memory(GiB)": 123.79,
+      "step": 43405,
+      "train_speed(iter/s)": 1.316419
+    },
+    {
+      "acc": 0.68820481,
+      "epoch": 1.1012176560121765,
+      "grad_norm": 5.28125,
+      "learning_rate": 4.577071712137012e-06,
+      "loss": 1.38880806,
+      "memory(GiB)": 123.79,
+      "step": 43410,
+      "train_speed(iter/s)": 1.316436
+    },
+    {
+      "acc": 0.69259896,
+      "epoch": 1.1013444951801117,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.576026855811893e-06,
+      "loss": 1.38166771,
+      "memory(GiB)": 123.79,
+      "step": 43415,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.69608412,
+      "epoch": 1.1014713343480467,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.5749820181350095e-06,
+      "loss": 1.37746677,
+      "memory(GiB)": 123.79,
+      "step": 43420,
+      "train_speed(iter/s)": 1.316469
+    },
+    {
+      "acc": 0.67662945,
+      "epoch": 1.1015981735159817,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.57393719915231e-06,
+      "loss": 1.38262787,
+      "memory(GiB)": 123.79,
+      "step": 43425,
+      "train_speed(iter/s)": 1.316485
+    },
+    {
+      "acc": 0.68372493,
+      "epoch": 1.1017250126839169,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.5728923989097604e-06,
+      "loss": 1.36218042,
+      "memory(GiB)": 123.79,
+      "step": 43430,
+      "train_speed(iter/s)": 1.3165
+    },
+    {
+      "acc": 0.68638482,
+      "epoch": 1.1018518518518519,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.571847617453306e-06,
+      "loss": 1.31903811,
+      "memory(GiB)": 123.79,
+      "step": 43435,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.68708448,
+      "epoch": 1.1019786910197868,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.570802854828906e-06,
+      "loss": 1.41692066,
+      "memory(GiB)": 123.79,
+      "step": 43440,
+      "train_speed(iter/s)": 1.316533
+    },
+    {
+      "acc": 0.68953032,
+      "epoch": 1.102105530187722,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.569758111082512e-06,
+      "loss": 1.35654478,
+      "memory(GiB)": 123.79,
+      "step": 43445,
+      "train_speed(iter/s)": 1.316548
+    },
+    {
+      "acc": 0.69562826,
+      "epoch": 1.102232369355657,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.568713386260078e-06,
+      "loss": 1.30741539,
+      "memory(GiB)": 123.79,
+      "step": 43450,
+      "train_speed(iter/s)": 1.316565
+    },
+    {
+      "acc": 0.70812531,
+      "epoch": 1.102359208523592,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.567668680407555e-06,
+      "loss": 1.34423046,
+      "memory(GiB)": 123.79,
+      "step": 43455,
+      "train_speed(iter/s)": 1.316582
+    },
+    {
+      "acc": 0.6767262,
+      "epoch": 1.102486047691527,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.566623993570893e-06,
+      "loss": 1.42684155,
+      "memory(GiB)": 123.79,
+      "step": 43460,
+      "train_speed(iter/s)": 1.316599
+    },
+    {
+      "acc": 0.66249371,
+      "epoch": 1.1026128868594622,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.565579325796043e-06,
+      "loss": 1.49520645,
+      "memory(GiB)": 123.79,
+      "step": 43465,
+      "train_speed(iter/s)": 1.316616
+    },
+    {
+      "acc": 0.69419136,
+      "epoch": 1.1027397260273972,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.564534677128954e-06,
+      "loss": 1.39936504,
+      "memory(GiB)": 123.79,
+      "step": 43470,
+      "train_speed(iter/s)": 1.316632
+    },
+    {
+      "acc": 0.6828742,
+      "epoch": 1.1028665651953322,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.563490047615574e-06,
+      "loss": 1.39724522,
+      "memory(GiB)": 123.79,
+      "step": 43475,
+      "train_speed(iter/s)": 1.316649
+    },
+    {
+      "acc": 0.68817501,
+      "epoch": 1.1029934043632674,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.56244543730185e-06,
+      "loss": 1.33778725,
+      "memory(GiB)": 123.79,
+      "step": 43480,
+      "train_speed(iter/s)": 1.316667
+    },
+    {
+      "acc": 0.69163389,
+      "epoch": 1.1031202435312024,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.561400846233729e-06,
+      "loss": 1.34562759,
+      "memory(GiB)": 123.79,
+      "step": 43485,
+      "train_speed(iter/s)": 1.316683
+    },
+    {
+      "acc": 0.68923159,
+      "epoch": 1.1032470826991374,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.56035627445716e-06,
+      "loss": 1.34048471,
+      "memory(GiB)": 123.79,
+      "step": 43490,
+      "train_speed(iter/s)": 1.3167
+    },
+    {
+      "acc": 0.67813106,
+      "epoch": 1.1033739218670726,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.55931172201808e-06,
+      "loss": 1.48742228,
+      "memory(GiB)": 123.79,
+      "step": 43495,
+      "train_speed(iter/s)": 1.316717
+    },
+    {
+      "acc": 0.68357601,
+      "epoch": 1.1035007610350076,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.558267188962441e-06,
+      "loss": 1.34261169,
+      "memory(GiB)": 123.79,
+      "step": 43500,
+      "train_speed(iter/s)": 1.316734
+    },
+    {
+      "acc": 0.68634477,
+      "epoch": 1.1036276002029426,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.557222675336182e-06,
+      "loss": 1.41457682,
+      "memory(GiB)": 123.79,
+      "step": 43505,
+      "train_speed(iter/s)": 1.31675
+    },
+    {
+      "acc": 0.68606029,
+      "epoch": 1.1037544393708778,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.556178181185249e-06,
+      "loss": 1.37355261,
+      "memory(GiB)": 123.79,
+      "step": 43510,
+      "train_speed(iter/s)": 1.316764
+    },
+    {
+      "acc": 0.67960572,
+      "epoch": 1.1038812785388128,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.555133706555579e-06,
+      "loss": 1.46122208,
+      "memory(GiB)": 123.79,
+      "step": 43515,
+      "train_speed(iter/s)": 1.316781
+    },
+    {
+      "acc": 0.68130903,
+      "epoch": 1.1040081177067478,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.554089251493115e-06,
+      "loss": 1.43087282,
+      "memory(GiB)": 123.79,
+      "step": 43520,
+      "train_speed(iter/s)": 1.316798
+    },
+    {
+      "acc": 0.69499092,
+      "epoch": 1.104134956874683,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.553044816043796e-06,
+      "loss": 1.33234749,
+      "memory(GiB)": 123.79,
+      "step": 43525,
+      "train_speed(iter/s)": 1.316814
+    },
+    {
+      "acc": 0.68837423,
+      "epoch": 1.104261796042618,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.552000400253563e-06,
+      "loss": 1.3915514,
+      "memory(GiB)": 123.79,
+      "step": 43530,
+      "train_speed(iter/s)": 1.316831
+    },
+    {
+      "acc": 0.69376879,
+      "epoch": 1.104388635210553,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.550956004168352e-06,
+      "loss": 1.39451008,
+      "memory(GiB)": 123.79,
+      "step": 43535,
+      "train_speed(iter/s)": 1.316848
+    },
+    {
+      "acc": 0.68394804,
+      "epoch": 1.1045154743784882,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.5499116278341e-06,
+      "loss": 1.43382874,
+      "memory(GiB)": 123.79,
+      "step": 43540,
+      "train_speed(iter/s)": 1.316864
+    },
+    {
+      "acc": 0.69223461,
+      "epoch": 1.1046423135464232,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.548867271296745e-06,
+      "loss": 1.41247005,
+      "memory(GiB)": 123.79,
+      "step": 43545,
+      "train_speed(iter/s)": 1.316882
+    },
+    {
+      "acc": 0.70650477,
+      "epoch": 1.1047691527143582,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.547822934602222e-06,
+      "loss": 1.31981249,
+      "memory(GiB)": 123.79,
+      "step": 43550,
+      "train_speed(iter/s)": 1.316899
+    },
+    {
+      "acc": 0.69320521,
+      "epoch": 1.1048959918822931,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.5467786177964635e-06,
+      "loss": 1.38711567,
+      "memory(GiB)": 123.79,
+      "step": 43555,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.67883048,
+      "epoch": 1.1050228310502284,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.545734320925406e-06,
+      "loss": 1.41694708,
+      "memory(GiB)": 123.79,
+      "step": 43560,
+      "train_speed(iter/s)": 1.316933
+    },
+    {
+      "acc": 0.69376721,
+      "epoch": 1.1051496702181633,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.544690044034981e-06,
+      "loss": 1.37914,
+      "memory(GiB)": 123.79,
+      "step": 43565,
+      "train_speed(iter/s)": 1.31695
+    },
+    {
+      "acc": 0.69017682,
+      "epoch": 1.1052765093860983,
+      "grad_norm": 2.875,
+      "learning_rate": 4.543645787171122e-06,
+      "loss": 1.38677197,
+      "memory(GiB)": 123.79,
+      "step": 43570,
+      "train_speed(iter/s)": 1.316967
+    },
+    {
+      "acc": 0.67823043,
+      "epoch": 1.1054033485540335,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.5426015503797565e-06,
+      "loss": 1.42175446,
+      "memory(GiB)": 123.79,
+      "step": 43575,
+      "train_speed(iter/s)": 1.316984
+    },
+    {
+      "acc": 0.68380299,
+      "epoch": 1.1055301877219685,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.5415573337068185e-06,
+      "loss": 1.40788765,
+      "memory(GiB)": 123.79,
+      "step": 43580,
+      "train_speed(iter/s)": 1.317
+    },
+    {
+      "acc": 0.68498621,
+      "epoch": 1.1056570268899035,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.540513137198233e-06,
+      "loss": 1.42718859,
+      "memory(GiB)": 123.79,
+      "step": 43585,
+      "train_speed(iter/s)": 1.317017
+    },
+    {
+      "acc": 0.69982595,
+      "epoch": 1.1057838660578387,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.539468960899936e-06,
+      "loss": 1.33059568,
+      "memory(GiB)": 123.79,
+      "step": 43590,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.67103744,
+      "epoch": 1.1059107052257737,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.538424804857847e-06,
+      "loss": 1.44552231,
+      "memory(GiB)": 123.79,
+      "step": 43595,
+      "train_speed(iter/s)": 1.31705
+    },
+    {
+      "acc": 0.68911109,
+      "epoch": 1.1060375443937087,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.537380669117896e-06,
+      "loss": 1.42401085,
+      "memory(GiB)": 123.79,
+      "step": 43600,
+      "train_speed(iter/s)": 1.317068
+    },
+    {
+      "acc": 0.69056258,
+      "epoch": 1.106164383561644,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.536336553726008e-06,
+      "loss": 1.36317196,
+      "memory(GiB)": 123.79,
+      "step": 43605,
+      "train_speed(iter/s)": 1.317085
+    },
+    {
+      "acc": 0.6959754,
+      "epoch": 1.106291222729579,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.535292458728112e-06,
+      "loss": 1.37806692,
+      "memory(GiB)": 123.79,
+      "step": 43610,
+      "train_speed(iter/s)": 1.317101
+    },
+    {
+      "acc": 0.67768645,
+      "epoch": 1.106418061897514,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.534248384170126e-06,
+      "loss": 1.41836157,
+      "memory(GiB)": 123.79,
+      "step": 43615,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.68610401,
+      "epoch": 1.106544901065449,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.533204330097974e-06,
+      "loss": 1.39499855,
+      "memory(GiB)": 123.79,
+      "step": 43620,
+      "train_speed(iter/s)": 1.317135
+    },
+    {
+      "acc": 0.69331541,
+      "epoch": 1.106671740233384,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.532160296557581e-06,
+      "loss": 1.39481306,
+      "memory(GiB)": 123.79,
+      "step": 43625,
+      "train_speed(iter/s)": 1.317153
+    },
+    {
+      "acc": 0.70045757,
+      "epoch": 1.106798579401319,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.531116283594868e-06,
+      "loss": 1.33658037,
+      "memory(GiB)": 123.79,
+      "step": 43630,
+      "train_speed(iter/s)": 1.31717
+    },
+    {
+      "acc": 0.70956612,
+      "epoch": 1.106925418569254,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.530072291255753e-06,
+      "loss": 1.35253086,
+      "memory(GiB)": 123.79,
+      "step": 43635,
+      "train_speed(iter/s)": 1.317186
+    },
+    {
+      "acc": 0.69169602,
+      "epoch": 1.1070522577371893,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.529028319586157e-06,
+      "loss": 1.39086008,
+      "memory(GiB)": 123.79,
+      "step": 43640,
+      "train_speed(iter/s)": 1.317203
+    },
+    {
+      "acc": 0.6794075,
+      "epoch": 1.1071790969051243,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.527984368631997e-06,
+      "loss": 1.47246609,
+      "memory(GiB)": 123.79,
+      "step": 43645,
+      "train_speed(iter/s)": 1.31722
+    },
+    {
+      "acc": 0.6814363,
+      "epoch": 1.1073059360730593,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.526940438439196e-06,
+      "loss": 1.38167982,
+      "memory(GiB)": 123.79,
+      "step": 43650,
+      "train_speed(iter/s)": 1.317235
+    },
+    {
+      "acc": 0.69239335,
+      "epoch": 1.1074327752409945,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.525896529053662e-06,
+      "loss": 1.41323357,
+      "memory(GiB)": 123.79,
+      "step": 43655,
+      "train_speed(iter/s)": 1.317252
+    },
+    {
+      "acc": 0.68634052,
+      "epoch": 1.1075596144089295,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.524852640521318e-06,
+      "loss": 1.36339407,
+      "memory(GiB)": 123.79,
+      "step": 43660,
+      "train_speed(iter/s)": 1.317269
+    },
+    {
+      "acc": 0.69031558,
+      "epoch": 1.1076864535768645,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.523808772888073e-06,
+      "loss": 1.33424234,
+      "memory(GiB)": 123.79,
+      "step": 43665,
+      "train_speed(iter/s)": 1.317286
+    },
+    {
+      "acc": 0.68554311,
+      "epoch": 1.1078132927447997,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.522764926199848e-06,
+      "loss": 1.39165287,
+      "memory(GiB)": 123.79,
+      "step": 43670,
+      "train_speed(iter/s)": 1.317303
+    },
+    {
+      "acc": 0.68598661,
+      "epoch": 1.1079401319127347,
+      "grad_norm": 3.375,
+      "learning_rate": 4.5217211005025516e-06,
+      "loss": 1.36643209,
+      "memory(GiB)": 123.79,
+      "step": 43675,
+      "train_speed(iter/s)": 1.317321
+    },
+    {
+      "acc": 0.68369484,
+      "epoch": 1.1080669710806696,
+      "grad_norm": 3.625,
+      "learning_rate": 4.520677295842095e-06,
+      "loss": 1.36842079,
+      "memory(GiB)": 123.79,
+      "step": 43680,
+      "train_speed(iter/s)": 1.317337
+    },
+    {
+      "acc": 0.70670161,
+      "epoch": 1.1081938102486049,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.5196335122643915e-06,
+      "loss": 1.25797491,
+      "memory(GiB)": 123.79,
+      "step": 43685,
+      "train_speed(iter/s)": 1.317354
+    },
+    {
+      "acc": 0.7011899,
+      "epoch": 1.1083206494165398,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.518589749815352e-06,
+      "loss": 1.34049416,
+      "memory(GiB)": 123.79,
+      "step": 43690,
+      "train_speed(iter/s)": 1.317371
+    },
+    {
+      "acc": 0.68848047,
+      "epoch": 1.1084474885844748,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.517546008540884e-06,
+      "loss": 1.37589226,
+      "memory(GiB)": 123.79,
+      "step": 43695,
+      "train_speed(iter/s)": 1.317388
+    },
+    {
+      "acc": 0.68585043,
+      "epoch": 1.10857432775241,
+      "grad_norm": 2.875,
+      "learning_rate": 4.5165022884868946e-06,
+      "loss": 1.37538633,
+      "memory(GiB)": 123.79,
+      "step": 43700,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.67004867,
+      "epoch": 1.108701166920345,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.515458589699295e-06,
+      "loss": 1.4798542,
+      "memory(GiB)": 123.79,
+      "step": 43705,
+      "train_speed(iter/s)": 1.317422
+    },
+    {
+      "acc": 0.68489532,
+      "epoch": 1.10882800608828,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.514414912223991e-06,
+      "loss": 1.39138727,
+      "memory(GiB)": 123.79,
+      "step": 43710,
+      "train_speed(iter/s)": 1.317439
+    },
+    {
+      "acc": 0.69269323,
+      "epoch": 1.108954845256215,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.513371256106885e-06,
+      "loss": 1.32027884,
+      "memory(GiB)": 123.79,
+      "step": 43715,
+      "train_speed(iter/s)": 1.317457
+    },
+    {
+      "acc": 0.6775835,
+      "epoch": 1.1090816844241502,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.512327621393885e-06,
+      "loss": 1.39567661,
+      "memory(GiB)": 123.79,
+      "step": 43720,
+      "train_speed(iter/s)": 1.317473
+    },
+    {
+      "acc": 0.70156474,
+      "epoch": 1.1092085235920852,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.511284008130892e-06,
+      "loss": 1.34130421,
+      "memory(GiB)": 123.79,
+      "step": 43725,
+      "train_speed(iter/s)": 1.31749
+    },
+    {
+      "acc": 0.68799076,
+      "epoch": 1.1093353627600202,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.510240416363813e-06,
+      "loss": 1.37351055,
+      "memory(GiB)": 123.79,
+      "step": 43730,
+      "train_speed(iter/s)": 1.317506
+    },
+    {
+      "acc": 0.67968316,
+      "epoch": 1.1094622019279554,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.5091968461385455e-06,
+      "loss": 1.4495079,
+      "memory(GiB)": 123.79,
+      "step": 43735,
+      "train_speed(iter/s)": 1.317524
+    },
+    {
+      "acc": 0.69002509,
+      "epoch": 1.1095890410958904,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.508153297500993e-06,
+      "loss": 1.38936205,
+      "memory(GiB)": 123.79,
+      "step": 43740,
+      "train_speed(iter/s)": 1.317541
+    },
+    {
+      "acc": 0.68589334,
+      "epoch": 1.1097158802638254,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.507109770497052e-06,
+      "loss": 1.36519537,
+      "memory(GiB)": 123.79,
+      "step": 43745,
+      "train_speed(iter/s)": 1.317557
+    },
+    {
+      "acc": 0.70544987,
+      "epoch": 1.1098427194317606,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.506066265172626e-06,
+      "loss": 1.33333549,
+      "memory(GiB)": 123.79,
+      "step": 43750,
+      "train_speed(iter/s)": 1.317575
+    },
+    {
+      "acc": 0.65945849,
+      "epoch": 1.1099695585996956,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.505022781573611e-06,
+      "loss": 1.44980764,
+      "memory(GiB)": 123.79,
+      "step": 43755,
+      "train_speed(iter/s)": 1.317591
+    },
+    {
+      "acc": 0.69747744,
+      "epoch": 1.1100963977676306,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.503979319745902e-06,
+      "loss": 1.38378801,
+      "memory(GiB)": 123.79,
+      "step": 43760,
+      "train_speed(iter/s)": 1.317606
+    },
+    {
+      "acc": 0.69123855,
+      "epoch": 1.1102232369355658,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.502935879735398e-06,
+      "loss": 1.36893463,
+      "memory(GiB)": 123.79,
+      "step": 43765,
+      "train_speed(iter/s)": 1.317622
+    },
+    {
+      "acc": 0.6862895,
+      "epoch": 1.1103500761035008,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.5018924615879956e-06,
+      "loss": 1.40643549,
+      "memory(GiB)": 123.79,
+      "step": 43770,
+      "train_speed(iter/s)": 1.317639
+    },
+    {
+      "acc": 0.69106245,
+      "epoch": 1.1104769152714358,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.500849065349584e-06,
+      "loss": 1.38642168,
+      "memory(GiB)": 123.79,
+      "step": 43775,
+      "train_speed(iter/s)": 1.317656
+    },
+    {
+      "acc": 0.67592139,
+      "epoch": 1.1106037544393708,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.499805691066059e-06,
+      "loss": 1.38001709,
+      "memory(GiB)": 123.79,
+      "step": 43780,
+      "train_speed(iter/s)": 1.317673
+    },
+    {
+      "acc": 0.68348451,
+      "epoch": 1.110730593607306,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.498762338783314e-06,
+      "loss": 1.38571625,
+      "memory(GiB)": 123.79,
+      "step": 43785,
+      "train_speed(iter/s)": 1.31769
+    },
+    {
+      "acc": 0.69398022,
+      "epoch": 1.110857432775241,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.49771900854724e-06,
+      "loss": 1.36240215,
+      "memory(GiB)": 123.79,
+      "step": 43790,
+      "train_speed(iter/s)": 1.317706
+    },
+    {
+      "acc": 0.67458134,
+      "epoch": 1.110984271943176,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.496675700403724e-06,
+      "loss": 1.47417793,
+      "memory(GiB)": 123.79,
+      "step": 43795,
+      "train_speed(iter/s)": 1.317722
+    },
+    {
+      "acc": 0.69218302,
+      "epoch": 1.1111111111111112,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.495632414398659e-06,
+      "loss": 1.44541378,
+      "memory(GiB)": 123.79,
+      "step": 43800,
+      "train_speed(iter/s)": 1.317739
+    },
+    {
+      "acc": 0.67136455,
+      "epoch": 1.1112379502790461,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.494589150577932e-06,
+      "loss": 1.48677197,
+      "memory(GiB)": 123.79,
+      "step": 43805,
+      "train_speed(iter/s)": 1.317757
+    },
+    {
+      "acc": 0.69725833,
+      "epoch": 1.1113647894469811,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.493545908987432e-06,
+      "loss": 1.40566826,
+      "memory(GiB)": 123.79,
+      "step": 43810,
+      "train_speed(iter/s)": 1.317774
+    },
+    {
+      "acc": 0.67147937,
+      "epoch": 1.1114916286149163,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.492502689673044e-06,
+      "loss": 1.48480749,
+      "memory(GiB)": 123.79,
+      "step": 43815,
+      "train_speed(iter/s)": 1.317792
+    },
+    {
+      "acc": 0.70274563,
+      "epoch": 1.1116184677828513,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.491459492680651e-06,
+      "loss": 1.36228199,
+      "memory(GiB)": 123.79,
+      "step": 43820,
+      "train_speed(iter/s)": 1.317808
+    },
+    {
+      "acc": 0.6901001,
+      "epoch": 1.1117453069507863,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.4904163180561425e-06,
+      "loss": 1.33092699,
+      "memory(GiB)": 123.79,
+      "step": 43825,
+      "train_speed(iter/s)": 1.317825
+    },
+    {
+      "acc": 0.66574907,
+      "epoch": 1.1118721461187215,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.4893731658453996e-06,
+      "loss": 1.40526209,
+      "memory(GiB)": 123.79,
+      "step": 43830,
+      "train_speed(iter/s)": 1.317842
+    },
+    {
+      "acc": 0.69010844,
+      "epoch": 1.1119989852866565,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.4883300360943035e-06,
+      "loss": 1.36197243,
+      "memory(GiB)": 123.79,
+      "step": 43835,
+      "train_speed(iter/s)": 1.31786
+    },
+    {
+      "acc": 0.68316922,
+      "epoch": 1.1121258244545915,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.4872869288487366e-06,
+      "loss": 1.40856895,
+      "memory(GiB)": 123.79,
+      "step": 43840,
+      "train_speed(iter/s)": 1.317876
+    },
+    {
+      "acc": 0.67949677,
+      "epoch": 1.1122526636225267,
+      "grad_norm": 2.625,
+      "learning_rate": 4.48624384415458e-06,
+      "loss": 1.32783947,
+      "memory(GiB)": 123.79,
+      "step": 43845,
+      "train_speed(iter/s)": 1.317893
+    },
+    {
+      "acc": 0.68177471,
+      "epoch": 1.1123795027904617,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.485200782057715e-06,
+      "loss": 1.44340563,
+      "memory(GiB)": 123.79,
+      "step": 43850,
+      "train_speed(iter/s)": 1.31791
+    },
+    {
+      "acc": 0.68121805,
+      "epoch": 1.1125063419583967,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.4841577426040145e-06,
+      "loss": 1.42093887,
+      "memory(GiB)": 123.79,
+      "step": 43855,
+      "train_speed(iter/s)": 1.317927
+    },
+    {
+      "acc": 0.68667078,
+      "epoch": 1.112633181126332,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.483114725839361e-06,
+      "loss": 1.37968769,
+      "memory(GiB)": 123.79,
+      "step": 43860,
+      "train_speed(iter/s)": 1.317944
+    },
+    {
+      "acc": 0.68661518,
+      "epoch": 1.112760020294267,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.482071731809629e-06,
+      "loss": 1.43673878,
+      "memory(GiB)": 123.79,
+      "step": 43865,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.67433934,
+      "epoch": 1.112886859462202,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.481028760560697e-06,
+      "loss": 1.43932705,
+      "memory(GiB)": 123.79,
+      "step": 43870,
+      "train_speed(iter/s)": 1.317975
+    },
+    {
+      "acc": 0.6917192,
+      "epoch": 1.1130136986301369,
+      "grad_norm": 3.25,
+      "learning_rate": 4.479985812138435e-06,
+      "loss": 1.40321579,
+      "memory(GiB)": 123.79,
+      "step": 43875,
+      "train_speed(iter/s)": 1.317991
+    },
+    {
+      "acc": 0.68217573,
+      "epoch": 1.113140537798072,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.478942886588719e-06,
+      "loss": 1.37511063,
+      "memory(GiB)": 123.79,
+      "step": 43880,
+      "train_speed(iter/s)": 1.318008
+    },
+    {
+      "acc": 0.69069128,
+      "epoch": 1.113267376966007,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.47789998395742e-06,
+      "loss": 1.37118645,
+      "memory(GiB)": 123.79,
+      "step": 43885,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.69970112,
+      "epoch": 1.113394216133942,
+      "grad_norm": 3.96875,
+      "learning_rate": 4.476857104290413e-06,
+      "loss": 1.40645676,
+      "memory(GiB)": 123.79,
+      "step": 43890,
+      "train_speed(iter/s)": 1.318039
+    },
+    {
+      "acc": 0.69786663,
+      "epoch": 1.1135210553018773,
+      "grad_norm": 3.5,
+      "learning_rate": 4.4758142476335655e-06,
+      "loss": 1.37978477,
+      "memory(GiB)": 123.79,
+      "step": 43895,
+      "train_speed(iter/s)": 1.318055
+    },
+    {
+      "acc": 0.70437288,
+      "epoch": 1.1136478944698123,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.474771414032747e-06,
+      "loss": 1.41813612,
+      "memory(GiB)": 123.79,
+      "step": 43900,
+      "train_speed(iter/s)": 1.318071
+    },
+    {
+      "acc": 0.69659224,
+      "epoch": 1.1137747336377473,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.473728603533827e-06,
+      "loss": 1.40724926,
+      "memory(GiB)": 123.79,
+      "step": 43905,
+      "train_speed(iter/s)": 1.318087
+    },
+    {
+      "acc": 0.68081036,
+      "epoch": 1.1139015728056825,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.472685816182674e-06,
+      "loss": 1.39291229,
+      "memory(GiB)": 123.79,
+      "step": 43910,
+      "train_speed(iter/s)": 1.318104
+    },
+    {
+      "acc": 0.68829679,
+      "epoch": 1.1140284119736175,
+      "grad_norm": 4.125,
+      "learning_rate": 4.471643052025152e-06,
+      "loss": 1.43146381,
+      "memory(GiB)": 123.79,
+      "step": 43915,
+      "train_speed(iter/s)": 1.31812
+    },
+    {
+      "acc": 0.67473211,
+      "epoch": 1.1141552511415524,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.470600311107127e-06,
+      "loss": 1.44199257,
+      "memory(GiB)": 123.79,
+      "step": 43920,
+      "train_speed(iter/s)": 1.318135
+    },
+    {
+      "acc": 0.67240291,
+      "epoch": 1.1142820903094877,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.469557593474464e-06,
+      "loss": 1.3627739,
+      "memory(GiB)": 123.79,
+      "step": 43925,
+      "train_speed(iter/s)": 1.318152
+    },
+    {
+      "acc": 0.67805071,
+      "epoch": 1.1144089294774226,
+      "grad_norm": 3.25,
+      "learning_rate": 4.468514899173027e-06,
+      "loss": 1.41665134,
+      "memory(GiB)": 123.79,
+      "step": 43930,
+      "train_speed(iter/s)": 1.318168
+    },
+    {
+      "acc": 0.7006588,
+      "epoch": 1.1145357686453576,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.4674722282486775e-06,
+      "loss": 1.35122719,
+      "memory(GiB)": 123.79,
+      "step": 43935,
+      "train_speed(iter/s)": 1.318184
+    },
+    {
+      "acc": 0.69479036,
+      "epoch": 1.1146626078132926,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.4664295807472765e-06,
+      "loss": 1.34772682,
+      "memory(GiB)": 123.79,
+      "step": 43940,
+      "train_speed(iter/s)": 1.3182
+    },
+    {
+      "acc": 0.69084024,
+      "epoch": 1.1147894469812278,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.465386956714684e-06,
+      "loss": 1.3900897,
+      "memory(GiB)": 123.79,
+      "step": 43945,
+      "train_speed(iter/s)": 1.318216
+    },
+    {
+      "acc": 0.68622169,
+      "epoch": 1.1149162861491628,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.4643443561967625e-06,
+      "loss": 1.37848825,
+      "memory(GiB)": 123.79,
+      "step": 43950,
+      "train_speed(iter/s)": 1.318232
+    },
+    {
+      "acc": 0.69514284,
+      "epoch": 1.1150431253170978,
+      "grad_norm": 3.5,
+      "learning_rate": 4.463301779239366e-06,
+      "loss": 1.28658199,
+      "memory(GiB)": 123.79,
+      "step": 43955,
+      "train_speed(iter/s)": 1.318249
+    },
+    {
+      "acc": 0.69927726,
+      "epoch": 1.115169964485033,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.462259225888354e-06,
+      "loss": 1.33332253,
+      "memory(GiB)": 123.79,
+      "step": 43960,
+      "train_speed(iter/s)": 1.318265
+    },
+    {
+      "acc": 0.67671881,
+      "epoch": 1.115296803652968,
+      "grad_norm": 3.0,
+      "learning_rate": 4.4612166961895805e-06,
+      "loss": 1.44549789,
+      "memory(GiB)": 123.79,
+      "step": 43965,
+      "train_speed(iter/s)": 1.318281
+    },
+    {
+      "acc": 0.68675103,
+      "epoch": 1.115423642820903,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.460174190188905e-06,
+      "loss": 1.39276562,
+      "memory(GiB)": 123.79,
+      "step": 43970,
+      "train_speed(iter/s)": 1.318298
+    },
+    {
+      "acc": 0.67915387,
+      "epoch": 1.1155504819888382,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.459131707932177e-06,
+      "loss": 1.46022463,
+      "memory(GiB)": 123.79,
+      "step": 43975,
+      "train_speed(iter/s)": 1.318315
+    },
+    {
+      "acc": 0.67823615,
+      "epoch": 1.1156773211567732,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.458089249465251e-06,
+      "loss": 1.42723093,
+      "memory(GiB)": 123.79,
+      "step": 43980,
+      "train_speed(iter/s)": 1.318332
+    },
+    {
+      "acc": 0.67197886,
+      "epoch": 1.1158041603247082,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.45704681483398e-06,
+      "loss": 1.44425163,
+      "memory(GiB)": 123.79,
+      "step": 43985,
+      "train_speed(iter/s)": 1.318349
+    },
+    {
+      "acc": 0.68334546,
+      "epoch": 1.1159309994926434,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.456004404084215e-06,
+      "loss": 1.40268326,
+      "memory(GiB)": 123.79,
+      "step": 43990,
+      "train_speed(iter/s)": 1.318366
+    },
+    {
+      "acc": 0.67071724,
+      "epoch": 1.1160578386605784,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.454962017261803e-06,
+      "loss": 1.46981297,
+      "memory(GiB)": 123.79,
+      "step": 43995,
+      "train_speed(iter/s)": 1.318383
+    },
+    {
+      "acc": 0.68493338,
+      "epoch": 1.1161846778285134,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.453919654412596e-06,
+      "loss": 1.3842207,
+      "memory(GiB)": 123.79,
+      "step": 44000,
+      "train_speed(iter/s)": 1.318401
+    },
+    {
+      "epoch": 1.1161846778285134,
+      "eval_acc": 0.6750507980461808,
+      "eval_loss": 1.3567947149276733,
+      "eval_runtime": 69.8424,
+      "eval_samples_per_second": 91.205,
+      "eval_steps_per_second": 22.808,
+      "step": 44000
+    },
+    {
+      "acc": 0.68502755,
+      "epoch": 1.1163115169964486,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.45287731558244e-06,
+      "loss": 1.35427408,
+      "memory(GiB)": 123.79,
+      "step": 44005,
+      "train_speed(iter/s)": 1.315169
+    },
+    {
+      "acc": 0.68645287,
+      "epoch": 1.1164383561643836,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.451835000817185e-06,
+      "loss": 1.36689062,
+      "memory(GiB)": 123.79,
+      "step": 44010,
+      "train_speed(iter/s)": 1.315185
+    },
+    {
+      "acc": 0.69929438,
+      "epoch": 1.1165651953323186,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.450792710162672e-06,
+      "loss": 1.41724796,
+      "memory(GiB)": 123.79,
+      "step": 44015,
+      "train_speed(iter/s)": 1.315201
+    },
+    {
+      "acc": 0.68035336,
+      "epoch": 1.1166920345002538,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.449750443664747e-06,
+      "loss": 1.3674305,
+      "memory(GiB)": 123.79,
+      "step": 44020,
+      "train_speed(iter/s)": 1.315218
+    },
+    {
+      "acc": 0.69054499,
+      "epoch": 1.1168188736681888,
+      "grad_norm": 2.75,
+      "learning_rate": 4.448708201369254e-06,
+      "loss": 1.37573147,
+      "memory(GiB)": 123.79,
+      "step": 44025,
+      "train_speed(iter/s)": 1.315235
+    },
+    {
+      "acc": 0.68291445,
+      "epoch": 1.1169457128361238,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.4476659833220374e-06,
+      "loss": 1.39577885,
+      "memory(GiB)": 123.79,
+      "step": 44030,
+      "train_speed(iter/s)": 1.315251
+    },
+    {
+      "acc": 0.69466343,
+      "epoch": 1.1170725520040587,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.4466237895689365e-06,
+      "loss": 1.36969852,
+      "memory(GiB)": 123.79,
+      "step": 44035,
+      "train_speed(iter/s)": 1.315267
+    },
+    {
+      "acc": 0.68663015,
+      "epoch": 1.117199391171994,
+      "grad_norm": 3.625,
+      "learning_rate": 4.44558162015579e-06,
+      "loss": 1.38666544,
+      "memory(GiB)": 123.79,
+      "step": 44040,
+      "train_speed(iter/s)": 1.315283
+    },
+    {
+      "acc": 0.68522701,
+      "epoch": 1.117326230339929,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.444539475128441e-06,
+      "loss": 1.42468414,
+      "memory(GiB)": 123.79,
+      "step": 44045,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.69099865,
+      "epoch": 1.117453069507864,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.443497354532726e-06,
+      "loss": 1.40976849,
+      "memory(GiB)": 123.79,
+      "step": 44050,
+      "train_speed(iter/s)": 1.315316
+    },
+    {
+      "acc": 0.69059267,
+      "epoch": 1.1175799086757991,
+      "grad_norm": 3.125,
+      "learning_rate": 4.442455258414482e-06,
+      "loss": 1.32812891,
+      "memory(GiB)": 123.79,
+      "step": 44055,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.6769742,
+      "epoch": 1.1177067478437341,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.441413186819543e-06,
+      "loss": 1.37363577,
+      "memory(GiB)": 123.79,
+      "step": 44060,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.6940773,
+      "epoch": 1.1178335870116691,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.440371139793747e-06,
+      "loss": 1.33532648,
+      "memory(GiB)": 123.79,
+      "step": 44065,
+      "train_speed(iter/s)": 1.315365
+    },
+    {
+      "acc": 0.69260001,
+      "epoch": 1.1179604261796043,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.43932911738293e-06,
+      "loss": 1.374224,
+      "memory(GiB)": 123.79,
+      "step": 44070,
+      "train_speed(iter/s)": 1.315381
+    },
+    {
+      "acc": 0.68936605,
+      "epoch": 1.1180872653475393,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.438287119632917e-06,
+      "loss": 1.40246658,
+      "memory(GiB)": 123.79,
+      "step": 44075,
+      "train_speed(iter/s)": 1.315397
+    },
+    {
+      "acc": 0.68651299,
+      "epoch": 1.1182141045154743,
+      "grad_norm": 5.21875,
+      "learning_rate": 4.4372451465895465e-06,
+      "loss": 1.37940922,
+      "memory(GiB)": 123.79,
+      "step": 44080,
+      "train_speed(iter/s)": 1.315414
+    },
+    {
+      "acc": 0.68315201,
+      "epoch": 1.1183409436834095,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.436203198298645e-06,
+      "loss": 1.44510689,
+      "memory(GiB)": 123.79,
+      "step": 44085,
+      "train_speed(iter/s)": 1.315431
+    },
+    {
+      "acc": 0.67803221,
+      "epoch": 1.1184677828513445,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.435161274806049e-06,
+      "loss": 1.38262825,
+      "memory(GiB)": 123.79,
+      "step": 44090,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.67560291,
+      "epoch": 1.1185946220192795,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.4341193761575765e-06,
+      "loss": 1.43631992,
+      "memory(GiB)": 123.79,
+      "step": 44095,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.69593487,
+      "epoch": 1.1187214611872145,
+      "grad_norm": 3.5,
+      "learning_rate": 4.433077502399063e-06,
+      "loss": 1.39699955,
+      "memory(GiB)": 123.79,
+      "step": 44100,
+      "train_speed(iter/s)": 1.315482
+    },
+    {
+      "acc": 0.6917418,
+      "epoch": 1.1188483003551497,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.43203565357633e-06,
+      "loss": 1.32135658,
+      "memory(GiB)": 123.79,
+      "step": 44105,
+      "train_speed(iter/s)": 1.315498
+    },
+    {
+      "acc": 0.69168625,
+      "epoch": 1.1189751395230847,
+      "grad_norm": 3.5,
+      "learning_rate": 4.430993829735208e-06,
+      "loss": 1.35649605,
+      "memory(GiB)": 123.79,
+      "step": 44110,
+      "train_speed(iter/s)": 1.315514
+    },
+    {
+      "acc": 0.69136515,
+      "epoch": 1.1191019786910197,
+      "grad_norm": 2.875,
+      "learning_rate": 4.429952030921516e-06,
+      "loss": 1.33421316,
+      "memory(GiB)": 123.79,
+      "step": 44115,
+      "train_speed(iter/s)": 1.315531
+    },
+    {
+      "acc": 0.70333672,
+      "epoch": 1.119228817858955,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.428910257181077e-06,
+      "loss": 1.40199718,
+      "memory(GiB)": 123.79,
+      "step": 44120,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.69504166,
+      "epoch": 1.1193556570268899,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.427868508559717e-06,
+      "loss": 1.34104595,
+      "memory(GiB)": 123.79,
+      "step": 44125,
+      "train_speed(iter/s)": 1.31556
+    },
+    {
+      "acc": 0.70401678,
+      "epoch": 1.1194824961948249,
+      "grad_norm": 3.5,
+      "learning_rate": 4.426826785103256e-06,
+      "loss": 1.31157703,
+      "memory(GiB)": 123.79,
+      "step": 44130,
+      "train_speed(iter/s)": 1.315572
+    },
+    {
+      "acc": 0.68614435,
+      "epoch": 1.11960933536276,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.425785086857509e-06,
+      "loss": 1.38516407,
+      "memory(GiB)": 123.79,
+      "step": 44135,
+      "train_speed(iter/s)": 1.315589
+    },
+    {
+      "acc": 0.68450642,
+      "epoch": 1.119736174530695,
+      "grad_norm": 3.375,
+      "learning_rate": 4.424743413868298e-06,
+      "loss": 1.4040081,
+      "memory(GiB)": 123.79,
+      "step": 44140,
+      "train_speed(iter/s)": 1.315605
+    },
+    {
+      "acc": 0.68769979,
+      "epoch": 1.11986301369863,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.42370176618144e-06,
+      "loss": 1.35496054,
+      "memory(GiB)": 123.79,
+      "step": 44145,
+      "train_speed(iter/s)": 1.315621
+    },
+    {
+      "acc": 0.67780404,
+      "epoch": 1.1199898528665653,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.422660143842753e-06,
+      "loss": 1.40053806,
+      "memory(GiB)": 123.79,
+      "step": 44150,
+      "train_speed(iter/s)": 1.315638
+    },
+    {
+      "acc": 0.68210621,
+      "epoch": 1.1201166920345003,
+      "grad_norm": 4.46875,
+      "learning_rate": 4.421618546898048e-06,
+      "loss": 1.43650436,
+      "memory(GiB)": 123.79,
+      "step": 44155,
+      "train_speed(iter/s)": 1.315654
+    },
+    {
+      "acc": 0.691465,
+      "epoch": 1.1202435312024352,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.420576975393143e-06,
+      "loss": 1.30375023,
+      "memory(GiB)": 123.79,
+      "step": 44160,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.70514784,
+      "epoch": 1.1203703703703705,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.4195354293738484e-06,
+      "loss": 1.32250423,
+      "memory(GiB)": 123.79,
+      "step": 44165,
+      "train_speed(iter/s)": 1.315687
+    },
+    {
+      "acc": 0.68147125,
+      "epoch": 1.1204972095383054,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.418493908885979e-06,
+      "loss": 1.33992386,
+      "memory(GiB)": 123.79,
+      "step": 44170,
+      "train_speed(iter/s)": 1.315703
+    },
+    {
+      "acc": 0.68880591,
+      "epoch": 1.1206240487062404,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.417452413975343e-06,
+      "loss": 1.37744236,
+      "memory(GiB)": 123.79,
+      "step": 44175,
+      "train_speed(iter/s)": 1.315719
+    },
+    {
+      "acc": 0.68996687,
+      "epoch": 1.1207508878741756,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.4164109446877514e-06,
+      "loss": 1.37222538,
+      "memory(GiB)": 123.79,
+      "step": 44180,
+      "train_speed(iter/s)": 1.315736
+    },
+    {
+      "acc": 0.69843893,
+      "epoch": 1.1208777270421106,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.41536950106901e-06,
+      "loss": 1.3553462,
+      "memory(GiB)": 123.79,
+      "step": 44185,
+      "train_speed(iter/s)": 1.315752
+    },
+    {
+      "acc": 0.69835768,
+      "epoch": 1.1210045662100456,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.414328083164931e-06,
+      "loss": 1.3424799,
+      "memory(GiB)": 123.79,
+      "step": 44190,
+      "train_speed(iter/s)": 1.315768
+    },
+    {
+      "acc": 0.67288246,
+      "epoch": 1.1211314053779806,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.4132866910213154e-06,
+      "loss": 1.42995434,
+      "memory(GiB)": 123.79,
+      "step": 44195,
+      "train_speed(iter/s)": 1.315785
+    },
+    {
+      "acc": 0.69501944,
+      "epoch": 1.1212582445459158,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.41224532468397e-06,
+      "loss": 1.28683004,
+      "memory(GiB)": 123.79,
+      "step": 44200,
+      "train_speed(iter/s)": 1.315795
+    },
+    {
+      "acc": 0.68517456,
+      "epoch": 1.1213850837138508,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.411203984198701e-06,
+      "loss": 1.39435272,
+      "memory(GiB)": 123.79,
+      "step": 44205,
+      "train_speed(iter/s)": 1.315812
+    },
+    {
+      "acc": 0.67191353,
+      "epoch": 1.1215119228817858,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.41016266961131e-06,
+      "loss": 1.39434566,
+      "memory(GiB)": 123.79,
+      "step": 44210,
+      "train_speed(iter/s)": 1.315829
+    },
+    {
+      "acc": 0.6901916,
+      "epoch": 1.121638762049721,
+      "grad_norm": 4.46875,
+      "learning_rate": 4.409121380967597e-06,
+      "loss": 1.41234016,
+      "memory(GiB)": 123.79,
+      "step": 44215,
+      "train_speed(iter/s)": 1.315846
+    },
+    {
+      "acc": 0.69546099,
+      "epoch": 1.121765601217656,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.408080118313364e-06,
+      "loss": 1.39184093,
+      "memory(GiB)": 123.79,
+      "step": 44220,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68884406,
+      "epoch": 1.121892440385591,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.40703888169441e-06,
+      "loss": 1.41692314,
+      "memory(GiB)": 123.79,
+      "step": 44225,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.69345551,
+      "epoch": 1.1220192795535262,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.4059976711565355e-06,
+      "loss": 1.32810106,
+      "memory(GiB)": 123.79,
+      "step": 44230,
+      "train_speed(iter/s)": 1.315896
+    },
+    {
+      "acc": 0.68595433,
+      "epoch": 1.1221461187214612,
+      "grad_norm": 3.375,
+      "learning_rate": 4.404956486745532e-06,
+      "loss": 1.34487743,
+      "memory(GiB)": 123.79,
+      "step": 44235,
+      "train_speed(iter/s)": 1.315913
+    },
+    {
+      "acc": 0.68213863,
+      "epoch": 1.1222729578893962,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.403915328507201e-06,
+      "loss": 1.4189002,
+      "memory(GiB)": 123.79,
+      "step": 44240,
+      "train_speed(iter/s)": 1.315929
+    },
+    {
+      "acc": 0.68107395,
+      "epoch": 1.1223997970573314,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.4028741964873334e-06,
+      "loss": 1.37263279,
+      "memory(GiB)": 123.79,
+      "step": 44245,
+      "train_speed(iter/s)": 1.315945
+    },
+    {
+      "acc": 0.67703028,
+      "epoch": 1.1225266362252664,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.4018330907317275e-06,
+      "loss": 1.40201073,
+      "memory(GiB)": 123.79,
+      "step": 44250,
+      "train_speed(iter/s)": 1.315962
+    },
+    {
+      "acc": 0.67064095,
+      "epoch": 1.1226534753932014,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.400792011286171e-06,
+      "loss": 1.52286472,
+      "memory(GiB)": 123.79,
+      "step": 44255,
+      "train_speed(iter/s)": 1.315979
+    },
+    {
+      "acc": 0.70142498,
+      "epoch": 1.1227803145611364,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.3997509581964566e-06,
+      "loss": 1.35715485,
+      "memory(GiB)": 123.79,
+      "step": 44260,
+      "train_speed(iter/s)": 1.315995
+    },
+    {
+      "acc": 0.68720355,
+      "epoch": 1.1229071537290716,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.398709931508376e-06,
+      "loss": 1.35623808,
+      "memory(GiB)": 123.79,
+      "step": 44265,
+      "train_speed(iter/s)": 1.316012
+    },
+    {
+      "acc": 0.68752427,
+      "epoch": 1.1230339928970066,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.397668931267718e-06,
+      "loss": 1.35138216,
+      "memory(GiB)": 123.79,
+      "step": 44270,
+      "train_speed(iter/s)": 1.316028
+    },
+    {
+      "acc": 0.68275328,
+      "epoch": 1.1231608320649416,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.396627957520269e-06,
+      "loss": 1.42025843,
+      "memory(GiB)": 123.79,
+      "step": 44275,
+      "train_speed(iter/s)": 1.316044
+    },
+    {
+      "acc": 0.67475595,
+      "epoch": 1.1232876712328768,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.395587010311815e-06,
+      "loss": 1.41068344,
+      "memory(GiB)": 123.79,
+      "step": 44280,
+      "train_speed(iter/s)": 1.316061
+    },
+    {
+      "acc": 0.69146595,
+      "epoch": 1.1234145104008117,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.394546089688143e-06,
+      "loss": 1.35986481,
+      "memory(GiB)": 123.79,
+      "step": 44285,
+      "train_speed(iter/s)": 1.316077
+    },
+    {
+      "acc": 0.68709278,
+      "epoch": 1.1235413495687467,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.3935051956950395e-06,
+      "loss": 1.40167503,
+      "memory(GiB)": 123.79,
+      "step": 44290,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.68393292,
+      "epoch": 1.123668188736682,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.3924643283782824e-06,
+      "loss": 1.39432096,
+      "memory(GiB)": 123.79,
+      "step": 44295,
+      "train_speed(iter/s)": 1.31611
+    },
+    {
+      "acc": 0.69746981,
+      "epoch": 1.123795027904617,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.391423487783657e-06,
+      "loss": 1.38423624,
+      "memory(GiB)": 123.79,
+      "step": 44300,
+      "train_speed(iter/s)": 1.316127
+    },
+    {
+      "acc": 0.69562736,
+      "epoch": 1.123921867072552,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.3903826739569444e-06,
+      "loss": 1.37772636,
+      "memory(GiB)": 123.79,
+      "step": 44305,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.69565191,
+      "epoch": 1.1240487062404871,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.389341886943926e-06,
+      "loss": 1.35308495,
+      "memory(GiB)": 123.79,
+      "step": 44310,
+      "train_speed(iter/s)": 1.316158
+    },
+    {
+      "acc": 0.68205233,
+      "epoch": 1.1241755454084221,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.388301126790374e-06,
+      "loss": 1.41279087,
+      "memory(GiB)": 123.79,
+      "step": 44315,
+      "train_speed(iter/s)": 1.316174
+    },
+    {
+      "acc": 0.6891046,
+      "epoch": 1.1243023845763571,
+      "grad_norm": 3.375,
+      "learning_rate": 4.387260393542071e-06,
+      "loss": 1.38738365,
+      "memory(GiB)": 123.79,
+      "step": 44320,
+      "train_speed(iter/s)": 1.316191
+    },
+    {
+      "acc": 0.67799597,
+      "epoch": 1.1244292237442923,
+      "grad_norm": 2.875,
+      "learning_rate": 4.38621968724479e-06,
+      "loss": 1.47014904,
+      "memory(GiB)": 123.79,
+      "step": 44325,
+      "train_speed(iter/s)": 1.316207
+    },
+    {
+      "acc": 0.68619518,
+      "epoch": 1.1245560629122273,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.385179007944311e-06,
+      "loss": 1.39784336,
+      "memory(GiB)": 123.79,
+      "step": 44330,
+      "train_speed(iter/s)": 1.31622
+    },
+    {
+      "acc": 0.67993684,
+      "epoch": 1.1246829020801623,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.384138355686402e-06,
+      "loss": 1.44926567,
+      "memory(GiB)": 123.79,
+      "step": 44335,
+      "train_speed(iter/s)": 1.316237
+    },
+    {
+      "acc": 0.68464088,
+      "epoch": 1.1248097412480975,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.383097730516837e-06,
+      "loss": 1.42654839,
+      "memory(GiB)": 123.79,
+      "step": 44340,
+      "train_speed(iter/s)": 1.316254
+    },
+    {
+      "acc": 0.69987473,
+      "epoch": 1.1249365804160325,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.382057132481389e-06,
+      "loss": 1.361059,
+      "memory(GiB)": 123.79,
+      "step": 44345,
+      "train_speed(iter/s)": 1.31627
+    },
+    {
+      "acc": 0.69897642,
+      "epoch": 1.1250634195839675,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.381016561625829e-06,
+      "loss": 1.34572992,
+      "memory(GiB)": 123.79,
+      "step": 44350,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.68386412,
+      "epoch": 1.1251902587519025,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.379976017995922e-06,
+      "loss": 1.46148643,
+      "memory(GiB)": 123.79,
+      "step": 44355,
+      "train_speed(iter/s)": 1.316301
+    },
+    {
+      "acc": 0.69493084,
+      "epoch": 1.1253170979198377,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.378935501637438e-06,
+      "loss": 1.43061457,
+      "memory(GiB)": 123.79,
+      "step": 44360,
+      "train_speed(iter/s)": 1.316317
+    },
+    {
+      "acc": 0.68408561,
+      "epoch": 1.1254439370877727,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.377895012596144e-06,
+      "loss": 1.37923031,
+      "memory(GiB)": 123.79,
+      "step": 44365,
+      "train_speed(iter/s)": 1.316335
+    },
+    {
+      "acc": 0.679426,
+      "epoch": 1.1255707762557077,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.376854550917805e-06,
+      "loss": 1.42480621,
+      "memory(GiB)": 123.79,
+      "step": 44370,
+      "train_speed(iter/s)": 1.316351
+    },
+    {
+      "acc": 0.68179979,
+      "epoch": 1.1256976154236429,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.375814116648184e-06,
+      "loss": 1.43418751,
+      "memory(GiB)": 123.79,
+      "step": 44375,
+      "train_speed(iter/s)": 1.316367
+    },
+    {
+      "acc": 0.69284229,
+      "epoch": 1.1258244545915779,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.374773709833045e-06,
+      "loss": 1.37290201,
+      "memory(GiB)": 123.79,
+      "step": 44380,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.68980627,
+      "epoch": 1.1259512937595129,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.37373333051815e-06,
+      "loss": 1.35519552,
+      "memory(GiB)": 123.79,
+      "step": 44385,
+      "train_speed(iter/s)": 1.316399
+    },
+    {
+      "acc": 0.67180042,
+      "epoch": 1.126078132927448,
+      "grad_norm": 3.125,
+      "learning_rate": 4.37269297874926e-06,
+      "loss": 1.45411797,
+      "memory(GiB)": 123.79,
+      "step": 44390,
+      "train_speed(iter/s)": 1.316417
+    },
+    {
+      "acc": 0.66853204,
+      "epoch": 1.126204972095383,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.371652654572134e-06,
+      "loss": 1.46197948,
+      "memory(GiB)": 123.79,
+      "step": 44395,
+      "train_speed(iter/s)": 1.316433
+    },
+    {
+      "acc": 0.69467735,
+      "epoch": 1.126331811263318,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.370612358032529e-06,
+      "loss": 1.33775196,
+      "memory(GiB)": 123.79,
+      "step": 44400,
+      "train_speed(iter/s)": 1.31645
+    },
+    {
+      "acc": 0.68481441,
+      "epoch": 1.1264586504312533,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.369572089176201e-06,
+      "loss": 1.35670929,
+      "memory(GiB)": 123.79,
+      "step": 44405,
+      "train_speed(iter/s)": 1.316467
+    },
+    {
+      "acc": 0.68118601,
+      "epoch": 1.1265854895991883,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.3685318480489095e-06,
+      "loss": 1.37951927,
+      "memory(GiB)": 123.79,
+      "step": 44410,
+      "train_speed(iter/s)": 1.316484
+    },
+    {
+      "acc": 0.69847908,
+      "epoch": 1.1267123287671232,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.367491634696405e-06,
+      "loss": 1.3390646,
+      "memory(GiB)": 123.79,
+      "step": 44415,
+      "train_speed(iter/s)": 1.316501
+    },
+    {
+      "acc": 0.68323603,
+      "epoch": 1.1268391679350582,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.366451449164442e-06,
+      "loss": 1.34926653,
+      "memory(GiB)": 123.79,
+      "step": 44420,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.68607678,
+      "epoch": 1.1269660071029934,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.365411291498774e-06,
+      "loss": 1.38637486,
+      "memory(GiB)": 123.79,
+      "step": 44425,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.68428149,
+      "epoch": 1.1270928462709284,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.364371161745151e-06,
+      "loss": 1.40479317,
+      "memory(GiB)": 123.79,
+      "step": 44430,
+      "train_speed(iter/s)": 1.316552
+    },
+    {
+      "acc": 0.68197479,
+      "epoch": 1.1272196854388636,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.363331059949321e-06,
+      "loss": 1.36972647,
+      "memory(GiB)": 123.79,
+      "step": 44435,
+      "train_speed(iter/s)": 1.316569
+    },
+    {
+      "acc": 0.68674521,
+      "epoch": 1.1273465246067986,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.362290986157034e-06,
+      "loss": 1.38047905,
+      "memory(GiB)": 123.79,
+      "step": 44440,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.69151607,
+      "epoch": 1.1274733637747336,
+      "grad_norm": 3.5,
+      "learning_rate": 4.361250940414036e-06,
+      "loss": 1.37562809,
+      "memory(GiB)": 123.79,
+      "step": 44445,
+      "train_speed(iter/s)": 1.316602
+    },
+    {
+      "acc": 0.68955822,
+      "epoch": 1.1276002029426686,
+      "grad_norm": 3.625,
+      "learning_rate": 4.360210922766076e-06,
+      "loss": 1.38506823,
+      "memory(GiB)": 123.79,
+      "step": 44450,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "acc": 0.66538181,
+      "epoch": 1.1277270421106038,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.359170933258893e-06,
+      "loss": 1.49087982,
+      "memory(GiB)": 123.79,
+      "step": 44455,
+      "train_speed(iter/s)": 1.316635
+    },
+    {
+      "acc": 0.69851322,
+      "epoch": 1.1278538812785388,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.358130971938235e-06,
+      "loss": 1.35860033,
+      "memory(GiB)": 123.79,
+      "step": 44460,
+      "train_speed(iter/s)": 1.316653
+    },
+    {
+      "acc": 0.67588162,
+      "epoch": 1.1279807204464738,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.357091038849841e-06,
+      "loss": 1.43883781,
+      "memory(GiB)": 123.79,
+      "step": 44465,
+      "train_speed(iter/s)": 1.31667
+    },
+    {
+      "acc": 0.6994586,
+      "epoch": 1.128107559614409,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.356051134039455e-06,
+      "loss": 1.35534534,
+      "memory(GiB)": 123.79,
+      "step": 44470,
+      "train_speed(iter/s)": 1.316687
+    },
+    {
+      "acc": 0.69010324,
+      "epoch": 1.128234398782344,
+      "grad_norm": 2.75,
+      "learning_rate": 4.3550112575528155e-06,
+      "loss": 1.35413055,
+      "memory(GiB)": 123.79,
+      "step": 44475,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.70430264,
+      "epoch": 1.128361237950279,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.353971409435659e-06,
+      "loss": 1.27094164,
+      "memory(GiB)": 123.79,
+      "step": 44480,
+      "train_speed(iter/s)": 1.31672
+    },
+    {
+      "acc": 0.69885311,
+      "epoch": 1.1284880771182142,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.352931589733725e-06,
+      "loss": 1.33113661,
+      "memory(GiB)": 123.79,
+      "step": 44485,
+      "train_speed(iter/s)": 1.316736
+    },
+    {
+      "acc": 0.68444219,
+      "epoch": 1.1286149162861492,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.35189179849275e-06,
+      "loss": 1.42936325,
+      "memory(GiB)": 123.79,
+      "step": 44490,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.68726006,
+      "epoch": 1.1287417554540842,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.350852035758466e-06,
+      "loss": 1.47977037,
+      "memory(GiB)": 123.79,
+      "step": 44495,
+      "train_speed(iter/s)": 1.316769
+    },
+    {
+      "acc": 0.68143129,
+      "epoch": 1.1288685946220194,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.3498123015766066e-06,
+      "loss": 1.38942947,
+      "memory(GiB)": 123.79,
+      "step": 44500,
+      "train_speed(iter/s)": 1.316786
+    },
+    {
+      "acc": 0.69526119,
+      "epoch": 1.1289954337899544,
+      "grad_norm": 4.125,
+      "learning_rate": 4.348772595992906e-06,
+      "loss": 1.3788456,
+      "memory(GiB)": 123.79,
+      "step": 44505,
+      "train_speed(iter/s)": 1.316802
+    },
+    {
+      "acc": 0.68287063,
+      "epoch": 1.1291222729578894,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.347732919053096e-06,
+      "loss": 1.38505516,
+      "memory(GiB)": 123.79,
+      "step": 44510,
+      "train_speed(iter/s)": 1.316818
+    },
+    {
+      "acc": 0.6791708,
+      "epoch": 1.1292491121258244,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.346693270802902e-06,
+      "loss": 1.41060638,
+      "memory(GiB)": 123.79,
+      "step": 44515,
+      "train_speed(iter/s)": 1.316834
+    },
+    {
+      "acc": 0.69244242,
+      "epoch": 1.1293759512937596,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.345653651288055e-06,
+      "loss": 1.33822784,
+      "memory(GiB)": 123.79,
+      "step": 44520,
+      "train_speed(iter/s)": 1.31685
+    },
+    {
+      "acc": 0.68143239,
+      "epoch": 1.1295027904616946,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.344614060554281e-06,
+      "loss": 1.37334003,
+      "memory(GiB)": 123.79,
+      "step": 44525,
+      "train_speed(iter/s)": 1.316867
+    },
+    {
+      "acc": 0.69207206,
+      "epoch": 1.1296296296296295,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.343574498647311e-06,
+      "loss": 1.39023485,
+      "memory(GiB)": 123.79,
+      "step": 44530,
+      "train_speed(iter/s)": 1.316885
+    },
+    {
+      "acc": 0.69315815,
+      "epoch": 1.1297564687975648,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.342534965612861e-06,
+      "loss": 1.32656479,
+      "memory(GiB)": 123.79,
+      "step": 44535,
+      "train_speed(iter/s)": 1.316902
+    },
+    {
+      "acc": 0.68727207,
+      "epoch": 1.1298833079654997,
+      "grad_norm": 3.25,
+      "learning_rate": 4.34149546149666e-06,
+      "loss": 1.37593985,
+      "memory(GiB)": 123.79,
+      "step": 44540,
+      "train_speed(iter/s)": 1.31692
+    },
+    {
+      "acc": 0.68395529,
+      "epoch": 1.1300101471334347,
+      "grad_norm": 2.875,
+      "learning_rate": 4.340455986344428e-06,
+      "loss": 1.4643014,
+      "memory(GiB)": 123.79,
+      "step": 44545,
+      "train_speed(iter/s)": 1.316937
+    },
+    {
+      "acc": 0.69232593,
+      "epoch": 1.13013698630137,
+      "grad_norm": 3.125,
+      "learning_rate": 4.3394165402018875e-06,
+      "loss": 1.42181416,
+      "memory(GiB)": 123.79,
+      "step": 44550,
+      "train_speed(iter/s)": 1.316954
+    },
+    {
+      "acc": 0.69723835,
+      "epoch": 1.130263825469305,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.338377123114757e-06,
+      "loss": 1.36976337,
+      "memory(GiB)": 123.79,
+      "step": 44555,
+      "train_speed(iter/s)": 1.316969
+    },
+    {
+      "acc": 0.67722206,
+      "epoch": 1.13039066463724,
+      "grad_norm": 2.75,
+      "learning_rate": 4.337337735128752e-06,
+      "loss": 1.43957958,
+      "memory(GiB)": 123.79,
+      "step": 44560,
+      "train_speed(iter/s)": 1.316987
+    },
+    {
+      "acc": 0.68419151,
+      "epoch": 1.1305175038051751,
+      "grad_norm": 2.375,
+      "learning_rate": 4.336298376289594e-06,
+      "loss": 1.40198345,
+      "memory(GiB)": 123.79,
+      "step": 44565,
+      "train_speed(iter/s)": 1.317003
+    },
+    {
+      "acc": 0.66863756,
+      "epoch": 1.1306443429731101,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.335259046642998e-06,
+      "loss": 1.47091522,
+      "memory(GiB)": 123.79,
+      "step": 44570,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.6854219,
+      "epoch": 1.130771182141045,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.334219746234675e-06,
+      "loss": 1.39533415,
+      "memory(GiB)": 123.79,
+      "step": 44575,
+      "train_speed(iter/s)": 1.317038
+    },
+    {
+      "acc": 0.6811182,
+      "epoch": 1.13089802130898,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.3331804751103395e-06,
+      "loss": 1.37467384,
+      "memory(GiB)": 123.79,
+      "step": 44580,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.67688131,
+      "epoch": 1.1310248604769153,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.332141233315705e-06,
+      "loss": 1.3966855,
+      "memory(GiB)": 123.79,
+      "step": 44585,
+      "train_speed(iter/s)": 1.317071
+    },
+    {
+      "acc": 0.69824147,
+      "epoch": 1.1311516996448503,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.331102020896482e-06,
+      "loss": 1.37992077,
+      "memory(GiB)": 123.79,
+      "step": 44590,
+      "train_speed(iter/s)": 1.317089
+    },
+    {
+      "acc": 0.70346117,
+      "epoch": 1.1312785388127855,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.330062837898376e-06,
+      "loss": 1.30969629,
+      "memory(GiB)": 123.79,
+      "step": 44595,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.6818316,
+      "epoch": 1.1314053779807205,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.3290236843670985e-06,
+      "loss": 1.41158543,
+      "memory(GiB)": 123.79,
+      "step": 44600,
+      "train_speed(iter/s)": 1.317122
+    },
+    {
+      "acc": 0.68802128,
+      "epoch": 1.1315322171486555,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.327984560348354e-06,
+      "loss": 1.3974287,
+      "memory(GiB)": 123.79,
+      "step": 44605,
+      "train_speed(iter/s)": 1.317139
+    },
+    {
+      "acc": 0.69894724,
+      "epoch": 1.1316590563165905,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.3269454658878516e-06,
+      "loss": 1.35712891,
+      "memory(GiB)": 123.79,
+      "step": 44610,
+      "train_speed(iter/s)": 1.317154
+    },
+    {
+      "acc": 0.69310689,
+      "epoch": 1.1317858954845257,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.325906401031291e-06,
+      "loss": 1.39916344,
+      "memory(GiB)": 123.79,
+      "step": 44615,
+      "train_speed(iter/s)": 1.31717
+    },
+    {
+      "acc": 0.68835516,
+      "epoch": 1.1319127346524607,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.324867365824376e-06,
+      "loss": 1.37143192,
+      "memory(GiB)": 123.79,
+      "step": 44620,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.69221935,
+      "epoch": 1.1320395738203957,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.323828360312809e-06,
+      "loss": 1.38473949,
+      "memory(GiB)": 123.79,
+      "step": 44625,
+      "train_speed(iter/s)": 1.317204
+    },
+    {
+      "acc": 0.68757277,
+      "epoch": 1.1321664129883309,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.32278938454229e-06,
+      "loss": 1.35827284,
+      "memory(GiB)": 123.79,
+      "step": 44630,
+      "train_speed(iter/s)": 1.31722
+    },
+    {
+      "acc": 0.68841977,
+      "epoch": 1.1322932521562659,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.321750438558517e-06,
+      "loss": 1.3176013,
+      "memory(GiB)": 123.79,
+      "step": 44635,
+      "train_speed(iter/s)": 1.317237
+    },
+    {
+      "acc": 0.67965827,
+      "epoch": 1.1324200913242009,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.3207115224071874e-06,
+      "loss": 1.3321579,
+      "memory(GiB)": 123.79,
+      "step": 44640,
+      "train_speed(iter/s)": 1.317253
+    },
+    {
+      "acc": 0.67016101,
+      "epoch": 1.132546930492136,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.319672636133998e-06,
+      "loss": 1.45132847,
+      "memory(GiB)": 123.79,
+      "step": 44645,
+      "train_speed(iter/s)": 1.317268
+    },
+    {
+      "acc": 0.69404073,
+      "epoch": 1.132673769660071,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.318633779784646e-06,
+      "loss": 1.41747665,
+      "memory(GiB)": 123.79,
+      "step": 44650,
+      "train_speed(iter/s)": 1.317284
+    },
+    {
+      "acc": 0.68027573,
+      "epoch": 1.132800608828006,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.317594953404818e-06,
+      "loss": 1.44843769,
+      "memory(GiB)": 123.79,
+      "step": 44655,
+      "train_speed(iter/s)": 1.3173
+    },
+    {
+      "acc": 0.68813629,
+      "epoch": 1.1329274479959413,
+      "grad_norm": 2.5,
+      "learning_rate": 4.316556157040213e-06,
+      "loss": 1.35786762,
+      "memory(GiB)": 123.79,
+      "step": 44660,
+      "train_speed(iter/s)": 1.317317
+    },
+    {
+      "acc": 0.6657959,
+      "epoch": 1.1330542871638762,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.315517390736519e-06,
+      "loss": 1.44658337,
+      "memory(GiB)": 123.79,
+      "step": 44665,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.67379928,
+      "epoch": 1.1331811263318112,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.314478654539429e-06,
+      "loss": 1.39759026,
+      "memory(GiB)": 123.79,
+      "step": 44670,
+      "train_speed(iter/s)": 1.317349
+    },
+    {
+      "acc": 0.68398809,
+      "epoch": 1.1333079654997462,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.313439948494625e-06,
+      "loss": 1.41816254,
+      "memory(GiB)": 123.79,
+      "step": 44675,
+      "train_speed(iter/s)": 1.317365
+    },
+    {
+      "acc": 0.69320297,
+      "epoch": 1.1334348046676814,
+      "grad_norm": 3.75,
+      "learning_rate": 4.312401272647799e-06,
+      "loss": 1.36620493,
+      "memory(GiB)": 123.79,
+      "step": 44680,
+      "train_speed(iter/s)": 1.31738
+    },
+    {
+      "acc": 0.68760409,
+      "epoch": 1.1335616438356164,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.311362627044633e-06,
+      "loss": 1.39485321,
+      "memory(GiB)": 123.79,
+      "step": 44685,
+      "train_speed(iter/s)": 1.317396
+    },
+    {
+      "acc": 0.68588495,
+      "epoch": 1.1336884830035514,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.310324011730816e-06,
+      "loss": 1.42441425,
+      "memory(GiB)": 123.79,
+      "step": 44690,
+      "train_speed(iter/s)": 1.317412
+    },
+    {
+      "acc": 0.68157635,
+      "epoch": 1.1338153221714866,
+      "grad_norm": 3.25,
+      "learning_rate": 4.309285426752027e-06,
+      "loss": 1.43535175,
+      "memory(GiB)": 123.79,
+      "step": 44695,
+      "train_speed(iter/s)": 1.317428
+    },
+    {
+      "acc": 0.68867235,
+      "epoch": 1.1339421613394216,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.308246872153947e-06,
+      "loss": 1.40884323,
+      "memory(GiB)": 123.79,
+      "step": 44700,
+      "train_speed(iter/s)": 1.317445
+    },
+    {
+      "acc": 0.69383783,
+      "epoch": 1.1340690005073566,
+      "grad_norm": 2.875,
+      "learning_rate": 4.307208347982259e-06,
+      "loss": 1.29440765,
+      "memory(GiB)": 123.79,
+      "step": 44705,
+      "train_speed(iter/s)": 1.317461
+    },
+    {
+      "acc": 0.69796276,
+      "epoch": 1.1341958396752918,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.306169854282643e-06,
+      "loss": 1.32976789,
+      "memory(GiB)": 123.79,
+      "step": 44710,
+      "train_speed(iter/s)": 1.317476
+    },
+    {
+      "acc": 0.69277043,
+      "epoch": 1.1343226788432268,
+      "grad_norm": 3.8125,
+      "learning_rate": 4.305131391100773e-06,
+      "loss": 1.38635321,
+      "memory(GiB)": 123.79,
+      "step": 44715,
+      "train_speed(iter/s)": 1.317492
+    },
+    {
+      "acc": 0.69891005,
+      "epoch": 1.1344495180111618,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.304092958482325e-06,
+      "loss": 1.35851326,
+      "memory(GiB)": 123.79,
+      "step": 44720,
+      "train_speed(iter/s)": 1.317508
+    },
+    {
+      "acc": 0.69107089,
+      "epoch": 1.134576357179097,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.303054556472978e-06,
+      "loss": 1.3697773,
+      "memory(GiB)": 123.79,
+      "step": 44725,
+      "train_speed(iter/s)": 1.317524
+    },
+    {
+      "acc": 0.68794098,
+      "epoch": 1.134703196347032,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.3020161851184036e-06,
+      "loss": 1.40984287,
+      "memory(GiB)": 123.79,
+      "step": 44730,
+      "train_speed(iter/s)": 1.31754
+    },
+    {
+      "acc": 0.67686095,
+      "epoch": 1.134830035514967,
+      "grad_norm": 3.84375,
+      "learning_rate": 4.300977844464273e-06,
+      "loss": 1.41458578,
+      "memory(GiB)": 123.79,
+      "step": 44735,
+      "train_speed(iter/s)": 1.317555
+    },
+    {
+      "acc": 0.68957667,
+      "epoch": 1.134956874682902,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.2999395345562564e-06,
+      "loss": 1.33998661,
+      "memory(GiB)": 123.79,
+      "step": 44740,
+      "train_speed(iter/s)": 1.317571
+    },
+    {
+      "acc": 0.68236275,
+      "epoch": 1.1350837138508372,
+      "grad_norm": 4.875,
+      "learning_rate": 4.298901255440025e-06,
+      "loss": 1.41009483,
+      "memory(GiB)": 123.79,
+      "step": 44745,
+      "train_speed(iter/s)": 1.317588
+    },
+    {
+      "acc": 0.68086467,
+      "epoch": 1.1352105530187722,
+      "grad_norm": 4.375,
+      "learning_rate": 4.297863007161249e-06,
+      "loss": 1.39772129,
+      "memory(GiB)": 123.79,
+      "step": 44750,
+      "train_speed(iter/s)": 1.317604
+    },
+    {
+      "acc": 0.68363042,
+      "epoch": 1.1353373921867074,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.29682478976559e-06,
+      "loss": 1.34798889,
+      "memory(GiB)": 123.79,
+      "step": 44755,
+      "train_speed(iter/s)": 1.31762
+    },
+    {
+      "acc": 0.68695006,
+      "epoch": 1.1354642313546424,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.295786603298717e-06,
+      "loss": 1.37538366,
+      "memory(GiB)": 123.79,
+      "step": 44760,
+      "train_speed(iter/s)": 1.317635
+    },
+    {
+      "acc": 0.68764162,
+      "epoch": 1.1355910705225774,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.294748447806293e-06,
+      "loss": 1.35435982,
+      "memory(GiB)": 123.79,
+      "step": 44765,
+      "train_speed(iter/s)": 1.317652
+    },
+    {
+      "acc": 0.68750396,
+      "epoch": 1.1357179096905123,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.293710323333983e-06,
+      "loss": 1.38184834,
+      "memory(GiB)": 123.79,
+      "step": 44770,
+      "train_speed(iter/s)": 1.317667
+    },
+    {
+      "acc": 0.69202542,
+      "epoch": 1.1358447488584476,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.292672229927445e-06,
+      "loss": 1.35606213,
+      "memory(GiB)": 123.79,
+      "step": 44775,
+      "train_speed(iter/s)": 1.317683
+    },
+    {
+      "acc": 0.69766822,
+      "epoch": 1.1359715880263825,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.2916341676323386e-06,
+      "loss": 1.36209202,
+      "memory(GiB)": 123.79,
+      "step": 44780,
+      "train_speed(iter/s)": 1.317699
+    },
+    {
+      "acc": 0.69854546,
+      "epoch": 1.1360984271943175,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.290596136494326e-06,
+      "loss": 1.31507454,
+      "memory(GiB)": 123.79,
+      "step": 44785,
+      "train_speed(iter/s)": 1.317714
+    },
+    {
+      "acc": 0.67641129,
+      "epoch": 1.1362252663622527,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.289558136559063e-06,
+      "loss": 1.46254234,
+      "memory(GiB)": 123.79,
+      "step": 44790,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "acc": 0.68940015,
+      "epoch": 1.1363521055301877,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.288520167872203e-06,
+      "loss": 1.36848831,
+      "memory(GiB)": 123.79,
+      "step": 44795,
+      "train_speed(iter/s)": 1.317747
+    },
+    {
+      "acc": 0.67278895,
+      "epoch": 1.1364789446981227,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.287482230479404e-06,
+      "loss": 1.49391994,
+      "memory(GiB)": 123.79,
+      "step": 44800,
+      "train_speed(iter/s)": 1.317764
+    },
+    {
+      "acc": 0.67567158,
+      "epoch": 1.136605783866058,
+      "grad_norm": 3.375,
+      "learning_rate": 4.286444324426318e-06,
+      "loss": 1.41341801,
+      "memory(GiB)": 123.79,
+      "step": 44805,
+      "train_speed(iter/s)": 1.31778
+    },
+    {
+      "acc": 0.68856001,
+      "epoch": 1.136732623033993,
+      "grad_norm": 2.875,
+      "learning_rate": 4.2854064497585964e-06,
+      "loss": 1.36586618,
+      "memory(GiB)": 123.79,
+      "step": 44810,
+      "train_speed(iter/s)": 1.317796
+    },
+    {
+      "acc": 0.68177586,
+      "epoch": 1.136859462201928,
+      "grad_norm": 3.0,
+      "learning_rate": 4.284368606521888e-06,
+      "loss": 1.44221058,
+      "memory(GiB)": 123.79,
+      "step": 44815,
+      "train_speed(iter/s)": 1.317812
+    },
+    {
+      "acc": 0.70322514,
+      "epoch": 1.1369863013698631,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.283330794761845e-06,
+      "loss": 1.3921011,
+      "memory(GiB)": 123.79,
+      "step": 44820,
+      "train_speed(iter/s)": 1.317828
+    },
+    {
+      "acc": 0.68967686,
+      "epoch": 1.137113140537798,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.282293014524112e-06,
+      "loss": 1.37903824,
+      "memory(GiB)": 123.79,
+      "step": 44825,
+      "train_speed(iter/s)": 1.317843
+    },
+    {
+      "acc": 0.6847971,
+      "epoch": 1.137239979705733,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.281255265854338e-06,
+      "loss": 1.38750858,
+      "memory(GiB)": 123.79,
+      "step": 44830,
+      "train_speed(iter/s)": 1.317859
+    },
+    {
+      "acc": 0.69452977,
+      "epoch": 1.137366818873668,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.280217548798166e-06,
+      "loss": 1.33333044,
+      "memory(GiB)": 123.79,
+      "step": 44835,
+      "train_speed(iter/s)": 1.317876
+    },
+    {
+      "acc": 0.6805563,
+      "epoch": 1.1374936580416033,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.279179863401239e-06,
+      "loss": 1.39424257,
+      "memory(GiB)": 123.79,
+      "step": 44840,
+      "train_speed(iter/s)": 1.317892
+    },
+    {
+      "acc": 0.71016798,
+      "epoch": 1.1376204972095383,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.278142209709199e-06,
+      "loss": 1.31807508,
+      "memory(GiB)": 123.79,
+      "step": 44845,
+      "train_speed(iter/s)": 1.317908
+    },
+    {
+      "acc": 0.6905901,
+      "epoch": 1.1377473363774733,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.277104587767691e-06,
+      "loss": 1.42355719,
+      "memory(GiB)": 123.79,
+      "step": 44850,
+      "train_speed(iter/s)": 1.317924
+    },
+    {
+      "acc": 0.6789341,
+      "epoch": 1.1378741755454085,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.276066997622348e-06,
+      "loss": 1.41117373,
+      "memory(GiB)": 123.79,
+      "step": 44855,
+      "train_speed(iter/s)": 1.317941
+    },
+    {
+      "acc": 0.70262585,
+      "epoch": 1.1380010147133435,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.27502943931881e-06,
+      "loss": 1.34011354,
+      "memory(GiB)": 123.79,
+      "step": 44860,
+      "train_speed(iter/s)": 1.317957
+    },
+    {
+      "acc": 0.67702789,
+      "epoch": 1.1381278538812785,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.273991912902716e-06,
+      "loss": 1.42313805,
+      "memory(GiB)": 123.79,
+      "step": 44865,
+      "train_speed(iter/s)": 1.317973
+    },
+    {
+      "acc": 0.67530322,
+      "epoch": 1.1382546930492137,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.272954418419699e-06,
+      "loss": 1.40532904,
+      "memory(GiB)": 123.79,
+      "step": 44870,
+      "train_speed(iter/s)": 1.317989
+    },
+    {
+      "acc": 0.6971293,
+      "epoch": 1.1383815322171487,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.2719169559153905e-06,
+      "loss": 1.32524452,
+      "memory(GiB)": 123.79,
+      "step": 44875,
+      "train_speed(iter/s)": 1.318005
+    },
+    {
+      "acc": 0.69921679,
+      "epoch": 1.1385083713850837,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.270879525435426e-06,
+      "loss": 1.37865133,
+      "memory(GiB)": 123.79,
+      "step": 44880,
+      "train_speed(iter/s)": 1.318019
+    },
+    {
+      "acc": 0.6916595,
+      "epoch": 1.1386352105530189,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.269842127025435e-06,
+      "loss": 1.4049943,
+      "memory(GiB)": 123.79,
+      "step": 44885,
+      "train_speed(iter/s)": 1.318035
+    },
+    {
+      "acc": 0.68622227,
+      "epoch": 1.1387620497209539,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.2688047607310504e-06,
+      "loss": 1.3929512,
+      "memory(GiB)": 123.79,
+      "step": 44890,
+      "train_speed(iter/s)": 1.31805
+    },
+    {
+      "acc": 0.67426171,
+      "epoch": 1.1388888888888888,
+      "grad_norm": 3.125,
+      "learning_rate": 4.267767426597893e-06,
+      "loss": 1.4453207,
+      "memory(GiB)": 123.79,
+      "step": 44895,
+      "train_speed(iter/s)": 1.318067
+    },
+    {
+      "acc": 0.68227086,
+      "epoch": 1.1390157280568238,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.266730124671594e-06,
+      "loss": 1.38309326,
+      "memory(GiB)": 123.79,
+      "step": 44900,
+      "train_speed(iter/s)": 1.318082
+    },
+    {
+      "acc": 0.69433851,
+      "epoch": 1.139142567224759,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.265692854997778e-06,
+      "loss": 1.35260563,
+      "memory(GiB)": 123.79,
+      "step": 44905,
+      "train_speed(iter/s)": 1.318098
+    },
+    {
+      "acc": 0.70649867,
+      "epoch": 1.139269406392694,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.2646556176220714e-06,
+      "loss": 1.34563456,
+      "memory(GiB)": 123.79,
+      "step": 44910,
+      "train_speed(iter/s)": 1.318115
+    },
+    {
+      "acc": 0.69332962,
+      "epoch": 1.1393962455606292,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.263618412590092e-06,
+      "loss": 1.34279594,
+      "memory(GiB)": 123.79,
+      "step": 44915,
+      "train_speed(iter/s)": 1.318131
+    },
+    {
+      "acc": 0.67388754,
+      "epoch": 1.1395230847285642,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.2625812399474604e-06,
+      "loss": 1.42907887,
+      "memory(GiB)": 123.79,
+      "step": 44920,
+      "train_speed(iter/s)": 1.318148
+    },
+    {
+      "acc": 0.69165812,
+      "epoch": 1.1396499238964992,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.2615440997398e-06,
+      "loss": 1.40199833,
+      "memory(GiB)": 123.79,
+      "step": 44925,
+      "train_speed(iter/s)": 1.318164
+    },
+    {
+      "acc": 0.69369798,
+      "epoch": 1.1397767630644342,
+      "grad_norm": 3.25,
+      "learning_rate": 4.2605069920127284e-06,
+      "loss": 1.3238596,
+      "memory(GiB)": 123.79,
+      "step": 44930,
+      "train_speed(iter/s)": 1.31818
+    },
+    {
+      "acc": 0.69295931,
+      "epoch": 1.1399036022323694,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.25946991681186e-06,
+      "loss": 1.3791604,
+      "memory(GiB)": 123.79,
+      "step": 44935,
+      "train_speed(iter/s)": 1.318197
+    },
+    {
+      "acc": 0.68274708,
+      "epoch": 1.1400304414003044,
+      "grad_norm": 3.0,
+      "learning_rate": 4.258432874182809e-06,
+      "loss": 1.399998,
+      "memory(GiB)": 123.79,
+      "step": 44940,
+      "train_speed(iter/s)": 1.318213
+    },
+    {
+      "acc": 0.68463926,
+      "epoch": 1.1401572805682394,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.2573958641711925e-06,
+      "loss": 1.41205635,
+      "memory(GiB)": 123.79,
+      "step": 44945,
+      "train_speed(iter/s)": 1.318229
+    },
+    {
+      "acc": 0.69745517,
+      "epoch": 1.1402841197361746,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.256358886822622e-06,
+      "loss": 1.32006426,
+      "memory(GiB)": 123.79,
+      "step": 44950,
+      "train_speed(iter/s)": 1.318245
+    },
+    {
+      "acc": 0.69571838,
+      "epoch": 1.1404109589041096,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.255321942182707e-06,
+      "loss": 1.35269909,
+      "memory(GiB)": 123.79,
+      "step": 44955,
+      "train_speed(iter/s)": 1.318261
+    },
+    {
+      "acc": 0.68500166,
+      "epoch": 1.1405377980720446,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.254285030297058e-06,
+      "loss": 1.40707035,
+      "memory(GiB)": 123.79,
+      "step": 44960,
+      "train_speed(iter/s)": 1.318277
+    },
+    {
+      "acc": 0.69269166,
+      "epoch": 1.1406646372399798,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.2532481512112814e-06,
+      "loss": 1.41774635,
+      "memory(GiB)": 123.79,
+      "step": 44965,
+      "train_speed(iter/s)": 1.318293
+    },
+    {
+      "acc": 0.68334055,
+      "epoch": 1.1407914764079148,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.252211304970988e-06,
+      "loss": 1.4231432,
+      "memory(GiB)": 123.79,
+      "step": 44970,
+      "train_speed(iter/s)": 1.31831
+    },
+    {
+      "acc": 0.68749862,
+      "epoch": 1.1409183155758498,
+      "grad_norm": 3.625,
+      "learning_rate": 4.251174491621778e-06,
+      "loss": 1.38407326,
+      "memory(GiB)": 123.79,
+      "step": 44975,
+      "train_speed(iter/s)": 1.318325
+    },
+    {
+      "acc": 0.68363562,
+      "epoch": 1.141045154743785,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.250137711209258e-06,
+      "loss": 1.41711092,
+      "memory(GiB)": 123.79,
+      "step": 44980,
+      "train_speed(iter/s)": 1.318341
+    },
+    {
+      "acc": 0.68469477,
+      "epoch": 1.14117199391172,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.249100963779028e-06,
+      "loss": 1.49483404,
+      "memory(GiB)": 123.79,
+      "step": 44985,
+      "train_speed(iter/s)": 1.318357
+    },
+    {
+      "acc": 0.69281292,
+      "epoch": 1.141298833079655,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.248064249376692e-06,
+      "loss": 1.38999939,
+      "memory(GiB)": 123.79,
+      "step": 44990,
+      "train_speed(iter/s)": 1.318374
+    },
+    {
+      "acc": 0.68423243,
+      "epoch": 1.14142567224759,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.2470275680478466e-06,
+      "loss": 1.37125912,
+      "memory(GiB)": 123.79,
+      "step": 44995,
+      "train_speed(iter/s)": 1.318389
+    },
+    {
+      "acc": 0.69599581,
+      "epoch": 1.1415525114155252,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.2459909198380886e-06,
+      "loss": 1.32638655,
+      "memory(GiB)": 123.79,
+      "step": 45000,
+      "train_speed(iter/s)": 1.318406
+    },
+    {
+      "epoch": 1.1415525114155252,
+      "eval_acc": 0.6750136261973825,
+      "eval_loss": 1.3565222024917603,
+      "eval_runtime": 70.1387,
+      "eval_samples_per_second": 90.82,
+      "eval_steps_per_second": 22.712,
+      "step": 45000
+    },
+    {
+      "acc": 0.6981957,
+      "epoch": 1.1416793505834602,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.244954304793019e-06,
+      "loss": 1.34153481,
+      "memory(GiB)": 123.79,
+      "step": 45005,
+      "train_speed(iter/s)": 1.315229
+    },
+    {
+      "acc": 0.68599043,
+      "epoch": 1.1418061897513951,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.2439177229582304e-06,
+      "loss": 1.4186615,
+      "memory(GiB)": 123.79,
+      "step": 45010,
+      "train_speed(iter/s)": 1.315244
+    },
+    {
+      "acc": 0.68576517,
+      "epoch": 1.1419330289193304,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.242881174379313e-06,
+      "loss": 1.38615646,
+      "memory(GiB)": 123.79,
+      "step": 45015,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.67807193,
+      "epoch": 1.1420598680872653,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.241844659101865e-06,
+      "loss": 1.38319998,
+      "memory(GiB)": 123.79,
+      "step": 45020,
+      "train_speed(iter/s)": 1.315274
+    },
+    {
+      "acc": 0.67076306,
+      "epoch": 1.1421867072552003,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.240808177171472e-06,
+      "loss": 1.50005827,
+      "memory(GiB)": 123.79,
+      "step": 45025,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.69184365,
+      "epoch": 1.1423135464231355,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.239771728633727e-06,
+      "loss": 1.35440102,
+      "memory(GiB)": 123.79,
+      "step": 45030,
+      "train_speed(iter/s)": 1.315306
+    },
+    {
+      "acc": 0.65329962,
+      "epoch": 1.1424403855910705,
+      "grad_norm": 2.75,
+      "learning_rate": 4.238735313534213e-06,
+      "loss": 1.49013233,
+      "memory(GiB)": 123.79,
+      "step": 45035,
+      "train_speed(iter/s)": 1.315321
+    },
+    {
+      "acc": 0.68418627,
+      "epoch": 1.1425672247590055,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.23769893191852e-06,
+      "loss": 1.42814922,
+      "memory(GiB)": 123.79,
+      "step": 45040,
+      "train_speed(iter/s)": 1.315336
+    },
+    {
+      "acc": 0.69346352,
+      "epoch": 1.1426940639269407,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.236662583832229e-06,
+      "loss": 1.31462154,
+      "memory(GiB)": 123.79,
+      "step": 45045,
+      "train_speed(iter/s)": 1.315351
+    },
+    {
+      "acc": 0.6822154,
+      "epoch": 1.1428209030948757,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.23562626932093e-06,
+      "loss": 1.51610203,
+      "memory(GiB)": 123.79,
+      "step": 45050,
+      "train_speed(iter/s)": 1.315368
+    },
+    {
+      "acc": 0.67797165,
+      "epoch": 1.1429477422628107,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.234589988430198e-06,
+      "loss": 1.42678404,
+      "memory(GiB)": 123.79,
+      "step": 45055,
+      "train_speed(iter/s)": 1.315381
+    },
+    {
+      "acc": 0.68139868,
+      "epoch": 1.1430745814307457,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.233553741205615e-06,
+      "loss": 1.43318367,
+      "memory(GiB)": 123.79,
+      "step": 45060,
+      "train_speed(iter/s)": 1.315397
+    },
+    {
+      "acc": 0.67469044,
+      "epoch": 1.143201420598681,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.2325175276927614e-06,
+      "loss": 1.38667507,
+      "memory(GiB)": 123.79,
+      "step": 45065,
+      "train_speed(iter/s)": 1.315413
+    },
+    {
+      "acc": 0.68681936,
+      "epoch": 1.143328259766616,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.231481347937214e-06,
+      "loss": 1.36011496,
+      "memory(GiB)": 123.79,
+      "step": 45070,
+      "train_speed(iter/s)": 1.315428
+    },
+    {
+      "acc": 0.68772659,
+      "epoch": 1.143455098934551,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.230445201984547e-06,
+      "loss": 1.35273914,
+      "memory(GiB)": 123.79,
+      "step": 45075,
+      "train_speed(iter/s)": 1.315438
+    },
+    {
+      "acc": 0.68294764,
+      "epoch": 1.143581938102486,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.229409089880336e-06,
+      "loss": 1.39196138,
+      "memory(GiB)": 123.79,
+      "step": 45080,
+      "train_speed(iter/s)": 1.315454
+    },
+    {
+      "acc": 0.68266687,
+      "epoch": 1.143708777270421,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.2283730116701535e-06,
+      "loss": 1.42054567,
+      "memory(GiB)": 123.79,
+      "step": 45085,
+      "train_speed(iter/s)": 1.31547
+    },
+    {
+      "acc": 0.68786201,
+      "epoch": 1.143835616438356,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.227336967399573e-06,
+      "loss": 1.39546289,
+      "memory(GiB)": 123.79,
+      "step": 45090,
+      "train_speed(iter/s)": 1.315487
+    },
+    {
+      "acc": 0.69755268,
+      "epoch": 1.1439624556062913,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.2263009571141585e-06,
+      "loss": 1.36000443,
+      "memory(GiB)": 123.79,
+      "step": 45095,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.69338527,
+      "epoch": 1.1440892947742263,
+      "grad_norm": 2.875,
+      "learning_rate": 4.225264980859485e-06,
+      "loss": 1.37565622,
+      "memory(GiB)": 123.79,
+      "step": 45100,
+      "train_speed(iter/s)": 1.315521
+    },
+    {
+      "acc": 0.68009253,
+      "epoch": 1.1442161339421613,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.224229038681115e-06,
+      "loss": 1.43576632,
+      "memory(GiB)": 123.79,
+      "step": 45105,
+      "train_speed(iter/s)": 1.315537
+    },
+    {
+      "acc": 0.67971902,
+      "epoch": 1.1443429731100965,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.223193130624619e-06,
+      "loss": 1.42862568,
+      "memory(GiB)": 123.79,
+      "step": 45110,
+      "train_speed(iter/s)": 1.315554
+    },
+    {
+      "acc": 0.70060105,
+      "epoch": 1.1444698122780315,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.222157256735553e-06,
+      "loss": 1.30654736,
+      "memory(GiB)": 123.79,
+      "step": 45115,
+      "train_speed(iter/s)": 1.31557
+    },
+    {
+      "acc": 0.70084791,
+      "epoch": 1.1445966514459665,
+      "grad_norm": 3.375,
+      "learning_rate": 4.2211214170594865e-06,
+      "loss": 1.38603172,
+      "memory(GiB)": 123.79,
+      "step": 45120,
+      "train_speed(iter/s)": 1.315586
+    },
+    {
+      "acc": 0.67407532,
+      "epoch": 1.1447234906139017,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.220085611641976e-06,
+      "loss": 1.41925898,
+      "memory(GiB)": 123.79,
+      "step": 45125,
+      "train_speed(iter/s)": 1.315601
+    },
+    {
+      "acc": 0.6852767,
+      "epoch": 1.1448503297818367,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.2190498405285826e-06,
+      "loss": 1.42612495,
+      "memory(GiB)": 123.79,
+      "step": 45130,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.68782563,
+      "epoch": 1.1449771689497716,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.218014103764865e-06,
+      "loss": 1.38401222,
+      "memory(GiB)": 123.79,
+      "step": 45135,
+      "train_speed(iter/s)": 1.315632
+    },
+    {
+      "acc": 0.67040863,
+      "epoch": 1.1451040081177069,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.216978401396376e-06,
+      "loss": 1.43275471,
+      "memory(GiB)": 123.79,
+      "step": 45140,
+      "train_speed(iter/s)": 1.315647
+    },
+    {
+      "acc": 0.67955132,
+      "epoch": 1.1452308472856418,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.215942733468675e-06,
+      "loss": 1.44383421,
+      "memory(GiB)": 123.79,
+      "step": 45145,
+      "train_speed(iter/s)": 1.315663
+    },
+    {
+      "acc": 0.70213118,
+      "epoch": 1.1453576864535768,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.2149071000273134e-06,
+      "loss": 1.27786207,
+      "memory(GiB)": 123.79,
+      "step": 45150,
+      "train_speed(iter/s)": 1.315679
+    },
+    {
+      "acc": 0.6891242,
+      "epoch": 1.1454845256215118,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.213871501117842e-06,
+      "loss": 1.33380394,
+      "memory(GiB)": 123.79,
+      "step": 45155,
+      "train_speed(iter/s)": 1.315695
+    },
+    {
+      "acc": 0.67328253,
+      "epoch": 1.145611364789447,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.212835936785811e-06,
+      "loss": 1.50762978,
+      "memory(GiB)": 123.79,
+      "step": 45160,
+      "train_speed(iter/s)": 1.315711
+    },
+    {
+      "acc": 0.69707718,
+      "epoch": 1.145738203957382,
+      "grad_norm": 3.25,
+      "learning_rate": 4.21180040707677e-06,
+      "loss": 1.31044827,
+      "memory(GiB)": 123.79,
+      "step": 45165,
+      "train_speed(iter/s)": 1.315727
+    },
+    {
+      "acc": 0.69098296,
+      "epoch": 1.145865043125317,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.2107649120362684e-06,
+      "loss": 1.43481874,
+      "memory(GiB)": 123.79,
+      "step": 45170,
+      "train_speed(iter/s)": 1.315742
+    },
+    {
+      "acc": 0.67197342,
+      "epoch": 1.1459918822932522,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.2097294517098465e-06,
+      "loss": 1.44384747,
+      "memory(GiB)": 123.79,
+      "step": 45175,
+      "train_speed(iter/s)": 1.315757
+    },
+    {
+      "acc": 0.68719168,
+      "epoch": 1.1461187214611872,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.208694026143054e-06,
+      "loss": 1.45148211,
+      "memory(GiB)": 123.79,
+      "step": 45180,
+      "train_speed(iter/s)": 1.315772
+    },
+    {
+      "acc": 0.68495326,
+      "epoch": 1.1462455606291222,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.2076586353814295e-06,
+      "loss": 1.49729061,
+      "memory(GiB)": 123.79,
+      "step": 45185,
+      "train_speed(iter/s)": 1.315787
+    },
+    {
+      "acc": 0.68479252,
+      "epoch": 1.1463723997970574,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.2066232794705174e-06,
+      "loss": 1.38879929,
+      "memory(GiB)": 123.79,
+      "step": 45190,
+      "train_speed(iter/s)": 1.315802
+    },
+    {
+      "acc": 0.6773613,
+      "epoch": 1.1464992389649924,
+      "grad_norm": 3.25,
+      "learning_rate": 4.205587958455854e-06,
+      "loss": 1.40050344,
+      "memory(GiB)": 123.79,
+      "step": 45195,
+      "train_speed(iter/s)": 1.315818
+    },
+    {
+      "acc": 0.68230982,
+      "epoch": 1.1466260781329274,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.204552672382981e-06,
+      "loss": 1.4386652,
+      "memory(GiB)": 123.79,
+      "step": 45200,
+      "train_speed(iter/s)": 1.315833
+    },
+    {
+      "acc": 0.68147488,
+      "epoch": 1.1467529173008626,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.203517421297431e-06,
+      "loss": 1.37783928,
+      "memory(GiB)": 123.79,
+      "step": 45205,
+      "train_speed(iter/s)": 1.315848
+    },
+    {
+      "acc": 0.68214035,
+      "epoch": 1.1468797564687976,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.202482205244742e-06,
+      "loss": 1.360322,
+      "memory(GiB)": 123.79,
+      "step": 45210,
+      "train_speed(iter/s)": 1.315864
+    },
+    {
+      "acc": 0.70379353,
+      "epoch": 1.1470065956367326,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.201447024270446e-06,
+      "loss": 1.36471195,
+      "memory(GiB)": 123.79,
+      "step": 45215,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.68238215,
+      "epoch": 1.1471334348046676,
+      "grad_norm": 3.0,
+      "learning_rate": 4.200411878420074e-06,
+      "loss": 1.45854778,
+      "memory(GiB)": 123.79,
+      "step": 45220,
+      "train_speed(iter/s)": 1.315896
+    },
+    {
+      "acc": 0.68510723,
+      "epoch": 1.1472602739726028,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.199376767739158e-06,
+      "loss": 1.37390881,
+      "memory(GiB)": 123.79,
+      "step": 45225,
+      "train_speed(iter/s)": 1.31591
+    },
+    {
+      "acc": 0.69886827,
+      "epoch": 1.1473871131405378,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.1983416922732276e-06,
+      "loss": 1.3619215,
+      "memory(GiB)": 123.79,
+      "step": 45230,
+      "train_speed(iter/s)": 1.315925
+    },
+    {
+      "acc": 0.68612709,
+      "epoch": 1.147513952308473,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.197306652067807e-06,
+      "loss": 1.32829952,
+      "memory(GiB)": 123.79,
+      "step": 45235,
+      "train_speed(iter/s)": 1.315941
+    },
+    {
+      "acc": 0.69911032,
+      "epoch": 1.147640791476408,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.196271647168425e-06,
+      "loss": 1.32271929,
+      "memory(GiB)": 123.79,
+      "step": 45240,
+      "train_speed(iter/s)": 1.315956
+    },
+    {
+      "acc": 0.67847395,
+      "epoch": 1.147767630644343,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.195236677620604e-06,
+      "loss": 1.47382278,
+      "memory(GiB)": 123.79,
+      "step": 45245,
+      "train_speed(iter/s)": 1.315972
+    },
+    {
+      "acc": 0.69827461,
+      "epoch": 1.147894469812278,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.1942017434698675e-06,
+      "loss": 1.27494497,
+      "memory(GiB)": 123.79,
+      "step": 45250,
+      "train_speed(iter/s)": 1.315987
+    },
+    {
+      "acc": 0.68690815,
+      "epoch": 1.1480213089802132,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.1931668447617346e-06,
+      "loss": 1.36770754,
+      "memory(GiB)": 123.79,
+      "step": 45255,
+      "train_speed(iter/s)": 1.316003
+    },
+    {
+      "acc": 0.69348001,
+      "epoch": 1.1481481481481481,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.192131981541727e-06,
+      "loss": 1.45657787,
+      "memory(GiB)": 123.79,
+      "step": 45260,
+      "train_speed(iter/s)": 1.316018
+    },
+    {
+      "acc": 0.69058537,
+      "epoch": 1.1482749873160831,
+      "grad_norm": 3.75,
+      "learning_rate": 4.19109715385536e-06,
+      "loss": 1.36884918,
+      "memory(GiB)": 123.79,
+      "step": 45265,
+      "train_speed(iter/s)": 1.316034
+    },
+    {
+      "acc": 0.67458324,
+      "epoch": 1.1484018264840183,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.190062361748154e-06,
+      "loss": 1.40287914,
+      "memory(GiB)": 123.79,
+      "step": 45270,
+      "train_speed(iter/s)": 1.316049
+    },
+    {
+      "acc": 0.67941256,
+      "epoch": 1.1485286656519533,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.189027605265621e-06,
+      "loss": 1.40429077,
+      "memory(GiB)": 123.79,
+      "step": 45275,
+      "train_speed(iter/s)": 1.316064
+    },
+    {
+      "acc": 0.68441,
+      "epoch": 1.1486555048198883,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.187992884453273e-06,
+      "loss": 1.40802689,
+      "memory(GiB)": 123.79,
+      "step": 45280,
+      "train_speed(iter/s)": 1.31608
+    },
+    {
+      "acc": 0.68220911,
+      "epoch": 1.1487823439878235,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.186958199356624e-06,
+      "loss": 1.40218649,
+      "memory(GiB)": 123.79,
+      "step": 45285,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.68088894,
+      "epoch": 1.1489091831557585,
+      "grad_norm": 3.71875,
+      "learning_rate": 4.185923550021185e-06,
+      "loss": 1.4226284,
+      "memory(GiB)": 123.79,
+      "step": 45290,
+      "train_speed(iter/s)": 1.316109
+    },
+    {
+      "acc": 0.68700085,
+      "epoch": 1.1490360223236935,
+      "grad_norm": 2.625,
+      "learning_rate": 4.1848889364924625e-06,
+      "loss": 1.37857132,
+      "memory(GiB)": 123.79,
+      "step": 45295,
+      "train_speed(iter/s)": 1.316124
+    },
+    {
+      "acc": 0.67414374,
+      "epoch": 1.1491628614916287,
+      "grad_norm": 3.5,
+      "learning_rate": 4.183854358815962e-06,
+      "loss": 1.45114412,
+      "memory(GiB)": 123.79,
+      "step": 45300,
+      "train_speed(iter/s)": 1.316139
+    },
+    {
+      "acc": 0.68297787,
+      "epoch": 1.1492897006595637,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.182819817037192e-06,
+      "loss": 1.37925434,
+      "memory(GiB)": 123.79,
+      "step": 45305,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.67915077,
+      "epoch": 1.1494165398274987,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.181785311201655e-06,
+      "loss": 1.41630125,
+      "memory(GiB)": 123.79,
+      "step": 45310,
+      "train_speed(iter/s)": 1.316169
+    },
+    {
+      "acc": 0.68128138,
+      "epoch": 1.1495433789954337,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.1807508413548515e-06,
+      "loss": 1.44780045,
+      "memory(GiB)": 123.79,
+      "step": 45315,
+      "train_speed(iter/s)": 1.316183
+    },
+    {
+      "acc": 0.7005331,
+      "epoch": 1.149670218163369,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.179716407542285e-06,
+      "loss": 1.32618418,
+      "memory(GiB)": 123.79,
+      "step": 45320,
+      "train_speed(iter/s)": 1.316199
+    },
+    {
+      "acc": 0.67879601,
+      "epoch": 1.1497970573313039,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.178682009809452e-06,
+      "loss": 1.44640656,
+      "memory(GiB)": 123.79,
+      "step": 45325,
+      "train_speed(iter/s)": 1.316214
+    },
+    {
+      "acc": 0.68534775,
+      "epoch": 1.1499238964992389,
+      "grad_norm": 2.625,
+      "learning_rate": 4.177647648201854e-06,
+      "loss": 1.38129435,
+      "memory(GiB)": 123.79,
+      "step": 45330,
+      "train_speed(iter/s)": 1.31623
+    },
+    {
+      "acc": 0.68118725,
+      "epoch": 1.150050735667174,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.1766133227649815e-06,
+      "loss": 1.36474266,
+      "memory(GiB)": 123.79,
+      "step": 45335,
+      "train_speed(iter/s)": 1.316245
+    },
+    {
+      "acc": 0.69624128,
+      "epoch": 1.150177574835109,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.175579033544332e-06,
+      "loss": 1.34841452,
+      "memory(GiB)": 123.79,
+      "step": 45340,
+      "train_speed(iter/s)": 1.31626
+    },
+    {
+      "acc": 0.67163916,
+      "epoch": 1.150304414003044,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.174544780585395e-06,
+      "loss": 1.39792032,
+      "memory(GiB)": 123.79,
+      "step": 45345,
+      "train_speed(iter/s)": 1.316277
+    },
+    {
+      "acc": 0.67631874,
+      "epoch": 1.1504312531709793,
+      "grad_norm": 5.84375,
+      "learning_rate": 4.1735105639336686e-06,
+      "loss": 1.49478989,
+      "memory(GiB)": 123.79,
+      "step": 45350,
+      "train_speed(iter/s)": 1.316291
+    },
+    {
+      "acc": 0.68187404,
+      "epoch": 1.1505580923389143,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.172476383634635e-06,
+      "loss": 1.417976,
+      "memory(GiB)": 123.79,
+      "step": 45355,
+      "train_speed(iter/s)": 1.316305
+    },
+    {
+      "acc": 0.68470178,
+      "epoch": 1.1506849315068493,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.171442239733783e-06,
+      "loss": 1.47236462,
+      "memory(GiB)": 123.79,
+      "step": 45360,
+      "train_speed(iter/s)": 1.316319
+    },
+    {
+      "acc": 0.6744319,
+      "epoch": 1.1508117706747845,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.170408132276603e-06,
+      "loss": 1.47713509,
+      "memory(GiB)": 123.79,
+      "step": 45365,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.67553434,
+      "epoch": 1.1509386098427195,
+      "grad_norm": 4.0,
+      "learning_rate": 4.1693740613085776e-06,
+      "loss": 1.42666903,
+      "memory(GiB)": 123.79,
+      "step": 45370,
+      "train_speed(iter/s)": 1.316349
+    },
+    {
+      "acc": 0.69746056,
+      "epoch": 1.1510654490106544,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.168340026875188e-06,
+      "loss": 1.36010084,
+      "memory(GiB)": 123.79,
+      "step": 45375,
+      "train_speed(iter/s)": 1.316364
+    },
+    {
+      "acc": 0.69218545,
+      "epoch": 1.1511922881785894,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.167306029021917e-06,
+      "loss": 1.38231068,
+      "memory(GiB)": 123.79,
+      "step": 45380,
+      "train_speed(iter/s)": 1.316379
+    },
+    {
+      "acc": 0.68194385,
+      "epoch": 1.1513191273465246,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.166272067794246e-06,
+      "loss": 1.42469807,
+      "memory(GiB)": 123.79,
+      "step": 45385,
+      "train_speed(iter/s)": 1.316394
+    },
+    {
+      "acc": 0.67665353,
+      "epoch": 1.1514459665144596,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.165238143237651e-06,
+      "loss": 1.43447704,
+      "memory(GiB)": 123.79,
+      "step": 45390,
+      "train_speed(iter/s)": 1.316409
+    },
+    {
+      "acc": 0.67396979,
+      "epoch": 1.1515728056823948,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.164204255397608e-06,
+      "loss": 1.390411,
+      "memory(GiB)": 123.79,
+      "step": 45395,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.67665911,
+      "epoch": 1.1516996448503298,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.163170404319596e-06,
+      "loss": 1.3985755,
+      "memory(GiB)": 123.79,
+      "step": 45400,
+      "train_speed(iter/s)": 1.316439
+    },
+    {
+      "acc": 0.69060931,
+      "epoch": 1.1518264840182648,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.1621365900490825e-06,
+      "loss": 1.39300222,
+      "memory(GiB)": 123.79,
+      "step": 45405,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.69839287,
+      "epoch": 1.1519533231861998,
+      "grad_norm": 3.875,
+      "learning_rate": 4.1611028126315455e-06,
+      "loss": 1.36603317,
+      "memory(GiB)": 123.79,
+      "step": 45410,
+      "train_speed(iter/s)": 1.316468
+    },
+    {
+      "acc": 0.68335905,
+      "epoch": 1.152080162354135,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.160069072112451e-06,
+      "loss": 1.3872735,
+      "memory(GiB)": 123.79,
+      "step": 45415,
+      "train_speed(iter/s)": 1.316483
+    },
+    {
+      "acc": 0.69090571,
+      "epoch": 1.15220700152207,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.1590353685372695e-06,
+      "loss": 1.32997131,
+      "memory(GiB)": 123.79,
+      "step": 45420,
+      "train_speed(iter/s)": 1.316498
+    },
+    {
+      "acc": 0.69850121,
+      "epoch": 1.152333840690005,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.158001701951465e-06,
+      "loss": 1.30805836,
+      "memory(GiB)": 123.79,
+      "step": 45425,
+      "train_speed(iter/s)": 1.316514
+    },
+    {
+      "acc": 0.68384809,
+      "epoch": 1.1524606798579402,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.156968072400508e-06,
+      "loss": 1.42039595,
+      "memory(GiB)": 123.79,
+      "step": 45430,
+      "train_speed(iter/s)": 1.31653
+    },
+    {
+      "acc": 0.68042479,
+      "epoch": 1.1525875190258752,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.155934479929858e-06,
+      "loss": 1.41392384,
+      "memory(GiB)": 123.79,
+      "step": 45435,
+      "train_speed(iter/s)": 1.316545
+    },
+    {
+      "acc": 0.71661558,
+      "epoch": 1.1527143581938102,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.154900924584976e-06,
+      "loss": 1.25272675,
+      "memory(GiB)": 123.79,
+      "step": 45440,
+      "train_speed(iter/s)": 1.31656
+    },
+    {
+      "acc": 0.67436671,
+      "epoch": 1.1528411973617454,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.153867406411327e-06,
+      "loss": 1.4692194,
+      "memory(GiB)": 123.79,
+      "step": 45445,
+      "train_speed(iter/s)": 1.316575
+    },
+    {
+      "acc": 0.6972208,
+      "epoch": 1.1529680365296804,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.152833925454367e-06,
+      "loss": 1.32464867,
+      "memory(GiB)": 123.79,
+      "step": 45450,
+      "train_speed(iter/s)": 1.31659
+    },
+    {
+      "acc": 0.69371243,
+      "epoch": 1.1530948756976154,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.1518004817595515e-06,
+      "loss": 1.33763638,
+      "memory(GiB)": 123.79,
+      "step": 45455,
+      "train_speed(iter/s)": 1.316602
+    },
+    {
+      "acc": 0.65727854,
+      "epoch": 1.1532217148655506,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.150767075372338e-06,
+      "loss": 1.47458649,
+      "memory(GiB)": 123.79,
+      "step": 45460,
+      "train_speed(iter/s)": 1.316617
+    },
+    {
+      "acc": 0.68773084,
+      "epoch": 1.1533485540334856,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.149733706338182e-06,
+      "loss": 1.35453806,
+      "memory(GiB)": 123.79,
+      "step": 45465,
+      "train_speed(iter/s)": 1.316632
+    },
+    {
+      "acc": 0.69446688,
+      "epoch": 1.1534753932014206,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.148700374702533e-06,
+      "loss": 1.37830868,
+      "memory(GiB)": 123.79,
+      "step": 45470,
+      "train_speed(iter/s)": 1.316648
+    },
+    {
+      "acc": 0.69824171,
+      "epoch": 1.1536022323693556,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.147667080510841e-06,
+      "loss": 1.38421354,
+      "memory(GiB)": 123.79,
+      "step": 45475,
+      "train_speed(iter/s)": 1.316663
+    },
+    {
+      "acc": 0.69662328,
+      "epoch": 1.1537290715372908,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.146633823808557e-06,
+      "loss": 1.34927902,
+      "memory(GiB)": 123.79,
+      "step": 45480,
+      "train_speed(iter/s)": 1.316677
+    },
+    {
+      "acc": 0.66687093,
+      "epoch": 1.1538559107052258,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.145600604641127e-06,
+      "loss": 1.41441679,
+      "memory(GiB)": 123.79,
+      "step": 45485,
+      "train_speed(iter/s)": 1.316692
+    },
+    {
+      "acc": 0.69420681,
+      "epoch": 1.1539827498731607,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.1445674230539985e-06,
+      "loss": 1.36401768,
+      "memory(GiB)": 123.79,
+      "step": 45490,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.70447025,
+      "epoch": 1.154109589041096,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.143534279092613e-06,
+      "loss": 1.36787701,
+      "memory(GiB)": 123.79,
+      "step": 45495,
+      "train_speed(iter/s)": 1.316721
+    },
+    {
+      "acc": 0.68115015,
+      "epoch": 1.154236428209031,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.142501172802412e-06,
+      "loss": 1.4749855,
+      "memory(GiB)": 123.79,
+      "step": 45500,
+      "train_speed(iter/s)": 1.316736
+    },
+    {
+      "acc": 0.67630424,
+      "epoch": 1.154363267376966,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.14146810422884e-06,
+      "loss": 1.44215717,
+      "memory(GiB)": 123.79,
+      "step": 45505,
+      "train_speed(iter/s)": 1.31675
+    },
+    {
+      "acc": 0.67666583,
+      "epoch": 1.1544901065449011,
+      "grad_norm": 4.0,
+      "learning_rate": 4.140435073417335e-06,
+      "loss": 1.34657192,
+      "memory(GiB)": 123.79,
+      "step": 45510,
+      "train_speed(iter/s)": 1.316764
+    },
+    {
+      "acc": 0.68973131,
+      "epoch": 1.1546169457128361,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.139402080413331e-06,
+      "loss": 1.40954227,
+      "memory(GiB)": 123.79,
+      "step": 45515,
+      "train_speed(iter/s)": 1.316778
+    },
+    {
+      "acc": 0.69047608,
+      "epoch": 1.1547437848807711,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.138369125262266e-06,
+      "loss": 1.40414867,
+      "memory(GiB)": 123.79,
+      "step": 45520,
+      "train_speed(iter/s)": 1.316793
+    },
+    {
+      "acc": 0.67874589,
+      "epoch": 1.1548706240487063,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.137336208009574e-06,
+      "loss": 1.39599276,
+      "memory(GiB)": 123.79,
+      "step": 45525,
+      "train_speed(iter/s)": 1.316808
+    },
+    {
+      "acc": 0.68029637,
+      "epoch": 1.1549974632166413,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.136303328700688e-06,
+      "loss": 1.43718853,
+      "memory(GiB)": 123.79,
+      "step": 45530,
+      "train_speed(iter/s)": 1.316823
+    },
+    {
+      "acc": 0.67238436,
+      "epoch": 1.1551243023845763,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.135270487381037e-06,
+      "loss": 1.45257864,
+      "memory(GiB)": 123.79,
+      "step": 45535,
+      "train_speed(iter/s)": 1.316838
+    },
+    {
+      "acc": 0.69569607,
+      "epoch": 1.1552511415525113,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.13423768409605e-06,
+      "loss": 1.36384249,
+      "memory(GiB)": 123.79,
+      "step": 45540,
+      "train_speed(iter/s)": 1.316853
+    },
+    {
+      "acc": 0.68546348,
+      "epoch": 1.1553779807204465,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.133204918891155e-06,
+      "loss": 1.40664501,
+      "memory(GiB)": 123.79,
+      "step": 45545,
+      "train_speed(iter/s)": 1.316868
+    },
+    {
+      "acc": 0.69057198,
+      "epoch": 1.1555048198883815,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.132172191811781e-06,
+      "loss": 1.4067008,
+      "memory(GiB)": 123.79,
+      "step": 45550,
+      "train_speed(iter/s)": 1.316883
+    },
+    {
+      "acc": 0.68559275,
+      "epoch": 1.1556316590563167,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.131139502903345e-06,
+      "loss": 1.39246159,
+      "memory(GiB)": 123.79,
+      "step": 45555,
+      "train_speed(iter/s)": 1.316898
+    },
+    {
+      "acc": 0.67754269,
+      "epoch": 1.1557584982242517,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.130106852211273e-06,
+      "loss": 1.42916622,
+      "memory(GiB)": 123.79,
+      "step": 45560,
+      "train_speed(iter/s)": 1.316913
+    },
+    {
+      "acc": 0.69174089,
+      "epoch": 1.1558853373921867,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.129074239780986e-06,
+      "loss": 1.3484499,
+      "memory(GiB)": 123.79,
+      "step": 45565,
+      "train_speed(iter/s)": 1.316928
+    },
+    {
+      "acc": 0.70376973,
+      "epoch": 1.1560121765601217,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.128041665657903e-06,
+      "loss": 1.38859377,
+      "memory(GiB)": 123.79,
+      "step": 45570,
+      "train_speed(iter/s)": 1.316944
+    },
+    {
+      "acc": 0.68930817,
+      "epoch": 1.1561390157280569,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.127009129887441e-06,
+      "loss": 1.37566986,
+      "memory(GiB)": 123.79,
+      "step": 45575,
+      "train_speed(iter/s)": 1.31696
+    },
+    {
+      "acc": 0.67835221,
+      "epoch": 1.1562658548959919,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.125976632515013e-06,
+      "loss": 1.41134567,
+      "memory(GiB)": 123.79,
+      "step": 45580,
+      "train_speed(iter/s)": 1.316976
+    },
+    {
+      "acc": 0.68422794,
+      "epoch": 1.1563926940639269,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.124944173586036e-06,
+      "loss": 1.40901318,
+      "memory(GiB)": 123.79,
+      "step": 45585,
+      "train_speed(iter/s)": 1.316991
+    },
+    {
+      "acc": 0.67981148,
+      "epoch": 1.156519533231862,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.123911753145922e-06,
+      "loss": 1.42894058,
+      "memory(GiB)": 123.79,
+      "step": 45590,
+      "train_speed(iter/s)": 1.317006
+    },
+    {
+      "acc": 0.68248911,
+      "epoch": 1.156646372399797,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.12287937124008e-06,
+      "loss": 1.40659027,
+      "memory(GiB)": 123.79,
+      "step": 45595,
+      "train_speed(iter/s)": 1.317023
+    },
+    {
+      "acc": 0.68482776,
+      "epoch": 1.156773211567732,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.121847027913918e-06,
+      "loss": 1.44514198,
+      "memory(GiB)": 123.79,
+      "step": 45600,
+      "train_speed(iter/s)": 1.317038
+    },
+    {
+      "acc": 0.69487147,
+      "epoch": 1.1569000507356673,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.1208147232128456e-06,
+      "loss": 1.37140274,
+      "memory(GiB)": 123.79,
+      "step": 45605,
+      "train_speed(iter/s)": 1.317053
+    },
+    {
+      "acc": 0.68251982,
+      "epoch": 1.1570268899036023,
+      "grad_norm": 3.125,
+      "learning_rate": 4.119782457182267e-06,
+      "loss": 1.42214441,
+      "memory(GiB)": 123.79,
+      "step": 45610,
+      "train_speed(iter/s)": 1.317065
+    },
+    {
+      "acc": 0.70235415,
+      "epoch": 1.1571537290715372,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.118750229867585e-06,
+      "loss": 1.36956892,
+      "memory(GiB)": 123.79,
+      "step": 45615,
+      "train_speed(iter/s)": 1.317081
+    },
+    {
+      "acc": 0.67645912,
+      "epoch": 1.1572805682394725,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.117718041314204e-06,
+      "loss": 1.40564003,
+      "memory(GiB)": 123.79,
+      "step": 45620,
+      "train_speed(iter/s)": 1.317097
+    },
+    {
+      "acc": 0.67594242,
+      "epoch": 1.1574074074074074,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.11668589156752e-06,
+      "loss": 1.42358932,
+      "memory(GiB)": 123.79,
+      "step": 45625,
+      "train_speed(iter/s)": 1.317109
+    },
+    {
+      "acc": 0.7018682,
+      "epoch": 1.1575342465753424,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.115653780672937e-06,
+      "loss": 1.3197958,
+      "memory(GiB)": 123.79,
+      "step": 45630,
+      "train_speed(iter/s)": 1.317125
+    },
+    {
+      "acc": 0.68428526,
+      "epoch": 1.1576610857432774,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.1146217086758475e-06,
+      "loss": 1.39758396,
+      "memory(GiB)": 123.79,
+      "step": 45635,
+      "train_speed(iter/s)": 1.317141
+    },
+    {
+      "acc": 0.70528994,
+      "epoch": 1.1577879249112126,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.113589675621649e-06,
+      "loss": 1.29141636,
+      "memory(GiB)": 123.79,
+      "step": 45640,
+      "train_speed(iter/s)": 1.317156
+    },
+    {
+      "acc": 0.69743242,
+      "epoch": 1.1579147640791476,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.112557681555733e-06,
+      "loss": 1.3412035,
+      "memory(GiB)": 123.79,
+      "step": 45645,
+      "train_speed(iter/s)": 1.317171
+    },
+    {
+      "acc": 0.68948951,
+      "epoch": 1.1580416032470826,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.111525726523494e-06,
+      "loss": 1.41584072,
+      "memory(GiB)": 123.79,
+      "step": 45650,
+      "train_speed(iter/s)": 1.317186
+    },
+    {
+      "acc": 0.69971981,
+      "epoch": 1.1581684424150178,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.110493810570319e-06,
+      "loss": 1.30422421,
+      "memory(GiB)": 123.79,
+      "step": 45655,
+      "train_speed(iter/s)": 1.317201
+    },
+    {
+      "acc": 0.68468213,
+      "epoch": 1.1582952815829528,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.109461933741598e-06,
+      "loss": 1.42376652,
+      "memory(GiB)": 123.79,
+      "step": 45660,
+      "train_speed(iter/s)": 1.317216
+    },
+    {
+      "acc": 0.67403955,
+      "epoch": 1.1584221207508878,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.108430096082716e-06,
+      "loss": 1.45837555,
+      "memory(GiB)": 123.79,
+      "step": 45665,
+      "train_speed(iter/s)": 1.317231
+    },
+    {
+      "acc": 0.67903857,
+      "epoch": 1.158548959918823,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.107398297639062e-06,
+      "loss": 1.42919407,
+      "memory(GiB)": 123.79,
+      "step": 45670,
+      "train_speed(iter/s)": 1.317244
+    },
+    {
+      "acc": 0.67498055,
+      "epoch": 1.158675799086758,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.106366538456013e-06,
+      "loss": 1.43192196,
+      "memory(GiB)": 123.79,
+      "step": 45675,
+      "train_speed(iter/s)": 1.31726
+    },
+    {
+      "acc": 0.70047808,
+      "epoch": 1.158802638254693,
+      "grad_norm": 3.0,
+      "learning_rate": 4.105334818578954e-06,
+      "loss": 1.33480797,
+      "memory(GiB)": 123.79,
+      "step": 45680,
+      "train_speed(iter/s)": 1.317275
+    },
+    {
+      "acc": 0.6913763,
+      "epoch": 1.1589294774226282,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.104303138053265e-06,
+      "loss": 1.43409233,
+      "memory(GiB)": 123.79,
+      "step": 45685,
+      "train_speed(iter/s)": 1.31729
+    },
+    {
+      "acc": 0.70663905,
+      "epoch": 1.1590563165905632,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.103271496924323e-06,
+      "loss": 1.31933727,
+      "memory(GiB)": 123.79,
+      "step": 45690,
+      "train_speed(iter/s)": 1.317304
+    },
+    {
+      "acc": 0.7021431,
+      "epoch": 1.1591831557584982,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.102239895237503e-06,
+      "loss": 1.41526356,
+      "memory(GiB)": 123.79,
+      "step": 45695,
+      "train_speed(iter/s)": 1.31732
+    },
+    {
+      "acc": 0.69910803,
+      "epoch": 1.1593099949264332,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.101208333038181e-06,
+      "loss": 1.32758961,
+      "memory(GiB)": 123.79,
+      "step": 45700,
+      "train_speed(iter/s)": 1.317335
+    },
+    {
+      "acc": 0.6825696,
+      "epoch": 1.1594368340943684,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.1001768103717285e-06,
+      "loss": 1.42787304,
+      "memory(GiB)": 123.79,
+      "step": 45705,
+      "train_speed(iter/s)": 1.317349
+    },
+    {
+      "acc": 0.68385682,
+      "epoch": 1.1595636732623034,
+      "grad_norm": 3.375,
+      "learning_rate": 4.09914532728352e-06,
+      "loss": 1.32888212,
+      "memory(GiB)": 123.79,
+      "step": 45710,
+      "train_speed(iter/s)": 1.317363
+    },
+    {
+      "acc": 0.67990074,
+      "epoch": 1.1596905124302386,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.09811388381892e-06,
+      "loss": 1.43546629,
+      "memory(GiB)": 123.79,
+      "step": 45715,
+      "train_speed(iter/s)": 1.317378
+    },
+    {
+      "acc": 0.69878626,
+      "epoch": 1.1598173515981736,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.097082480023298e-06,
+      "loss": 1.38847561,
+      "memory(GiB)": 123.79,
+      "step": 45720,
+      "train_speed(iter/s)": 1.317393
+    },
+    {
+      "acc": 0.67653847,
+      "epoch": 1.1599441907661086,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.09605111594202e-06,
+      "loss": 1.4448288,
+      "memory(GiB)": 123.79,
+      "step": 45725,
+      "train_speed(iter/s)": 1.317409
+    },
+    {
+      "acc": 0.68940239,
+      "epoch": 1.1600710299340435,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.095019791620451e-06,
+      "loss": 1.39786024,
+      "memory(GiB)": 123.79,
+      "step": 45730,
+      "train_speed(iter/s)": 1.317424
+    },
+    {
+      "acc": 0.6897841,
+      "epoch": 1.1601978691019788,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.093988507103951e-06,
+      "loss": 1.34509668,
+      "memory(GiB)": 123.79,
+      "step": 45735,
+      "train_speed(iter/s)": 1.317439
+    },
+    {
+      "acc": 0.695576,
+      "epoch": 1.1603247082699137,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.09295726243788e-06,
+      "loss": 1.33213024,
+      "memory(GiB)": 123.79,
+      "step": 45740,
+      "train_speed(iter/s)": 1.317453
+    },
+    {
+      "acc": 0.69687591,
+      "epoch": 1.1604515474378487,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.091926057667601e-06,
+      "loss": 1.41617184,
+      "memory(GiB)": 123.79,
+      "step": 45745,
+      "train_speed(iter/s)": 1.317468
+    },
+    {
+      "acc": 0.6763505,
+      "epoch": 1.160578386605784,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.0908948928384675e-06,
+      "loss": 1.38995943,
+      "memory(GiB)": 123.79,
+      "step": 45750,
+      "train_speed(iter/s)": 1.317483
+    },
+    {
+      "acc": 0.69374599,
+      "epoch": 1.160705225773719,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.089863767995835e-06,
+      "loss": 1.41296387,
+      "memory(GiB)": 123.79,
+      "step": 45755,
+      "train_speed(iter/s)": 1.317498
+    },
+    {
+      "acc": 0.68025465,
+      "epoch": 1.160832064941654,
+      "grad_norm": 3.25,
+      "learning_rate": 4.088832683185057e-06,
+      "loss": 1.39562159,
+      "memory(GiB)": 123.79,
+      "step": 45760,
+      "train_speed(iter/s)": 1.317513
+    },
+    {
+      "acc": 0.67591534,
+      "epoch": 1.1609589041095891,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.087801638451485e-06,
+      "loss": 1.45251007,
+      "memory(GiB)": 123.79,
+      "step": 45765,
+      "train_speed(iter/s)": 1.317527
+    },
+    {
+      "acc": 0.68624325,
+      "epoch": 1.1610857432775241,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.086770633840472e-06,
+      "loss": 1.41757965,
+      "memory(GiB)": 123.79,
+      "step": 45770,
+      "train_speed(iter/s)": 1.317542
+    },
+    {
+      "acc": 0.68631544,
+      "epoch": 1.161212582445459,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.085739669397362e-06,
+      "loss": 1.37343616,
+      "memory(GiB)": 123.79,
+      "step": 45775,
+      "train_speed(iter/s)": 1.317557
+    },
+    {
+      "acc": 0.68964047,
+      "epoch": 1.1613394216133943,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.084708745167504e-06,
+      "loss": 1.33205833,
+      "memory(GiB)": 123.79,
+      "step": 45780,
+      "train_speed(iter/s)": 1.317572
+    },
+    {
+      "acc": 0.66956196,
+      "epoch": 1.1614662607813293,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.08367786119624e-06,
+      "loss": 1.50565224,
+      "memory(GiB)": 123.79,
+      "step": 45785,
+      "train_speed(iter/s)": 1.317587
+    },
+    {
+      "acc": 0.69795618,
+      "epoch": 1.1615930999492643,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.082647017528918e-06,
+      "loss": 1.37456417,
+      "memory(GiB)": 123.79,
+      "step": 45790,
+      "train_speed(iter/s)": 1.3176
+    },
+    {
+      "acc": 0.68809133,
+      "epoch": 1.1617199391171993,
+      "grad_norm": 4.0,
+      "learning_rate": 4.081616214210874e-06,
+      "loss": 1.39899845,
+      "memory(GiB)": 123.79,
+      "step": 45795,
+      "train_speed(iter/s)": 1.317615
+    },
+    {
+      "acc": 0.68654189,
+      "epoch": 1.1618467782851345,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.0805854512874485e-06,
+      "loss": 1.39001102,
+      "memory(GiB)": 123.79,
+      "step": 45800,
+      "train_speed(iter/s)": 1.317629
+    },
+    {
+      "acc": 0.69130616,
+      "epoch": 1.1619736174530695,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.079554728803981e-06,
+      "loss": 1.36553402,
+      "memory(GiB)": 123.79,
+      "step": 45805,
+      "train_speed(iter/s)": 1.317642
+    },
+    {
+      "acc": 0.69461293,
+      "epoch": 1.1621004566210045,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.078524046805806e-06,
+      "loss": 1.37904758,
+      "memory(GiB)": 123.79,
+      "step": 45810,
+      "train_speed(iter/s)": 1.317657
+    },
+    {
+      "acc": 0.68545065,
+      "epoch": 1.1622272957889397,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.0774934053382576e-06,
+      "loss": 1.37098885,
+      "memory(GiB)": 123.79,
+      "step": 45815,
+      "train_speed(iter/s)": 1.317673
+    },
+    {
+      "acc": 0.70048203,
+      "epoch": 1.1623541349568747,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.076462804446667e-06,
+      "loss": 1.39020672,
+      "memory(GiB)": 123.79,
+      "step": 45820,
+      "train_speed(iter/s)": 1.317688
+    },
+    {
+      "acc": 0.68481212,
+      "epoch": 1.1624809741248097,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.0754322441763654e-06,
+      "loss": 1.41859913,
+      "memory(GiB)": 123.79,
+      "step": 45825,
+      "train_speed(iter/s)": 1.317704
+    },
+    {
+      "acc": 0.68014879,
+      "epoch": 1.1626078132927449,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.0744017245726834e-06,
+      "loss": 1.42403555,
+      "memory(GiB)": 123.79,
+      "step": 45830,
+      "train_speed(iter/s)": 1.31772
+    },
+    {
+      "acc": 0.68008909,
+      "epoch": 1.1627346524606799,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.073371245680944e-06,
+      "loss": 1.43026066,
+      "memory(GiB)": 123.79,
+      "step": 45835,
+      "train_speed(iter/s)": 1.317736
+    },
+    {
+      "acc": 0.67746315,
+      "epoch": 1.1628614916286149,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.0723408075464754e-06,
+      "loss": 1.46493549,
+      "memory(GiB)": 123.79,
+      "step": 45840,
+      "train_speed(iter/s)": 1.317752
+    },
+    {
+      "acc": 0.67914867,
+      "epoch": 1.16298833079655,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.071310410214598e-06,
+      "loss": 1.43710566,
+      "memory(GiB)": 123.79,
+      "step": 45845,
+      "train_speed(iter/s)": 1.317769
+    },
+    {
+      "acc": 0.67857656,
+      "epoch": 1.163115169964485,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.070280053730639e-06,
+      "loss": 1.46736279,
+      "memory(GiB)": 123.79,
+      "step": 45850,
+      "train_speed(iter/s)": 1.317784
+    },
+    {
+      "acc": 0.6840786,
+      "epoch": 1.16324200913242,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.069249738139911e-06,
+      "loss": 1.4192215,
+      "memory(GiB)": 123.79,
+      "step": 45855,
+      "train_speed(iter/s)": 1.317801
+    },
+    {
+      "acc": 0.6904552,
+      "epoch": 1.163368848300355,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.068219463487736e-06,
+      "loss": 1.38512068,
+      "memory(GiB)": 123.79,
+      "step": 45860,
+      "train_speed(iter/s)": 1.317814
+    },
+    {
+      "acc": 0.69474821,
+      "epoch": 1.1634956874682902,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.0671892298194286e-06,
+      "loss": 1.39879656,
+      "memory(GiB)": 123.79,
+      "step": 45865,
+      "train_speed(iter/s)": 1.317828
+    },
+    {
+      "acc": 0.69241438,
+      "epoch": 1.1636225266362252,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.066159037180304e-06,
+      "loss": 1.3827961,
+      "memory(GiB)": 123.79,
+      "step": 45870,
+      "train_speed(iter/s)": 1.317843
+    },
+    {
+      "acc": 0.68904166,
+      "epoch": 1.1637493658041604,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.065128885615674e-06,
+      "loss": 1.42926331,
+      "memory(GiB)": 123.79,
+      "step": 45875,
+      "train_speed(iter/s)": 1.317859
+    },
+    {
+      "acc": 0.66630263,
+      "epoch": 1.1638762049720954,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.064098775170849e-06,
+      "loss": 1.43703814,
+      "memory(GiB)": 123.79,
+      "step": 45880,
+      "train_speed(iter/s)": 1.317875
+    },
+    {
+      "acc": 0.68354082,
+      "epoch": 1.1640030441400304,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.063068705891139e-06,
+      "loss": 1.42268534,
+      "memory(GiB)": 123.79,
+      "step": 45885,
+      "train_speed(iter/s)": 1.31789
+    },
+    {
+      "acc": 0.68718853,
+      "epoch": 1.1641298833079654,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.062038677821852e-06,
+      "loss": 1.40927505,
+      "memory(GiB)": 123.79,
+      "step": 45890,
+      "train_speed(iter/s)": 1.317906
+    },
+    {
+      "acc": 0.69282885,
+      "epoch": 1.1642567224759006,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.061008691008289e-06,
+      "loss": 1.39045753,
+      "memory(GiB)": 123.79,
+      "step": 45895,
+      "train_speed(iter/s)": 1.31792
+    },
+    {
+      "acc": 0.69573812,
+      "epoch": 1.1643835616438356,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.059978745495757e-06,
+      "loss": 1.34949389,
+      "memory(GiB)": 123.79,
+      "step": 45900,
+      "train_speed(iter/s)": 1.317935
+    },
+    {
+      "acc": 0.68532677,
+      "epoch": 1.1645104008117706,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.058948841329557e-06,
+      "loss": 1.38636713,
+      "memory(GiB)": 123.79,
+      "step": 45905,
+      "train_speed(iter/s)": 1.31795
+    },
+    {
+      "acc": 0.68174739,
+      "epoch": 1.1646372399797058,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.057918978554989e-06,
+      "loss": 1.45145035,
+      "memory(GiB)": 123.79,
+      "step": 45910,
+      "train_speed(iter/s)": 1.317965
+    },
+    {
+      "acc": 0.69317331,
+      "epoch": 1.1647640791476408,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.056889157217348e-06,
+      "loss": 1.35580311,
+      "memory(GiB)": 123.79,
+      "step": 45915,
+      "train_speed(iter/s)": 1.317979
+    },
+    {
+      "acc": 0.66876521,
+      "epoch": 1.1648909183155758,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.0558593773619346e-06,
+      "loss": 1.46598358,
+      "memory(GiB)": 123.79,
+      "step": 45920,
+      "train_speed(iter/s)": 1.317994
+    },
+    {
+      "acc": 0.68588009,
+      "epoch": 1.165017757483511,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.05482963903404e-06,
+      "loss": 1.39615088,
+      "memory(GiB)": 123.79,
+      "step": 45925,
+      "train_speed(iter/s)": 1.318009
+    },
+    {
+      "acc": 0.67679834,
+      "epoch": 1.165144596651446,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.05379994227896e-06,
+      "loss": 1.45531912,
+      "memory(GiB)": 123.79,
+      "step": 45930,
+      "train_speed(iter/s)": 1.318024
+    },
+    {
+      "acc": 0.67247615,
+      "epoch": 1.165271435819381,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.052770287141981e-06,
+      "loss": 1.43992491,
+      "memory(GiB)": 123.79,
+      "step": 45935,
+      "train_speed(iter/s)": 1.318039
+    },
+    {
+      "acc": 0.69854884,
+      "epoch": 1.1653982749873162,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.051740673668393e-06,
+      "loss": 1.30989609,
+      "memory(GiB)": 123.79,
+      "step": 45940,
+      "train_speed(iter/s)": 1.318054
+    },
+    {
+      "acc": 0.68191824,
+      "epoch": 1.1655251141552512,
+      "grad_norm": 3.375,
+      "learning_rate": 4.0507111019034855e-06,
+      "loss": 1.38589439,
+      "memory(GiB)": 123.79,
+      "step": 45945,
+      "train_speed(iter/s)": 1.318069
+    },
+    {
+      "acc": 0.68668313,
+      "epoch": 1.1656519533231862,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.049681571892543e-06,
+      "loss": 1.3813612,
+      "memory(GiB)": 123.79,
+      "step": 45950,
+      "train_speed(iter/s)": 1.318084
+    },
+    {
+      "acc": 0.68318481,
+      "epoch": 1.1657787924911212,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.048652083680847e-06,
+      "loss": 1.38119726,
+      "memory(GiB)": 123.79,
+      "step": 45955,
+      "train_speed(iter/s)": 1.3181
+    },
+    {
+      "acc": 0.66721325,
+      "epoch": 1.1659056316590564,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.047622637313678e-06,
+      "loss": 1.52647276,
+      "memory(GiB)": 123.79,
+      "step": 45960,
+      "train_speed(iter/s)": 1.318115
+    },
+    {
+      "acc": 0.69518356,
+      "epoch": 1.1660324708269914,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.046593232836319e-06,
+      "loss": 1.32981434,
+      "memory(GiB)": 123.79,
+      "step": 45965,
+      "train_speed(iter/s)": 1.31813
+    },
+    {
+      "acc": 0.69214373,
+      "epoch": 1.1661593099949263,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.045563870294047e-06,
+      "loss": 1.41568584,
+      "memory(GiB)": 123.79,
+      "step": 45970,
+      "train_speed(iter/s)": 1.318147
+    },
+    {
+      "acc": 0.69361534,
+      "epoch": 1.1662861491628616,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.044534549732135e-06,
+      "loss": 1.31671305,
+      "memory(GiB)": 123.79,
+      "step": 45975,
+      "train_speed(iter/s)": 1.318161
+    },
+    {
+      "acc": 0.70563288,
+      "epoch": 1.1664129883307965,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.043505271195861e-06,
+      "loss": 1.3029088,
+      "memory(GiB)": 123.79,
+      "step": 45980,
+      "train_speed(iter/s)": 1.318177
+    },
+    {
+      "acc": 0.68688612,
+      "epoch": 1.1665398274987315,
+      "grad_norm": 3.765625,
+      "learning_rate": 4.042476034730494e-06,
+      "loss": 1.44093838,
+      "memory(GiB)": 123.79,
+      "step": 45985,
+      "train_speed(iter/s)": 1.318192
+    },
+    {
+      "acc": 0.67958918,
+      "epoch": 1.1666666666666667,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.041446840381309e-06,
+      "loss": 1.44980307,
+      "memory(GiB)": 123.79,
+      "step": 45990,
+      "train_speed(iter/s)": 1.318208
+    },
+    {
+      "acc": 0.69142952,
+      "epoch": 1.1667935058346017,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.040417688193569e-06,
+      "loss": 1.35890064,
+      "memory(GiB)": 123.79,
+      "step": 45995,
+      "train_speed(iter/s)": 1.318223
+    },
+    {
+      "acc": 0.68804264,
+      "epoch": 1.1669203450025367,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.039388578212545e-06,
+      "loss": 1.36881084,
+      "memory(GiB)": 123.79,
+      "step": 46000,
+      "train_speed(iter/s)": 1.318239
+    },
+    {
+      "epoch": 1.1669203450025367,
+      "eval_acc": 0.6750194734544969,
+      "eval_loss": 1.3563450574874878,
+      "eval_runtime": 70.058,
+      "eval_samples_per_second": 90.925,
+      "eval_steps_per_second": 22.738,
+      "step": 46000
+    },
+    {
+      "acc": 0.6886405,
+      "epoch": 1.167047184170472,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.0383595104834975e-06,
+      "loss": 1.38294601,
+      "memory(GiB)": 123.79,
+      "step": 46005,
+      "train_speed(iter/s)": 1.31514
+    },
+    {
+      "acc": 0.68025026,
+      "epoch": 1.167174023338407,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.037330485051695e-06,
+      "loss": 1.40198269,
+      "memory(GiB)": 123.79,
+      "step": 46010,
+      "train_speed(iter/s)": 1.315156
+    },
+    {
+      "acc": 0.67081661,
+      "epoch": 1.167300862506342,
+      "grad_norm": 2.59375,
+      "learning_rate": 4.0363015019623955e-06,
+      "loss": 1.43777447,
+      "memory(GiB)": 123.79,
+      "step": 46015,
+      "train_speed(iter/s)": 1.315172
+    },
+    {
+      "acc": 0.69931707,
+      "epoch": 1.167427701674277,
+      "grad_norm": 2.875,
+      "learning_rate": 4.0352725612608565e-06,
+      "loss": 1.3551877,
+      "memory(GiB)": 123.79,
+      "step": 46020,
+      "train_speed(iter/s)": 1.315187
+    },
+    {
+      "acc": 0.68764849,
+      "epoch": 1.167554540842212,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.0342436629923385e-06,
+      "loss": 1.42353859,
+      "memory(GiB)": 123.79,
+      "step": 46025,
+      "train_speed(iter/s)": 1.315203
+    },
+    {
+      "acc": 0.69143248,
+      "epoch": 1.167681380010147,
+      "grad_norm": 3.125,
+      "learning_rate": 4.033214807202098e-06,
+      "loss": 1.37995329,
+      "memory(GiB)": 123.79,
+      "step": 46030,
+      "train_speed(iter/s)": 1.315219
+    },
+    {
+      "acc": 0.68799467,
+      "epoch": 1.1678082191780823,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.032185993935385e-06,
+      "loss": 1.40867033,
+      "memory(GiB)": 123.79,
+      "step": 46035,
+      "train_speed(iter/s)": 1.315235
+    },
+    {
+      "acc": 0.68975654,
+      "epoch": 1.1679350583460173,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.031157223237452e-06,
+      "loss": 1.32615929,
+      "memory(GiB)": 123.79,
+      "step": 46040,
+      "train_speed(iter/s)": 1.315251
+    },
+    {
+      "acc": 0.70576105,
+      "epoch": 1.1680618975139523,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.0301284951535504e-06,
+      "loss": 1.32221155,
+      "memory(GiB)": 123.79,
+      "step": 46045,
+      "train_speed(iter/s)": 1.315267
+    },
+    {
+      "acc": 0.70255899,
+      "epoch": 1.1681887366818873,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.029099809728929e-06,
+      "loss": 1.39370823,
+      "memory(GiB)": 123.79,
+      "step": 46050,
+      "train_speed(iter/s)": 1.315283
+    },
+    {
+      "acc": 0.68594308,
+      "epoch": 1.1683155758498225,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.028071167008831e-06,
+      "loss": 1.44583464,
+      "memory(GiB)": 123.79,
+      "step": 46055,
+      "train_speed(iter/s)": 1.315299
+    },
+    {
+      "acc": 0.69413481,
+      "epoch": 1.1684424150177575,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.027042567038503e-06,
+      "loss": 1.32699165,
+      "memory(GiB)": 123.79,
+      "step": 46060,
+      "train_speed(iter/s)": 1.315316
+    },
+    {
+      "acc": 0.69040112,
+      "epoch": 1.1685692541856925,
+      "grad_norm": 2.71875,
+      "learning_rate": 4.026014009863186e-06,
+      "loss": 1.41153765,
+      "memory(GiB)": 123.79,
+      "step": 46065,
+      "train_speed(iter/s)": 1.315332
+    },
+    {
+      "acc": 0.68756304,
+      "epoch": 1.1686960933536277,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.024985495528124e-06,
+      "loss": 1.39067278,
+      "memory(GiB)": 123.79,
+      "step": 46070,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.68923655,
+      "epoch": 1.1688229325215627,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.023957024078552e-06,
+      "loss": 1.39215889,
+      "memory(GiB)": 123.79,
+      "step": 46075,
+      "train_speed(iter/s)": 1.315364
+    },
+    {
+      "acc": 0.69175301,
+      "epoch": 1.1689497716894977,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.022928595559707e-06,
+      "loss": 1.35366402,
+      "memory(GiB)": 123.79,
+      "step": 46080,
+      "train_speed(iter/s)": 1.315379
+    },
+    {
+      "acc": 0.68999367,
+      "epoch": 1.1690766108574329,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.021900210016824e-06,
+      "loss": 1.39180822,
+      "memory(GiB)": 123.79,
+      "step": 46085,
+      "train_speed(iter/s)": 1.315393
+    },
+    {
+      "acc": 0.69375062,
+      "epoch": 1.1692034500253679,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.020871867495139e-06,
+      "loss": 1.32546978,
+      "memory(GiB)": 123.79,
+      "step": 46090,
+      "train_speed(iter/s)": 1.315408
+    },
+    {
+      "acc": 0.69424324,
+      "epoch": 1.1693302891933028,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.01984356803988e-06,
+      "loss": 1.37738619,
+      "memory(GiB)": 123.79,
+      "step": 46095,
+      "train_speed(iter/s)": 1.315423
+    },
+    {
+      "acc": 0.70074587,
+      "epoch": 1.169457128361238,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.018815311696274e-06,
+      "loss": 1.39723997,
+      "memory(GiB)": 123.79,
+      "step": 46100,
+      "train_speed(iter/s)": 1.315439
+    },
+    {
+      "acc": 0.68125954,
+      "epoch": 1.169583967529173,
+      "grad_norm": 3.75,
+      "learning_rate": 4.017787098509555e-06,
+      "loss": 1.41632061,
+      "memory(GiB)": 123.79,
+      "step": 46105,
+      "train_speed(iter/s)": 1.315454
+    },
+    {
+      "acc": 0.67929301,
+      "epoch": 1.169710806697108,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.016758928524944e-06,
+      "loss": 1.38288555,
+      "memory(GiB)": 123.79,
+      "step": 46110,
+      "train_speed(iter/s)": 1.31547
+    },
+    {
+      "acc": 0.69875021,
+      "epoch": 1.169837645865043,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.015730801787663e-06,
+      "loss": 1.31708069,
+      "memory(GiB)": 123.79,
+      "step": 46115,
+      "train_speed(iter/s)": 1.315486
+    },
+    {
+      "acc": 0.67909226,
+      "epoch": 1.1699644850329782,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.014702718342938e-06,
+      "loss": 1.43501778,
+      "memory(GiB)": 123.79,
+      "step": 46120,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.6750679,
+      "epoch": 1.1700913242009132,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.013674678235985e-06,
+      "loss": 1.42419767,
+      "memory(GiB)": 123.79,
+      "step": 46125,
+      "train_speed(iter/s)": 1.315517
+    },
+    {
+      "acc": 0.67542315,
+      "epoch": 1.1702181633688482,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.012646681512026e-06,
+      "loss": 1.37643614,
+      "memory(GiB)": 123.79,
+      "step": 46130,
+      "train_speed(iter/s)": 1.315533
+    },
+    {
+      "acc": 0.68341904,
+      "epoch": 1.1703450025367834,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.011618728216271e-06,
+      "loss": 1.42209787,
+      "memory(GiB)": 123.79,
+      "step": 46135,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.68731632,
+      "epoch": 1.1704718417047184,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.010590818393938e-06,
+      "loss": 1.43263721,
+      "memory(GiB)": 123.79,
+      "step": 46140,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.70752087,
+      "epoch": 1.1705986808726534,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.009562952090238e-06,
+      "loss": 1.36190987,
+      "memory(GiB)": 123.79,
+      "step": 46145,
+      "train_speed(iter/s)": 1.315579
+    },
+    {
+      "acc": 0.6873282,
+      "epoch": 1.1707255200405886,
+      "grad_norm": 4.625,
+      "learning_rate": 4.0085351293503825e-06,
+      "loss": 1.36181602,
+      "memory(GiB)": 123.79,
+      "step": 46150,
+      "train_speed(iter/s)": 1.315595
+    },
+    {
+      "acc": 0.67735825,
+      "epoch": 1.1708523592085236,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.007507350219578e-06,
+      "loss": 1.48379707,
+      "memory(GiB)": 123.79,
+      "step": 46155,
+      "train_speed(iter/s)": 1.31561
+    },
+    {
+      "acc": 0.69402981,
+      "epoch": 1.1709791983764586,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.0064796147430305e-06,
+      "loss": 1.32116776,
+      "memory(GiB)": 123.79,
+      "step": 46160,
+      "train_speed(iter/s)": 1.315625
+    },
+    {
+      "acc": 0.70193424,
+      "epoch": 1.1711060375443938,
+      "grad_norm": 3.203125,
+      "learning_rate": 4.005451922965946e-06,
+      "loss": 1.39412718,
+      "memory(GiB)": 123.79,
+      "step": 46165,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.69144163,
+      "epoch": 1.1712328767123288,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.0044242749335285e-06,
+      "loss": 1.37276068,
+      "memory(GiB)": 123.79,
+      "step": 46170,
+      "train_speed(iter/s)": 1.315656
+    },
+    {
+      "acc": 0.67860532,
+      "epoch": 1.1713597158802638,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.0033966706909735e-06,
+      "loss": 1.48303432,
+      "memory(GiB)": 123.79,
+      "step": 46175,
+      "train_speed(iter/s)": 1.315671
+    },
+    {
+      "acc": 0.68229084,
+      "epoch": 1.1714865550481988,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.002369110283482e-06,
+      "loss": 1.37227736,
+      "memory(GiB)": 123.79,
+      "step": 46180,
+      "train_speed(iter/s)": 1.315683
+    },
+    {
+      "acc": 0.67065573,
+      "epoch": 1.171613394216134,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.001341593756253e-06,
+      "loss": 1.41720181,
+      "memory(GiB)": 123.79,
+      "step": 46185,
+      "train_speed(iter/s)": 1.315698
+    },
+    {
+      "acc": 0.6919796,
+      "epoch": 1.171740233384069,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.00031412115448e-06,
+      "loss": 1.40133896,
+      "memory(GiB)": 123.79,
+      "step": 46190,
+      "train_speed(iter/s)": 1.315713
+    },
+    {
+      "acc": 0.69482179,
+      "epoch": 1.1718670725520042,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.999286692523352e-06,
+      "loss": 1.34431648,
+      "memory(GiB)": 123.79,
+      "step": 46195,
+      "train_speed(iter/s)": 1.315729
+    },
+    {
+      "acc": 0.70324526,
+      "epoch": 1.1719939117199392,
+      "grad_norm": 2.875,
+      "learning_rate": 3.998259307908065e-06,
+      "loss": 1.32480764,
+      "memory(GiB)": 123.79,
+      "step": 46200,
+      "train_speed(iter/s)": 1.315745
+    },
+    {
+      "acc": 0.6730999,
+      "epoch": 1.1721207508878742,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.997231967353806e-06,
+      "loss": 1.40325079,
+      "memory(GiB)": 123.79,
+      "step": 46205,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.69416161,
+      "epoch": 1.1722475900558091,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.996204670905765e-06,
+      "loss": 1.36294975,
+      "memory(GiB)": 123.79,
+      "step": 46210,
+      "train_speed(iter/s)": 1.315769
+    },
+    {
+      "acc": 0.69891138,
+      "epoch": 1.1723744292237444,
+      "grad_norm": 3.0,
+      "learning_rate": 3.9951774186091195e-06,
+      "loss": 1.37695999,
+      "memory(GiB)": 123.79,
+      "step": 46215,
+      "train_speed(iter/s)": 1.315785
+    },
+    {
+      "acc": 0.71004887,
+      "epoch": 1.1725012683916793,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.9941502105090594e-06,
+      "loss": 1.32646275,
+      "memory(GiB)": 123.79,
+      "step": 46220,
+      "train_speed(iter/s)": 1.3158
+    },
+    {
+      "acc": 0.70500069,
+      "epoch": 1.1726281075596143,
+      "grad_norm": 3.125,
+      "learning_rate": 3.9931230466507634e-06,
+      "loss": 1.311273,
+      "memory(GiB)": 123.79,
+      "step": 46225,
+      "train_speed(iter/s)": 1.315815
+    },
+    {
+      "acc": 0.68890343,
+      "epoch": 1.1727549467275495,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.992095927079412e-06,
+      "loss": 1.44526882,
+      "memory(GiB)": 123.79,
+      "step": 46230,
+      "train_speed(iter/s)": 1.31583
+    },
+    {
+      "acc": 0.6979578,
+      "epoch": 1.1728817858954845,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.991068851840182e-06,
+      "loss": 1.32800379,
+      "memory(GiB)": 123.79,
+      "step": 46235,
+      "train_speed(iter/s)": 1.315846
+    },
+    {
+      "acc": 0.70595922,
+      "epoch": 1.1730086250634195,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.990041820978246e-06,
+      "loss": 1.31691494,
+      "memory(GiB)": 123.79,
+      "step": 46240,
+      "train_speed(iter/s)": 1.315861
+    },
+    {
+      "acc": 0.70193048,
+      "epoch": 1.1731354642313547,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.989014834538782e-06,
+      "loss": 1.3297081,
+      "memory(GiB)": 123.79,
+      "step": 46245,
+      "train_speed(iter/s)": 1.315877
+    },
+    {
+      "acc": 0.70072007,
+      "epoch": 1.1732623033992897,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.987987892566959e-06,
+      "loss": 1.33913116,
+      "memory(GiB)": 123.79,
+      "step": 46250,
+      "train_speed(iter/s)": 1.315893
+    },
+    {
+      "acc": 0.70038643,
+      "epoch": 1.1733891425672247,
+      "grad_norm": 2.75,
+      "learning_rate": 3.986960995107948e-06,
+      "loss": 1.35314789,
+      "memory(GiB)": 123.79,
+      "step": 46255,
+      "train_speed(iter/s)": 1.315908
+    },
+    {
+      "acc": 0.69886546,
+      "epoch": 1.17351598173516,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.985934142206912e-06,
+      "loss": 1.36158285,
+      "memory(GiB)": 123.79,
+      "step": 46260,
+      "train_speed(iter/s)": 1.315924
+    },
+    {
+      "acc": 0.69428949,
+      "epoch": 1.173642820903095,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.984907333909022e-06,
+      "loss": 1.3845644,
+      "memory(GiB)": 123.79,
+      "step": 46265,
+      "train_speed(iter/s)": 1.315939
+    },
+    {
+      "acc": 0.68058228,
+      "epoch": 1.17376966007103,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.983880570259441e-06,
+      "loss": 1.41612444,
+      "memory(GiB)": 123.79,
+      "step": 46270,
+      "train_speed(iter/s)": 1.315953
+    },
+    {
+      "acc": 0.69105368,
+      "epoch": 1.1738964992389649,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.982853851303327e-06,
+      "loss": 1.3674016,
+      "memory(GiB)": 123.79,
+      "step": 46275,
+      "train_speed(iter/s)": 1.315969
+    },
+    {
+      "acc": 0.67598543,
+      "epoch": 1.1740233384069,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.981827177085842e-06,
+      "loss": 1.44841805,
+      "memory(GiB)": 123.79,
+      "step": 46280,
+      "train_speed(iter/s)": 1.315984
+    },
+    {
+      "acc": 0.68981256,
+      "epoch": 1.174150177574835,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.980800547652143e-06,
+      "loss": 1.39773436,
+      "memory(GiB)": 123.79,
+      "step": 46285,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.66973696,
+      "epoch": 1.17427701674277,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.979773963047388e-06,
+      "loss": 1.52064171,
+      "memory(GiB)": 123.79,
+      "step": 46290,
+      "train_speed(iter/s)": 1.316017
+    },
+    {
+      "acc": 0.69034891,
+      "epoch": 1.1744038559107053,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.978747423316729e-06,
+      "loss": 1.35198698,
+      "memory(GiB)": 123.79,
+      "step": 46295,
+      "train_speed(iter/s)": 1.316032
+    },
+    {
+      "acc": 0.67752824,
+      "epoch": 1.1745306950786403,
+      "grad_norm": 2.875,
+      "learning_rate": 3.977720928505317e-06,
+      "loss": 1.45171394,
+      "memory(GiB)": 123.79,
+      "step": 46300,
+      "train_speed(iter/s)": 1.316048
+    },
+    {
+      "acc": 0.68747358,
+      "epoch": 1.1746575342465753,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.976694478658301e-06,
+      "loss": 1.45554571,
+      "memory(GiB)": 123.79,
+      "step": 46305,
+      "train_speed(iter/s)": 1.316064
+    },
+    {
+      "acc": 0.69581242,
+      "epoch": 1.1747843734145105,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.975668073820834e-06,
+      "loss": 1.34425526,
+      "memory(GiB)": 123.79,
+      "step": 46310,
+      "train_speed(iter/s)": 1.31608
+    },
+    {
+      "acc": 0.7152668,
+      "epoch": 1.1749112125824455,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.9746417140380576e-06,
+      "loss": 1.29026814,
+      "memory(GiB)": 123.79,
+      "step": 46315,
+      "train_speed(iter/s)": 1.316096
+    },
+    {
+      "acc": 0.68632259,
+      "epoch": 1.1750380517503805,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.973615399355114e-06,
+      "loss": 1.40670242,
+      "memory(GiB)": 123.79,
+      "step": 46320,
+      "train_speed(iter/s)": 1.316112
+    },
+    {
+      "acc": 0.68977466,
+      "epoch": 1.1751648909183157,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.97258912981715e-06,
+      "loss": 1.42138824,
+      "memory(GiB)": 123.79,
+      "step": 46325,
+      "train_speed(iter/s)": 1.316127
+    },
+    {
+      "acc": 0.69036579,
+      "epoch": 1.1752917300862507,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.9715629054693035e-06,
+      "loss": 1.35348377,
+      "memory(GiB)": 123.79,
+      "step": 46330,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.68883395,
+      "epoch": 1.1754185692541856,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.970536726356711e-06,
+      "loss": 1.37392406,
+      "memory(GiB)": 123.79,
+      "step": 46335,
+      "train_speed(iter/s)": 1.316158
+    },
+    {
+      "acc": 0.67591915,
+      "epoch": 1.1755454084221206,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.969510592524509e-06,
+      "loss": 1.40891504,
+      "memory(GiB)": 123.79,
+      "step": 46340,
+      "train_speed(iter/s)": 1.316173
+    },
+    {
+      "acc": 0.68295174,
+      "epoch": 1.1756722475900558,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.968484504017833e-06,
+      "loss": 1.43147678,
+      "memory(GiB)": 123.79,
+      "step": 46345,
+      "train_speed(iter/s)": 1.316188
+    },
+    {
+      "acc": 0.70099678,
+      "epoch": 1.1757990867579908,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.967458460881815e-06,
+      "loss": 1.31867065,
+      "memory(GiB)": 123.79,
+      "step": 46350,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.70284929,
+      "epoch": 1.175925925925926,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.96643246316158e-06,
+      "loss": 1.35255203,
+      "memory(GiB)": 123.79,
+      "step": 46355,
+      "train_speed(iter/s)": 1.316218
+    },
+    {
+      "acc": 0.69663553,
+      "epoch": 1.176052765093861,
+      "grad_norm": 2.75,
+      "learning_rate": 3.965406510902263e-06,
+      "loss": 1.32249594,
+      "memory(GiB)": 123.79,
+      "step": 46360,
+      "train_speed(iter/s)": 1.316233
+    },
+    {
+      "acc": 0.70267863,
+      "epoch": 1.176179604261796,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.9643806041489855e-06,
+      "loss": 1.34707756,
+      "memory(GiB)": 123.79,
+      "step": 46365,
+      "train_speed(iter/s)": 1.316249
+    },
+    {
+      "acc": 0.68232479,
+      "epoch": 1.176306443429731,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.963354742946874e-06,
+      "loss": 1.39652615,
+      "memory(GiB)": 123.79,
+      "step": 46370,
+      "train_speed(iter/s)": 1.316264
+    },
+    {
+      "acc": 0.68228378,
+      "epoch": 1.1764332825976662,
+      "grad_norm": 2.5,
+      "learning_rate": 3.962328927341048e-06,
+      "loss": 1.41256332,
+      "memory(GiB)": 123.79,
+      "step": 46375,
+      "train_speed(iter/s)": 1.31628
+    },
+    {
+      "acc": 0.70742941,
+      "epoch": 1.1765601217656012,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.961303157376628e-06,
+      "loss": 1.33966236,
+      "memory(GiB)": 123.79,
+      "step": 46380,
+      "train_speed(iter/s)": 1.316297
+    },
+    {
+      "acc": 0.68903456,
+      "epoch": 1.1766869609335362,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.960277433098734e-06,
+      "loss": 1.34593372,
+      "memory(GiB)": 123.79,
+      "step": 46385,
+      "train_speed(iter/s)": 1.316313
+    },
+    {
+      "acc": 0.67740593,
+      "epoch": 1.1768138001014714,
+      "grad_norm": 3.625,
+      "learning_rate": 3.959251754552481e-06,
+      "loss": 1.44443722,
+      "memory(GiB)": 123.79,
+      "step": 46390,
+      "train_speed(iter/s)": 1.31633
+    },
+    {
+      "acc": 0.6791018,
+      "epoch": 1.1769406392694064,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.958226121782982e-06,
+      "loss": 1.41794205,
+      "memory(GiB)": 123.79,
+      "step": 46395,
+      "train_speed(iter/s)": 1.316346
+    },
+    {
+      "acc": 0.68538756,
+      "epoch": 1.1770674784373414,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.9572005348353486e-06,
+      "loss": 1.36181803,
+      "memory(GiB)": 123.79,
+      "step": 46400,
+      "train_speed(iter/s)": 1.316359
+    },
+    {
+      "acc": 0.67814035,
+      "epoch": 1.1771943176052766,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.956174993754691e-06,
+      "loss": 1.42213707,
+      "memory(GiB)": 123.79,
+      "step": 46405,
+      "train_speed(iter/s)": 1.316375
+    },
+    {
+      "acc": 0.70679693,
+      "epoch": 1.1773211567732116,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.955149498586119e-06,
+      "loss": 1.27241516,
+      "memory(GiB)": 123.79,
+      "step": 46410,
+      "train_speed(iter/s)": 1.316391
+    },
+    {
+      "acc": 0.67840056,
+      "epoch": 1.1774479959411466,
+      "grad_norm": 3.984375,
+      "learning_rate": 3.954124049374736e-06,
+      "loss": 1.41511841,
+      "memory(GiB)": 123.79,
+      "step": 46415,
+      "train_speed(iter/s)": 1.316407
+    },
+    {
+      "acc": 0.70020328,
+      "epoch": 1.1775748351090818,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.9530986461656465e-06,
+      "loss": 1.3452486,
+      "memory(GiB)": 123.79,
+      "step": 46420,
+      "train_speed(iter/s)": 1.316423
+    },
+    {
+      "acc": 0.68538132,
+      "epoch": 1.1777016742770168,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.952073289003953e-06,
+      "loss": 1.36403179,
+      "memory(GiB)": 123.79,
+      "step": 46425,
+      "train_speed(iter/s)": 1.316438
+    },
+    {
+      "acc": 0.69631033,
+      "epoch": 1.1778285134449518,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.9510479779347566e-06,
+      "loss": 1.28096142,
+      "memory(GiB)": 123.79,
+      "step": 46430,
+      "train_speed(iter/s)": 1.316455
+    },
+    {
+      "acc": 0.68468819,
+      "epoch": 1.1779553526128868,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.950022713003151e-06,
+      "loss": 1.39174051,
+      "memory(GiB)": 123.79,
+      "step": 46435,
+      "train_speed(iter/s)": 1.31647
+    },
+    {
+      "acc": 0.68139601,
+      "epoch": 1.178082191780822,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.9489974942542355e-06,
+      "loss": 1.39609261,
+      "memory(GiB)": 123.79,
+      "step": 46440,
+      "train_speed(iter/s)": 1.316486
+    },
+    {
+      "acc": 0.68623004,
+      "epoch": 1.178209030948757,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.947972321733101e-06,
+      "loss": 1.38534365,
+      "memory(GiB)": 123.79,
+      "step": 46445,
+      "train_speed(iter/s)": 1.316501
+    },
+    {
+      "acc": 0.68749619,
+      "epoch": 1.178335870116692,
+      "grad_norm": 2.875,
+      "learning_rate": 3.946947195484843e-06,
+      "loss": 1.40095329,
+      "memory(GiB)": 123.79,
+      "step": 46450,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.68422561,
+      "epoch": 1.1784627092846272,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.945922115554548e-06,
+      "loss": 1.40433054,
+      "memory(GiB)": 123.79,
+      "step": 46455,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.69569354,
+      "epoch": 1.1785895484525621,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.944897081987303e-06,
+      "loss": 1.43838787,
+      "memory(GiB)": 123.79,
+      "step": 46460,
+      "train_speed(iter/s)": 1.31655
+    },
+    {
+      "acc": 0.68480082,
+      "epoch": 1.1787163876204971,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.943872094828197e-06,
+      "loss": 1.42381802,
+      "memory(GiB)": 123.79,
+      "step": 46465,
+      "train_speed(iter/s)": 1.316566
+    },
+    {
+      "acc": 0.70209532,
+      "epoch": 1.1788432267884323,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.942847154122312e-06,
+      "loss": 1.31154842,
+      "memory(GiB)": 123.79,
+      "step": 46470,
+      "train_speed(iter/s)": 1.316581
+    },
+    {
+      "acc": 0.70131817,
+      "epoch": 1.1789700659563673,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.941822259914728e-06,
+      "loss": 1.31707096,
+      "memory(GiB)": 123.79,
+      "step": 46475,
+      "train_speed(iter/s)": 1.316597
+    },
+    {
+      "acc": 0.69371557,
+      "epoch": 1.1790969051243023,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.940797412250524e-06,
+      "loss": 1.39058342,
+      "memory(GiB)": 123.79,
+      "step": 46480,
+      "train_speed(iter/s)": 1.316613
+    },
+    {
+      "acc": 0.69420447,
+      "epoch": 1.1792237442922375,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.93977261117478e-06,
+      "loss": 1.39268932,
+      "memory(GiB)": 123.79,
+      "step": 46485,
+      "train_speed(iter/s)": 1.316628
+    },
+    {
+      "acc": 0.68110576,
+      "epoch": 1.1793505834601725,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.938747856732572e-06,
+      "loss": 1.38896456,
+      "memory(GiB)": 123.79,
+      "step": 46490,
+      "train_speed(iter/s)": 1.316644
+    },
+    {
+      "acc": 0.70160961,
+      "epoch": 1.1794774226281075,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.9377231489689685e-06,
+      "loss": 1.42218685,
+      "memory(GiB)": 123.79,
+      "step": 46495,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.68103342,
+      "epoch": 1.1796042617960425,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.936698487929045e-06,
+      "loss": 1.40069466,
+      "memory(GiB)": 123.79,
+      "step": 46500,
+      "train_speed(iter/s)": 1.316676
+    },
+    {
+      "acc": 0.69986963,
+      "epoch": 1.1797311009639777,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.935673873657868e-06,
+      "loss": 1.33889294,
+      "memory(GiB)": 123.79,
+      "step": 46505,
+      "train_speed(iter/s)": 1.316692
+    },
+    {
+      "acc": 0.68571315,
+      "epoch": 1.1798579401319127,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.934649306200508e-06,
+      "loss": 1.44221973,
+      "memory(GiB)": 123.79,
+      "step": 46510,
+      "train_speed(iter/s)": 1.316708
+    },
+    {
+      "acc": 0.68953629,
+      "epoch": 1.179984779299848,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.933624785602027e-06,
+      "loss": 1.3811203,
+      "memory(GiB)": 123.79,
+      "step": 46515,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.67757158,
+      "epoch": 1.180111618467783,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.932600311907489e-06,
+      "loss": 1.39437084,
+      "memory(GiB)": 123.79,
+      "step": 46520,
+      "train_speed(iter/s)": 1.31674
+    },
+    {
+      "acc": 0.68357458,
+      "epoch": 1.1802384576357179,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.931575885161955e-06,
+      "loss": 1.43806973,
+      "memory(GiB)": 123.79,
+      "step": 46525,
+      "train_speed(iter/s)": 1.316756
+    },
+    {
+      "acc": 0.68467393,
+      "epoch": 1.1803652968036529,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.930551505410484e-06,
+      "loss": 1.38076906,
+      "memory(GiB)": 123.79,
+      "step": 46530,
+      "train_speed(iter/s)": 1.316772
+    },
+    {
+      "acc": 0.68761339,
+      "epoch": 1.180492135971588,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.929527172698132e-06,
+      "loss": 1.38539085,
+      "memory(GiB)": 123.79,
+      "step": 46535,
+      "train_speed(iter/s)": 1.316788
+    },
+    {
+      "acc": 0.67799931,
+      "epoch": 1.180618975139523,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.928502887069954e-06,
+      "loss": 1.40552845,
+      "memory(GiB)": 123.79,
+      "step": 46540,
+      "train_speed(iter/s)": 1.316804
+    },
+    {
+      "acc": 0.70636544,
+      "epoch": 1.180745814307458,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.927478648571003e-06,
+      "loss": 1.27560921,
+      "memory(GiB)": 123.79,
+      "step": 46545,
+      "train_speed(iter/s)": 1.316819
+    },
+    {
+      "acc": 0.68867836,
+      "epoch": 1.1808726534753933,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.926454457246331e-06,
+      "loss": 1.37207851,
+      "memory(GiB)": 123.79,
+      "step": 46550,
+      "train_speed(iter/s)": 1.316835
+    },
+    {
+      "acc": 0.68349924,
+      "epoch": 1.1809994926433283,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.9254303131409834e-06,
+      "loss": 1.41963758,
+      "memory(GiB)": 123.79,
+      "step": 46555,
+      "train_speed(iter/s)": 1.316851
+    },
+    {
+      "acc": 0.68591766,
+      "epoch": 1.1811263318112633,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.924406216300009e-06,
+      "loss": 1.34128351,
+      "memory(GiB)": 123.79,
+      "step": 46560,
+      "train_speed(iter/s)": 1.316866
+    },
+    {
+      "acc": 0.69796562,
+      "epoch": 1.1812531709791985,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.92338216676845e-06,
+      "loss": 1.36704254,
+      "memory(GiB)": 123.79,
+      "step": 46565,
+      "train_speed(iter/s)": 1.31688
+    },
+    {
+      "acc": 0.68189445,
+      "epoch": 1.1813800101471335,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.922358164591353e-06,
+      "loss": 1.43811722,
+      "memory(GiB)": 123.79,
+      "step": 46570,
+      "train_speed(iter/s)": 1.316895
+    },
+    {
+      "acc": 0.68546963,
+      "epoch": 1.1815068493150684,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.921334209813752e-06,
+      "loss": 1.42886915,
+      "memory(GiB)": 123.79,
+      "step": 46575,
+      "train_speed(iter/s)": 1.31691
+    },
+    {
+      "acc": 0.68037677,
+      "epoch": 1.1816336884830037,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.92031030248069e-06,
+      "loss": 1.42016954,
+      "memory(GiB)": 123.79,
+      "step": 46580,
+      "train_speed(iter/s)": 1.316925
+    },
+    {
+      "acc": 0.67311115,
+      "epoch": 1.1817605276509386,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.9192864426372e-06,
+      "loss": 1.44830227,
+      "memory(GiB)": 123.79,
+      "step": 46585,
+      "train_speed(iter/s)": 1.31694
+    },
+    {
+      "acc": 0.70228877,
+      "epoch": 1.1818873668188736,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.918262630328319e-06,
+      "loss": 1.34001236,
+      "memory(GiB)": 123.79,
+      "step": 46590,
+      "train_speed(iter/s)": 1.316955
+    },
+    {
+      "acc": 0.6912776,
+      "epoch": 1.1820142059868086,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.917238865599077e-06,
+      "loss": 1.39114113,
+      "memory(GiB)": 123.79,
+      "step": 46595,
+      "train_speed(iter/s)": 1.316971
+    },
+    {
+      "acc": 0.7116147,
+      "epoch": 1.1821410451547438,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.916215148494502e-06,
+      "loss": 1.29096909,
+      "memory(GiB)": 123.79,
+      "step": 46600,
+      "train_speed(iter/s)": 1.316986
+    },
+    {
+      "acc": 0.67344971,
+      "epoch": 1.1822678843226788,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.9151914790596255e-06,
+      "loss": 1.44645576,
+      "memory(GiB)": 123.79,
+      "step": 46605,
+      "train_speed(iter/s)": 1.317002
+    },
+    {
+      "acc": 0.69331203,
+      "epoch": 1.1823947234906138,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.914167857339472e-06,
+      "loss": 1.37187243,
+      "memory(GiB)": 123.79,
+      "step": 46610,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.6796361,
+      "epoch": 1.182521562658549,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.913144283379061e-06,
+      "loss": 1.41992416,
+      "memory(GiB)": 123.79,
+      "step": 46615,
+      "train_speed(iter/s)": 1.317033
+    },
+    {
+      "acc": 0.68675184,
+      "epoch": 1.182648401826484,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.912120757223418e-06,
+      "loss": 1.46268253,
+      "memory(GiB)": 123.79,
+      "step": 46620,
+      "train_speed(iter/s)": 1.317048
+    },
+    {
+      "acc": 0.68871031,
+      "epoch": 1.182775240994419,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.911097278917561e-06,
+      "loss": 1.40493126,
+      "memory(GiB)": 123.79,
+      "step": 46625,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.69593263,
+      "epoch": 1.1829020801623542,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.91007384850651e-06,
+      "loss": 1.35826244,
+      "memory(GiB)": 123.79,
+      "step": 46630,
+      "train_speed(iter/s)": 1.317074
+    },
+    {
+      "acc": 0.69268327,
+      "epoch": 1.1830289193302892,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.909050466035274e-06,
+      "loss": 1.32035313,
+      "memory(GiB)": 123.79,
+      "step": 46635,
+      "train_speed(iter/s)": 1.31709
+    },
+    {
+      "acc": 0.69288874,
+      "epoch": 1.1831557584982242,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.90802713154887e-06,
+      "loss": 1.38254499,
+      "memory(GiB)": 123.79,
+      "step": 46640,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.70009661,
+      "epoch": 1.1832825976661594,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.9070038450923074e-06,
+      "loss": 1.30189257,
+      "memory(GiB)": 123.79,
+      "step": 46645,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.68188705,
+      "epoch": 1.1834094368340944,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.9059806067105985e-06,
+      "loss": 1.41820021,
+      "memory(GiB)": 123.79,
+      "step": 46650,
+      "train_speed(iter/s)": 1.317135
+    },
+    {
+      "acc": 0.68026419,
+      "epoch": 1.1835362760020294,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.904957416448744e-06,
+      "loss": 1.44503593,
+      "memory(GiB)": 123.79,
+      "step": 46655,
+      "train_speed(iter/s)": 1.317151
+    },
+    {
+      "acc": 0.70046358,
+      "epoch": 1.1836631151699644,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.903934274351753e-06,
+      "loss": 1.35609798,
+      "memory(GiB)": 123.79,
+      "step": 46660,
+      "train_speed(iter/s)": 1.317166
+    },
+    {
+      "acc": 0.69337754,
+      "epoch": 1.1837899543378996,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.9029111804646245e-06,
+      "loss": 1.37340412,
+      "memory(GiB)": 123.79,
+      "step": 46665,
+      "train_speed(iter/s)": 1.317181
+    },
+    {
+      "acc": 0.68453789,
+      "epoch": 1.1839167935058346,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.9018881348323626e-06,
+      "loss": 1.3929678,
+      "memory(GiB)": 123.79,
+      "step": 46670,
+      "train_speed(iter/s)": 1.317197
+    },
+    {
+      "acc": 0.69915466,
+      "epoch": 1.1840436326737698,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.9008651374999615e-06,
+      "loss": 1.35928965,
+      "memory(GiB)": 123.79,
+      "step": 46675,
+      "train_speed(iter/s)": 1.317212
+    },
+    {
+      "acc": 0.68748212,
+      "epoch": 1.1841704718417048,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.899842188512419e-06,
+      "loss": 1.37234077,
+      "memory(GiB)": 123.79,
+      "step": 46680,
+      "train_speed(iter/s)": 1.317227
+    },
+    {
+      "acc": 0.68510275,
+      "epoch": 1.1842973110096398,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.898819287914729e-06,
+      "loss": 1.33716507,
+      "memory(GiB)": 123.79,
+      "step": 46685,
+      "train_speed(iter/s)": 1.317242
+    },
+    {
+      "acc": 0.68305631,
+      "epoch": 1.1844241501775747,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.897796435751885e-06,
+      "loss": 1.40004253,
+      "memory(GiB)": 123.79,
+      "step": 46690,
+      "train_speed(iter/s)": 1.317258
+    },
+    {
+      "acc": 0.67488976,
+      "epoch": 1.18455098934551,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.896773632068873e-06,
+      "loss": 1.44177647,
+      "memory(GiB)": 123.79,
+      "step": 46695,
+      "train_speed(iter/s)": 1.317272
+    },
+    {
+      "acc": 0.67761774,
+      "epoch": 1.184677828513445,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.8957508769106825e-06,
+      "loss": 1.41289406,
+      "memory(GiB)": 123.79,
+      "step": 46700,
+      "train_speed(iter/s)": 1.317288
+    },
+    {
+      "acc": 0.69372339,
+      "epoch": 1.18480466768138,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.894728170322298e-06,
+      "loss": 1.41407156,
+      "memory(GiB)": 123.79,
+      "step": 46705,
+      "train_speed(iter/s)": 1.317303
+    },
+    {
+      "acc": 0.70280962,
+      "epoch": 1.1849315068493151,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.893705512348705e-06,
+      "loss": 1.31024981,
+      "memory(GiB)": 123.79,
+      "step": 46710,
+      "train_speed(iter/s)": 1.317319
+    },
+    {
+      "acc": 0.70112958,
+      "epoch": 1.1850583460172501,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.89268290303488e-06,
+      "loss": 1.39733953,
+      "memory(GiB)": 123.79,
+      "step": 46715,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.69863043,
+      "epoch": 1.1851851851851851,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.891660342425807e-06,
+      "loss": 1.30222616,
+      "memory(GiB)": 123.79,
+      "step": 46720,
+      "train_speed(iter/s)": 1.317348
+    },
+    {
+      "acc": 0.7052403,
+      "epoch": 1.1853120243531203,
+      "grad_norm": 3.75,
+      "learning_rate": 3.890637830566459e-06,
+      "loss": 1.30658674,
+      "memory(GiB)": 123.79,
+      "step": 46725,
+      "train_speed(iter/s)": 1.317363
+    },
+    {
+      "acc": 0.68069739,
+      "epoch": 1.1854388635210553,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.889615367501815e-06,
+      "loss": 1.43001318,
+      "memory(GiB)": 123.79,
+      "step": 46730,
+      "train_speed(iter/s)": 1.317377
+    },
+    {
+      "acc": 0.69352465,
+      "epoch": 1.1855657026889903,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.888592953276842e-06,
+      "loss": 1.3906004,
+      "memory(GiB)": 123.79,
+      "step": 46735,
+      "train_speed(iter/s)": 1.317392
+    },
+    {
+      "acc": 0.701333,
+      "epoch": 1.1856925418569255,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.8875705879365135e-06,
+      "loss": 1.28786316,
+      "memory(GiB)": 123.79,
+      "step": 46740,
+      "train_speed(iter/s)": 1.317407
+    },
+    {
+      "acc": 0.68821754,
+      "epoch": 1.1858193810248605,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.886548271525797e-06,
+      "loss": 1.34163456,
+      "memory(GiB)": 123.79,
+      "step": 46745,
+      "train_speed(iter/s)": 1.317423
+    },
+    {
+      "acc": 0.69338322,
+      "epoch": 1.1859462201927955,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.88552600408966e-06,
+      "loss": 1.33013563,
+      "memory(GiB)": 123.79,
+      "step": 46750,
+      "train_speed(iter/s)": 1.317438
+    },
+    {
+      "acc": 0.68133354,
+      "epoch": 1.1860730593607305,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.8845037856730646e-06,
+      "loss": 1.44606552,
+      "memory(GiB)": 123.79,
+      "step": 46755,
+      "train_speed(iter/s)": 1.317454
+    },
+    {
+      "acc": 0.69158363,
+      "epoch": 1.1861998985286657,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.883481616320972e-06,
+      "loss": 1.42699051,
+      "memory(GiB)": 123.79,
+      "step": 46760,
+      "train_speed(iter/s)": 1.31747
+    },
+    {
+      "acc": 0.69240856,
+      "epoch": 1.1863267376966007,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.882459496078343e-06,
+      "loss": 1.37620754,
+      "memory(GiB)": 123.79,
+      "step": 46765,
+      "train_speed(iter/s)": 1.317485
+    },
+    {
+      "acc": 0.69316225,
+      "epoch": 1.1864535768645357,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.881437424990137e-06,
+      "loss": 1.35088587,
+      "memory(GiB)": 123.79,
+      "step": 46770,
+      "train_speed(iter/s)": 1.317501
+    },
+    {
+      "acc": 0.68690495,
+      "epoch": 1.1865804160324709,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.880415403101304e-06,
+      "loss": 1.39860163,
+      "memory(GiB)": 123.79,
+      "step": 46775,
+      "train_speed(iter/s)": 1.317516
+    },
+    {
+      "acc": 0.7002502,
+      "epoch": 1.1867072552004059,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.879393430456801e-06,
+      "loss": 1.30195341,
+      "memory(GiB)": 123.79,
+      "step": 46780,
+      "train_speed(iter/s)": 1.31753
+    },
+    {
+      "acc": 0.68429499,
+      "epoch": 1.1868340943683409,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.87837150710158e-06,
+      "loss": 1.41855764,
+      "memory(GiB)": 123.79,
+      "step": 46785,
+      "train_speed(iter/s)": 1.317546
+    },
+    {
+      "acc": 0.6979497,
+      "epoch": 1.186960933536276,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.877349633080587e-06,
+      "loss": 1.30125418,
+      "memory(GiB)": 123.79,
+      "step": 46790,
+      "train_speed(iter/s)": 1.317562
+    },
+    {
+      "acc": 0.68779621,
+      "epoch": 1.187087772704211,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.876327808438767e-06,
+      "loss": 1.40306034,
+      "memory(GiB)": 123.79,
+      "step": 46795,
+      "train_speed(iter/s)": 1.317577
+    },
+    {
+      "acc": 0.68719692,
+      "epoch": 1.187214611872146,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.875306033221069e-06,
+      "loss": 1.3426218,
+      "memory(GiB)": 123.79,
+      "step": 46800,
+      "train_speed(iter/s)": 1.317593
+    },
+    {
+      "acc": 0.67520261,
+      "epoch": 1.1873414510400813,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.874284307472432e-06,
+      "loss": 1.45037403,
+      "memory(GiB)": 123.79,
+      "step": 46805,
+      "train_speed(iter/s)": 1.317607
+    },
+    {
+      "acc": 0.68502216,
+      "epoch": 1.1874682902080163,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.873262631237799e-06,
+      "loss": 1.43445721,
+      "memory(GiB)": 123.79,
+      "step": 46810,
+      "train_speed(iter/s)": 1.317623
+    },
+    {
+      "acc": 0.67124805,
+      "epoch": 1.1875951293759512,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.872241004562105e-06,
+      "loss": 1.44717197,
+      "memory(GiB)": 123.79,
+      "step": 46815,
+      "train_speed(iter/s)": 1.317639
+    },
+    {
+      "acc": 0.70101442,
+      "epoch": 1.1877219685438862,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.871219427490285e-06,
+      "loss": 1.34332638,
+      "memory(GiB)": 123.79,
+      "step": 46820,
+      "train_speed(iter/s)": 1.317655
+    },
+    {
+      "acc": 0.67918711,
+      "epoch": 1.1878488077118214,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.870197900067276e-06,
+      "loss": 1.43145742,
+      "memory(GiB)": 123.79,
+      "step": 46825,
+      "train_speed(iter/s)": 1.31767
+    },
+    {
+      "acc": 0.69239693,
+      "epoch": 1.1879756468797564,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.869176422338009e-06,
+      "loss": 1.34182491,
+      "memory(GiB)": 123.79,
+      "step": 46830,
+      "train_speed(iter/s)": 1.317686
+    },
+    {
+      "acc": 0.69784698,
+      "epoch": 1.1881024860476916,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.868154994347409e-06,
+      "loss": 1.37059307,
+      "memory(GiB)": 123.79,
+      "step": 46835,
+      "train_speed(iter/s)": 1.317702
+    },
+    {
+      "acc": 0.69796863,
+      "epoch": 1.1882293252156266,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.867133616140406e-06,
+      "loss": 1.32877769,
+      "memory(GiB)": 123.79,
+      "step": 46840,
+      "train_speed(iter/s)": 1.317717
+    },
+    {
+      "acc": 0.69145765,
+      "epoch": 1.1883561643835616,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.866112287761926e-06,
+      "loss": 1.42933407,
+      "memory(GiB)": 123.79,
+      "step": 46845,
+      "train_speed(iter/s)": 1.317729
+    },
+    {
+      "acc": 0.67695403,
+      "epoch": 1.1884830035514966,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.86509100925689e-06,
+      "loss": 1.42509699,
+      "memory(GiB)": 123.79,
+      "step": 46850,
+      "train_speed(iter/s)": 1.317745
+    },
+    {
+      "acc": 0.68063402,
+      "epoch": 1.1886098427194318,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.8640697806702166e-06,
+      "loss": 1.37240639,
+      "memory(GiB)": 123.79,
+      "step": 46855,
+      "train_speed(iter/s)": 1.31776
+    },
+    {
+      "acc": 0.68602352,
+      "epoch": 1.1887366818873668,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.8630486020468265e-06,
+      "loss": 1.43724575,
+      "memory(GiB)": 123.79,
+      "step": 46860,
+      "train_speed(iter/s)": 1.317775
+    },
+    {
+      "acc": 0.68565378,
+      "epoch": 1.1888635210553018,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.862027473431634e-06,
+      "loss": 1.39058533,
+      "memory(GiB)": 123.79,
+      "step": 46865,
+      "train_speed(iter/s)": 1.317791
+    },
+    {
+      "acc": 0.6967289,
+      "epoch": 1.188990360223237,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.861006394869558e-06,
+      "loss": 1.3247036,
+      "memory(GiB)": 123.79,
+      "step": 46870,
+      "train_speed(iter/s)": 1.317808
+    },
+    {
+      "acc": 0.68256283,
+      "epoch": 1.189117199391172,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.859985366405502e-06,
+      "loss": 1.39694843,
+      "memory(GiB)": 123.79,
+      "step": 46875,
+      "train_speed(iter/s)": 1.317823
+    },
+    {
+      "acc": 0.68075562,
+      "epoch": 1.189244038559107,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.85896438808438e-06,
+      "loss": 1.38305054,
+      "memory(GiB)": 123.79,
+      "step": 46880,
+      "train_speed(iter/s)": 1.317838
+    },
+    {
+      "acc": 0.67779598,
+      "epoch": 1.1893708777270422,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.857943459951099e-06,
+      "loss": 1.39909115,
+      "memory(GiB)": 123.79,
+      "step": 46885,
+      "train_speed(iter/s)": 1.317853
+    },
+    {
+      "acc": 0.67816582,
+      "epoch": 1.1894977168949772,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.856922582050565e-06,
+      "loss": 1.41627207,
+      "memory(GiB)": 123.79,
+      "step": 46890,
+      "train_speed(iter/s)": 1.317868
+    },
+    {
+      "acc": 0.71086202,
+      "epoch": 1.1896245560629122,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.855901754427678e-06,
+      "loss": 1.30012217,
+      "memory(GiB)": 123.79,
+      "step": 46895,
+      "train_speed(iter/s)": 1.317883
+    },
+    {
+      "acc": 0.67456226,
+      "epoch": 1.1897513952308474,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.854880977127339e-06,
+      "loss": 1.45954714,
+      "memory(GiB)": 123.79,
+      "step": 46900,
+      "train_speed(iter/s)": 1.317898
+    },
+    {
+      "acc": 0.68508897,
+      "epoch": 1.1898782343987824,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.8538602501944475e-06,
+      "loss": 1.36161146,
+      "memory(GiB)": 123.79,
+      "step": 46905,
+      "train_speed(iter/s)": 1.317914
+    },
+    {
+      "acc": 0.68501606,
+      "epoch": 1.1900050735667174,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.852839573673902e-06,
+      "loss": 1.36325655,
+      "memory(GiB)": 123.79,
+      "step": 46910,
+      "train_speed(iter/s)": 1.317929
+    },
+    {
+      "acc": 0.69204941,
+      "epoch": 1.1901319127346524,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.851818947610591e-06,
+      "loss": 1.34662895,
+      "memory(GiB)": 123.79,
+      "step": 46915,
+      "train_speed(iter/s)": 1.317944
+    },
+    {
+      "acc": 0.68728132,
+      "epoch": 1.1902587519025876,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.850798372049409e-06,
+      "loss": 1.3652626,
+      "memory(GiB)": 123.79,
+      "step": 46920,
+      "train_speed(iter/s)": 1.31796
+    },
+    {
+      "acc": 0.68730698,
+      "epoch": 1.1903855910705226,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.849777847035246e-06,
+      "loss": 1.36479778,
+      "memory(GiB)": 123.79,
+      "step": 46925,
+      "train_speed(iter/s)": 1.317975
+    },
+    {
+      "acc": 0.70046406,
+      "epoch": 1.1905124302384575,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.84875737261299e-06,
+      "loss": 1.29375982,
+      "memory(GiB)": 123.79,
+      "step": 46930,
+      "train_speed(iter/s)": 1.317991
+    },
+    {
+      "acc": 0.68175383,
+      "epoch": 1.1906392694063928,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.847736948827523e-06,
+      "loss": 1.42400389,
+      "memory(GiB)": 123.79,
+      "step": 46935,
+      "train_speed(iter/s)": 1.318007
+    },
+    {
+      "acc": 0.70417972,
+      "epoch": 1.1907661085743277,
+      "grad_norm": 3.625,
+      "learning_rate": 3.846716575723729e-06,
+      "loss": 1.34331245,
+      "memory(GiB)": 123.79,
+      "step": 46940,
+      "train_speed(iter/s)": 1.318022
+    },
+    {
+      "acc": 0.6856678,
+      "epoch": 1.1908929477422627,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.845696253346489e-06,
+      "loss": 1.42058449,
+      "memory(GiB)": 123.79,
+      "step": 46945,
+      "train_speed(iter/s)": 1.318039
+    },
+    {
+      "acc": 0.69816961,
+      "epoch": 1.191019786910198,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.8446759817406835e-06,
+      "loss": 1.37235098,
+      "memory(GiB)": 123.79,
+      "step": 46950,
+      "train_speed(iter/s)": 1.318054
+    },
+    {
+      "acc": 0.69119949,
+      "epoch": 1.191146626078133,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.8436557609511856e-06,
+      "loss": 1.41429176,
+      "memory(GiB)": 123.79,
+      "step": 46955,
+      "train_speed(iter/s)": 1.31807
+    },
+    {
+      "acc": 0.71074991,
+      "epoch": 1.191273465246068,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.842635591022869e-06,
+      "loss": 1.24012508,
+      "memory(GiB)": 123.79,
+      "step": 46960,
+      "train_speed(iter/s)": 1.318085
+    },
+    {
+      "acc": 0.688763,
+      "epoch": 1.1914003044140031,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.8416154720006065e-06,
+      "loss": 1.3917942,
+      "memory(GiB)": 123.79,
+      "step": 46965,
+      "train_speed(iter/s)": 1.318098
+    },
+    {
+      "acc": 0.68746405,
+      "epoch": 1.1915271435819381,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.840595403929269e-06,
+      "loss": 1.39678249,
+      "memory(GiB)": 123.79,
+      "step": 46970,
+      "train_speed(iter/s)": 1.318114
+    },
+    {
+      "acc": 0.67695346,
+      "epoch": 1.191653982749873,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.839575386853721e-06,
+      "loss": 1.34611855,
+      "memory(GiB)": 123.79,
+      "step": 46975,
+      "train_speed(iter/s)": 1.318129
+    },
+    {
+      "acc": 0.68064098,
+      "epoch": 1.191780821917808,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.838555420818827e-06,
+      "loss": 1.3653739,
+      "memory(GiB)": 123.79,
+      "step": 46980,
+      "train_speed(iter/s)": 1.318145
+    },
+    {
+      "acc": 0.69069147,
+      "epoch": 1.1919076610857433,
+      "grad_norm": 3.25,
+      "learning_rate": 3.837535505869453e-06,
+      "loss": 1.38198204,
+      "memory(GiB)": 123.79,
+      "step": 46985,
+      "train_speed(iter/s)": 1.318161
+    },
+    {
+      "acc": 0.68272495,
+      "epoch": 1.1920345002536783,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.836515642050458e-06,
+      "loss": 1.37022676,
+      "memory(GiB)": 123.79,
+      "step": 46990,
+      "train_speed(iter/s)": 1.318175
+    },
+    {
+      "acc": 0.67018447,
+      "epoch": 1.1921613394216135,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.835495829406698e-06,
+      "loss": 1.3952795,
+      "memory(GiB)": 123.79,
+      "step": 46995,
+      "train_speed(iter/s)": 1.31819
+    },
+    {
+      "acc": 0.69471574,
+      "epoch": 1.1922881785895485,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.834476067983031e-06,
+      "loss": 1.40057621,
+      "memory(GiB)": 123.79,
+      "step": 47000,
+      "train_speed(iter/s)": 1.318205
+    },
+    {
+      "epoch": 1.1922881785895485,
+      "eval_acc": 0.6750311679687255,
+      "eval_loss": 1.3565434217453003,
+      "eval_runtime": 69.7838,
+      "eval_samples_per_second": 91.282,
+      "eval_steps_per_second": 22.828,
+      "step": 47000
+    },
+    {
+      "acc": 0.68570247,
+      "epoch": 1.1924150177574835,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.83345635782431e-06,
+      "loss": 1.44495525,
+      "memory(GiB)": 123.79,
+      "step": 47005,
+      "train_speed(iter/s)": 1.315184
+    },
+    {
+      "acc": 0.68767991,
+      "epoch": 1.1925418569254185,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.832436698975388e-06,
+      "loss": 1.41195889,
+      "memory(GiB)": 123.79,
+      "step": 47010,
+      "train_speed(iter/s)": 1.315199
+    },
+    {
+      "acc": 0.70514317,
+      "epoch": 1.1926686960933537,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.831417091481111e-06,
+      "loss": 1.32060776,
+      "memory(GiB)": 123.79,
+      "step": 47015,
+      "train_speed(iter/s)": 1.315215
+    },
+    {
+      "acc": 0.70139217,
+      "epoch": 1.1927955352612887,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.830397535386328e-06,
+      "loss": 1.36042089,
+      "memory(GiB)": 123.79,
+      "step": 47020,
+      "train_speed(iter/s)": 1.31523
+    },
+    {
+      "acc": 0.68572769,
+      "epoch": 1.1929223744292237,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.829378030735883e-06,
+      "loss": 1.42774363,
+      "memory(GiB)": 123.79,
+      "step": 47025,
+      "train_speed(iter/s)": 1.315246
+    },
+    {
+      "acc": 0.69216671,
+      "epoch": 1.1930492135971589,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.82835857757462e-06,
+      "loss": 1.3743022,
+      "memory(GiB)": 123.79,
+      "step": 47030,
+      "train_speed(iter/s)": 1.315261
+    },
+    {
+      "acc": 0.69256554,
+      "epoch": 1.1931760527650939,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.827339175947378e-06,
+      "loss": 1.40611544,
+      "memory(GiB)": 123.79,
+      "step": 47035,
+      "train_speed(iter/s)": 1.315276
+    },
+    {
+      "acc": 0.68874617,
+      "epoch": 1.1933028919330289,
+      "grad_norm": 3.984375,
+      "learning_rate": 3.826319825898992e-06,
+      "loss": 1.35073814,
+      "memory(GiB)": 123.79,
+      "step": 47040,
+      "train_speed(iter/s)": 1.315288
+    },
+    {
+      "acc": 0.69822893,
+      "epoch": 1.193429731100964,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.825300527474302e-06,
+      "loss": 1.38428402,
+      "memory(GiB)": 123.79,
+      "step": 47045,
+      "train_speed(iter/s)": 1.315303
+    },
+    {
+      "acc": 0.68123035,
+      "epoch": 1.193556570268899,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.824281280718141e-06,
+      "loss": 1.43733149,
+      "memory(GiB)": 123.79,
+      "step": 47050,
+      "train_speed(iter/s)": 1.315318
+    },
+    {
+      "acc": 0.68804789,
+      "epoch": 1.193683409436834,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.823262085675337e-06,
+      "loss": 1.39404716,
+      "memory(GiB)": 123.79,
+      "step": 47055,
+      "train_speed(iter/s)": 1.315332
+    },
+    {
+      "acc": 0.67473507,
+      "epoch": 1.1938102486047693,
+      "grad_norm": 3.25,
+      "learning_rate": 3.822242942390718e-06,
+      "loss": 1.4237402,
+      "memory(GiB)": 123.79,
+      "step": 47060,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.68365278,
+      "epoch": 1.1939370877727042,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.821223850909115e-06,
+      "loss": 1.3954339,
+      "memory(GiB)": 123.79,
+      "step": 47065,
+      "train_speed(iter/s)": 1.315363
+    },
+    {
+      "acc": 0.68141813,
+      "epoch": 1.1940639269406392,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.820204811275351e-06,
+      "loss": 1.41110382,
+      "memory(GiB)": 123.79,
+      "step": 47070,
+      "train_speed(iter/s)": 1.315378
+    },
+    {
+      "acc": 0.69604435,
+      "epoch": 1.1941907661085742,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.8191858235342446e-06,
+      "loss": 1.40834408,
+      "memory(GiB)": 123.79,
+      "step": 47075,
+      "train_speed(iter/s)": 1.315393
+    },
+    {
+      "acc": 0.676933,
+      "epoch": 1.1943176052765094,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.818166887730618e-06,
+      "loss": 1.42672796,
+      "memory(GiB)": 123.79,
+      "step": 47080,
+      "train_speed(iter/s)": 1.315403
+    },
+    {
+      "acc": 0.67716846,
+      "epoch": 1.1944444444444444,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.817148003909288e-06,
+      "loss": 1.47487631,
+      "memory(GiB)": 123.79,
+      "step": 47085,
+      "train_speed(iter/s)": 1.31542
+    },
+    {
+      "acc": 0.68170547,
+      "epoch": 1.1945712836123794,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.816129172115073e-06,
+      "loss": 1.3607707,
+      "memory(GiB)": 123.79,
+      "step": 47090,
+      "train_speed(iter/s)": 1.315437
+    },
+    {
+      "acc": 0.67571688,
+      "epoch": 1.1946981227803146,
+      "grad_norm": 3.375,
+      "learning_rate": 3.815110392392778e-06,
+      "loss": 1.41263161,
+      "memory(GiB)": 123.79,
+      "step": 47095,
+      "train_speed(iter/s)": 1.315453
+    },
+    {
+      "acc": 0.68398867,
+      "epoch": 1.1948249619482496,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.8140916647872204e-06,
+      "loss": 1.38151512,
+      "memory(GiB)": 123.79,
+      "step": 47100,
+      "train_speed(iter/s)": 1.31547
+    },
+    {
+      "acc": 0.67479076,
+      "epoch": 1.1949518011161846,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.813072989343205e-06,
+      "loss": 1.44116669,
+      "memory(GiB)": 123.79,
+      "step": 47105,
+      "train_speed(iter/s)": 1.315484
+    },
+    {
+      "acc": 0.69245458,
+      "epoch": 1.1950786402841198,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.812054366105541e-06,
+      "loss": 1.40434284,
+      "memory(GiB)": 123.79,
+      "step": 47110,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.69579844,
+      "epoch": 1.1952054794520548,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.8110357951190284e-06,
+      "loss": 1.29485149,
+      "memory(GiB)": 123.79,
+      "step": 47115,
+      "train_speed(iter/s)": 1.315517
+    },
+    {
+      "acc": 0.65101814,
+      "epoch": 1.1953323186199898,
+      "grad_norm": 2.875,
+      "learning_rate": 3.8100172764284694e-06,
+      "loss": 1.4805294,
+      "memory(GiB)": 123.79,
+      "step": 47120,
+      "train_speed(iter/s)": 1.315533
+    },
+    {
+      "acc": 0.7004693,
+      "epoch": 1.195459157787925,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.8089988100786635e-06,
+      "loss": 1.36900215,
+      "memory(GiB)": 123.79,
+      "step": 47125,
+      "train_speed(iter/s)": 1.315549
+    },
+    {
+      "acc": 0.68246679,
+      "epoch": 1.19558599695586,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.807980396114409e-06,
+      "loss": 1.34576483,
+      "memory(GiB)": 123.79,
+      "step": 47130,
+      "train_speed(iter/s)": 1.315566
+    },
+    {
+      "acc": 0.68249049,
+      "epoch": 1.195712836123795,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.8069620345804974e-06,
+      "loss": 1.37773075,
+      "memory(GiB)": 123.79,
+      "step": 47135,
+      "train_speed(iter/s)": 1.315582
+    },
+    {
+      "acc": 0.68060131,
+      "epoch": 1.19583967529173,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.8059437255217214e-06,
+      "loss": 1.4526413,
+      "memory(GiB)": 123.79,
+      "step": 47140,
+      "train_speed(iter/s)": 1.315599
+    },
+    {
+      "acc": 0.69666085,
+      "epoch": 1.1959665144596652,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.8049254689828723e-06,
+      "loss": 1.36011114,
+      "memory(GiB)": 123.79,
+      "step": 47145,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.68587399,
+      "epoch": 1.1960933536276002,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.8039072650087377e-06,
+      "loss": 1.38161058,
+      "memory(GiB)": 123.79,
+      "step": 47150,
+      "train_speed(iter/s)": 1.315633
+    },
+    {
+      "acc": 0.68398695,
+      "epoch": 1.1962201927955354,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.8028891136440994e-06,
+      "loss": 1.35231304,
+      "memory(GiB)": 123.79,
+      "step": 47155,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.6809021,
+      "epoch": 1.1963470319634704,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.801871014933744e-06,
+      "loss": 1.37239246,
+      "memory(GiB)": 123.79,
+      "step": 47160,
+      "train_speed(iter/s)": 1.315668
+    },
+    {
+      "acc": 0.6816515,
+      "epoch": 1.1964738711314054,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.8008529689224493e-06,
+      "loss": 1.39733696,
+      "memory(GiB)": 123.79,
+      "step": 47165,
+      "train_speed(iter/s)": 1.315685
+    },
+    {
+      "acc": 0.68210244,
+      "epoch": 1.1966007102993403,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.7998349756549974e-06,
+      "loss": 1.39027348,
+      "memory(GiB)": 123.79,
+      "step": 47170,
+      "train_speed(iter/s)": 1.315703
+    },
+    {
+      "acc": 0.66712747,
+      "epoch": 1.1967275494672756,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.79881703517616e-06,
+      "loss": 1.53939991,
+      "memory(GiB)": 123.79,
+      "step": 47175,
+      "train_speed(iter/s)": 1.31572
+    },
+    {
+      "acc": 0.69128332,
+      "epoch": 1.1968543886352105,
+      "grad_norm": 3.75,
+      "learning_rate": 3.797799147530713e-06,
+      "loss": 1.41155043,
+      "memory(GiB)": 123.79,
+      "step": 47180,
+      "train_speed(iter/s)": 1.315736
+    },
+    {
+      "acc": 0.68504009,
+      "epoch": 1.1969812278031455,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.796781312763425e-06,
+      "loss": 1.38062487,
+      "memory(GiB)": 123.79,
+      "step": 47185,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.68970695,
+      "epoch": 1.1971080669710807,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.79576353091907e-06,
+      "loss": 1.30252628,
+      "memory(GiB)": 123.79,
+      "step": 47190,
+      "train_speed(iter/s)": 1.315771
+    },
+    {
+      "acc": 0.70286713,
+      "epoch": 1.1972349061390157,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.7947458020424094e-06,
+      "loss": 1.37550888,
+      "memory(GiB)": 123.79,
+      "step": 47195,
+      "train_speed(iter/s)": 1.315788
+    },
+    {
+      "acc": 0.68862305,
+      "epoch": 1.1973617453069507,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.793728126178209e-06,
+      "loss": 1.40170555,
+      "memory(GiB)": 123.79,
+      "step": 47200,
+      "train_speed(iter/s)": 1.315805
+    },
+    {
+      "acc": 0.69008288,
+      "epoch": 1.197488584474886,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.792710503371232e-06,
+      "loss": 1.30398283,
+      "memory(GiB)": 123.79,
+      "step": 47205,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.67904649,
+      "epoch": 1.197615423642821,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.7916929336662386e-06,
+      "loss": 1.48701839,
+      "memory(GiB)": 123.79,
+      "step": 47210,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.69200549,
+      "epoch": 1.197742262810756,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.790675417107982e-06,
+      "loss": 1.35443497,
+      "memory(GiB)": 123.79,
+      "step": 47215,
+      "train_speed(iter/s)": 1.315855
+    },
+    {
+      "acc": 0.69027004,
+      "epoch": 1.1978691019786911,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.7896579537412213e-06,
+      "loss": 1.35362816,
+      "memory(GiB)": 123.79,
+      "step": 47220,
+      "train_speed(iter/s)": 1.315873
+    },
+    {
+      "acc": 0.66468811,
+      "epoch": 1.197995941146626,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.7886405436107076e-06,
+      "loss": 1.48134623,
+      "memory(GiB)": 123.79,
+      "step": 47225,
+      "train_speed(iter/s)": 1.31589
+    },
+    {
+      "acc": 0.68096118,
+      "epoch": 1.198122780314561,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.7876231867611917e-06,
+      "loss": 1.38240366,
+      "memory(GiB)": 123.79,
+      "step": 47230,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.6788435,
+      "epoch": 1.198249619482496,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.7866058832374197e-06,
+      "loss": 1.43434477,
+      "memory(GiB)": 123.79,
+      "step": 47235,
+      "train_speed(iter/s)": 1.315924
+    },
+    {
+      "acc": 0.68780222,
+      "epoch": 1.1983764586504313,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.7855886330841383e-06,
+      "loss": 1.43762436,
+      "memory(GiB)": 123.79,
+      "step": 47240,
+      "train_speed(iter/s)": 1.315941
+    },
+    {
+      "acc": 0.68719659,
+      "epoch": 1.1985032978183663,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.7845714363460908e-06,
+      "loss": 1.39664373,
+      "memory(GiB)": 123.79,
+      "step": 47245,
+      "train_speed(iter/s)": 1.315958
+    },
+    {
+      "acc": 0.69053478,
+      "epoch": 1.1986301369863013,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.78355429306802e-06,
+      "loss": 1.40081081,
+      "memory(GiB)": 123.79,
+      "step": 47250,
+      "train_speed(iter/s)": 1.315974
+    },
+    {
+      "acc": 0.70031009,
+      "epoch": 1.1987569761542365,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.7825372032946605e-06,
+      "loss": 1.33680143,
+      "memory(GiB)": 123.79,
+      "step": 47255,
+      "train_speed(iter/s)": 1.315816
+    },
+    {
+      "acc": 0.69741964,
+      "epoch": 1.1988838153221715,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.7815201670707502e-06,
+      "loss": 1.37175446,
+      "memory(GiB)": 123.79,
+      "step": 47260,
+      "train_speed(iter/s)": 1.315832
+    },
+    {
+      "acc": 0.68196521,
+      "epoch": 1.1990106544901065,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.7805031844410235e-06,
+      "loss": 1.46498842,
+      "memory(GiB)": 123.79,
+      "step": 47265,
+      "train_speed(iter/s)": 1.315849
+    },
+    {
+      "acc": 0.69391022,
+      "epoch": 1.1991374936580417,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.7794862554502126e-06,
+      "loss": 1.34095116,
+      "memory(GiB)": 123.79,
+      "step": 47270,
+      "train_speed(iter/s)": 1.315866
+    },
+    {
+      "acc": 0.68666973,
+      "epoch": 1.1992643328259767,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.778469380143045e-06,
+      "loss": 1.37411633,
+      "memory(GiB)": 123.79,
+      "step": 47275,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.69654546,
+      "epoch": 1.1993911719939117,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.777452558564246e-06,
+      "loss": 1.35450649,
+      "memory(GiB)": 123.79,
+      "step": 47280,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.69542885,
+      "epoch": 1.1995180111618469,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.776435790758543e-06,
+      "loss": 1.33428555,
+      "memory(GiB)": 123.79,
+      "step": 47285,
+      "train_speed(iter/s)": 1.315917
+    },
+    {
+      "acc": 0.69420252,
+      "epoch": 1.1996448503297819,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.7754190767706577e-06,
+      "loss": 1.29079905,
+      "memory(GiB)": 123.79,
+      "step": 47290,
+      "train_speed(iter/s)": 1.315933
+    },
+    {
+      "acc": 0.68233833,
+      "epoch": 1.1997716894977168,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.774402416645307e-06,
+      "loss": 1.41129742,
+      "memory(GiB)": 123.79,
+      "step": 47295,
+      "train_speed(iter/s)": 1.315949
+    },
+    {
+      "acc": 0.68157101,
+      "epoch": 1.1998985286656518,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.77338581042721e-06,
+      "loss": 1.39553013,
+      "memory(GiB)": 123.79,
+      "step": 47300,
+      "train_speed(iter/s)": 1.315966
+    },
+    {
+      "acc": 0.69736481,
+      "epoch": 1.200025367833587,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.7723692581610817e-06,
+      "loss": 1.41804628,
+      "memory(GiB)": 123.79,
+      "step": 47305,
+      "train_speed(iter/s)": 1.315983
+    },
+    {
+      "acc": 0.69314933,
+      "epoch": 1.200152207001522,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.771352759891637e-06,
+      "loss": 1.46361132,
+      "memory(GiB)": 123.79,
+      "step": 47310,
+      "train_speed(iter/s)": 1.315999
+    },
+    {
+      "acc": 0.68415785,
+      "epoch": 1.2002790461694572,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.7703363156635807e-06,
+      "loss": 1.45798855,
+      "memory(GiB)": 123.79,
+      "step": 47315,
+      "train_speed(iter/s)": 1.316016
+    },
+    {
+      "acc": 0.68936958,
+      "epoch": 1.2004058853373922,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.769319925521624e-06,
+      "loss": 1.40314007,
+      "memory(GiB)": 123.79,
+      "step": 47320,
+      "train_speed(iter/s)": 1.316033
+    },
+    {
+      "acc": 0.70114393,
+      "epoch": 1.2005327245053272,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.76830358951047e-06,
+      "loss": 1.37243452,
+      "memory(GiB)": 123.79,
+      "step": 47325,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.67994385,
+      "epoch": 1.2006595636732622,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.767287307674826e-06,
+      "loss": 1.34424419,
+      "memory(GiB)": 123.79,
+      "step": 47330,
+      "train_speed(iter/s)": 1.316066
+    },
+    {
+      "acc": 0.69034357,
+      "epoch": 1.2007864028411974,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.766271080059389e-06,
+      "loss": 1.37329807,
+      "memory(GiB)": 123.79,
+      "step": 47335,
+      "train_speed(iter/s)": 1.316082
+    },
+    {
+      "acc": 0.69524565,
+      "epoch": 1.2009132420091324,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.7652549067088568e-06,
+      "loss": 1.34921551,
+      "memory(GiB)": 123.79,
+      "step": 47340,
+      "train_speed(iter/s)": 1.316097
+    },
+    {
+      "acc": 0.69535151,
+      "epoch": 1.2010400811770674,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.7642387876679275e-06,
+      "loss": 1.39548321,
+      "memory(GiB)": 123.79,
+      "step": 47345,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.68172607,
+      "epoch": 1.2011669203450026,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.7632227229812947e-06,
+      "loss": 1.39823837,
+      "memory(GiB)": 123.79,
+      "step": 47350,
+      "train_speed(iter/s)": 1.316129
+    },
+    {
+      "acc": 0.68232775,
+      "epoch": 1.2012937595129376,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.7622067126936475e-06,
+      "loss": 1.42358313,
+      "memory(GiB)": 123.79,
+      "step": 47355,
+      "train_speed(iter/s)": 1.316145
+    },
+    {
+      "acc": 0.69257059,
+      "epoch": 1.2014205986808726,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.761190756849674e-06,
+      "loss": 1.33240299,
+      "memory(GiB)": 123.79,
+      "step": 47360,
+      "train_speed(iter/s)": 1.316161
+    },
+    {
+      "acc": 0.68693008,
+      "epoch": 1.2015474378488078,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.7601748554940633e-06,
+      "loss": 1.40203791,
+      "memory(GiB)": 123.79,
+      "step": 47365,
+      "train_speed(iter/s)": 1.316178
+    },
+    {
+      "acc": 0.68698964,
+      "epoch": 1.2016742770167428,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.7591590086714984e-06,
+      "loss": 1.32557468,
+      "memory(GiB)": 123.79,
+      "step": 47370,
+      "train_speed(iter/s)": 1.316194
+    },
+    {
+      "acc": 0.67961588,
+      "epoch": 1.2018011161846778,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.7581432164266587e-06,
+      "loss": 1.35970535,
+      "memory(GiB)": 123.79,
+      "step": 47375,
+      "train_speed(iter/s)": 1.316209
+    },
+    {
+      "acc": 0.68733597,
+      "epoch": 1.201927955352613,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.7571274788042255e-06,
+      "loss": 1.38626871,
+      "memory(GiB)": 123.79,
+      "step": 47380,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.68030734,
+      "epoch": 1.202054794520548,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.756111795848874e-06,
+      "loss": 1.415909,
+      "memory(GiB)": 123.79,
+      "step": 47385,
+      "train_speed(iter/s)": 1.316243
+    },
+    {
+      "acc": 0.68833361,
+      "epoch": 1.202181633688483,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.755096167605281e-06,
+      "loss": 1.43931866,
+      "memory(GiB)": 123.79,
+      "step": 47390,
+      "train_speed(iter/s)": 1.316259
+    },
+    {
+      "acc": 0.69432564,
+      "epoch": 1.202308472856418,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.7540805941181165e-06,
+      "loss": 1.34429588,
+      "memory(GiB)": 123.79,
+      "step": 47395,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.68730869,
+      "epoch": 1.2024353120243532,
+      "grad_norm": 2.875,
+      "learning_rate": 3.7530650754320492e-06,
+      "loss": 1.45030737,
+      "memory(GiB)": 123.79,
+      "step": 47400,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.70553913,
+      "epoch": 1.2025621511922882,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.752049611591746e-06,
+      "loss": 1.30255909,
+      "memory(GiB)": 123.79,
+      "step": 47405,
+      "train_speed(iter/s)": 1.31631
+    },
+    {
+      "acc": 0.68752508,
+      "epoch": 1.2026889903602231,
+      "grad_norm": 3.125,
+      "learning_rate": 3.7510342026418756e-06,
+      "loss": 1.41732635,
+      "memory(GiB)": 123.79,
+      "step": 47410,
+      "train_speed(iter/s)": 1.316327
+    },
+    {
+      "acc": 0.68342004,
+      "epoch": 1.2028158295281584,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.7500188486270948e-06,
+      "loss": 1.37456598,
+      "memory(GiB)": 123.79,
+      "step": 47415,
+      "train_speed(iter/s)": 1.316343
+    },
+    {
+      "acc": 0.67312336,
+      "epoch": 1.2029426686960933,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.7490035495920664e-06,
+      "loss": 1.46519032,
+      "memory(GiB)": 123.79,
+      "step": 47420,
+      "train_speed(iter/s)": 1.31636
+    },
+    {
+      "acc": 0.69665251,
+      "epoch": 1.2030695078640283,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.747988305581447e-06,
+      "loss": 1.32705002,
+      "memory(GiB)": 123.79,
+      "step": 47425,
+      "train_speed(iter/s)": 1.316376
+    },
+    {
+      "acc": 0.69082909,
+      "epoch": 1.2031963470319635,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.7469731166398933e-06,
+      "loss": 1.40223999,
+      "memory(GiB)": 123.79,
+      "step": 47430,
+      "train_speed(iter/s)": 1.316393
+    },
+    {
+      "acc": 0.67911673,
+      "epoch": 1.2033231861998985,
+      "grad_norm": 3.0,
+      "learning_rate": 3.745957982812054e-06,
+      "loss": 1.43095922,
+      "memory(GiB)": 123.79,
+      "step": 47435,
+      "train_speed(iter/s)": 1.316409
+    },
+    {
+      "acc": 0.67972155,
+      "epoch": 1.2034500253678335,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.744942904142582e-06,
+      "loss": 1.40173941,
+      "memory(GiB)": 123.79,
+      "step": 47440,
+      "train_speed(iter/s)": 1.316426
+    },
+    {
+      "acc": 0.70277781,
+      "epoch": 1.2035768645357687,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.743927880676125e-06,
+      "loss": 1.32712231,
+      "memory(GiB)": 123.79,
+      "step": 47445,
+      "train_speed(iter/s)": 1.316442
+    },
+    {
+      "acc": 0.69517813,
+      "epoch": 1.2037037037037037,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.742912912457329e-06,
+      "loss": 1.32609215,
+      "memory(GiB)": 123.79,
+      "step": 47450,
+      "train_speed(iter/s)": 1.316457
+    },
+    {
+      "acc": 0.68184967,
+      "epoch": 1.2038305428716387,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.7418979995308336e-06,
+      "loss": 1.38403339,
+      "memory(GiB)": 123.79,
+      "step": 47455,
+      "train_speed(iter/s)": 1.316474
+    },
+    {
+      "acc": 0.68577294,
+      "epoch": 1.2039573820395737,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.740883141941282e-06,
+      "loss": 1.341819,
+      "memory(GiB)": 123.79,
+      "step": 47460,
+      "train_speed(iter/s)": 1.31649
+    },
+    {
+      "acc": 0.69348516,
+      "epoch": 1.204084221207509,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.7398683397333103e-06,
+      "loss": 1.40005589,
+      "memory(GiB)": 123.79,
+      "step": 47465,
+      "train_speed(iter/s)": 1.316507
+    },
+    {
+      "acc": 0.68947082,
+      "epoch": 1.204211060375444,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.7388535929515573e-06,
+      "loss": 1.35264053,
+      "memory(GiB)": 123.79,
+      "step": 47470,
+      "train_speed(iter/s)": 1.316524
+    },
+    {
+      "acc": 0.70040731,
+      "epoch": 1.204337899543379,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.737838901640653e-06,
+      "loss": 1.31586533,
+      "memory(GiB)": 123.79,
+      "step": 47475,
+      "train_speed(iter/s)": 1.31654
+    },
+    {
+      "acc": 0.68619766,
+      "epoch": 1.204464738711314,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.736824265845228e-06,
+      "loss": 1.37791796,
+      "memory(GiB)": 123.79,
+      "step": 47480,
+      "train_speed(iter/s)": 1.316556
+    },
+    {
+      "acc": 0.69650393,
+      "epoch": 1.204591577879249,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.7358096856099118e-06,
+      "loss": 1.3682663,
+      "memory(GiB)": 123.79,
+      "step": 47485,
+      "train_speed(iter/s)": 1.316573
+    },
+    {
+      "acc": 0.68393307,
+      "epoch": 1.204718417047184,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.7347951609793315e-06,
+      "loss": 1.35722342,
+      "memory(GiB)": 123.79,
+      "step": 47490,
+      "train_speed(iter/s)": 1.316589
+    },
+    {
+      "acc": 0.67662296,
+      "epoch": 1.2048452562151193,
+      "grad_norm": 3.375,
+      "learning_rate": 3.7337806919981077e-06,
+      "loss": 1.39755106,
+      "memory(GiB)": 123.79,
+      "step": 47495,
+      "train_speed(iter/s)": 1.316604
+    },
+    {
+      "acc": 0.68899784,
+      "epoch": 1.2049720953830543,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.732766278710861e-06,
+      "loss": 1.35088663,
+      "memory(GiB)": 123.79,
+      "step": 47500,
+      "train_speed(iter/s)": 1.31662
+    },
+    {
+      "acc": 0.68931932,
+      "epoch": 1.2050989345509893,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.7317519211622123e-06,
+      "loss": 1.37022038,
+      "memory(GiB)": 123.79,
+      "step": 47505,
+      "train_speed(iter/s)": 1.316635
+    },
+    {
+      "acc": 0.68426485,
+      "epoch": 1.2052257737189245,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.7307376193967772e-06,
+      "loss": 1.42370462,
+      "memory(GiB)": 123.79,
+      "step": 47510,
+      "train_speed(iter/s)": 1.316651
+    },
+    {
+      "acc": 0.68699932,
+      "epoch": 1.2053526128868595,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.7297233734591664e-06,
+      "loss": 1.37633324,
+      "memory(GiB)": 123.79,
+      "step": 47515,
+      "train_speed(iter/s)": 1.316666
+    },
+    {
+      "acc": 0.68282409,
+      "epoch": 1.2054794520547945,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.7287091833939948e-06,
+      "loss": 1.43018532,
+      "memory(GiB)": 123.79,
+      "step": 47520,
+      "train_speed(iter/s)": 1.316682
+    },
+    {
+      "acc": 0.70161619,
+      "epoch": 1.2056062912227297,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.7276950492458675e-06,
+      "loss": 1.31400747,
+      "memory(GiB)": 123.79,
+      "step": 47525,
+      "train_speed(iter/s)": 1.316699
+    },
+    {
+      "acc": 0.68629332,
+      "epoch": 1.2057331303906647,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.7266809710593956e-06,
+      "loss": 1.39027872,
+      "memory(GiB)": 123.79,
+      "step": 47530,
+      "train_speed(iter/s)": 1.316715
+    },
+    {
+      "acc": 0.68605232,
+      "epoch": 1.2058599695585996,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.7256669488791763e-06,
+      "loss": 1.35467863,
+      "memory(GiB)": 123.79,
+      "step": 47535,
+      "train_speed(iter/s)": 1.316732
+    },
+    {
+      "acc": 0.69284878,
+      "epoch": 1.2059868087265349,
+      "grad_norm": 3.0,
+      "learning_rate": 3.7246529827498156e-06,
+      "loss": 1.29874859,
+      "memory(GiB)": 123.79,
+      "step": 47540,
+      "train_speed(iter/s)": 1.316748
+    },
+    {
+      "acc": 0.68880839,
+      "epoch": 1.2061136478944698,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.7236390727159094e-06,
+      "loss": 1.38092899,
+      "memory(GiB)": 123.79,
+      "step": 47545,
+      "train_speed(iter/s)": 1.316763
+    },
+    {
+      "acc": 0.70263624,
+      "epoch": 1.2062404870624048,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.7226252188220573e-06,
+      "loss": 1.32123966,
+      "memory(GiB)": 123.79,
+      "step": 47550,
+      "train_speed(iter/s)": 1.31678
+    },
+    {
+      "acc": 0.68473845,
+      "epoch": 1.2063673262303398,
+      "grad_norm": 3.5,
+      "learning_rate": 3.7216114211128505e-06,
+      "loss": 1.38485565,
+      "memory(GiB)": 123.79,
+      "step": 47555,
+      "train_speed(iter/s)": 1.316797
+    },
+    {
+      "acc": 0.69433718,
+      "epoch": 1.206494165398275,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.720597679632879e-06,
+      "loss": 1.35598822,
+      "memory(GiB)": 123.79,
+      "step": 47560,
+      "train_speed(iter/s)": 1.316813
+    },
+    {
+      "acc": 0.6825531,
+      "epoch": 1.20662100456621,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.7195839944267357e-06,
+      "loss": 1.46842051,
+      "memory(GiB)": 123.79,
+      "step": 47565,
+      "train_speed(iter/s)": 1.31683
+    },
+    {
+      "acc": 0.70370245,
+      "epoch": 1.206747843734145,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.718570365539006e-06,
+      "loss": 1.30391941,
+      "memory(GiB)": 123.79,
+      "step": 47570,
+      "train_speed(iter/s)": 1.316846
+    },
+    {
+      "acc": 0.69335709,
+      "epoch": 1.2068746829020802,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.717556793014271e-06,
+      "loss": 1.3881547,
+      "memory(GiB)": 123.79,
+      "step": 47575,
+      "train_speed(iter/s)": 1.316862
+    },
+    {
+      "acc": 0.68512459,
+      "epoch": 1.2070015220700152,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.716543276897113e-06,
+      "loss": 1.44762983,
+      "memory(GiB)": 123.79,
+      "step": 47580,
+      "train_speed(iter/s)": 1.316879
+    },
+    {
+      "acc": 0.67872005,
+      "epoch": 1.2071283612379502,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.715529817232114e-06,
+      "loss": 1.36858902,
+      "memory(GiB)": 123.79,
+      "step": 47585,
+      "train_speed(iter/s)": 1.316895
+    },
+    {
+      "acc": 0.68635817,
+      "epoch": 1.2072552004058854,
+      "grad_norm": 3.75,
+      "learning_rate": 3.7145164140638483e-06,
+      "loss": 1.41550179,
+      "memory(GiB)": 123.79,
+      "step": 47590,
+      "train_speed(iter/s)": 1.316739
+    },
+    {
+      "acc": 0.6856451,
+      "epoch": 1.2073820395738204,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.713503067436889e-06,
+      "loss": 1.40874939,
+      "memory(GiB)": 123.79,
+      "step": 47595,
+      "train_speed(iter/s)": 1.316755
+    },
+    {
+      "acc": 0.69128795,
+      "epoch": 1.2075088787417554,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.7124897773958084e-06,
+      "loss": 1.35197449,
+      "memory(GiB)": 123.79,
+      "step": 47600,
+      "train_speed(iter/s)": 1.316772
+    },
+    {
+      "acc": 0.68742151,
+      "epoch": 1.2076357179096906,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.7114765439851752e-06,
+      "loss": 1.47605705,
+      "memory(GiB)": 123.79,
+      "step": 47605,
+      "train_speed(iter/s)": 1.316789
+    },
+    {
+      "acc": 0.69079084,
+      "epoch": 1.2077625570776256,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.7104633672495584e-06,
+      "loss": 1.34183807,
+      "memory(GiB)": 123.79,
+      "step": 47610,
+      "train_speed(iter/s)": 1.316805
+    },
+    {
+      "acc": 0.70254602,
+      "epoch": 1.2078893962455606,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.709450247233519e-06,
+      "loss": 1.30172749,
+      "memory(GiB)": 123.79,
+      "step": 47615,
+      "train_speed(iter/s)": 1.316822
+    },
+    {
+      "acc": 0.66807075,
+      "epoch": 1.2080162354134956,
+      "grad_norm": 5.46875,
+      "learning_rate": 3.7084371839816204e-06,
+      "loss": 1.45520058,
+      "memory(GiB)": 123.79,
+      "step": 47620,
+      "train_speed(iter/s)": 1.316838
+    },
+    {
+      "acc": 0.70202579,
+      "epoch": 1.2081430745814308,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.707424177538419e-06,
+      "loss": 1.33953819,
+      "memory(GiB)": 123.79,
+      "step": 47625,
+      "train_speed(iter/s)": 1.316854
+    },
+    {
+      "acc": 0.6915381,
+      "epoch": 1.2082699137493658,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.7064112279484753e-06,
+      "loss": 1.35487652,
+      "memory(GiB)": 123.79,
+      "step": 47630,
+      "train_speed(iter/s)": 1.31687
+    },
+    {
+      "acc": 0.70049319,
+      "epoch": 1.208396752917301,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.7053983352563407e-06,
+      "loss": 1.37490807,
+      "memory(GiB)": 123.79,
+      "step": 47635,
+      "train_speed(iter/s)": 1.316886
+    },
+    {
+      "acc": 0.69003305,
+      "epoch": 1.208523592085236,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.704385499506565e-06,
+      "loss": 1.33898201,
+      "memory(GiB)": 123.79,
+      "step": 47640,
+      "train_speed(iter/s)": 1.316901
+    },
+    {
+      "acc": 0.68164306,
+      "epoch": 1.208650431253171,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.703372720743702e-06,
+      "loss": 1.40215778,
+      "memory(GiB)": 123.79,
+      "step": 47645,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.68881888,
+      "epoch": 1.208777270421106,
+      "grad_norm": 3.375,
+      "learning_rate": 3.7023599990122966e-06,
+      "loss": 1.3709959,
+      "memory(GiB)": 123.79,
+      "step": 47650,
+      "train_speed(iter/s)": 1.316931
+    },
+    {
+      "acc": 0.68063641,
+      "epoch": 1.2089041095890412,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.7013473343568897e-06,
+      "loss": 1.40966969,
+      "memory(GiB)": 123.79,
+      "step": 47655,
+      "train_speed(iter/s)": 1.316946
+    },
+    {
+      "acc": 0.69405813,
+      "epoch": 1.2090309487569761,
+      "grad_norm": 3.375,
+      "learning_rate": 3.700334726822026e-06,
+      "loss": 1.35717964,
+      "memory(GiB)": 123.79,
+      "step": 47660,
+      "train_speed(iter/s)": 1.31696
+    },
+    {
+      "acc": 0.7016305,
+      "epoch": 1.2091577879249111,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.6993221764522435e-06,
+      "loss": 1.30642538,
+      "memory(GiB)": 123.79,
+      "step": 47665,
+      "train_speed(iter/s)": 1.316975
+    },
+    {
+      "acc": 0.68871837,
+      "epoch": 1.2092846270928463,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.6983096832920806e-06,
+      "loss": 1.3883667,
+      "memory(GiB)": 123.79,
+      "step": 47670,
+      "train_speed(iter/s)": 1.316988
+    },
+    {
+      "acc": 0.69693232,
+      "epoch": 1.2094114662607813,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.697297247386066e-06,
+      "loss": 1.39302006,
+      "memory(GiB)": 123.79,
+      "step": 47675,
+      "train_speed(iter/s)": 1.317003
+    },
+    {
+      "acc": 0.68774104,
+      "epoch": 1.2095383054287163,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.6962848687787365e-06,
+      "loss": 1.47111244,
+      "memory(GiB)": 123.79,
+      "step": 47680,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.68781071,
+      "epoch": 1.2096651445966515,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.6952725475146183e-06,
+      "loss": 1.35147915,
+      "memory(GiB)": 123.79,
+      "step": 47685,
+      "train_speed(iter/s)": 1.317032
+    },
+    {
+      "acc": 0.69202671,
+      "epoch": 1.2097919837645865,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.69426028363824e-06,
+      "loss": 1.35338516,
+      "memory(GiB)": 123.79,
+      "step": 47690,
+      "train_speed(iter/s)": 1.317046
+    },
+    {
+      "acc": 0.69900589,
+      "epoch": 1.2099188229325215,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.6932480771941237e-06,
+      "loss": 1.34180098,
+      "memory(GiB)": 123.79,
+      "step": 47695,
+      "train_speed(iter/s)": 1.317061
+    },
+    {
+      "acc": 0.6893239,
+      "epoch": 1.2100456621004567,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.6922359282267904e-06,
+      "loss": 1.35596123,
+      "memory(GiB)": 123.79,
+      "step": 47700,
+      "train_speed(iter/s)": 1.317075
+    },
+    {
+      "acc": 0.69062881,
+      "epoch": 1.2101725012683917,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.6912238367807606e-06,
+      "loss": 1.37966242,
+      "memory(GiB)": 123.79,
+      "step": 47705,
+      "train_speed(iter/s)": 1.31709
+    },
+    {
+      "acc": 0.67905817,
+      "epoch": 1.2102993404363267,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.6902118029005507e-06,
+      "loss": 1.43434629,
+      "memory(GiB)": 123.79,
+      "step": 47710,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.69146533,
+      "epoch": 1.2104261796042617,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.6891998266306717e-06,
+      "loss": 1.41324606,
+      "memory(GiB)": 123.79,
+      "step": 47715,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.68963156,
+      "epoch": 1.210553018772197,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.688187908015636e-06,
+      "loss": 1.37166004,
+      "memory(GiB)": 123.79,
+      "step": 47720,
+      "train_speed(iter/s)": 1.316949
+    },
+    {
+      "acc": 0.69608822,
+      "epoch": 1.2106798579401319,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.6871760470999546e-06,
+      "loss": 1.39814129,
+      "memory(GiB)": 123.79,
+      "step": 47725,
+      "train_speed(iter/s)": 1.316963
+    },
+    {
+      "acc": 0.68548985,
+      "epoch": 1.2108066971080669,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.6861642439281325e-06,
+      "loss": 1.35446472,
+      "memory(GiB)": 123.79,
+      "step": 47730,
+      "train_speed(iter/s)": 1.316978
+    },
+    {
+      "acc": 0.68055048,
+      "epoch": 1.210933536276002,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.6851524985446707e-06,
+      "loss": 1.44110079,
+      "memory(GiB)": 123.79,
+      "step": 47735,
+      "train_speed(iter/s)": 1.316992
+    },
+    {
+      "acc": 0.68506551,
+      "epoch": 1.211060375443937,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.6841408109940737e-06,
+      "loss": 1.36590958,
+      "memory(GiB)": 123.79,
+      "step": 47740,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.68764138,
+      "epoch": 1.211187214611872,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.6831291813208377e-06,
+      "loss": 1.40292082,
+      "memory(GiB)": 123.79,
+      "step": 47745,
+      "train_speed(iter/s)": 1.31702
+    },
+    {
+      "acc": 0.67074461,
+      "epoch": 1.2113140537798073,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.682117609569462e-06,
+      "loss": 1.46697998,
+      "memory(GiB)": 123.79,
+      "step": 47750,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.67342482,
+      "epoch": 1.2114408929477423,
+      "grad_norm": 2.75,
+      "learning_rate": 3.681106095784436e-06,
+      "loss": 1.4422699,
+      "memory(GiB)": 123.79,
+      "step": 47755,
+      "train_speed(iter/s)": 1.317049
+    },
+    {
+      "acc": 0.71161451,
+      "epoch": 1.2115677321156773,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.6800946400102522e-06,
+      "loss": 1.27904902,
+      "memory(GiB)": 123.79,
+      "step": 47760,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.69257526,
+      "epoch": 1.2116945712836125,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.6790832422913984e-06,
+      "loss": 1.39704208,
+      "memory(GiB)": 123.79,
+      "step": 47765,
+      "train_speed(iter/s)": 1.317078
+    },
+    {
+      "acc": 0.69933553,
+      "epoch": 1.2118214104515475,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.6780719026723632e-06,
+      "loss": 1.37889423,
+      "memory(GiB)": 123.79,
+      "step": 47770,
+      "train_speed(iter/s)": 1.317093
+    },
+    {
+      "acc": 0.6722259,
+      "epoch": 1.2119482496194824,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.677060621197627e-06,
+      "loss": 1.48362837,
+      "memory(GiB)": 123.79,
+      "step": 47775,
+      "train_speed(iter/s)": 1.317109
+    },
+    {
+      "acc": 0.67947559,
+      "epoch": 1.2120750887874174,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.6760493979116696e-06,
+      "loss": 1.40217628,
+      "memory(GiB)": 123.79,
+      "step": 47780,
+      "train_speed(iter/s)": 1.317122
+    },
+    {
+      "acc": 0.69738717,
+      "epoch": 1.2122019279553526,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.6750382328589725e-06,
+      "loss": 1.37983112,
+      "memory(GiB)": 123.79,
+      "step": 47785,
+      "train_speed(iter/s)": 1.317136
+    },
+    {
+      "acc": 0.67330327,
+      "epoch": 1.2123287671232876,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.67402712608401e-06,
+      "loss": 1.43579874,
+      "memory(GiB)": 123.79,
+      "step": 47790,
+      "train_speed(iter/s)": 1.317151
+    },
+    {
+      "acc": 0.68917975,
+      "epoch": 1.2124556062912228,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.673016077631253e-06,
+      "loss": 1.4186552,
+      "memory(GiB)": 123.79,
+      "step": 47795,
+      "train_speed(iter/s)": 1.317166
+    },
+    {
+      "acc": 0.67808018,
+      "epoch": 1.2125824454591578,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.672005087545173e-06,
+      "loss": 1.40485687,
+      "memory(GiB)": 123.79,
+      "step": 47800,
+      "train_speed(iter/s)": 1.31718
+    },
+    {
+      "acc": 0.69169016,
+      "epoch": 1.2127092846270928,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.6709941558702393e-06,
+      "loss": 1.39416895,
+      "memory(GiB)": 123.79,
+      "step": 47805,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.68399038,
+      "epoch": 1.2128361237950278,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.6699832826509174e-06,
+      "loss": 1.43679705,
+      "memory(GiB)": 123.79,
+      "step": 47810,
+      "train_speed(iter/s)": 1.317208
+    },
+    {
+      "acc": 0.67450571,
+      "epoch": 1.212962962962963,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.6689724679316665e-06,
+      "loss": 1.46796131,
+      "memory(GiB)": 123.79,
+      "step": 47815,
+      "train_speed(iter/s)": 1.317222
+    },
+    {
+      "acc": 0.67701468,
+      "epoch": 1.213089802130898,
+      "grad_norm": 3.375,
+      "learning_rate": 3.66796171175695e-06,
+      "loss": 1.42726688,
+      "memory(GiB)": 123.79,
+      "step": 47820,
+      "train_speed(iter/s)": 1.317236
+    },
+    {
+      "acc": 0.69001088,
+      "epoch": 1.213216641298833,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.666951014171224e-06,
+      "loss": 1.38512955,
+      "memory(GiB)": 123.79,
+      "step": 47825,
+      "train_speed(iter/s)": 1.317251
+    },
+    {
+      "acc": 0.67955585,
+      "epoch": 1.2133434804667682,
+      "grad_norm": 2.75,
+      "learning_rate": 3.6659403752189453e-06,
+      "loss": 1.37370071,
+      "memory(GiB)": 123.79,
+      "step": 47830,
+      "train_speed(iter/s)": 1.317266
+    },
+    {
+      "acc": 0.68921213,
+      "epoch": 1.2134703196347032,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.664929794944565e-06,
+      "loss": 1.34929161,
+      "memory(GiB)": 123.79,
+      "step": 47835,
+      "train_speed(iter/s)": 1.31728
+    },
+    {
+      "acc": 0.69864473,
+      "epoch": 1.2135971588026382,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.663919273392532e-06,
+      "loss": 1.34605999,
+      "memory(GiB)": 123.79,
+      "step": 47840,
+      "train_speed(iter/s)": 1.317294
+    },
+    {
+      "acc": 0.70537729,
+      "epoch": 1.2137239979705734,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.662908810607294e-06,
+      "loss": 1.26703396,
+      "memory(GiB)": 123.79,
+      "step": 47845,
+      "train_speed(iter/s)": 1.317308
+    },
+    {
+      "acc": 0.68701835,
+      "epoch": 1.2138508371385084,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.6618984066332986e-06,
+      "loss": 1.43439283,
+      "memory(GiB)": 123.79,
+      "step": 47850,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.70064068,
+      "epoch": 1.2139776763064434,
+      "grad_norm": 3.75,
+      "learning_rate": 3.660888061514984e-06,
+      "loss": 1.3661871,
+      "memory(GiB)": 123.79,
+      "step": 47855,
+      "train_speed(iter/s)": 1.317336
+    },
+    {
+      "acc": 0.69279885,
+      "epoch": 1.2141045154743786,
+      "grad_norm": 3.625,
+      "learning_rate": 3.6598777752967896e-06,
+      "loss": 1.41982069,
+      "memory(GiB)": 123.79,
+      "step": 47860,
+      "train_speed(iter/s)": 1.317349
+    },
+    {
+      "acc": 0.69118085,
+      "epoch": 1.2142313546423136,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.658867548023156e-06,
+      "loss": 1.39560108,
+      "memory(GiB)": 123.79,
+      "step": 47865,
+      "train_speed(iter/s)": 1.317363
+    },
+    {
+      "acc": 0.69031029,
+      "epoch": 1.2143581938102486,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.657857379738515e-06,
+      "loss": 1.41943474,
+      "memory(GiB)": 123.79,
+      "step": 47870,
+      "train_speed(iter/s)": 1.317377
+    },
+    {
+      "acc": 0.68079853,
+      "epoch": 1.2144850329781836,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.656847270487298e-06,
+      "loss": 1.43467731,
+      "memory(GiB)": 123.79,
+      "step": 47875,
+      "train_speed(iter/s)": 1.317392
+    },
+    {
+      "acc": 0.6831327,
+      "epoch": 1.2146118721461188,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.655837220313936e-06,
+      "loss": 1.38601818,
+      "memory(GiB)": 123.79,
+      "step": 47880,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.67153807,
+      "epoch": 1.2147387113140538,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.654827229262852e-06,
+      "loss": 1.42481661,
+      "memory(GiB)": 123.79,
+      "step": 47885,
+      "train_speed(iter/s)": 1.317419
+    },
+    {
+      "acc": 0.70659771,
+      "epoch": 1.2148655504819887,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.653817297378476e-06,
+      "loss": 1.30689058,
+      "memory(GiB)": 123.79,
+      "step": 47890,
+      "train_speed(iter/s)": 1.317433
+    },
+    {
+      "acc": 0.7038435,
+      "epoch": 1.214992389649924,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.6528074247052225e-06,
+      "loss": 1.32323837,
+      "memory(GiB)": 123.79,
+      "step": 47895,
+      "train_speed(iter/s)": 1.317448
+    },
+    {
+      "acc": 0.69012537,
+      "epoch": 1.215119228817859,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.651797611287514e-06,
+      "loss": 1.32443647,
+      "memory(GiB)": 123.79,
+      "step": 47900,
+      "train_speed(iter/s)": 1.317461
+    },
+    {
+      "acc": 0.70672011,
+      "epoch": 1.215246067985794,
+      "grad_norm": 4.0,
+      "learning_rate": 3.6507878571697646e-06,
+      "loss": 1.34627066,
+      "memory(GiB)": 123.79,
+      "step": 47905,
+      "train_speed(iter/s)": 1.317475
+    },
+    {
+      "acc": 0.689674,
+      "epoch": 1.2153729071537291,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.6497781623963915e-06,
+      "loss": 1.4042408,
+      "memory(GiB)": 123.79,
+      "step": 47910,
+      "train_speed(iter/s)": 1.317489
+    },
+    {
+      "acc": 0.68503766,
+      "epoch": 1.2154997463216641,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.648768527011802e-06,
+      "loss": 1.40739756,
+      "memory(GiB)": 123.79,
+      "step": 47915,
+      "train_speed(iter/s)": 1.317502
+    },
+    {
+      "acc": 0.68634605,
+      "epoch": 1.2156265854895991,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.6477589510604044e-06,
+      "loss": 1.4152441,
+      "memory(GiB)": 123.79,
+      "step": 47920,
+      "train_speed(iter/s)": 1.317517
+    },
+    {
+      "acc": 0.68250942,
+      "epoch": 1.2157534246575343,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.646749434586607e-06,
+      "loss": 1.40162373,
+      "memory(GiB)": 123.79,
+      "step": 47925,
+      "train_speed(iter/s)": 1.317531
+    },
+    {
+      "acc": 0.70750914,
+      "epoch": 1.2158802638254693,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.645739977634811e-06,
+      "loss": 1.33274975,
+      "memory(GiB)": 123.79,
+      "step": 47930,
+      "train_speed(iter/s)": 1.317546
+    },
+    {
+      "acc": 0.69461718,
+      "epoch": 1.2160071029934043,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.6447305802494177e-06,
+      "loss": 1.362397,
+      "memory(GiB)": 123.79,
+      "step": 47935,
+      "train_speed(iter/s)": 1.317561
+    },
+    {
+      "acc": 0.69143882,
+      "epoch": 1.2161339421613393,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.6437212424748227e-06,
+      "loss": 1.42918024,
+      "memory(GiB)": 123.79,
+      "step": 47940,
+      "train_speed(iter/s)": 1.317574
+    },
+    {
+      "acc": 0.68263102,
+      "epoch": 1.2162607813292745,
+      "grad_norm": 3.875,
+      "learning_rate": 3.642711964355423e-06,
+      "loss": 1.41964417,
+      "memory(GiB)": 123.79,
+      "step": 47945,
+      "train_speed(iter/s)": 1.317587
+    },
+    {
+      "acc": 0.67496328,
+      "epoch": 1.2163876204972095,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.6417027459356134e-06,
+      "loss": 1.41706429,
+      "memory(GiB)": 123.79,
+      "step": 47950,
+      "train_speed(iter/s)": 1.3176
+    },
+    {
+      "acc": 0.66628571,
+      "epoch": 1.2165144596651447,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.640693587259778e-06,
+      "loss": 1.48682728,
+      "memory(GiB)": 123.79,
+      "step": 47955,
+      "train_speed(iter/s)": 1.317613
+    },
+    {
+      "acc": 0.6784461,
+      "epoch": 1.2166412988330797,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.6396844883723092e-06,
+      "loss": 1.46479044,
+      "memory(GiB)": 123.79,
+      "step": 47960,
+      "train_speed(iter/s)": 1.317626
+    },
+    {
+      "acc": 0.6897141,
+      "epoch": 1.2167681380010147,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.6386754493175893e-06,
+      "loss": 1.42705154,
+      "memory(GiB)": 123.79,
+      "step": 47965,
+      "train_speed(iter/s)": 1.317639
+    },
+    {
+      "acc": 0.68736696,
+      "epoch": 1.2168949771689497,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.637666470140003e-06,
+      "loss": 1.38035526,
+      "memory(GiB)": 123.79,
+      "step": 47970,
+      "train_speed(iter/s)": 1.317652
+    },
+    {
+      "acc": 0.69977298,
+      "epoch": 1.2170218163368849,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.6366575508839265e-06,
+      "loss": 1.37779608,
+      "memory(GiB)": 123.79,
+      "step": 47975,
+      "train_speed(iter/s)": 1.317665
+    },
+    {
+      "acc": 0.67913566,
+      "epoch": 1.2171486555048199,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.635648691593737e-06,
+      "loss": 1.46924133,
+      "memory(GiB)": 123.79,
+      "step": 47980,
+      "train_speed(iter/s)": 1.317678
+    },
+    {
+      "acc": 0.68254604,
+      "epoch": 1.2172754946727549,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.6346398923138094e-06,
+      "loss": 1.39035473,
+      "memory(GiB)": 123.79,
+      "step": 47985,
+      "train_speed(iter/s)": 1.317692
+    },
+    {
+      "acc": 0.68774838,
+      "epoch": 1.21740233384069,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.633631153088517e-06,
+      "loss": 1.36825447,
+      "memory(GiB)": 123.79,
+      "step": 47990,
+      "train_speed(iter/s)": 1.317705
+    },
+    {
+      "acc": 0.67930584,
+      "epoch": 1.217529173008625,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.6326224739622255e-06,
+      "loss": 1.46894703,
+      "memory(GiB)": 123.79,
+      "step": 47995,
+      "train_speed(iter/s)": 1.317718
+    },
+    {
+      "acc": 0.68701181,
+      "epoch": 1.21765601217656,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.6316138549793024e-06,
+      "loss": 1.38347464,
+      "memory(GiB)": 123.79,
+      "step": 48000,
+      "train_speed(iter/s)": 1.317731
+    },
+    {
+      "epoch": 1.21765601217656,
+      "eval_acc": 0.6750578982869625,
+      "eval_loss": 1.3565645217895508,
+      "eval_runtime": 70.275,
+      "eval_samples_per_second": 90.644,
+      "eval_steps_per_second": 22.668,
+      "step": 48000
+    },
+    {
+      "acc": 0.69145312,
+      "epoch": 1.2177828513444953,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.630605296184111e-06,
+      "loss": 1.32736435,
+      "memory(GiB)": 123.79,
+      "step": 48005,
+      "train_speed(iter/s)": 1.314756
+    },
+    {
+      "acc": 0.67754345,
+      "epoch": 1.2179096905124303,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.6295967976210146e-06,
+      "loss": 1.46567793,
+      "memory(GiB)": 123.79,
+      "step": 48010,
+      "train_speed(iter/s)": 1.314771
+    },
+    {
+      "acc": 0.67283978,
+      "epoch": 1.2180365296803652,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.6285883593343685e-06,
+      "loss": 1.46546497,
+      "memory(GiB)": 123.79,
+      "step": 48015,
+      "train_speed(iter/s)": 1.314787
+    },
+    {
+      "acc": 0.68891792,
+      "epoch": 1.2181633688483005,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.6275799813685274e-06,
+      "loss": 1.403234,
+      "memory(GiB)": 123.79,
+      "step": 48020,
+      "train_speed(iter/s)": 1.314802
+    },
+    {
+      "acc": 0.68499718,
+      "epoch": 1.2182902080162354,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.6265716637678484e-06,
+      "loss": 1.39685659,
+      "memory(GiB)": 123.79,
+      "step": 48025,
+      "train_speed(iter/s)": 1.314817
+    },
+    {
+      "acc": 0.68472896,
+      "epoch": 1.2184170471841704,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.62556340657668e-06,
+      "loss": 1.39544353,
+      "memory(GiB)": 123.79,
+      "step": 48030,
+      "train_speed(iter/s)": 1.314833
+    },
+    {
+      "acc": 0.68809309,
+      "epoch": 1.2185438863521054,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.6245552098393665e-06,
+      "loss": 1.42021198,
+      "memory(GiB)": 123.79,
+      "step": 48035,
+      "train_speed(iter/s)": 1.314847
+    },
+    {
+      "acc": 0.68333468,
+      "epoch": 1.2186707255200406,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.6235470736002576e-06,
+      "loss": 1.37483883,
+      "memory(GiB)": 123.79,
+      "step": 48040,
+      "train_speed(iter/s)": 1.314863
+    },
+    {
+      "acc": 0.69466128,
+      "epoch": 1.2187975646879756,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.622538997903693e-06,
+      "loss": 1.39033537,
+      "memory(GiB)": 123.79,
+      "step": 48045,
+      "train_speed(iter/s)": 1.314878
+    },
+    {
+      "acc": 0.68471017,
+      "epoch": 1.2189244038559106,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.621530982794015e-06,
+      "loss": 1.42275124,
+      "memory(GiB)": 123.79,
+      "step": 48050,
+      "train_speed(iter/s)": 1.314889
+    },
+    {
+      "acc": 0.67727628,
+      "epoch": 1.2190512430238458,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.620523028315558e-06,
+      "loss": 1.43586235,
+      "memory(GiB)": 123.79,
+      "step": 48055,
+      "train_speed(iter/s)": 1.314903
+    },
+    {
+      "acc": 0.68040075,
+      "epoch": 1.2191780821917808,
+      "grad_norm": 3.5,
+      "learning_rate": 3.6195151345126556e-06,
+      "loss": 1.48255424,
+      "memory(GiB)": 123.79,
+      "step": 48060,
+      "train_speed(iter/s)": 1.314918
+    },
+    {
+      "acc": 0.69546161,
+      "epoch": 1.2193049213597158,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.6185073014296425e-06,
+      "loss": 1.36032295,
+      "memory(GiB)": 123.79,
+      "step": 48065,
+      "train_speed(iter/s)": 1.314932
+    },
+    {
+      "acc": 0.69554558,
+      "epoch": 1.219431760527651,
+      "grad_norm": 3.375,
+      "learning_rate": 3.6174995291108474e-06,
+      "loss": 1.33845959,
+      "memory(GiB)": 123.79,
+      "step": 48070,
+      "train_speed(iter/s)": 1.314948
+    },
+    {
+      "acc": 0.67682023,
+      "epoch": 1.219558599695586,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.6164918176005937e-06,
+      "loss": 1.42498217,
+      "memory(GiB)": 123.79,
+      "step": 48075,
+      "train_speed(iter/s)": 1.314962
+    },
+    {
+      "acc": 0.67665963,
+      "epoch": 1.219685438863521,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.6154841669432062e-06,
+      "loss": 1.39108486,
+      "memory(GiB)": 123.79,
+      "step": 48080,
+      "train_speed(iter/s)": 1.314975
+    },
+    {
+      "acc": 0.69170933,
+      "epoch": 1.2198122780314562,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.614476577183007e-06,
+      "loss": 1.35249147,
+      "memory(GiB)": 123.79,
+      "step": 48085,
+      "train_speed(iter/s)": 1.314991
+    },
+    {
+      "acc": 0.69536877,
+      "epoch": 1.2199391171993912,
+      "grad_norm": 2.625,
+      "learning_rate": 3.6134690483643154e-06,
+      "loss": 1.35454044,
+      "memory(GiB)": 123.79,
+      "step": 48090,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.67689238,
+      "epoch": 1.2200659563673262,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.6124615805314434e-06,
+      "loss": 1.41409388,
+      "memory(GiB)": 123.79,
+      "step": 48095,
+      "train_speed(iter/s)": 1.315022
+    },
+    {
+      "acc": 0.67691483,
+      "epoch": 1.2201927955352612,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.611454173728707e-06,
+      "loss": 1.41126022,
+      "memory(GiB)": 123.79,
+      "step": 48100,
+      "train_speed(iter/s)": 1.315035
+    },
+    {
+      "acc": 0.69201241,
+      "epoch": 1.2203196347031964,
+      "grad_norm": 2.875,
+      "learning_rate": 3.610446828000414e-06,
+      "loss": 1.37089787,
+      "memory(GiB)": 123.79,
+      "step": 48105,
+      "train_speed(iter/s)": 1.31505
+    },
+    {
+      "acc": 0.68527231,
+      "epoch": 1.2204464738711314,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.609439543390877e-06,
+      "loss": 1.39703245,
+      "memory(GiB)": 123.79,
+      "step": 48110,
+      "train_speed(iter/s)": 1.315066
+    },
+    {
+      "acc": 0.68633375,
+      "epoch": 1.2205733130390666,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.608432319944394e-06,
+      "loss": 1.42561684,
+      "memory(GiB)": 123.79,
+      "step": 48115,
+      "train_speed(iter/s)": 1.315083
+    },
+    {
+      "acc": 0.686133,
+      "epoch": 1.2207001522070016,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.607425157705271e-06,
+      "loss": 1.37715492,
+      "memory(GiB)": 123.79,
+      "step": 48120,
+      "train_speed(iter/s)": 1.315098
+    },
+    {
+      "acc": 0.68778868,
+      "epoch": 1.2208269913749366,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.6064180567178064e-06,
+      "loss": 1.39617023,
+      "memory(GiB)": 123.79,
+      "step": 48125,
+      "train_speed(iter/s)": 1.315114
+    },
+    {
+      "acc": 0.69525461,
+      "epoch": 1.2209538305428715,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.6054110170263002e-06,
+      "loss": 1.37408228,
+      "memory(GiB)": 123.79,
+      "step": 48130,
+      "train_speed(iter/s)": 1.315128
+    },
+    {
+      "acc": 0.68511753,
+      "epoch": 1.2210806697108068,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.6044040386750423e-06,
+      "loss": 1.37790527,
+      "memory(GiB)": 123.79,
+      "step": 48135,
+      "train_speed(iter/s)": 1.315143
+    },
+    {
+      "acc": 0.69316006,
+      "epoch": 1.2212075088787417,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.6033971217083242e-06,
+      "loss": 1.35095406,
+      "memory(GiB)": 123.79,
+      "step": 48140,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.6982698,
+      "epoch": 1.2213343480466767,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.602390266170438e-06,
+      "loss": 1.33794937,
+      "memory(GiB)": 123.79,
+      "step": 48145,
+      "train_speed(iter/s)": 1.315174
+    },
+    {
+      "acc": 0.68695927,
+      "epoch": 1.221461187214612,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.6013834721056683e-06,
+      "loss": 1.38550959,
+      "memory(GiB)": 123.79,
+      "step": 48150,
+      "train_speed(iter/s)": 1.315189
+    },
+    {
+      "acc": 0.70043836,
+      "epoch": 1.221588026382547,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.6003767395582967e-06,
+      "loss": 1.24636955,
+      "memory(GiB)": 123.79,
+      "step": 48155,
+      "train_speed(iter/s)": 1.315204
+    },
+    {
+      "acc": 0.69565434,
+      "epoch": 1.221714865550482,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.599370068572604e-06,
+      "loss": 1.31676483,
+      "memory(GiB)": 123.79,
+      "step": 48160,
+      "train_speed(iter/s)": 1.315218
+    },
+    {
+      "acc": 0.68985424,
+      "epoch": 1.2218417047184171,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.5983634591928705e-06,
+      "loss": 1.39373617,
+      "memory(GiB)": 123.79,
+      "step": 48165,
+      "train_speed(iter/s)": 1.315234
+    },
+    {
+      "acc": 0.70241313,
+      "epoch": 1.2219685438863521,
+      "grad_norm": 3.0,
+      "learning_rate": 3.5973569114633704e-06,
+      "loss": 1.37351265,
+      "memory(GiB)": 123.79,
+      "step": 48170,
+      "train_speed(iter/s)": 1.315249
+    },
+    {
+      "acc": 0.69338279,
+      "epoch": 1.222095383054287,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.5963504254283743e-06,
+      "loss": 1.41969662,
+      "memory(GiB)": 123.79,
+      "step": 48175,
+      "train_speed(iter/s)": 1.315265
+    },
+    {
+      "acc": 0.69232788,
+      "epoch": 1.2222222222222223,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.595344001132154e-06,
+      "loss": 1.34384232,
+      "memory(GiB)": 123.79,
+      "step": 48180,
+      "train_speed(iter/s)": 1.315282
+    },
+    {
+      "acc": 0.68589187,
+      "epoch": 1.2223490613901573,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.5943376386189744e-06,
+      "loss": 1.37854357,
+      "memory(GiB)": 123.79,
+      "step": 48185,
+      "train_speed(iter/s)": 1.315297
+    },
+    {
+      "acc": 0.69741154,
+      "epoch": 1.2224759005580923,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.5933313379331047e-06,
+      "loss": 1.37605076,
+      "memory(GiB)": 123.79,
+      "step": 48190,
+      "train_speed(iter/s)": 1.315313
+    },
+    {
+      "acc": 0.68021722,
+      "epoch": 1.2226027397260273,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.5923250991188e-06,
+      "loss": 1.42415037,
+      "memory(GiB)": 123.79,
+      "step": 48195,
+      "train_speed(iter/s)": 1.315325
+    },
+    {
+      "acc": 0.68360462,
+      "epoch": 1.2227295788939625,
+      "grad_norm": 3.5,
+      "learning_rate": 3.591318922220324e-06,
+      "loss": 1.41765966,
+      "memory(GiB)": 123.79,
+      "step": 48200,
+      "train_speed(iter/s)": 1.315341
+    },
+    {
+      "acc": 0.68797493,
+      "epoch": 1.2228564180618975,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.5903128072819287e-06,
+      "loss": 1.41077499,
+      "memory(GiB)": 123.79,
+      "step": 48205,
+      "train_speed(iter/s)": 1.315357
+    },
+    {
+      "acc": 0.68597474,
+      "epoch": 1.2229832572298325,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.5893067543478733e-06,
+      "loss": 1.4212306,
+      "memory(GiB)": 123.79,
+      "step": 48210,
+      "train_speed(iter/s)": 1.315367
+    },
+    {
+      "acc": 0.67030363,
+      "epoch": 1.2231100963977677,
+      "grad_norm": 3.25,
+      "learning_rate": 3.5883007634624033e-06,
+      "loss": 1.46044998,
+      "memory(GiB)": 123.79,
+      "step": 48215,
+      "train_speed(iter/s)": 1.315382
+    },
+    {
+      "acc": 0.67842598,
+      "epoch": 1.2232369355657027,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.5872948346697676e-06,
+      "loss": 1.36625538,
+      "memory(GiB)": 123.79,
+      "step": 48220,
+      "train_speed(iter/s)": 1.315397
+    },
+    {
+      "acc": 0.71178894,
+      "epoch": 1.2233637747336377,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.5862889680142133e-06,
+      "loss": 1.33835926,
+      "memory(GiB)": 123.79,
+      "step": 48225,
+      "train_speed(iter/s)": 1.315412
+    },
+    {
+      "acc": 0.69050665,
+      "epoch": 1.2234906139015729,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5852831635399833e-06,
+      "loss": 1.40191498,
+      "memory(GiB)": 123.79,
+      "step": 48230,
+      "train_speed(iter/s)": 1.315427
+    },
+    {
+      "acc": 0.69082336,
+      "epoch": 1.2236174530695079,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.5842774212913144e-06,
+      "loss": 1.40843019,
+      "memory(GiB)": 123.79,
+      "step": 48235,
+      "train_speed(iter/s)": 1.315442
+    },
+    {
+      "acc": 0.68530717,
+      "epoch": 1.2237442922374429,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.583271741312445e-06,
+      "loss": 1.39761829,
+      "memory(GiB)": 123.79,
+      "step": 48240,
+      "train_speed(iter/s)": 1.315457
+    },
+    {
+      "acc": 0.68591995,
+      "epoch": 1.223871131405378,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.58226612364761e-06,
+      "loss": 1.3771637,
+      "memory(GiB)": 123.79,
+      "step": 48245,
+      "train_speed(iter/s)": 1.315472
+    },
+    {
+      "acc": 0.70019436,
+      "epoch": 1.223997970573313,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.581260568341042e-06,
+      "loss": 1.35260429,
+      "memory(GiB)": 123.79,
+      "step": 48250,
+      "train_speed(iter/s)": 1.315488
+    },
+    {
+      "acc": 0.6786067,
+      "epoch": 1.224124809741248,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.580255075436967e-06,
+      "loss": 1.44767084,
+      "memory(GiB)": 123.79,
+      "step": 48255,
+      "train_speed(iter/s)": 1.315503
+    },
+    {
+      "acc": 0.68265567,
+      "epoch": 1.224251648909183,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.5792496449796127e-06,
+      "loss": 1.37257767,
+      "memory(GiB)": 123.79,
+      "step": 48260,
+      "train_speed(iter/s)": 1.315518
+    },
+    {
+      "acc": 0.70072403,
+      "epoch": 1.2243784880771182,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.578244277013201e-06,
+      "loss": 1.35742817,
+      "memory(GiB)": 123.79,
+      "step": 48265,
+      "train_speed(iter/s)": 1.315533
+    },
+    {
+      "acc": 0.69705434,
+      "epoch": 1.2245053272450532,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.5772389715819568e-06,
+      "loss": 1.4048151,
+      "memory(GiB)": 123.79,
+      "step": 48270,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.69499326,
+      "epoch": 1.2246321664129884,
+      "grad_norm": 3.25,
+      "learning_rate": 3.5762337287300925e-06,
+      "loss": 1.3643364,
+      "memory(GiB)": 123.79,
+      "step": 48275,
+      "train_speed(iter/s)": 1.315563
+    },
+    {
+      "acc": 0.68018174,
+      "epoch": 1.2247590055809234,
+      "grad_norm": 2.875,
+      "learning_rate": 3.575228548501825e-06,
+      "loss": 1.40918331,
+      "memory(GiB)": 123.79,
+      "step": 48280,
+      "train_speed(iter/s)": 1.315579
+    },
+    {
+      "acc": 0.6808672,
+      "epoch": 1.2248858447488584,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.574223430941368e-06,
+      "loss": 1.39239063,
+      "memory(GiB)": 123.79,
+      "step": 48285,
+      "train_speed(iter/s)": 1.315595
+    },
+    {
+      "acc": 0.67498159,
+      "epoch": 1.2250126839167934,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.573218376092932e-06,
+      "loss": 1.49157505,
+      "memory(GiB)": 123.79,
+      "step": 48290,
+      "train_speed(iter/s)": 1.31561
+    },
+    {
+      "acc": 0.68999772,
+      "epoch": 1.2251395230847286,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.5722133840007197e-06,
+      "loss": 1.39511976,
+      "memory(GiB)": 123.79,
+      "step": 48295,
+      "train_speed(iter/s)": 1.315626
+    },
+    {
+      "acc": 0.67661104,
+      "epoch": 1.2252663622526636,
+      "grad_norm": 4.0,
+      "learning_rate": 3.5712084547089367e-06,
+      "loss": 1.37702875,
+      "memory(GiB)": 123.79,
+      "step": 48300,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.69126825,
+      "epoch": 1.2253932014205986,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.5702035882617857e-06,
+      "loss": 1.38214359,
+      "memory(GiB)": 123.79,
+      "step": 48305,
+      "train_speed(iter/s)": 1.315657
+    },
+    {
+      "acc": 0.678617,
+      "epoch": 1.2255200405885338,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.5691987847034667e-06,
+      "loss": 1.43832893,
+      "memory(GiB)": 123.79,
+      "step": 48310,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.68248425,
+      "epoch": 1.2256468797564688,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.5681940440781705e-06,
+      "loss": 1.41406765,
+      "memory(GiB)": 123.79,
+      "step": 48315,
+      "train_speed(iter/s)": 1.315686
+    },
+    {
+      "acc": 0.6951088,
+      "epoch": 1.2257737189244038,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.5671893664300934e-06,
+      "loss": 1.35248089,
+      "memory(GiB)": 123.79,
+      "step": 48320,
+      "train_speed(iter/s)": 1.315701
+    },
+    {
+      "acc": 0.70528193,
+      "epoch": 1.225900558092339,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.5661847518034244e-06,
+      "loss": 1.27089405,
+      "memory(GiB)": 123.79,
+      "step": 48325,
+      "train_speed(iter/s)": 1.315549
+    },
+    {
+      "acc": 0.6690742,
+      "epoch": 1.226027397260274,
+      "grad_norm": 3.0,
+      "learning_rate": 3.5651802002423543e-06,
+      "loss": 1.41595621,
+      "memory(GiB)": 123.79,
+      "step": 48330,
+      "train_speed(iter/s)": 1.315565
+    },
+    {
+      "acc": 0.70144553,
+      "epoch": 1.226154236428209,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.5641757117910625e-06,
+      "loss": 1.35079174,
+      "memory(GiB)": 123.79,
+      "step": 48335,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.67324533,
+      "epoch": 1.2262810755961442,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.563171286493734e-06,
+      "loss": 1.43154125,
+      "memory(GiB)": 123.79,
+      "step": 48340,
+      "train_speed(iter/s)": 1.315593
+    },
+    {
+      "acc": 0.66737852,
+      "epoch": 1.2264079147640792,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.5621669243945457e-06,
+      "loss": 1.42523136,
+      "memory(GiB)": 123.79,
+      "step": 48345,
+      "train_speed(iter/s)": 1.315608
+    },
+    {
+      "acc": 0.67858572,
+      "epoch": 1.2265347539320142,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.5611626255376785e-06,
+      "loss": 1.38544569,
+      "memory(GiB)": 123.79,
+      "step": 48350,
+      "train_speed(iter/s)": 1.315623
+    },
+    {
+      "acc": 0.70395803,
+      "epoch": 1.2266615930999492,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.560158389967302e-06,
+      "loss": 1.39370384,
+      "memory(GiB)": 123.79,
+      "step": 48355,
+      "train_speed(iter/s)": 1.315639
+    },
+    {
+      "acc": 0.68628974,
+      "epoch": 1.2267884322678844,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.559154217727586e-06,
+      "loss": 1.44643974,
+      "memory(GiB)": 123.79,
+      "step": 48360,
+      "train_speed(iter/s)": 1.315655
+    },
+    {
+      "acc": 0.68406143,
+      "epoch": 1.2269152714358194,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.5581501088627026e-06,
+      "loss": 1.47227955,
+      "memory(GiB)": 123.79,
+      "step": 48365,
+      "train_speed(iter/s)": 1.315671
+    },
+    {
+      "acc": 0.69209609,
+      "epoch": 1.2270421106037543,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.557146063416815e-06,
+      "loss": 1.38476276,
+      "memory(GiB)": 123.79,
+      "step": 48370,
+      "train_speed(iter/s)": 1.315686
+    },
+    {
+      "acc": 0.68276939,
+      "epoch": 1.2271689497716896,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.5561420814340843e-06,
+      "loss": 1.42034369,
+      "memory(GiB)": 123.79,
+      "step": 48375,
+      "train_speed(iter/s)": 1.315702
+    },
+    {
+      "acc": 0.67057562,
+      "epoch": 1.2272957889396245,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.555138162958671e-06,
+      "loss": 1.45262432,
+      "memory(GiB)": 123.79,
+      "step": 48380,
+      "train_speed(iter/s)": 1.315716
+    },
+    {
+      "acc": 0.69372263,
+      "epoch": 1.2274226281075595,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.5541343080347325e-06,
+      "loss": 1.33766222,
+      "memory(GiB)": 123.79,
+      "step": 48385,
+      "train_speed(iter/s)": 1.315733
+    },
+    {
+      "acc": 0.68357234,
+      "epoch": 1.2275494672754947,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.5531305167064234e-06,
+      "loss": 1.40117083,
+      "memory(GiB)": 123.79,
+      "step": 48390,
+      "train_speed(iter/s)": 1.315746
+    },
+    {
+      "acc": 0.69915843,
+      "epoch": 1.2276763064434297,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5521267890178922e-06,
+      "loss": 1.35043707,
+      "memory(GiB)": 123.79,
+      "step": 48395,
+      "train_speed(iter/s)": 1.315762
+    },
+    {
+      "acc": 0.68469019,
+      "epoch": 1.2278031456113647,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.5511231250132905e-06,
+      "loss": 1.37441978,
+      "memory(GiB)": 123.79,
+      "step": 48400,
+      "train_speed(iter/s)": 1.315778
+    },
+    {
+      "acc": 0.69495363,
+      "epoch": 1.2279299847793,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.550119524736761e-06,
+      "loss": 1.42608118,
+      "memory(GiB)": 123.79,
+      "step": 48405,
+      "train_speed(iter/s)": 1.315794
+    },
+    {
+      "acc": 0.67604914,
+      "epoch": 1.228056823947235,
+      "grad_norm": 3.625,
+      "learning_rate": 3.5491159882324513e-06,
+      "loss": 1.39135237,
+      "memory(GiB)": 123.79,
+      "step": 48410,
+      "train_speed(iter/s)": 1.31581
+    },
+    {
+      "acc": 0.681003,
+      "epoch": 1.22818366311517,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.548112515544495e-06,
+      "loss": 1.37822895,
+      "memory(GiB)": 123.79,
+      "step": 48415,
+      "train_speed(iter/s)": 1.315826
+    },
+    {
+      "acc": 0.68865347,
+      "epoch": 1.228310502283105,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.547109106717034e-06,
+      "loss": 1.37661304,
+      "memory(GiB)": 123.79,
+      "step": 48420,
+      "train_speed(iter/s)": 1.315841
+    },
+    {
+      "acc": 0.69412036,
+      "epoch": 1.22843734145104,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.546105761794199e-06,
+      "loss": 1.40649986,
+      "memory(GiB)": 123.79,
+      "step": 48425,
+      "train_speed(iter/s)": 1.315857
+    },
+    {
+      "acc": 0.68723993,
+      "epoch": 1.228564180618975,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.5451024808201268e-06,
+      "loss": 1.38172808,
+      "memory(GiB)": 123.79,
+      "step": 48430,
+      "train_speed(iter/s)": 1.315873
+    },
+    {
+      "acc": 0.66570454,
+      "epoch": 1.2286910197869103,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.5440992638389417e-06,
+      "loss": 1.46403122,
+      "memory(GiB)": 123.79,
+      "step": 48435,
+      "train_speed(iter/s)": 1.315888
+    },
+    {
+      "acc": 0.69214063,
+      "epoch": 1.2288178589548453,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.5430961108947705e-06,
+      "loss": 1.40066051,
+      "memory(GiB)": 123.79,
+      "step": 48440,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.67167783,
+      "epoch": 1.2289446981227803,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.5420930220317373e-06,
+      "loss": 1.40316105,
+      "memory(GiB)": 123.79,
+      "step": 48445,
+      "train_speed(iter/s)": 1.315919
+    },
+    {
+      "acc": 0.70234079,
+      "epoch": 1.2290715372907153,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.541089997293964e-06,
+      "loss": 1.3170352,
+      "memory(GiB)": 123.79,
+      "step": 48450,
+      "train_speed(iter/s)": 1.315935
+    },
+    {
+      "acc": 0.69275885,
+      "epoch": 1.2291983764586505,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.5400870367255635e-06,
+      "loss": 1.34080095,
+      "memory(GiB)": 123.79,
+      "step": 48455,
+      "train_speed(iter/s)": 1.315951
+    },
+    {
+      "acc": 0.67969232,
+      "epoch": 1.2293252156265855,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.539084140370654e-06,
+      "loss": 1.36142969,
+      "memory(GiB)": 123.79,
+      "step": 48460,
+      "train_speed(iter/s)": 1.315966
+    },
+    {
+      "acc": 0.68257318,
+      "epoch": 1.2294520547945205,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.538081308273347e-06,
+      "loss": 1.42273083,
+      "memory(GiB)": 123.79,
+      "step": 48465,
+      "train_speed(iter/s)": 1.315982
+    },
+    {
+      "acc": 0.68610229,
+      "epoch": 1.2295788939624557,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.537078540477752e-06,
+      "loss": 1.36970854,
+      "memory(GiB)": 123.79,
+      "step": 48470,
+      "train_speed(iter/s)": 1.315997
+    },
+    {
+      "acc": 0.67042041,
+      "epoch": 1.2297057331303907,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5360758370279722e-06,
+      "loss": 1.43707457,
+      "memory(GiB)": 123.79,
+      "step": 48475,
+      "train_speed(iter/s)": 1.316012
+    },
+    {
+      "acc": 0.69555264,
+      "epoch": 1.2298325722983257,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.535073197968114e-06,
+      "loss": 1.32841892,
+      "memory(GiB)": 123.79,
+      "step": 48480,
+      "train_speed(iter/s)": 1.316027
+    },
+    {
+      "acc": 0.69528365,
+      "epoch": 1.2299594114662609,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.5340706233422763e-06,
+      "loss": 1.44435902,
+      "memory(GiB)": 123.79,
+      "step": 48485,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.67965326,
+      "epoch": 1.2300862506341959,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.5330681131945588e-06,
+      "loss": 1.41683807,
+      "memory(GiB)": 123.79,
+      "step": 48490,
+      "train_speed(iter/s)": 1.316058
+    },
+    {
+      "acc": 0.68033776,
+      "epoch": 1.2302130898021308,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.5320656675690546e-06,
+      "loss": 1.3572298,
+      "memory(GiB)": 123.79,
+      "step": 48495,
+      "train_speed(iter/s)": 1.316074
+    },
+    {
+      "acc": 0.68317952,
+      "epoch": 1.230339928970066,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.531063286509855e-06,
+      "loss": 1.35816507,
+      "memory(GiB)": 123.79,
+      "step": 48500,
+      "train_speed(iter/s)": 1.316089
+    },
+    {
+      "acc": 0.69259863,
+      "epoch": 1.230466768138001,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.530060970061051e-06,
+      "loss": 1.3884861,
+      "memory(GiB)": 123.79,
+      "step": 48505,
+      "train_speed(iter/s)": 1.316105
+    },
+    {
+      "acc": 0.68143206,
+      "epoch": 1.230593607305936,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.52905871826673e-06,
+      "loss": 1.44006691,
+      "memory(GiB)": 123.79,
+      "step": 48510,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.68638163,
+      "epoch": 1.230720446473871,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5280565311709725e-06,
+      "loss": 1.34303017,
+      "memory(GiB)": 123.79,
+      "step": 48515,
+      "train_speed(iter/s)": 1.316134
+    },
+    {
+      "acc": 0.68623743,
+      "epoch": 1.2308472856418062,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.5270544088178597e-06,
+      "loss": 1.43079681,
+      "memory(GiB)": 123.79,
+      "step": 48520,
+      "train_speed(iter/s)": 1.31615
+    },
+    {
+      "acc": 0.68245859,
+      "epoch": 1.2309741248097412,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.526052351251471e-06,
+      "loss": 1.42166023,
+      "memory(GiB)": 123.79,
+      "step": 48525,
+      "train_speed(iter/s)": 1.316166
+    },
+    {
+      "acc": 0.68724737,
+      "epoch": 1.2311009639776762,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.5250503585158825e-06,
+      "loss": 1.3966341,
+      "memory(GiB)": 123.79,
+      "step": 48530,
+      "train_speed(iter/s)": 1.31618
+    },
+    {
+      "acc": 0.68907108,
+      "epoch": 1.2312278031456114,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.5240484306551615e-06,
+      "loss": 1.37895889,
+      "memory(GiB)": 123.79,
+      "step": 48535,
+      "train_speed(iter/s)": 1.316195
+    },
+    {
+      "acc": 0.68116317,
+      "epoch": 1.2313546423135464,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.5230465677133813e-06,
+      "loss": 1.42932816,
+      "memory(GiB)": 123.79,
+      "step": 48540,
+      "train_speed(iter/s)": 1.31621
+    },
+    {
+      "acc": 0.67847414,
+      "epoch": 1.2314814814814814,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.5220447697346063e-06,
+      "loss": 1.42969971,
+      "memory(GiB)": 123.79,
+      "step": 48545,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.67745857,
+      "epoch": 1.2316083206494166,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.521043036762903e-06,
+      "loss": 1.46325293,
+      "memory(GiB)": 123.79,
+      "step": 48550,
+      "train_speed(iter/s)": 1.316241
+    },
+    {
+      "acc": 0.67848434,
+      "epoch": 1.2317351598173516,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.5200413688423284e-06,
+      "loss": 1.40226955,
+      "memory(GiB)": 123.79,
+      "step": 48555,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.67864466,
+      "epoch": 1.2318619989852866,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.519039766016943e-06,
+      "loss": 1.44644489,
+      "memory(GiB)": 123.79,
+      "step": 48560,
+      "train_speed(iter/s)": 1.316269
+    },
+    {
+      "acc": 0.67903299,
+      "epoch": 1.2319888381532218,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.5180382283307983e-06,
+      "loss": 1.48660364,
+      "memory(GiB)": 123.79,
+      "step": 48565,
+      "train_speed(iter/s)": 1.316285
+    },
+    {
+      "acc": 0.7019558,
+      "epoch": 1.2321156773211568,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.517036755827952e-06,
+      "loss": 1.33902979,
+      "memory(GiB)": 123.79,
+      "step": 48570,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.69828959,
+      "epoch": 1.2322425164890918,
+      "grad_norm": 2.875,
+      "learning_rate": 3.516035348552449e-06,
+      "loss": 1.39469452,
+      "memory(GiB)": 123.79,
+      "step": 48575,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.69062195,
+      "epoch": 1.2323693556570268,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.515034006548335e-06,
+      "loss": 1.38383141,
+      "memory(GiB)": 123.79,
+      "step": 48580,
+      "train_speed(iter/s)": 1.316332
+    },
+    {
+      "acc": 0.69692945,
+      "epoch": 1.232496194824962,
+      "grad_norm": 3.375,
+      "learning_rate": 3.5140327298596565e-06,
+      "loss": 1.35930233,
+      "memory(GiB)": 123.79,
+      "step": 48585,
+      "train_speed(iter/s)": 1.316348
+    },
+    {
+      "acc": 0.68339415,
+      "epoch": 1.232623033992897,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.5130315185304547e-06,
+      "loss": 1.36553555,
+      "memory(GiB)": 123.79,
+      "step": 48590,
+      "train_speed(iter/s)": 1.316364
+    },
+    {
+      "acc": 0.67796698,
+      "epoch": 1.2327498731608322,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.5120303726047642e-06,
+      "loss": 1.3989481,
+      "memory(GiB)": 123.79,
+      "step": 48595,
+      "train_speed(iter/s)": 1.31638
+    },
+    {
+      "acc": 0.68065348,
+      "epoch": 1.2328767123287672,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.51102929212662e-06,
+      "loss": 1.48995562,
+      "memory(GiB)": 123.79,
+      "step": 48600,
+      "train_speed(iter/s)": 1.316396
+    },
+    {
+      "acc": 0.67995105,
+      "epoch": 1.2330035514967022,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.5100282771400563e-06,
+      "loss": 1.46037207,
+      "memory(GiB)": 123.79,
+      "step": 48605,
+      "train_speed(iter/s)": 1.316411
+    },
+    {
+      "acc": 0.67676978,
+      "epoch": 1.2331303906646371,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.5090273276891023e-06,
+      "loss": 1.38063889,
+      "memory(GiB)": 123.79,
+      "step": 48610,
+      "train_speed(iter/s)": 1.316427
+    },
+    {
+      "acc": 0.69172473,
+      "epoch": 1.2332572298325724,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5080264438177815e-06,
+      "loss": 1.36000233,
+      "memory(GiB)": 123.79,
+      "step": 48615,
+      "train_speed(iter/s)": 1.316443
+    },
+    {
+      "acc": 0.68538623,
+      "epoch": 1.2333840690005073,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.50702562557012e-06,
+      "loss": 1.45851049,
+      "memory(GiB)": 123.79,
+      "step": 48620,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.68180699,
+      "epoch": 1.2335109081684423,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.506024872990135e-06,
+      "loss": 1.41042261,
+      "memory(GiB)": 123.79,
+      "step": 48625,
+      "train_speed(iter/s)": 1.316475
+    },
+    {
+      "acc": 0.68254061,
+      "epoch": 1.2336377473363775,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.5050241861218493e-06,
+      "loss": 1.3605979,
+      "memory(GiB)": 123.79,
+      "step": 48630,
+      "train_speed(iter/s)": 1.31649
+    },
+    {
+      "acc": 0.68955774,
+      "epoch": 1.2337645865043125,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.5040235650092725e-06,
+      "loss": 1.36167297,
+      "memory(GiB)": 123.79,
+      "step": 48635,
+      "train_speed(iter/s)": 1.316506
+    },
+    {
+      "acc": 0.71143866,
+      "epoch": 1.2338914256722475,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.503023009696419e-06,
+      "loss": 1.25961018,
+      "memory(GiB)": 123.79,
+      "step": 48640,
+      "train_speed(iter/s)": 1.316522
+    },
+    {
+      "acc": 0.68219547,
+      "epoch": 1.2340182648401827,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.5020225202272963e-06,
+      "loss": 1.42125607,
+      "memory(GiB)": 123.79,
+      "step": 48645,
+      "train_speed(iter/s)": 1.316538
+    },
+    {
+      "acc": 0.67172904,
+      "epoch": 1.2341451040081177,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.501022096645913e-06,
+      "loss": 1.45821791,
+      "memory(GiB)": 123.79,
+      "step": 48650,
+      "train_speed(iter/s)": 1.316553
+    },
+    {
+      "acc": 0.68451128,
+      "epoch": 1.2342719431760527,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.5000217389962685e-06,
+      "loss": 1.4035429,
+      "memory(GiB)": 123.79,
+      "step": 48655,
+      "train_speed(iter/s)": 1.316569
+    },
+    {
+      "acc": 0.68741875,
+      "epoch": 1.234398782343988,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.499021447322365e-06,
+      "loss": 1.44102964,
+      "memory(GiB)": 123.79,
+      "step": 48660,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.68898978,
+      "epoch": 1.234525621511923,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.4980212216681997e-06,
+      "loss": 1.39804201,
+      "memory(GiB)": 123.79,
+      "step": 48665,
+      "train_speed(iter/s)": 1.316601
+    },
+    {
+      "acc": 0.69288244,
+      "epoch": 1.234652460679858,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.4970210620777687e-06,
+      "loss": 1.3766674,
+      "memory(GiB)": 123.79,
+      "step": 48670,
+      "train_speed(iter/s)": 1.316617
+    },
+    {
+      "acc": 0.69369316,
+      "epoch": 1.2347792998477929,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.496020968595059e-06,
+      "loss": 1.40409346,
+      "memory(GiB)": 123.79,
+      "step": 48675,
+      "train_speed(iter/s)": 1.316633
+    },
+    {
+      "acc": 0.69877067,
+      "epoch": 1.234906139015728,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.4950209412640634e-06,
+      "loss": 1.34862309,
+      "memory(GiB)": 123.79,
+      "step": 48680,
+      "train_speed(iter/s)": 1.316649
+    },
+    {
+      "acc": 0.70059967,
+      "epoch": 1.235032978183663,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.494020980128766e-06,
+      "loss": 1.31959438,
+      "memory(GiB)": 123.79,
+      "step": 48685,
+      "train_speed(iter/s)": 1.316665
+    },
+    {
+      "acc": 0.69480877,
+      "epoch": 1.235159817351598,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.4930210852331505e-06,
+      "loss": 1.32086897,
+      "memory(GiB)": 123.79,
+      "step": 48690,
+      "train_speed(iter/s)": 1.316682
+    },
+    {
+      "acc": 0.68851457,
+      "epoch": 1.2352866565195333,
+      "grad_norm": 3.75,
+      "learning_rate": 3.4920212566211943e-06,
+      "loss": 1.30462112,
+      "memory(GiB)": 123.79,
+      "step": 48695,
+      "train_speed(iter/s)": 1.316698
+    },
+    {
+      "acc": 0.68046741,
+      "epoch": 1.2354134956874683,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.491021494336876e-06,
+      "loss": 1.40131826,
+      "memory(GiB)": 123.79,
+      "step": 48700,
+      "train_speed(iter/s)": 1.316714
+    },
+    {
+      "acc": 0.67889948,
+      "epoch": 1.2355403348554033,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.4900217984241692e-06,
+      "loss": 1.39122858,
+      "memory(GiB)": 123.79,
+      "step": 48705,
+      "train_speed(iter/s)": 1.31673
+    },
+    {
+      "acc": 0.68529983,
+      "epoch": 1.2356671740233385,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.4890221689270466e-06,
+      "loss": 1.37001343,
+      "memory(GiB)": 123.79,
+      "step": 48710,
+      "train_speed(iter/s)": 1.316745
+    },
+    {
+      "acc": 0.69437385,
+      "epoch": 1.2357940131912735,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.488022605889475e-06,
+      "loss": 1.36767559,
+      "memory(GiB)": 123.79,
+      "step": 48715,
+      "train_speed(iter/s)": 1.316761
+    },
+    {
+      "acc": 0.68888159,
+      "epoch": 1.2359208523592085,
+      "grad_norm": 3.75,
+      "learning_rate": 3.4870231093554172e-06,
+      "loss": 1.33056412,
+      "memory(GiB)": 123.79,
+      "step": 48720,
+      "train_speed(iter/s)": 1.316777
+    },
+    {
+      "acc": 0.6819293,
+      "epoch": 1.2360476915271437,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.4860236793688407e-06,
+      "loss": 1.41893682,
+      "memory(GiB)": 123.79,
+      "step": 48725,
+      "train_speed(iter/s)": 1.316792
+    },
+    {
+      "acc": 0.66923165,
+      "epoch": 1.2361745306950787,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.4850243159737024e-06,
+      "loss": 1.37715311,
+      "memory(GiB)": 123.79,
+      "step": 48730,
+      "train_speed(iter/s)": 1.316808
+    },
+    {
+      "acc": 0.68321562,
+      "epoch": 1.2363013698630136,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.4840250192139574e-06,
+      "loss": 1.41738129,
+      "memory(GiB)": 123.79,
+      "step": 48735,
+      "train_speed(iter/s)": 1.316824
+    },
+    {
+      "acc": 0.6932868,
+      "epoch": 1.2364282090309486,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.4830257891335595e-06,
+      "loss": 1.3259408,
+      "memory(GiB)": 123.79,
+      "step": 48740,
+      "train_speed(iter/s)": 1.316841
+    },
+    {
+      "acc": 0.69005165,
+      "epoch": 1.2365550481988838,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.4820266257764613e-06,
+      "loss": 1.298913,
+      "memory(GiB)": 123.79,
+      "step": 48745,
+      "train_speed(iter/s)": 1.316857
+    },
+    {
+      "acc": 0.69845524,
+      "epoch": 1.2366818873668188,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.4810275291866103e-06,
+      "loss": 1.33631659,
+      "memory(GiB)": 123.79,
+      "step": 48750,
+      "train_speed(iter/s)": 1.316872
+    },
+    {
+      "acc": 0.68244252,
+      "epoch": 1.236808726534754,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.4800284994079487e-06,
+      "loss": 1.36031322,
+      "memory(GiB)": 123.79,
+      "step": 48755,
+      "train_speed(iter/s)": 1.316888
+    },
+    {
+      "acc": 0.67464304,
+      "epoch": 1.236935565702689,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.4790295364844207e-06,
+      "loss": 1.41869717,
+      "memory(GiB)": 123.79,
+      "step": 48760,
+      "train_speed(iter/s)": 1.316904
+    },
+    {
+      "acc": 0.68000522,
+      "epoch": 1.237062404870624,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.4780306404599628e-06,
+      "loss": 1.38055458,
+      "memory(GiB)": 123.79,
+      "step": 48765,
+      "train_speed(iter/s)": 1.31692
+    },
+    {
+      "acc": 0.67888803,
+      "epoch": 1.237189244038559,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.4770318113785164e-06,
+      "loss": 1.45627537,
+      "memory(GiB)": 123.79,
+      "step": 48770,
+      "train_speed(iter/s)": 1.316937
+    },
+    {
+      "acc": 0.68422403,
+      "epoch": 1.2373160832064942,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.4760330492840065e-06,
+      "loss": 1.44147701,
+      "memory(GiB)": 123.79,
+      "step": 48775,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.68401899,
+      "epoch": 1.2374429223744292,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.4750343542203684e-06,
+      "loss": 1.42131433,
+      "memory(GiB)": 123.79,
+      "step": 48780,
+      "train_speed(iter/s)": 1.316969
+    },
+    {
+      "acc": 0.67721577,
+      "epoch": 1.2375697615423642,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.474035726231527e-06,
+      "loss": 1.43257637,
+      "memory(GiB)": 123.79,
+      "step": 48785,
+      "train_speed(iter/s)": 1.316986
+    },
+    {
+      "acc": 0.69231644,
+      "epoch": 1.2376966007102994,
+      "grad_norm": 3.0,
+      "learning_rate": 3.473037165361409e-06,
+      "loss": 1.42220097,
+      "memory(GiB)": 123.79,
+      "step": 48790,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.6898396,
+      "epoch": 1.2378234398782344,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.4720386716539333e-06,
+      "loss": 1.37502146,
+      "memory(GiB)": 123.79,
+      "step": 48795,
+      "train_speed(iter/s)": 1.317017
+    },
+    {
+      "acc": 0.68126826,
+      "epoch": 1.2379502790461694,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.471040245153018e-06,
+      "loss": 1.38136425,
+      "memory(GiB)": 123.79,
+      "step": 48800,
+      "train_speed(iter/s)": 1.317034
+    },
+    {
+      "acc": 0.68909016,
+      "epoch": 1.2380771182141046,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.4700418859025793e-06,
+      "loss": 1.39895267,
+      "memory(GiB)": 123.79,
+      "step": 48805,
+      "train_speed(iter/s)": 1.31705
+    },
+    {
+      "acc": 0.69593782,
+      "epoch": 1.2382039573820396,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.4690435939465307e-06,
+      "loss": 1.3621089,
+      "memory(GiB)": 123.79,
+      "step": 48810,
+      "train_speed(iter/s)": 1.317066
+    },
+    {
+      "acc": 0.6843636,
+      "epoch": 1.2383307965499746,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.4680453693287786e-06,
+      "loss": 1.34784746,
+      "memory(GiB)": 123.79,
+      "step": 48815,
+      "train_speed(iter/s)": 1.317082
+    },
+    {
+      "acc": 0.68529949,
+      "epoch": 1.2384576357179098,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.4670472120932297e-06,
+      "loss": 1.41956739,
+      "memory(GiB)": 123.79,
+      "step": 48820,
+      "train_speed(iter/s)": 1.317097
+    },
+    {
+      "acc": 0.68156528,
+      "epoch": 1.2385844748858448,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.46604912228379e-06,
+      "loss": 1.4393158,
+      "memory(GiB)": 123.79,
+      "step": 48825,
+      "train_speed(iter/s)": 1.317112
+    },
+    {
+      "acc": 0.69001589,
+      "epoch": 1.2387113140537798,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.46505109994436e-06,
+      "loss": 1.37690868,
+      "memory(GiB)": 123.79,
+      "step": 48830,
+      "train_speed(iter/s)": 1.317128
+    },
+    {
+      "acc": 0.68568583,
+      "epoch": 1.2388381532217148,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.464053145118833e-06,
+      "loss": 1.32290363,
+      "memory(GiB)": 123.79,
+      "step": 48835,
+      "train_speed(iter/s)": 1.31714
+    },
+    {
+      "acc": 0.69155502,
+      "epoch": 1.23896499238965,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.4630552578511073e-06,
+      "loss": 1.4088913,
+      "memory(GiB)": 123.79,
+      "step": 48840,
+      "train_speed(iter/s)": 1.317156
+    },
+    {
+      "acc": 0.67745395,
+      "epoch": 1.239091831557585,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.4620574381850723e-06,
+      "loss": 1.39952278,
+      "memory(GiB)": 123.79,
+      "step": 48845,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.6842906,
+      "epoch": 1.23921867072552,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.4610596861646194e-06,
+      "loss": 1.42337627,
+      "memory(GiB)": 123.79,
+      "step": 48850,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.6994482,
+      "epoch": 1.2393455098934552,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.460062001833632e-06,
+      "loss": 1.34961166,
+      "memory(GiB)": 123.79,
+      "step": 48855,
+      "train_speed(iter/s)": 1.317203
+    },
+    {
+      "acc": 0.7078186,
+      "epoch": 1.2394723490613901,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.459064385235993e-06,
+      "loss": 1.35426998,
+      "memory(GiB)": 123.79,
+      "step": 48860,
+      "train_speed(iter/s)": 1.317219
+    },
+    {
+      "acc": 0.68557014,
+      "epoch": 1.2395991882293251,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.45806683641558e-06,
+      "loss": 1.40326443,
+      "memory(GiB)": 123.79,
+      "step": 48865,
+      "train_speed(iter/s)": 1.317235
+    },
+    {
+      "acc": 0.69523468,
+      "epoch": 1.2397260273972603,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.457069355416275e-06,
+      "loss": 1.30042553,
+      "memory(GiB)": 123.79,
+      "step": 48870,
+      "train_speed(iter/s)": 1.317251
+    },
+    {
+      "acc": 0.68580523,
+      "epoch": 1.2398528665651953,
+      "grad_norm": 3.25,
+      "learning_rate": 3.456071942281947e-06,
+      "loss": 1.40476246,
+      "memory(GiB)": 123.79,
+      "step": 48875,
+      "train_speed(iter/s)": 1.317267
+    },
+    {
+      "acc": 0.67203407,
+      "epoch": 1.2399797057331303,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.455074597056467e-06,
+      "loss": 1.46301556,
+      "memory(GiB)": 123.79,
+      "step": 48880,
+      "train_speed(iter/s)": 1.317283
+    },
+    {
+      "acc": 0.69376831,
+      "epoch": 1.2401065449010655,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.454077319783705e-06,
+      "loss": 1.29272938,
+      "memory(GiB)": 123.79,
+      "step": 48885,
+      "train_speed(iter/s)": 1.317299
+    },
+    {
+      "acc": 0.67079768,
+      "epoch": 1.2402333840690005,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.4530801105075257e-06,
+      "loss": 1.46030235,
+      "memory(GiB)": 123.79,
+      "step": 48890,
+      "train_speed(iter/s)": 1.317315
+    },
+    {
+      "acc": 0.68451476,
+      "epoch": 1.2403602232369355,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.4520829692717874e-06,
+      "loss": 1.42303352,
+      "memory(GiB)": 123.79,
+      "step": 48895,
+      "train_speed(iter/s)": 1.317332
+    },
+    {
+      "acc": 0.6790307,
+      "epoch": 1.2404870624048705,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.451085896120352e-06,
+      "loss": 1.40469227,
+      "memory(GiB)": 123.79,
+      "step": 48900,
+      "train_speed(iter/s)": 1.317348
+    },
+    {
+      "acc": 0.67860746,
+      "epoch": 1.2406139015728057,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.450088891097074e-06,
+      "loss": 1.39200821,
+      "memory(GiB)": 123.79,
+      "step": 48905,
+      "train_speed(iter/s)": 1.317364
+    },
+    {
+      "acc": 0.6871295,
+      "epoch": 1.2407407407407407,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.4490919542458085e-06,
+      "loss": 1.38279085,
+      "memory(GiB)": 123.79,
+      "step": 48910,
+      "train_speed(iter/s)": 1.31738
+    },
+    {
+      "acc": 0.66834469,
+      "epoch": 1.240867579908676,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.4480950856104002e-06,
+      "loss": 1.45705223,
+      "memory(GiB)": 123.79,
+      "step": 48915,
+      "train_speed(iter/s)": 1.317396
+    },
+    {
+      "acc": 0.68532,
+      "epoch": 1.240994419076611,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.4470982852347e-06,
+      "loss": 1.37796841,
+      "memory(GiB)": 123.79,
+      "step": 48920,
+      "train_speed(iter/s)": 1.317413
+    },
+    {
+      "acc": 0.67081165,
+      "epoch": 1.2411212582445459,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.44610155316255e-06,
+      "loss": 1.43677855,
+      "memory(GiB)": 123.79,
+      "step": 48925,
+      "train_speed(iter/s)": 1.317429
+    },
+    {
+      "acc": 0.69486113,
+      "epoch": 1.2412480974124809,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.4451048894377925e-06,
+      "loss": 1.36870937,
+      "memory(GiB)": 123.79,
+      "step": 48930,
+      "train_speed(iter/s)": 1.317445
+    },
+    {
+      "acc": 0.69189444,
+      "epoch": 1.241374936580416,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.444108294104264e-06,
+      "loss": 1.37422256,
+      "memory(GiB)": 123.79,
+      "step": 48935,
+      "train_speed(iter/s)": 1.317462
+    },
+    {
+      "acc": 0.69252834,
+      "epoch": 1.241501775748351,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.443111767205797e-06,
+      "loss": 1.40982714,
+      "memory(GiB)": 123.79,
+      "step": 48940,
+      "train_speed(iter/s)": 1.317478
+    },
+    {
+      "acc": 0.67697968,
+      "epoch": 1.241628614916286,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.442115308786227e-06,
+      "loss": 1.40287552,
+      "memory(GiB)": 123.79,
+      "step": 48945,
+      "train_speed(iter/s)": 1.317494
+    },
+    {
+      "acc": 0.68451777,
+      "epoch": 1.2417554540842213,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.4411189188893822e-06,
+      "loss": 1.4238987,
+      "memory(GiB)": 123.79,
+      "step": 48950,
+      "train_speed(iter/s)": 1.317511
+    },
+    {
+      "acc": 0.68585644,
+      "epoch": 1.2418822932521563,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.4401225975590867e-06,
+      "loss": 1.42983618,
+      "memory(GiB)": 123.79,
+      "step": 48955,
+      "train_speed(iter/s)": 1.317526
+    },
+    {
+      "acc": 0.66453829,
+      "epoch": 1.2420091324200913,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.439126344839163e-06,
+      "loss": 1.39545441,
+      "memory(GiB)": 123.79,
+      "step": 48960,
+      "train_speed(iter/s)": 1.317542
+    },
+    {
+      "acc": 0.69441805,
+      "epoch": 1.2421359715880265,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.438130160773431e-06,
+      "loss": 1.3894062,
+      "memory(GiB)": 123.79,
+      "step": 48965,
+      "train_speed(iter/s)": 1.317559
+    },
+    {
+      "acc": 0.68044853,
+      "epoch": 1.2422628107559615,
+      "grad_norm": 2.625,
+      "learning_rate": 3.43713404540571e-06,
+      "loss": 1.4593256,
+      "memory(GiB)": 123.79,
+      "step": 48970,
+      "train_speed(iter/s)": 1.317575
+    },
+    {
+      "acc": 0.68851986,
+      "epoch": 1.2423896499238964,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.4361379987798094e-06,
+      "loss": 1.43368626,
+      "memory(GiB)": 123.79,
+      "step": 48975,
+      "train_speed(iter/s)": 1.317591
+    },
+    {
+      "acc": 0.69399252,
+      "epoch": 1.2425164890918317,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.435142020939542e-06,
+      "loss": 1.35203304,
+      "memory(GiB)": 123.79,
+      "step": 48980,
+      "train_speed(iter/s)": 1.317607
+    },
+    {
+      "acc": 0.69257321,
+      "epoch": 1.2426433282597666,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.4341461119287144e-06,
+      "loss": 1.34114456,
+      "memory(GiB)": 123.79,
+      "step": 48985,
+      "train_speed(iter/s)": 1.317623
+    },
+    {
+      "acc": 0.69750586,
+      "epoch": 1.2427701674277016,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.433150271791135e-06,
+      "loss": 1.38676014,
+      "memory(GiB)": 123.79,
+      "step": 48990,
+      "train_speed(iter/s)": 1.31764
+    },
+    {
+      "acc": 0.69282665,
+      "epoch": 1.2428970065956366,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.432154500570599e-06,
+      "loss": 1.3377883,
+      "memory(GiB)": 123.79,
+      "step": 48995,
+      "train_speed(iter/s)": 1.317656
+    },
+    {
+      "acc": 0.69151878,
+      "epoch": 1.2430238457635718,
+      "grad_norm": 2.75,
+      "learning_rate": 3.431158798310909e-06,
+      "loss": 1.35900669,
+      "memory(GiB)": 123.79,
+      "step": 49000,
+      "train_speed(iter/s)": 1.317673
+    },
+    {
+      "epoch": 1.2430238457635718,
+      "eval_acc": 0.6750286620013908,
+      "eval_loss": 1.3564714193344116,
+      "eval_runtime": 70.196,
+      "eval_samples_per_second": 90.746,
+      "eval_steps_per_second": 22.694,
+      "step": 49000
+    },
+    {
+      "acc": 0.67792015,
+      "epoch": 1.2431506849315068,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.4301631650558588e-06,
+      "loss": 1.42772017,
+      "memory(GiB)": 123.79,
+      "step": 49005,
+      "train_speed(iter/s)": 1.314762
+    },
+    {
+      "acc": 0.67711039,
+      "epoch": 1.2432775240994418,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.4291676008492424e-06,
+      "loss": 1.3692667,
+      "memory(GiB)": 123.79,
+      "step": 49010,
+      "train_speed(iter/s)": 1.314779
+    },
+    {
+      "acc": 0.69878731,
+      "epoch": 1.243404363267377,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.428172105734848e-06,
+      "loss": 1.35373211,
+      "memory(GiB)": 123.79,
+      "step": 49015,
+      "train_speed(iter/s)": 1.314795
+    },
+    {
+      "acc": 0.69000711,
+      "epoch": 1.243531202435312,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.4271766797564608e-06,
+      "loss": 1.40897923,
+      "memory(GiB)": 123.79,
+      "step": 49020,
+      "train_speed(iter/s)": 1.314812
+    },
+    {
+      "acc": 0.68347311,
+      "epoch": 1.243658041603247,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.4261813229578665e-06,
+      "loss": 1.41115875,
+      "memory(GiB)": 123.79,
+      "step": 49025,
+      "train_speed(iter/s)": 1.314828
+    },
+    {
+      "acc": 0.70136113,
+      "epoch": 1.2437848807711822,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.425186035382846e-06,
+      "loss": 1.35383587,
+      "memory(GiB)": 123.79,
+      "step": 49030,
+      "train_speed(iter/s)": 1.314845
+    },
+    {
+      "acc": 0.6882247,
+      "epoch": 1.2439117199391172,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.4241908170751727e-06,
+      "loss": 1.3584281,
+      "memory(GiB)": 123.79,
+      "step": 49035,
+      "train_speed(iter/s)": 1.314861
+    },
+    {
+      "acc": 0.69195032,
+      "epoch": 1.2440385591070522,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.4231956680786217e-06,
+      "loss": 1.35145149,
+      "memory(GiB)": 123.79,
+      "step": 49040,
+      "train_speed(iter/s)": 1.314876
+    },
+    {
+      "acc": 0.68775339,
+      "epoch": 1.2441653982749874,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.422200588436967e-06,
+      "loss": 1.40792027,
+      "memory(GiB)": 123.79,
+      "step": 49045,
+      "train_speed(iter/s)": 1.314891
+    },
+    {
+      "acc": 0.67319016,
+      "epoch": 1.2442922374429224,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.4212055781939744e-06,
+      "loss": 1.43557644,
+      "memory(GiB)": 123.79,
+      "step": 49050,
+      "train_speed(iter/s)": 1.314907
+    },
+    {
+      "acc": 0.68416119,
+      "epoch": 1.2444190766108574,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.4202106373934085e-06,
+      "loss": 1.46757011,
+      "memory(GiB)": 123.79,
+      "step": 49055,
+      "train_speed(iter/s)": 1.314922
+    },
+    {
+      "acc": 0.69155436,
+      "epoch": 1.2445459157787924,
+      "grad_norm": 3.375,
+      "learning_rate": 3.4192157660790324e-06,
+      "loss": 1.4278986,
+      "memory(GiB)": 123.79,
+      "step": 49060,
+      "train_speed(iter/s)": 1.314937
+    },
+    {
+      "acc": 0.67673392,
+      "epoch": 1.2446727549467276,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.418220964294604e-06,
+      "loss": 1.42740135,
+      "memory(GiB)": 123.79,
+      "step": 49065,
+      "train_speed(iter/s)": 1.314952
+    },
+    {
+      "acc": 0.67560782,
+      "epoch": 1.2447995941146626,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.417226232083881e-06,
+      "loss": 1.42541714,
+      "memory(GiB)": 123.79,
+      "step": 49070,
+      "train_speed(iter/s)": 1.314968
+    },
+    {
+      "acc": 0.67996244,
+      "epoch": 1.2449264332825978,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.416231569490615e-06,
+      "loss": 1.4427,
+      "memory(GiB)": 123.79,
+      "step": 49075,
+      "train_speed(iter/s)": 1.314984
+    },
+    {
+      "acc": 0.71778326,
+      "epoch": 1.2450532724505328,
+      "grad_norm": 3.75,
+      "learning_rate": 3.4152369765585545e-06,
+      "loss": 1.22108116,
+      "memory(GiB)": 123.79,
+      "step": 49080,
+      "train_speed(iter/s)": 1.314995
+    },
+    {
+      "acc": 0.6905396,
+      "epoch": 1.2451801116184678,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.4142424533314474e-06,
+      "loss": 1.36725454,
+      "memory(GiB)": 123.79,
+      "step": 49085,
+      "train_speed(iter/s)": 1.315012
+    },
+    {
+      "acc": 0.67600641,
+      "epoch": 1.2453069507864027,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.4132479998530383e-06,
+      "loss": 1.43576536,
+      "memory(GiB)": 123.79,
+      "step": 49090,
+      "train_speed(iter/s)": 1.315028
+    },
+    {
+      "acc": 0.67962017,
+      "epoch": 1.245433789954338,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.4122536161670656e-06,
+      "loss": 1.38022785,
+      "memory(GiB)": 123.79,
+      "step": 49095,
+      "train_speed(iter/s)": 1.315044
+    },
+    {
+      "acc": 0.69517646,
+      "epoch": 1.245560629122273,
+      "grad_norm": 4.25,
+      "learning_rate": 3.411259302317267e-06,
+      "loss": 1.3441061,
+      "memory(GiB)": 123.79,
+      "step": 49100,
+      "train_speed(iter/s)": 1.31506
+    },
+    {
+      "acc": 0.68574448,
+      "epoch": 1.245687468290208,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.410265058347378e-06,
+      "loss": 1.3369339,
+      "memory(GiB)": 123.79,
+      "step": 49105,
+      "train_speed(iter/s)": 1.315075
+    },
+    {
+      "acc": 0.69658513,
+      "epoch": 1.2458143074581431,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.4092708843011303e-06,
+      "loss": 1.35721893,
+      "memory(GiB)": 123.79,
+      "step": 49110,
+      "train_speed(iter/s)": 1.315091
+    },
+    {
+      "acc": 0.67891707,
+      "epoch": 1.2459411466260781,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.4082767802222493e-06,
+      "loss": 1.4237277,
+      "memory(GiB)": 123.79,
+      "step": 49115,
+      "train_speed(iter/s)": 1.315106
+    },
+    {
+      "acc": 0.68618674,
+      "epoch": 1.2460679857940131,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.4072827461544635e-06,
+      "loss": 1.40007668,
+      "memory(GiB)": 123.79,
+      "step": 49120,
+      "train_speed(iter/s)": 1.315122
+    },
+    {
+      "acc": 0.69099917,
+      "epoch": 1.2461948249619483,
+      "grad_norm": 3.125,
+      "learning_rate": 3.4062887821414935e-06,
+      "loss": 1.36871252,
+      "memory(GiB)": 123.79,
+      "step": 49125,
+      "train_speed(iter/s)": 1.315138
+    },
+    {
+      "acc": 0.6854043,
+      "epoch": 1.2463216641298833,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.4052948882270585e-06,
+      "loss": 1.42477303,
+      "memory(GiB)": 123.79,
+      "step": 49130,
+      "train_speed(iter/s)": 1.315155
+    },
+    {
+      "acc": 0.68058248,
+      "epoch": 1.2464485032978183,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.404301064454873e-06,
+      "loss": 1.48579082,
+      "memory(GiB)": 123.79,
+      "step": 49135,
+      "train_speed(iter/s)": 1.31517
+    },
+    {
+      "acc": 0.6845542,
+      "epoch": 1.2465753424657535,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.4033073108686515e-06,
+      "loss": 1.39174747,
+      "memory(GiB)": 123.79,
+      "step": 49140,
+      "train_speed(iter/s)": 1.315187
+    },
+    {
+      "acc": 0.67544804,
+      "epoch": 1.2467021816336885,
+      "grad_norm": 3.0,
+      "learning_rate": 3.4023136275121026e-06,
+      "loss": 1.42041569,
+      "memory(GiB)": 123.79,
+      "step": 49145,
+      "train_speed(iter/s)": 1.315204
+    },
+    {
+      "acc": 0.67212653,
+      "epoch": 1.2468290208016235,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.401320014428935e-06,
+      "loss": 1.50306149,
+      "memory(GiB)": 123.79,
+      "step": 49150,
+      "train_speed(iter/s)": 1.315221
+    },
+    {
+      "acc": 0.6912549,
+      "epoch": 1.2469558599695585,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.40032647166285e-06,
+      "loss": 1.36702709,
+      "memory(GiB)": 123.79,
+      "step": 49155,
+      "train_speed(iter/s)": 1.315237
+    },
+    {
+      "acc": 0.69600682,
+      "epoch": 1.2470826991374937,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.3993329992575473e-06,
+      "loss": 1.34101696,
+      "memory(GiB)": 123.79,
+      "step": 49160,
+      "train_speed(iter/s)": 1.315254
+    },
+    {
+      "acc": 0.68353758,
+      "epoch": 1.2472095383054287,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.3983395972567277e-06,
+      "loss": 1.41810408,
+      "memory(GiB)": 123.79,
+      "step": 49165,
+      "train_speed(iter/s)": 1.31527
+    },
+    {
+      "acc": 0.69159355,
+      "epoch": 1.2473363774733637,
+      "grad_norm": 3.25,
+      "learning_rate": 3.397346265704084e-06,
+      "loss": 1.3530489,
+      "memory(GiB)": 123.79,
+      "step": 49170,
+      "train_speed(iter/s)": 1.315286
+    },
+    {
+      "acc": 0.70063477,
+      "epoch": 1.2474632166412989,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.396353004643306e-06,
+      "loss": 1.33307028,
+      "memory(GiB)": 123.79,
+      "step": 49175,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.68336153,
+      "epoch": 1.2475900558092339,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.3953598141180817e-06,
+      "loss": 1.39711761,
+      "memory(GiB)": 123.79,
+      "step": 49180,
+      "train_speed(iter/s)": 1.315318
+    },
+    {
+      "acc": 0.69652286,
+      "epoch": 1.2477168949771689,
+      "grad_norm": 5.125,
+      "learning_rate": 3.3943666941720978e-06,
+      "loss": 1.36140947,
+      "memory(GiB)": 123.79,
+      "step": 49185,
+      "train_speed(iter/s)": 1.315334
+    },
+    {
+      "acc": 0.68614082,
+      "epoch": 1.247843734145104,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.3933736448490363e-06,
+      "loss": 1.37711773,
+      "memory(GiB)": 123.79,
+      "step": 49190,
+      "train_speed(iter/s)": 1.315351
+    },
+    {
+      "acc": 0.68422503,
+      "epoch": 1.247970573313039,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.392380666192573e-06,
+      "loss": 1.39492865,
+      "memory(GiB)": 123.79,
+      "step": 49195,
+      "train_speed(iter/s)": 1.315366
+    },
+    {
+      "acc": 0.6799324,
+      "epoch": 1.248097412480974,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.391387758246386e-06,
+      "loss": 1.46893129,
+      "memory(GiB)": 123.79,
+      "step": 49200,
+      "train_speed(iter/s)": 1.315382
+    },
+    {
+      "acc": 0.70344915,
+      "epoch": 1.2482242516489093,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.3903949210541477e-06,
+      "loss": 1.30144396,
+      "memory(GiB)": 123.79,
+      "step": 49205,
+      "train_speed(iter/s)": 1.315398
+    },
+    {
+      "acc": 0.68692398,
+      "epoch": 1.2483510908168443,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.389402154659529e-06,
+      "loss": 1.41370029,
+      "memory(GiB)": 123.79,
+      "step": 49210,
+      "train_speed(iter/s)": 1.315414
+    },
+    {
+      "acc": 0.67696257,
+      "epoch": 1.2484779299847792,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.388409459106192e-06,
+      "loss": 1.44286909,
+      "memory(GiB)": 123.79,
+      "step": 49215,
+      "train_speed(iter/s)": 1.31543
+    },
+    {
+      "acc": 0.68312855,
+      "epoch": 1.2486047691527142,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.3874168344378024e-06,
+      "loss": 1.36281309,
+      "memory(GiB)": 123.79,
+      "step": 49220,
+      "train_speed(iter/s)": 1.315445
+    },
+    {
+      "acc": 0.68847036,
+      "epoch": 1.2487316083206494,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.38642428069802e-06,
+      "loss": 1.37341452,
+      "memory(GiB)": 123.79,
+      "step": 49225,
+      "train_speed(iter/s)": 1.315461
+    },
+    {
+      "acc": 0.70220447,
+      "epoch": 1.2488584474885844,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.385431797930503e-06,
+      "loss": 1.37936964,
+      "memory(GiB)": 123.79,
+      "step": 49230,
+      "train_speed(iter/s)": 1.315477
+    },
+    {
+      "acc": 0.67941513,
+      "epoch": 1.2489852866565196,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.3844393861789036e-06,
+      "loss": 1.45647945,
+      "memory(GiB)": 123.79,
+      "step": 49235,
+      "train_speed(iter/s)": 1.315492
+    },
+    {
+      "acc": 0.68445902,
+      "epoch": 1.2491121258244546,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.383447045486872e-06,
+      "loss": 1.43839855,
+      "memory(GiB)": 123.79,
+      "step": 49240,
+      "train_speed(iter/s)": 1.315507
+    },
+    {
+      "acc": 0.66726747,
+      "epoch": 1.2492389649923896,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.382454775898057e-06,
+      "loss": 1.45143681,
+      "memory(GiB)": 123.79,
+      "step": 49245,
+      "train_speed(iter/s)": 1.315523
+    },
+    {
+      "acc": 0.69290824,
+      "epoch": 1.2493658041603246,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.381462577456104e-06,
+      "loss": 1.38542709,
+      "memory(GiB)": 123.79,
+      "step": 49250,
+      "train_speed(iter/s)": 1.315538
+    },
+    {
+      "acc": 0.69522853,
+      "epoch": 1.2494926433282598,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.3804704502046527e-06,
+      "loss": 1.3591404,
+      "memory(GiB)": 123.79,
+      "step": 49255,
+      "train_speed(iter/s)": 1.315554
+    },
+    {
+      "acc": 0.69890437,
+      "epoch": 1.2496194824961948,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.3794783941873406e-06,
+      "loss": 1.34319324,
+      "memory(GiB)": 123.79,
+      "step": 49260,
+      "train_speed(iter/s)": 1.315569
+    },
+    {
+      "acc": 0.68694038,
+      "epoch": 1.2497463216641298,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.3784864094478044e-06,
+      "loss": 1.41043816,
+      "memory(GiB)": 123.79,
+      "step": 49265,
+      "train_speed(iter/s)": 1.315584
+    },
+    {
+      "acc": 0.68504171,
+      "epoch": 1.249873160832065,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.377494496029677e-06,
+      "loss": 1.37806759,
+      "memory(GiB)": 123.79,
+      "step": 49270,
+      "train_speed(iter/s)": 1.315596
+    },
+    {
+      "acc": 0.69536443,
+      "epoch": 1.25,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.3765026539765832e-06,
+      "loss": 1.39709892,
+      "memory(GiB)": 123.79,
+      "step": 49275,
+      "train_speed(iter/s)": 1.315612
+    },
+    {
+      "acc": 0.68437467,
+      "epoch": 1.250126839167935,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.375510883332152e-06,
+      "loss": 1.42909555,
+      "memory(GiB)": 123.79,
+      "step": 49280,
+      "train_speed(iter/s)": 1.315628
+    },
+    {
+      "acc": 0.67556663,
+      "epoch": 1.2502536783358702,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.3745191841400037e-06,
+      "loss": 1.41798277,
+      "memory(GiB)": 123.79,
+      "step": 49285,
+      "train_speed(iter/s)": 1.315643
+    },
+    {
+      "acc": 0.67978373,
+      "epoch": 1.2503805175038052,
+      "grad_norm": 2.75,
+      "learning_rate": 3.373527556443762e-06,
+      "loss": 1.43855362,
+      "memory(GiB)": 123.79,
+      "step": 49290,
+      "train_speed(iter/s)": 1.315659
+    },
+    {
+      "acc": 0.68621244,
+      "epoch": 1.2505073566717402,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.372536000287038e-06,
+      "loss": 1.3847641,
+      "memory(GiB)": 123.79,
+      "step": 49295,
+      "train_speed(iter/s)": 1.315675
+    },
+    {
+      "acc": 0.6961751,
+      "epoch": 1.2506341958396754,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.3715445157134474e-06,
+      "loss": 1.3661787,
+      "memory(GiB)": 123.79,
+      "step": 49300,
+      "train_speed(iter/s)": 1.315691
+    },
+    {
+      "acc": 0.68552957,
+      "epoch": 1.2507610350076104,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.370553102766598e-06,
+      "loss": 1.36154537,
+      "memory(GiB)": 123.79,
+      "step": 49305,
+      "train_speed(iter/s)": 1.315704
+    },
+    {
+      "acc": 0.67920513,
+      "epoch": 1.2508878741755454,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.369561761490101e-06,
+      "loss": 1.42817173,
+      "memory(GiB)": 123.79,
+      "step": 49310,
+      "train_speed(iter/s)": 1.315719
+    },
+    {
+      "acc": 0.68724494,
+      "epoch": 1.2510147133434804,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.3685704919275553e-06,
+      "loss": 1.45010891,
+      "memory(GiB)": 123.79,
+      "step": 49315,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.69933128,
+      "epoch": 1.2511415525114156,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.3675792941225625e-06,
+      "loss": 1.29750051,
+      "memory(GiB)": 123.79,
+      "step": 49320,
+      "train_speed(iter/s)": 1.315748
+    },
+    {
+      "acc": 0.69696641,
+      "epoch": 1.2512683916793506,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.3665881681187214e-06,
+      "loss": 1.41890068,
+      "memory(GiB)": 123.79,
+      "step": 49325,
+      "train_speed(iter/s)": 1.315764
+    },
+    {
+      "acc": 0.69687271,
+      "epoch": 1.2513952308472858,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.3655971139596265e-06,
+      "loss": 1.3841074,
+      "memory(GiB)": 123.79,
+      "step": 49330,
+      "train_speed(iter/s)": 1.315779
+    },
+    {
+      "acc": 0.67977571,
+      "epoch": 1.2515220700152208,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.3646061316888655e-06,
+      "loss": 1.42730408,
+      "memory(GiB)": 123.79,
+      "step": 49335,
+      "train_speed(iter/s)": 1.315795
+    },
+    {
+      "acc": 0.67661147,
+      "epoch": 1.2516489091831557,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.3636152213500295e-06,
+      "loss": 1.45919752,
+      "memory(GiB)": 123.79,
+      "step": 49340,
+      "train_speed(iter/s)": 1.31581
+    },
+    {
+      "acc": 0.68188739,
+      "epoch": 1.2517757483510907,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.362624382986702e-06,
+      "loss": 1.4208744,
+      "memory(GiB)": 123.79,
+      "step": 49345,
+      "train_speed(iter/s)": 1.315825
+    },
+    {
+      "acc": 0.69152241,
+      "epoch": 1.251902587519026,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.3616336166424653e-06,
+      "loss": 1.39279461,
+      "memory(GiB)": 123.79,
+      "step": 49350,
+      "train_speed(iter/s)": 1.31584
+    },
+    {
+      "acc": 0.69774532,
+      "epoch": 1.252029426686961,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.360642922360895e-06,
+      "loss": 1.30648708,
+      "memory(GiB)": 123.79,
+      "step": 49355,
+      "train_speed(iter/s)": 1.315856
+    },
+    {
+      "acc": 0.67897167,
+      "epoch": 1.252156265854896,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.3596523001855684e-06,
+      "loss": 1.49160557,
+      "memory(GiB)": 123.79,
+      "step": 49360,
+      "train_speed(iter/s)": 1.315871
+    },
+    {
+      "acc": 0.70405941,
+      "epoch": 1.2522831050228311,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.358661750160057e-06,
+      "loss": 1.29277849,
+      "memory(GiB)": 123.79,
+      "step": 49365,
+      "train_speed(iter/s)": 1.315886
+    },
+    {
+      "acc": 0.68667154,
+      "epoch": 1.2524099441907661,
+      "grad_norm": 2.625,
+      "learning_rate": 3.3576712723279326e-06,
+      "loss": 1.37205257,
+      "memory(GiB)": 123.79,
+      "step": 49370,
+      "train_speed(iter/s)": 1.315901
+    },
+    {
+      "acc": 0.67706509,
+      "epoch": 1.2525367833587011,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.3566808667327566e-06,
+      "loss": 1.4852293,
+      "memory(GiB)": 123.79,
+      "step": 49375,
+      "train_speed(iter/s)": 1.315917
+    },
+    {
+      "acc": 0.70028,
+      "epoch": 1.252663622526636,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.355690533418091e-06,
+      "loss": 1.37342491,
+      "memory(GiB)": 123.79,
+      "step": 49380,
+      "train_speed(iter/s)": 1.315932
+    },
+    {
+      "acc": 0.67300949,
+      "epoch": 1.2527904616945713,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.354700272427499e-06,
+      "loss": 1.43308258,
+      "memory(GiB)": 123.79,
+      "step": 49385,
+      "train_speed(iter/s)": 1.315948
+    },
+    {
+      "acc": 0.68939471,
+      "epoch": 1.2529173008625063,
+      "grad_norm": 3.375,
+      "learning_rate": 3.3537100838045356e-06,
+      "loss": 1.45250797,
+      "memory(GiB)": 123.79,
+      "step": 49390,
+      "train_speed(iter/s)": 1.315963
+    },
+    {
+      "acc": 0.68646564,
+      "epoch": 1.2530441400304415,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.3527199675927526e-06,
+      "loss": 1.43254871,
+      "memory(GiB)": 123.79,
+      "step": 49395,
+      "train_speed(iter/s)": 1.315979
+    },
+    {
+      "acc": 0.6760087,
+      "epoch": 1.2531709791983765,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.3517299238356982e-06,
+      "loss": 1.46244678,
+      "memory(GiB)": 123.79,
+      "step": 49400,
+      "train_speed(iter/s)": 1.315994
+    },
+    {
+      "acc": 0.6870316,
+      "epoch": 1.2532978183663115,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.3507399525769214e-06,
+      "loss": 1.37252064,
+      "memory(GiB)": 123.79,
+      "step": 49405,
+      "train_speed(iter/s)": 1.316009
+    },
+    {
+      "acc": 0.67345304,
+      "epoch": 1.2534246575342465,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.3497500538599664e-06,
+      "loss": 1.45911322,
+      "memory(GiB)": 123.79,
+      "step": 49410,
+      "train_speed(iter/s)": 1.316025
+    },
+    {
+      "acc": 0.68466167,
+      "epoch": 1.2535514967021817,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.34876022772837e-06,
+      "loss": 1.44073429,
+      "memory(GiB)": 123.79,
+      "step": 49415,
+      "train_speed(iter/s)": 1.31604
+    },
+    {
+      "acc": 0.68463607,
+      "epoch": 1.2536783358701167,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.347770474225672e-06,
+      "loss": 1.45353184,
+      "memory(GiB)": 123.79,
+      "step": 49420,
+      "train_speed(iter/s)": 1.316055
+    },
+    {
+      "acc": 0.68512292,
+      "epoch": 1.2538051750380519,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.3467807933954034e-06,
+      "loss": 1.38176384,
+      "memory(GiB)": 123.79,
+      "step": 49425,
+      "train_speed(iter/s)": 1.316071
+    },
+    {
+      "acc": 0.67883239,
+      "epoch": 1.2539320142059869,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.345791185281101e-06,
+      "loss": 1.39674072,
+      "memory(GiB)": 123.79,
+      "step": 49430,
+      "train_speed(iter/s)": 1.316086
+    },
+    {
+      "acc": 0.69432507,
+      "epoch": 1.2540588533739219,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.3448016499262836e-06,
+      "loss": 1.36602468,
+      "memory(GiB)": 123.79,
+      "step": 49435,
+      "train_speed(iter/s)": 1.316102
+    },
+    {
+      "acc": 0.68615198,
+      "epoch": 1.2541856925418569,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.3438121873744812e-06,
+      "loss": 1.4281126,
+      "memory(GiB)": 123.79,
+      "step": 49440,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.68872623,
+      "epoch": 1.254312531709792,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.342822797669212e-06,
+      "loss": 1.39673662,
+      "memory(GiB)": 123.79,
+      "step": 49445,
+      "train_speed(iter/s)": 1.316132
+    },
+    {
+      "acc": 0.68267102,
+      "epoch": 1.254439370877727,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.3418334808539966e-06,
+      "loss": 1.45211258,
+      "memory(GiB)": 123.79,
+      "step": 49450,
+      "train_speed(iter/s)": 1.316147
+    },
+    {
+      "acc": 0.66850281,
+      "epoch": 1.254566210045662,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.340844236972347e-06,
+      "loss": 1.46021404,
+      "memory(GiB)": 123.79,
+      "step": 49455,
+      "train_speed(iter/s)": 1.316163
+    },
+    {
+      "acc": 0.69172649,
+      "epoch": 1.2546930492135973,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.3398550660677748e-06,
+      "loss": 1.39709187,
+      "memory(GiB)": 123.79,
+      "step": 49460,
+      "train_speed(iter/s)": 1.316179
+    },
+    {
+      "acc": 0.67930079,
+      "epoch": 1.2548198883815322,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.3388659681837898e-06,
+      "loss": 1.40474014,
+      "memory(GiB)": 123.79,
+      "step": 49465,
+      "train_speed(iter/s)": 1.316194
+    },
+    {
+      "acc": 0.67451434,
+      "epoch": 1.2549467275494672,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.3378769433638965e-06,
+      "loss": 1.45593758,
+      "memory(GiB)": 123.79,
+      "step": 49470,
+      "train_speed(iter/s)": 1.31621
+    },
+    {
+      "acc": 0.68246875,
+      "epoch": 1.2550735667174022,
+      "grad_norm": 3.5,
+      "learning_rate": 3.336887991651595e-06,
+      "loss": 1.43471966,
+      "memory(GiB)": 123.79,
+      "step": 49475,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.67841415,
+      "epoch": 1.2552004058853374,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.3358991130903845e-06,
+      "loss": 1.41554842,
+      "memory(GiB)": 123.79,
+      "step": 49480,
+      "train_speed(iter/s)": 1.316241
+    },
+    {
+      "acc": 0.71508179,
+      "epoch": 1.2553272450532724,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.334910307723761e-06,
+      "loss": 1.31578808,
+      "memory(GiB)": 123.79,
+      "step": 49485,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.69575386,
+      "epoch": 1.2554540842212076,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.333921575595218e-06,
+      "loss": 1.37432938,
+      "memory(GiB)": 123.79,
+      "step": 49490,
+      "train_speed(iter/s)": 1.316273
+    },
+    {
+      "acc": 0.6669076,
+      "epoch": 1.2555809233891426,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.3329329167482404e-06,
+      "loss": 1.44430733,
+      "memory(GiB)": 123.79,
+      "step": 49495,
+      "train_speed(iter/s)": 1.316289
+    },
+    {
+      "acc": 0.69827366,
+      "epoch": 1.2557077625570776,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.331944331226317e-06,
+      "loss": 1.37205248,
+      "memory(GiB)": 123.79,
+      "step": 49500,
+      "train_speed(iter/s)": 1.316305
+    },
+    {
+      "acc": 0.69878912,
+      "epoch": 1.2558346017250126,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.330955819072928e-06,
+      "loss": 1.38755684,
+      "memory(GiB)": 123.79,
+      "step": 49505,
+      "train_speed(iter/s)": 1.31632
+    },
+    {
+      "acc": 0.71663675,
+      "epoch": 1.2559614408929478,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.329967380331556e-06,
+      "loss": 1.29256859,
+      "memory(GiB)": 123.79,
+      "step": 49510,
+      "train_speed(iter/s)": 1.316336
+    },
+    {
+      "acc": 0.68179989,
+      "epoch": 1.2560882800608828,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.3289790150456737e-06,
+      "loss": 1.39944725,
+      "memory(GiB)": 123.79,
+      "step": 49515,
+      "train_speed(iter/s)": 1.316352
+    },
+    {
+      "acc": 0.68528605,
+      "epoch": 1.2562151192288178,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.327990723258755e-06,
+      "loss": 1.39441462,
+      "memory(GiB)": 123.79,
+      "step": 49520,
+      "train_speed(iter/s)": 1.316368
+    },
+    {
+      "acc": 0.68423061,
+      "epoch": 1.256341958396753,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.3270025050142684e-06,
+      "loss": 1.36617241,
+      "memory(GiB)": 123.79,
+      "step": 49525,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.68005862,
+      "epoch": 1.256468797564688,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.3260143603556827e-06,
+      "loss": 1.37680693,
+      "memory(GiB)": 123.79,
+      "step": 49530,
+      "train_speed(iter/s)": 1.316399
+    },
+    {
+      "acc": 0.6728652,
+      "epoch": 1.256595636732623,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.3250262893264583e-06,
+      "loss": 1.4109663,
+      "memory(GiB)": 123.79,
+      "step": 49535,
+      "train_speed(iter/s)": 1.316415
+    },
+    {
+      "acc": 0.6784708,
+      "epoch": 1.256722475900558,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.3240382919700555e-06,
+      "loss": 1.39795609,
+      "memory(GiB)": 123.79,
+      "step": 49540,
+      "train_speed(iter/s)": 1.31643
+    },
+    {
+      "acc": 0.68597684,
+      "epoch": 1.2568493150684932,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.3230503683299316e-06,
+      "loss": 1.37250633,
+      "memory(GiB)": 123.79,
+      "step": 49545,
+      "train_speed(iter/s)": 1.316446
+    },
+    {
+      "acc": 0.69125919,
+      "epoch": 1.2569761542364282,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.3220625184495404e-06,
+      "loss": 1.39199438,
+      "memory(GiB)": 123.79,
+      "step": 49550,
+      "train_speed(iter/s)": 1.316462
+    },
+    {
+      "acc": 0.68209572,
+      "epoch": 1.2571029934043634,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.3210747423723293e-06,
+      "loss": 1.39421368,
+      "memory(GiB)": 123.79,
+      "step": 49555,
+      "train_speed(iter/s)": 1.316478
+    },
+    {
+      "acc": 0.68924193,
+      "epoch": 1.2572298325722984,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.3200870401417486e-06,
+      "loss": 1.42029667,
+      "memory(GiB)": 123.79,
+      "step": 49560,
+      "train_speed(iter/s)": 1.316494
+    },
+    {
+      "acc": 0.69270754,
+      "epoch": 1.2573566717402334,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.3190994118012387e-06,
+      "loss": 1.41978922,
+      "memory(GiB)": 123.79,
+      "step": 49565,
+      "train_speed(iter/s)": 1.316509
+    },
+    {
+      "acc": 0.68721056,
+      "epoch": 1.2574835109081683,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.318111857394244e-06,
+      "loss": 1.43536816,
+      "memory(GiB)": 123.79,
+      "step": 49570,
+      "train_speed(iter/s)": 1.316524
+    },
+    {
+      "acc": 0.69954157,
+      "epoch": 1.2576103500761036,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.3171243769641957e-06,
+      "loss": 1.34415255,
+      "memory(GiB)": 123.79,
+      "step": 49575,
+      "train_speed(iter/s)": 1.316539
+    },
+    {
+      "acc": 0.68009758,
+      "epoch": 1.2577371892440385,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.316136970554532e-06,
+      "loss": 1.40394878,
+      "memory(GiB)": 123.79,
+      "step": 49580,
+      "train_speed(iter/s)": 1.316555
+    },
+    {
+      "acc": 0.69119902,
+      "epoch": 1.2578640284119738,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.315149638208681e-06,
+      "loss": 1.40020704,
+      "memory(GiB)": 123.79,
+      "step": 49585,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.68815575,
+      "epoch": 1.2579908675799087,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.3141623799700738e-06,
+      "loss": 1.38637447,
+      "memory(GiB)": 123.79,
+      "step": 49590,
+      "train_speed(iter/s)": 1.316585
+    },
+    {
+      "acc": 0.67137346,
+      "epoch": 1.2581177067478437,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.3131751958821313e-06,
+      "loss": 1.47260494,
+      "memory(GiB)": 123.79,
+      "step": 49595,
+      "train_speed(iter/s)": 1.316601
+    },
+    {
+      "acc": 0.6931644,
+      "epoch": 1.2582445459157787,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.312188085988273e-06,
+      "loss": 1.34570637,
+      "memory(GiB)": 123.79,
+      "step": 49600,
+      "train_speed(iter/s)": 1.316616
+    },
+    {
+      "acc": 0.69595203,
+      "epoch": 1.258371385083714,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.311201050331919e-06,
+      "loss": 1.39326439,
+      "memory(GiB)": 123.79,
+      "step": 49605,
+      "train_speed(iter/s)": 1.316631
+    },
+    {
+      "acc": 0.69879208,
+      "epoch": 1.258498224251649,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.310214088956485e-06,
+      "loss": 1.36737108,
+      "memory(GiB)": 123.79,
+      "step": 49610,
+      "train_speed(iter/s)": 1.316646
+    },
+    {
+      "acc": 0.70502138,
+      "epoch": 1.258625063419584,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.3092272019053773e-06,
+      "loss": 1.26593418,
+      "memory(GiB)": 123.79,
+      "step": 49615,
+      "train_speed(iter/s)": 1.316662
+    },
+    {
+      "acc": 0.70848742,
+      "epoch": 1.2587519025875191,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.308240389222006e-06,
+      "loss": 1.29491739,
+      "memory(GiB)": 123.79,
+      "step": 49620,
+      "train_speed(iter/s)": 1.316675
+    },
+    {
+      "acc": 0.69726381,
+      "epoch": 1.2588787417554541,
+      "grad_norm": 3.0,
+      "learning_rate": 3.3072536509497762e-06,
+      "loss": 1.35185575,
+      "memory(GiB)": 123.79,
+      "step": 49625,
+      "train_speed(iter/s)": 1.31669
+    },
+    {
+      "acc": 0.6898694,
+      "epoch": 1.259005580923389,
+      "grad_norm": 3.0,
+      "learning_rate": 3.306266987132089e-06,
+      "loss": 1.3693944,
+      "memory(GiB)": 123.79,
+      "step": 49630,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.70139084,
+      "epoch": 1.259132420091324,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.3052803978123405e-06,
+      "loss": 1.33223972,
+      "memory(GiB)": 123.79,
+      "step": 49635,
+      "train_speed(iter/s)": 1.316721
+    },
+    {
+      "acc": 0.69427676,
+      "epoch": 1.2592592592592593,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.3042938830339264e-06,
+      "loss": 1.35460424,
+      "memory(GiB)": 123.79,
+      "step": 49640,
+      "train_speed(iter/s)": 1.316737
+    },
+    {
+      "acc": 0.68797317,
+      "epoch": 1.2593860984271943,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.303307442840238e-06,
+      "loss": 1.32739487,
+      "memory(GiB)": 123.79,
+      "step": 49645,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.69710093,
+      "epoch": 1.2595129375951295,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.302321077274666e-06,
+      "loss": 1.32194824,
+      "memory(GiB)": 123.79,
+      "step": 49650,
+      "train_speed(iter/s)": 1.316769
+    },
+    {
+      "acc": 0.66969156,
+      "epoch": 1.2596397767630645,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.30133478638059e-06,
+      "loss": 1.48328381,
+      "memory(GiB)": 123.79,
+      "step": 49655,
+      "train_speed(iter/s)": 1.316783
+    },
+    {
+      "acc": 0.68651099,
+      "epoch": 1.2597666159309995,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.300348570201395e-06,
+      "loss": 1.36187363,
+      "memory(GiB)": 123.79,
+      "step": 49660,
+      "train_speed(iter/s)": 1.316799
+    },
+    {
+      "acc": 0.67692852,
+      "epoch": 1.2598934550989345,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.299362428780457e-06,
+      "loss": 1.4051403,
+      "memory(GiB)": 123.79,
+      "step": 49665,
+      "train_speed(iter/s)": 1.316813
+    },
+    {
+      "acc": 0.70158544,
+      "epoch": 1.2600202942668697,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.298376362161154e-06,
+      "loss": 1.25096807,
+      "memory(GiB)": 123.79,
+      "step": 49670,
+      "train_speed(iter/s)": 1.316829
+    },
+    {
+      "acc": 0.69683275,
+      "epoch": 1.2601471334348047,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.297390370386856e-06,
+      "loss": 1.39224215,
+      "memory(GiB)": 123.79,
+      "step": 49675,
+      "train_speed(iter/s)": 1.316844
+    },
+    {
+      "acc": 0.67365303,
+      "epoch": 1.2602739726027397,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.2964044535009288e-06,
+      "loss": 1.44717693,
+      "memory(GiB)": 123.79,
+      "step": 49680,
+      "train_speed(iter/s)": 1.316859
+    },
+    {
+      "acc": 0.68957567,
+      "epoch": 1.2604008117706749,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.2954186115467412e-06,
+      "loss": 1.35678158,
+      "memory(GiB)": 123.79,
+      "step": 49685,
+      "train_speed(iter/s)": 1.316873
+    },
+    {
+      "acc": 0.67847433,
+      "epoch": 1.2605276509386099,
+      "grad_norm": 3.375,
+      "learning_rate": 3.2944328445676543e-06,
+      "loss": 1.42500868,
+      "memory(GiB)": 123.79,
+      "step": 49690,
+      "train_speed(iter/s)": 1.316889
+    },
+    {
+      "acc": 0.6726758,
+      "epoch": 1.2606544901065448,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.2934471526070254e-06,
+      "loss": 1.4611392,
+      "memory(GiB)": 123.79,
+      "step": 49695,
+      "train_speed(iter/s)": 1.316905
+    },
+    {
+      "acc": 0.69498291,
+      "epoch": 1.2607813292744798,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.2924615357082078e-06,
+      "loss": 1.36073694,
+      "memory(GiB)": 123.79,
+      "step": 49700,
+      "train_speed(iter/s)": 1.31692
+    },
+    {
+      "acc": 0.67643385,
+      "epoch": 1.260908168442415,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.2914759939145574e-06,
+      "loss": 1.44776897,
+      "memory(GiB)": 123.79,
+      "step": 49705,
+      "train_speed(iter/s)": 1.316936
+    },
+    {
+      "acc": 0.68110795,
+      "epoch": 1.26103500761035,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.2904905272694214e-06,
+      "loss": 1.41250334,
+      "memory(GiB)": 123.79,
+      "step": 49710,
+      "train_speed(iter/s)": 1.316951
+    },
+    {
+      "acc": 0.6916182,
+      "epoch": 1.2611618467782852,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.289505135816142e-06,
+      "loss": 1.36522617,
+      "memory(GiB)": 123.79,
+      "step": 49715,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.67769599,
+      "epoch": 1.2612886859462202,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.2885198195980653e-06,
+      "loss": 1.41503077,
+      "memory(GiB)": 123.79,
+      "step": 49720,
+      "train_speed(iter/s)": 1.316981
+    },
+    {
+      "acc": 0.68016977,
+      "epoch": 1.2614155251141552,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.287534578658527e-06,
+      "loss": 1.39468937,
+      "memory(GiB)": 123.79,
+      "step": 49725,
+      "train_speed(iter/s)": 1.316997
+    },
+    {
+      "acc": 0.68108549,
+      "epoch": 1.2615423642820902,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.2865494130408657e-06,
+      "loss": 1.39297371,
+      "memory(GiB)": 123.79,
+      "step": 49730,
+      "train_speed(iter/s)": 1.317012
+    },
+    {
+      "acc": 0.68715057,
+      "epoch": 1.2616692034500254,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.2855643227884097e-06,
+      "loss": 1.43165817,
+      "memory(GiB)": 123.79,
+      "step": 49735,
+      "train_speed(iter/s)": 1.317028
+    },
+    {
+      "acc": 0.69680839,
+      "epoch": 1.2617960426179604,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.28457930794449e-06,
+      "loss": 1.33851824,
+      "memory(GiB)": 123.79,
+      "step": 49740,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.70563779,
+      "epoch": 1.2619228817858956,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.283594368552429e-06,
+      "loss": 1.33497105,
+      "memory(GiB)": 123.79,
+      "step": 49745,
+      "train_speed(iter/s)": 1.317059
+    },
+    {
+      "acc": 0.68610735,
+      "epoch": 1.2620497209538306,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.282609504655554e-06,
+      "loss": 1.41489,
+      "memory(GiB)": 123.79,
+      "step": 49750,
+      "train_speed(iter/s)": 1.317074
+    },
+    {
+      "acc": 0.69251904,
+      "epoch": 1.2621765601217656,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.281624716297179e-06,
+      "loss": 1.35518761,
+      "memory(GiB)": 123.79,
+      "step": 49755,
+      "train_speed(iter/s)": 1.31709
+    },
+    {
+      "acc": 0.68263388,
+      "epoch": 1.2623033992897006,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.28064000352062e-06,
+      "loss": 1.40777321,
+      "memory(GiB)": 123.79,
+      "step": 49760,
+      "train_speed(iter/s)": 1.317105
+    },
+    {
+      "acc": 0.68216672,
+      "epoch": 1.2624302384576358,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.279655366369191e-06,
+      "loss": 1.37200346,
+      "memory(GiB)": 123.79,
+      "step": 49765,
+      "train_speed(iter/s)": 1.317121
+    },
+    {
+      "acc": 0.67183948,
+      "epoch": 1.2625570776255708,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.2786708048862e-06,
+      "loss": 1.43136072,
+      "memory(GiB)": 123.79,
+      "step": 49770,
+      "train_speed(iter/s)": 1.317137
+    },
+    {
+      "acc": 0.69471383,
+      "epoch": 1.2626839167935058,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.2776863191149517e-06,
+      "loss": 1.39729738,
+      "memory(GiB)": 123.79,
+      "step": 49775,
+      "train_speed(iter/s)": 1.317153
+    },
+    {
+      "acc": 0.69171982,
+      "epoch": 1.262810755961441,
+      "grad_norm": 3.75,
+      "learning_rate": 3.2767019090987483e-06,
+      "loss": 1.4181757,
+      "memory(GiB)": 123.79,
+      "step": 49780,
+      "train_speed(iter/s)": 1.317169
+    },
+    {
+      "acc": 0.67376394,
+      "epoch": 1.262937595129376,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.27571757488089e-06,
+      "loss": 1.47137547,
+      "memory(GiB)": 123.79,
+      "step": 49785,
+      "train_speed(iter/s)": 1.317185
+    },
+    {
+      "acc": 0.68956418,
+      "epoch": 1.263064434297311,
+      "grad_norm": 3.125,
+      "learning_rate": 3.274733316504672e-06,
+      "loss": 1.34969215,
+      "memory(GiB)": 123.79,
+      "step": 49790,
+      "train_speed(iter/s)": 1.317201
+    },
+    {
+      "acc": 0.69484224,
+      "epoch": 1.263191273465246,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.273749134013383e-06,
+      "loss": 1.31773376,
+      "memory(GiB)": 123.79,
+      "step": 49795,
+      "train_speed(iter/s)": 1.317217
+    },
+    {
+      "acc": 0.69045339,
+      "epoch": 1.2633181126331812,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.2727650274503154e-06,
+      "loss": 1.35714626,
+      "memory(GiB)": 123.79,
+      "step": 49800,
+      "train_speed(iter/s)": 1.317233
+    },
+    {
+      "acc": 0.68662038,
+      "epoch": 1.2634449518011162,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.2717809968587523e-06,
+      "loss": 1.36376133,
+      "memory(GiB)": 123.79,
+      "step": 49805,
+      "train_speed(iter/s)": 1.31725
+    },
+    {
+      "acc": 0.69004025,
+      "epoch": 1.2635717909690514,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.270797042281979e-06,
+      "loss": 1.35855379,
+      "memory(GiB)": 123.79,
+      "step": 49810,
+      "train_speed(iter/s)": 1.317265
+    },
+    {
+      "acc": 0.68146462,
+      "epoch": 1.2636986301369864,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.269813163763271e-06,
+      "loss": 1.47490654,
+      "memory(GiB)": 123.79,
+      "step": 49815,
+      "train_speed(iter/s)": 1.317282
+    },
+    {
+      "acc": 0.6875958,
+      "epoch": 1.2638254693049213,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.268829361345904e-06,
+      "loss": 1.3992981,
+      "memory(GiB)": 123.79,
+      "step": 49820,
+      "train_speed(iter/s)": 1.317293
+    },
+    {
+      "acc": 0.69493694,
+      "epoch": 1.2639523084728563,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.2678456350731526e-06,
+      "loss": 1.34825306,
+      "memory(GiB)": 123.79,
+      "step": 49825,
+      "train_speed(iter/s)": 1.317308
+    },
+    {
+      "acc": 0.70604444,
+      "epoch": 1.2640791476407915,
+      "grad_norm": 3.875,
+      "learning_rate": 3.266861984988283e-06,
+      "loss": 1.31684437,
+      "memory(GiB)": 123.79,
+      "step": 49830,
+      "train_speed(iter/s)": 1.317325
+    },
+    {
+      "acc": 0.69376049,
+      "epoch": 1.2642059868087265,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.2658784111345614e-06,
+      "loss": 1.41014614,
+      "memory(GiB)": 123.79,
+      "step": 49835,
+      "train_speed(iter/s)": 1.317341
+    },
+    {
+      "acc": 0.68500886,
+      "epoch": 1.2643328259766615,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.2648949135552482e-06,
+      "loss": 1.33638639,
+      "memory(GiB)": 123.79,
+      "step": 49840,
+      "train_speed(iter/s)": 1.317357
+    },
+    {
+      "acc": 0.68798833,
+      "epoch": 1.2644596651445967,
+      "grad_norm": 4.875,
+      "learning_rate": 3.2639114922936045e-06,
+      "loss": 1.41568718,
+      "memory(GiB)": 123.79,
+      "step": 49845,
+      "train_speed(iter/s)": 1.317372
+    },
+    {
+      "acc": 0.70513973,
+      "epoch": 1.2645865043125317,
+      "grad_norm": 4.15625,
+      "learning_rate": 3.2629281473928855e-06,
+      "loss": 1.30088396,
+      "memory(GiB)": 123.79,
+      "step": 49850,
+      "train_speed(iter/s)": 1.317389
+    },
+    {
+      "acc": 0.68833303,
+      "epoch": 1.2647133434804667,
+      "grad_norm": 2.59375,
+      "learning_rate": 3.26194487889634e-06,
+      "loss": 1.36789989,
+      "memory(GiB)": 123.79,
+      "step": 49855,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.69932609,
+      "epoch": 1.2648401826484017,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.2609616868472192e-06,
+      "loss": 1.32524757,
+      "memory(GiB)": 123.79,
+      "step": 49860,
+      "train_speed(iter/s)": 1.317421
+    },
+    {
+      "acc": 0.6896771,
+      "epoch": 1.264967021816337,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.259978571288767e-06,
+      "loss": 1.33707533,
+      "memory(GiB)": 123.79,
+      "step": 49865,
+      "train_speed(iter/s)": 1.317438
+    },
+    {
+      "acc": 0.68024864,
+      "epoch": 1.265093860984272,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.2589955322642293e-06,
+      "loss": 1.4830986,
+      "memory(GiB)": 123.79,
+      "step": 49870,
+      "train_speed(iter/s)": 1.317454
+    },
+    {
+      "acc": 0.67946081,
+      "epoch": 1.2652207001522071,
+      "grad_norm": 2.40625,
+      "learning_rate": 3.2580125698168376e-06,
+      "loss": 1.37173004,
+      "memory(GiB)": 123.79,
+      "step": 49875,
+      "train_speed(iter/s)": 1.317469
+    },
+    {
+      "acc": 0.6831851,
+      "epoch": 1.265347539320142,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2570296839898314e-06,
+      "loss": 1.44209385,
+      "memory(GiB)": 123.79,
+      "step": 49880,
+      "train_speed(iter/s)": 1.317485
+    },
+    {
+      "acc": 0.69068332,
+      "epoch": 1.265474378488077,
+      "grad_norm": 3.671875,
+      "learning_rate": 3.2560468748264405e-06,
+      "loss": 1.35473175,
+      "memory(GiB)": 123.79,
+      "step": 49885,
+      "train_speed(iter/s)": 1.317502
+    },
+    {
+      "acc": 0.67624679,
+      "epoch": 1.265601217656012,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.2550641423698965e-06,
+      "loss": 1.40610504,
+      "memory(GiB)": 123.79,
+      "step": 49890,
+      "train_speed(iter/s)": 1.317518
+    },
+    {
+      "acc": 0.67327042,
+      "epoch": 1.2657280568239473,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.2540814866634206e-06,
+      "loss": 1.45098495,
+      "memory(GiB)": 123.79,
+      "step": 49895,
+      "train_speed(iter/s)": 1.317534
+    },
+    {
+      "acc": 0.68298707,
+      "epoch": 1.2658548959918823,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.2530989077502355e-06,
+      "loss": 1.36742191,
+      "memory(GiB)": 123.79,
+      "step": 49900,
+      "train_speed(iter/s)": 1.317548
+    },
+    {
+      "acc": 0.70411358,
+      "epoch": 1.2659817351598175,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.252116405673561e-06,
+      "loss": 1.30304146,
+      "memory(GiB)": 123.79,
+      "step": 49905,
+      "train_speed(iter/s)": 1.317564
+    },
+    {
+      "acc": 0.69563093,
+      "epoch": 1.2661085743277525,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.2511339804766107e-06,
+      "loss": 1.33266907,
+      "memory(GiB)": 123.79,
+      "step": 49910,
+      "train_speed(iter/s)": 1.31758
+    },
+    {
+      "acc": 0.70208406,
+      "epoch": 1.2662354134956875,
+      "grad_norm": 3.375,
+      "learning_rate": 3.250151632202596e-06,
+      "loss": 1.36240168,
+      "memory(GiB)": 123.79,
+      "step": 49915,
+      "train_speed(iter/s)": 1.317596
+    },
+    {
+      "acc": 0.69488149,
+      "epoch": 1.2663622526636225,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.249169360894724e-06,
+      "loss": 1.39081926,
+      "memory(GiB)": 123.79,
+      "step": 49920,
+      "train_speed(iter/s)": 1.317611
+    },
+    {
+      "acc": 0.68222566,
+      "epoch": 1.2664890918315577,
+      "grad_norm": 4.375,
+      "learning_rate": 3.2481871665962006e-06,
+      "loss": 1.45608501,
+      "memory(GiB)": 123.79,
+      "step": 49925,
+      "train_speed(iter/s)": 1.317627
+    },
+    {
+      "acc": 0.70200748,
+      "epoch": 1.2666159309994927,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.2472050493502282e-06,
+      "loss": 1.33430748,
+      "memory(GiB)": 123.79,
+      "step": 49930,
+      "train_speed(iter/s)": 1.317643
+    },
+    {
+      "acc": 0.6981297,
+      "epoch": 1.2667427701674276,
+      "grad_norm": 3.625,
+      "learning_rate": 3.2462230092000017e-06,
+      "loss": 1.33379326,
+      "memory(GiB)": 123.79,
+      "step": 49935,
+      "train_speed(iter/s)": 1.317658
+    },
+    {
+      "acc": 0.6901206,
+      "epoch": 1.2668696093353629,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.2452410461887184e-06,
+      "loss": 1.35601864,
+      "memory(GiB)": 123.79,
+      "step": 49940,
+      "train_speed(iter/s)": 1.317672
+    },
+    {
+      "acc": 0.69675188,
+      "epoch": 1.2669964485032978,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.244259160359567e-06,
+      "loss": 1.34037209,
+      "memory(GiB)": 123.79,
+      "step": 49945,
+      "train_speed(iter/s)": 1.317687
+    },
+    {
+      "acc": 0.68958364,
+      "epoch": 1.2671232876712328,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.2432773517557385e-06,
+      "loss": 1.40222511,
+      "memory(GiB)": 123.79,
+      "step": 49950,
+      "train_speed(iter/s)": 1.317701
+    },
+    {
+      "acc": 0.68883209,
+      "epoch": 1.2672501268391678,
+      "grad_norm": 2.625,
+      "learning_rate": 3.2422956204204147e-06,
+      "loss": 1.37415295,
+      "memory(GiB)": 123.79,
+      "step": 49955,
+      "train_speed(iter/s)": 1.317717
+    },
+    {
+      "acc": 0.6922627,
+      "epoch": 1.267376966007103,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.2413139663967763e-06,
+      "loss": 1.42112293,
+      "memory(GiB)": 123.79,
+      "step": 49960,
+      "train_speed(iter/s)": 1.317733
+    },
+    {
+      "acc": 0.70452404,
+      "epoch": 1.267503805175038,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.2403323897280013e-06,
+      "loss": 1.2918335,
+      "memory(GiB)": 123.79,
+      "step": 49965,
+      "train_speed(iter/s)": 1.317748
+    },
+    {
+      "acc": 0.67864318,
+      "epoch": 1.2676306443429732,
+      "grad_norm": 3.5,
+      "learning_rate": 3.2393508904572663e-06,
+      "loss": 1.34728804,
+      "memory(GiB)": 123.79,
+      "step": 49970,
+      "train_speed(iter/s)": 1.317763
+    },
+    {
+      "acc": 0.67950573,
+      "epoch": 1.2677574835109082,
+      "grad_norm": 2.5,
+      "learning_rate": 3.2383694686277382e-06,
+      "loss": 1.42837753,
+      "memory(GiB)": 123.79,
+      "step": 49975,
+      "train_speed(iter/s)": 1.317777
+    },
+    {
+      "acc": 0.68147507,
+      "epoch": 1.2678843226788432,
+      "grad_norm": 2.875,
+      "learning_rate": 3.2373881242825857e-06,
+      "loss": 1.40407248,
+      "memory(GiB)": 123.79,
+      "step": 49980,
+      "train_speed(iter/s)": 1.317792
+    },
+    {
+      "acc": 0.69365258,
+      "epoch": 1.2680111618467782,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.236406857464973e-06,
+      "loss": 1.35703211,
+      "memory(GiB)": 123.79,
+      "step": 49985,
+      "train_speed(iter/s)": 1.317807
+    },
+    {
+      "acc": 0.6825757,
+      "epoch": 1.2681380010147134,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.235425668218063e-06,
+      "loss": 1.39851265,
+      "memory(GiB)": 123.79,
+      "step": 49990,
+      "train_speed(iter/s)": 1.317823
+    },
+    {
+      "acc": 0.67411141,
+      "epoch": 1.2682648401826484,
+      "grad_norm": 2.75,
+      "learning_rate": 3.234444556585007e-06,
+      "loss": 1.39433117,
+      "memory(GiB)": 123.79,
+      "step": 49995,
+      "train_speed(iter/s)": 1.317838
+    },
+    {
+      "acc": 0.68292556,
+      "epoch": 1.2683916793505834,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.233463522608964e-06,
+      "loss": 1.37989254,
+      "memory(GiB)": 123.79,
+      "step": 50000,
+      "train_speed(iter/s)": 1.317853
+    },
+    {
+      "epoch": 1.2683916793505834,
+      "eval_acc": 0.675053721674738,
+      "eval_loss": 1.3563916683197021,
+      "eval_runtime": 70.1174,
+      "eval_samples_per_second": 90.848,
+      "eval_steps_per_second": 22.719,
+      "step": 50000
+    },
+    {
+      "acc": 0.70686474,
+      "epoch": 1.2685185185185186,
+      "grad_norm": 3.5,
+      "learning_rate": 3.2324825663330818e-06,
+      "loss": 1.37456703,
+      "memory(GiB)": 123.79,
+      "step": 50005,
+      "train_speed(iter/s)": 1.314997
+    },
+    {
+      "acc": 0.69431629,
+      "epoch": 1.2686453576864536,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.231501687800509e-06,
+      "loss": 1.32858257,
+      "memory(GiB)": 123.79,
+      "step": 50010,
+      "train_speed(iter/s)": 1.315012
+    },
+    {
+      "acc": 0.68566046,
+      "epoch": 1.2687721968543886,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.2305208870543857e-06,
+      "loss": 1.37563677,
+      "memory(GiB)": 123.79,
+      "step": 50015,
+      "train_speed(iter/s)": 1.315027
+    },
+    {
+      "acc": 0.69429693,
+      "epoch": 1.2688990360223236,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.2295401641378544e-06,
+      "loss": 1.37805653,
+      "memory(GiB)": 123.79,
+      "step": 50020,
+      "train_speed(iter/s)": 1.315041
+    },
+    {
+      "acc": 0.69102297,
+      "epoch": 1.2690258751902588,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.2285595190940513e-06,
+      "loss": 1.39908895,
+      "memory(GiB)": 123.79,
+      "step": 50025,
+      "train_speed(iter/s)": 1.315056
+    },
+    {
+      "acc": 0.68174133,
+      "epoch": 1.2691527143581938,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.2275789519661103e-06,
+      "loss": 1.49699402,
+      "memory(GiB)": 123.79,
+      "step": 50030,
+      "train_speed(iter/s)": 1.315071
+    },
+    {
+      "acc": 0.68851724,
+      "epoch": 1.269279553526129,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.2265984627971595e-06,
+      "loss": 1.43530216,
+      "memory(GiB)": 123.79,
+      "step": 50035,
+      "train_speed(iter/s)": 1.315085
+    },
+    {
+      "acc": 0.67727652,
+      "epoch": 1.269406392694064,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.225618051630326e-06,
+      "loss": 1.44272289,
+      "memory(GiB)": 123.79,
+      "step": 50040,
+      "train_speed(iter/s)": 1.3151
+    },
+    {
+      "acc": 0.69443502,
+      "epoch": 1.269533231861999,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.2246377185087325e-06,
+      "loss": 1.40185528,
+      "memory(GiB)": 123.79,
+      "step": 50045,
+      "train_speed(iter/s)": 1.315114
+    },
+    {
+      "acc": 0.67718077,
+      "epoch": 1.269660071029934,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.2236574634755003e-06,
+      "loss": 1.45323639,
+      "memory(GiB)": 123.79,
+      "step": 50050,
+      "train_speed(iter/s)": 1.315128
+    },
+    {
+      "acc": 0.67408953,
+      "epoch": 1.2697869101978692,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.222677286573742e-06,
+      "loss": 1.497682,
+      "memory(GiB)": 123.79,
+      "step": 50055,
+      "train_speed(iter/s)": 1.315143
+    },
+    {
+      "acc": 0.68370695,
+      "epoch": 1.2699137493658041,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.221697187846571e-06,
+      "loss": 1.42055655,
+      "memory(GiB)": 123.79,
+      "step": 50060,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.69747787,
+      "epoch": 1.2700405885337394,
+      "grad_norm": 3.875,
+      "learning_rate": 3.2207171673370984e-06,
+      "loss": 1.40705357,
+      "memory(GiB)": 123.79,
+      "step": 50065,
+      "train_speed(iter/s)": 1.315173
+    },
+    {
+      "acc": 0.68376727,
+      "epoch": 1.2701674277016743,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2197372250884295e-06,
+      "loss": 1.39879513,
+      "memory(GiB)": 123.79,
+      "step": 50070,
+      "train_speed(iter/s)": 1.315187
+    },
+    {
+      "acc": 0.68568349,
+      "epoch": 1.2702942668696093,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.218757361143664e-06,
+      "loss": 1.4099102,
+      "memory(GiB)": 123.79,
+      "step": 50075,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.69270916,
+      "epoch": 1.2704211060375443,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.2177775755459034e-06,
+      "loss": 1.41359444,
+      "memory(GiB)": 123.79,
+      "step": 50080,
+      "train_speed(iter/s)": 1.315217
+    },
+    {
+      "acc": 0.68150573,
+      "epoch": 1.2705479452054795,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.216797868338241e-06,
+      "loss": 1.40406466,
+      "memory(GiB)": 123.79,
+      "step": 50085,
+      "train_speed(iter/s)": 1.315231
+    },
+    {
+      "acc": 0.69292021,
+      "epoch": 1.2706747843734145,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.215818239563773e-06,
+      "loss": 1.3891243,
+      "memory(GiB)": 123.79,
+      "step": 50090,
+      "train_speed(iter/s)": 1.315245
+    },
+    {
+      "acc": 0.69773803,
+      "epoch": 1.2708016235413495,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.2148386892655814e-06,
+      "loss": 1.32612867,
+      "memory(GiB)": 123.79,
+      "step": 50095,
+      "train_speed(iter/s)": 1.31526
+    },
+    {
+      "acc": 0.68598528,
+      "epoch": 1.2709284627092847,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.2138592174867556e-06,
+      "loss": 1.44817114,
+      "memory(GiB)": 123.79,
+      "step": 50100,
+      "train_speed(iter/s)": 1.315271
+    },
+    {
+      "acc": 0.69730225,
+      "epoch": 1.2710553018772197,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.2128798242703745e-06,
+      "loss": 1.37429237,
+      "memory(GiB)": 123.79,
+      "step": 50105,
+      "train_speed(iter/s)": 1.315285
+    },
+    {
+      "acc": 0.69153948,
+      "epoch": 1.2711821410451547,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.2119005096595203e-06,
+      "loss": 1.40134735,
+      "memory(GiB)": 123.79,
+      "step": 50110,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.69769955,
+      "epoch": 1.2713089802130897,
+      "grad_norm": 2.625,
+      "learning_rate": 3.2109212736972636e-06,
+      "loss": 1.37647276,
+      "memory(GiB)": 123.79,
+      "step": 50115,
+      "train_speed(iter/s)": 1.315315
+    },
+    {
+      "acc": 0.69678445,
+      "epoch": 1.271435819381025,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.2099421164266758e-06,
+      "loss": 1.32535849,
+      "memory(GiB)": 123.79,
+      "step": 50120,
+      "train_speed(iter/s)": 1.31533
+    },
+    {
+      "acc": 0.68714452,
+      "epoch": 1.27156265854896,
+      "grad_norm": 4.125,
+      "learning_rate": 3.2089630378908264e-06,
+      "loss": 1.40088806,
+      "memory(GiB)": 123.79,
+      "step": 50125,
+      "train_speed(iter/s)": 1.315345
+    },
+    {
+      "acc": 0.69626708,
+      "epoch": 1.271689497716895,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.207984038132781e-06,
+      "loss": 1.34399204,
+      "memory(GiB)": 123.79,
+      "step": 50130,
+      "train_speed(iter/s)": 1.31536
+    },
+    {
+      "acc": 0.6900857,
+      "epoch": 1.27181633688483,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.2070051171955966e-06,
+      "loss": 1.40446854,
+      "memory(GiB)": 123.79,
+      "step": 50135,
+      "train_speed(iter/s)": 1.315374
+    },
+    {
+      "acc": 0.70573063,
+      "epoch": 1.271943176052765,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.206026275122332e-06,
+      "loss": 1.28745995,
+      "memory(GiB)": 123.79,
+      "step": 50140,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.70785484,
+      "epoch": 1.2720700152207,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.205047511956042e-06,
+      "loss": 1.31126289,
+      "memory(GiB)": 123.79,
+      "step": 50145,
+      "train_speed(iter/s)": 1.315404
+    },
+    {
+      "acc": 0.68193798,
+      "epoch": 1.2721968543886353,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.204068827739777e-06,
+      "loss": 1.48745937,
+      "memory(GiB)": 123.79,
+      "step": 50150,
+      "train_speed(iter/s)": 1.31542
+    },
+    {
+      "acc": 0.68512659,
+      "epoch": 1.2723236935565703,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.2030902225165814e-06,
+      "loss": 1.40879574,
+      "memory(GiB)": 123.79,
+      "step": 50155,
+      "train_speed(iter/s)": 1.315434
+    },
+    {
+      "acc": 0.69656782,
+      "epoch": 1.2724505327245053,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.2021116963295016e-06,
+      "loss": 1.36458559,
+      "memory(GiB)": 123.79,
+      "step": 50160,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.68290195,
+      "epoch": 1.2725773718924405,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.2011332492215753e-06,
+      "loss": 1.38791714,
+      "memory(GiB)": 123.79,
+      "step": 50165,
+      "train_speed(iter/s)": 1.315463
+    },
+    {
+      "acc": 0.68982086,
+      "epoch": 1.2727042110603755,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.200154881235842e-06,
+      "loss": 1.38322821,
+      "memory(GiB)": 123.79,
+      "step": 50170,
+      "train_speed(iter/s)": 1.315478
+    },
+    {
+      "acc": 0.69578071,
+      "epoch": 1.2728310502283104,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.1991765924153316e-06,
+      "loss": 1.39011803,
+      "memory(GiB)": 123.79,
+      "step": 50175,
+      "train_speed(iter/s)": 1.315493
+    },
+    {
+      "acc": 0.68190804,
+      "epoch": 1.2729578893962454,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.198198382803075e-06,
+      "loss": 1.37884865,
+      "memory(GiB)": 123.79,
+      "step": 50180,
+      "train_speed(iter/s)": 1.315508
+    },
+    {
+      "acc": 0.67135649,
+      "epoch": 1.2730847285641806,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.197220252442097e-06,
+      "loss": 1.40162392,
+      "memory(GiB)": 123.79,
+      "step": 50185,
+      "train_speed(iter/s)": 1.315522
+    },
+    {
+      "acc": 0.6804193,
+      "epoch": 1.2732115677321156,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.1962422013754237e-06,
+      "loss": 1.40489941,
+      "memory(GiB)": 123.79,
+      "step": 50190,
+      "train_speed(iter/s)": 1.315537
+    },
+    {
+      "acc": 0.7061182,
+      "epoch": 1.2733384069000508,
+      "grad_norm": 3.75,
+      "learning_rate": 3.1952642296460696e-06,
+      "loss": 1.32408504,
+      "memory(GiB)": 123.79,
+      "step": 50195,
+      "train_speed(iter/s)": 1.315551
+    },
+    {
+      "acc": 0.66987567,
+      "epoch": 1.2734652460679858,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.194286337297051e-06,
+      "loss": 1.45099392,
+      "memory(GiB)": 123.79,
+      "step": 50200,
+      "train_speed(iter/s)": 1.315566
+    },
+    {
+      "acc": 0.6824163,
+      "epoch": 1.2735920852359208,
+      "grad_norm": 4.875,
+      "learning_rate": 3.1933085243713837e-06,
+      "loss": 1.32153425,
+      "memory(GiB)": 123.79,
+      "step": 50205,
+      "train_speed(iter/s)": 1.315582
+    },
+    {
+      "acc": 0.69767704,
+      "epoch": 1.2737189244038558,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.1923307909120736e-06,
+      "loss": 1.43704853,
+      "memory(GiB)": 123.79,
+      "step": 50210,
+      "train_speed(iter/s)": 1.31559
+    },
+    {
+      "acc": 0.68600941,
+      "epoch": 1.273845763571791,
+      "grad_norm": 2.875,
+      "learning_rate": 3.1913531369621243e-06,
+      "loss": 1.38604403,
+      "memory(GiB)": 123.79,
+      "step": 50215,
+      "train_speed(iter/s)": 1.315604
+    },
+    {
+      "acc": 0.69032316,
+      "epoch": 1.273972602739726,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.19037556256454e-06,
+      "loss": 1.39983273,
+      "memory(GiB)": 123.79,
+      "step": 50220,
+      "train_speed(iter/s)": 1.315618
+    },
+    {
+      "acc": 0.67998986,
+      "epoch": 1.2740994419076612,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.189398067762318e-06,
+      "loss": 1.43501863,
+      "memory(GiB)": 123.79,
+      "step": 50225,
+      "train_speed(iter/s)": 1.315633
+    },
+    {
+      "acc": 0.68600922,
+      "epoch": 1.2742262810755962,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.1884206525984535e-06,
+      "loss": 1.41212101,
+      "memory(GiB)": 123.79,
+      "step": 50230,
+      "train_speed(iter/s)": 1.315647
+    },
+    {
+      "acc": 0.68334007,
+      "epoch": 1.2743531202435312,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.1874433171159348e-06,
+      "loss": 1.41574345,
+      "memory(GiB)": 123.79,
+      "step": 50235,
+      "train_speed(iter/s)": 1.315662
+    },
+    {
+      "acc": 0.68703623,
+      "epoch": 1.2744799594114662,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.1864660613577523e-06,
+      "loss": 1.4326107,
+      "memory(GiB)": 123.79,
+      "step": 50240,
+      "train_speed(iter/s)": 1.315676
+    },
+    {
+      "acc": 0.6919662,
+      "epoch": 1.2746067985794014,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.185488885366889e-06,
+      "loss": 1.34295444,
+      "memory(GiB)": 123.79,
+      "step": 50245,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.68592792,
+      "epoch": 1.2747336377473364,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.1845117891863274e-06,
+      "loss": 1.3444046,
+      "memory(GiB)": 123.79,
+      "step": 50250,
+      "train_speed(iter/s)": 1.315704
+    },
+    {
+      "acc": 0.68605046,
+      "epoch": 1.2748604769152714,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.1835347728590414e-06,
+      "loss": 1.38392534,
+      "memory(GiB)": 123.79,
+      "step": 50255,
+      "train_speed(iter/s)": 1.315717
+    },
+    {
+      "acc": 0.68020611,
+      "epoch": 1.2749873160832066,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.1825578364280064e-06,
+      "loss": 1.34833775,
+      "memory(GiB)": 123.79,
+      "step": 50260,
+      "train_speed(iter/s)": 1.315732
+    },
+    {
+      "acc": 0.69401608,
+      "epoch": 1.2751141552511416,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.181580979936192e-06,
+      "loss": 1.37993956,
+      "memory(GiB)": 123.79,
+      "step": 50265,
+      "train_speed(iter/s)": 1.315747
+    },
+    {
+      "acc": 0.67708979,
+      "epoch": 1.2752409944190766,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.1806042034265656e-06,
+      "loss": 1.41852131,
+      "memory(GiB)": 123.79,
+      "step": 50270,
+      "train_speed(iter/s)": 1.315763
+    },
+    {
+      "acc": 0.68285542,
+      "epoch": 1.2753678335870116,
+      "grad_norm": 3.25,
+      "learning_rate": 3.179627506942089e-06,
+      "loss": 1.3956028,
+      "memory(GiB)": 123.79,
+      "step": 50275,
+      "train_speed(iter/s)": 1.315778
+    },
+    {
+      "acc": 0.69682293,
+      "epoch": 1.2754946727549468,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.17865089052572e-06,
+      "loss": 1.40759315,
+      "memory(GiB)": 123.79,
+      "step": 50280,
+      "train_speed(iter/s)": 1.315793
+    },
+    {
+      "acc": 0.69649534,
+      "epoch": 1.2756215119228818,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.177674354220418e-06,
+      "loss": 1.33130341,
+      "memory(GiB)": 123.79,
+      "step": 50285,
+      "train_speed(iter/s)": 1.315807
+    },
+    {
+      "acc": 0.69491072,
+      "epoch": 1.275748351090817,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.1766978980691355e-06,
+      "loss": 1.37753735,
+      "memory(GiB)": 123.79,
+      "step": 50290,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.69962611,
+      "epoch": 1.275875190258752,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.1757215221148173e-06,
+      "loss": 1.39538317,
+      "memory(GiB)": 123.79,
+      "step": 50295,
+      "train_speed(iter/s)": 1.315836
+    },
+    {
+      "acc": 0.68635807,
+      "epoch": 1.276002029426687,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.174745226400413e-06,
+      "loss": 1.39155893,
+      "memory(GiB)": 123.79,
+      "step": 50300,
+      "train_speed(iter/s)": 1.315851
+    },
+    {
+      "acc": 0.69474607,
+      "epoch": 1.276128868594622,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.1737690109688613e-06,
+      "loss": 1.3515008,
+      "memory(GiB)": 123.79,
+      "step": 50305,
+      "train_speed(iter/s)": 1.315866
+    },
+    {
+      "acc": 0.69400716,
+      "epoch": 1.2762557077625571,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.1727928758631054e-06,
+      "loss": 1.35591736,
+      "memory(GiB)": 123.79,
+      "step": 50310,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.68430982,
+      "epoch": 1.2763825469304921,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.1718168211260734e-06,
+      "loss": 1.44450932,
+      "memory(GiB)": 123.79,
+      "step": 50315,
+      "train_speed(iter/s)": 1.315895
+    },
+    {
+      "acc": 0.67636027,
+      "epoch": 1.2765093860984271,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.1708408468007014e-06,
+      "loss": 1.45370903,
+      "memory(GiB)": 123.79,
+      "step": 50320,
+      "train_speed(iter/s)": 1.315909
+    },
+    {
+      "acc": 0.68153772,
+      "epoch": 1.2766362252663623,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.169864952929914e-06,
+      "loss": 1.42336769,
+      "memory(GiB)": 123.79,
+      "step": 50325,
+      "train_speed(iter/s)": 1.315924
+    },
+    {
+      "acc": 0.68533678,
+      "epoch": 1.2767630644342973,
+      "grad_norm": 2.75,
+      "learning_rate": 3.168889139556639e-06,
+      "loss": 1.38543806,
+      "memory(GiB)": 123.79,
+      "step": 50330,
+      "train_speed(iter/s)": 1.315938
+    },
+    {
+      "acc": 0.68750124,
+      "epoch": 1.2768899036022323,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.1679134067237942e-06,
+      "loss": 1.39943218,
+      "memory(GiB)": 123.79,
+      "step": 50335,
+      "train_speed(iter/s)": 1.315953
+    },
+    {
+      "acc": 0.68907366,
+      "epoch": 1.2770167427701673,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.166937754474296e-06,
+      "loss": 1.37794685,
+      "memory(GiB)": 123.79,
+      "step": 50340,
+      "train_speed(iter/s)": 1.315967
+    },
+    {
+      "acc": 0.6994884,
+      "epoch": 1.2771435819381025,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.165962182851061e-06,
+      "loss": 1.35776672,
+      "memory(GiB)": 123.79,
+      "step": 50345,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.69674511,
+      "epoch": 1.2772704211060375,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.1649866918969984e-06,
+      "loss": 1.36292295,
+      "memory(GiB)": 123.79,
+      "step": 50350,
+      "train_speed(iter/s)": 1.315996
+    },
+    {
+      "acc": 0.68187709,
+      "epoch": 1.2773972602739727,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.164011281655013e-06,
+      "loss": 1.3911768,
+      "memory(GiB)": 123.79,
+      "step": 50355,
+      "train_speed(iter/s)": 1.316011
+    },
+    {
+      "acc": 0.67852235,
+      "epoch": 1.2775240994419077,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.163035952168007e-06,
+      "loss": 1.44871159,
+      "memory(GiB)": 123.79,
+      "step": 50360,
+      "train_speed(iter/s)": 1.316026
+    },
+    {
+      "acc": 0.69361773,
+      "epoch": 1.2776509386098427,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.1620607034788832e-06,
+      "loss": 1.44648361,
+      "memory(GiB)": 123.79,
+      "step": 50365,
+      "train_speed(iter/s)": 1.316042
+    },
+    {
+      "acc": 0.6830225,
+      "epoch": 1.2777777777777777,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.1610855356305354e-06,
+      "loss": 1.42387342,
+      "memory(GiB)": 123.79,
+      "step": 50370,
+      "train_speed(iter/s)": 1.316057
+    },
+    {
+      "acc": 0.67918587,
+      "epoch": 1.277904616945713,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.160110448665854e-06,
+      "loss": 1.39270077,
+      "memory(GiB)": 123.79,
+      "step": 50375,
+      "train_speed(iter/s)": 1.316073
+    },
+    {
+      "acc": 0.69644761,
+      "epoch": 1.2780314561136479,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.159135442627731e-06,
+      "loss": 1.33313866,
+      "memory(GiB)": 123.79,
+      "step": 50380,
+      "train_speed(iter/s)": 1.316088
+    },
+    {
+      "acc": 0.69518242,
+      "epoch": 1.278158295281583,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.158160517559049e-06,
+      "loss": 1.39483204,
+      "memory(GiB)": 123.79,
+      "step": 50385,
+      "train_speed(iter/s)": 1.316103
+    },
+    {
+      "acc": 0.68544908,
+      "epoch": 1.278285134449518,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.157185673502693e-06,
+      "loss": 1.41471043,
+      "memory(GiB)": 123.79,
+      "step": 50390,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.68075304,
+      "epoch": 1.278411973617453,
+      "grad_norm": 3.875,
+      "learning_rate": 3.156210910501537e-06,
+      "loss": 1.38386011,
+      "memory(GiB)": 123.79,
+      "step": 50395,
+      "train_speed(iter/s)": 1.316134
+    },
+    {
+      "acc": 0.68380809,
+      "epoch": 1.278538812785388,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.155236228598457e-06,
+      "loss": 1.38529472,
+      "memory(GiB)": 123.79,
+      "step": 50400,
+      "train_speed(iter/s)": 1.31615
+    },
+    {
+      "acc": 0.6771719,
+      "epoch": 1.2786656519533233,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.1542616278363238e-06,
+      "loss": 1.39909458,
+      "memory(GiB)": 123.79,
+      "step": 50405,
+      "train_speed(iter/s)": 1.316165
+    },
+    {
+      "acc": 0.67865429,
+      "epoch": 1.2787924911212583,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.1532871082580064e-06,
+      "loss": 1.44227667,
+      "memory(GiB)": 123.79,
+      "step": 50410,
+      "train_speed(iter/s)": 1.31618
+    },
+    {
+      "acc": 0.6863719,
+      "epoch": 1.2789193302891932,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.152312669906366e-06,
+      "loss": 1.39114838,
+      "memory(GiB)": 123.79,
+      "step": 50415,
+      "train_speed(iter/s)": 1.316194
+    },
+    {
+      "acc": 0.70192842,
+      "epoch": 1.2790461694571285,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.1513383128242624e-06,
+      "loss": 1.32451572,
+      "memory(GiB)": 123.79,
+      "step": 50420,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.69704757,
+      "epoch": 1.2791730086250634,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.150364037054555e-06,
+      "loss": 1.4183672,
+      "memory(GiB)": 123.79,
+      "step": 50425,
+      "train_speed(iter/s)": 1.316222
+    },
+    {
+      "acc": 0.68504763,
+      "epoch": 1.2792998477929984,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.149389842640096e-06,
+      "loss": 1.34909687,
+      "memory(GiB)": 123.79,
+      "step": 50430,
+      "train_speed(iter/s)": 1.316237
+    },
+    {
+      "acc": 0.69679756,
+      "epoch": 1.2794266869609334,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.148415729623732e-06,
+      "loss": 1.35730295,
+      "memory(GiB)": 123.79,
+      "step": 50435,
+      "train_speed(iter/s)": 1.316253
+    },
+    {
+      "acc": 0.69536104,
+      "epoch": 1.2795535261288686,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.1474416980483126e-06,
+      "loss": 1.41899414,
+      "memory(GiB)": 123.79,
+      "step": 50440,
+      "train_speed(iter/s)": 1.316269
+    },
+    {
+      "acc": 0.70443912,
+      "epoch": 1.2796803652968036,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.1464677479566774e-06,
+      "loss": 1.37579746,
+      "memory(GiB)": 123.79,
+      "step": 50445,
+      "train_speed(iter/s)": 1.316284
+    },
+    {
+      "acc": 0.69151845,
+      "epoch": 1.2798072044647388,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.1454938793916677e-06,
+      "loss": 1.41720219,
+      "memory(GiB)": 123.79,
+      "step": 50450,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.67604609,
+      "epoch": 1.2799340436326738,
+      "grad_norm": 3.25,
+      "learning_rate": 3.144520092396115e-06,
+      "loss": 1.40367155,
+      "memory(GiB)": 123.79,
+      "step": 50455,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.67795563,
+      "epoch": 1.2800608828006088,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.1435463870128536e-06,
+      "loss": 1.40713968,
+      "memory(GiB)": 123.79,
+      "step": 50460,
+      "train_speed(iter/s)": 1.316332
+    },
+    {
+      "acc": 0.70455594,
+      "epoch": 1.2801877219685438,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.142572763284709e-06,
+      "loss": 1.33942633,
+      "memory(GiB)": 123.79,
+      "step": 50465,
+      "train_speed(iter/s)": 1.316348
+    },
+    {
+      "acc": 0.67838507,
+      "epoch": 1.280314561136479,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.14159922125451e-06,
+      "loss": 1.42127781,
+      "memory(GiB)": 123.79,
+      "step": 50470,
+      "train_speed(iter/s)": 1.316362
+    },
+    {
+      "acc": 0.68559895,
+      "epoch": 1.280441400304414,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.1406257609650724e-06,
+      "loss": 1.38236971,
+      "memory(GiB)": 123.79,
+      "step": 50475,
+      "train_speed(iter/s)": 1.316376
+    },
+    {
+      "acc": 0.69400315,
+      "epoch": 1.280568239472349,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.139652382459215e-06,
+      "loss": 1.39679451,
+      "memory(GiB)": 123.79,
+      "step": 50480,
+      "train_speed(iter/s)": 1.316392
+    },
+    {
+      "acc": 0.68375177,
+      "epoch": 1.2806950786402842,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.1386790857797535e-06,
+      "loss": 1.43279686,
+      "memory(GiB)": 123.79,
+      "step": 50485,
+      "train_speed(iter/s)": 1.316409
+    },
+    {
+      "acc": 0.7089066,
+      "epoch": 1.2808219178082192,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.1377058709694957e-06,
+      "loss": 1.33624477,
+      "memory(GiB)": 123.79,
+      "step": 50490,
+      "train_speed(iter/s)": 1.316425
+    },
+    {
+      "acc": 0.68319874,
+      "epoch": 1.2809487569761542,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.1367327380712477e-06,
+      "loss": 1.33415108,
+      "memory(GiB)": 123.79,
+      "step": 50495,
+      "train_speed(iter/s)": 1.316439
+    },
+    {
+      "acc": 0.68028193,
+      "epoch": 1.2810755961440892,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.135759687127812e-06,
+      "loss": 1.38145037,
+      "memory(GiB)": 123.79,
+      "step": 50500,
+      "train_speed(iter/s)": 1.316455
+    },
+    {
+      "acc": 0.68857555,
+      "epoch": 1.2812024353120244,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.134786718181989e-06,
+      "loss": 1.39810419,
+      "memory(GiB)": 123.79,
+      "step": 50505,
+      "train_speed(iter/s)": 1.316471
+    },
+    {
+      "acc": 0.68865781,
+      "epoch": 1.2813292744799594,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.133813831276575e-06,
+      "loss": 1.37684965,
+      "memory(GiB)": 123.79,
+      "step": 50510,
+      "train_speed(iter/s)": 1.316486
+    },
+    {
+      "acc": 0.68894424,
+      "epoch": 1.2814561136478946,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.1328410264543585e-06,
+      "loss": 1.39525824,
+      "memory(GiB)": 123.79,
+      "step": 50515,
+      "train_speed(iter/s)": 1.316502
+    },
+    {
+      "acc": 0.69181156,
+      "epoch": 1.2815829528158296,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.131868303758131e-06,
+      "loss": 1.32657528,
+      "memory(GiB)": 123.79,
+      "step": 50520,
+      "train_speed(iter/s)": 1.316518
+    },
+    {
+      "acc": 0.68778257,
+      "epoch": 1.2817097919837646,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.1308956632306754e-06,
+      "loss": 1.34472876,
+      "memory(GiB)": 123.79,
+      "step": 50525,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.6815568,
+      "epoch": 1.2818366311516995,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.129923104914776e-06,
+      "loss": 1.42387571,
+      "memory(GiB)": 123.79,
+      "step": 50530,
+      "train_speed(iter/s)": 1.316549
+    },
+    {
+      "acc": 0.69763417,
+      "epoch": 1.2819634703196348,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.1289506288532045e-06,
+      "loss": 1.27621698,
+      "memory(GiB)": 123.79,
+      "step": 50535,
+      "train_speed(iter/s)": 1.316565
+    },
+    {
+      "acc": 0.67434783,
+      "epoch": 1.2820903094875697,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.12797823508874e-06,
+      "loss": 1.42120094,
+      "memory(GiB)": 123.79,
+      "step": 50540,
+      "train_speed(iter/s)": 1.316581
+    },
+    {
+      "acc": 0.68955212,
+      "epoch": 1.282217148655505,
+      "grad_norm": 3.0,
+      "learning_rate": 3.127005923664149e-06,
+      "loss": 1.38281107,
+      "memory(GiB)": 123.79,
+      "step": 50545,
+      "train_speed(iter/s)": 1.316596
+    },
+    {
+      "acc": 0.68788066,
+      "epoch": 1.28234398782344,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.1260336946222014e-06,
+      "loss": 1.34870014,
+      "memory(GiB)": 123.79,
+      "step": 50550,
+      "train_speed(iter/s)": 1.316613
+    },
+    {
+      "acc": 0.70075235,
+      "epoch": 1.282470826991375,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.1250615480056584e-06,
+      "loss": 1.30393181,
+      "memory(GiB)": 123.79,
+      "step": 50555,
+      "train_speed(iter/s)": 1.316628
+    },
+    {
+      "acc": 0.70217204,
+      "epoch": 1.28259766615931,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.124089483857278e-06,
+      "loss": 1.28944302,
+      "memory(GiB)": 123.79,
+      "step": 50560,
+      "train_speed(iter/s)": 1.316644
+    },
+    {
+      "acc": 0.68183699,
+      "epoch": 1.2827245053272451,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.123117502219819e-06,
+      "loss": 1.40885773,
+      "memory(GiB)": 123.79,
+      "step": 50565,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.69565916,
+      "epoch": 1.2828513444951801,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.122145603136032e-06,
+      "loss": 1.38029518,
+      "memory(GiB)": 123.79,
+      "step": 50570,
+      "train_speed(iter/s)": 1.316676
+    },
+    {
+      "acc": 0.69337883,
+      "epoch": 1.2829781836631151,
+      "grad_norm": 3.625,
+      "learning_rate": 3.1211737866486653e-06,
+      "loss": 1.40116081,
+      "memory(GiB)": 123.79,
+      "step": 50575,
+      "train_speed(iter/s)": 1.316691
+    },
+    {
+      "acc": 0.6875504,
+      "epoch": 1.2831050228310503,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.1202020528004627e-06,
+      "loss": 1.36462135,
+      "memory(GiB)": 123.79,
+      "step": 50580,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.6853447,
+      "epoch": 1.2832318619989853,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.119230401634167e-06,
+      "loss": 1.39190855,
+      "memory(GiB)": 123.79,
+      "step": 50585,
+      "train_speed(iter/s)": 1.316719
+    },
+    {
+      "acc": 0.68897581,
+      "epoch": 1.2833587011669203,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.118258833192517e-06,
+      "loss": 1.3572525,
+      "memory(GiB)": 123.79,
+      "step": 50590,
+      "train_speed(iter/s)": 1.316734
+    },
+    {
+      "acc": 0.67970514,
+      "epoch": 1.2834855403348553,
+      "grad_norm": 4.625,
+      "learning_rate": 3.117287347518242e-06,
+      "loss": 1.37578697,
+      "memory(GiB)": 123.79,
+      "step": 50595,
+      "train_speed(iter/s)": 1.316749
+    },
+    {
+      "acc": 0.68490448,
+      "epoch": 1.2836123795027905,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.116315944654077e-06,
+      "loss": 1.46556358,
+      "memory(GiB)": 123.79,
+      "step": 50600,
+      "train_speed(iter/s)": 1.316764
+    },
+    {
+      "acc": 0.67988324,
+      "epoch": 1.2837392186707255,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.115344624642745e-06,
+      "loss": 1.41442146,
+      "memory(GiB)": 123.79,
+      "step": 50605,
+      "train_speed(iter/s)": 1.316779
+    },
+    {
+      "acc": 0.70045309,
+      "epoch": 1.2838660578386607,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.1143733875269734e-06,
+      "loss": 1.39620609,
+      "memory(GiB)": 123.79,
+      "step": 50610,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.70757627,
+      "epoch": 1.2839928970065957,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.1134022333494774e-06,
+      "loss": 1.30080204,
+      "memory(GiB)": 123.79,
+      "step": 50615,
+      "train_speed(iter/s)": 1.316809
+    },
+    {
+      "acc": 0.70600867,
+      "epoch": 1.2841197361745307,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.1124311621529744e-06,
+      "loss": 1.31500378,
+      "memory(GiB)": 123.79,
+      "step": 50620,
+      "train_speed(iter/s)": 1.316824
+    },
+    {
+      "acc": 0.69138489,
+      "epoch": 1.2842465753424657,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.111460173980175e-06,
+      "loss": 1.4101099,
+      "memory(GiB)": 123.79,
+      "step": 50625,
+      "train_speed(iter/s)": 1.316838
+    },
+    {
+      "acc": 0.68616848,
+      "epoch": 1.2843734145104009,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.110489268873792e-06,
+      "loss": 1.35473328,
+      "memory(GiB)": 123.79,
+      "step": 50630,
+      "train_speed(iter/s)": 1.316854
+    },
+    {
+      "acc": 0.69971013,
+      "epoch": 1.2845002536783359,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.1095184468765248e-06,
+      "loss": 1.39091549,
+      "memory(GiB)": 123.79,
+      "step": 50635,
+      "train_speed(iter/s)": 1.316868
+    },
+    {
+      "acc": 0.68075552,
+      "epoch": 1.2846270928462709,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.1085477080310766e-06,
+      "loss": 1.40203648,
+      "memory(GiB)": 123.79,
+      "step": 50640,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.69189711,
+      "epoch": 1.284753932014206,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.1075770523801453e-06,
+      "loss": 1.34335442,
+      "memory(GiB)": 123.79,
+      "step": 50645,
+      "train_speed(iter/s)": 1.316899
+    },
+    {
+      "acc": 0.69268913,
+      "epoch": 1.284880771182141,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.106606479966426e-06,
+      "loss": 1.36213827,
+      "memory(GiB)": 123.79,
+      "step": 50650,
+      "train_speed(iter/s)": 1.316914
+    },
+    {
+      "acc": 0.70698032,
+      "epoch": 1.285007610350076,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.1056359908326044e-06,
+      "loss": 1.3669342,
+      "memory(GiB)": 123.79,
+      "step": 50655,
+      "train_speed(iter/s)": 1.316928
+    },
+    {
+      "acc": 0.6803369,
+      "epoch": 1.285134449518011,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.1046655850213707e-06,
+      "loss": 1.36918192,
+      "memory(GiB)": 123.79,
+      "step": 50660,
+      "train_speed(iter/s)": 1.316943
+    },
+    {
+      "acc": 0.68831902,
+      "epoch": 1.2852612886859462,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.103695262575407e-06,
+      "loss": 1.41130638,
+      "memory(GiB)": 123.79,
+      "step": 50665,
+      "train_speed(iter/s)": 1.316958
+    },
+    {
+      "acc": 0.68349743,
+      "epoch": 1.2853881278538812,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.102725023537393e-06,
+      "loss": 1.419804,
+      "memory(GiB)": 123.79,
+      "step": 50670,
+      "train_speed(iter/s)": 1.316973
+    },
+    {
+      "acc": 0.6915772,
+      "epoch": 1.2855149670218164,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.1017548679500008e-06,
+      "loss": 1.31835327,
+      "memory(GiB)": 123.79,
+      "step": 50675,
+      "train_speed(iter/s)": 1.316987
+    },
+    {
+      "acc": 0.68071222,
+      "epoch": 1.2856418061897514,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.1007847958559057e-06,
+      "loss": 1.3948266,
+      "memory(GiB)": 123.79,
+      "step": 50680,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.67809849,
+      "epoch": 1.2857686453576864,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.099814807297774e-06,
+      "loss": 1.42551451,
+      "memory(GiB)": 123.79,
+      "step": 50685,
+      "train_speed(iter/s)": 1.317016
+    },
+    {
+      "acc": 0.70019569,
+      "epoch": 1.2858954845256214,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.098844902318272e-06,
+      "loss": 1.32650976,
+      "memory(GiB)": 123.79,
+      "step": 50690,
+      "train_speed(iter/s)": 1.317029
+    },
+    {
+      "acc": 0.69929914,
+      "epoch": 1.2860223236935566,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.0978750809600596e-06,
+      "loss": 1.39469471,
+      "memory(GiB)": 123.79,
+      "step": 50695,
+      "train_speed(iter/s)": 1.317042
+    },
+    {
+      "acc": 0.70184002,
+      "epoch": 1.2861491628614916,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.0969053432657913e-06,
+      "loss": 1.35714102,
+      "memory(GiB)": 123.79,
+      "step": 50700,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.68932819,
+      "epoch": 1.2862760020294268,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.0959356892781246e-06,
+      "loss": 1.3605895,
+      "memory(GiB)": 123.79,
+      "step": 50705,
+      "train_speed(iter/s)": 1.317069
+    },
+    {
+      "acc": 0.69317827,
+      "epoch": 1.2864028411973618,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.0949661190397072e-06,
+      "loss": 1.38179083,
+      "memory(GiB)": 123.79,
+      "step": 50710,
+      "train_speed(iter/s)": 1.317083
+    },
+    {
+      "acc": 0.68638849,
+      "epoch": 1.2865296803652968,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.0939966325931852e-06,
+      "loss": 1.43539505,
+      "memory(GiB)": 123.79,
+      "step": 50715,
+      "train_speed(iter/s)": 1.317097
+    },
+    {
+      "acc": 0.66994386,
+      "epoch": 1.2866565195332318,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.0930272299812e-06,
+      "loss": 1.46657133,
+      "memory(GiB)": 123.79,
+      "step": 50720,
+      "train_speed(iter/s)": 1.317111
+    },
+    {
+      "acc": 0.69124012,
+      "epoch": 1.286783358701167,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.0920579112463916e-06,
+      "loss": 1.36563034,
+      "memory(GiB)": 123.79,
+      "step": 50725,
+      "train_speed(iter/s)": 1.317126
+    },
+    {
+      "acc": 0.68307209,
+      "epoch": 1.286910197869102,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.0910886764313964e-06,
+      "loss": 1.46760283,
+      "memory(GiB)": 123.79,
+      "step": 50730,
+      "train_speed(iter/s)": 1.31714
+    },
+    {
+      "acc": 0.69823451,
+      "epoch": 1.287037037037037,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.0901195255788406e-06,
+      "loss": 1.33880463,
+      "memory(GiB)": 123.79,
+      "step": 50735,
+      "train_speed(iter/s)": 1.317154
+    },
+    {
+      "acc": 0.70412674,
+      "epoch": 1.2871638762049722,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.089150458731357e-06,
+      "loss": 1.31993523,
+      "memory(GiB)": 123.79,
+      "step": 50740,
+      "train_speed(iter/s)": 1.317168
+    },
+    {
+      "acc": 0.70886536,
+      "epoch": 1.2872907153729072,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.0881814759315666e-06,
+      "loss": 1.31508331,
+      "memory(GiB)": 123.79,
+      "step": 50745,
+      "train_speed(iter/s)": 1.317182
+    },
+    {
+      "acc": 0.6850421,
+      "epoch": 1.2874175545408422,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.0872125772220934e-06,
+      "loss": 1.4144125,
+      "memory(GiB)": 123.79,
+      "step": 50750,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.68786397,
+      "epoch": 1.2875443937087772,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.0862437626455483e-06,
+      "loss": 1.42987423,
+      "memory(GiB)": 123.79,
+      "step": 50755,
+      "train_speed(iter/s)": 1.317208
+    },
+    {
+      "acc": 0.69711757,
+      "epoch": 1.2876712328767124,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.0852750322445473e-06,
+      "loss": 1.28881283,
+      "memory(GiB)": 123.79,
+      "step": 50760,
+      "train_speed(iter/s)": 1.317221
+    },
+    {
+      "acc": 0.69364853,
+      "epoch": 1.2877980720446474,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.0843063860616982e-06,
+      "loss": 1.36598091,
+      "memory(GiB)": 123.79,
+      "step": 50765,
+      "train_speed(iter/s)": 1.317234
+    },
+    {
+      "acc": 0.69027472,
+      "epoch": 1.2879249112125826,
+      "grad_norm": 3.0,
+      "learning_rate": 3.0833378241396094e-06,
+      "loss": 1.37518702,
+      "memory(GiB)": 123.79,
+      "step": 50770,
+      "train_speed(iter/s)": 1.317249
+    },
+    {
+      "acc": 0.66805325,
+      "epoch": 1.2880517503805176,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.0823693465208794e-06,
+      "loss": 1.40359669,
+      "memory(GiB)": 123.79,
+      "step": 50775,
+      "train_speed(iter/s)": 1.317264
+    },
+    {
+      "acc": 0.68770103,
+      "epoch": 1.2881785895484525,
+      "grad_norm": 4.125,
+      "learning_rate": 3.081400953248106e-06,
+      "loss": 1.37292404,
+      "memory(GiB)": 123.79,
+      "step": 50780,
+      "train_speed(iter/s)": 1.317278
+    },
+    {
+      "acc": 0.68596478,
+      "epoch": 1.2883054287163875,
+      "grad_norm": 4.78125,
+      "learning_rate": 3.0804326443638854e-06,
+      "loss": 1.40479746,
+      "memory(GiB)": 123.79,
+      "step": 50785,
+      "train_speed(iter/s)": 1.317293
+    },
+    {
+      "acc": 0.67775173,
+      "epoch": 1.2884322678843227,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.0794644199108087e-06,
+      "loss": 1.46759491,
+      "memory(GiB)": 123.79,
+      "step": 50790,
+      "train_speed(iter/s)": 1.317307
+    },
+    {
+      "acc": 0.68267312,
+      "epoch": 1.2885591070522577,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.07849627993146e-06,
+      "loss": 1.41319046,
+      "memory(GiB)": 123.79,
+      "step": 50795,
+      "train_speed(iter/s)": 1.317322
+    },
+    {
+      "acc": 0.68657064,
+      "epoch": 1.2886859462201927,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.0775282244684233e-06,
+      "loss": 1.35004921,
+      "memory(GiB)": 123.79,
+      "step": 50800,
+      "train_speed(iter/s)": 1.317336
+    },
+    {
+      "acc": 0.67630706,
+      "epoch": 1.288812785388128,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.076560253564279e-06,
+      "loss": 1.42235413,
+      "memory(GiB)": 123.79,
+      "step": 50805,
+      "train_speed(iter/s)": 1.317352
+    },
+    {
+      "acc": 0.68486295,
+      "epoch": 1.288939624556063,
+      "grad_norm": 4.125,
+      "learning_rate": 3.0755923672616038e-06,
+      "loss": 1.43169937,
+      "memory(GiB)": 123.79,
+      "step": 50810,
+      "train_speed(iter/s)": 1.317366
+    },
+    {
+      "acc": 0.68012247,
+      "epoch": 1.289066463723998,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.074624565602966e-06,
+      "loss": 1.4564683,
+      "memory(GiB)": 123.79,
+      "step": 50815,
+      "train_speed(iter/s)": 1.317381
+    },
+    {
+      "acc": 0.67825861,
+      "epoch": 1.289193302891933,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.073656848630937e-06,
+      "loss": 1.37528009,
+      "memory(GiB)": 123.79,
+      "step": 50820,
+      "train_speed(iter/s)": 1.317395
+    },
+    {
+      "acc": 0.67828979,
+      "epoch": 1.2893201420598681,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.0726892163880784e-06,
+      "loss": 1.41723022,
+      "memory(GiB)": 123.79,
+      "step": 50825,
+      "train_speed(iter/s)": 1.317408
+    },
+    {
+      "acc": 0.68103957,
+      "epoch": 1.289446981227803,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.071721668916956e-06,
+      "loss": 1.41865559,
+      "memory(GiB)": 123.79,
+      "step": 50830,
+      "train_speed(iter/s)": 1.317423
+    },
+    {
+      "acc": 0.69846172,
+      "epoch": 1.2895738203957383,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.0707542062601225e-06,
+      "loss": 1.3118927,
+      "memory(GiB)": 123.79,
+      "step": 50835,
+      "train_speed(iter/s)": 1.317436
+    },
+    {
+      "acc": 0.66992531,
+      "epoch": 1.2897006595636733,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.0697868284601323e-06,
+      "loss": 1.40255241,
+      "memory(GiB)": 123.79,
+      "step": 50840,
+      "train_speed(iter/s)": 1.317451
+    },
+    {
+      "acc": 0.70335178,
+      "epoch": 1.2898274987316083,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.068819535559534e-06,
+      "loss": 1.30707397,
+      "memory(GiB)": 123.79,
+      "step": 50845,
+      "train_speed(iter/s)": 1.317466
+    },
+    {
+      "acc": 0.69343519,
+      "epoch": 1.2899543378995433,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.0678523276008774e-06,
+      "loss": 1.32936563,
+      "memory(GiB)": 123.79,
+      "step": 50850,
+      "train_speed(iter/s)": 1.317482
+    },
+    {
+      "acc": 0.69261875,
+      "epoch": 1.2900811770674785,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.0668852046267e-06,
+      "loss": 1.34188089,
+      "memory(GiB)": 123.79,
+      "step": 50855,
+      "train_speed(iter/s)": 1.317496
+    },
+    {
+      "acc": 0.69076352,
+      "epoch": 1.2902080162354135,
+      "grad_norm": 3.5,
+      "learning_rate": 3.0659181666795413e-06,
+      "loss": 1.36771317,
+      "memory(GiB)": 123.79,
+      "step": 50860,
+      "train_speed(iter/s)": 1.317511
+    },
+    {
+      "acc": 0.67900429,
+      "epoch": 1.2903348554033487,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.0649512138019376e-06,
+      "loss": 1.40426636,
+      "memory(GiB)": 123.79,
+      "step": 50865,
+      "train_speed(iter/s)": 1.317526
+    },
+    {
+      "acc": 0.68986592,
+      "epoch": 1.2904616945712837,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.0639843460364203e-06,
+      "loss": 1.36651764,
+      "memory(GiB)": 123.79,
+      "step": 50870,
+      "train_speed(iter/s)": 1.31754
+    },
+    {
+      "acc": 0.678723,
+      "epoch": 1.2905885337392187,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.0630175634255134e-06,
+      "loss": 1.42659702,
+      "memory(GiB)": 123.79,
+      "step": 50875,
+      "train_speed(iter/s)": 1.317555
+    },
+    {
+      "acc": 0.67353754,
+      "epoch": 1.2907153729071537,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.062050866011742e-06,
+      "loss": 1.4521904,
+      "memory(GiB)": 123.79,
+      "step": 50880,
+      "train_speed(iter/s)": 1.31757
+    },
+    {
+      "acc": 0.6710845,
+      "epoch": 1.2908422120750889,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.0610842538376264e-06,
+      "loss": 1.49274464,
+      "memory(GiB)": 123.79,
+      "step": 50885,
+      "train_speed(iter/s)": 1.317586
+    },
+    {
+      "acc": 0.69419131,
+      "epoch": 1.2909690512430239,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.060117726945683e-06,
+      "loss": 1.40482645,
+      "memory(GiB)": 123.79,
+      "step": 50890,
+      "train_speed(iter/s)": 1.317601
+    },
+    {
+      "acc": 0.66727638,
+      "epoch": 1.2910958904109588,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.059151285378421e-06,
+      "loss": 1.48829794,
+      "memory(GiB)": 123.79,
+      "step": 50895,
+      "train_speed(iter/s)": 1.317617
+    },
+    {
+      "acc": 0.69770012,
+      "epoch": 1.291222729578894,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.0581849291783518e-06,
+      "loss": 1.33823318,
+      "memory(GiB)": 123.79,
+      "step": 50900,
+      "train_speed(iter/s)": 1.317632
+    },
+    {
+      "acc": 0.69320498,
+      "epoch": 1.291349568746829,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.057218658387977e-06,
+      "loss": 1.35901756,
+      "memory(GiB)": 123.79,
+      "step": 50905,
+      "train_speed(iter/s)": 1.317648
+    },
+    {
+      "acc": 0.69119291,
+      "epoch": 1.291476407914764,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.0562524730498023e-06,
+      "loss": 1.41455221,
+      "memory(GiB)": 123.79,
+      "step": 50910,
+      "train_speed(iter/s)": 1.317663
+    },
+    {
+      "acc": 0.67523069,
+      "epoch": 1.291603247082699,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.055286373206321e-06,
+      "loss": 1.45950279,
+      "memory(GiB)": 123.79,
+      "step": 50915,
+      "train_speed(iter/s)": 1.317679
+    },
+    {
+      "acc": 0.68980365,
+      "epoch": 1.2917300862506342,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.054320358900027e-06,
+      "loss": 1.4115263,
+      "memory(GiB)": 123.79,
+      "step": 50920,
+      "train_speed(iter/s)": 1.317694
+    },
+    {
+      "acc": 0.67438135,
+      "epoch": 1.2918569254185692,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.053354430173411e-06,
+      "loss": 1.45292664,
+      "memory(GiB)": 123.79,
+      "step": 50925,
+      "train_speed(iter/s)": 1.31771
+    },
+    {
+      "acc": 0.70824289,
+      "epoch": 1.2919837645865044,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.0523885870689595e-06,
+      "loss": 1.32400913,
+      "memory(GiB)": 123.79,
+      "step": 50930,
+      "train_speed(iter/s)": 1.317725
+    },
+    {
+      "acc": 0.70425572,
+      "epoch": 1.2921106037544394,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.051422829629152e-06,
+      "loss": 1.32013359,
+      "memory(GiB)": 123.79,
+      "step": 50935,
+      "train_speed(iter/s)": 1.317737
+    },
+    {
+      "acc": 0.68207674,
+      "epoch": 1.2922374429223744,
+      "grad_norm": 2.625,
+      "learning_rate": 3.050457157896467e-06,
+      "loss": 1.39699469,
+      "memory(GiB)": 123.79,
+      "step": 50940,
+      "train_speed(iter/s)": 1.317753
+    },
+    {
+      "acc": 0.69187918,
+      "epoch": 1.2923642820903094,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.049491571913382e-06,
+      "loss": 1.36470814,
+      "memory(GiB)": 123.79,
+      "step": 50945,
+      "train_speed(iter/s)": 1.317766
+    },
+    {
+      "acc": 0.67735252,
+      "epoch": 1.2924911212582446,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.048526071722367e-06,
+      "loss": 1.39466114,
+      "memory(GiB)": 123.79,
+      "step": 50950,
+      "train_speed(iter/s)": 1.317782
+    },
+    {
+      "acc": 0.68719711,
+      "epoch": 1.2926179604261796,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.047560657365886e-06,
+      "loss": 1.37448978,
+      "memory(GiB)": 123.79,
+      "step": 50955,
+      "train_speed(iter/s)": 1.317798
+    },
+    {
+      "acc": 0.67388,
+      "epoch": 1.2927447995941146,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.046595328886405e-06,
+      "loss": 1.47020473,
+      "memory(GiB)": 123.79,
+      "step": 50960,
+      "train_speed(iter/s)": 1.317814
+    },
+    {
+      "acc": 0.68857889,
+      "epoch": 1.2928716387620498,
+      "grad_norm": 3.0,
+      "learning_rate": 3.0456300863263815e-06,
+      "loss": 1.36125479,
+      "memory(GiB)": 123.79,
+      "step": 50965,
+      "train_speed(iter/s)": 1.31783
+    },
+    {
+      "acc": 0.69268694,
+      "epoch": 1.2929984779299848,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.044664929728276e-06,
+      "loss": 1.34742413,
+      "memory(GiB)": 123.79,
+      "step": 50970,
+      "train_speed(iter/s)": 1.317845
+    },
+    {
+      "acc": 0.69901867,
+      "epoch": 1.2931253170979198,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.0436998591345336e-06,
+      "loss": 1.31997261,
+      "memory(GiB)": 123.79,
+      "step": 50975,
+      "train_speed(iter/s)": 1.317861
+    },
+    {
+      "acc": 0.68553658,
+      "epoch": 1.2932521562658548,
+      "grad_norm": 4.375,
+      "learning_rate": 3.042734874587607e-06,
+      "loss": 1.40237846,
+      "memory(GiB)": 123.79,
+      "step": 50980,
+      "train_speed(iter/s)": 1.317877
+    },
+    {
+      "acc": 0.67587261,
+      "epoch": 1.29337899543379,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.041769976129938e-06,
+      "loss": 1.37244415,
+      "memory(GiB)": 123.79,
+      "step": 50985,
+      "train_speed(iter/s)": 1.317892
+    },
+    {
+      "acc": 0.69669428,
+      "epoch": 1.293505834601725,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.0408051638039697e-06,
+      "loss": 1.39012518,
+      "memory(GiB)": 123.79,
+      "step": 50990,
+      "train_speed(iter/s)": 1.317906
+    },
+    {
+      "acc": 0.6691288,
+      "epoch": 1.2936326737696602,
+      "grad_norm": 5.28125,
+      "learning_rate": 3.039840437652137e-06,
+      "loss": 1.49621277,
+      "memory(GiB)": 123.79,
+      "step": 50995,
+      "train_speed(iter/s)": 1.317922
+    },
+    {
+      "acc": 0.68641911,
+      "epoch": 1.2937595129375952,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.0388757977168724e-06,
+      "loss": 1.40725918,
+      "memory(GiB)": 123.79,
+      "step": 51000,
+      "train_speed(iter/s)": 1.317937
+    },
+    {
+      "epoch": 1.2937595129375952,
+      "eval_acc": 0.6749614185445759,
+      "eval_loss": 1.356402039527893,
+      "eval_runtime": 69.9922,
+      "eval_samples_per_second": 91.01,
+      "eval_steps_per_second": 22.76,
+      "step": 51000
+    },
+    {
+      "acc": 0.69109054,
+      "epoch": 1.2938863521055302,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.0379112440406066e-06,
+      "loss": 1.3444561,
+      "memory(GiB)": 123.79,
+      "step": 51005,
+      "train_speed(iter/s)": 1.315144
+    },
+    {
+      "acc": 0.68707986,
+      "epoch": 1.2940131912734651,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.036946776665766e-06,
+      "loss": 1.37526712,
+      "memory(GiB)": 123.79,
+      "step": 51010,
+      "train_speed(iter/s)": 1.315159
+    },
+    {
+      "acc": 0.69347315,
+      "epoch": 1.2941400304414004,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.0359823956347695e-06,
+      "loss": 1.34619427,
+      "memory(GiB)": 123.79,
+      "step": 51015,
+      "train_speed(iter/s)": 1.315173
+    },
+    {
+      "acc": 0.6870575,
+      "epoch": 1.2942668696093353,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.0350181009900347e-06,
+      "loss": 1.48419924,
+      "memory(GiB)": 123.79,
+      "step": 51020,
+      "train_speed(iter/s)": 1.315188
+    },
+    {
+      "acc": 0.70295162,
+      "epoch": 1.2943937087772706,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.0340538927739784e-06,
+      "loss": 1.32783184,
+      "memory(GiB)": 123.79,
+      "step": 51025,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.69854889,
+      "epoch": 1.2945205479452055,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.0330897710290093e-06,
+      "loss": 1.38829632,
+      "memory(GiB)": 123.79,
+      "step": 51030,
+      "train_speed(iter/s)": 1.315217
+    },
+    {
+      "acc": 0.67779737,
+      "epoch": 1.2946473871131405,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.032125735797532e-06,
+      "loss": 1.40619535,
+      "memory(GiB)": 123.79,
+      "step": 51035,
+      "train_speed(iter/s)": 1.31523
+    },
+    {
+      "acc": 0.66990366,
+      "epoch": 1.2947742262810755,
+      "grad_norm": 3.578125,
+      "learning_rate": 3.031161787121952e-06,
+      "loss": 1.47477703,
+      "memory(GiB)": 123.79,
+      "step": 51040,
+      "train_speed(iter/s)": 1.315244
+    },
+    {
+      "acc": 0.70030298,
+      "epoch": 1.2949010654490107,
+      "grad_norm": 2.625,
+      "learning_rate": 3.0301979250446655e-06,
+      "loss": 1.4248354,
+      "memory(GiB)": 123.79,
+      "step": 51045,
+      "train_speed(iter/s)": 1.315257
+    },
+    {
+      "acc": 0.71145225,
+      "epoch": 1.2950279046169457,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.029234149608071e-06,
+      "loss": 1.3104044,
+      "memory(GiB)": 123.79,
+      "step": 51050,
+      "train_speed(iter/s)": 1.31527
+    },
+    {
+      "acc": 0.69086237,
+      "epoch": 1.2951547437848807,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.0282704608545566e-06,
+      "loss": 1.3598278,
+      "memory(GiB)": 123.79,
+      "step": 51055,
+      "train_speed(iter/s)": 1.315284
+    },
+    {
+      "acc": 0.69291701,
+      "epoch": 1.295281582952816,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.0273068588265097e-06,
+      "loss": 1.40059118,
+      "memory(GiB)": 123.79,
+      "step": 51060,
+      "train_speed(iter/s)": 1.315298
+    },
+    {
+      "acc": 0.67956047,
+      "epoch": 1.295408422120751,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.0263433435663143e-06,
+      "loss": 1.43732147,
+      "memory(GiB)": 123.79,
+      "step": 51065,
+      "train_speed(iter/s)": 1.315312
+    },
+    {
+      "acc": 0.68060203,
+      "epoch": 1.295535261288686,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.0253799151163522e-06,
+      "loss": 1.41801271,
+      "memory(GiB)": 123.79,
+      "step": 51070,
+      "train_speed(iter/s)": 1.315326
+    },
+    {
+      "acc": 0.67670479,
+      "epoch": 1.295662100456621,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.0244165735189967e-06,
+      "loss": 1.38369083,
+      "memory(GiB)": 123.79,
+      "step": 51075,
+      "train_speed(iter/s)": 1.31534
+    },
+    {
+      "acc": 0.67065735,
+      "epoch": 1.295788939624556,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.023453318816619e-06,
+      "loss": 1.45980301,
+      "memory(GiB)": 123.79,
+      "step": 51080,
+      "train_speed(iter/s)": 1.315354
+    },
+    {
+      "acc": 0.68016319,
+      "epoch": 1.295915778792491,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.022490151051591e-06,
+      "loss": 1.35722713,
+      "memory(GiB)": 123.79,
+      "step": 51085,
+      "train_speed(iter/s)": 1.315362
+    },
+    {
+      "acc": 0.68602552,
+      "epoch": 1.2960426179604263,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.0215270702662753e-06,
+      "loss": 1.41570673,
+      "memory(GiB)": 123.79,
+      "step": 51090,
+      "train_speed(iter/s)": 1.315376
+    },
+    {
+      "acc": 0.70702677,
+      "epoch": 1.2961694571283613,
+      "grad_norm": 3.125,
+      "learning_rate": 3.020564076503031e-06,
+      "loss": 1.29644089,
+      "memory(GiB)": 123.79,
+      "step": 51095,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.68300905,
+      "epoch": 1.2962962962962963,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.019601169804216e-06,
+      "loss": 1.41149588,
+      "memory(GiB)": 123.79,
+      "step": 51100,
+      "train_speed(iter/s)": 1.315403
+    },
+    {
+      "acc": 0.68797131,
+      "epoch": 1.2964231354642313,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.018638350212184e-06,
+      "loss": 1.41163406,
+      "memory(GiB)": 123.79,
+      "step": 51105,
+      "train_speed(iter/s)": 1.315417
+    },
+    {
+      "acc": 0.68231859,
+      "epoch": 1.2965499746321665,
+      "grad_norm": 2.75,
+      "learning_rate": 3.0176756177692845e-06,
+      "loss": 1.37774076,
+      "memory(GiB)": 123.79,
+      "step": 51110,
+      "train_speed(iter/s)": 1.31543
+    },
+    {
+      "acc": 0.68159685,
+      "epoch": 1.2966768138001015,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.01671297251786e-06,
+      "loss": 1.42150173,
+      "memory(GiB)": 123.79,
+      "step": 51115,
+      "train_speed(iter/s)": 1.315444
+    },
+    {
+      "acc": 0.69333177,
+      "epoch": 1.2968036529680365,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.0157504145002546e-06,
+      "loss": 1.36196918,
+      "memory(GiB)": 123.79,
+      "step": 51120,
+      "train_speed(iter/s)": 1.315458
+    },
+    {
+      "acc": 0.67773485,
+      "epoch": 1.2969304921359717,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.0147879437588046e-06,
+      "loss": 1.40001106,
+      "memory(GiB)": 123.79,
+      "step": 51125,
+      "train_speed(iter/s)": 1.315472
+    },
+    {
+      "acc": 0.68686109,
+      "epoch": 1.2970573313039067,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.013825560335845e-06,
+      "loss": 1.39173336,
+      "memory(GiB)": 123.79,
+      "step": 51130,
+      "train_speed(iter/s)": 1.315486
+    },
+    {
+      "acc": 0.69321752,
+      "epoch": 1.2971841704718416,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.0128632642737044e-06,
+      "loss": 1.34399605,
+      "memory(GiB)": 123.79,
+      "step": 51135,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.6989666,
+      "epoch": 1.2973110096397766,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.0119010556147088e-06,
+      "loss": 1.35710173,
+      "memory(GiB)": 123.79,
+      "step": 51140,
+      "train_speed(iter/s)": 1.315514
+    },
+    {
+      "acc": 0.69257803,
+      "epoch": 1.2974378488077118,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.0109389344011813e-06,
+      "loss": 1.38999443,
+      "memory(GiB)": 123.79,
+      "step": 51145,
+      "train_speed(iter/s)": 1.315528
+    },
+    {
+      "acc": 0.69167318,
+      "epoch": 1.2975646879756468,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.0099769006754415e-06,
+      "loss": 1.36759472,
+      "memory(GiB)": 123.79,
+      "step": 51150,
+      "train_speed(iter/s)": 1.315543
+    },
+    {
+      "acc": 0.68013024,
+      "epoch": 1.297691527143582,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.0090149544798007e-06,
+      "loss": 1.37232227,
+      "memory(GiB)": 123.79,
+      "step": 51155,
+      "train_speed(iter/s)": 1.315556
+    },
+    {
+      "acc": 0.66940179,
+      "epoch": 1.297818366311517,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.0080530958565712e-06,
+      "loss": 1.42072048,
+      "memory(GiB)": 123.79,
+      "step": 51160,
+      "train_speed(iter/s)": 1.31557
+    },
+    {
+      "acc": 0.68651958,
+      "epoch": 1.297945205479452,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.0070913248480602e-06,
+      "loss": 1.37532387,
+      "memory(GiB)": 123.79,
+      "step": 51165,
+      "train_speed(iter/s)": 1.315584
+    },
+    {
+      "acc": 0.6889967,
+      "epoch": 1.298072044647387,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.0061296414965724e-06,
+      "loss": 1.36997232,
+      "memory(GiB)": 123.79,
+      "step": 51170,
+      "train_speed(iter/s)": 1.315598
+    },
+    {
+      "acc": 0.69900241,
+      "epoch": 1.2981988838153222,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.005168045844402e-06,
+      "loss": 1.39212475,
+      "memory(GiB)": 123.79,
+      "step": 51175,
+      "train_speed(iter/s)": 1.315611
+    },
+    {
+      "acc": 0.68599367,
+      "epoch": 1.2983257229832572,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.0042065379338486e-06,
+      "loss": 1.39576244,
+      "memory(GiB)": 123.79,
+      "step": 51180,
+      "train_speed(iter/s)": 1.315625
+    },
+    {
+      "acc": 0.68938932,
+      "epoch": 1.2984525621511924,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.003245117807201e-06,
+      "loss": 1.39584179,
+      "memory(GiB)": 123.79,
+      "step": 51185,
+      "train_speed(iter/s)": 1.315639
+    },
+    {
+      "acc": 0.70036955,
+      "epoch": 1.2985794013191274,
+      "grad_norm": 2.90625,
+      "learning_rate": 3.0022837855067514e-06,
+      "loss": 1.32278328,
+      "memory(GiB)": 123.79,
+      "step": 51190,
+      "train_speed(iter/s)": 1.315652
+    },
+    {
+      "acc": 0.69193726,
+      "epoch": 1.2987062404870624,
+      "grad_norm": 2.828125,
+      "learning_rate": 3.0013225410747772e-06,
+      "loss": 1.32884426,
+      "memory(GiB)": 123.79,
+      "step": 51195,
+      "train_speed(iter/s)": 1.315665
+    },
+    {
+      "acc": 0.67440634,
+      "epoch": 1.2988330796549974,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.0003613845535617e-06,
+      "loss": 1.40839939,
+      "memory(GiB)": 123.79,
+      "step": 51200,
+      "train_speed(iter/s)": 1.315679
+    },
+    {
+      "acc": 0.69026465,
+      "epoch": 1.2989599188229326,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.9994003159853793e-06,
+      "loss": 1.3951643,
+      "memory(GiB)": 123.79,
+      "step": 51205,
+      "train_speed(iter/s)": 1.315692
+    },
+    {
+      "acc": 0.688239,
+      "epoch": 1.2990867579908676,
+      "grad_norm": 3.125,
+      "learning_rate": 2.998439335412505e-06,
+      "loss": 1.33031807,
+      "memory(GiB)": 123.79,
+      "step": 51210,
+      "train_speed(iter/s)": 1.315706
+    },
+    {
+      "acc": 0.68563786,
+      "epoch": 1.2992135971588026,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.9974784428772043e-06,
+      "loss": 1.36286106,
+      "memory(GiB)": 123.79,
+      "step": 51215,
+      "train_speed(iter/s)": 1.315718
+    },
+    {
+      "acc": 0.68703599,
+      "epoch": 1.2993404363267378,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.996517638421741e-06,
+      "loss": 1.41041851,
+      "memory(GiB)": 123.79,
+      "step": 51220,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.69210348,
+      "epoch": 1.2994672754946728,
+      "grad_norm": 3.0,
+      "learning_rate": 2.9955569220883777e-06,
+      "loss": 1.3820837,
+      "memory(GiB)": 123.79,
+      "step": 51225,
+      "train_speed(iter/s)": 1.315744
+    },
+    {
+      "acc": 0.67931695,
+      "epoch": 1.2995941146626078,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.9945962939193718e-06,
+      "loss": 1.43700027,
+      "memory(GiB)": 123.79,
+      "step": 51230,
+      "train_speed(iter/s)": 1.315756
+    },
+    {
+      "acc": 0.69204082,
+      "epoch": 1.2997209538305428,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.9936357539569728e-06,
+      "loss": 1.3498189,
+      "memory(GiB)": 123.79,
+      "step": 51235,
+      "train_speed(iter/s)": 1.31577
+    },
+    {
+      "acc": 0.69736729,
+      "epoch": 1.299847792998478,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.9926753022434306e-06,
+      "loss": 1.32810326,
+      "memory(GiB)": 123.79,
+      "step": 51240,
+      "train_speed(iter/s)": 1.315785
+    },
+    {
+      "acc": 0.70321884,
+      "epoch": 1.299974632166413,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.9917149388209908e-06,
+      "loss": 1.3649744,
+      "memory(GiB)": 123.79,
+      "step": 51245,
+      "train_speed(iter/s)": 1.315798
+    },
+    {
+      "acc": 0.68449879,
+      "epoch": 1.3001014713343482,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.9907546637318964e-06,
+      "loss": 1.39512596,
+      "memory(GiB)": 123.79,
+      "step": 51250,
+      "train_speed(iter/s)": 1.315812
+    },
+    {
+      "acc": 0.69715891,
+      "epoch": 1.3002283105022832,
+      "grad_norm": 4.125,
+      "learning_rate": 2.98979447701838e-06,
+      "loss": 1.38848524,
+      "memory(GiB)": 123.79,
+      "step": 51255,
+      "train_speed(iter/s)": 1.315826
+    },
+    {
+      "acc": 0.69000673,
+      "epoch": 1.3003551496702181,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.988834378722679e-06,
+      "loss": 1.40893402,
+      "memory(GiB)": 123.79,
+      "step": 51260,
+      "train_speed(iter/s)": 1.31584
+    },
+    {
+      "acc": 0.67955952,
+      "epoch": 1.3004819888381531,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.9878743688870193e-06,
+      "loss": 1.38416624,
+      "memory(GiB)": 123.79,
+      "step": 51265,
+      "train_speed(iter/s)": 1.315853
+    },
+    {
+      "acc": 0.68544369,
+      "epoch": 1.3006088280060883,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.9869144475536306e-06,
+      "loss": 1.39540215,
+      "memory(GiB)": 123.79,
+      "step": 51270,
+      "train_speed(iter/s)": 1.315866
+    },
+    {
+      "acc": 0.67609034,
+      "epoch": 1.3007356671740233,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.9859546147647316e-06,
+      "loss": 1.48314533,
+      "memory(GiB)": 123.79,
+      "step": 51275,
+      "train_speed(iter/s)": 1.315878
+    },
+    {
+      "acc": 0.68033657,
+      "epoch": 1.3008625063419583,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.984994870562541e-06,
+      "loss": 1.4366478,
+      "memory(GiB)": 123.79,
+      "step": 51280,
+      "train_speed(iter/s)": 1.315891
+    },
+    {
+      "acc": 0.68175502,
+      "epoch": 1.3009893455098935,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.9840352149892703e-06,
+      "loss": 1.38007975,
+      "memory(GiB)": 123.79,
+      "step": 51285,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.69627786,
+      "epoch": 1.3011161846778285,
+      "grad_norm": 3.25,
+      "learning_rate": 2.9830756480871343e-06,
+      "loss": 1.41945114,
+      "memory(GiB)": 123.79,
+      "step": 51290,
+      "train_speed(iter/s)": 1.315917
+    },
+    {
+      "acc": 0.68936691,
+      "epoch": 1.3012430238457635,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.9821161698983347e-06,
+      "loss": 1.41086979,
+      "memory(GiB)": 123.79,
+      "step": 51295,
+      "train_speed(iter/s)": 1.31593
+    },
+    {
+      "acc": 0.68453741,
+      "epoch": 1.3013698630136985,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.9811567804650747e-06,
+      "loss": 1.36874161,
+      "memory(GiB)": 123.79,
+      "step": 51300,
+      "train_speed(iter/s)": 1.315944
+    },
+    {
+      "acc": 0.68822374,
+      "epoch": 1.3014967021816337,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.980197479829554e-06,
+      "loss": 1.3874321,
+      "memory(GiB)": 123.79,
+      "step": 51305,
+      "train_speed(iter/s)": 1.315957
+    },
+    {
+      "acc": 0.69870415,
+      "epoch": 1.3016235413495687,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.9792382680339666e-06,
+      "loss": 1.34825439,
+      "memory(GiB)": 123.79,
+      "step": 51310,
+      "train_speed(iter/s)": 1.31597
+    },
+    {
+      "acc": 0.67699599,
+      "epoch": 1.301750380517504,
+      "grad_norm": 3.375,
+      "learning_rate": 2.9782791451205e-06,
+      "loss": 1.40640306,
+      "memory(GiB)": 123.79,
+      "step": 51315,
+      "train_speed(iter/s)": 1.315983
+    },
+    {
+      "acc": 0.66457844,
+      "epoch": 1.301877219685439,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.9773201111313444e-06,
+      "loss": 1.44783449,
+      "memory(GiB)": 123.79,
+      "step": 51320,
+      "train_speed(iter/s)": 1.315996
+    },
+    {
+      "acc": 0.69665499,
+      "epoch": 1.302004058853374,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.9763611661086806e-06,
+      "loss": 1.3841897,
+      "memory(GiB)": 123.79,
+      "step": 51325,
+      "train_speed(iter/s)": 1.316009
+    },
+    {
+      "acc": 0.70393915,
+      "epoch": 1.3021308980213089,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.975402310094689e-06,
+      "loss": 1.34063654,
+      "memory(GiB)": 123.79,
+      "step": 51330,
+      "train_speed(iter/s)": 1.316023
+    },
+    {
+      "acc": 0.67695918,
+      "epoch": 1.302257737189244,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.9744435431315403e-06,
+      "loss": 1.4083499,
+      "memory(GiB)": 123.79,
+      "step": 51335,
+      "train_speed(iter/s)": 1.316037
+    },
+    {
+      "acc": 0.67411308,
+      "epoch": 1.302384576357179,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.9734848652614097e-06,
+      "loss": 1.47034101,
+      "memory(GiB)": 123.79,
+      "step": 51340,
+      "train_speed(iter/s)": 1.316051
+    },
+    {
+      "acc": 0.69708314,
+      "epoch": 1.3025114155251143,
+      "grad_norm": 3.5,
+      "learning_rate": 2.972526276526461e-06,
+      "loss": 1.39351559,
+      "memory(GiB)": 123.79,
+      "step": 51345,
+      "train_speed(iter/s)": 1.316064
+    },
+    {
+      "acc": 0.70033169,
+      "epoch": 1.3026382546930493,
+      "grad_norm": 3.5,
+      "learning_rate": 2.971567776968861e-06,
+      "loss": 1.35356741,
+      "memory(GiB)": 123.79,
+      "step": 51350,
+      "train_speed(iter/s)": 1.316077
+    },
+    {
+      "acc": 0.70498886,
+      "epoch": 1.3027650938609843,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.9706093666307654e-06,
+      "loss": 1.35856609,
+      "memory(GiB)": 123.79,
+      "step": 51355,
+      "train_speed(iter/s)": 1.316091
+    },
+    {
+      "acc": 0.6833941,
+      "epoch": 1.3028919330289193,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.969651045554329e-06,
+      "loss": 1.38425961,
+      "memory(GiB)": 123.79,
+      "step": 51360,
+      "train_speed(iter/s)": 1.316105
+    },
+    {
+      "acc": 0.68422375,
+      "epoch": 1.3030187721968545,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.968692813781706e-06,
+      "loss": 1.40608339,
+      "memory(GiB)": 123.79,
+      "step": 51365,
+      "train_speed(iter/s)": 1.316119
+    },
+    {
+      "acc": 0.6752058,
+      "epoch": 1.3031456113647895,
+      "grad_norm": 3.25,
+      "learning_rate": 2.9677346713550437e-06,
+      "loss": 1.40216293,
+      "memory(GiB)": 123.79,
+      "step": 51370,
+      "train_speed(iter/s)": 1.316133
+    },
+    {
+      "acc": 0.70362239,
+      "epoch": 1.3032724505327244,
+      "grad_norm": 2.3125,
+      "learning_rate": 2.966776618316482e-06,
+      "loss": 1.31744213,
+      "memory(GiB)": 123.79,
+      "step": 51375,
+      "train_speed(iter/s)": 1.316147
+    },
+    {
+      "acc": 0.68093014,
+      "epoch": 1.3033992897006597,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.9658186547081612e-06,
+      "loss": 1.42307539,
+      "memory(GiB)": 123.79,
+      "step": 51380,
+      "train_speed(iter/s)": 1.31616
+    },
+    {
+      "acc": 0.69912176,
+      "epoch": 1.3035261288685946,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.9648607805722197e-06,
+      "loss": 1.34896755,
+      "memory(GiB)": 123.79,
+      "step": 51385,
+      "train_speed(iter/s)": 1.316173
+    },
+    {
+      "acc": 0.6928669,
+      "epoch": 1.3036529680365296,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.963902995950788e-06,
+      "loss": 1.34664783,
+      "memory(GiB)": 123.79,
+      "step": 51390,
+      "train_speed(iter/s)": 1.316187
+    },
+    {
+      "acc": 0.67922821,
+      "epoch": 1.3037798072044646,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.962945300885991e-06,
+      "loss": 1.43638592,
+      "memory(GiB)": 123.79,
+      "step": 51395,
+      "train_speed(iter/s)": 1.316201
+    },
+    {
+      "acc": 0.68692398,
+      "epoch": 1.3039066463723998,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.9619876954199557e-06,
+      "loss": 1.35723219,
+      "memory(GiB)": 123.79,
+      "step": 51400,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.70281816,
+      "epoch": 1.3040334855403348,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.9610301795947992e-06,
+      "loss": 1.32908306,
+      "memory(GiB)": 123.79,
+      "step": 51405,
+      "train_speed(iter/s)": 1.316227
+    },
+    {
+      "acc": 0.67699938,
+      "epoch": 1.30416032470827,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.9600727534526417e-06,
+      "loss": 1.43352852,
+      "memory(GiB)": 123.79,
+      "step": 51410,
+      "train_speed(iter/s)": 1.316241
+    },
+    {
+      "acc": 0.68762178,
+      "epoch": 1.304287163876205,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.9591154170355895e-06,
+      "loss": 1.47830639,
+      "memory(GiB)": 123.79,
+      "step": 51415,
+      "train_speed(iter/s)": 1.316255
+    },
+    {
+      "acc": 0.6698905,
+      "epoch": 1.30441400304414,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.9581581703857545e-06,
+      "loss": 1.4466074,
+      "memory(GiB)": 123.79,
+      "step": 51420,
+      "train_speed(iter/s)": 1.316268
+    },
+    {
+      "acc": 0.68843069,
+      "epoch": 1.304540842212075,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.9572010135452377e-06,
+      "loss": 1.34955177,
+      "memory(GiB)": 123.79,
+      "step": 51425,
+      "train_speed(iter/s)": 1.316281
+    },
+    {
+      "acc": 0.69389844,
+      "epoch": 1.3046676813800102,
+      "grad_norm": 4.375,
+      "learning_rate": 2.9562439465561425e-06,
+      "loss": 1.41687107,
+      "memory(GiB)": 123.79,
+      "step": 51430,
+      "train_speed(iter/s)": 1.316295
+    },
+    {
+      "acc": 0.69343624,
+      "epoch": 1.3047945205479452,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.955286969460563e-06,
+      "loss": 1.35065613,
+      "memory(GiB)": 123.79,
+      "step": 51435,
+      "train_speed(iter/s)": 1.316307
+    },
+    {
+      "acc": 0.67917347,
+      "epoch": 1.3049213597158802,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.9543300823005903e-06,
+      "loss": 1.45483437,
+      "memory(GiB)": 123.79,
+      "step": 51440,
+      "train_speed(iter/s)": 1.31632
+    },
+    {
+      "acc": 0.67050853,
+      "epoch": 1.3050481988838154,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.953373285118315e-06,
+      "loss": 1.40040293,
+      "memory(GiB)": 123.79,
+      "step": 51445,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.6784296,
+      "epoch": 1.3051750380517504,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.9524165779558206e-06,
+      "loss": 1.43199129,
+      "memory(GiB)": 123.79,
+      "step": 51450,
+      "train_speed(iter/s)": 1.316347
+    },
+    {
+      "acc": 0.68837404,
+      "epoch": 1.3053018772196854,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.9514599608551865e-06,
+      "loss": 1.38478107,
+      "memory(GiB)": 123.79,
+      "step": 51455,
+      "train_speed(iter/s)": 1.316359
+    },
+    {
+      "acc": 0.68288021,
+      "epoch": 1.3054287163876204,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.9505034338584882e-06,
+      "loss": 1.47642202,
+      "memory(GiB)": 123.79,
+      "step": 51460,
+      "train_speed(iter/s)": 1.316372
+    },
+    {
+      "acc": 0.7079906,
+      "epoch": 1.3055555555555556,
+      "grad_norm": 3.0,
+      "learning_rate": 2.9495469970078e-06,
+      "loss": 1.38053017,
+      "memory(GiB)": 123.79,
+      "step": 51465,
+      "train_speed(iter/s)": 1.316385
+    },
+    {
+      "acc": 0.67833533,
+      "epoch": 1.3056823947234906,
+      "grad_norm": 3.125,
+      "learning_rate": 2.9485906503451907e-06,
+      "loss": 1.46272659,
+      "memory(GiB)": 123.79,
+      "step": 51470,
+      "train_speed(iter/s)": 1.316398
+    },
+    {
+      "acc": 0.67378807,
+      "epoch": 1.3058092338914258,
+      "grad_norm": 2.875,
+      "learning_rate": 2.9476343939127217e-06,
+      "loss": 1.40789661,
+      "memory(GiB)": 123.79,
+      "step": 51475,
+      "train_speed(iter/s)": 1.316411
+    },
+    {
+      "acc": 0.7006278,
+      "epoch": 1.3059360730593608,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.9466782277524554e-06,
+      "loss": 1.29343071,
+      "memory(GiB)": 123.79,
+      "step": 51480,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.68178072,
+      "epoch": 1.3060629122272958,
+      "grad_norm": 5.5,
+      "learning_rate": 2.9457221519064477e-06,
+      "loss": 1.4539156,
+      "memory(GiB)": 123.79,
+      "step": 51485,
+      "train_speed(iter/s)": 1.316438
+    },
+    {
+      "acc": 0.6943079,
+      "epoch": 1.3061897513952307,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.944766166416754e-06,
+      "loss": 1.37872553,
+      "memory(GiB)": 123.79,
+      "step": 51490,
+      "train_speed(iter/s)": 1.31645
+    },
+    {
+      "acc": 0.68344541,
+      "epoch": 1.306316590563166,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.943810271325418e-06,
+      "loss": 1.39360571,
+      "memory(GiB)": 123.79,
+      "step": 51495,
+      "train_speed(iter/s)": 1.316462
+    },
+    {
+      "acc": 0.68019648,
+      "epoch": 1.306443429731101,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.9428544666744873e-06,
+      "loss": 1.37045975,
+      "memory(GiB)": 123.79,
+      "step": 51500,
+      "train_speed(iter/s)": 1.316475
+    },
+    {
+      "acc": 0.67614851,
+      "epoch": 1.3065702688990362,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.9418987525060004e-06,
+      "loss": 1.49906769,
+      "memory(GiB)": 123.79,
+      "step": 51505,
+      "train_speed(iter/s)": 1.316488
+    },
+    {
+      "acc": 0.6958684,
+      "epoch": 1.3066971080669711,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.9409431288619973e-06,
+      "loss": 1.41059589,
+      "memory(GiB)": 123.79,
+      "step": 51510,
+      "train_speed(iter/s)": 1.3165
+    },
+    {
+      "acc": 0.67219439,
+      "epoch": 1.3068239472349061,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.939987595784507e-06,
+      "loss": 1.44730148,
+      "memory(GiB)": 123.79,
+      "step": 51515,
+      "train_speed(iter/s)": 1.316514
+    },
+    {
+      "acc": 0.68489528,
+      "epoch": 1.3069507864028411,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.9390321533155585e-06,
+      "loss": 1.48603535,
+      "memory(GiB)": 123.79,
+      "step": 51520,
+      "train_speed(iter/s)": 1.316526
+    },
+    {
+      "acc": 0.69480534,
+      "epoch": 1.3070776255707763,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.9380768014971794e-06,
+      "loss": 1.42535763,
+      "memory(GiB)": 123.79,
+      "step": 51525,
+      "train_speed(iter/s)": 1.316539
+    },
+    {
+      "acc": 0.70161085,
+      "epoch": 1.3072044647387113,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.937121540371389e-06,
+      "loss": 1.31185989,
+      "memory(GiB)": 123.79,
+      "step": 51530,
+      "train_speed(iter/s)": 1.316553
+    },
+    {
+      "acc": 0.69368067,
+      "epoch": 1.3073313039066463,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.9361663699802007e-06,
+      "loss": 1.40655003,
+      "memory(GiB)": 123.79,
+      "step": 51535,
+      "train_speed(iter/s)": 1.316566
+    },
+    {
+      "acc": 0.68537698,
+      "epoch": 1.3074581430745815,
+      "grad_norm": 3.125,
+      "learning_rate": 2.9352112903656315e-06,
+      "loss": 1.38037357,
+      "memory(GiB)": 123.79,
+      "step": 51540,
+      "train_speed(iter/s)": 1.31658
+    },
+    {
+      "acc": 0.68321619,
+      "epoch": 1.3075849822425165,
+      "grad_norm": 2.875,
+      "learning_rate": 2.9342563015696866e-06,
+      "loss": 1.48945866,
+      "memory(GiB)": 123.79,
+      "step": 51545,
+      "train_speed(iter/s)": 1.316594
+    },
+    {
+      "acc": 0.70756745,
+      "epoch": 1.3077118214104515,
+      "grad_norm": 3.25,
+      "learning_rate": 2.9333014036343765e-06,
+      "loss": 1.32381401,
+      "memory(GiB)": 123.79,
+      "step": 51550,
+      "train_speed(iter/s)": 1.316609
+    },
+    {
+      "acc": 0.68653126,
+      "epoch": 1.3078386605783865,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.932346596601694e-06,
+      "loss": 1.37032356,
+      "memory(GiB)": 123.79,
+      "step": 51555,
+      "train_speed(iter/s)": 1.316623
+    },
+    {
+      "acc": 0.67875509,
+      "epoch": 1.3079654997463217,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.931391880513641e-06,
+      "loss": 1.38046551,
+      "memory(GiB)": 123.79,
+      "step": 51560,
+      "train_speed(iter/s)": 1.316637
+    },
+    {
+      "acc": 0.67478971,
+      "epoch": 1.3080923389142567,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.9304372554122074e-06,
+      "loss": 1.46071873,
+      "memory(GiB)": 123.79,
+      "step": 51565,
+      "train_speed(iter/s)": 1.316651
+    },
+    {
+      "acc": 0.67710619,
+      "epoch": 1.308219178082192,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.9294827213393857e-06,
+      "loss": 1.42307053,
+      "memory(GiB)": 123.79,
+      "step": 51570,
+      "train_speed(iter/s)": 1.316665
+    },
+    {
+      "acc": 0.69887562,
+      "epoch": 1.308346017250127,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.9285282783371567e-06,
+      "loss": 1.3136034,
+      "memory(GiB)": 123.79,
+      "step": 51575,
+      "train_speed(iter/s)": 1.316679
+    },
+    {
+      "acc": 0.68885365,
+      "epoch": 1.3084728564180619,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.9275739264475013e-06,
+      "loss": 1.36433516,
+      "memory(GiB)": 123.79,
+      "step": 51580,
+      "train_speed(iter/s)": 1.316693
+    },
+    {
+      "acc": 0.68468618,
+      "epoch": 1.3085996955859969,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.926619665712399e-06,
+      "loss": 1.35429535,
+      "memory(GiB)": 123.79,
+      "step": 51585,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.67780991,
+      "epoch": 1.308726534753932,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.9256654961738217e-06,
+      "loss": 1.41925526,
+      "memory(GiB)": 123.79,
+      "step": 51590,
+      "train_speed(iter/s)": 1.316718
+    },
+    {
+      "acc": 0.69920158,
+      "epoch": 1.308853373921867,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.9247114178737356e-06,
+      "loss": 1.32093353,
+      "memory(GiB)": 123.79,
+      "step": 51595,
+      "train_speed(iter/s)": 1.316731
+    },
+    {
+      "acc": 0.68491774,
+      "epoch": 1.308980213089802,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.9237574308541063e-06,
+      "loss": 1.40622463,
+      "memory(GiB)": 123.79,
+      "step": 51600,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.68101568,
+      "epoch": 1.3091070522577373,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.9228035351568955e-06,
+      "loss": 1.41004801,
+      "memory(GiB)": 123.79,
+      "step": 51605,
+      "train_speed(iter/s)": 1.316756
+    },
+    {
+      "acc": 0.70197897,
+      "epoch": 1.3092338914256723,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.92184973082406e-06,
+      "loss": 1.26596279,
+      "memory(GiB)": 123.79,
+      "step": 51610,
+      "train_speed(iter/s)": 1.316768
+    },
+    {
+      "acc": 0.68693247,
+      "epoch": 1.3093607305936072,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.920896017897551e-06,
+      "loss": 1.34906712,
+      "memory(GiB)": 123.79,
+      "step": 51615,
+      "train_speed(iter/s)": 1.316781
+    },
+    {
+      "acc": 0.67879963,
+      "epoch": 1.3094875697615422,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.9199423964193176e-06,
+      "loss": 1.39209938,
+      "memory(GiB)": 123.79,
+      "step": 51620,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.68716378,
+      "epoch": 1.3096144089294774,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.9189888664313045e-06,
+      "loss": 1.38035116,
+      "memory(GiB)": 123.79,
+      "step": 51625,
+      "train_speed(iter/s)": 1.316809
+    },
+    {
+      "acc": 0.68621626,
+      "epoch": 1.3097412480974124,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.9180354279754517e-06,
+      "loss": 1.40311928,
+      "memory(GiB)": 123.79,
+      "step": 51630,
+      "train_speed(iter/s)": 1.316823
+    },
+    {
+      "acc": 0.6897356,
+      "epoch": 1.3098680872653476,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.9170820810936968e-06,
+      "loss": 1.37581062,
+      "memory(GiB)": 123.79,
+      "step": 51635,
+      "train_speed(iter/s)": 1.316837
+    },
+    {
+      "acc": 0.68099718,
+      "epoch": 1.3099949264332826,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.9161288258279715e-06,
+      "loss": 1.39937916,
+      "memory(GiB)": 123.79,
+      "step": 51640,
+      "train_speed(iter/s)": 1.316852
+    },
+    {
+      "acc": 0.68567476,
+      "epoch": 1.3101217656012176,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.9151756622202037e-06,
+      "loss": 1.39833679,
+      "memory(GiB)": 123.79,
+      "step": 51645,
+      "train_speed(iter/s)": 1.316866
+    },
+    {
+      "acc": 0.68069587,
+      "epoch": 1.3102486047691526,
+      "grad_norm": 4.96875,
+      "learning_rate": 2.914222590312319e-06,
+      "loss": 1.40492668,
+      "memory(GiB)": 123.79,
+      "step": 51650,
+      "train_speed(iter/s)": 1.31688
+    },
+    {
+      "acc": 0.67799549,
+      "epoch": 1.3103754439370878,
+      "grad_norm": 4.5,
+      "learning_rate": 2.9132696101462366e-06,
+      "loss": 1.40147533,
+      "memory(GiB)": 123.79,
+      "step": 51655,
+      "train_speed(iter/s)": 1.316896
+    },
+    {
+      "acc": 0.67159057,
+      "epoch": 1.3105022831050228,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.912316721763874e-06,
+      "loss": 1.48790054,
+      "memory(GiB)": 123.79,
+      "step": 51660,
+      "train_speed(iter/s)": 1.31691
+    },
+    {
+      "acc": 0.70285597,
+      "epoch": 1.310629122272958,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.9113639252071395e-06,
+      "loss": 1.28495178,
+      "memory(GiB)": 123.79,
+      "step": 51665,
+      "train_speed(iter/s)": 1.316925
+    },
+    {
+      "acc": 0.68226581,
+      "epoch": 1.310755961440893,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.91041122051795e-06,
+      "loss": 1.39442463,
+      "memory(GiB)": 123.79,
+      "step": 51670,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.67341194,
+      "epoch": 1.310882800608828,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.9094586077382016e-06,
+      "loss": 1.42815056,
+      "memory(GiB)": 123.79,
+      "step": 51675,
+      "train_speed(iter/s)": 1.316954
+    },
+    {
+      "acc": 0.68180962,
+      "epoch": 1.311009639776763,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.9085060869097977e-06,
+      "loss": 1.39236422,
+      "memory(GiB)": 123.79,
+      "step": 51680,
+      "train_speed(iter/s)": 1.316968
+    },
+    {
+      "acc": 0.69159341,
+      "epoch": 1.3111364789446982,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.907553658074631e-06,
+      "loss": 1.30911751,
+      "memory(GiB)": 123.79,
+      "step": 51685,
+      "train_speed(iter/s)": 1.316981
+    },
+    {
+      "acc": 0.67791238,
+      "epoch": 1.3112633181126332,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.906601321274601e-06,
+      "loss": 1.42874279,
+      "memory(GiB)": 123.79,
+      "step": 51690,
+      "train_speed(iter/s)": 1.316994
+    },
+    {
+      "acc": 0.69077921,
+      "epoch": 1.3113901572805682,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.90564907655159e-06,
+      "loss": 1.42081642,
+      "memory(GiB)": 123.79,
+      "step": 51695,
+      "train_speed(iter/s)": 1.317007
+    },
+    {
+      "acc": 0.69574618,
+      "epoch": 1.3115169964485034,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.9046969239474808e-06,
+      "loss": 1.33189316,
+      "memory(GiB)": 123.79,
+      "step": 51700,
+      "train_speed(iter/s)": 1.31702
+    },
+    {
+      "acc": 0.68475418,
+      "epoch": 1.3116438356164384,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.9037448635041574e-06,
+      "loss": 1.3830904,
+      "memory(GiB)": 123.79,
+      "step": 51705,
+      "train_speed(iter/s)": 1.317033
+    },
+    {
+      "acc": 0.68495293,
+      "epoch": 1.3117706747843734,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.9027928952634964e-06,
+      "loss": 1.36726046,
+      "memory(GiB)": 123.79,
+      "step": 51710,
+      "train_speed(iter/s)": 1.317046
+    },
+    {
+      "acc": 0.69661674,
+      "epoch": 1.3118975139523084,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.901841019267363e-06,
+      "loss": 1.31627913,
+      "memory(GiB)": 123.79,
+      "step": 51715,
+      "train_speed(iter/s)": 1.317059
+    },
+    {
+      "acc": 0.69039402,
+      "epoch": 1.3120243531202436,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.900889235557631e-06,
+      "loss": 1.41589365,
+      "memory(GiB)": 123.79,
+      "step": 51720,
+      "train_speed(iter/s)": 1.317072
+    },
+    {
+      "acc": 0.69703655,
+      "epoch": 1.3121511922881786,
+      "grad_norm": 3.0,
+      "learning_rate": 2.8999375441761627e-06,
+      "loss": 1.35728073,
+      "memory(GiB)": 123.79,
+      "step": 51725,
+      "train_speed(iter/s)": 1.317085
+    },
+    {
+      "acc": 0.69157619,
+      "epoch": 1.3122780314561138,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8989859451648193e-06,
+      "loss": 1.34618702,
+      "memory(GiB)": 123.79,
+      "step": 51730,
+      "train_speed(iter/s)": 1.317099
+    },
+    {
+      "acc": 0.69639044,
+      "epoch": 1.3124048706240488,
+      "grad_norm": 4.0,
+      "learning_rate": 2.89803443856545e-06,
+      "loss": 1.31489868,
+      "memory(GiB)": 123.79,
+      "step": 51735,
+      "train_speed(iter/s)": 1.317109
+    },
+    {
+      "acc": 0.68362541,
+      "epoch": 1.3125317097919837,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.897083024419913e-06,
+      "loss": 1.4210537,
+      "memory(GiB)": 123.79,
+      "step": 51740,
+      "train_speed(iter/s)": 1.317122
+    },
+    {
+      "acc": 0.68635163,
+      "epoch": 1.3126585489599187,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.8961317027700534e-06,
+      "loss": 1.36939468,
+      "memory(GiB)": 123.79,
+      "step": 51745,
+      "train_speed(iter/s)": 1.317135
+    },
+    {
+      "acc": 0.68372383,
+      "epoch": 1.312785388127854,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.8951804736577148e-06,
+      "loss": 1.40163908,
+      "memory(GiB)": 123.79,
+      "step": 51750,
+      "train_speed(iter/s)": 1.317147
+    },
+    {
+      "acc": 0.70055828,
+      "epoch": 1.312912227295789,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.894229337124736e-06,
+      "loss": 1.39590368,
+      "memory(GiB)": 123.79,
+      "step": 51755,
+      "train_speed(iter/s)": 1.317161
+    },
+    {
+      "acc": 0.67040343,
+      "epoch": 1.313039066463724,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.8932782932129524e-06,
+      "loss": 1.41568279,
+      "memory(GiB)": 123.79,
+      "step": 51760,
+      "train_speed(iter/s)": 1.317175
+    },
+    {
+      "acc": 0.69429045,
+      "epoch": 1.3131659056316591,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.8923273419641956e-06,
+      "loss": 1.30204086,
+      "memory(GiB)": 123.79,
+      "step": 51765,
+      "train_speed(iter/s)": 1.317188
+    },
+    {
+      "acc": 0.68835244,
+      "epoch": 1.3132927447995941,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.891376483420292e-06,
+      "loss": 1.3779458,
+      "memory(GiB)": 123.79,
+      "step": 51770,
+      "train_speed(iter/s)": 1.317201
+    },
+    {
+      "acc": 0.68862581,
+      "epoch": 1.3134195839675291,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.8904257176230655e-06,
+      "loss": 1.35679131,
+      "memory(GiB)": 123.79,
+      "step": 51775,
+      "train_speed(iter/s)": 1.317214
+    },
+    {
+      "acc": 0.69881763,
+      "epoch": 1.313546423135464,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.8894750446143345e-06,
+      "loss": 1.311623,
+      "memory(GiB)": 123.79,
+      "step": 51780,
+      "train_speed(iter/s)": 1.317227
+    },
+    {
+      "acc": 0.67789507,
+      "epoch": 1.3136732623033993,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.8885244644359134e-06,
+      "loss": 1.39413948,
+      "memory(GiB)": 123.79,
+      "step": 51785,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.68009152,
+      "epoch": 1.3138001014713343,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.887573977129614e-06,
+      "loss": 1.42859545,
+      "memory(GiB)": 123.79,
+      "step": 51790,
+      "train_speed(iter/s)": 1.317254
+    },
+    {
+      "acc": 0.69931822,
+      "epoch": 1.3139269406392695,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.886623582737242e-06,
+      "loss": 1.36566648,
+      "memory(GiB)": 123.79,
+      "step": 51795,
+      "train_speed(iter/s)": 1.317266
+    },
+    {
+      "acc": 0.68675961,
+      "epoch": 1.3140537798072045,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.8856732813006007e-06,
+      "loss": 1.43733025,
+      "memory(GiB)": 123.79,
+      "step": 51800,
+      "train_speed(iter/s)": 1.31728
+    },
+    {
+      "acc": 0.68043494,
+      "epoch": 1.3141806189751395,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8847230728614854e-06,
+      "loss": 1.4684185,
+      "memory(GiB)": 123.79,
+      "step": 51805,
+      "train_speed(iter/s)": 1.317292
+    },
+    {
+      "acc": 0.7053853,
+      "epoch": 1.3143074581430745,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.883772957461698e-06,
+      "loss": 1.29718447,
+      "memory(GiB)": 123.79,
+      "step": 51810,
+      "train_speed(iter/s)": 1.317304
+    },
+    {
+      "acc": 0.6767087,
+      "epoch": 1.3144342973110097,
+      "grad_norm": 4.0,
+      "learning_rate": 2.8828229351430224e-06,
+      "loss": 1.4553154,
+      "memory(GiB)": 123.79,
+      "step": 51815,
+      "train_speed(iter/s)": 1.317318
+    },
+    {
+      "acc": 0.6955554,
+      "epoch": 1.3145611364789447,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.881873005947247e-06,
+      "loss": 1.3644886,
+      "memory(GiB)": 123.79,
+      "step": 51820,
+      "train_speed(iter/s)": 1.317331
+    },
+    {
+      "acc": 0.68134923,
+      "epoch": 1.31468797564688,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.88092316991615e-06,
+      "loss": 1.412041,
+      "memory(GiB)": 123.79,
+      "step": 51825,
+      "train_speed(iter/s)": 1.317345
+    },
+    {
+      "acc": 0.69630027,
+      "epoch": 1.3148148148148149,
+      "grad_norm": 3.5,
+      "learning_rate": 2.879973427091518e-06,
+      "loss": 1.35848103,
+      "memory(GiB)": 123.79,
+      "step": 51830,
+      "train_speed(iter/s)": 1.317358
+    },
+    {
+      "acc": 0.68976493,
+      "epoch": 1.3149416539827499,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.879023777515118e-06,
+      "loss": 1.34791241,
+      "memory(GiB)": 123.79,
+      "step": 51835,
+      "train_speed(iter/s)": 1.317372
+    },
+    {
+      "acc": 0.68058319,
+      "epoch": 1.3150684931506849,
+      "grad_norm": 2.75,
+      "learning_rate": 2.8780742212287192e-06,
+      "loss": 1.37314472,
+      "memory(GiB)": 123.79,
+      "step": 51840,
+      "train_speed(iter/s)": 1.317386
+    },
+    {
+      "acc": 0.67521768,
+      "epoch": 1.31519533231862,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.8771247582740924e-06,
+      "loss": 1.43070507,
+      "memory(GiB)": 123.79,
+      "step": 51845,
+      "train_speed(iter/s)": 1.3174
+    },
+    {
+      "acc": 0.6885684,
+      "epoch": 1.315322171486555,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.876175388692999e-06,
+      "loss": 1.37378731,
+      "memory(GiB)": 123.79,
+      "step": 51850,
+      "train_speed(iter/s)": 1.317413
+    },
+    {
+      "acc": 0.68127813,
+      "epoch": 1.31544901065449,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.875226112527192e-06,
+      "loss": 1.46218519,
+      "memory(GiB)": 123.79,
+      "step": 51855,
+      "train_speed(iter/s)": 1.317427
+    },
+    {
+      "acc": 0.68930798,
+      "epoch": 1.3155758498224253,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.8742769298184246e-06,
+      "loss": 1.40834675,
+      "memory(GiB)": 123.79,
+      "step": 51860,
+      "train_speed(iter/s)": 1.31744
+    },
+    {
+      "acc": 0.68824463,
+      "epoch": 1.3157026889903602,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.8733278406084507e-06,
+      "loss": 1.40348816,
+      "memory(GiB)": 123.79,
+      "step": 51865,
+      "train_speed(iter/s)": 1.317453
+    },
+    {
+      "acc": 0.6852057,
+      "epoch": 1.3158295281582952,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.872378844939015e-06,
+      "loss": 1.39827099,
+      "memory(GiB)": 123.79,
+      "step": 51870,
+      "train_speed(iter/s)": 1.317466
+    },
+    {
+      "acc": 0.69061899,
+      "epoch": 1.3159563673262302,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.871429942851853e-06,
+      "loss": 1.43699064,
+      "memory(GiB)": 123.79,
+      "step": 51875,
+      "train_speed(iter/s)": 1.317479
+    },
+    {
+      "acc": 0.70256896,
+      "epoch": 1.3160832064941654,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.8704811343887075e-06,
+      "loss": 1.30971603,
+      "memory(GiB)": 123.79,
+      "step": 51880,
+      "train_speed(iter/s)": 1.317492
+    },
+    {
+      "acc": 0.697118,
+      "epoch": 1.3162100456621004,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.86953241959131e-06,
+      "loss": 1.37755842,
+      "memory(GiB)": 123.79,
+      "step": 51885,
+      "train_speed(iter/s)": 1.317504
+    },
+    {
+      "acc": 0.69469519,
+      "epoch": 1.3163368848300356,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8685837985013874e-06,
+      "loss": 1.35584164,
+      "memory(GiB)": 123.79,
+      "step": 51890,
+      "train_speed(iter/s)": 1.317517
+    },
+    {
+      "acc": 0.69711342,
+      "epoch": 1.3164637239979706,
+      "grad_norm": 3.375,
+      "learning_rate": 2.867635271160666e-06,
+      "loss": 1.3523243,
+      "memory(GiB)": 123.79,
+      "step": 51895,
+      "train_speed(iter/s)": 1.31753
+    },
+    {
+      "acc": 0.67565084,
+      "epoch": 1.3165905631659056,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.8666868376108658e-06,
+      "loss": 1.51250839,
+      "memory(GiB)": 123.79,
+      "step": 51900,
+      "train_speed(iter/s)": 1.317543
+    },
+    {
+      "acc": 0.68315401,
+      "epoch": 1.3167174023338406,
+      "grad_norm": 3.625,
+      "learning_rate": 2.865738497893703e-06,
+      "loss": 1.37322845,
+      "memory(GiB)": 123.79,
+      "step": 51905,
+      "train_speed(iter/s)": 1.317554
+    },
+    {
+      "acc": 0.68113871,
+      "epoch": 1.3168442415017758,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.8647902520508896e-06,
+      "loss": 1.37749481,
+      "memory(GiB)": 123.79,
+      "step": 51910,
+      "train_speed(iter/s)": 1.317566
+    },
+    {
+      "acc": 0.68111157,
+      "epoch": 1.3169710806697108,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.8638421001241346e-06,
+      "loss": 1.4174118,
+      "memory(GiB)": 123.79,
+      "step": 51915,
+      "train_speed(iter/s)": 1.317578
+    },
+    {
+      "acc": 0.68603735,
+      "epoch": 1.3170979198376458,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.8628940421551404e-06,
+      "loss": 1.40427818,
+      "memory(GiB)": 123.79,
+      "step": 51920,
+      "train_speed(iter/s)": 1.31759
+    },
+    {
+      "acc": 0.67365198,
+      "epoch": 1.317224759005581,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.861946078185608e-06,
+      "loss": 1.46111231,
+      "memory(GiB)": 123.79,
+      "step": 51925,
+      "train_speed(iter/s)": 1.317603
+    },
+    {
+      "acc": 0.69016886,
+      "epoch": 1.317351598173516,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.860998208257233e-06,
+      "loss": 1.35820408,
+      "memory(GiB)": 123.79,
+      "step": 51930,
+      "train_speed(iter/s)": 1.317616
+    },
+    {
+      "acc": 0.67577305,
+      "epoch": 1.317478437341451,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.860050432411707e-06,
+      "loss": 1.44253635,
+      "memory(GiB)": 123.79,
+      "step": 51935,
+      "train_speed(iter/s)": 1.317628
+    },
+    {
+      "acc": 0.67791052,
+      "epoch": 1.317605276509386,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.8591027506907167e-06,
+      "loss": 1.45078144,
+      "memory(GiB)": 123.79,
+      "step": 51940,
+      "train_speed(iter/s)": 1.317641
+    },
+    {
+      "acc": 0.69667983,
+      "epoch": 1.3177321156773212,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.858155163135946e-06,
+      "loss": 1.39243774,
+      "memory(GiB)": 123.79,
+      "step": 51945,
+      "train_speed(iter/s)": 1.317654
+    },
+    {
+      "acc": 0.67467403,
+      "epoch": 1.3178589548452562,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.857207669789074e-06,
+      "loss": 1.42770529,
+      "memory(GiB)": 123.79,
+      "step": 51950,
+      "train_speed(iter/s)": 1.317667
+    },
+    {
+      "acc": 0.67010832,
+      "epoch": 1.3179857940131914,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.8562602706917754e-06,
+      "loss": 1.43131618,
+      "memory(GiB)": 123.79,
+      "step": 51955,
+      "train_speed(iter/s)": 1.31768
+    },
+    {
+      "acc": 0.69217563,
+      "epoch": 1.3181126331811264,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.8553129658857215e-06,
+      "loss": 1.40090656,
+      "memory(GiB)": 123.79,
+      "step": 51960,
+      "train_speed(iter/s)": 1.317694
+    },
+    {
+      "acc": 0.68102617,
+      "epoch": 1.3182394723490614,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.854365755412576e-06,
+      "loss": 1.37965469,
+      "memory(GiB)": 123.79,
+      "step": 51965,
+      "train_speed(iter/s)": 1.317705
+    },
+    {
+      "acc": 0.69247484,
+      "epoch": 1.3183663115169963,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.8534186393140083e-06,
+      "loss": 1.32502604,
+      "memory(GiB)": 123.79,
+      "step": 51970,
+      "train_speed(iter/s)": 1.317718
+    },
+    {
+      "acc": 0.68345637,
+      "epoch": 1.3184931506849316,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.8524716176316715e-06,
+      "loss": 1.35907688,
+      "memory(GiB)": 123.79,
+      "step": 51975,
+      "train_speed(iter/s)": 1.317729
+    },
+    {
+      "acc": 0.66318083,
+      "epoch": 1.3186199898528665,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.851524690407218e-06,
+      "loss": 1.4982769,
+      "memory(GiB)": 123.79,
+      "step": 51980,
+      "train_speed(iter/s)": 1.317742
+    },
+    {
+      "acc": 0.69560175,
+      "epoch": 1.3187468290208018,
+      "grad_norm": 2.5,
+      "learning_rate": 2.8505778576823036e-06,
+      "loss": 1.40236254,
+      "memory(GiB)": 123.79,
+      "step": 51985,
+      "train_speed(iter/s)": 1.317755
+    },
+    {
+      "acc": 0.68556781,
+      "epoch": 1.3188736681887367,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.849631119498573e-06,
+      "loss": 1.45272846,
+      "memory(GiB)": 123.79,
+      "step": 51990,
+      "train_speed(iter/s)": 1.317768
+    },
+    {
+      "acc": 0.67509241,
+      "epoch": 1.3190005073566717,
+      "grad_norm": 2.875,
+      "learning_rate": 2.8486844758976652e-06,
+      "loss": 1.41575069,
+      "memory(GiB)": 123.79,
+      "step": 51995,
+      "train_speed(iter/s)": 1.317782
+    },
+    {
+      "acc": 0.6890892,
+      "epoch": 1.3191273465246067,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.8477379269212157e-06,
+      "loss": 1.35246906,
+      "memory(GiB)": 123.79,
+      "step": 52000,
+      "train_speed(iter/s)": 1.317797
+    },
+    {
+      "epoch": 1.3191273465246067,
+      "eval_acc": 0.6750127908749376,
+      "eval_loss": 1.3564318418502808,
+      "eval_runtime": 69.9824,
+      "eval_samples_per_second": 91.023,
+      "eval_steps_per_second": 22.763,
+      "step": 52000
+    },
+    {
+      "acc": 0.69056149,
+      "epoch": 1.319254185692542,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.846791472610865e-06,
+      "loss": 1.38662806,
+      "memory(GiB)": 123.79,
+      "step": 52005,
+      "train_speed(iter/s)": 1.315057
+    },
+    {
+      "acc": 0.67639866,
+      "epoch": 1.319381024860477,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.845845113008239e-06,
+      "loss": 1.44246826,
+      "memory(GiB)": 123.79,
+      "step": 52010,
+      "train_speed(iter/s)": 1.315071
+    },
+    {
+      "acc": 0.68495045,
+      "epoch": 1.319507864028412,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.84489884815496e-06,
+      "loss": 1.39037447,
+      "memory(GiB)": 123.79,
+      "step": 52015,
+      "train_speed(iter/s)": 1.315086
+    },
+    {
+      "acc": 0.67363939,
+      "epoch": 1.3196347031963471,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.843952678092653e-06,
+      "loss": 1.37814426,
+      "memory(GiB)": 123.79,
+      "step": 52020,
+      "train_speed(iter/s)": 1.3151
+    },
+    {
+      "acc": 0.67422543,
+      "epoch": 1.3197615423642821,
+      "grad_norm": 3.375,
+      "learning_rate": 2.8430066028629328e-06,
+      "loss": 1.42852907,
+      "memory(GiB)": 123.79,
+      "step": 52025,
+      "train_speed(iter/s)": 1.315114
+    },
+    {
+      "acc": 0.69208984,
+      "epoch": 1.319888381532217,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.842060622507415e-06,
+      "loss": 1.3405652,
+      "memory(GiB)": 123.79,
+      "step": 52030,
+      "train_speed(iter/s)": 1.315129
+    },
+    {
+      "acc": 0.69847207,
+      "epoch": 1.320015220700152,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.841114737067702e-06,
+      "loss": 1.33204517,
+      "memory(GiB)": 123.79,
+      "step": 52035,
+      "train_speed(iter/s)": 1.315143
+    },
+    {
+      "acc": 0.70513539,
+      "epoch": 1.3201420598680873,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.840168946585402e-06,
+      "loss": 1.30893679,
+      "memory(GiB)": 123.79,
+      "step": 52040,
+      "train_speed(iter/s)": 1.315158
+    },
+    {
+      "acc": 0.70496206,
+      "epoch": 1.3202688990360223,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.8392232511021158e-06,
+      "loss": 1.34583874,
+      "memory(GiB)": 123.79,
+      "step": 52045,
+      "train_speed(iter/s)": 1.315171
+    },
+    {
+      "acc": 0.67996178,
+      "epoch": 1.3203957382039575,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.8382776506594385e-06,
+      "loss": 1.42509155,
+      "memory(GiB)": 123.79,
+      "step": 52050,
+      "train_speed(iter/s)": 1.315186
+    },
+    {
+      "acc": 0.68600597,
+      "epoch": 1.3205225773718925,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.837332145298961e-06,
+      "loss": 1.38524675,
+      "memory(GiB)": 123.79,
+      "step": 52055,
+      "train_speed(iter/s)": 1.3152
+    },
+    {
+      "acc": 0.68800001,
+      "epoch": 1.3206494165398275,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.836386735062271e-06,
+      "loss": 1.41822224,
+      "memory(GiB)": 123.79,
+      "step": 52060,
+      "train_speed(iter/s)": 1.315215
+    },
+    {
+      "acc": 0.70401163,
+      "epoch": 1.3207762557077625,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.835441419990953e-06,
+      "loss": 1.32788124,
+      "memory(GiB)": 123.79,
+      "step": 52065,
+      "train_speed(iter/s)": 1.31523
+    },
+    {
+      "acc": 0.67553091,
+      "epoch": 1.3209030948756977,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.834496200126585e-06,
+      "loss": 1.40081539,
+      "memory(GiB)": 123.79,
+      "step": 52070,
+      "train_speed(iter/s)": 1.315245
+    },
+    {
+      "acc": 0.67101707,
+      "epoch": 1.3210299340436327,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.8335510755107426e-06,
+      "loss": 1.47983646,
+      "memory(GiB)": 123.79,
+      "step": 52075,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.70056524,
+      "epoch": 1.3211567732115677,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.8326060461849966e-06,
+      "loss": 1.39145765,
+      "memory(GiB)": 123.79,
+      "step": 52080,
+      "train_speed(iter/s)": 1.315273
+    },
+    {
+      "acc": 0.69278316,
+      "epoch": 1.3212836123795029,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.8316611121909126e-06,
+      "loss": 1.38074245,
+      "memory(GiB)": 123.79,
+      "step": 52085,
+      "train_speed(iter/s)": 1.315286
+    },
+    {
+      "acc": 0.68780508,
+      "epoch": 1.3214104515474379,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.8307162735700544e-06,
+      "loss": 1.48178043,
+      "memory(GiB)": 123.79,
+      "step": 52090,
+      "train_speed(iter/s)": 1.315299
+    },
+    {
+      "acc": 0.69709463,
+      "epoch": 1.3215372907153728,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.8297715303639796e-06,
+      "loss": 1.30964966,
+      "memory(GiB)": 123.79,
+      "step": 52095,
+      "train_speed(iter/s)": 1.315313
+    },
+    {
+      "acc": 0.66800966,
+      "epoch": 1.3216641298833078,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.8288268826142423e-06,
+      "loss": 1.46517506,
+      "memory(GiB)": 123.79,
+      "step": 52100,
+      "train_speed(iter/s)": 1.315327
+    },
+    {
+      "acc": 0.68520889,
+      "epoch": 1.321790969051243,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.8278823303623905e-06,
+      "loss": 1.39057169,
+      "memory(GiB)": 123.79,
+      "step": 52105,
+      "train_speed(iter/s)": 1.315341
+    },
+    {
+      "acc": 0.67915106,
+      "epoch": 1.321917808219178,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.8269378736499754e-06,
+      "loss": 1.43110256,
+      "memory(GiB)": 123.79,
+      "step": 52110,
+      "train_speed(iter/s)": 1.315355
+    },
+    {
+      "acc": 0.68864269,
+      "epoch": 1.3220446473871132,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.8259935125185323e-06,
+      "loss": 1.45848866,
+      "memory(GiB)": 123.79,
+      "step": 52115,
+      "train_speed(iter/s)": 1.315369
+    },
+    {
+      "acc": 0.6745841,
+      "epoch": 1.3221714865550482,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.8250492470096008e-06,
+      "loss": 1.50222569,
+      "memory(GiB)": 123.79,
+      "step": 52120,
+      "train_speed(iter/s)": 1.315383
+    },
+    {
+      "acc": 0.69313684,
+      "epoch": 1.3222983257229832,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.824105077164712e-06,
+      "loss": 1.35686035,
+      "memory(GiB)": 123.79,
+      "step": 52125,
+      "train_speed(iter/s)": 1.315396
+    },
+    {
+      "acc": 0.68311329,
+      "epoch": 1.3224251648909182,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.823161003025401e-06,
+      "loss": 1.41585102,
+      "memory(GiB)": 123.79,
+      "step": 52130,
+      "train_speed(iter/s)": 1.315409
+    },
+    {
+      "acc": 0.68213663,
+      "epoch": 1.3225520040588534,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.822217024633186e-06,
+      "loss": 1.46138487,
+      "memory(GiB)": 123.79,
+      "step": 52135,
+      "train_speed(iter/s)": 1.315422
+    },
+    {
+      "acc": 0.68760839,
+      "epoch": 1.3226788432267884,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.821273142029587e-06,
+      "loss": 1.3996623,
+      "memory(GiB)": 123.79,
+      "step": 52140,
+      "train_speed(iter/s)": 1.315436
+    },
+    {
+      "acc": 0.67086172,
+      "epoch": 1.3228056823947236,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.820329355256124e-06,
+      "loss": 1.40486631,
+      "memory(GiB)": 123.79,
+      "step": 52145,
+      "train_speed(iter/s)": 1.315449
+    },
+    {
+      "acc": 0.68904457,
+      "epoch": 1.3229325215626586,
+      "grad_norm": 3.125,
+      "learning_rate": 2.8193856643543106e-06,
+      "loss": 1.37571268,
+      "memory(GiB)": 123.79,
+      "step": 52150,
+      "train_speed(iter/s)": 1.315461
+    },
+    {
+      "acc": 0.68482041,
+      "epoch": 1.3230593607305936,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.8184420693656468e-06,
+      "loss": 1.40327892,
+      "memory(GiB)": 123.79,
+      "step": 52155,
+      "train_speed(iter/s)": 1.315474
+    },
+    {
+      "acc": 0.6879663,
+      "epoch": 1.3231861998985286,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.817498570331643e-06,
+      "loss": 1.35149002,
+      "memory(GiB)": 123.79,
+      "step": 52160,
+      "train_speed(iter/s)": 1.315487
+    },
+    {
+      "acc": 0.6920156,
+      "epoch": 1.3233130390664638,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.816555167293795e-06,
+      "loss": 1.34224367,
+      "memory(GiB)": 123.79,
+      "step": 52165,
+      "train_speed(iter/s)": 1.3155
+    },
+    {
+      "acc": 0.68776455,
+      "epoch": 1.3234398782343988,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.815611860293603e-06,
+      "loss": 1.44082985,
+      "memory(GiB)": 123.79,
+      "step": 52170,
+      "train_speed(iter/s)": 1.315514
+    },
+    {
+      "acc": 0.69168663,
+      "epoch": 1.3235667174023338,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.814668649372549e-06,
+      "loss": 1.38727608,
+      "memory(GiB)": 123.79,
+      "step": 52175,
+      "train_speed(iter/s)": 1.315527
+    },
+    {
+      "acc": 0.68494925,
+      "epoch": 1.323693556570269,
+      "grad_norm": 3.375,
+      "learning_rate": 2.8137255345721266e-06,
+      "loss": 1.43359947,
+      "memory(GiB)": 123.79,
+      "step": 52180,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.68794608,
+      "epoch": 1.323820395738204,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.8127825159338163e-06,
+      "loss": 1.40852251,
+      "memory(GiB)": 123.79,
+      "step": 52185,
+      "train_speed(iter/s)": 1.315553
+    },
+    {
+      "acc": 0.67642679,
+      "epoch": 1.323947234906139,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.8118395934990962e-06,
+      "loss": 1.46933212,
+      "memory(GiB)": 123.79,
+      "step": 52190,
+      "train_speed(iter/s)": 1.315566
+    },
+    {
+      "acc": 0.67630148,
+      "epoch": 1.324074074074074,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.81089676730944e-06,
+      "loss": 1.41298971,
+      "memory(GiB)": 123.79,
+      "step": 52195,
+      "train_speed(iter/s)": 1.315579
+    },
+    {
+      "acc": 0.69483285,
+      "epoch": 1.3242009132420092,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.8099540374063185e-06,
+      "loss": 1.34244385,
+      "memory(GiB)": 123.79,
+      "step": 52200,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.67765312,
+      "epoch": 1.3243277524099442,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.8090114038311956e-06,
+      "loss": 1.43833303,
+      "memory(GiB)": 123.79,
+      "step": 52205,
+      "train_speed(iter/s)": 1.315604
+    },
+    {
+      "acc": 0.69137163,
+      "epoch": 1.3244545915778794,
+      "grad_norm": 3.375,
+      "learning_rate": 2.8080688666255328e-06,
+      "loss": 1.36296482,
+      "memory(GiB)": 123.79,
+      "step": 52210,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.68432426,
+      "epoch": 1.3245814307458144,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.8071264258307884e-06,
+      "loss": 1.37327518,
+      "memory(GiB)": 123.79,
+      "step": 52215,
+      "train_speed(iter/s)": 1.315624
+    },
+    {
+      "acc": 0.69749174,
+      "epoch": 1.3247082699137493,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.8061840814884133e-06,
+      "loss": 1.35052547,
+      "memory(GiB)": 123.79,
+      "step": 52220,
+      "train_speed(iter/s)": 1.315636
+    },
+    {
+      "acc": 0.70277777,
+      "epoch": 1.3248351090816843,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.805241833639858e-06,
+      "loss": 1.36738663,
+      "memory(GiB)": 123.79,
+      "step": 52225,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.68867912,
+      "epoch": 1.3249619482496195,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.804299682326565e-06,
+      "loss": 1.42226849,
+      "memory(GiB)": 123.79,
+      "step": 52230,
+      "train_speed(iter/s)": 1.315663
+    },
+    {
+      "acc": 0.68784108,
+      "epoch": 1.3250887874175545,
+      "grad_norm": 3.125,
+      "learning_rate": 2.8033576275899752e-06,
+      "loss": 1.42595024,
+      "memory(GiB)": 123.79,
+      "step": 52235,
+      "train_speed(iter/s)": 1.315675
+    },
+    {
+      "acc": 0.68225622,
+      "epoch": 1.3252156265854895,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.8024156694715242e-06,
+      "loss": 1.41932411,
+      "memory(GiB)": 123.79,
+      "step": 52240,
+      "train_speed(iter/s)": 1.315688
+    },
+    {
+      "acc": 0.68133335,
+      "epoch": 1.3253424657534247,
+      "grad_norm": 5.4375,
+      "learning_rate": 2.8014738080126424e-06,
+      "loss": 1.4143549,
+      "memory(GiB)": 123.79,
+      "step": 52245,
+      "train_speed(iter/s)": 1.315702
+    },
+    {
+      "acc": 0.68855147,
+      "epoch": 1.3254693049213597,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.8005320432547612e-06,
+      "loss": 1.37909689,
+      "memory(GiB)": 123.79,
+      "step": 52250,
+      "train_speed(iter/s)": 1.315715
+    },
+    {
+      "acc": 0.67353106,
+      "epoch": 1.3255961440892947,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.7995903752392993e-06,
+      "loss": 1.42325096,
+      "memory(GiB)": 123.79,
+      "step": 52255,
+      "train_speed(iter/s)": 1.315728
+    },
+    {
+      "acc": 0.70194321,
+      "epoch": 1.3257229832572297,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.7986488040076764e-06,
+      "loss": 1.29567528,
+      "memory(GiB)": 123.79,
+      "step": 52260,
+      "train_speed(iter/s)": 1.315741
+    },
+    {
+      "acc": 0.68232632,
+      "epoch": 1.325849822425165,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.797707329601306e-06,
+      "loss": 1.45316267,
+      "memory(GiB)": 123.79,
+      "step": 52265,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.68453836,
+      "epoch": 1.3259766615931,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.7967659520616032e-06,
+      "loss": 1.41194468,
+      "memory(GiB)": 123.79,
+      "step": 52270,
+      "train_speed(iter/s)": 1.315767
+    },
+    {
+      "acc": 0.6894412,
+      "epoch": 1.3261035007610351,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.7958246714299685e-06,
+      "loss": 1.2769701,
+      "memory(GiB)": 123.79,
+      "step": 52275,
+      "train_speed(iter/s)": 1.315779
+    },
+    {
+      "acc": 0.67871408,
+      "epoch": 1.32623033992897,
+      "grad_norm": 2.625,
+      "learning_rate": 2.7948834877478035e-06,
+      "loss": 1.43768568,
+      "memory(GiB)": 123.79,
+      "step": 52280,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.6740149,
+      "epoch": 1.326357179096905,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.7939424010565107e-06,
+      "loss": 1.46065464,
+      "memory(GiB)": 123.79,
+      "step": 52285,
+      "train_speed(iter/s)": 1.315805
+    },
+    {
+      "acc": 0.69023042,
+      "epoch": 1.32648401826484,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.793001411397482e-06,
+      "loss": 1.42140617,
+      "memory(GiB)": 123.79,
+      "step": 52290,
+      "train_speed(iter/s)": 1.315818
+    },
+    {
+      "acc": 0.68713007,
+      "epoch": 1.3266108574327753,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.792060518812103e-06,
+      "loss": 1.43324604,
+      "memory(GiB)": 123.79,
+      "step": 52295,
+      "train_speed(iter/s)": 1.315831
+    },
+    {
+      "acc": 0.69359455,
+      "epoch": 1.3267376966007103,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.7911197233417574e-06,
+      "loss": 1.38897934,
+      "memory(GiB)": 123.79,
+      "step": 52300,
+      "train_speed(iter/s)": 1.315842
+    },
+    {
+      "acc": 0.68230324,
+      "epoch": 1.3268645357686455,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.790179025027831e-06,
+      "loss": 1.41757126,
+      "memory(GiB)": 123.79,
+      "step": 52305,
+      "train_speed(iter/s)": 1.315855
+    },
+    {
+      "acc": 0.67121887,
+      "epoch": 1.3269913749365805,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.789238423911699e-06,
+      "loss": 1.43695869,
+      "memory(GiB)": 123.79,
+      "step": 52310,
+      "train_speed(iter/s)": 1.315868
+    },
+    {
+      "acc": 0.69618077,
+      "epoch": 1.3271182141045155,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.788297920034727e-06,
+      "loss": 1.32963362,
+      "memory(GiB)": 123.79,
+      "step": 52315,
+      "train_speed(iter/s)": 1.315881
+    },
+    {
+      "acc": 0.69617786,
+      "epoch": 1.3272450532724505,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.78735751343829e-06,
+      "loss": 1.39536686,
+      "memory(GiB)": 123.79,
+      "step": 52320,
+      "train_speed(iter/s)": 1.315895
+    },
+    {
+      "acc": 0.68156867,
+      "epoch": 1.3273718924403857,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.786417204163748e-06,
+      "loss": 1.41127605,
+      "memory(GiB)": 123.79,
+      "step": 52325,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.686377,
+      "epoch": 1.3274987316083207,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.7854769922524593e-06,
+      "loss": 1.38984632,
+      "memory(GiB)": 123.79,
+      "step": 52330,
+      "train_speed(iter/s)": 1.31592
+    },
+    {
+      "acc": 0.69540157,
+      "epoch": 1.3276255707762556,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.7845368777457803e-06,
+      "loss": 1.37560158,
+      "memory(GiB)": 123.79,
+      "step": 52335,
+      "train_speed(iter/s)": 1.315934
+    },
+    {
+      "acc": 0.69196777,
+      "epoch": 1.3277524099441909,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.7835968606850616e-06,
+      "loss": 1.39781466,
+      "memory(GiB)": 123.79,
+      "step": 52340,
+      "train_speed(iter/s)": 1.315947
+    },
+    {
+      "acc": 0.67486711,
+      "epoch": 1.3278792491121258,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.782656941111648e-06,
+      "loss": 1.39675665,
+      "memory(GiB)": 123.79,
+      "step": 52345,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.69965143,
+      "epoch": 1.3280060882800608,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.7817171190668812e-06,
+      "loss": 1.30428524,
+      "memory(GiB)": 123.79,
+      "step": 52350,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.68432207,
+      "epoch": 1.3281329274479958,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.7807773945921e-06,
+      "loss": 1.40506573,
+      "memory(GiB)": 123.79,
+      "step": 52355,
+      "train_speed(iter/s)": 1.315985
+    },
+    {
+      "acc": 0.70020366,
+      "epoch": 1.328259766615931,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.7798377677286363e-06,
+      "loss": 1.36089716,
+      "memory(GiB)": 123.79,
+      "step": 52360,
+      "train_speed(iter/s)": 1.315999
+    },
+    {
+      "acc": 0.69100537,
+      "epoch": 1.328386605783866,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.778898238517821e-06,
+      "loss": 1.33872375,
+      "memory(GiB)": 123.79,
+      "step": 52365,
+      "train_speed(iter/s)": 1.316012
+    },
+    {
+      "acc": 0.68433194,
+      "epoch": 1.3285134449518012,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.7779588070009767e-06,
+      "loss": 1.382335,
+      "memory(GiB)": 123.79,
+      "step": 52370,
+      "train_speed(iter/s)": 1.316025
+    },
+    {
+      "acc": 0.69053559,
+      "epoch": 1.3286402841197362,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.7770194732194256e-06,
+      "loss": 1.37317457,
+      "memory(GiB)": 123.79,
+      "step": 52375,
+      "train_speed(iter/s)": 1.316037
+    },
+    {
+      "acc": 0.6826582,
+      "epoch": 1.3287671232876712,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.7760802372144825e-06,
+      "loss": 1.44281778,
+      "memory(GiB)": 123.79,
+      "step": 52380,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.69302921,
+      "epoch": 1.3288939624556062,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.7751410990274596e-06,
+      "loss": 1.36305923,
+      "memory(GiB)": 123.79,
+      "step": 52385,
+      "train_speed(iter/s)": 1.316063
+    },
+    {
+      "acc": 0.68749285,
+      "epoch": 1.3290208016235414,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.774202058699664e-06,
+      "loss": 1.31151571,
+      "memory(GiB)": 123.79,
+      "step": 52390,
+      "train_speed(iter/s)": 1.316076
+    },
+    {
+      "acc": 0.69001026,
+      "epoch": 1.3291476407914764,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.7732631162724005e-06,
+      "loss": 1.36788187,
+      "memory(GiB)": 123.79,
+      "step": 52395,
+      "train_speed(iter/s)": 1.316089
+    },
+    {
+      "acc": 0.68895464,
+      "epoch": 1.3292744799594114,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.772324271786966e-06,
+      "loss": 1.38487644,
+      "memory(GiB)": 123.79,
+      "step": 52400,
+      "train_speed(iter/s)": 1.316103
+    },
+    {
+      "acc": 0.69406166,
+      "epoch": 1.3294013191273466,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.7713855252846545e-06,
+      "loss": 1.39635544,
+      "memory(GiB)": 123.79,
+      "step": 52405,
+      "train_speed(iter/s)": 1.316116
+    },
+    {
+      "acc": 0.69451656,
+      "epoch": 1.3295281582952816,
+      "grad_norm": 4.0,
+      "learning_rate": 2.7704468768067616e-06,
+      "loss": 1.33245058,
+      "memory(GiB)": 123.79,
+      "step": 52410,
+      "train_speed(iter/s)": 1.316129
+    },
+    {
+      "acc": 0.67426085,
+      "epoch": 1.3296549974632166,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.7695083263945664e-06,
+      "loss": 1.44291153,
+      "memory(GiB)": 123.79,
+      "step": 52415,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.67191234,
+      "epoch": 1.3297818366311516,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.7685698740893516e-06,
+      "loss": 1.47976885,
+      "memory(GiB)": 123.79,
+      "step": 52420,
+      "train_speed(iter/s)": 1.316156
+    },
+    {
+      "acc": 0.69363337,
+      "epoch": 1.3299086757990868,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.7676315199323995e-06,
+      "loss": 1.36431179,
+      "memory(GiB)": 123.79,
+      "step": 52425,
+      "train_speed(iter/s)": 1.31617
+    },
+    {
+      "acc": 0.7018199,
+      "epoch": 1.3300355149670218,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.7666932639649814e-06,
+      "loss": 1.29041786,
+      "memory(GiB)": 123.79,
+      "step": 52430,
+      "train_speed(iter/s)": 1.316185
+    },
+    {
+      "acc": 0.6749074,
+      "epoch": 1.330162354134957,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.765755106228362e-06,
+      "loss": 1.41115189,
+      "memory(GiB)": 123.79,
+      "step": 52435,
+      "train_speed(iter/s)": 1.316198
+    },
+    {
+      "acc": 0.68368344,
+      "epoch": 1.330289193302892,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.764817046763807e-06,
+      "loss": 1.35275383,
+      "memory(GiB)": 123.79,
+      "step": 52440,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.69181805,
+      "epoch": 1.330416032470827,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.7638790856125786e-06,
+      "loss": 1.43436785,
+      "memory(GiB)": 123.79,
+      "step": 52445,
+      "train_speed(iter/s)": 1.316227
+    },
+    {
+      "acc": 0.68040094,
+      "epoch": 1.330542871638762,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.7629412228159346e-06,
+      "loss": 1.44236069,
+      "memory(GiB)": 123.79,
+      "step": 52450,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.68554983,
+      "epoch": 1.3306697108066972,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.762003458415119e-06,
+      "loss": 1.42011528,
+      "memory(GiB)": 123.79,
+      "step": 52455,
+      "train_speed(iter/s)": 1.316252
+    },
+    {
+      "acc": 0.69088154,
+      "epoch": 1.3307965499746321,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.7610657924513853e-06,
+      "loss": 1.34611282,
+      "memory(GiB)": 123.79,
+      "step": 52460,
+      "train_speed(iter/s)": 1.316265
+    },
+    {
+      "acc": 0.69214921,
+      "epoch": 1.3309233891425674,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.7601282249659737e-06,
+      "loss": 1.40886831,
+      "memory(GiB)": 123.79,
+      "step": 52465,
+      "train_speed(iter/s)": 1.316279
+    },
+    {
+      "acc": 0.67116671,
+      "epoch": 1.3310502283105023,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.759190756000126e-06,
+      "loss": 1.45605183,
+      "memory(GiB)": 123.79,
+      "step": 52470,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.67722349,
+      "epoch": 1.3311770674784373,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.7582533855950687e-06,
+      "loss": 1.42449942,
+      "memory(GiB)": 123.79,
+      "step": 52475,
+      "train_speed(iter/s)": 1.316307
+    },
+    {
+      "acc": 0.6865098,
+      "epoch": 1.3313039066463723,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.757316113792038e-06,
+      "loss": 1.42120018,
+      "memory(GiB)": 123.79,
+      "step": 52480,
+      "train_speed(iter/s)": 1.316321
+    },
+    {
+      "acc": 0.67401023,
+      "epoch": 1.3314307458143075,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.756378940632258e-06,
+      "loss": 1.44939213,
+      "memory(GiB)": 123.79,
+      "step": 52485,
+      "train_speed(iter/s)": 1.316334
+    },
+    {
+      "acc": 0.6775342,
+      "epoch": 1.3315575849822425,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.755441866156949e-06,
+      "loss": 1.41888294,
+      "memory(GiB)": 123.79,
+      "step": 52490,
+      "train_speed(iter/s)": 1.316348
+    },
+    {
+      "acc": 0.68969612,
+      "epoch": 1.3316844241501775,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.7545048904073278e-06,
+      "loss": 1.41563225,
+      "memory(GiB)": 123.79,
+      "step": 52495,
+      "train_speed(iter/s)": 1.316362
+    },
+    {
+      "acc": 0.69622617,
+      "epoch": 1.3318112633181127,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.7535680134246067e-06,
+      "loss": 1.37454739,
+      "memory(GiB)": 123.79,
+      "step": 52500,
+      "train_speed(iter/s)": 1.316376
+    },
+    {
+      "acc": 0.7032156,
+      "epoch": 1.3319381024860477,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.752631235249995e-06,
+      "loss": 1.41585941,
+      "memory(GiB)": 123.79,
+      "step": 52505,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.69330139,
+      "epoch": 1.3320649416539827,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.7516945559246945e-06,
+      "loss": 1.30695724,
+      "memory(GiB)": 123.79,
+      "step": 52510,
+      "train_speed(iter/s)": 1.316403
+    },
+    {
+      "acc": 0.68663383,
+      "epoch": 1.3321917808219177,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.7507579754899053e-06,
+      "loss": 1.36303539,
+      "memory(GiB)": 123.79,
+      "step": 52515,
+      "train_speed(iter/s)": 1.316416
+    },
+    {
+      "acc": 0.70446262,
+      "epoch": 1.332318619989853,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.749821493986823e-06,
+      "loss": 1.30782852,
+      "memory(GiB)": 123.79,
+      "step": 52520,
+      "train_speed(iter/s)": 1.316429
+    },
+    {
+      "acc": 0.69472494,
+      "epoch": 1.332445459157788,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.748885111456637e-06,
+      "loss": 1.40398169,
+      "memory(GiB)": 123.79,
+      "step": 52525,
+      "train_speed(iter/s)": 1.316442
+    },
+    {
+      "acc": 0.69785137,
+      "epoch": 1.332572298325723,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.7479488279405354e-06,
+      "loss": 1.38479366,
+      "memory(GiB)": 123.79,
+      "step": 52530,
+      "train_speed(iter/s)": 1.316455
+    },
+    {
+      "acc": 0.68926105,
+      "epoch": 1.332699137493658,
+      "grad_norm": 3.5,
+      "learning_rate": 2.7470126434796984e-06,
+      "loss": 1.34963884,
+      "memory(GiB)": 123.79,
+      "step": 52535,
+      "train_speed(iter/s)": 1.316468
+    },
+    {
+      "acc": 0.67761536,
+      "epoch": 1.332825976661593,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.746076558115304e-06,
+      "loss": 1.38288918,
+      "memory(GiB)": 123.79,
+      "step": 52540,
+      "train_speed(iter/s)": 1.316481
+    },
+    {
+      "acc": 0.70159855,
+      "epoch": 1.332952815829528,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.7451405718885237e-06,
+      "loss": 1.35336666,
+      "memory(GiB)": 123.79,
+      "step": 52545,
+      "train_speed(iter/s)": 1.316494
+    },
+    {
+      "acc": 0.68092327,
+      "epoch": 1.3330796549974633,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.7442046848405328e-06,
+      "loss": 1.37207098,
+      "memory(GiB)": 123.79,
+      "step": 52550,
+      "train_speed(iter/s)": 1.316507
+    },
+    {
+      "acc": 0.69755468,
+      "epoch": 1.3332064941653983,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.743268897012489e-06,
+      "loss": 1.34275875,
+      "memory(GiB)": 123.79,
+      "step": 52555,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.68145118,
+      "epoch": 1.3333333333333333,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.7423332084455543e-06,
+      "loss": 1.40957699,
+      "memory(GiB)": 123.79,
+      "step": 52560,
+      "train_speed(iter/s)": 1.31653
+    },
+    {
+      "acc": 0.70475955,
+      "epoch": 1.3334601725012685,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.741397619180883e-06,
+      "loss": 1.2985405,
+      "memory(GiB)": 123.79,
+      "step": 52565,
+      "train_speed(iter/s)": 1.316542
+    },
+    {
+      "acc": 0.6902338,
+      "epoch": 1.3335870116692035,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.740462129259633e-06,
+      "loss": 1.42648926,
+      "memory(GiB)": 123.79,
+      "step": 52570,
+      "train_speed(iter/s)": 1.316556
+    },
+    {
+      "acc": 0.68520231,
+      "epoch": 1.3337138508371384,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.739526738722944e-06,
+      "loss": 1.44880934,
+      "memory(GiB)": 123.79,
+      "step": 52575,
+      "train_speed(iter/s)": 1.316568
+    },
+    {
+      "acc": 0.68655977,
+      "epoch": 1.3338406900050734,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.738591447611959e-06,
+      "loss": 1.3527523,
+      "memory(GiB)": 123.79,
+      "step": 52580,
+      "train_speed(iter/s)": 1.316581
+    },
+    {
+      "acc": 0.67407994,
+      "epoch": 1.3339675291730086,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.7376562559678214e-06,
+      "loss": 1.41503134,
+      "memory(GiB)": 123.79,
+      "step": 52585,
+      "train_speed(iter/s)": 1.316593
+    },
+    {
+      "acc": 0.68632364,
+      "epoch": 1.3340943683409436,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.7367211638316637e-06,
+      "loss": 1.44091721,
+      "memory(GiB)": 123.79,
+      "step": 52590,
+      "train_speed(iter/s)": 1.316605
+    },
+    {
+      "acc": 0.68836594,
+      "epoch": 1.3342212075088788,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.735786171244611e-06,
+      "loss": 1.39250183,
+      "memory(GiB)": 123.79,
+      "step": 52595,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "acc": 0.67527037,
+      "epoch": 1.3343480466768138,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.7348512782477922e-06,
+      "loss": 1.40485697,
+      "memory(GiB)": 123.79,
+      "step": 52600,
+      "train_speed(iter/s)": 1.316631
+    },
+    {
+      "acc": 0.69200335,
+      "epoch": 1.3344748858447488,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.7339164848823287e-06,
+      "loss": 1.37035294,
+      "memory(GiB)": 123.79,
+      "step": 52605,
+      "train_speed(iter/s)": 1.316643
+    },
+    {
+      "acc": 0.68674688,
+      "epoch": 1.3346017250126838,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.7329817911893365e-06,
+      "loss": 1.40934925,
+      "memory(GiB)": 123.79,
+      "step": 52610,
+      "train_speed(iter/s)": 1.316655
+    },
+    {
+      "acc": 0.69060159,
+      "epoch": 1.334728564180619,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.7320471972099226e-06,
+      "loss": 1.38381786,
+      "memory(GiB)": 123.79,
+      "step": 52615,
+      "train_speed(iter/s)": 1.316667
+    },
+    {
+      "acc": 0.68918042,
+      "epoch": 1.334855403348554,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.7311127029852007e-06,
+      "loss": 1.37982826,
+      "memory(GiB)": 123.79,
+      "step": 52620,
+      "train_speed(iter/s)": 1.316681
+    },
+    {
+      "acc": 0.67197247,
+      "epoch": 1.3349822425164892,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.7301783085562726e-06,
+      "loss": 1.46099072,
+      "memory(GiB)": 123.79,
+      "step": 52625,
+      "train_speed(iter/s)": 1.316693
+    },
+    {
+      "acc": 0.69112296,
+      "epoch": 1.3351090816844242,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.7292440139642364e-06,
+      "loss": 1.38340616,
+      "memory(GiB)": 123.79,
+      "step": 52630,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.67885466,
+      "epoch": 1.3352359208523592,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.7283098192501855e-06,
+      "loss": 1.38984241,
+      "memory(GiB)": 123.79,
+      "step": 52635,
+      "train_speed(iter/s)": 1.316719
+    },
+    {
+      "acc": 0.68890796,
+      "epoch": 1.3353627600202942,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.7273757244552124e-06,
+      "loss": 1.33036032,
+      "memory(GiB)": 123.79,
+      "step": 52640,
+      "train_speed(iter/s)": 1.316732
+    },
+    {
+      "acc": 0.6920537,
+      "epoch": 1.3354895991882294,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.726441729620401e-06,
+      "loss": 1.36489315,
+      "memory(GiB)": 123.79,
+      "step": 52645,
+      "train_speed(iter/s)": 1.316745
+    },
+    {
+      "acc": 0.6895741,
+      "epoch": 1.3356164383561644,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.725507834786833e-06,
+      "loss": 1.44250526,
+      "memory(GiB)": 123.79,
+      "step": 52650,
+      "train_speed(iter/s)": 1.31676
+    },
+    {
+      "acc": 0.69478531,
+      "epoch": 1.3357432775240994,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.7245740399955857e-06,
+      "loss": 1.3776741,
+      "memory(GiB)": 123.79,
+      "step": 52655,
+      "train_speed(iter/s)": 1.316774
+    },
+    {
+      "acc": 0.69223027,
+      "epoch": 1.3358701166920346,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.72364034528773e-06,
+      "loss": 1.37574577,
+      "memory(GiB)": 123.79,
+      "step": 52660,
+      "train_speed(iter/s)": 1.316788
+    },
+    {
+      "acc": 0.69693832,
+      "epoch": 1.3359969558599696,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.722706750704337e-06,
+      "loss": 1.39412689,
+      "memory(GiB)": 123.79,
+      "step": 52665,
+      "train_speed(iter/s)": 1.316803
+    },
+    {
+      "acc": 0.6810956,
+      "epoch": 1.3361237950279046,
+      "grad_norm": 4.125,
+      "learning_rate": 2.7217732562864673e-06,
+      "loss": 1.40777483,
+      "memory(GiB)": 123.79,
+      "step": 52670,
+      "train_speed(iter/s)": 1.316816
+    },
+    {
+      "acc": 0.66618099,
+      "epoch": 1.3362506341958396,
+      "grad_norm": 4.65625,
+      "learning_rate": 2.720839862075181e-06,
+      "loss": 1.44954739,
+      "memory(GiB)": 123.79,
+      "step": 52675,
+      "train_speed(iter/s)": 1.31683
+    },
+    {
+      "acc": 0.68482924,
+      "epoch": 1.3363774733637748,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.7199065681115344e-06,
+      "loss": 1.42054892,
+      "memory(GiB)": 123.79,
+      "step": 52680,
+      "train_speed(iter/s)": 1.316843
+    },
+    {
+      "acc": 0.69247565,
+      "epoch": 1.3365043125317098,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.7189733744365742e-06,
+      "loss": 1.39429913,
+      "memory(GiB)": 123.79,
+      "step": 52685,
+      "train_speed(iter/s)": 1.316857
+    },
+    {
+      "acc": 0.69628677,
+      "epoch": 1.336631151699645,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.718040281091353e-06,
+      "loss": 1.31434402,
+      "memory(GiB)": 123.79,
+      "step": 52690,
+      "train_speed(iter/s)": 1.316871
+    },
+    {
+      "acc": 0.68659172,
+      "epoch": 1.33675799086758,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.717107288116906e-06,
+      "loss": 1.35462961,
+      "memory(GiB)": 123.79,
+      "step": 52695,
+      "train_speed(iter/s)": 1.316885
+    },
+    {
+      "acc": 0.67585459,
+      "epoch": 1.336884830035515,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.716174395554274e-06,
+      "loss": 1.42162399,
+      "memory(GiB)": 123.79,
+      "step": 52700,
+      "train_speed(iter/s)": 1.316897
+    },
+    {
+      "acc": 0.68870463,
+      "epoch": 1.33701166920345,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.715241603444486e-06,
+      "loss": 1.43152132,
+      "memory(GiB)": 123.79,
+      "step": 52705,
+      "train_speed(iter/s)": 1.31691
+    },
+    {
+      "acc": 0.69573221,
+      "epoch": 1.3371385083713851,
+      "grad_norm": 3.5,
+      "learning_rate": 2.714308911828577e-06,
+      "loss": 1.35548248,
+      "memory(GiB)": 123.79,
+      "step": 52710,
+      "train_speed(iter/s)": 1.316923
+    },
+    {
+      "acc": 0.68079762,
+      "epoch": 1.3372653475393201,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.713376320747565e-06,
+      "loss": 1.45898809,
+      "memory(GiB)": 123.79,
+      "step": 52715,
+      "train_speed(iter/s)": 1.316936
+    },
+    {
+      "acc": 0.6808733,
+      "epoch": 1.3373921867072551,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.7124438302424696e-06,
+      "loss": 1.37184696,
+      "memory(GiB)": 123.79,
+      "step": 52720,
+      "train_speed(iter/s)": 1.316947
+    },
+    {
+      "acc": 0.68057499,
+      "epoch": 1.3375190258751903,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.711511440354309e-06,
+      "loss": 1.42393055,
+      "memory(GiB)": 123.79,
+      "step": 52725,
+      "train_speed(iter/s)": 1.31696
+    },
+    {
+      "acc": 0.68273978,
+      "epoch": 1.3376458650431253,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.710579151124095e-06,
+      "loss": 1.41211681,
+      "memory(GiB)": 123.79,
+      "step": 52730,
+      "train_speed(iter/s)": 1.316972
+    },
+    {
+      "acc": 0.68655839,
+      "epoch": 1.3377727042110603,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.70964696259283e-06,
+      "loss": 1.33895321,
+      "memory(GiB)": 123.79,
+      "step": 52735,
+      "train_speed(iter/s)": 1.316985
+    },
+    {
+      "acc": 0.68330364,
+      "epoch": 1.3378995433789953,
+      "grad_norm": 3.0,
+      "learning_rate": 2.7087148748015146e-06,
+      "loss": 1.36223383,
+      "memory(GiB)": 123.79,
+      "step": 52740,
+      "train_speed(iter/s)": 1.316997
+    },
+    {
+      "acc": 0.66706924,
+      "epoch": 1.3380263825469305,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.7077828877911517e-06,
+      "loss": 1.53029013,
+      "memory(GiB)": 123.79,
+      "step": 52745,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.68813562,
+      "epoch": 1.3381532217148655,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.706851001602733e-06,
+      "loss": 1.33446531,
+      "memory(GiB)": 123.79,
+      "step": 52750,
+      "train_speed(iter/s)": 1.317022
+    },
+    {
+      "acc": 0.69400454,
+      "epoch": 1.3382800608828007,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.7059192162772407e-06,
+      "loss": 1.37846651,
+      "memory(GiB)": 123.79,
+      "step": 52755,
+      "train_speed(iter/s)": 1.317036
+    },
+    {
+      "acc": 0.68525519,
+      "epoch": 1.3384069000507357,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.704987531855666e-06,
+      "loss": 1.41410713,
+      "memory(GiB)": 123.79,
+      "step": 52760,
+      "train_speed(iter/s)": 1.31705
+    },
+    {
+      "acc": 0.69304414,
+      "epoch": 1.3385337392186707,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.704055948378986e-06,
+      "loss": 1.37823677,
+      "memory(GiB)": 123.79,
+      "step": 52765,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.68713245,
+      "epoch": 1.3386605783866057,
+      "grad_norm": 3.25,
+      "learning_rate": 2.7031244658881773e-06,
+      "loss": 1.34919729,
+      "memory(GiB)": 123.79,
+      "step": 52770,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.69628048,
+      "epoch": 1.338787417554541,
+      "grad_norm": 4.25,
+      "learning_rate": 2.7021930844242085e-06,
+      "loss": 1.39723988,
+      "memory(GiB)": 123.79,
+      "step": 52775,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.6980176,
+      "epoch": 1.3389142567224759,
+      "grad_norm": 3.75,
+      "learning_rate": 2.7012618040280463e-06,
+      "loss": 1.39282055,
+      "memory(GiB)": 123.79,
+      "step": 52780,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.70868587,
+      "epoch": 1.339041095890411,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.7003306247406536e-06,
+      "loss": 1.3580595,
+      "memory(GiB)": 123.79,
+      "step": 52785,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.67572412,
+      "epoch": 1.339167935058346,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.6993995466029877e-06,
+      "loss": 1.47459564,
+      "memory(GiB)": 123.79,
+      "step": 52790,
+      "train_speed(iter/s)": 1.317132
+    },
+    {
+      "acc": 0.67073703,
+      "epoch": 1.339294774226281,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.6984685696560002e-06,
+      "loss": 1.41324654,
+      "memory(GiB)": 123.79,
+      "step": 52795,
+      "train_speed(iter/s)": 1.317146
+    },
+    {
+      "acc": 0.68775845,
+      "epoch": 1.339421613394216,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.6975376939406418e-06,
+      "loss": 1.35009127,
+      "memory(GiB)": 123.79,
+      "step": 52800,
+      "train_speed(iter/s)": 1.317159
+    },
+    {
+      "acc": 0.68969183,
+      "epoch": 1.3395484525621513,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.6966069194978537e-06,
+      "loss": 1.35583687,
+      "memory(GiB)": 123.79,
+      "step": 52805,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.67478461,
+      "epoch": 1.3396752917300863,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.6956762463685787e-06,
+      "loss": 1.4478323,
+      "memory(GiB)": 123.79,
+      "step": 52810,
+      "train_speed(iter/s)": 1.317183
+    },
+    {
+      "acc": 0.68638806,
+      "epoch": 1.3398021308980212,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.69474567459375e-06,
+      "loss": 1.42727528,
+      "memory(GiB)": 123.79,
+      "step": 52815,
+      "train_speed(iter/s)": 1.317196
+    },
+    {
+      "acc": 0.68904014,
+      "epoch": 1.3399289700659565,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.693815204214299e-06,
+      "loss": 1.41663342,
+      "memory(GiB)": 123.79,
+      "step": 52820,
+      "train_speed(iter/s)": 1.317208
+    },
+    {
+      "acc": 0.6901938,
+      "epoch": 1.3400558092338914,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.692884835271151e-06,
+      "loss": 1.38577681,
+      "memory(GiB)": 123.79,
+      "step": 52825,
+      "train_speed(iter/s)": 1.317221
+    },
+    {
+      "acc": 0.69903088,
+      "epoch": 1.3401826484018264,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.6919545678052296e-06,
+      "loss": 1.39155903,
+      "memory(GiB)": 123.79,
+      "step": 52830,
+      "train_speed(iter/s)": 1.317235
+    },
+    {
+      "acc": 0.70395985,
+      "epoch": 1.3403094875697614,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.69102440185745e-06,
+      "loss": 1.29296875,
+      "memory(GiB)": 123.79,
+      "step": 52835,
+      "train_speed(iter/s)": 1.317249
+    },
+    {
+      "acc": 0.66930208,
+      "epoch": 1.3404363267376966,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.690094337468726e-06,
+      "loss": 1.51962404,
+      "memory(GiB)": 123.79,
+      "step": 52840,
+      "train_speed(iter/s)": 1.317263
+    },
+    {
+      "acc": 0.68249025,
+      "epoch": 1.3405631659056316,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.6891643746799643e-06,
+      "loss": 1.41371498,
+      "memory(GiB)": 123.79,
+      "step": 52845,
+      "train_speed(iter/s)": 1.317276
+    },
+    {
+      "acc": 0.65968733,
+      "epoch": 1.3406900050735668,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.6882345135320753e-06,
+      "loss": 1.44856415,
+      "memory(GiB)": 123.79,
+      "step": 52850,
+      "train_speed(iter/s)": 1.317289
+    },
+    {
+      "acc": 0.69344158,
+      "epoch": 1.3408168442415018,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.68730475406595e-06,
+      "loss": 1.42400646,
+      "memory(GiB)": 123.79,
+      "step": 52855,
+      "train_speed(iter/s)": 1.317301
+    },
+    {
+      "acc": 0.67300463,
+      "epoch": 1.3409436834094368,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.6863750963224867e-06,
+      "loss": 1.45316696,
+      "memory(GiB)": 123.79,
+      "step": 52860,
+      "train_speed(iter/s)": 1.317315
+    },
+    {
+      "acc": 0.67424693,
+      "epoch": 1.3410705225773718,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.685445540342577e-06,
+      "loss": 1.43508873,
+      "memory(GiB)": 123.79,
+      "step": 52865,
+      "train_speed(iter/s)": 1.317328
+    },
+    {
+      "acc": 0.67169552,
+      "epoch": 1.341197361745307,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.6845160861671094e-06,
+      "loss": 1.50075788,
+      "memory(GiB)": 123.79,
+      "step": 52870,
+      "train_speed(iter/s)": 1.317341
+    },
+    {
+      "acc": 0.68319778,
+      "epoch": 1.341324200913242,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.6835867338369593e-06,
+      "loss": 1.33822756,
+      "memory(GiB)": 123.79,
+      "step": 52875,
+      "train_speed(iter/s)": 1.317354
+    },
+    {
+      "acc": 0.68251853,
+      "epoch": 1.341451040081177,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.6826574833930053e-06,
+      "loss": 1.32429504,
+      "memory(GiB)": 123.79,
+      "step": 52880,
+      "train_speed(iter/s)": 1.317366
+    },
+    {
+      "acc": 0.67929506,
+      "epoch": 1.3415778792491122,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.681728334876123e-06,
+      "loss": 1.39523592,
+      "memory(GiB)": 123.79,
+      "step": 52885,
+      "train_speed(iter/s)": 1.317379
+    },
+    {
+      "acc": 0.69840369,
+      "epoch": 1.3417047184170472,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.6807992883271806e-06,
+      "loss": 1.30645018,
+      "memory(GiB)": 123.79,
+      "step": 52890,
+      "train_speed(iter/s)": 1.317392
+    },
+    {
+      "acc": 0.69635248,
+      "epoch": 1.3418315575849822,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.6798703437870364e-06,
+      "loss": 1.35013676,
+      "memory(GiB)": 123.79,
+      "step": 52895,
+      "train_speed(iter/s)": 1.317405
+    },
+    {
+      "acc": 0.68865538,
+      "epoch": 1.3419583967529172,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.678941501296555e-06,
+      "loss": 1.39388151,
+      "memory(GiB)": 123.79,
+      "step": 52900,
+      "train_speed(iter/s)": 1.317417
+    },
+    {
+      "acc": 0.70021582,
+      "epoch": 1.3420852359208524,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.6780127608965896e-06,
+      "loss": 1.37401581,
+      "memory(GiB)": 123.79,
+      "step": 52905,
+      "train_speed(iter/s)": 1.31743
+    },
+    {
+      "acc": 0.69441137,
+      "epoch": 1.3422120750887874,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.677084122627991e-06,
+      "loss": 1.39432974,
+      "memory(GiB)": 123.79,
+      "step": 52910,
+      "train_speed(iter/s)": 1.317443
+    },
+    {
+      "acc": 0.68034067,
+      "epoch": 1.3423389142567226,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.6761555865316003e-06,
+      "loss": 1.44757357,
+      "memory(GiB)": 123.79,
+      "step": 52915,
+      "train_speed(iter/s)": 1.317456
+    },
+    {
+      "acc": 0.69903479,
+      "epoch": 1.3424657534246576,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.6752271526482644e-06,
+      "loss": 1.39556942,
+      "memory(GiB)": 123.79,
+      "step": 52920,
+      "train_speed(iter/s)": 1.317469
+    },
+    {
+      "acc": 0.67908654,
+      "epoch": 1.3425925925925926,
+      "grad_norm": 3.125,
+      "learning_rate": 2.674298821018817e-06,
+      "loss": 1.43544369,
+      "memory(GiB)": 123.79,
+      "step": 52925,
+      "train_speed(iter/s)": 1.317482
+    },
+    {
+      "acc": 0.68439198,
+      "epoch": 1.3427194317605275,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.673370591684091e-06,
+      "loss": 1.46527052,
+      "memory(GiB)": 123.79,
+      "step": 52930,
+      "train_speed(iter/s)": 1.317495
+    },
+    {
+      "acc": 0.67733092,
+      "epoch": 1.3428462709284628,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.672442464684915e-06,
+      "loss": 1.45739269,
+      "memory(GiB)": 123.79,
+      "step": 52935,
+      "train_speed(iter/s)": 1.317507
+    },
+    {
+      "acc": 0.69509816,
+      "epoch": 1.3429731100963977,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.671514440062111e-06,
+      "loss": 1.37007875,
+      "memory(GiB)": 123.79,
+      "step": 52940,
+      "train_speed(iter/s)": 1.31752
+    },
+    {
+      "acc": 0.68707395,
+      "epoch": 1.343099949264333,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.6705865178564973e-06,
+      "loss": 1.39138908,
+      "memory(GiB)": 123.79,
+      "step": 52945,
+      "train_speed(iter/s)": 1.317533
+    },
+    {
+      "acc": 0.68251972,
+      "epoch": 1.343226788432268,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.6696586981088886e-06,
+      "loss": 1.43199177,
+      "memory(GiB)": 123.79,
+      "step": 52950,
+      "train_speed(iter/s)": 1.317546
+    },
+    {
+      "acc": 0.6754209,
+      "epoch": 1.343353627600203,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.6687309808600947e-06,
+      "loss": 1.43430376,
+      "memory(GiB)": 123.79,
+      "step": 52955,
+      "train_speed(iter/s)": 1.31756
+    },
+    {
+      "acc": 0.67985616,
+      "epoch": 1.343480466768138,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.6678033661509208e-06,
+      "loss": 1.42243414,
+      "memory(GiB)": 123.79,
+      "step": 52960,
+      "train_speed(iter/s)": 1.317573
+    },
+    {
+      "acc": 0.67924204,
+      "epoch": 1.3436073059360731,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.6668758540221665e-06,
+      "loss": 1.44858131,
+      "memory(GiB)": 123.79,
+      "step": 52965,
+      "train_speed(iter/s)": 1.317587
+    },
+    {
+      "acc": 0.68387775,
+      "epoch": 1.3437341451040081,
+      "grad_norm": 3.5,
+      "learning_rate": 2.66594844451463e-06,
+      "loss": 1.34687281,
+      "memory(GiB)": 123.79,
+      "step": 52970,
+      "train_speed(iter/s)": 1.3176
+    },
+    {
+      "acc": 0.68199029,
+      "epoch": 1.3438609842719431,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.6650211376691008e-06,
+      "loss": 1.49907341,
+      "memory(GiB)": 123.79,
+      "step": 52975,
+      "train_speed(iter/s)": 1.317614
+    },
+    {
+      "acc": 0.68935709,
+      "epoch": 1.3439878234398783,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.664093933526368e-06,
+      "loss": 1.3996006,
+      "memory(GiB)": 123.79,
+      "step": 52980,
+      "train_speed(iter/s)": 1.317628
+    },
+    {
+      "acc": 0.68135605,
+      "epoch": 1.3441146626078133,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.6631668321272097e-06,
+      "loss": 1.4306078,
+      "memory(GiB)": 123.79,
+      "step": 52985,
+      "train_speed(iter/s)": 1.317642
+    },
+    {
+      "acc": 0.68078451,
+      "epoch": 1.3442415017757483,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.6622398335124122e-06,
+      "loss": 1.3964056,
+      "memory(GiB)": 123.79,
+      "step": 52990,
+      "train_speed(iter/s)": 1.317655
+    },
+    {
+      "acc": 0.68013902,
+      "epoch": 1.3443683409436833,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.661312937722742e-06,
+      "loss": 1.4276638,
+      "memory(GiB)": 123.79,
+      "step": 52995,
+      "train_speed(iter/s)": 1.31767
+    },
+    {
+      "acc": 0.6908257,
+      "epoch": 1.3444951801116185,
+      "grad_norm": 2.75,
+      "learning_rate": 2.6603861447989703e-06,
+      "loss": 1.35646105,
+      "memory(GiB)": 123.79,
+      "step": 53000,
+      "train_speed(iter/s)": 1.317683
+    },
+    {
+      "epoch": 1.3444951801116185,
+      "eval_acc": 0.6750520510298481,
+      "eval_loss": 1.3563165664672852,
+      "eval_runtime": 70.1033,
+      "eval_samples_per_second": 90.866,
+      "eval_steps_per_second": 22.724,
+      "step": 53000
+    },
+    {
+      "acc": 0.69015112,
+      "epoch": 1.3446220192795535,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.65945945478186e-06,
+      "loss": 1.33994493,
+      "memory(GiB)": 123.79,
+      "step": 53005,
+      "train_speed(iter/s)": 1.314996
+    },
+    {
+      "acc": 0.69377322,
+      "epoch": 1.3447488584474887,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.658532867712176e-06,
+      "loss": 1.36638441,
+      "memory(GiB)": 123.79,
+      "step": 53010,
+      "train_speed(iter/s)": 1.31501
+    },
+    {
+      "acc": 0.69489603,
+      "epoch": 1.3448756976154237,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.6576063836306687e-06,
+      "loss": 1.32014532,
+      "memory(GiB)": 123.79,
+      "step": 53015,
+      "train_speed(iter/s)": 1.315023
+    },
+    {
+      "acc": 0.68809662,
+      "epoch": 1.3450025367833587,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.656680002578088e-06,
+      "loss": 1.37621717,
+      "memory(GiB)": 123.79,
+      "step": 53020,
+      "train_speed(iter/s)": 1.315038
+    },
+    {
+      "acc": 0.69219885,
+      "epoch": 1.3451293759512937,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.655753724595186e-06,
+      "loss": 1.36331234,
+      "memory(GiB)": 123.79,
+      "step": 53025,
+      "train_speed(iter/s)": 1.315052
+    },
+    {
+      "acc": 0.68133745,
+      "epoch": 1.3452562151192289,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.6548275497227028e-06,
+      "loss": 1.37769909,
+      "memory(GiB)": 123.79,
+      "step": 53030,
+      "train_speed(iter/s)": 1.315065
+    },
+    {
+      "acc": 0.69720287,
+      "epoch": 1.3453830542871639,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6539014780013707e-06,
+      "loss": 1.36336164,
+      "memory(GiB)": 123.79,
+      "step": 53035,
+      "train_speed(iter/s)": 1.315078
+    },
+    {
+      "acc": 0.68487797,
+      "epoch": 1.3455098934550989,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.6529755094719276e-06,
+      "loss": 1.43824043,
+      "memory(GiB)": 123.79,
+      "step": 53040,
+      "train_speed(iter/s)": 1.315092
+    },
+    {
+      "acc": 0.68083258,
+      "epoch": 1.345636732623034,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.652049644175101e-06,
+      "loss": 1.42174988,
+      "memory(GiB)": 123.79,
+      "step": 53045,
+      "train_speed(iter/s)": 1.315106
+    },
+    {
+      "acc": 0.68765163,
+      "epoch": 1.345763571790969,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.6511238821516154e-06,
+      "loss": 1.36698112,
+      "memory(GiB)": 123.79,
+      "step": 53050,
+      "train_speed(iter/s)": 1.315119
+    },
+    {
+      "acc": 0.68989072,
+      "epoch": 1.345890410958904,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.650198223442185e-06,
+      "loss": 1.35336819,
+      "memory(GiB)": 123.79,
+      "step": 53055,
+      "train_speed(iter/s)": 1.315133
+    },
+    {
+      "acc": 0.68308806,
+      "epoch": 1.346017250126839,
+      "grad_norm": 3.125,
+      "learning_rate": 2.6492726680875296e-06,
+      "loss": 1.39762449,
+      "memory(GiB)": 123.79,
+      "step": 53060,
+      "train_speed(iter/s)": 1.315146
+    },
+    {
+      "acc": 0.7077178,
+      "epoch": 1.3461440892947742,
+      "grad_norm": 3.375,
+      "learning_rate": 2.6483472161283576e-06,
+      "loss": 1.36591206,
+      "memory(GiB)": 123.79,
+      "step": 53065,
+      "train_speed(iter/s)": 1.31516
+    },
+    {
+      "acc": 0.68132725,
+      "epoch": 1.3462709284627092,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.647421867605374e-06,
+      "loss": 1.36679029,
+      "memory(GiB)": 123.79,
+      "step": 53070,
+      "train_speed(iter/s)": 1.315174
+    },
+    {
+      "acc": 0.69073291,
+      "epoch": 1.3463977676306444,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.6464966225592804e-06,
+      "loss": 1.39212255,
+      "memory(GiB)": 123.79,
+      "step": 53075,
+      "train_speed(iter/s)": 1.315187
+    },
+    {
+      "acc": 0.68064051,
+      "epoch": 1.3465246067985794,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.645571481030773e-06,
+      "loss": 1.42473488,
+      "memory(GiB)": 123.79,
+      "step": 53080,
+      "train_speed(iter/s)": 1.315201
+    },
+    {
+      "acc": 0.69555645,
+      "epoch": 1.3466514459665144,
+      "grad_norm": 4.0,
+      "learning_rate": 2.6446464430605434e-06,
+      "loss": 1.37743359,
+      "memory(GiB)": 123.79,
+      "step": 53085,
+      "train_speed(iter/s)": 1.315214
+    },
+    {
+      "acc": 0.68600779,
+      "epoch": 1.3467782851344494,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.6437215086892797e-06,
+      "loss": 1.36748629,
+      "memory(GiB)": 123.79,
+      "step": 53090,
+      "train_speed(iter/s)": 1.315223
+    },
+    {
+      "acc": 0.70258775,
+      "epoch": 1.3469051243023846,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.642796677957664e-06,
+      "loss": 1.34542103,
+      "memory(GiB)": 123.79,
+      "step": 53095,
+      "train_speed(iter/s)": 1.315235
+    },
+    {
+      "acc": 0.67891541,
+      "epoch": 1.3470319634703196,
+      "grad_norm": 2.875,
+      "learning_rate": 2.641871950906374e-06,
+      "loss": 1.4388957,
+      "memory(GiB)": 123.79,
+      "step": 53100,
+      "train_speed(iter/s)": 1.315249
+    },
+    {
+      "acc": 0.68793001,
+      "epoch": 1.3471588026382548,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6409473275760843e-06,
+      "loss": 1.40168705,
+      "memory(GiB)": 123.79,
+      "step": 53105,
+      "train_speed(iter/s)": 1.315262
+    },
+    {
+      "acc": 0.68719773,
+      "epoch": 1.3472856418061898,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.640022808007463e-06,
+      "loss": 1.46098652,
+      "memory(GiB)": 123.79,
+      "step": 53110,
+      "train_speed(iter/s)": 1.315276
+    },
+    {
+      "acc": 0.68243361,
+      "epoch": 1.3474124809741248,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.639098392241176e-06,
+      "loss": 1.43724804,
+      "memory(GiB)": 123.79,
+      "step": 53115,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.67400932,
+      "epoch": 1.3475393201420598,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.6381740803178826e-06,
+      "loss": 1.39507236,
+      "memory(GiB)": 123.79,
+      "step": 53120,
+      "train_speed(iter/s)": 1.315303
+    },
+    {
+      "acc": 0.68782129,
+      "epoch": 1.347666159309995,
+      "grad_norm": 3.0,
+      "learning_rate": 2.6372498722782346e-06,
+      "loss": 1.3664958,
+      "memory(GiB)": 123.79,
+      "step": 53125,
+      "train_speed(iter/s)": 1.315317
+    },
+    {
+      "acc": 0.6811388,
+      "epoch": 1.34779299847793,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.6363257681628907e-06,
+      "loss": 1.42355499,
+      "memory(GiB)": 123.79,
+      "step": 53130,
+      "train_speed(iter/s)": 1.315331
+    },
+    {
+      "acc": 0.69451299,
+      "epoch": 1.347919837645865,
+      "grad_norm": 4.75,
+      "learning_rate": 2.635401768012491e-06,
+      "loss": 1.33726501,
+      "memory(GiB)": 123.79,
+      "step": 53135,
+      "train_speed(iter/s)": 1.315345
+    },
+    {
+      "acc": 0.69081421,
+      "epoch": 1.3480466768138002,
+      "grad_norm": 3.75,
+      "learning_rate": 2.6344778718676783e-06,
+      "loss": 1.36917114,
+      "memory(GiB)": 123.79,
+      "step": 53140,
+      "train_speed(iter/s)": 1.315358
+    },
+    {
+      "acc": 0.67924299,
+      "epoch": 1.3481735159817352,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.6335540797690886e-06,
+      "loss": 1.41132269,
+      "memory(GiB)": 123.79,
+      "step": 53145,
+      "train_speed(iter/s)": 1.315371
+    },
+    {
+      "acc": 0.70342808,
+      "epoch": 1.3483003551496702,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.632630391757359e-06,
+      "loss": 1.34490166,
+      "memory(GiB)": 123.79,
+      "step": 53150,
+      "train_speed(iter/s)": 1.315385
+    },
+    {
+      "acc": 0.67444487,
+      "epoch": 1.3484271943176052,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.6317068078731126e-06,
+      "loss": 1.33738079,
+      "memory(GiB)": 123.79,
+      "step": 53155,
+      "train_speed(iter/s)": 1.315398
+    },
+    {
+      "acc": 0.7003119,
+      "epoch": 1.3485540334855404,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.630783328156973e-06,
+      "loss": 1.30622492,
+      "memory(GiB)": 123.79,
+      "step": 53160,
+      "train_speed(iter/s)": 1.315412
+    },
+    {
+      "acc": 0.67989063,
+      "epoch": 1.3486808726534754,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.629859952649562e-06,
+      "loss": 1.3627532,
+      "memory(GiB)": 123.79,
+      "step": 53165,
+      "train_speed(iter/s)": 1.315426
+    },
+    {
+      "acc": 0.7031044,
+      "epoch": 1.3488077118214106,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.628936681391494e-06,
+      "loss": 1.3304184,
+      "memory(GiB)": 123.79,
+      "step": 53170,
+      "train_speed(iter/s)": 1.315439
+    },
+    {
+      "acc": 0.68990955,
+      "epoch": 1.3489345509893456,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.628013514423375e-06,
+      "loss": 1.45575943,
+      "memory(GiB)": 123.79,
+      "step": 53175,
+      "train_speed(iter/s)": 1.315452
+    },
+    {
+      "acc": 0.67254758,
+      "epoch": 1.3490613901572805,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.6270904517858102e-06,
+      "loss": 1.44330559,
+      "memory(GiB)": 123.79,
+      "step": 53180,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.68673706,
+      "epoch": 1.3491882293252155,
+      "grad_norm": 3.75,
+      "learning_rate": 2.6261674935194036e-06,
+      "loss": 1.41335144,
+      "memory(GiB)": 123.79,
+      "step": 53185,
+      "train_speed(iter/s)": 1.315478
+    },
+    {
+      "acc": 0.69454508,
+      "epoch": 1.3493150684931507,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.6252446396647503e-06,
+      "loss": 1.3129406,
+      "memory(GiB)": 123.79,
+      "step": 53190,
+      "train_speed(iter/s)": 1.315491
+    },
+    {
+      "acc": 0.67281146,
+      "epoch": 1.3494419076610857,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.6243218902624367e-06,
+      "loss": 1.46150322,
+      "memory(GiB)": 123.79,
+      "step": 53195,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.68821077,
+      "epoch": 1.3495687468290207,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.6233992453530555e-06,
+      "loss": 1.36474237,
+      "memory(GiB)": 123.79,
+      "step": 53200,
+      "train_speed(iter/s)": 1.315517
+    },
+    {
+      "acc": 0.70305367,
+      "epoch": 1.349695585996956,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6224767049771856e-06,
+      "loss": 1.32370853,
+      "memory(GiB)": 123.79,
+      "step": 53205,
+      "train_speed(iter/s)": 1.315531
+    },
+    {
+      "acc": 0.66413865,
+      "epoch": 1.349822425164891,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.621554269175405e-06,
+      "loss": 1.46112862,
+      "memory(GiB)": 123.79,
+      "step": 53210,
+      "train_speed(iter/s)": 1.315544
+    },
+    {
+      "acc": 0.66841412,
+      "epoch": 1.349949264332826,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.620631937988287e-06,
+      "loss": 1.46394291,
+      "memory(GiB)": 123.79,
+      "step": 53215,
+      "train_speed(iter/s)": 1.315556
+    },
+    {
+      "acc": 0.67868381,
+      "epoch": 1.350076103500761,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.6197097114564e-06,
+      "loss": 1.41206217,
+      "memory(GiB)": 123.79,
+      "step": 53220,
+      "train_speed(iter/s)": 1.315568
+    },
+    {
+      "acc": 0.66823659,
+      "epoch": 1.3502029426686961,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.618787589620306e-06,
+      "loss": 1.43915462,
+      "memory(GiB)": 123.79,
+      "step": 53225,
+      "train_speed(iter/s)": 1.315582
+    },
+    {
+      "acc": 0.69888058,
+      "epoch": 1.350329781836631,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.6178655725205653e-06,
+      "loss": 1.34482708,
+      "memory(GiB)": 123.79,
+      "step": 53230,
+      "train_speed(iter/s)": 1.315596
+    },
+    {
+      "acc": 0.69999962,
+      "epoch": 1.3504566210045663,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.6169436601977325e-06,
+      "loss": 1.34628963,
+      "memory(GiB)": 123.79,
+      "step": 53235,
+      "train_speed(iter/s)": 1.315609
+    },
+    {
+      "acc": 0.67783794,
+      "epoch": 1.3505834601725013,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.6160218526923576e-06,
+      "loss": 1.42238903,
+      "memory(GiB)": 123.79,
+      "step": 53240,
+      "train_speed(iter/s)": 1.315622
+    },
+    {
+      "acc": 0.66965899,
+      "epoch": 1.3507102993404363,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.6151001500449847e-06,
+      "loss": 1.48088303,
+      "memory(GiB)": 123.79,
+      "step": 53245,
+      "train_speed(iter/s)": 1.315636
+    },
+    {
+      "acc": 0.69103656,
+      "epoch": 1.3508371385083713,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.614178552296155e-06,
+      "loss": 1.43980703,
+      "memory(GiB)": 123.79,
+      "step": 53250,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.68572426,
+      "epoch": 1.3509639776763065,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.6132570594864047e-06,
+      "loss": 1.41894474,
+      "memory(GiB)": 123.79,
+      "step": 53255,
+      "train_speed(iter/s)": 1.315663
+    },
+    {
+      "acc": 0.70023994,
+      "epoch": 1.3510908168442415,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.612335671656265e-06,
+      "loss": 1.40612936,
+      "memory(GiB)": 123.79,
+      "step": 53260,
+      "train_speed(iter/s)": 1.315676
+    },
+    {
+      "acc": 0.70750055,
+      "epoch": 1.3512176560121767,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.6114143888462607e-06,
+      "loss": 1.32630749,
+      "memory(GiB)": 123.79,
+      "step": 53265,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.69146867,
+      "epoch": 1.3513444951801117,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.6104932110969195e-06,
+      "loss": 1.39576454,
+      "memory(GiB)": 123.79,
+      "step": 53270,
+      "train_speed(iter/s)": 1.315702
+    },
+    {
+      "acc": 0.69121923,
+      "epoch": 1.3514713343480467,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.609572138448753e-06,
+      "loss": 1.34918957,
+      "memory(GiB)": 123.79,
+      "step": 53275,
+      "train_speed(iter/s)": 1.315715
+    },
+    {
+      "acc": 0.69072618,
+      "epoch": 1.3515981735159817,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.608651170942277e-06,
+      "loss": 1.41395741,
+      "memory(GiB)": 123.79,
+      "step": 53280,
+      "train_speed(iter/s)": 1.315728
+    },
+    {
+      "acc": 0.68933058,
+      "epoch": 1.3517250126839169,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.607730308617997e-06,
+      "loss": 1.35758495,
+      "memory(GiB)": 123.79,
+      "step": 53285,
+      "train_speed(iter/s)": 1.315742
+    },
+    {
+      "acc": 0.7080821,
+      "epoch": 1.3518518518518519,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.6068095515164226e-06,
+      "loss": 1.33737965,
+      "memory(GiB)": 123.79,
+      "step": 53290,
+      "train_speed(iter/s)": 1.315755
+    },
+    {
+      "acc": 0.67550616,
+      "epoch": 1.3519786910197868,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.605888899678047e-06,
+      "loss": 1.42466049,
+      "memory(GiB)": 123.79,
+      "step": 53295,
+      "train_speed(iter/s)": 1.315768
+    },
+    {
+      "acc": 0.69402676,
+      "epoch": 1.352105530187722,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.6049683531433645e-06,
+      "loss": 1.39694939,
+      "memory(GiB)": 123.79,
+      "step": 53300,
+      "train_speed(iter/s)": 1.315782
+    },
+    {
+      "acc": 0.68159466,
+      "epoch": 1.352232369355657,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.6040479119528683e-06,
+      "loss": 1.42445431,
+      "memory(GiB)": 123.79,
+      "step": 53305,
+      "train_speed(iter/s)": 1.315795
+    },
+    {
+      "acc": 0.69287643,
+      "epoch": 1.352359208523592,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.6031275761470447e-06,
+      "loss": 1.43568325,
+      "memory(GiB)": 123.79,
+      "step": 53310,
+      "train_speed(iter/s)": 1.315808
+    },
+    {
+      "acc": 0.66502056,
+      "epoch": 1.352486047691527,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.60220734576637e-06,
+      "loss": 1.44069347,
+      "memory(GiB)": 123.79,
+      "step": 53315,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.66762228,
+      "epoch": 1.3526128868594622,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.601287220851319e-06,
+      "loss": 1.51645317,
+      "memory(GiB)": 123.79,
+      "step": 53320,
+      "train_speed(iter/s)": 1.315835
+    },
+    {
+      "acc": 0.70687699,
+      "epoch": 1.3527397260273972,
+      "grad_norm": 5.375,
+      "learning_rate": 2.6003672014423677e-06,
+      "loss": 1.38650398,
+      "memory(GiB)": 123.79,
+      "step": 53325,
+      "train_speed(iter/s)": 1.315849
+    },
+    {
+      "acc": 0.68731537,
+      "epoch": 1.3528665651953324,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.5994472875799827e-06,
+      "loss": 1.39111042,
+      "memory(GiB)": 123.79,
+      "step": 53330,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.6861104,
+      "epoch": 1.3529934043632674,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.598527479304619e-06,
+      "loss": 1.40468969,
+      "memory(GiB)": 123.79,
+      "step": 53335,
+      "train_speed(iter/s)": 1.315876
+    },
+    {
+      "acc": 0.70185885,
+      "epoch": 1.3531202435312024,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.597607776656741e-06,
+      "loss": 1.35604191,
+      "memory(GiB)": 123.79,
+      "step": 53340,
+      "train_speed(iter/s)": 1.31589
+    },
+    {
+      "acc": 0.69341764,
+      "epoch": 1.3532470826991374,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.5966881796767984e-06,
+      "loss": 1.38673191,
+      "memory(GiB)": 123.79,
+      "step": 53345,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.67918353,
+      "epoch": 1.3533739218670726,
+      "grad_norm": 4.90625,
+      "learning_rate": 2.5957686884052423e-06,
+      "loss": 1.42754784,
+      "memory(GiB)": 123.79,
+      "step": 53350,
+      "train_speed(iter/s)": 1.315918
+    },
+    {
+      "acc": 0.68063641,
+      "epoch": 1.3535007610350076,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.5948493028825093e-06,
+      "loss": 1.43764133,
+      "memory(GiB)": 123.79,
+      "step": 53355,
+      "train_speed(iter/s)": 1.315932
+    },
+    {
+      "acc": 0.68629684,
+      "epoch": 1.3536276002029426,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.593930023149044e-06,
+      "loss": 1.40691414,
+      "memory(GiB)": 123.79,
+      "step": 53360,
+      "train_speed(iter/s)": 1.315945
+    },
+    {
+      "acc": 0.68449254,
+      "epoch": 1.3537544393708778,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.593010849245279e-06,
+      "loss": 1.35859089,
+      "memory(GiB)": 123.79,
+      "step": 53365,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.67830014,
+      "epoch": 1.3538812785388128,
+      "grad_norm": 3.25,
+      "learning_rate": 2.592091781211643e-06,
+      "loss": 1.38214369,
+      "memory(GiB)": 123.79,
+      "step": 53370,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.70915217,
+      "epoch": 1.3540081177067478,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.591172819088561e-06,
+      "loss": 1.26345024,
+      "memory(GiB)": 123.79,
+      "step": 53375,
+      "train_speed(iter/s)": 1.315987
+    },
+    {
+      "acc": 0.67990618,
+      "epoch": 1.3541349568746828,
+      "grad_norm": 2.5,
+      "learning_rate": 2.590253962916453e-06,
+      "loss": 1.40558386,
+      "memory(GiB)": 123.79,
+      "step": 53380,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.68696251,
+      "epoch": 1.354261796042618,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.5893352127357347e-06,
+      "loss": 1.34858837,
+      "memory(GiB)": 123.79,
+      "step": 53385,
+      "train_speed(iter/s)": 1.316014
+    },
+    {
+      "acc": 0.68771439,
+      "epoch": 1.354388635210553,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.5884165685868164e-06,
+      "loss": 1.36122541,
+      "memory(GiB)": 123.79,
+      "step": 53390,
+      "train_speed(iter/s)": 1.316027
+    },
+    {
+      "acc": 0.67364697,
+      "epoch": 1.3545154743784882,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.5874980305101045e-06,
+      "loss": 1.42915401,
+      "memory(GiB)": 123.79,
+      "step": 53395,
+      "train_speed(iter/s)": 1.316041
+    },
+    {
+      "acc": 0.68372784,
+      "epoch": 1.3546423135464232,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.586579598546e-06,
+      "loss": 1.43396893,
+      "memory(GiB)": 123.79,
+      "step": 53400,
+      "train_speed(iter/s)": 1.316055
+    },
+    {
+      "acc": 0.70202651,
+      "epoch": 1.3547691527143582,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.5856612727348995e-06,
+      "loss": 1.31178904,
+      "memory(GiB)": 123.79,
+      "step": 53405,
+      "train_speed(iter/s)": 1.316069
+    },
+    {
+      "acc": 0.69734707,
+      "epoch": 1.3548959918822931,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.584743053117196e-06,
+      "loss": 1.36333904,
+      "memory(GiB)": 123.79,
+      "step": 53410,
+      "train_speed(iter/s)": 1.316083
+    },
+    {
+      "acc": 0.69944115,
+      "epoch": 1.3550228310502284,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.583824939733277e-06,
+      "loss": 1.30543098,
+      "memory(GiB)": 123.79,
+      "step": 53415,
+      "train_speed(iter/s)": 1.316098
+    },
+    {
+      "acc": 0.67760162,
+      "epoch": 1.3551496702181633,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.5829069326235234e-06,
+      "loss": 1.45841017,
+      "memory(GiB)": 123.79,
+      "step": 53420,
+      "train_speed(iter/s)": 1.316112
+    },
+    {
+      "acc": 0.67277641,
+      "epoch": 1.3552765093860986,
+      "grad_norm": 3.375,
+      "learning_rate": 2.5819890318283137e-06,
+      "loss": 1.38530636,
+      "memory(GiB)": 123.79,
+      "step": 53425,
+      "train_speed(iter/s)": 1.316126
+    },
+    {
+      "acc": 0.69435773,
+      "epoch": 1.3554033485540335,
+      "grad_norm": 3.375,
+      "learning_rate": 2.5810712373880253e-06,
+      "loss": 1.43850241,
+      "memory(GiB)": 123.79,
+      "step": 53430,
+      "train_speed(iter/s)": 1.31614
+    },
+    {
+      "acc": 0.69542642,
+      "epoch": 1.3555301877219685,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.5801535493430215e-06,
+      "loss": 1.36879816,
+      "memory(GiB)": 123.79,
+      "step": 53435,
+      "train_speed(iter/s)": 1.316153
+    },
+    {
+      "acc": 0.7061842,
+      "epoch": 1.3556570268899035,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.5792359677336685e-06,
+      "loss": 1.329006,
+      "memory(GiB)": 123.79,
+      "step": 53440,
+      "train_speed(iter/s)": 1.316167
+    },
+    {
+      "acc": 0.69272461,
+      "epoch": 1.3557838660578387,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.5783184926003237e-06,
+      "loss": 1.37790956,
+      "memory(GiB)": 123.79,
+      "step": 53445,
+      "train_speed(iter/s)": 1.316181
+    },
+    {
+      "acc": 0.67072654,
+      "epoch": 1.3559107052257737,
+      "grad_norm": 3.5,
+      "learning_rate": 2.5774011239833473e-06,
+      "loss": 1.41719837,
+      "memory(GiB)": 123.79,
+      "step": 53450,
+      "train_speed(iter/s)": 1.316195
+    },
+    {
+      "acc": 0.69249172,
+      "epoch": 1.3560375443937087,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.5764838619230843e-06,
+      "loss": 1.39586754,
+      "memory(GiB)": 123.79,
+      "step": 53455,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.7079628,
+      "epoch": 1.356164383561644,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.575566706459879e-06,
+      "loss": 1.3167943,
+      "memory(GiB)": 123.79,
+      "step": 53460,
+      "train_speed(iter/s)": 1.316222
+    },
+    {
+      "acc": 0.68013334,
+      "epoch": 1.356291222729579,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.574649657634076e-06,
+      "loss": 1.41894913,
+      "memory(GiB)": 123.79,
+      "step": 53465,
+      "train_speed(iter/s)": 1.316236
+    },
+    {
+      "acc": 0.69031315,
+      "epoch": 1.356418061897514,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.5737327154860116e-06,
+      "loss": 1.41239243,
+      "memory(GiB)": 123.79,
+      "step": 53470,
+      "train_speed(iter/s)": 1.31625
+    },
+    {
+      "acc": 0.67405872,
+      "epoch": 1.356544901065449,
+      "grad_norm": 3.25,
+      "learning_rate": 2.572815880056011e-06,
+      "loss": 1.47538929,
+      "memory(GiB)": 123.79,
+      "step": 53475,
+      "train_speed(iter/s)": 1.316263
+    },
+    {
+      "acc": 0.69886971,
+      "epoch": 1.356671740233384,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.571899151384406e-06,
+      "loss": 1.32982464,
+      "memory(GiB)": 123.79,
+      "step": 53480,
+      "train_speed(iter/s)": 1.316277
+    },
+    {
+      "acc": 0.68400717,
+      "epoch": 1.356798579401319,
+      "grad_norm": 2.875,
+      "learning_rate": 2.5709825295115178e-06,
+      "loss": 1.42054939,
+      "memory(GiB)": 123.79,
+      "step": 53485,
+      "train_speed(iter/s)": 1.31629
+    },
+    {
+      "acc": 0.68040667,
+      "epoch": 1.3569254185692543,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.5700660144776647e-06,
+      "loss": 1.44608097,
+      "memory(GiB)": 123.79,
+      "step": 53490,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.68086472,
+      "epoch": 1.3570522577371893,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.5691496063231527e-06,
+      "loss": 1.42451019,
+      "memory(GiB)": 123.79,
+      "step": 53495,
+      "train_speed(iter/s)": 1.316318
+    },
+    {
+      "acc": 0.69166422,
+      "epoch": 1.3571790969051243,
+      "grad_norm": 5.625,
+      "learning_rate": 2.568233305088296e-06,
+      "loss": 1.33209362,
+      "memory(GiB)": 123.79,
+      "step": 53500,
+      "train_speed(iter/s)": 1.31633
+    },
+    {
+      "acc": 0.69313116,
+      "epoch": 1.3573059360730593,
+      "grad_norm": 3.0,
+      "learning_rate": 2.5673171108133956e-06,
+      "loss": 1.38070698,
+      "memory(GiB)": 123.79,
+      "step": 53505,
+      "train_speed(iter/s)": 1.316342
+    },
+    {
+      "acc": 0.67460432,
+      "epoch": 1.3574327752409945,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.5664010235387503e-06,
+      "loss": 1.40221338,
+      "memory(GiB)": 123.79,
+      "step": 53510,
+      "train_speed(iter/s)": 1.316356
+    },
+    {
+      "acc": 0.68939161,
+      "epoch": 1.3575596144089295,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.565485043304653e-06,
+      "loss": 1.36241217,
+      "memory(GiB)": 123.79,
+      "step": 53515,
+      "train_speed(iter/s)": 1.31637
+    },
+    {
+      "acc": 0.67894568,
+      "epoch": 1.3576864535768645,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.564569170151392e-06,
+      "loss": 1.3898859,
+      "memory(GiB)": 123.79,
+      "step": 53520,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.6953805,
+      "epoch": 1.3578132927447997,
+      "grad_norm": 2.875,
+      "learning_rate": 2.5636534041192534e-06,
+      "loss": 1.32067833,
+      "memory(GiB)": 123.79,
+      "step": 53525,
+      "train_speed(iter/s)": 1.316397
+    },
+    {
+      "acc": 0.68666229,
+      "epoch": 1.3579401319127347,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.5627377452485153e-06,
+      "loss": 1.38915539,
+      "memory(GiB)": 123.79,
+      "step": 53530,
+      "train_speed(iter/s)": 1.316411
+    },
+    {
+      "acc": 0.70260859,
+      "epoch": 1.3580669710806696,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.561822193579453e-06,
+      "loss": 1.32845516,
+      "memory(GiB)": 123.79,
+      "step": 53535,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.69760551,
+      "epoch": 1.3581938102486046,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.560906749152335e-06,
+      "loss": 1.38981352,
+      "memory(GiB)": 123.79,
+      "step": 53540,
+      "train_speed(iter/s)": 1.316438
+    },
+    {
+      "acc": 0.68336773,
+      "epoch": 1.3583206494165398,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.55999141200743e-06,
+      "loss": 1.36985226,
+      "memory(GiB)": 123.79,
+      "step": 53545,
+      "train_speed(iter/s)": 1.316452
+    },
+    {
+      "acc": 0.69905534,
+      "epoch": 1.3584474885844748,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.5590761821849954e-06,
+      "loss": 1.28756237,
+      "memory(GiB)": 123.79,
+      "step": 53550,
+      "train_speed(iter/s)": 1.316464
+    },
+    {
+      "acc": 0.69499397,
+      "epoch": 1.35857432775241,
+      "grad_norm": 3.375,
+      "learning_rate": 2.5581610597252883e-06,
+      "loss": 1.3728056,
+      "memory(GiB)": 123.79,
+      "step": 53555,
+      "train_speed(iter/s)": 1.316478
+    },
+    {
+      "acc": 0.68827677,
+      "epoch": 1.358701166920345,
+      "grad_norm": 2.875,
+      "learning_rate": 2.5572460446685593e-06,
+      "loss": 1.33030643,
+      "memory(GiB)": 123.79,
+      "step": 53560,
+      "train_speed(iter/s)": 1.316491
+    },
+    {
+      "acc": 0.68185024,
+      "epoch": 1.35882800608828,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.5563311370550535e-06,
+      "loss": 1.38023386,
+      "memory(GiB)": 123.79,
+      "step": 53565,
+      "train_speed(iter/s)": 1.316505
+    },
+    {
+      "acc": 0.67928972,
+      "epoch": 1.358954845256215,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.5554163369250194e-06,
+      "loss": 1.46420946,
+      "memory(GiB)": 123.79,
+      "step": 53570,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.70064812,
+      "epoch": 1.3590816844241502,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.5545016443186867e-06,
+      "loss": 1.34822426,
+      "memory(GiB)": 123.79,
+      "step": 53575,
+      "train_speed(iter/s)": 1.316531
+    },
+    {
+      "acc": 0.68211508,
+      "epoch": 1.3592085235920852,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.55358705927629e-06,
+      "loss": 1.36270266,
+      "memory(GiB)": 123.79,
+      "step": 53580,
+      "train_speed(iter/s)": 1.316543
+    },
+    {
+      "acc": 0.68376079,
+      "epoch": 1.3593353627600204,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.552672581838055e-06,
+      "loss": 1.4079545,
+      "memory(GiB)": 123.79,
+      "step": 53585,
+      "train_speed(iter/s)": 1.316556
+    },
+    {
+      "acc": 0.70933828,
+      "epoch": 1.3594622019279554,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.5517582120442095e-06,
+      "loss": 1.3324605,
+      "memory(GiB)": 123.79,
+      "step": 53590,
+      "train_speed(iter/s)": 1.316569
+    },
+    {
+      "acc": 0.67414055,
+      "epoch": 1.3595890410958904,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.5508439499349675e-06,
+      "loss": 1.41540051,
+      "memory(GiB)": 123.79,
+      "step": 53595,
+      "train_speed(iter/s)": 1.316583
+    },
+    {
+      "acc": 0.66641684,
+      "epoch": 1.3597158802638254,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.549929795550541e-06,
+      "loss": 1.42224941,
+      "memory(GiB)": 123.79,
+      "step": 53600,
+      "train_speed(iter/s)": 1.316597
+    },
+    {
+      "acc": 0.6974278,
+      "epoch": 1.3598427194317606,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.549015748931143e-06,
+      "loss": 1.34594755,
+      "memory(GiB)": 123.79,
+      "step": 53605,
+      "train_speed(iter/s)": 1.31661
+    },
+    {
+      "acc": 0.69177294,
+      "epoch": 1.3599695585996956,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.5481018101169763e-06,
+      "loss": 1.39675684,
+      "memory(GiB)": 123.79,
+      "step": 53610,
+      "train_speed(iter/s)": 1.316624
+    },
+    {
+      "acc": 0.68323545,
+      "epoch": 1.3600963977676306,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.547187979148238e-06,
+      "loss": 1.35551109,
+      "memory(GiB)": 123.79,
+      "step": 53615,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.69782443,
+      "epoch": 1.3602232369355658,
+      "grad_norm": 2.75,
+      "learning_rate": 2.546274256065121e-06,
+      "loss": 1.38915396,
+      "memory(GiB)": 123.79,
+      "step": 53620,
+      "train_speed(iter/s)": 1.316651
+    },
+    {
+      "acc": 0.68560185,
+      "epoch": 1.3603500761035008,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.545360640907819e-06,
+      "loss": 1.36477537,
+      "memory(GiB)": 123.79,
+      "step": 53625,
+      "train_speed(iter/s)": 1.316664
+    },
+    {
+      "acc": 0.6752964,
+      "epoch": 1.3604769152714358,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.544447133716518e-06,
+      "loss": 1.3598361,
+      "memory(GiB)": 123.79,
+      "step": 53630,
+      "train_speed(iter/s)": 1.316678
+    },
+    {
+      "acc": 0.66815734,
+      "epoch": 1.3606037544393708,
+      "grad_norm": 2.875,
+      "learning_rate": 2.5435337345313904e-06,
+      "loss": 1.43831778,
+      "memory(GiB)": 123.79,
+      "step": 53635,
+      "train_speed(iter/s)": 1.316692
+    },
+    {
+      "acc": 0.68133597,
+      "epoch": 1.360730593607306,
+      "grad_norm": 3.5,
+      "learning_rate": 2.5426204433926194e-06,
+      "loss": 1.36630726,
+      "memory(GiB)": 123.79,
+      "step": 53640,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.68147902,
+      "epoch": 1.360857432775241,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.541707260340372e-06,
+      "loss": 1.43255634,
+      "memory(GiB)": 123.79,
+      "step": 53645,
+      "train_speed(iter/s)": 1.316718
+    },
+    {
+      "acc": 0.69196897,
+      "epoch": 1.3609842719431762,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.5407941854148156e-06,
+      "loss": 1.40492802,
+      "memory(GiB)": 123.79,
+      "step": 53650,
+      "train_speed(iter/s)": 1.316732
+    },
+    {
+      "acc": 0.67023411,
+      "epoch": 1.3611111111111112,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.5398812186561095e-06,
+      "loss": 1.46028433,
+      "memory(GiB)": 123.79,
+      "step": 53655,
+      "train_speed(iter/s)": 1.316746
+    },
+    {
+      "acc": 0.67529144,
+      "epoch": 1.3612379502790461,
+      "grad_norm": 2.625,
+      "learning_rate": 2.5389683601044114e-06,
+      "loss": 1.41044426,
+      "memory(GiB)": 123.79,
+      "step": 53660,
+      "train_speed(iter/s)": 1.316759
+    },
+    {
+      "acc": 0.69094892,
+      "epoch": 1.3613647894469811,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.538055609799873e-06,
+      "loss": 1.38699684,
+      "memory(GiB)": 123.79,
+      "step": 53665,
+      "train_speed(iter/s)": 1.316774
+    },
+    {
+      "acc": 0.69063425,
+      "epoch": 1.3614916286149163,
+      "grad_norm": 3.125,
+      "learning_rate": 2.5371429677826397e-06,
+      "loss": 1.37549877,
+      "memory(GiB)": 123.79,
+      "step": 53670,
+      "train_speed(iter/s)": 1.316787
+    },
+    {
+      "acc": 0.70494218,
+      "epoch": 1.3616184677828513,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.5362304340928556e-06,
+      "loss": 1.31567574,
+      "memory(GiB)": 123.79,
+      "step": 53675,
+      "train_speed(iter/s)": 1.316802
+    },
+    {
+      "acc": 0.67744064,
+      "epoch": 1.3617453069507863,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.535318008770656e-06,
+      "loss": 1.43843021,
+      "memory(GiB)": 123.79,
+      "step": 53680,
+      "train_speed(iter/s)": 1.316817
+    },
+    {
+      "acc": 0.68647475,
+      "epoch": 1.3618721461187215,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.534405691856175e-06,
+      "loss": 1.37245235,
+      "memory(GiB)": 123.79,
+      "step": 53685,
+      "train_speed(iter/s)": 1.316831
+    },
+    {
+      "acc": 0.69825778,
+      "epoch": 1.3619989852866565,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.5334934833895396e-06,
+      "loss": 1.33778267,
+      "memory(GiB)": 123.79,
+      "step": 53690,
+      "train_speed(iter/s)": 1.316846
+    },
+    {
+      "acc": 0.68046446,
+      "epoch": 1.3621258244545915,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.5325813834108724e-06,
+      "loss": 1.40830421,
+      "memory(GiB)": 123.79,
+      "step": 53695,
+      "train_speed(iter/s)": 1.31686
+    },
+    {
+      "acc": 0.70060081,
+      "epoch": 1.3622526636225265,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.531669391960293e-06,
+      "loss": 1.35744972,
+      "memory(GiB)": 123.79,
+      "step": 53700,
+      "train_speed(iter/s)": 1.316873
+    },
+    {
+      "acc": 0.69007111,
+      "epoch": 1.3623795027904617,
+      "grad_norm": 2.75,
+      "learning_rate": 2.5307575090779125e-06,
+      "loss": 1.43064041,
+      "memory(GiB)": 123.79,
+      "step": 53705,
+      "train_speed(iter/s)": 1.316888
+    },
+    {
+      "acc": 0.68203702,
+      "epoch": 1.3625063419583967,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.529845734803844e-06,
+      "loss": 1.38009701,
+      "memory(GiB)": 123.79,
+      "step": 53710,
+      "train_speed(iter/s)": 1.316901
+    },
+    {
+      "acc": 0.68918009,
+      "epoch": 1.362633181126332,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.5289340691781872e-06,
+      "loss": 1.44889841,
+      "memory(GiB)": 123.79,
+      "step": 53715,
+      "train_speed(iter/s)": 1.316916
+    },
+    {
+      "acc": 0.69381199,
+      "epoch": 1.362760020294267,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.528022512241042e-06,
+      "loss": 1.36936111,
+      "memory(GiB)": 123.79,
+      "step": 53720,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.69807944,
+      "epoch": 1.362886859462202,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.5271110640325013e-06,
+      "loss": 1.37255678,
+      "memory(GiB)": 123.79,
+      "step": 53725,
+      "train_speed(iter/s)": 1.316946
+    },
+    {
+      "acc": 0.68467951,
+      "epoch": 1.3630136986301369,
+      "grad_norm": 2.75,
+      "learning_rate": 2.5261997245926612e-06,
+      "loss": 1.39462757,
+      "memory(GiB)": 123.79,
+      "step": 53730,
+      "train_speed(iter/s)": 1.31696
+    },
+    {
+      "acc": 0.67952147,
+      "epoch": 1.363140537798072,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.5252884939615995e-06,
+      "loss": 1.40867853,
+      "memory(GiB)": 123.79,
+      "step": 53735,
+      "train_speed(iter/s)": 1.316975
+    },
+    {
+      "acc": 0.68658981,
+      "epoch": 1.363267376966007,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.5243773721793973e-06,
+      "loss": 1.38068609,
+      "memory(GiB)": 123.79,
+      "step": 53740,
+      "train_speed(iter/s)": 1.316989
+    },
+    {
+      "acc": 0.69521952,
+      "epoch": 1.3633942161339423,
+      "grad_norm": 4.0,
+      "learning_rate": 2.5234663592861325e-06,
+      "loss": 1.33480577,
+      "memory(GiB)": 123.79,
+      "step": 53745,
+      "train_speed(iter/s)": 1.317003
+    },
+    {
+      "acc": 0.68367305,
+      "epoch": 1.3635210553018773,
+      "grad_norm": 3.25,
+      "learning_rate": 2.522555455321876e-06,
+      "loss": 1.37150908,
+      "memory(GiB)": 123.79,
+      "step": 53750,
+      "train_speed(iter/s)": 1.317018
+    },
+    {
+      "acc": 0.68528919,
+      "epoch": 1.3636478944698123,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.52164466032669e-06,
+      "loss": 1.36773434,
+      "memory(GiB)": 123.79,
+      "step": 53755,
+      "train_speed(iter/s)": 1.317032
+    },
+    {
+      "acc": 0.69914083,
+      "epoch": 1.3637747336377473,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.5207339743406344e-06,
+      "loss": 1.40404072,
+      "memory(GiB)": 123.79,
+      "step": 53760,
+      "train_speed(iter/s)": 1.317046
+    },
+    {
+      "acc": 0.67771368,
+      "epoch": 1.3639015728056825,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.5198233974037705e-06,
+      "loss": 1.3976346,
+      "memory(GiB)": 123.79,
+      "step": 53765,
+      "train_speed(iter/s)": 1.31706
+    },
+    {
+      "acc": 0.67936611,
+      "epoch": 1.3640284119736175,
+      "grad_norm": 3.5,
+      "learning_rate": 2.5189129295561486e-06,
+      "loss": 1.4215086,
+      "memory(GiB)": 123.79,
+      "step": 53770,
+      "train_speed(iter/s)": 1.317073
+    },
+    {
+      "acc": 0.68730226,
+      "epoch": 1.3641552511415524,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.518002570837809e-06,
+      "loss": 1.38631868,
+      "memory(GiB)": 123.79,
+      "step": 53775,
+      "train_speed(iter/s)": 1.317087
+    },
+    {
+      "acc": 0.67824421,
+      "epoch": 1.3642820903094877,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.5170923212887997e-06,
+      "loss": 1.47904902,
+      "memory(GiB)": 123.79,
+      "step": 53780,
+      "train_speed(iter/s)": 1.317102
+    },
+    {
+      "acc": 0.7089448,
+      "epoch": 1.3644089294774226,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.5161821809491554e-06,
+      "loss": 1.34575939,
+      "memory(GiB)": 123.79,
+      "step": 53785,
+      "train_speed(iter/s)": 1.317116
+    },
+    {
+      "acc": 0.69152641,
+      "epoch": 1.3645357686453576,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.5152721498589104e-06,
+      "loss": 1.36558723,
+      "memory(GiB)": 123.79,
+      "step": 53790,
+      "train_speed(iter/s)": 1.31713
+    },
+    {
+      "acc": 0.68140373,
+      "epoch": 1.3646626078132926,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.514362228058086e-06,
+      "loss": 1.41569595,
+      "memory(GiB)": 123.79,
+      "step": 53795,
+      "train_speed(iter/s)": 1.317144
+    },
+    {
+      "acc": 0.68139963,
+      "epoch": 1.3647894469812278,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.51345241558671e-06,
+      "loss": 1.45141335,
+      "memory(GiB)": 123.79,
+      "step": 53800,
+      "train_speed(iter/s)": 1.317158
+    },
+    {
+      "acc": 0.67651868,
+      "epoch": 1.3649162861491628,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.5125427124847985e-06,
+      "loss": 1.44712715,
+      "memory(GiB)": 123.79,
+      "step": 53805,
+      "train_speed(iter/s)": 1.317172
+    },
+    {
+      "acc": 0.68899894,
+      "epoch": 1.365043125317098,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.5116331187923645e-06,
+      "loss": 1.34351397,
+      "memory(GiB)": 123.79,
+      "step": 53810,
+      "train_speed(iter/s)": 1.317184
+    },
+    {
+      "acc": 0.69195557,
+      "epoch": 1.365169964485033,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.510723634549415e-06,
+      "loss": 1.37948341,
+      "memory(GiB)": 123.79,
+      "step": 53815,
+      "train_speed(iter/s)": 1.317198
+    },
+    {
+      "acc": 0.69596186,
+      "epoch": 1.365296803652968,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.509814259795954e-06,
+      "loss": 1.3823226,
+      "memory(GiB)": 123.79,
+      "step": 53820,
+      "train_speed(iter/s)": 1.317212
+    },
+    {
+      "acc": 0.70106568,
+      "epoch": 1.365423642820903,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.50890499457198e-06,
+      "loss": 1.3799674,
+      "memory(GiB)": 123.79,
+      "step": 53825,
+      "train_speed(iter/s)": 1.317226
+    },
+    {
+      "acc": 0.70093393,
+      "epoch": 1.3655504819888382,
+      "grad_norm": 3.0,
+      "learning_rate": 2.5079958389174865e-06,
+      "loss": 1.31725464,
+      "memory(GiB)": 123.79,
+      "step": 53830,
+      "train_speed(iter/s)": 1.31724
+    },
+    {
+      "acc": 0.68027024,
+      "epoch": 1.3656773211567732,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.5070867928724618e-06,
+      "loss": 1.46044941,
+      "memory(GiB)": 123.79,
+      "step": 53835,
+      "train_speed(iter/s)": 1.317253
+    },
+    {
+      "acc": 0.6915431,
+      "epoch": 1.3658041603247082,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.50617785647689e-06,
+      "loss": 1.33997192,
+      "memory(GiB)": 123.79,
+      "step": 53840,
+      "train_speed(iter/s)": 1.317268
+    },
+    {
+      "acc": 0.67869577,
+      "epoch": 1.3659309994926434,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.5052690297707506e-06,
+      "loss": 1.38716946,
+      "memory(GiB)": 123.79,
+      "step": 53845,
+      "train_speed(iter/s)": 1.317278
+    },
+    {
+      "acc": 0.69586716,
+      "epoch": 1.3660578386605784,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.5043603127940164e-06,
+      "loss": 1.28512774,
+      "memory(GiB)": 123.79,
+      "step": 53850,
+      "train_speed(iter/s)": 1.317292
+    },
+    {
+      "acc": 0.68662014,
+      "epoch": 1.3661846778285134,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.503451705586659e-06,
+      "loss": 1.39796886,
+      "memory(GiB)": 123.79,
+      "step": 53855,
+      "train_speed(iter/s)": 1.317306
+    },
+    {
+      "acc": 0.7063426,
+      "epoch": 1.3663115169964484,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.5025432081886412e-06,
+      "loss": 1.33705235,
+      "memory(GiB)": 123.79,
+      "step": 53860,
+      "train_speed(iter/s)": 1.317319
+    },
+    {
+      "acc": 0.6861691,
+      "epoch": 1.3664383561643836,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.5016348206399215e-06,
+      "loss": 1.34778671,
+      "memory(GiB)": 123.79,
+      "step": 53865,
+      "train_speed(iter/s)": 1.317333
+    },
+    {
+      "acc": 0.68676643,
+      "epoch": 1.3665651953323186,
+      "grad_norm": 3.625,
+      "learning_rate": 2.500726542980461e-06,
+      "loss": 1.43621998,
+      "memory(GiB)": 123.79,
+      "step": 53870,
+      "train_speed(iter/s)": 1.317346
+    },
+    {
+      "acc": 0.68296537,
+      "epoch": 1.3666920345002538,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.499818375250204e-06,
+      "loss": 1.38799438,
+      "memory(GiB)": 123.79,
+      "step": 53875,
+      "train_speed(iter/s)": 1.31736
+    },
+    {
+      "acc": 0.68620815,
+      "epoch": 1.3668188736681888,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.4989103174890946e-06,
+      "loss": 1.4169322,
+      "memory(GiB)": 123.79,
+      "step": 53880,
+      "train_speed(iter/s)": 1.317373
+    },
+    {
+      "acc": 0.68614726,
+      "epoch": 1.3669457128361238,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.498002369737078e-06,
+      "loss": 1.43029575,
+      "memory(GiB)": 123.79,
+      "step": 53885,
+      "train_speed(iter/s)": 1.317386
+    },
+    {
+      "acc": 0.69539785,
+      "epoch": 1.3670725520040587,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.49709453203409e-06,
+      "loss": 1.36367474,
+      "memory(GiB)": 123.79,
+      "step": 53890,
+      "train_speed(iter/s)": 1.3174
+    },
+    {
+      "acc": 0.69157085,
+      "epoch": 1.367199391171994,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.496186804420057e-06,
+      "loss": 1.41330719,
+      "memory(GiB)": 123.79,
+      "step": 53895,
+      "train_speed(iter/s)": 1.317413
+    },
+    {
+      "acc": 0.69561028,
+      "epoch": 1.367326230339929,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.4952791869349056e-06,
+      "loss": 1.39806709,
+      "memory(GiB)": 123.79,
+      "step": 53900,
+      "train_speed(iter/s)": 1.317426
+    },
+    {
+      "acc": 0.69665518,
+      "epoch": 1.3674530695078642,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.4943716796185603e-06,
+      "loss": 1.38740273,
+      "memory(GiB)": 123.79,
+      "step": 53905,
+      "train_speed(iter/s)": 1.31744
+    },
+    {
+      "acc": 0.69113088,
+      "epoch": 1.3675799086757991,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.493464282510937e-06,
+      "loss": 1.4586647,
+      "memory(GiB)": 123.79,
+      "step": 53910,
+      "train_speed(iter/s)": 1.317453
+    },
+    {
+      "acc": 0.6968606,
+      "epoch": 1.3677067478437341,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.4925569956519414e-06,
+      "loss": 1.3459343,
+      "memory(GiB)": 123.79,
+      "step": 53915,
+      "train_speed(iter/s)": 1.317467
+    },
+    {
+      "acc": 0.67923212,
+      "epoch": 1.3678335870116691,
+      "grad_norm": 3.0,
+      "learning_rate": 2.491649819081486e-06,
+      "loss": 1.4503397,
+      "memory(GiB)": 123.79,
+      "step": 53920,
+      "train_speed(iter/s)": 1.317481
+    },
+    {
+      "acc": 0.69901128,
+      "epoch": 1.3679604261796043,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.490742752839471e-06,
+      "loss": 1.42300186,
+      "memory(GiB)": 123.79,
+      "step": 53925,
+      "train_speed(iter/s)": 1.317494
+    },
+    {
+      "acc": 0.69033575,
+      "epoch": 1.3680872653475393,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.4898357969657943e-06,
+      "loss": 1.352244,
+      "memory(GiB)": 123.79,
+      "step": 53930,
+      "train_speed(iter/s)": 1.317508
+    },
+    {
+      "acc": 0.69464235,
+      "epoch": 1.3682141045154743,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.4889289515003425e-06,
+      "loss": 1.38238802,
+      "memory(GiB)": 123.79,
+      "step": 53935,
+      "train_speed(iter/s)": 1.317522
+    },
+    {
+      "acc": 0.70109968,
+      "epoch": 1.3683409436834095,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.4880222164830085e-06,
+      "loss": 1.37545748,
+      "memory(GiB)": 123.79,
+      "step": 53940,
+      "train_speed(iter/s)": 1.317536
+    },
+    {
+      "acc": 0.69525261,
+      "epoch": 1.3684677828513445,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.4871155919536725e-06,
+      "loss": 1.38153124,
+      "memory(GiB)": 123.79,
+      "step": 53945,
+      "train_speed(iter/s)": 1.317549
+    },
+    {
+      "acc": 0.7016181,
+      "epoch": 1.3685946220192795,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.486209077952212e-06,
+      "loss": 1.34313898,
+      "memory(GiB)": 123.79,
+      "step": 53950,
+      "train_speed(iter/s)": 1.317561
+    },
+    {
+      "acc": 0.69838939,
+      "epoch": 1.3687214611872145,
+      "grad_norm": 4.5,
+      "learning_rate": 2.4853026745185e-06,
+      "loss": 1.38035736,
+      "memory(GiB)": 123.79,
+      "step": 53955,
+      "train_speed(iter/s)": 1.317575
+    },
+    {
+      "acc": 0.68897057,
+      "epoch": 1.3688483003551497,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.4843963816924035e-06,
+      "loss": 1.35681629,
+      "memory(GiB)": 123.79,
+      "step": 53960,
+      "train_speed(iter/s)": 1.317589
+    },
+    {
+      "acc": 0.69347792,
+      "epoch": 1.3689751395230847,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.483490199513785e-06,
+      "loss": 1.34610357,
+      "memory(GiB)": 123.79,
+      "step": 53965,
+      "train_speed(iter/s)": 1.317602
+    },
+    {
+      "acc": 0.69565086,
+      "epoch": 1.36910197869102,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.4825841280225033e-06,
+      "loss": 1.39764023,
+      "memory(GiB)": 123.79,
+      "step": 53970,
+      "train_speed(iter/s)": 1.317616
+    },
+    {
+      "acc": 0.67833509,
+      "epoch": 1.369228817858955,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.4816781672584107e-06,
+      "loss": 1.40487766,
+      "memory(GiB)": 123.79,
+      "step": 53975,
+      "train_speed(iter/s)": 1.31763
+    },
+    {
+      "acc": 0.67561154,
+      "epoch": 1.3693556570268899,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.480772317261356e-06,
+      "loss": 1.42353344,
+      "memory(GiB)": 123.79,
+      "step": 53980,
+      "train_speed(iter/s)": 1.317643
+    },
+    {
+      "acc": 0.68738203,
+      "epoch": 1.3694824961948249,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.479866578071183e-06,
+      "loss": 1.38177986,
+      "memory(GiB)": 123.79,
+      "step": 53985,
+      "train_speed(iter/s)": 1.317657
+    },
+    {
+      "acc": 0.6860899,
+      "epoch": 1.36960933536276,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.4789609497277284e-06,
+      "loss": 1.36563473,
+      "memory(GiB)": 123.79,
+      "step": 53990,
+      "train_speed(iter/s)": 1.317671
+    },
+    {
+      "acc": 0.69874821,
+      "epoch": 1.369736174530695,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.478055432270828e-06,
+      "loss": 1.38967876,
+      "memory(GiB)": 123.79,
+      "step": 53995,
+      "train_speed(iter/s)": 1.317685
+    },
+    {
+      "acc": 0.71971107,
+      "epoch": 1.36986301369863,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.4771500257403086e-06,
+      "loss": 1.22254562,
+      "memory(GiB)": 123.79,
+      "step": 54000,
+      "train_speed(iter/s)": 1.317698
+    },
+    {
+      "epoch": 1.36986301369863,
+      "eval_acc": 0.6750791990093076,
+      "eval_loss": 1.3563014268875122,
+      "eval_runtime": 69.3627,
+      "eval_samples_per_second": 91.836,
+      "eval_steps_per_second": 22.966,
+      "step": 54000
+    },
+    {
+      "acc": 0.69036818,
+      "epoch": 1.3699898528665653,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.476244730175993e-06,
+      "loss": 1.36506958,
+      "memory(GiB)": 123.79,
+      "step": 54005,
+      "train_speed(iter/s)": 1.315088
+    },
+    {
+      "acc": 0.69059286,
+      "epoch": 1.3701166920345003,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.4753395456177056e-06,
+      "loss": 1.40792503,
+      "memory(GiB)": 123.79,
+      "step": 54010,
+      "train_speed(iter/s)": 1.315103
+    },
+    {
+      "acc": 0.68379898,
+      "epoch": 1.3702435312024352,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.474434472105255e-06,
+      "loss": 1.4085125,
+      "memory(GiB)": 123.79,
+      "step": 54015,
+      "train_speed(iter/s)": 1.315118
+    },
+    {
+      "acc": 0.68128648,
+      "epoch": 1.3703703703703702,
+      "grad_norm": 3.375,
+      "learning_rate": 2.473529509678452e-06,
+      "loss": 1.34597626,
+      "memory(GiB)": 123.79,
+      "step": 54020,
+      "train_speed(iter/s)": 1.315132
+    },
+    {
+      "acc": 0.68188243,
+      "epoch": 1.3704972095383054,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.4726246583770996e-06,
+      "loss": 1.45718889,
+      "memory(GiB)": 123.79,
+      "step": 54025,
+      "train_speed(iter/s)": 1.315146
+    },
+    {
+      "acc": 0.68799047,
+      "epoch": 1.3706240487062404,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.4717199182410025e-06,
+      "loss": 1.34009838,
+      "memory(GiB)": 123.79,
+      "step": 54030,
+      "train_speed(iter/s)": 1.315161
+    },
+    {
+      "acc": 0.70171509,
+      "epoch": 1.3707508878741756,
+      "grad_norm": 4.0,
+      "learning_rate": 2.4708152893099493e-06,
+      "loss": 1.36220255,
+      "memory(GiB)": 123.79,
+      "step": 54035,
+      "train_speed(iter/s)": 1.315175
+    },
+    {
+      "acc": 0.69608078,
+      "epoch": 1.3708777270421106,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.4699107716237293e-06,
+      "loss": 1.35186014,
+      "memory(GiB)": 123.79,
+      "step": 54040,
+      "train_speed(iter/s)": 1.31519
+    },
+    {
+      "acc": 0.69050941,
+      "epoch": 1.3710045662100456,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.469006365222132e-06,
+      "loss": 1.34314346,
+      "memory(GiB)": 123.79,
+      "step": 54045,
+      "train_speed(iter/s)": 1.315204
+    },
+    {
+      "acc": 0.67837458,
+      "epoch": 1.3711314053779806,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.4681020701449365e-06,
+      "loss": 1.39586048,
+      "memory(GiB)": 123.79,
+      "step": 54050,
+      "train_speed(iter/s)": 1.315218
+    },
+    {
+      "acc": 0.67923427,
+      "epoch": 1.3712582445459158,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.4671978864319123e-06,
+      "loss": 1.39309635,
+      "memory(GiB)": 123.79,
+      "step": 54055,
+      "train_speed(iter/s)": 1.315233
+    },
+    {
+      "acc": 0.68539691,
+      "epoch": 1.3713850837138508,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.466293814122835e-06,
+      "loss": 1.43935318,
+      "memory(GiB)": 123.79,
+      "step": 54060,
+      "train_speed(iter/s)": 1.315246
+    },
+    {
+      "acc": 0.70563259,
+      "epoch": 1.371511922881786,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.4653898532574684e-06,
+      "loss": 1.34350185,
+      "memory(GiB)": 123.79,
+      "step": 54065,
+      "train_speed(iter/s)": 1.315261
+    },
+    {
+      "acc": 0.68878965,
+      "epoch": 1.371638762049721,
+      "grad_norm": 3.125,
+      "learning_rate": 2.4644860038755737e-06,
+      "loss": 1.3831687,
+      "memory(GiB)": 123.79,
+      "step": 54070,
+      "train_speed(iter/s)": 1.315275
+    },
+    {
+      "acc": 0.6845644,
+      "epoch": 1.371765601217656,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.4635822660169007e-06,
+      "loss": 1.42964497,
+      "memory(GiB)": 123.79,
+      "step": 54075,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.67684107,
+      "epoch": 1.371892440385591,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.4626786397212065e-06,
+      "loss": 1.47139263,
+      "memory(GiB)": 123.79,
+      "step": 54080,
+      "train_speed(iter/s)": 1.315305
+    },
+    {
+      "acc": 0.70195026,
+      "epoch": 1.3720192795535262,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.461775125028234e-06,
+      "loss": 1.39478159,
+      "memory(GiB)": 123.79,
+      "step": 54085,
+      "train_speed(iter/s)": 1.315318
+    },
+    {
+      "acc": 0.6743392,
+      "epoch": 1.3721461187214612,
+      "grad_norm": 5.25,
+      "learning_rate": 2.4608717219777236e-06,
+      "loss": 1.4395709,
+      "memory(GiB)": 123.79,
+      "step": 54090,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.68653293,
+      "epoch": 1.3722729578893962,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.459968430609411e-06,
+      "loss": 1.36380644,
+      "memory(GiB)": 123.79,
+      "step": 54095,
+      "train_speed(iter/s)": 1.315348
+    },
+    {
+      "acc": 0.69929075,
+      "epoch": 1.3723997970573314,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.459065250963028e-06,
+      "loss": 1.35112953,
+      "memory(GiB)": 123.79,
+      "step": 54100,
+      "train_speed(iter/s)": 1.315363
+    },
+    {
+      "acc": 0.70308485,
+      "epoch": 1.3725266362252664,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.458162183078299e-06,
+      "loss": 1.35125465,
+      "memory(GiB)": 123.79,
+      "step": 54105,
+      "train_speed(iter/s)": 1.315377
+    },
+    {
+      "acc": 0.70484037,
+      "epoch": 1.3726534753932014,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.4572592269949464e-06,
+      "loss": 1.29116278,
+      "memory(GiB)": 123.79,
+      "step": 54110,
+      "train_speed(iter/s)": 1.315392
+    },
+    {
+      "acc": 0.68918328,
+      "epoch": 1.3727803145611364,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.4563563827526848e-06,
+      "loss": 1.38746738,
+      "memory(GiB)": 123.79,
+      "step": 54115,
+      "train_speed(iter/s)": 1.315407
+    },
+    {
+      "acc": 0.67821798,
+      "epoch": 1.3729071537290716,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.455453650391226e-06,
+      "loss": 1.3972681,
+      "memory(GiB)": 123.79,
+      "step": 54120,
+      "train_speed(iter/s)": 1.315421
+    },
+    {
+      "acc": 0.69075174,
+      "epoch": 1.3730339928970066,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.454551029950277e-06,
+      "loss": 1.39884996,
+      "memory(GiB)": 123.79,
+      "step": 54125,
+      "train_speed(iter/s)": 1.315436
+    },
+    {
+      "acc": 0.70569344,
+      "epoch": 1.3731608320649418,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.4536485214695377e-06,
+      "loss": 1.28682289,
+      "memory(GiB)": 123.79,
+      "step": 54130,
+      "train_speed(iter/s)": 1.315451
+    },
+    {
+      "acc": 0.68599157,
+      "epoch": 1.3732876712328768,
+      "grad_norm": 2.5,
+      "learning_rate": 2.4527461249887054e-06,
+      "loss": 1.38165741,
+      "memory(GiB)": 123.79,
+      "step": 54135,
+      "train_speed(iter/s)": 1.315463
+    },
+    {
+      "acc": 0.68783159,
+      "epoch": 1.3734145104008117,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.451843840547471e-06,
+      "loss": 1.35843697,
+      "memory(GiB)": 123.79,
+      "step": 54140,
+      "train_speed(iter/s)": 1.315478
+    },
+    {
+      "acc": 0.69824858,
+      "epoch": 1.3735413495687467,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.4509416681855193e-06,
+      "loss": 1.40946703,
+      "memory(GiB)": 123.79,
+      "step": 54145,
+      "train_speed(iter/s)": 1.315492
+    },
+    {
+      "acc": 0.69145069,
+      "epoch": 1.373668188736682,
+      "grad_norm": 3.375,
+      "learning_rate": 2.4500396079425377e-06,
+      "loss": 1.41933346,
+      "memory(GiB)": 123.79,
+      "step": 54150,
+      "train_speed(iter/s)": 1.315507
+    },
+    {
+      "acc": 0.68985677,
+      "epoch": 1.373795027904617,
+      "grad_norm": 4.0,
+      "learning_rate": 2.4491376598581967e-06,
+      "loss": 1.40192366,
+      "memory(GiB)": 123.79,
+      "step": 54155,
+      "train_speed(iter/s)": 1.315521
+    },
+    {
+      "acc": 0.68556771,
+      "epoch": 1.373921867072552,
+      "grad_norm": 3.75,
+      "learning_rate": 2.4482358239721704e-06,
+      "loss": 1.4029582,
+      "memory(GiB)": 123.79,
+      "step": 54160,
+      "train_speed(iter/s)": 1.315536
+    },
+    {
+      "acc": 0.68768988,
+      "epoch": 1.3740487062404871,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.4473341003241234e-06,
+      "loss": 1.4031353,
+      "memory(GiB)": 123.79,
+      "step": 54165,
+      "train_speed(iter/s)": 1.31555
+    },
+    {
+      "acc": 0.69896297,
+      "epoch": 1.3741755454084221,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.446432488953724e-06,
+      "loss": 1.25903015,
+      "memory(GiB)": 123.79,
+      "step": 54170,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.67429247,
+      "epoch": 1.3743023845763571,
+      "grad_norm": 3.75,
+      "learning_rate": 2.445530989900622e-06,
+      "loss": 1.38193407,
+      "memory(GiB)": 123.79,
+      "step": 54175,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.68684044,
+      "epoch": 1.374429223744292,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.4446296032044697e-06,
+      "loss": 1.42919941,
+      "memory(GiB)": 123.79,
+      "step": 54180,
+      "train_speed(iter/s)": 1.315592
+    },
+    {
+      "acc": 0.69304352,
+      "epoch": 1.3745560629122273,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.443728328904919e-06,
+      "loss": 1.3794795,
+      "memory(GiB)": 123.79,
+      "step": 54185,
+      "train_speed(iter/s)": 1.315607
+    },
+    {
+      "acc": 0.67522106,
+      "epoch": 1.3746829020801623,
+      "grad_norm": 3.625,
+      "learning_rate": 2.442827167041611e-06,
+      "loss": 1.42459145,
+      "memory(GiB)": 123.79,
+      "step": 54190,
+      "train_speed(iter/s)": 1.315621
+    },
+    {
+      "acc": 0.69755211,
+      "epoch": 1.3748097412480975,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.441926117654179e-06,
+      "loss": 1.38699703,
+      "memory(GiB)": 123.79,
+      "step": 54195,
+      "train_speed(iter/s)": 1.315635
+    },
+    {
+      "acc": 0.68502579,
+      "epoch": 1.3749365804160325,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.4410251807822555e-06,
+      "loss": 1.44422359,
+      "memory(GiB)": 123.79,
+      "step": 54200,
+      "train_speed(iter/s)": 1.315649
+    },
+    {
+      "acc": 0.69530497,
+      "epoch": 1.3750634195839675,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.4401243564654713e-06,
+      "loss": 1.3812252,
+      "memory(GiB)": 123.79,
+      "step": 54205,
+      "train_speed(iter/s)": 1.315662
+    },
+    {
+      "acc": 0.69776263,
+      "epoch": 1.3751902587519025,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.4392236447434494e-06,
+      "loss": 1.33877125,
+      "memory(GiB)": 123.79,
+      "step": 54210,
+      "train_speed(iter/s)": 1.315676
+    },
+    {
+      "acc": 0.69369884,
+      "epoch": 1.3753170979198377,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.4383230456558005e-06,
+      "loss": 1.34390602,
+      "memory(GiB)": 123.79,
+      "step": 54215,
+      "train_speed(iter/s)": 1.31569
+    },
+    {
+      "acc": 0.68946524,
+      "epoch": 1.3754439370877727,
+      "grad_norm": 3.5,
+      "learning_rate": 2.437422559242143e-06,
+      "loss": 1.35553064,
+      "memory(GiB)": 123.79,
+      "step": 54220,
+      "train_speed(iter/s)": 1.315698
+    },
+    {
+      "acc": 0.69186964,
+      "epoch": 1.375570776255708,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.4365221855420822e-06,
+      "loss": 1.42359438,
+      "memory(GiB)": 123.79,
+      "step": 54225,
+      "train_speed(iter/s)": 1.31555
+    },
+    {
+      "acc": 0.68716626,
+      "epoch": 1.3756976154236429,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.435621924595221e-06,
+      "loss": 1.34418945,
+      "memory(GiB)": 123.79,
+      "step": 54230,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.69533901,
+      "epoch": 1.3758244545915779,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.4347217764411567e-06,
+      "loss": 1.43547726,
+      "memory(GiB)": 123.79,
+      "step": 54235,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.70341287,
+      "epoch": 1.3759512937595129,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.433821741119482e-06,
+      "loss": 1.33537722,
+      "memory(GiB)": 123.79,
+      "step": 54240,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.67288103,
+      "epoch": 1.376078132927448,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.432921818669784e-06,
+      "loss": 1.42710266,
+      "memory(GiB)": 123.79,
+      "step": 54245,
+      "train_speed(iter/s)": 1.315605
+    },
+    {
+      "acc": 0.69891434,
+      "epoch": 1.376204972095383,
+      "grad_norm": 2.875,
+      "learning_rate": 2.432022009131646e-06,
+      "loss": 1.35497761,
+      "memory(GiB)": 123.79,
+      "step": 54250,
+      "train_speed(iter/s)": 1.315619
+    },
+    {
+      "acc": 0.6766572,
+      "epoch": 1.376331811263318,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.4311223125446447e-06,
+      "loss": 1.47100792,
+      "memory(GiB)": 123.79,
+      "step": 54255,
+      "train_speed(iter/s)": 1.315633
+    },
+    {
+      "acc": 0.67889333,
+      "epoch": 1.3764586504312533,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.4302227289483537e-06,
+      "loss": 1.45989428,
+      "memory(GiB)": 123.79,
+      "step": 54260,
+      "train_speed(iter/s)": 1.315647
+    },
+    {
+      "acc": 0.68261886,
+      "epoch": 1.3765854895991883,
+      "grad_norm": 3.5,
+      "learning_rate": 2.42932325838234e-06,
+      "loss": 1.40986633,
+      "memory(GiB)": 123.79,
+      "step": 54265,
+      "train_speed(iter/s)": 1.31566
+    },
+    {
+      "acc": 0.69695244,
+      "epoch": 1.3767123287671232,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.4284239008861665e-06,
+      "loss": 1.39106445,
+      "memory(GiB)": 123.79,
+      "step": 54270,
+      "train_speed(iter/s)": 1.315674
+    },
+    {
+      "acc": 0.68901601,
+      "epoch": 1.3768391679350582,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.4275246564993917e-06,
+      "loss": 1.41770134,
+      "memory(GiB)": 123.79,
+      "step": 54275,
+      "train_speed(iter/s)": 1.315687
+    },
+    {
+      "acc": 0.6949491,
+      "epoch": 1.3769660071029934,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.426625525261567e-06,
+      "loss": 1.33218346,
+      "memory(GiB)": 123.79,
+      "step": 54280,
+      "train_speed(iter/s)": 1.315701
+    },
+    {
+      "acc": 0.68982277,
+      "epoch": 1.3770928462709284,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.425726507212242e-06,
+      "loss": 1.39835873,
+      "memory(GiB)": 123.79,
+      "step": 54285,
+      "train_speed(iter/s)": 1.315714
+    },
+    {
+      "acc": 0.68697152,
+      "epoch": 1.3772196854388636,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.424827602390958e-06,
+      "loss": 1.40499077,
+      "memory(GiB)": 123.79,
+      "step": 54290,
+      "train_speed(iter/s)": 1.315727
+    },
+    {
+      "acc": 0.68678532,
+      "epoch": 1.3773465246067986,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.4239288108372534e-06,
+      "loss": 1.38048077,
+      "memory(GiB)": 123.79,
+      "step": 54295,
+      "train_speed(iter/s)": 1.31574
+    },
+    {
+      "acc": 0.69017353,
+      "epoch": 1.3774733637747336,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.4230301325906606e-06,
+      "loss": 1.40288296,
+      "memory(GiB)": 123.79,
+      "step": 54300,
+      "train_speed(iter/s)": 1.315753
+    },
+    {
+      "acc": 0.67857103,
+      "epoch": 1.3776002029426686,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.4221315676907066e-06,
+      "loss": 1.44622784,
+      "memory(GiB)": 123.79,
+      "step": 54305,
+      "train_speed(iter/s)": 1.315767
+    },
+    {
+      "acc": 0.6801291,
+      "epoch": 1.3777270421106038,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.4212331161769194e-06,
+      "loss": 1.45228357,
+      "memory(GiB)": 123.79,
+      "step": 54310,
+      "train_speed(iter/s)": 1.315781
+    },
+    {
+      "acc": 0.67899756,
+      "epoch": 1.3778538812785388,
+      "grad_norm": 3.125,
+      "learning_rate": 2.420334778088811e-06,
+      "loss": 1.38245029,
+      "memory(GiB)": 123.79,
+      "step": 54315,
+      "train_speed(iter/s)": 1.315794
+    },
+    {
+      "acc": 0.68667803,
+      "epoch": 1.3779807204464738,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.4194365534658944e-06,
+      "loss": 1.40612288,
+      "memory(GiB)": 123.79,
+      "step": 54320,
+      "train_speed(iter/s)": 1.315807
+    },
+    {
+      "acc": 0.67749619,
+      "epoch": 1.378107559614409,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.4185384423476817e-06,
+      "loss": 1.47033072,
+      "memory(GiB)": 123.79,
+      "step": 54325,
+      "train_speed(iter/s)": 1.315821
+    },
+    {
+      "acc": 0.67776699,
+      "epoch": 1.378234398782344,
+      "grad_norm": 3.25,
+      "learning_rate": 2.4176404447736758e-06,
+      "loss": 1.44758291,
+      "memory(GiB)": 123.79,
+      "step": 54330,
+      "train_speed(iter/s)": 1.315835
+    },
+    {
+      "acc": 0.68075638,
+      "epoch": 1.378361237950279,
+      "grad_norm": 3.625,
+      "learning_rate": 2.41674256078337e-06,
+      "loss": 1.40389185,
+      "memory(GiB)": 123.79,
+      "step": 54335,
+      "train_speed(iter/s)": 1.315848
+    },
+    {
+      "acc": 0.66817045,
+      "epoch": 1.378488077118214,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.4158447904162585e-06,
+      "loss": 1.46915598,
+      "memory(GiB)": 123.79,
+      "step": 54340,
+      "train_speed(iter/s)": 1.315713
+    },
+    {
+      "acc": 0.68722873,
+      "epoch": 1.3786149162861492,
+      "grad_norm": 3.5,
+      "learning_rate": 2.414947133711832e-06,
+      "loss": 1.41209908,
+      "memory(GiB)": 123.79,
+      "step": 54345,
+      "train_speed(iter/s)": 1.315726
+    },
+    {
+      "acc": 0.67792263,
+      "epoch": 1.3787417554540842,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.414049590709574e-06,
+      "loss": 1.39639759,
+      "memory(GiB)": 123.79,
+      "step": 54350,
+      "train_speed(iter/s)": 1.315739
+    },
+    {
+      "acc": 0.67485151,
+      "epoch": 1.3788685946220194,
+      "grad_norm": 3.25,
+      "learning_rate": 2.4131521614489567e-06,
+      "loss": 1.4565073,
+      "memory(GiB)": 123.79,
+      "step": 54355,
+      "train_speed(iter/s)": 1.315752
+    },
+    {
+      "acc": 0.69579592,
+      "epoch": 1.3789954337899544,
+      "grad_norm": 2.875,
+      "learning_rate": 2.412254845969459e-06,
+      "loss": 1.31920099,
+      "memory(GiB)": 123.79,
+      "step": 54360,
+      "train_speed(iter/s)": 1.315765
+    },
+    {
+      "acc": 0.69966211,
+      "epoch": 1.3791222729578894,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.4113576443105464e-06,
+      "loss": 1.35281935,
+      "memory(GiB)": 123.79,
+      "step": 54365,
+      "train_speed(iter/s)": 1.315779
+    },
+    {
+      "acc": 0.67871952,
+      "epoch": 1.3792491121258244,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.410460556511684e-06,
+      "loss": 1.37878628,
+      "memory(GiB)": 123.79,
+      "step": 54370,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.68559642,
+      "epoch": 1.3793759512937596,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.4095635826123235e-06,
+      "loss": 1.36727133,
+      "memory(GiB)": 123.79,
+      "step": 54375,
+      "train_speed(iter/s)": 1.315806
+    },
+    {
+      "acc": 0.69469929,
+      "epoch": 1.3795027904616946,
+      "grad_norm": 2.875,
+      "learning_rate": 2.4086667226519245e-06,
+      "loss": 1.38245602,
+      "memory(GiB)": 123.79,
+      "step": 54380,
+      "train_speed(iter/s)": 1.31582
+    },
+    {
+      "acc": 0.68273993,
+      "epoch": 1.3796296296296298,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.4077699766699323e-06,
+      "loss": 1.43133316,
+      "memory(GiB)": 123.79,
+      "step": 54385,
+      "train_speed(iter/s)": 1.315834
+    },
+    {
+      "acc": 0.69737701,
+      "epoch": 1.3797564687975648,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.4068733447057903e-06,
+      "loss": 1.36391945,
+      "memory(GiB)": 123.79,
+      "step": 54390,
+      "train_speed(iter/s)": 1.315848
+    },
+    {
+      "acc": 0.66476059,
+      "epoch": 1.3798833079654997,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.405976826798936e-06,
+      "loss": 1.46229467,
+      "memory(GiB)": 123.79,
+      "step": 54395,
+      "train_speed(iter/s)": 1.315861
+    },
+    {
+      "acc": 0.7016048,
+      "epoch": 1.3800101471334347,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.405080422988802e-06,
+      "loss": 1.3858593,
+      "memory(GiB)": 123.79,
+      "step": 54400,
+      "train_speed(iter/s)": 1.315875
+    },
+    {
+      "acc": 0.68854709,
+      "epoch": 1.38013698630137,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.404184133314817e-06,
+      "loss": 1.45913448,
+      "memory(GiB)": 123.79,
+      "step": 54405,
+      "train_speed(iter/s)": 1.315889
+    },
+    {
+      "acc": 0.69063783,
+      "epoch": 1.380263825469305,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.4032879578164027e-06,
+      "loss": 1.39904242,
+      "memory(GiB)": 123.79,
+      "step": 54410,
+      "train_speed(iter/s)": 1.315902
+    },
+    {
+      "acc": 0.6883481,
+      "epoch": 1.38039066463724,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.402391896532978e-06,
+      "loss": 1.37436323,
+      "memory(GiB)": 123.79,
+      "step": 54415,
+      "train_speed(iter/s)": 1.315916
+    },
+    {
+      "acc": 0.67206883,
+      "epoch": 1.3805175038051751,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.4014959495039548e-06,
+      "loss": 1.48643456,
+      "memory(GiB)": 123.79,
+      "step": 54420,
+      "train_speed(iter/s)": 1.31593
+    },
+    {
+      "acc": 0.68953018,
+      "epoch": 1.3806443429731101,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.4006001167687416e-06,
+      "loss": 1.42466488,
+      "memory(GiB)": 123.79,
+      "step": 54425,
+      "train_speed(iter/s)": 1.315943
+    },
+    {
+      "acc": 0.69013481,
+      "epoch": 1.380771182141045,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.39970439836674e-06,
+      "loss": 1.38594189,
+      "memory(GiB)": 123.79,
+      "step": 54430,
+      "train_speed(iter/s)": 1.315957
+    },
+    {
+      "acc": 0.68203425,
+      "epoch": 1.38089802130898,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.3988087943373497e-06,
+      "loss": 1.41334257,
+      "memory(GiB)": 123.79,
+      "step": 54435,
+      "train_speed(iter/s)": 1.315971
+    },
+    {
+      "acc": 0.69978504,
+      "epoch": 1.3810248604769153,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.397913304719961e-06,
+      "loss": 1.31369781,
+      "memory(GiB)": 123.79,
+      "step": 54440,
+      "train_speed(iter/s)": 1.315985
+    },
+    {
+      "acc": 0.69186153,
+      "epoch": 1.3811516996448503,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.397017929553961e-06,
+      "loss": 1.34875746,
+      "memory(GiB)": 123.79,
+      "step": 54445,
+      "train_speed(iter/s)": 1.315999
+    },
+    {
+      "acc": 0.69214272,
+      "epoch": 1.3812785388127855,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.396122668878738e-06,
+      "loss": 1.34788771,
+      "memory(GiB)": 123.79,
+      "step": 54450,
+      "train_speed(iter/s)": 1.316013
+    },
+    {
+      "acc": 0.6674047,
+      "epoch": 1.3814053779807205,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.3952275227336636e-06,
+      "loss": 1.51948242,
+      "memory(GiB)": 123.79,
+      "step": 54455,
+      "train_speed(iter/s)": 1.316027
+    },
+    {
+      "acc": 0.68360305,
+      "epoch": 1.3815322171486555,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.3943324911581117e-06,
+      "loss": 1.36331444,
+      "memory(GiB)": 123.79,
+      "step": 54460,
+      "train_speed(iter/s)": 1.316041
+    },
+    {
+      "acc": 0.69211121,
+      "epoch": 1.3816590563165905,
+      "grad_norm": 2.75,
+      "learning_rate": 2.393437574191449e-06,
+      "loss": 1.35150299,
+      "memory(GiB)": 123.79,
+      "step": 54465,
+      "train_speed(iter/s)": 1.315886
+    },
+    {
+      "acc": 0.69119163,
+      "epoch": 1.3817858954845257,
+      "grad_norm": 2.875,
+      "learning_rate": 2.3925427718730426e-06,
+      "loss": 1.37416716,
+      "memory(GiB)": 123.79,
+      "step": 54470,
+      "train_speed(iter/s)": 1.315899
+    },
+    {
+      "acc": 0.70462036,
+      "epoch": 1.3819127346524607,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.391648084242245e-06,
+      "loss": 1.31132145,
+      "memory(GiB)": 123.79,
+      "step": 54475,
+      "train_speed(iter/s)": 1.315913
+    },
+    {
+      "acc": 0.68382001,
+      "epoch": 1.3820395738203957,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.3907535113384084e-06,
+      "loss": 1.44915409,
+      "memory(GiB)": 123.79,
+      "step": 54480,
+      "train_speed(iter/s)": 1.315927
+    },
+    {
+      "acc": 0.68248262,
+      "epoch": 1.3821664129883309,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.389859053200883e-06,
+      "loss": 1.40803661,
+      "memory(GiB)": 123.79,
+      "step": 54485,
+      "train_speed(iter/s)": 1.31594
+    },
+    {
+      "acc": 0.69132538,
+      "epoch": 1.3822932521562659,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.3889647098690127e-06,
+      "loss": 1.38787899,
+      "memory(GiB)": 123.79,
+      "step": 54490,
+      "train_speed(iter/s)": 1.315953
+    },
+    {
+      "acc": 0.69003406,
+      "epoch": 1.3824200913242009,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.3880704813821275e-06,
+      "loss": 1.39045296,
+      "memory(GiB)": 123.79,
+      "step": 54495,
+      "train_speed(iter/s)": 1.315967
+    },
+    {
+      "acc": 0.68174982,
+      "epoch": 1.3825469304921358,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.3871763677795656e-06,
+      "loss": 1.42156506,
+      "memory(GiB)": 123.79,
+      "step": 54500,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.69257927,
+      "epoch": 1.382673769660071,
+      "grad_norm": 3.375,
+      "learning_rate": 2.386282369100653e-06,
+      "loss": 1.36417809,
+      "memory(GiB)": 123.79,
+      "step": 54505,
+      "train_speed(iter/s)": 1.315995
+    },
+    {
+      "acc": 0.68542919,
+      "epoch": 1.382800608828006,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.385388485384713e-06,
+      "loss": 1.45136299,
+      "memory(GiB)": 123.79,
+      "step": 54510,
+      "train_speed(iter/s)": 1.316009
+    },
+    {
+      "acc": 0.69689775,
+      "epoch": 1.3829274479959413,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.384494716671057e-06,
+      "loss": 1.38458462,
+      "memory(GiB)": 123.79,
+      "step": 54515,
+      "train_speed(iter/s)": 1.316023
+    },
+    {
+      "acc": 0.70986061,
+      "epoch": 1.3830542871638762,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.3836010629990027e-06,
+      "loss": 1.28581543,
+      "memory(GiB)": 123.79,
+      "step": 54520,
+      "train_speed(iter/s)": 1.316037
+    },
+    {
+      "acc": 0.69169321,
+      "epoch": 1.3831811263318112,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.382707524407855e-06,
+      "loss": 1.41653748,
+      "memory(GiB)": 123.79,
+      "step": 54525,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.68773174,
+      "epoch": 1.3833079654997462,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.3818141009369155e-06,
+      "loss": 1.4079834,
+      "memory(GiB)": 123.79,
+      "step": 54530,
+      "train_speed(iter/s)": 1.316063
+    },
+    {
+      "acc": 0.68963485,
+      "epoch": 1.3834348046676814,
+      "grad_norm": 2.875,
+      "learning_rate": 2.3809207926254813e-06,
+      "loss": 1.41931314,
+      "memory(GiB)": 123.79,
+      "step": 54535,
+      "train_speed(iter/s)": 1.316077
+    },
+    {
+      "acc": 0.69652505,
+      "epoch": 1.3835616438356164,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.380027599512844e-06,
+      "loss": 1.38090162,
+      "memory(GiB)": 123.79,
+      "step": 54540,
+      "train_speed(iter/s)": 1.31609
+    },
+    {
+      "acc": 0.67945213,
+      "epoch": 1.3836884830035516,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.3791345216382906e-06,
+      "loss": 1.4006155,
+      "memory(GiB)": 123.79,
+      "step": 54545,
+      "train_speed(iter/s)": 1.316103
+    },
+    {
+      "acc": 0.679743,
+      "epoch": 1.3838153221714866,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.378241559041102e-06,
+      "loss": 1.4367979,
+      "memory(GiB)": 123.79,
+      "step": 54550,
+      "train_speed(iter/s)": 1.316116
+    },
+    {
+      "acc": 0.68734078,
+      "epoch": 1.3839421613394216,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.377348711760555e-06,
+      "loss": 1.44747992,
+      "memory(GiB)": 123.79,
+      "step": 54555,
+      "train_speed(iter/s)": 1.316129
+    },
+    {
+      "acc": 0.68048058,
+      "epoch": 1.3840690005073566,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.3764559798359204e-06,
+      "loss": 1.4227293,
+      "memory(GiB)": 123.79,
+      "step": 54560,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.69473996,
+      "epoch": 1.3841958396752918,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.3755633633064658e-06,
+      "loss": 1.31575108,
+      "memory(GiB)": 123.79,
+      "step": 54565,
+      "train_speed(iter/s)": 1.316155
+    },
+    {
+      "acc": 0.66006565,
+      "epoch": 1.3843226788432268,
+      "grad_norm": 2.625,
+      "learning_rate": 2.374670862211451e-06,
+      "loss": 1.50530739,
+      "memory(GiB)": 123.79,
+      "step": 54570,
+      "train_speed(iter/s)": 1.316168
+    },
+    {
+      "acc": 0.69427719,
+      "epoch": 1.3844495180111618,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.373778476590134e-06,
+      "loss": 1.35998068,
+      "memory(GiB)": 123.79,
+      "step": 54575,
+      "train_speed(iter/s)": 1.316181
+    },
+    {
+      "acc": 0.68909378,
+      "epoch": 1.384576357179097,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.372886206481764e-06,
+      "loss": 1.36487103,
+      "memory(GiB)": 123.79,
+      "step": 54580,
+      "train_speed(iter/s)": 1.316194
+    },
+    {
+      "acc": 0.68430443,
+      "epoch": 1.384703196347032,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.3719940519255864e-06,
+      "loss": 1.42526808,
+      "memory(GiB)": 123.79,
+      "step": 54585,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.6789052,
+      "epoch": 1.384830035514967,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.371102012960847e-06,
+      "loss": 1.44993038,
+      "memory(GiB)": 123.79,
+      "step": 54590,
+      "train_speed(iter/s)": 1.316222
+    },
+    {
+      "acc": 0.69261875,
+      "epoch": 1.384956874682902,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.3702100896267767e-06,
+      "loss": 1.35456772,
+      "memory(GiB)": 123.79,
+      "step": 54595,
+      "train_speed(iter/s)": 1.316236
+    },
+    {
+      "acc": 0.69543085,
+      "epoch": 1.3850837138508372,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.3693182819626077e-06,
+      "loss": 1.37765989,
+      "memory(GiB)": 123.79,
+      "step": 54600,
+      "train_speed(iter/s)": 1.316249
+    },
+    {
+      "acc": 0.68543601,
+      "epoch": 1.3852105530187722,
+      "grad_norm": 4.40625,
+      "learning_rate": 2.3684265900075637e-06,
+      "loss": 1.38914585,
+      "memory(GiB)": 123.79,
+      "step": 54605,
+      "train_speed(iter/s)": 1.316262
+    },
+    {
+      "acc": 0.68875065,
+      "epoch": 1.3853373921867074,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.3675350138008714e-06,
+      "loss": 1.40131607,
+      "memory(GiB)": 123.79,
+      "step": 54610,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.69554896,
+      "epoch": 1.3854642313546424,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.3666435533817406e-06,
+      "loss": 1.41583614,
+      "memory(GiB)": 123.79,
+      "step": 54615,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.68852234,
+      "epoch": 1.3855910705225774,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.3657522087893806e-06,
+      "loss": 1.40838299,
+      "memory(GiB)": 123.79,
+      "step": 54620,
+      "train_speed(iter/s)": 1.316299
+    },
+    {
+      "acc": 0.69598155,
+      "epoch": 1.3857179096905123,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.3648609800630022e-06,
+      "loss": 1.3804533,
+      "memory(GiB)": 123.79,
+      "step": 54625,
+      "train_speed(iter/s)": 1.316313
+    },
+    {
+      "acc": 0.68263359,
+      "epoch": 1.3858447488584476,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.363969867241805e-06,
+      "loss": 1.38920431,
+      "memory(GiB)": 123.79,
+      "step": 54630,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.67343969,
+      "epoch": 1.3859715880263825,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.36307887036498e-06,
+      "loss": 1.4058342,
+      "memory(GiB)": 123.79,
+      "step": 54635,
+      "train_speed(iter/s)": 1.31634
+    },
+    {
+      "acc": 0.68675175,
+      "epoch": 1.3860984271943175,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.3621879894717177e-06,
+      "loss": 1.36840229,
+      "memory(GiB)": 123.79,
+      "step": 54640,
+      "train_speed(iter/s)": 1.316352
+    },
+    {
+      "acc": 0.70845027,
+      "epoch": 1.3862252663622527,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.361297224601206e-06,
+      "loss": 1.34604073,
+      "memory(GiB)": 123.79,
+      "step": 54645,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.70674276,
+      "epoch": 1.3863521055301877,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.360406575792625e-06,
+      "loss": 1.32245817,
+      "memory(GiB)": 123.79,
+      "step": 54650,
+      "train_speed(iter/s)": 1.316378
+    },
+    {
+      "acc": 0.70045309,
+      "epoch": 1.3864789446981227,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.3595160430851445e-06,
+      "loss": 1.29130135,
+      "memory(GiB)": 123.79,
+      "step": 54655,
+      "train_speed(iter/s)": 1.316391
+    },
+    {
+      "acc": 0.67982512,
+      "epoch": 1.3866057838660577,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.3586256265179392e-06,
+      "loss": 1.3951272,
+      "memory(GiB)": 123.79,
+      "step": 54660,
+      "train_speed(iter/s)": 1.316405
+    },
+    {
+      "acc": 0.69848542,
+      "epoch": 1.386732623033993,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.3577353261301715e-06,
+      "loss": 1.30467186,
+      "memory(GiB)": 123.79,
+      "step": 54665,
+      "train_speed(iter/s)": 1.316416
+    },
+    {
+      "acc": 0.6832366,
+      "epoch": 1.386859462201928,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.356845141961001e-06,
+      "loss": 1.43852463,
+      "memory(GiB)": 123.79,
+      "step": 54670,
+      "train_speed(iter/s)": 1.316429
+    },
+    {
+      "acc": 0.68121204,
+      "epoch": 1.3869863013698631,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.355955074049582e-06,
+      "loss": 1.36367369,
+      "memory(GiB)": 123.79,
+      "step": 54675,
+      "train_speed(iter/s)": 1.316442
+    },
+    {
+      "acc": 0.69475298,
+      "epoch": 1.387113140537798,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.355065122435064e-06,
+      "loss": 1.37832546,
+      "memory(GiB)": 123.79,
+      "step": 54680,
+      "train_speed(iter/s)": 1.316455
+    },
+    {
+      "acc": 0.68011875,
+      "epoch": 1.387239979705733,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.3541752871565902e-06,
+      "loss": 1.41487408,
+      "memory(GiB)": 123.79,
+      "step": 54685,
+      "train_speed(iter/s)": 1.316469
+    },
+    {
+      "acc": 0.69272113,
+      "epoch": 1.387366818873668,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.3532855682533003e-06,
+      "loss": 1.38975773,
+      "memory(GiB)": 123.79,
+      "step": 54690,
+      "train_speed(iter/s)": 1.316482
+    },
+    {
+      "acc": 0.69410501,
+      "epoch": 1.3874936580416033,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.352395965764328e-06,
+      "loss": 1.38086395,
+      "memory(GiB)": 123.79,
+      "step": 54695,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.67730398,
+      "epoch": 1.3876204972095383,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.3515064797288013e-06,
+      "loss": 1.38679457,
+      "memory(GiB)": 123.79,
+      "step": 54700,
+      "train_speed(iter/s)": 1.316508
+    },
+    {
+      "acc": 0.68206396,
+      "epoch": 1.3877473363774735,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.350617110185845e-06,
+      "loss": 1.44019499,
+      "memory(GiB)": 123.79,
+      "step": 54705,
+      "train_speed(iter/s)": 1.316521
+    },
+    {
+      "acc": 0.69135318,
+      "epoch": 1.3878741755454085,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.3497278571745763e-06,
+      "loss": 1.3691987,
+      "memory(GiB)": 123.79,
+      "step": 54710,
+      "train_speed(iter/s)": 1.316535
+    },
+    {
+      "acc": 0.69799509,
+      "epoch": 1.3880010147133435,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.348838720734109e-06,
+      "loss": 1.32966862,
+      "memory(GiB)": 123.79,
+      "step": 54715,
+      "train_speed(iter/s)": 1.316548
+    },
+    {
+      "acc": 0.6954607,
+      "epoch": 1.3881278538812785,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.347949700903552e-06,
+      "loss": 1.41505756,
+      "memory(GiB)": 123.79,
+      "step": 54720,
+      "train_speed(iter/s)": 1.316563
+    },
+    {
+      "acc": 0.67905064,
+      "epoch": 1.3882546930492137,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.3470607977220066e-06,
+      "loss": 1.35540257,
+      "memory(GiB)": 123.79,
+      "step": 54725,
+      "train_speed(iter/s)": 1.316576
+    },
+    {
+      "acc": 0.69237795,
+      "epoch": 1.3883815322171487,
+      "grad_norm": 2.875,
+      "learning_rate": 2.346172011228573e-06,
+      "loss": 1.40292187,
+      "memory(GiB)": 123.79,
+      "step": 54730,
+      "train_speed(iter/s)": 1.316589
+    },
+    {
+      "acc": 0.70056934,
+      "epoch": 1.3885083713850837,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.345283341462342e-06,
+      "loss": 1.34099808,
+      "memory(GiB)": 123.79,
+      "step": 54735,
+      "train_speed(iter/s)": 1.316603
+    },
+    {
+      "acc": 0.67614174,
+      "epoch": 1.3886352105530189,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.3443947884624026e-06,
+      "loss": 1.43590145,
+      "memory(GiB)": 123.79,
+      "step": 54740,
+      "train_speed(iter/s)": 1.316616
+    },
+    {
+      "acc": 0.69416676,
+      "epoch": 1.3887620497209539,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.3435063522678346e-06,
+      "loss": 1.33584146,
+      "memory(GiB)": 123.79,
+      "step": 54745,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.680865,
+      "epoch": 1.3888888888888888,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.3426180329177217e-06,
+      "loss": 1.397715,
+      "memory(GiB)": 123.79,
+      "step": 54750,
+      "train_speed(iter/s)": 1.316643
+    },
+    {
+      "acc": 0.68621454,
+      "epoch": 1.3890157280568238,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.3417298304511297e-06,
+      "loss": 1.39168816,
+      "memory(GiB)": 123.79,
+      "step": 54755,
+      "train_speed(iter/s)": 1.316657
+    },
+    {
+      "acc": 0.66551337,
+      "epoch": 1.389142567224759,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.340841744907127e-06,
+      "loss": 1.49233932,
+      "memory(GiB)": 123.79,
+      "step": 54760,
+      "train_speed(iter/s)": 1.31667
+    },
+    {
+      "acc": 0.6905993,
+      "epoch": 1.389269406392694,
+      "grad_norm": 3.625,
+      "learning_rate": 2.3399537763247783e-06,
+      "loss": 1.35386887,
+      "memory(GiB)": 123.79,
+      "step": 54765,
+      "train_speed(iter/s)": 1.316684
+    },
+    {
+      "acc": 0.68259096,
+      "epoch": 1.3893962455606292,
+      "grad_norm": 3.0,
+      "learning_rate": 2.3390659247431404e-06,
+      "loss": 1.3930769,
+      "memory(GiB)": 123.79,
+      "step": 54770,
+      "train_speed(iter/s)": 1.316697
+    },
+    {
+      "acc": 0.69542751,
+      "epoch": 1.3895230847285642,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.338178190201261e-06,
+      "loss": 1.34512348,
+      "memory(GiB)": 123.79,
+      "step": 54775,
+      "train_speed(iter/s)": 1.31671
+    },
+    {
+      "acc": 0.68322482,
+      "epoch": 1.3896499238964992,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.3372905727381877e-06,
+      "loss": 1.38282051,
+      "memory(GiB)": 123.79,
+      "step": 54780,
+      "train_speed(iter/s)": 1.316724
+    },
+    {
+      "acc": 0.70003967,
+      "epoch": 1.3897767630644342,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.3364030723929647e-06,
+      "loss": 1.33545237,
+      "memory(GiB)": 123.79,
+      "step": 54785,
+      "train_speed(iter/s)": 1.316736
+    },
+    {
+      "acc": 0.70695777,
+      "epoch": 1.3899036022323694,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.335515689204629e-06,
+      "loss": 1.33437557,
+      "memory(GiB)": 123.79,
+      "step": 54790,
+      "train_speed(iter/s)": 1.316749
+    },
+    {
+      "acc": 0.69581699,
+      "epoch": 1.3900304414003044,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.334628423212206e-06,
+      "loss": 1.36868076,
+      "memory(GiB)": 123.79,
+      "step": 54795,
+      "train_speed(iter/s)": 1.316761
+    },
+    {
+      "acc": 0.68954744,
+      "epoch": 1.3901572805682394,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.3337412744547256e-06,
+      "loss": 1.37608337,
+      "memory(GiB)": 123.79,
+      "step": 54800,
+      "train_speed(iter/s)": 1.316775
+    },
+    {
+      "acc": 0.7021162,
+      "epoch": 1.3902841197361746,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.332854242971209e-06,
+      "loss": 1.35530052,
+      "memory(GiB)": 123.79,
+      "step": 54805,
+      "train_speed(iter/s)": 1.316788
+    },
+    {
+      "acc": 0.68430262,
+      "epoch": 1.3904109589041096,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.331967328800672e-06,
+      "loss": 1.3721467,
+      "memory(GiB)": 123.79,
+      "step": 54810,
+      "train_speed(iter/s)": 1.3168
+    },
+    {
+      "acc": 0.69017844,
+      "epoch": 1.3905377980720446,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.33108053198212e-06,
+      "loss": 1.31197987,
+      "memory(GiB)": 123.79,
+      "step": 54815,
+      "train_speed(iter/s)": 1.316813
+    },
+    {
+      "acc": 0.68667927,
+      "epoch": 1.3906646372399796,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.330193852554564e-06,
+      "loss": 1.36660023,
+      "memory(GiB)": 123.79,
+      "step": 54820,
+      "train_speed(iter/s)": 1.316825
+    },
+    {
+      "acc": 0.69133005,
+      "epoch": 1.3907914764079148,
+      "grad_norm": 3.828125,
+      "learning_rate": 2.3293072905570024e-06,
+      "loss": 1.38841352,
+      "memory(GiB)": 123.79,
+      "step": 54825,
+      "train_speed(iter/s)": 1.316838
+    },
+    {
+      "acc": 0.69646387,
+      "epoch": 1.3909183155758498,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.3284208460284303e-06,
+      "loss": 1.38324795,
+      "memory(GiB)": 123.79,
+      "step": 54830,
+      "train_speed(iter/s)": 1.316849
+    },
+    {
+      "acc": 0.69509392,
+      "epoch": 1.391045154743785,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.3275345190078364e-06,
+      "loss": 1.39507608,
+      "memory(GiB)": 123.79,
+      "step": 54835,
+      "train_speed(iter/s)": 1.316862
+    },
+    {
+      "acc": 0.66744623,
+      "epoch": 1.39117199391172,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.3266483095342064e-06,
+      "loss": 1.42742395,
+      "memory(GiB)": 123.79,
+      "step": 54840,
+      "train_speed(iter/s)": 1.316875
+    },
+    {
+      "acc": 0.69170446,
+      "epoch": 1.391298833079655,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.3257622176465194e-06,
+      "loss": 1.34066019,
+      "memory(GiB)": 123.79,
+      "step": 54845,
+      "train_speed(iter/s)": 1.316887
+    },
+    {
+      "acc": 0.68661022,
+      "epoch": 1.39142567224759,
+      "grad_norm": 3.5,
+      "learning_rate": 2.3248762433837494e-06,
+      "loss": 1.43311119,
+      "memory(GiB)": 123.79,
+      "step": 54850,
+      "train_speed(iter/s)": 1.316901
+    },
+    {
+      "acc": 0.68581104,
+      "epoch": 1.3915525114155252,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.323990386784867e-06,
+      "loss": 1.37072535,
+      "memory(GiB)": 123.79,
+      "step": 54855,
+      "train_speed(iter/s)": 1.316914
+    },
+    {
+      "acc": 0.69311562,
+      "epoch": 1.3916793505834602,
+      "grad_norm": 5.03125,
+      "learning_rate": 2.3231046478888335e-06,
+      "loss": 1.38127613,
+      "memory(GiB)": 123.79,
+      "step": 54860,
+      "train_speed(iter/s)": 1.316927
+    },
+    {
+      "acc": 0.68480644,
+      "epoch": 1.3918061897513954,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.3222190267346094e-06,
+      "loss": 1.38210011,
+      "memory(GiB)": 123.79,
+      "step": 54865,
+      "train_speed(iter/s)": 1.31694
+    },
+    {
+      "acc": 0.69333239,
+      "epoch": 1.3919330289193304,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.3213335233611484e-06,
+      "loss": 1.37096539,
+      "memory(GiB)": 123.79,
+      "step": 54870,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.68425274,
+      "epoch": 1.3920598680872653,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.320448137807398e-06,
+      "loss": 1.37470665,
+      "memory(GiB)": 123.79,
+      "step": 54875,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.68307638,
+      "epoch": 1.3921867072552003,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.3195628701123017e-06,
+      "loss": 1.43672295,
+      "memory(GiB)": 123.79,
+      "step": 54880,
+      "train_speed(iter/s)": 1.316978
+    },
+    {
+      "acc": 0.68963728,
+      "epoch": 1.3923135464231355,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.3186777203147964e-06,
+      "loss": 1.38413801,
+      "memory(GiB)": 123.79,
+      "step": 54885,
+      "train_speed(iter/s)": 1.316988
+    },
+    {
+      "acc": 0.69380093,
+      "epoch": 1.3924403855910705,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.3177926884538193e-06,
+      "loss": 1.44251614,
+      "memory(GiB)": 123.79,
+      "step": 54890,
+      "train_speed(iter/s)": 1.317001
+    },
+    {
+      "acc": 0.68418827,
+      "epoch": 1.3925672247590055,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.3169077745682933e-06,
+      "loss": 1.38747396,
+      "memory(GiB)": 123.79,
+      "step": 54895,
+      "train_speed(iter/s)": 1.317014
+    },
+    {
+      "acc": 0.68532457,
+      "epoch": 1.3926940639269407,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.316022978697143e-06,
+      "loss": 1.37393713,
+      "memory(GiB)": 123.79,
+      "step": 54900,
+      "train_speed(iter/s)": 1.317026
+    },
+    {
+      "acc": 0.68320136,
+      "epoch": 1.3928209030948757,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.3151383008792826e-06,
+      "loss": 1.38707275,
+      "memory(GiB)": 123.79,
+      "step": 54905,
+      "train_speed(iter/s)": 1.317039
+    },
+    {
+      "acc": 0.69189215,
+      "epoch": 1.3929477422628107,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.314253741153631e-06,
+      "loss": 1.36993647,
+      "memory(GiB)": 123.79,
+      "step": 54910,
+      "train_speed(iter/s)": 1.317051
+    },
+    {
+      "acc": 0.68246574,
+      "epoch": 1.3930745814307457,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.313369299559088e-06,
+      "loss": 1.34385967,
+      "memory(GiB)": 123.79,
+      "step": 54915,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.69657946,
+      "epoch": 1.393201420598681,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.3124849761345576e-06,
+      "loss": 1.28749905,
+      "memory(GiB)": 123.79,
+      "step": 54920,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.69343734,
+      "epoch": 1.393328259766616,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.311600770918938e-06,
+      "loss": 1.38159065,
+      "memory(GiB)": 123.79,
+      "step": 54925,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.70128269,
+      "epoch": 1.393455098934551,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.310716683951122e-06,
+      "loss": 1.37426348,
+      "memory(GiB)": 123.79,
+      "step": 54930,
+      "train_speed(iter/s)": 1.317103
+    },
+    {
+      "acc": 0.68947139,
+      "epoch": 1.393581938102486,
+      "grad_norm": 4.125,
+      "learning_rate": 2.3098327152699884e-06,
+      "loss": 1.44080248,
+      "memory(GiB)": 123.79,
+      "step": 54935,
+      "train_speed(iter/s)": 1.317117
+    },
+    {
+      "acc": 0.69652176,
+      "epoch": 1.393708777270421,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.308948864914425e-06,
+      "loss": 1.38417053,
+      "memory(GiB)": 123.79,
+      "step": 54940,
+      "train_speed(iter/s)": 1.317129
+    },
+    {
+      "acc": 0.68279324,
+      "epoch": 1.393835616438356,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.308065132923305e-06,
+      "loss": 1.41336365,
+      "memory(GiB)": 123.79,
+      "step": 54945,
+      "train_speed(iter/s)": 1.317142
+    },
+    {
+      "acc": 0.69626665,
+      "epoch": 1.3939624556062913,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.3071815193355005e-06,
+      "loss": 1.34639206,
+      "memory(GiB)": 123.79,
+      "step": 54950,
+      "train_speed(iter/s)": 1.317155
+    },
+    {
+      "acc": 0.69004927,
+      "epoch": 1.3940892947742263,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.3062980241898725e-06,
+      "loss": 1.37048206,
+      "memory(GiB)": 123.79,
+      "step": 54955,
+      "train_speed(iter/s)": 1.317168
+    },
+    {
+      "acc": 0.68081121,
+      "epoch": 1.3942161339421613,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.3054146475252852e-06,
+      "loss": 1.42191706,
+      "memory(GiB)": 123.79,
+      "step": 54960,
+      "train_speed(iter/s)": 1.317181
+    },
+    {
+      "acc": 0.7064868,
+      "epoch": 1.3943429731100965,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.3045313893805926e-06,
+      "loss": 1.25896511,
+      "memory(GiB)": 123.79,
+      "step": 54965,
+      "train_speed(iter/s)": 1.317194
+    },
+    {
+      "acc": 0.69210749,
+      "epoch": 1.3944698122780315,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.303648249794644e-06,
+      "loss": 1.42695942,
+      "memory(GiB)": 123.79,
+      "step": 54970,
+      "train_speed(iter/s)": 1.317207
+    },
+    {
+      "acc": 0.68750243,
+      "epoch": 1.3945966514459665,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.302765228806283e-06,
+      "loss": 1.35646324,
+      "memory(GiB)": 123.79,
+      "step": 54975,
+      "train_speed(iter/s)": 1.31722
+    },
+    {
+      "acc": 0.6784133,
+      "epoch": 1.3947234906139014,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.30188232645435e-06,
+      "loss": 1.44901857,
+      "memory(GiB)": 123.79,
+      "step": 54980,
+      "train_speed(iter/s)": 1.317232
+    },
+    {
+      "acc": 0.67573109,
+      "epoch": 1.3948503297818367,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.300999542777678e-06,
+      "loss": 1.46661472,
+      "memory(GiB)": 123.79,
+      "step": 54985,
+      "train_speed(iter/s)": 1.317245
+    },
+    {
+      "acc": 0.68846455,
+      "epoch": 1.3949771689497716,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.300116877815097e-06,
+      "loss": 1.34070463,
+      "memory(GiB)": 123.79,
+      "step": 54990,
+      "train_speed(iter/s)": 1.317258
+    },
+    {
+      "acc": 0.67877188,
+      "epoch": 1.3951040081177069,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.2992343316054296e-06,
+      "loss": 1.40916758,
+      "memory(GiB)": 123.79,
+      "step": 54995,
+      "train_speed(iter/s)": 1.317271
+    },
+    {
+      "acc": 0.6903996,
+      "epoch": 1.3952308472856418,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.298351904187494e-06,
+      "loss": 1.37601528,
+      "memory(GiB)": 123.79,
+      "step": 55000,
+      "train_speed(iter/s)": 1.317283
+    },
+    {
+      "epoch": 1.3952308472856418,
+      "eval_acc": 0.6750716811073034,
+      "eval_loss": 1.3562896251678467,
+      "eval_runtime": 69.823,
+      "eval_samples_per_second": 91.231,
+      "eval_steps_per_second": 22.815,
+      "step": 55000
+    },
+    {
+      "acc": 0.68706694,
+      "epoch": 1.3953576864535768,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.297469595600104e-06,
+      "loss": 1.38833551,
+      "memory(GiB)": 123.79,
+      "step": 55005,
+      "train_speed(iter/s)": 1.314705
+    },
+    {
+      "acc": 0.68058071,
+      "epoch": 1.3954845256215118,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.2965874058820668e-06,
+      "loss": 1.42935057,
+      "memory(GiB)": 123.79,
+      "step": 55010,
+      "train_speed(iter/s)": 1.314718
+    },
+    {
+      "acc": 0.68703451,
+      "epoch": 1.395611364789447,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.2957053350721857e-06,
+      "loss": 1.38195019,
+      "memory(GiB)": 123.79,
+      "step": 55015,
+      "train_speed(iter/s)": 1.314731
+    },
+    {
+      "acc": 0.69250436,
+      "epoch": 1.395738203957382,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.294823383209258e-06,
+      "loss": 1.39863672,
+      "memory(GiB)": 123.79,
+      "step": 55020,
+      "train_speed(iter/s)": 1.314745
+    },
+    {
+      "acc": 0.69131174,
+      "epoch": 1.3958650431253172,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.2939415503320733e-06,
+      "loss": 1.37214661,
+      "memory(GiB)": 123.79,
+      "step": 55025,
+      "train_speed(iter/s)": 1.314759
+    },
+    {
+      "acc": 0.6681818,
+      "epoch": 1.3959918822932522,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.293059836479425e-06,
+      "loss": 1.44235649,
+      "memory(GiB)": 123.79,
+      "step": 55030,
+      "train_speed(iter/s)": 1.314773
+    },
+    {
+      "acc": 0.6937315,
+      "epoch": 1.3961187214611872,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.2921782416900883e-06,
+      "loss": 1.35098143,
+      "memory(GiB)": 123.79,
+      "step": 55035,
+      "train_speed(iter/s)": 1.314786
+    },
+    {
+      "acc": 0.69466305,
+      "epoch": 1.3962455606291222,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.2912967660028425e-06,
+      "loss": 1.3613163,
+      "memory(GiB)": 123.79,
+      "step": 55040,
+      "train_speed(iter/s)": 1.314799
+    },
+    {
+      "acc": 0.70118065,
+      "epoch": 1.3963723997970574,
+      "grad_norm": 2.375,
+      "learning_rate": 2.2904154094564568e-06,
+      "loss": 1.33877888,
+      "memory(GiB)": 123.79,
+      "step": 55045,
+      "train_speed(iter/s)": 1.314813
+    },
+    {
+      "acc": 0.67949591,
+      "epoch": 1.3964992389649924,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.2895341720897018e-06,
+      "loss": 1.38823299,
+      "memory(GiB)": 123.79,
+      "step": 55050,
+      "train_speed(iter/s)": 1.314827
+    },
+    {
+      "acc": 0.68089375,
+      "epoch": 1.3966260781329274,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.2886530539413336e-06,
+      "loss": 1.45139284,
+      "memory(GiB)": 123.79,
+      "step": 55055,
+      "train_speed(iter/s)": 1.314841
+    },
+    {
+      "acc": 0.69086313,
+      "epoch": 1.3967529173008626,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.2877720550501082e-06,
+      "loss": 1.38493404,
+      "memory(GiB)": 123.79,
+      "step": 55060,
+      "train_speed(iter/s)": 1.314854
+    },
+    {
+      "acc": 0.70966749,
+      "epoch": 1.3968797564687976,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.2868911754547783e-06,
+      "loss": 1.33687172,
+      "memory(GiB)": 123.79,
+      "step": 55065,
+      "train_speed(iter/s)": 1.314867
+    },
+    {
+      "acc": 0.68959141,
+      "epoch": 1.3970065956367326,
+      "grad_norm": 3.0,
+      "learning_rate": 2.28601041519409e-06,
+      "loss": 1.34483547,
+      "memory(GiB)": 123.79,
+      "step": 55070,
+      "train_speed(iter/s)": 1.31488
+    },
+    {
+      "acc": 0.68050013,
+      "epoch": 1.3971334348046676,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.2851297743067786e-06,
+      "loss": 1.39953136,
+      "memory(GiB)": 123.79,
+      "step": 55075,
+      "train_speed(iter/s)": 1.314894
+    },
+    {
+      "acc": 0.69219656,
+      "epoch": 1.3972602739726028,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.2842492528315784e-06,
+      "loss": 1.32479153,
+      "memory(GiB)": 123.79,
+      "step": 55080,
+      "train_speed(iter/s)": 1.314908
+    },
+    {
+      "acc": 0.6866024,
+      "epoch": 1.3973871131405378,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.283368850807223e-06,
+      "loss": 1.36863308,
+      "memory(GiB)": 123.79,
+      "step": 55085,
+      "train_speed(iter/s)": 1.314921
+    },
+    {
+      "acc": 0.697965,
+      "epoch": 1.397513952308473,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.282488568272437e-06,
+      "loss": 1.37563286,
+      "memory(GiB)": 123.79,
+      "step": 55090,
+      "train_speed(iter/s)": 1.314935
+    },
+    {
+      "acc": 0.68313046,
+      "epoch": 1.397640791476408,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.2816084052659326e-06,
+      "loss": 1.41056595,
+      "memory(GiB)": 123.79,
+      "step": 55095,
+      "train_speed(iter/s)": 1.31494
+    },
+    {
+      "acc": 0.69194512,
+      "epoch": 1.397767630644343,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.2807283618264288e-06,
+      "loss": 1.42459164,
+      "memory(GiB)": 123.79,
+      "step": 55100,
+      "train_speed(iter/s)": 1.314952
+    },
+    {
+      "acc": 0.69258413,
+      "epoch": 1.397894469812278,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.2798484379926324e-06,
+      "loss": 1.34171591,
+      "memory(GiB)": 123.79,
+      "step": 55105,
+      "train_speed(iter/s)": 1.314965
+    },
+    {
+      "acc": 0.69532261,
+      "epoch": 1.3980213089802132,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.278968633803246e-06,
+      "loss": 1.31534309,
+      "memory(GiB)": 123.79,
+      "step": 55110,
+      "train_speed(iter/s)": 1.314979
+    },
+    {
+      "acc": 0.68706303,
+      "epoch": 1.3981481481481481,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.2780889492969684e-06,
+      "loss": 1.3635067,
+      "memory(GiB)": 123.79,
+      "step": 55115,
+      "train_speed(iter/s)": 1.314992
+    },
+    {
+      "acc": 0.69860072,
+      "epoch": 1.3982749873160831,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.277209384512491e-06,
+      "loss": 1.33157101,
+      "memory(GiB)": 123.79,
+      "step": 55120,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.67715173,
+      "epoch": 1.3984018264840183,
+      "grad_norm": 4.6875,
+      "learning_rate": 2.2763299394885013e-06,
+      "loss": 1.39676456,
+      "memory(GiB)": 123.79,
+      "step": 55125,
+      "train_speed(iter/s)": 1.31502
+    },
+    {
+      "acc": 0.68688803,
+      "epoch": 1.3985286656519533,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.2754506142636808e-06,
+      "loss": 1.41091881,
+      "memory(GiB)": 123.79,
+      "step": 55130,
+      "train_speed(iter/s)": 1.315033
+    },
+    {
+      "acc": 0.70049129,
+      "epoch": 1.3986555048198883,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.274571408876707e-06,
+      "loss": 1.34924402,
+      "memory(GiB)": 123.79,
+      "step": 55135,
+      "train_speed(iter/s)": 1.315047
+    },
+    {
+      "acc": 0.70090866,
+      "epoch": 1.3987823439878233,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.2736923233662504e-06,
+      "loss": 1.38380938,
+      "memory(GiB)": 123.79,
+      "step": 55140,
+      "train_speed(iter/s)": 1.315061
+    },
+    {
+      "acc": 0.68861523,
+      "epoch": 1.3989091831557585,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.2728133577709776e-06,
+      "loss": 1.43820057,
+      "memory(GiB)": 123.79,
+      "step": 55145,
+      "train_speed(iter/s)": 1.315074
+    },
+    {
+      "acc": 0.69283762,
+      "epoch": 1.3990360223236935,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.2719345121295495e-06,
+      "loss": 1.38672686,
+      "memory(GiB)": 123.79,
+      "step": 55150,
+      "train_speed(iter/s)": 1.315087
+    },
+    {
+      "acc": 0.67185879,
+      "epoch": 1.3991628614916287,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.2710557864806214e-06,
+      "loss": 1.47298393,
+      "memory(GiB)": 123.79,
+      "step": 55155,
+      "train_speed(iter/s)": 1.315101
+    },
+    {
+      "acc": 0.68662815,
+      "epoch": 1.3992897006595637,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.2701771808628438e-06,
+      "loss": 1.38605518,
+      "memory(GiB)": 123.79,
+      "step": 55160,
+      "train_speed(iter/s)": 1.315115
+    },
+    {
+      "acc": 0.68341622,
+      "epoch": 1.3994165398274987,
+      "grad_norm": 3.125,
+      "learning_rate": 2.269298695314861e-06,
+      "loss": 1.42699642,
+      "memory(GiB)": 123.79,
+      "step": 55165,
+      "train_speed(iter/s)": 1.315129
+    },
+    {
+      "acc": 0.69113302,
+      "epoch": 1.3995433789954337,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.268420329875314e-06,
+      "loss": 1.38120594,
+      "memory(GiB)": 123.79,
+      "step": 55170,
+      "train_speed(iter/s)": 1.315142
+    },
+    {
+      "acc": 0.69226665,
+      "epoch": 1.399670218163369,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.2675420845828363e-06,
+      "loss": 1.33262415,
+      "memory(GiB)": 123.79,
+      "step": 55175,
+      "train_speed(iter/s)": 1.315156
+    },
+    {
+      "acc": 0.68437424,
+      "epoch": 1.3997970573313039,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.266663959476057e-06,
+      "loss": 1.35464334,
+      "memory(GiB)": 123.79,
+      "step": 55180,
+      "train_speed(iter/s)": 1.31517
+    },
+    {
+      "acc": 0.6920249,
+      "epoch": 1.399923896499239,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.265785954593598e-06,
+      "loss": 1.36020803,
+      "memory(GiB)": 123.79,
+      "step": 55185,
+      "train_speed(iter/s)": 1.315184
+    },
+    {
+      "acc": 0.68036385,
+      "epoch": 1.400050735667174,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.264908069974085e-06,
+      "loss": 1.37538013,
+      "memory(GiB)": 123.79,
+      "step": 55190,
+      "train_speed(iter/s)": 1.315198
+    },
+    {
+      "acc": 0.69422083,
+      "epoch": 1.400177574835109,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.2640303056561236e-06,
+      "loss": 1.33755856,
+      "memory(GiB)": 123.79,
+      "step": 55195,
+      "train_speed(iter/s)": 1.315211
+    },
+    {
+      "acc": 0.68369808,
+      "epoch": 1.400304414003044,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.2631526616783234e-06,
+      "loss": 1.3736084,
+      "memory(GiB)": 123.79,
+      "step": 55200,
+      "train_speed(iter/s)": 1.315224
+    },
+    {
+      "acc": 0.69404039,
+      "epoch": 1.4004312531709793,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.2622751380792896e-06,
+      "loss": 1.36947365,
+      "memory(GiB)": 123.79,
+      "step": 55205,
+      "train_speed(iter/s)": 1.315237
+    },
+    {
+      "acc": 0.67511692,
+      "epoch": 1.4005580923389143,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.26139773489762e-06,
+      "loss": 1.41539326,
+      "memory(GiB)": 123.79,
+      "step": 55210,
+      "train_speed(iter/s)": 1.31525
+    },
+    {
+      "acc": 0.68647957,
+      "epoch": 1.4006849315068493,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.260520452171904e-06,
+      "loss": 1.36703176,
+      "memory(GiB)": 123.79,
+      "step": 55215,
+      "train_speed(iter/s)": 1.315263
+    },
+    {
+      "acc": 0.68879762,
+      "epoch": 1.4008117706747845,
+      "grad_norm": 2.875,
+      "learning_rate": 2.259643289940727e-06,
+      "loss": 1.34317369,
+      "memory(GiB)": 123.79,
+      "step": 55220,
+      "train_speed(iter/s)": 1.315276
+    },
+    {
+      "acc": 0.68837652,
+      "epoch": 1.4009386098427195,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.2587662482426748e-06,
+      "loss": 1.39144192,
+      "memory(GiB)": 123.79,
+      "step": 55225,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.68515463,
+      "epoch": 1.4010654490106544,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.2578893271163234e-06,
+      "loss": 1.44587154,
+      "memory(GiB)": 123.79,
+      "step": 55230,
+      "train_speed(iter/s)": 1.315304
+    },
+    {
+      "acc": 0.68948946,
+      "epoch": 1.4011922881785894,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.2570125266002385e-06,
+      "loss": 1.37312737,
+      "memory(GiB)": 123.79,
+      "step": 55235,
+      "train_speed(iter/s)": 1.315317
+    },
+    {
+      "acc": 0.69162941,
+      "epoch": 1.4013191273465246,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.2561358467329907e-06,
+      "loss": 1.39574022,
+      "memory(GiB)": 123.79,
+      "step": 55240,
+      "train_speed(iter/s)": 1.315331
+    },
+    {
+      "acc": 0.68811121,
+      "epoch": 1.4014459665144596,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.2552592875531397e-06,
+      "loss": 1.39665174,
+      "memory(GiB)": 123.79,
+      "step": 55245,
+      "train_speed(iter/s)": 1.315344
+    },
+    {
+      "acc": 0.68928595,
+      "epoch": 1.4015728056823948,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.254382849099241e-06,
+      "loss": 1.36791067,
+      "memory(GiB)": 123.79,
+      "step": 55250,
+      "train_speed(iter/s)": 1.315358
+    },
+    {
+      "acc": 0.69327297,
+      "epoch": 1.4016996448503298,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.253506531409839e-06,
+      "loss": 1.33099689,
+      "memory(GiB)": 123.79,
+      "step": 55255,
+      "train_speed(iter/s)": 1.315372
+    },
+    {
+      "acc": 0.68083305,
+      "epoch": 1.4018264840182648,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.252630334523484e-06,
+      "loss": 1.43080072,
+      "memory(GiB)": 123.79,
+      "step": 55260,
+      "train_speed(iter/s)": 1.315386
+    },
+    {
+      "acc": 0.67528191,
+      "epoch": 1.4019533231861998,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.2517542584787134e-06,
+      "loss": 1.44187908,
+      "memory(GiB)": 123.79,
+      "step": 55265,
+      "train_speed(iter/s)": 1.315397
+    },
+    {
+      "acc": 0.68743176,
+      "epoch": 1.402080162354135,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.2508783033140596e-06,
+      "loss": 1.37462959,
+      "memory(GiB)": 123.79,
+      "step": 55270,
+      "train_speed(iter/s)": 1.31541
+    },
+    {
+      "acc": 0.67067504,
+      "epoch": 1.40220700152207,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.2500024690680528e-06,
+      "loss": 1.47615995,
+      "memory(GiB)": 123.79,
+      "step": 55275,
+      "train_speed(iter/s)": 1.315424
+    },
+    {
+      "acc": 0.70082536,
+      "epoch": 1.402333840690005,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.249126755779215e-06,
+      "loss": 1.31324835,
+      "memory(GiB)": 123.79,
+      "step": 55280,
+      "train_speed(iter/s)": 1.315437
+    },
+    {
+      "acc": 0.68216825,
+      "epoch": 1.4024606798579402,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.2482511634860645e-06,
+      "loss": 1.38415861,
+      "memory(GiB)": 123.79,
+      "step": 55285,
+      "train_speed(iter/s)": 1.315451
+    },
+    {
+      "acc": 0.68180647,
+      "epoch": 1.4025875190258752,
+      "grad_norm": 3.875,
+      "learning_rate": 2.247375692227113e-06,
+      "loss": 1.39667978,
+      "memory(GiB)": 123.79,
+      "step": 55290,
+      "train_speed(iter/s)": 1.315465
+    },
+    {
+      "acc": 0.69224739,
+      "epoch": 1.4027143581938102,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.2465003420408683e-06,
+      "loss": 1.41183891,
+      "memory(GiB)": 123.79,
+      "step": 55295,
+      "train_speed(iter/s)": 1.315479
+    },
+    {
+      "acc": 0.69432287,
+      "epoch": 1.4028411973617452,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.2456251129658325e-06,
+      "loss": 1.39540272,
+      "memory(GiB)": 123.79,
+      "step": 55300,
+      "train_speed(iter/s)": 1.315493
+    },
+    {
+      "acc": 0.67321892,
+      "epoch": 1.4029680365296804,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.2447500050405008e-06,
+      "loss": 1.43270817,
+      "memory(GiB)": 123.79,
+      "step": 55305,
+      "train_speed(iter/s)": 1.315505
+    },
+    {
+      "acc": 0.69507141,
+      "epoch": 1.4030948756976154,
+      "grad_norm": 5.25,
+      "learning_rate": 2.2438750183033657e-06,
+      "loss": 1.3779974,
+      "memory(GiB)": 123.79,
+      "step": 55310,
+      "train_speed(iter/s)": 1.315519
+    },
+    {
+      "acc": 0.69507704,
+      "epoch": 1.4032217148655506,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.2430001527929123e-06,
+      "loss": 1.38477659,
+      "memory(GiB)": 123.79,
+      "step": 55315,
+      "train_speed(iter/s)": 1.315533
+    },
+    {
+      "acc": 0.68554626,
+      "epoch": 1.4033485540334856,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.242125408547622e-06,
+      "loss": 1.43616962,
+      "memory(GiB)": 123.79,
+      "step": 55320,
+      "train_speed(iter/s)": 1.315547
+    },
+    {
+      "acc": 0.69014144,
+      "epoch": 1.4034753932014206,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.2412507856059667e-06,
+      "loss": 1.43440323,
+      "memory(GiB)": 123.79,
+      "step": 55325,
+      "train_speed(iter/s)": 1.315561
+    },
+    {
+      "acc": 0.68304811,
+      "epoch": 1.4036022323693556,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.2403762840064223e-06,
+      "loss": 1.38072319,
+      "memory(GiB)": 123.79,
+      "step": 55330,
+      "train_speed(iter/s)": 1.315575
+    },
+    {
+      "acc": 0.69819746,
+      "epoch": 1.4037290715372908,
+      "grad_norm": 2.578125,
+      "learning_rate": 2.239501903787448e-06,
+      "loss": 1.37253342,
+      "memory(GiB)": 123.79,
+      "step": 55335,
+      "train_speed(iter/s)": 1.315589
+    },
+    {
+      "acc": 0.67529993,
+      "epoch": 1.4038559107052258,
+      "grad_norm": 4.25,
+      "learning_rate": 2.2386276449875057e-06,
+      "loss": 1.41210604,
+      "memory(GiB)": 123.79,
+      "step": 55340,
+      "train_speed(iter/s)": 1.315602
+    },
+    {
+      "acc": 0.68256311,
+      "epoch": 1.403982749873161,
+      "grad_norm": 3.625,
+      "learning_rate": 2.2377535076450452e-06,
+      "loss": 1.40513058,
+      "memory(GiB)": 123.79,
+      "step": 55345,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.6910862,
+      "epoch": 1.404109589041096,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.236879491798522e-06,
+      "loss": 1.3423172,
+      "memory(GiB)": 123.79,
+      "step": 55350,
+      "train_speed(iter/s)": 1.31563
+    },
+    {
+      "acc": 0.69341006,
+      "epoch": 1.404236428209031,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.2360055974863737e-06,
+      "loss": 1.41101971,
+      "memory(GiB)": 123.79,
+      "step": 55355,
+      "train_speed(iter/s)": 1.315644
+    },
+    {
+      "acc": 0.69336042,
+      "epoch": 1.404363267376966,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.2351318247470376e-06,
+      "loss": 1.36735172,
+      "memory(GiB)": 123.79,
+      "step": 55360,
+      "train_speed(iter/s)": 1.315658
+    },
+    {
+      "acc": 0.67347536,
+      "epoch": 1.4044901065449011,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.2342581736189496e-06,
+      "loss": 1.42774868,
+      "memory(GiB)": 123.79,
+      "step": 55365,
+      "train_speed(iter/s)": 1.315671
+    },
+    {
+      "acc": 0.68873186,
+      "epoch": 1.4046169457128361,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.233384644140537e-06,
+      "loss": 1.36989021,
+      "memory(GiB)": 123.79,
+      "step": 55370,
+      "train_speed(iter/s)": 1.315685
+    },
+    {
+      "acc": 0.7029336,
+      "epoch": 1.4047437848807711,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.2325112363502167e-06,
+      "loss": 1.30552092,
+      "memory(GiB)": 123.79,
+      "step": 55375,
+      "train_speed(iter/s)": 1.315699
+    },
+    {
+      "acc": 0.6957983,
+      "epoch": 1.4048706240487063,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.23163795028641e-06,
+      "loss": 1.33761902,
+      "memory(GiB)": 123.79,
+      "step": 55380,
+      "train_speed(iter/s)": 1.315712
+    },
+    {
+      "acc": 0.68806634,
+      "epoch": 1.4049974632166413,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.230764785987526e-06,
+      "loss": 1.3546648,
+      "memory(GiB)": 123.79,
+      "step": 55385,
+      "train_speed(iter/s)": 1.315727
+    },
+    {
+      "acc": 0.68293142,
+      "epoch": 1.4051243023845763,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.229891743491972e-06,
+      "loss": 1.33387165,
+      "memory(GiB)": 123.79,
+      "step": 55390,
+      "train_speed(iter/s)": 1.315741
+    },
+    {
+      "acc": 0.68538551,
+      "epoch": 1.4052511415525113,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.2290188228381434e-06,
+      "loss": 1.36645746,
+      "memory(GiB)": 123.79,
+      "step": 55395,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.69175534,
+      "epoch": 1.4053779807204465,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.2281460240644397e-06,
+      "loss": 1.36432114,
+      "memory(GiB)": 123.79,
+      "step": 55400,
+      "train_speed(iter/s)": 1.315767
+    },
+    {
+      "acc": 0.67598591,
+      "epoch": 1.4055048198883815,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.22727334720925e-06,
+      "loss": 1.40812263,
+      "memory(GiB)": 123.79,
+      "step": 55405,
+      "train_speed(iter/s)": 1.315781
+    },
+    {
+      "acc": 0.69352961,
+      "epoch": 1.4056316590563167,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.2264007923109575e-06,
+      "loss": 1.38438435,
+      "memory(GiB)": 123.79,
+      "step": 55410,
+      "train_speed(iter/s)": 1.315794
+    },
+    {
+      "acc": 0.68837876,
+      "epoch": 1.4057584982242517,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.225528359407942e-06,
+      "loss": 1.39103336,
+      "memory(GiB)": 123.79,
+      "step": 55415,
+      "train_speed(iter/s)": 1.315808
+    },
+    {
+      "acc": 0.68658671,
+      "epoch": 1.4058853373921867,
+      "grad_norm": 3.375,
+      "learning_rate": 2.2246560485385756e-06,
+      "loss": 1.35154934,
+      "memory(GiB)": 123.79,
+      "step": 55420,
+      "train_speed(iter/s)": 1.315822
+    },
+    {
+      "acc": 0.68753028,
+      "epoch": 1.4060121765601217,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.2237838597412277e-06,
+      "loss": 1.40279331,
+      "memory(GiB)": 123.79,
+      "step": 55425,
+      "train_speed(iter/s)": 1.315836
+    },
+    {
+      "acc": 0.70232878,
+      "epoch": 1.4061390157280569,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.22291179305426e-06,
+      "loss": 1.2995573,
+      "memory(GiB)": 123.79,
+      "step": 55430,
+      "train_speed(iter/s)": 1.31585
+    },
+    {
+      "acc": 0.69888506,
+      "epoch": 1.4062658548959919,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.222039848516031e-06,
+      "loss": 1.36029129,
+      "memory(GiB)": 123.79,
+      "step": 55435,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68292122,
+      "epoch": 1.4063926940639269,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.2211680261648918e-06,
+      "loss": 1.38064241,
+      "memory(GiB)": 123.79,
+      "step": 55440,
+      "train_speed(iter/s)": 1.315876
+    },
+    {
+      "acc": 0.6908679,
+      "epoch": 1.406519533231862,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.22029632603919e-06,
+      "loss": 1.38230343,
+      "memory(GiB)": 123.79,
+      "step": 55445,
+      "train_speed(iter/s)": 1.315888
+    },
+    {
+      "acc": 0.68749905,
+      "epoch": 1.406646372399797,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.2194247481772652e-06,
+      "loss": 1.3810298,
+      "memory(GiB)": 123.79,
+      "step": 55450,
+      "train_speed(iter/s)": 1.315901
+    },
+    {
+      "acc": 0.68488483,
+      "epoch": 1.406773211567732,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.218553292617455e-06,
+      "loss": 1.39145603,
+      "memory(GiB)": 123.79,
+      "step": 55455,
+      "train_speed(iter/s)": 1.315915
+    },
+    {
+      "acc": 0.68846245,
+      "epoch": 1.406900050735667,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.2176819593980892e-06,
+      "loss": 1.3410244,
+      "memory(GiB)": 123.79,
+      "step": 55460,
+      "train_speed(iter/s)": 1.315928
+    },
+    {
+      "acc": 0.68635931,
+      "epoch": 1.4070268899036023,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.2168107485574914e-06,
+      "loss": 1.46407013,
+      "memory(GiB)": 123.79,
+      "step": 55465,
+      "train_speed(iter/s)": 1.31594
+    },
+    {
+      "acc": 0.67884779,
+      "epoch": 1.4071537290715372,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.215939660133986e-06,
+      "loss": 1.38015652,
+      "memory(GiB)": 123.79,
+      "step": 55470,
+      "train_speed(iter/s)": 1.315954
+    },
+    {
+      "acc": 0.69428482,
+      "epoch": 1.4072805682394725,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.215068694165883e-06,
+      "loss": 1.38207989,
+      "memory(GiB)": 123.79,
+      "step": 55475,
+      "train_speed(iter/s)": 1.315967
+    },
+    {
+      "acc": 0.69503651,
+      "epoch": 1.4074074074074074,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.2141978506914922e-06,
+      "loss": 1.30736189,
+      "memory(GiB)": 123.79,
+      "step": 55480,
+      "train_speed(iter/s)": 1.31598
+    },
+    {
+      "acc": 0.67160244,
+      "epoch": 1.4075342465753424,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.2133271297491165e-06,
+      "loss": 1.49964437,
+      "memory(GiB)": 123.79,
+      "step": 55485,
+      "train_speed(iter/s)": 1.315994
+    },
+    {
+      "acc": 0.69069662,
+      "epoch": 1.4076610857432774,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.2124565313770584e-06,
+      "loss": 1.36058969,
+      "memory(GiB)": 123.79,
+      "step": 55490,
+      "train_speed(iter/s)": 1.316008
+    },
+    {
+      "acc": 0.69822931,
+      "epoch": 1.4077879249112126,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.211586055613606e-06,
+      "loss": 1.36194735,
+      "memory(GiB)": 123.79,
+      "step": 55495,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.67670631,
+      "epoch": 1.4079147640791476,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.210715702497046e-06,
+      "loss": 1.3740387,
+      "memory(GiB)": 123.79,
+      "step": 55500,
+      "train_speed(iter/s)": 1.315887
+    },
+    {
+      "acc": 0.69019527,
+      "epoch": 1.4080416032470828,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.2098454720656647e-06,
+      "loss": 1.42246046,
+      "memory(GiB)": 123.79,
+      "step": 55505,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.69619651,
+      "epoch": 1.4081684424150178,
+      "grad_norm": 3.75,
+      "learning_rate": 2.2089753643577384e-06,
+      "loss": 1.33554411,
+      "memory(GiB)": 123.79,
+      "step": 55510,
+      "train_speed(iter/s)": 1.315914
+    },
+    {
+      "acc": 0.68870387,
+      "epoch": 1.4082952815829528,
+      "grad_norm": 2.875,
+      "learning_rate": 2.208105379411535e-06,
+      "loss": 1.36124468,
+      "memory(GiB)": 123.79,
+      "step": 55515,
+      "train_speed(iter/s)": 1.315927
+    },
+    {
+      "acc": 0.69206152,
+      "epoch": 1.4084221207508878,
+      "grad_norm": 3.25,
+      "learning_rate": 2.2072355172653197e-06,
+      "loss": 1.34864635,
+      "memory(GiB)": 123.79,
+      "step": 55520,
+      "train_speed(iter/s)": 1.315941
+    },
+    {
+      "acc": 0.68659,
+      "epoch": 1.408548959918823,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.2063657779573573e-06,
+      "loss": 1.37310944,
+      "memory(GiB)": 123.79,
+      "step": 55525,
+      "train_speed(iter/s)": 1.315955
+    },
+    {
+      "acc": 0.69201078,
+      "epoch": 1.408675799086758,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.2054961615259023e-06,
+      "loss": 1.3647438,
+      "memory(GiB)": 123.79,
+      "step": 55530,
+      "train_speed(iter/s)": 1.315967
+    },
+    {
+      "acc": 0.7063035,
+      "epoch": 1.408802638254693,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.2046266680091994e-06,
+      "loss": 1.30466747,
+      "memory(GiB)": 123.79,
+      "step": 55535,
+      "train_speed(iter/s)": 1.31598
+    },
+    {
+      "acc": 0.67739782,
+      "epoch": 1.4089294774226282,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.2037572974454974e-06,
+      "loss": 1.46743145,
+      "memory(GiB)": 123.79,
+      "step": 55540,
+      "train_speed(iter/s)": 1.315994
+    },
+    {
+      "acc": 0.69471602,
+      "epoch": 1.4090563165905632,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.202888049873034e-06,
+      "loss": 1.36299706,
+      "memory(GiB)": 123.79,
+      "step": 55545,
+      "train_speed(iter/s)": 1.316007
+    },
+    {
+      "acc": 0.68213105,
+      "epoch": 1.4091831557584982,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.2020189253300428e-06,
+      "loss": 1.51539488,
+      "memory(GiB)": 123.79,
+      "step": 55550,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.68350549,
+      "epoch": 1.4093099949264332,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.2011499238547506e-06,
+      "loss": 1.44181852,
+      "memory(GiB)": 123.79,
+      "step": 55555,
+      "train_speed(iter/s)": 1.316035
+    },
+    {
+      "acc": 0.67652779,
+      "epoch": 1.4094368340943684,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.2002810454853813e-06,
+      "loss": 1.42891436,
+      "memory(GiB)": 123.79,
+      "step": 55560,
+      "train_speed(iter/s)": 1.316048
+    },
+    {
+      "acc": 0.68574657,
+      "epoch": 1.4095636732623034,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.1994122902601513e-06,
+      "loss": 1.40437756,
+      "memory(GiB)": 123.79,
+      "step": 55565,
+      "train_speed(iter/s)": 1.316062
+    },
+    {
+      "acc": 0.67775412,
+      "epoch": 1.4096905124302386,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.1985436582172724e-06,
+      "loss": 1.37767878,
+      "memory(GiB)": 123.79,
+      "step": 55570,
+      "train_speed(iter/s)": 1.316074
+    },
+    {
+      "acc": 0.68955832,
+      "epoch": 1.4098173515981736,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.1976751493949512e-06,
+      "loss": 1.4349905,
+      "memory(GiB)": 123.79,
+      "step": 55575,
+      "train_speed(iter/s)": 1.316087
+    },
+    {
+      "acc": 0.68387938,
+      "epoch": 1.4099441907661086,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.196806763831388e-06,
+      "loss": 1.38492603,
+      "memory(GiB)": 123.79,
+      "step": 55580,
+      "train_speed(iter/s)": 1.316099
+    },
+    {
+      "acc": 0.69192333,
+      "epoch": 1.4100710299340435,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.1959385015647794e-06,
+      "loss": 1.44215822,
+      "memory(GiB)": 123.79,
+      "step": 55585,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.68634362,
+      "epoch": 1.4101978691019788,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.195070362633314e-06,
+      "loss": 1.42065868,
+      "memory(GiB)": 123.79,
+      "step": 55590,
+      "train_speed(iter/s)": 1.316126
+    },
+    {
+      "acc": 0.68693285,
+      "epoch": 1.4103247082699137,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.194202347075178e-06,
+      "loss": 1.41807346,
+      "memory(GiB)": 123.79,
+      "step": 55595,
+      "train_speed(iter/s)": 1.31614
+    },
+    {
+      "acc": 0.68465776,
+      "epoch": 1.4104515474378487,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.1933344549285485e-06,
+      "loss": 1.41313162,
+      "memory(GiB)": 123.79,
+      "step": 55600,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.68413897,
+      "epoch": 1.410578386605784,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.1924666862316015e-06,
+      "loss": 1.42288132,
+      "memory(GiB)": 123.79,
+      "step": 55605,
+      "train_speed(iter/s)": 1.316167
+    },
+    {
+      "acc": 0.69222231,
+      "epoch": 1.410705225773719,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.191599041022504e-06,
+      "loss": 1.35732422,
+      "memory(GiB)": 123.79,
+      "step": 55610,
+      "train_speed(iter/s)": 1.316181
+    },
+    {
+      "acc": 0.68634415,
+      "epoch": 1.410832064941654,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.19073151933942e-06,
+      "loss": 1.33407421,
+      "memory(GiB)": 123.79,
+      "step": 55615,
+      "train_speed(iter/s)": 1.316194
+    },
+    {
+      "acc": 0.68650208,
+      "epoch": 1.410958904109589,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.1898641212205053e-06,
+      "loss": 1.36848583,
+      "memory(GiB)": 123.79,
+      "step": 55620,
+      "train_speed(iter/s)": 1.316208
+    },
+    {
+      "acc": 0.6825654,
+      "epoch": 1.4110857432775241,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.1889968467039114e-06,
+      "loss": 1.39177132,
+      "memory(GiB)": 123.79,
+      "step": 55625,
+      "train_speed(iter/s)": 1.316221
+    },
+    {
+      "acc": 0.68298311,
+      "epoch": 1.411212582445459,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.1881296958277897e-06,
+      "loss": 1.38504696,
+      "memory(GiB)": 123.79,
+      "step": 55630,
+      "train_speed(iter/s)": 1.316235
+    },
+    {
+      "acc": 0.69938955,
+      "epoch": 1.4113394216133943,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.1872626686302767e-06,
+      "loss": 1.35593224,
+      "memory(GiB)": 123.79,
+      "step": 55635,
+      "train_speed(iter/s)": 1.316248
+    },
+    {
+      "acc": 0.68985682,
+      "epoch": 1.4114662607813293,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.186395765149508e-06,
+      "loss": 1.31730328,
+      "memory(GiB)": 123.79,
+      "step": 55640,
+      "train_speed(iter/s)": 1.316262
+    },
+    {
+      "acc": 0.67659903,
+      "epoch": 1.4115930999492643,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.1855289854236165e-06,
+      "loss": 1.46002588,
+      "memory(GiB)": 123.79,
+      "step": 55645,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.68207769,
+      "epoch": 1.4117199391171993,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.184662329490728e-06,
+      "loss": 1.3868001,
+      "memory(GiB)": 123.79,
+      "step": 55650,
+      "train_speed(iter/s)": 1.316289
+    },
+    {
+      "acc": 0.68595576,
+      "epoch": 1.4118467782851345,
+      "grad_norm": 3.75,
+      "learning_rate": 2.1837957973889584e-06,
+      "loss": 1.38571472,
+      "memory(GiB)": 123.79,
+      "step": 55655,
+      "train_speed(iter/s)": 1.316303
+    },
+    {
+      "acc": 0.67001777,
+      "epoch": 1.4119736174530695,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.1829293891564212e-06,
+      "loss": 1.43563271,
+      "memory(GiB)": 123.79,
+      "step": 55660,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.67991586,
+      "epoch": 1.4121004566210047,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.182063104831228e-06,
+      "loss": 1.39147367,
+      "memory(GiB)": 123.79,
+      "step": 55665,
+      "train_speed(iter/s)": 1.31633
+    },
+    {
+      "acc": 0.6868741,
+      "epoch": 1.4122272957889397,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.181196944451483e-06,
+      "loss": 1.41299667,
+      "memory(GiB)": 123.79,
+      "step": 55670,
+      "train_speed(iter/s)": 1.316343
+    },
+    {
+      "acc": 0.68559656,
+      "epoch": 1.4123541349568747,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.180330908055278e-06,
+      "loss": 1.44224777,
+      "memory(GiB)": 123.79,
+      "step": 55675,
+      "train_speed(iter/s)": 1.316356
+    },
+    {
+      "acc": 0.68013277,
+      "epoch": 1.4124809741248097,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.1794649956807094e-06,
+      "loss": 1.38997564,
+      "memory(GiB)": 123.79,
+      "step": 55680,
+      "train_speed(iter/s)": 1.316369
+    },
+    {
+      "acc": 0.70416002,
+      "epoch": 1.4126078132927449,
+      "grad_norm": 3.0,
+      "learning_rate": 2.178599207365864e-06,
+      "loss": 1.28425484,
+      "memory(GiB)": 123.79,
+      "step": 55685,
+      "train_speed(iter/s)": 1.316382
+    },
+    {
+      "acc": 0.7020462,
+      "epoch": 1.4127346524606799,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.1777335431488234e-06,
+      "loss": 1.35322132,
+      "memory(GiB)": 123.79,
+      "step": 55690,
+      "train_speed(iter/s)": 1.316396
+    },
+    {
+      "acc": 0.69266858,
+      "epoch": 1.4128614916286149,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.176868003067659e-06,
+      "loss": 1.37402439,
+      "memory(GiB)": 123.79,
+      "step": 55695,
+      "train_speed(iter/s)": 1.31641
+    },
+    {
+      "acc": 0.68447828,
+      "epoch": 1.41298833079655,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.1760025871604456e-06,
+      "loss": 1.3848423,
+      "memory(GiB)": 123.79,
+      "step": 55700,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.69621954,
+      "epoch": 1.413115169964485,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.175137295465247e-06,
+      "loss": 1.35642147,
+      "memory(GiB)": 123.79,
+      "step": 55705,
+      "train_speed(iter/s)": 1.316438
+    },
+    {
+      "acc": 0.68841548,
+      "epoch": 1.41324200913242,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.174272128020122e-06,
+      "loss": 1.41045799,
+      "memory(GiB)": 123.79,
+      "step": 55710,
+      "train_speed(iter/s)": 1.316452
+    },
+    {
+      "acc": 0.69583216,
+      "epoch": 1.413368848300355,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.1734070848631245e-06,
+      "loss": 1.37444935,
+      "memory(GiB)": 123.79,
+      "step": 55715,
+      "train_speed(iter/s)": 1.316466
+    },
+    {
+      "acc": 0.69922929,
+      "epoch": 1.4134956874682902,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.172542166032303e-06,
+      "loss": 1.33042812,
+      "memory(GiB)": 123.79,
+      "step": 55720,
+      "train_speed(iter/s)": 1.31648
+    },
+    {
+      "acc": 0.68940821,
+      "epoch": 1.4136225266362252,
+      "grad_norm": 2.875,
+      "learning_rate": 2.171677371565701e-06,
+      "loss": 1.40651674,
+      "memory(GiB)": 123.79,
+      "step": 55725,
+      "train_speed(iter/s)": 1.316494
+    },
+    {
+      "acc": 0.68423786,
+      "epoch": 1.4137493658041604,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.1708127015013565e-06,
+      "loss": 1.41795635,
+      "memory(GiB)": 123.79,
+      "step": 55730,
+      "train_speed(iter/s)": 1.316508
+    },
+    {
+      "acc": 0.68065267,
+      "epoch": 1.4138762049720954,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.1699481558773e-06,
+      "loss": 1.36756058,
+      "memory(GiB)": 123.79,
+      "step": 55735,
+      "train_speed(iter/s)": 1.316522
+    },
+    {
+      "acc": 0.68295193,
+      "epoch": 1.4140030441400304,
+      "grad_norm": 2.453125,
+      "learning_rate": 2.1690837347315597e-06,
+      "loss": 1.39802713,
+      "memory(GiB)": 123.79,
+      "step": 55740,
+      "train_speed(iter/s)": 1.316536
+    },
+    {
+      "acc": 0.695438,
+      "epoch": 1.4141298833079654,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.168219438102155e-06,
+      "loss": 1.36860924,
+      "memory(GiB)": 123.79,
+      "step": 55745,
+      "train_speed(iter/s)": 1.31655
+    },
+    {
+      "acc": 0.67690868,
+      "epoch": 1.4142567224759006,
+      "grad_norm": 3.875,
+      "learning_rate": 2.167355266027103e-06,
+      "loss": 1.39063244,
+      "memory(GiB)": 123.79,
+      "step": 55750,
+      "train_speed(iter/s)": 1.316564
+    },
+    {
+      "acc": 0.6976006,
+      "epoch": 1.4143835616438356,
+      "grad_norm": 3.125,
+      "learning_rate": 2.1664912185444127e-06,
+      "loss": 1.32012615,
+      "memory(GiB)": 123.79,
+      "step": 55755,
+      "train_speed(iter/s)": 1.316579
+    },
+    {
+      "acc": 0.68242249,
+      "epoch": 1.4145104008117706,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.16562729569209e-06,
+      "loss": 1.48290777,
+      "memory(GiB)": 123.79,
+      "step": 55760,
+      "train_speed(iter/s)": 1.316593
+    },
+    {
+      "acc": 0.69190707,
+      "epoch": 1.4146372399797058,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.164763497508131e-06,
+      "loss": 1.35224571,
+      "memory(GiB)": 123.79,
+      "step": 55765,
+      "train_speed(iter/s)": 1.316607
+    },
+    {
+      "acc": 0.70310335,
+      "epoch": 1.4147640791476408,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.1638998240305355e-06,
+      "loss": 1.37673893,
+      "memory(GiB)": 123.79,
+      "step": 55770,
+      "train_speed(iter/s)": 1.31662
+    },
+    {
+      "acc": 0.670821,
+      "epoch": 1.4148909183155758,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.163036275297286e-06,
+      "loss": 1.42577648,
+      "memory(GiB)": 123.79,
+      "step": 55775,
+      "train_speed(iter/s)": 1.316634
+    },
+    {
+      "acc": 0.68508081,
+      "epoch": 1.4150177574835108,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.162172851346368e-06,
+      "loss": 1.37617569,
+      "memory(GiB)": 123.79,
+      "step": 55780,
+      "train_speed(iter/s)": 1.316648
+    },
+    {
+      "acc": 0.69986706,
+      "epoch": 1.415144596651446,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.1613095522157557e-06,
+      "loss": 1.34027157,
+      "memory(GiB)": 123.79,
+      "step": 55785,
+      "train_speed(iter/s)": 1.316662
+    },
+    {
+      "acc": 0.68531332,
+      "epoch": 1.415271435819381,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.1604463779434267e-06,
+      "loss": 1.35829849,
+      "memory(GiB)": 123.79,
+      "step": 55790,
+      "train_speed(iter/s)": 1.316676
+    },
+    {
+      "acc": 0.68574443,
+      "epoch": 1.4153982749873162,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.159583328567342e-06,
+      "loss": 1.42619495,
+      "memory(GiB)": 123.79,
+      "step": 55795,
+      "train_speed(iter/s)": 1.31669
+    },
+    {
+      "acc": 0.69439197,
+      "epoch": 1.4155251141552512,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.158720404125462e-06,
+      "loss": 1.33611488,
+      "memory(GiB)": 123.79,
+      "step": 55800,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.69395323,
+      "epoch": 1.4156519533231862,
+      "grad_norm": 3.0,
+      "learning_rate": 2.1578576046557463e-06,
+      "loss": 1.36924877,
+      "memory(GiB)": 123.79,
+      "step": 55805,
+      "train_speed(iter/s)": 1.316717
+    },
+    {
+      "acc": 0.67783561,
+      "epoch": 1.4157787924911212,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.156994930196144e-06,
+      "loss": 1.39638033,
+      "memory(GiB)": 123.79,
+      "step": 55810,
+      "train_speed(iter/s)": 1.316731
+    },
+    {
+      "acc": 0.6986444,
+      "epoch": 1.4159056316590564,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.156132380784594e-06,
+      "loss": 1.32507792,
+      "memory(GiB)": 123.79,
+      "step": 55815,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.68330994,
+      "epoch": 1.4160324708269914,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.155269956459041e-06,
+      "loss": 1.38858852,
+      "memory(GiB)": 123.79,
+      "step": 55820,
+      "train_speed(iter/s)": 1.316757
+    },
+    {
+      "acc": 0.69316235,
+      "epoch": 1.4161593099949266,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.1544076572574156e-06,
+      "loss": 1.38385592,
+      "memory(GiB)": 123.79,
+      "step": 55825,
+      "train_speed(iter/s)": 1.31677
+    },
+    {
+      "acc": 0.68401656,
+      "epoch": 1.4162861491628616,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.1535454832176482e-06,
+      "loss": 1.37950287,
+      "memory(GiB)": 123.79,
+      "step": 55830,
+      "train_speed(iter/s)": 1.316784
+    },
+    {
+      "acc": 0.6772254,
+      "epoch": 1.4164129883307965,
+      "grad_norm": 3.25,
+      "learning_rate": 2.1526834343776556e-06,
+      "loss": 1.47482681,
+      "memory(GiB)": 123.79,
+      "step": 55835,
+      "train_speed(iter/s)": 1.316798
+    },
+    {
+      "acc": 0.66476498,
+      "epoch": 1.4165398274987315,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.1518215107753593e-06,
+      "loss": 1.45860939,
+      "memory(GiB)": 123.79,
+      "step": 55840,
+      "train_speed(iter/s)": 1.316813
+    },
+    {
+      "acc": 0.67576447,
+      "epoch": 1.4166666666666667,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.1509597124486693e-06,
+      "loss": 1.41761732,
+      "memory(GiB)": 123.79,
+      "step": 55845,
+      "train_speed(iter/s)": 1.316827
+    },
+    {
+      "acc": 0.69761109,
+      "epoch": 1.4167935058346017,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.1500980394354907e-06,
+      "loss": 1.36998463,
+      "memory(GiB)": 123.79,
+      "step": 55850,
+      "train_speed(iter/s)": 1.316841
+    },
+    {
+      "acc": 0.68558164,
+      "epoch": 1.4169203450025367,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.1492364917737252e-06,
+      "loss": 1.39210548,
+      "memory(GiB)": 123.79,
+      "step": 55855,
+      "train_speed(iter/s)": 1.316855
+    },
+    {
+      "acc": 0.68774633,
+      "epoch": 1.417047184170472,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.148375069501266e-06,
+      "loss": 1.36669359,
+      "memory(GiB)": 123.79,
+      "step": 55860,
+      "train_speed(iter/s)": 1.316869
+    },
+    {
+      "acc": 0.68651609,
+      "epoch": 1.417174023338407,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.147513772656003e-06,
+      "loss": 1.3602129,
+      "memory(GiB)": 123.79,
+      "step": 55865,
+      "train_speed(iter/s)": 1.316884
+    },
+    {
+      "acc": 0.68023262,
+      "epoch": 1.417300862506342,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.1466526012758194e-06,
+      "loss": 1.37627001,
+      "memory(GiB)": 123.79,
+      "step": 55870,
+      "train_speed(iter/s)": 1.316897
+    },
+    {
+      "acc": 0.69408178,
+      "epoch": 1.417427701674277,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.145791555398594e-06,
+      "loss": 1.3316,
+      "memory(GiB)": 123.79,
+      "step": 55875,
+      "train_speed(iter/s)": 1.316911
+    },
+    {
+      "acc": 0.69626002,
+      "epoch": 1.417554540842212,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.144930635062199e-06,
+      "loss": 1.31375618,
+      "memory(GiB)": 123.79,
+      "step": 55880,
+      "train_speed(iter/s)": 1.316925
+    },
+    {
+      "acc": 0.67081203,
+      "epoch": 1.417681380010147,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.144069840304502e-06,
+      "loss": 1.46562414,
+      "memory(GiB)": 123.79,
+      "step": 55885,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.69597087,
+      "epoch": 1.4178082191780823,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.1432091711633634e-06,
+      "loss": 1.43155918,
+      "memory(GiB)": 123.79,
+      "step": 55890,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.68830357,
+      "epoch": 1.4179350583460173,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.142348627676641e-06,
+      "loss": 1.39231052,
+      "memory(GiB)": 123.79,
+      "step": 55895,
+      "train_speed(iter/s)": 1.316967
+    },
+    {
+      "acc": 0.68894463,
+      "epoch": 1.4180618975139523,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.1414882098821836e-06,
+      "loss": 1.37421494,
+      "memory(GiB)": 123.79,
+      "step": 55900,
+      "train_speed(iter/s)": 1.316982
+    },
+    {
+      "acc": 0.69165001,
+      "epoch": 1.4181887366818873,
+      "grad_norm": 3.5,
+      "learning_rate": 2.1406279178178355e-06,
+      "loss": 1.40720167,
+      "memory(GiB)": 123.79,
+      "step": 55905,
+      "train_speed(iter/s)": 1.316995
+    },
+    {
+      "acc": 0.69104557,
+      "epoch": 1.4183155758498225,
+      "grad_norm": 3.25,
+      "learning_rate": 2.1397677515214422e-06,
+      "loss": 1.4011734,
+      "memory(GiB)": 123.79,
+      "step": 55910,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.68155251,
+      "epoch": 1.4184424150177575,
+      "grad_norm": 4.25,
+      "learning_rate": 2.1389077110308304e-06,
+      "loss": 1.43139095,
+      "memory(GiB)": 123.79,
+      "step": 55915,
+      "train_speed(iter/s)": 1.317024
+    },
+    {
+      "acc": 0.6849514,
+      "epoch": 1.4185692541856925,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.138047796383832e-06,
+      "loss": 1.43350353,
+      "memory(GiB)": 123.79,
+      "step": 55920,
+      "train_speed(iter/s)": 1.317037
+    },
+    {
+      "acc": 0.68276534,
+      "epoch": 1.4186960933536277,
+      "grad_norm": 4.96875,
+      "learning_rate": 2.1371880076182666e-06,
+      "loss": 1.44260092,
+      "memory(GiB)": 123.79,
+      "step": 55925,
+      "train_speed(iter/s)": 1.317051
+    },
+    {
+      "acc": 0.67162776,
+      "epoch": 1.4188229325215627,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.1363283447719584e-06,
+      "loss": 1.45988312,
+      "memory(GiB)": 123.79,
+      "step": 55930,
+      "train_speed(iter/s)": 1.317064
+    },
+    {
+      "acc": 0.703054,
+      "epoch": 1.4189497716894977,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.135468807882713e-06,
+      "loss": 1.34570837,
+      "memory(GiB)": 123.79,
+      "step": 55935,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.68416438,
+      "epoch": 1.4190766108574326,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.1346093969883367e-06,
+      "loss": 1.40640783,
+      "memory(GiB)": 123.79,
+      "step": 55940,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.6919158,
+      "epoch": 1.4192034500253679,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.1337501121266345e-06,
+      "loss": 1.34455929,
+      "memory(GiB)": 123.79,
+      "step": 55945,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.67222404,
+      "epoch": 1.4193302891933028,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.132890953335401e-06,
+      "loss": 1.48116798,
+      "memory(GiB)": 123.79,
+      "step": 55950,
+      "train_speed(iter/s)": 1.317118
+    },
+    {
+      "acc": 0.68330793,
+      "epoch": 1.419457128361238,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.1320319206524215e-06,
+      "loss": 1.48487892,
+      "memory(GiB)": 123.79,
+      "step": 55955,
+      "train_speed(iter/s)": 1.317132
+    },
+    {
+      "acc": 0.68964419,
+      "epoch": 1.419583967529173,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.1311730141154813e-06,
+      "loss": 1.40166855,
+      "memory(GiB)": 123.79,
+      "step": 55960,
+      "train_speed(iter/s)": 1.317145
+    },
+    {
+      "acc": 0.68018484,
+      "epoch": 1.419710806697108,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.1303142337623623e-06,
+      "loss": 1.40666418,
+      "memory(GiB)": 123.79,
+      "step": 55965,
+      "train_speed(iter/s)": 1.317158
+    },
+    {
+      "acc": 0.69077668,
+      "epoch": 1.419837645865043,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.1294555796308375e-06,
+      "loss": 1.35629158,
+      "memory(GiB)": 123.79,
+      "step": 55970,
+      "train_speed(iter/s)": 1.317171
+    },
+    {
+      "acc": 0.6881958,
+      "epoch": 1.4199644850329782,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.1285970517586686e-06,
+      "loss": 1.45337648,
+      "memory(GiB)": 123.79,
+      "step": 55975,
+      "train_speed(iter/s)": 1.317184
+    },
+    {
+      "acc": 0.67429686,
+      "epoch": 1.4200913242009132,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.127738650183623e-06,
+      "loss": 1.39563522,
+      "memory(GiB)": 123.79,
+      "step": 55980,
+      "train_speed(iter/s)": 1.317197
+    },
+    {
+      "acc": 0.68323689,
+      "epoch": 1.4202181633688484,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.1268803749434546e-06,
+      "loss": 1.40102301,
+      "memory(GiB)": 123.79,
+      "step": 55985,
+      "train_speed(iter/s)": 1.31721
+    },
+    {
+      "acc": 0.67266111,
+      "epoch": 1.4203450025367834,
+      "grad_norm": 4.25,
+      "learning_rate": 2.1260222260759158e-06,
+      "loss": 1.39251194,
+      "memory(GiB)": 123.79,
+      "step": 55990,
+      "train_speed(iter/s)": 1.317223
+    },
+    {
+      "acc": 0.69091678,
+      "epoch": 1.4204718417047184,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.1251642036187502e-06,
+      "loss": 1.383992,
+      "memory(GiB)": 123.79,
+      "step": 55995,
+      "train_speed(iter/s)": 1.317236
+    },
+    {
+      "acc": 0.69714251,
+      "epoch": 1.4205986808726534,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.124306307609699e-06,
+      "loss": 1.3786377,
+      "memory(GiB)": 123.79,
+      "step": 56000,
+      "train_speed(iter/s)": 1.317248
+    },
+    {
+      "epoch": 1.4205986808726534,
+      "eval_acc": 0.6751531250456817,
+      "eval_loss": 1.3562688827514648,
+      "eval_runtime": 69.76,
+      "eval_samples_per_second": 91.313,
+      "eval_steps_per_second": 22.835,
+      "step": 56000
+    },
+    {
+      "acc": 0.666431,
+      "epoch": 1.4207255200405886,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.123448538086495e-06,
+      "loss": 1.47786427,
+      "memory(GiB)": 123.79,
+      "step": 56005,
+      "train_speed(iter/s)": 1.314716
+    },
+    {
+      "acc": 0.68027191,
+      "epoch": 1.4208523592085236,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.122590895086867e-06,
+      "loss": 1.46946268,
+      "memory(GiB)": 123.79,
+      "step": 56010,
+      "train_speed(iter/s)": 1.314729
+    },
+    {
+      "acc": 0.67946844,
+      "epoch": 1.4209791983764586,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.1217333786485385e-06,
+      "loss": 1.38717403,
+      "memory(GiB)": 123.79,
+      "step": 56015,
+      "train_speed(iter/s)": 1.314742
+    },
+    {
+      "acc": 0.68018808,
+      "epoch": 1.4211060375443938,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.120875988809226e-06,
+      "loss": 1.4174963,
+      "memory(GiB)": 123.79,
+      "step": 56020,
+      "train_speed(iter/s)": 1.314754
+    },
+    {
+      "acc": 0.68874807,
+      "epoch": 1.4212328767123288,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.1200187256066425e-06,
+      "loss": 1.34745493,
+      "memory(GiB)": 123.79,
+      "step": 56025,
+      "train_speed(iter/s)": 1.314767
+    },
+    {
+      "acc": 0.69278545,
+      "epoch": 1.4213597158802638,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.119161589078493e-06,
+      "loss": 1.40420818,
+      "memory(GiB)": 123.79,
+      "step": 56030,
+      "train_speed(iter/s)": 1.314779
+    },
+    {
+      "acc": 0.67073421,
+      "epoch": 1.4214865550481988,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.118304579262479e-06,
+      "loss": 1.48980999,
+      "memory(GiB)": 123.79,
+      "step": 56035,
+      "train_speed(iter/s)": 1.31479
+    },
+    {
+      "acc": 0.69836907,
+      "epoch": 1.421613394216134,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.1174476961962957e-06,
+      "loss": 1.39218693,
+      "memory(GiB)": 123.79,
+      "step": 56040,
+      "train_speed(iter/s)": 1.314802
+    },
+    {
+      "acc": 0.68105536,
+      "epoch": 1.421740233384069,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.1165909399176328e-06,
+      "loss": 1.43116236,
+      "memory(GiB)": 123.79,
+      "step": 56045,
+      "train_speed(iter/s)": 1.314814
+    },
+    {
+      "acc": 0.67027235,
+      "epoch": 1.4218670725520042,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.1157343104641733e-06,
+      "loss": 1.44433842,
+      "memory(GiB)": 123.79,
+      "step": 56050,
+      "train_speed(iter/s)": 1.314826
+    },
+    {
+      "acc": 0.68499508,
+      "epoch": 1.4219939117199392,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.114877807873596e-06,
+      "loss": 1.34307699,
+      "memory(GiB)": 123.79,
+      "step": 56055,
+      "train_speed(iter/s)": 1.314838
+    },
+    {
+      "acc": 0.675804,
+      "epoch": 1.4221207508878742,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.114021432183574e-06,
+      "loss": 1.40075569,
+      "memory(GiB)": 123.79,
+      "step": 56060,
+      "train_speed(iter/s)": 1.314851
+    },
+    {
+      "acc": 0.67890606,
+      "epoch": 1.4222475900558091,
+      "grad_norm": 3.0,
+      "learning_rate": 2.113165183431773e-06,
+      "loss": 1.39930048,
+      "memory(GiB)": 123.79,
+      "step": 56065,
+      "train_speed(iter/s)": 1.314863
+    },
+    {
+      "acc": 0.70614786,
+      "epoch": 1.4223744292237444,
+      "grad_norm": 3.875,
+      "learning_rate": 2.112309061655859e-06,
+      "loss": 1.31234179,
+      "memory(GiB)": 123.79,
+      "step": 56070,
+      "train_speed(iter/s)": 1.314876
+    },
+    {
+      "acc": 0.71512856,
+      "epoch": 1.4225012683916793,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.1114530668934836e-06,
+      "loss": 1.3192564,
+      "memory(GiB)": 123.79,
+      "step": 56075,
+      "train_speed(iter/s)": 1.314888
+    },
+    {
+      "acc": 0.68810754,
+      "epoch": 1.4226281075596143,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.1105971991822966e-06,
+      "loss": 1.38840866,
+      "memory(GiB)": 123.79,
+      "step": 56080,
+      "train_speed(iter/s)": 1.314902
+    },
+    {
+      "acc": 0.67952433,
+      "epoch": 1.4227549467275495,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.1097414585599474e-06,
+      "loss": 1.4265542,
+      "memory(GiB)": 123.79,
+      "step": 56085,
+      "train_speed(iter/s)": 1.314915
+    },
+    {
+      "acc": 0.69460554,
+      "epoch": 1.4228817858954845,
+      "grad_norm": 3.625,
+      "learning_rate": 2.1088858450640743e-06,
+      "loss": 1.42996025,
+      "memory(GiB)": 123.79,
+      "step": 56090,
+      "train_speed(iter/s)": 1.314928
+    },
+    {
+      "acc": 0.69292679,
+      "epoch": 1.4230086250634195,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.108030358732308e-06,
+      "loss": 1.36631718,
+      "memory(GiB)": 123.79,
+      "step": 56095,
+      "train_speed(iter/s)": 1.314941
+    },
+    {
+      "acc": 0.69123931,
+      "epoch": 1.4231354642313545,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.107174999602277e-06,
+      "loss": 1.4099884,
+      "memory(GiB)": 123.79,
+      "step": 56100,
+      "train_speed(iter/s)": 1.314953
+    },
+    {
+      "acc": 0.69435148,
+      "epoch": 1.4232623033992897,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.106319767711606e-06,
+      "loss": 1.35719738,
+      "memory(GiB)": 123.79,
+      "step": 56105,
+      "train_speed(iter/s)": 1.314967
+    },
+    {
+      "acc": 0.67061882,
+      "epoch": 1.4233891425672247,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.105464663097913e-06,
+      "loss": 1.43601418,
+      "memory(GiB)": 123.79,
+      "step": 56110,
+      "train_speed(iter/s)": 1.314979
+    },
+    {
+      "acc": 0.69668965,
+      "epoch": 1.42351598173516,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.1046096857988033e-06,
+      "loss": 1.34624319,
+      "memory(GiB)": 123.79,
+      "step": 56115,
+      "train_speed(iter/s)": 1.314992
+    },
+    {
+      "acc": 0.68485909,
+      "epoch": 1.423642820903095,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.103754835851889e-06,
+      "loss": 1.37191181,
+      "memory(GiB)": 123.79,
+      "step": 56120,
+      "train_speed(iter/s)": 1.315004
+    },
+    {
+      "acc": 0.69770374,
+      "epoch": 1.42376966007103,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.102900113294768e-06,
+      "loss": 1.3349308,
+      "memory(GiB)": 123.79,
+      "step": 56125,
+      "train_speed(iter/s)": 1.315017
+    },
+    {
+      "acc": 0.68396454,
+      "epoch": 1.4238964992389649,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.1020455181650356e-06,
+      "loss": 1.38061295,
+      "memory(GiB)": 123.79,
+      "step": 56130,
+      "train_speed(iter/s)": 1.315026
+    },
+    {
+      "acc": 0.69020357,
+      "epoch": 1.4240233384069,
+      "grad_norm": 3.375,
+      "learning_rate": 2.101191050500277e-06,
+      "loss": 1.3850956,
+      "memory(GiB)": 123.79,
+      "step": 56135,
+      "train_speed(iter/s)": 1.315038
+    },
+    {
+      "acc": 0.67661762,
+      "epoch": 1.424150177574835,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.1003367103380797e-06,
+      "loss": 1.39387093,
+      "memory(GiB)": 123.79,
+      "step": 56140,
+      "train_speed(iter/s)": 1.31505
+    },
+    {
+      "acc": 0.6901176,
+      "epoch": 1.4242770167427703,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.09948249771602e-06,
+      "loss": 1.45708504,
+      "memory(GiB)": 123.79,
+      "step": 56145,
+      "train_speed(iter/s)": 1.315062
+    },
+    {
+      "acc": 0.68536549,
+      "epoch": 1.4244038559107053,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.09862841267167e-06,
+      "loss": 1.32048779,
+      "memory(GiB)": 123.79,
+      "step": 56150,
+      "train_speed(iter/s)": 1.315074
+    },
+    {
+      "acc": 0.68639469,
+      "epoch": 1.4245306950786403,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.097774455242596e-06,
+      "loss": 1.41350718,
+      "memory(GiB)": 123.79,
+      "step": 56155,
+      "train_speed(iter/s)": 1.315086
+    },
+    {
+      "acc": 0.68810263,
+      "epoch": 1.4246575342465753,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.096920625466359e-06,
+      "loss": 1.36223412,
+      "memory(GiB)": 123.79,
+      "step": 56160,
+      "train_speed(iter/s)": 1.315098
+    },
+    {
+      "acc": 0.67519855,
+      "epoch": 1.4247843734145105,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.0960669233805136e-06,
+      "loss": 1.45412302,
+      "memory(GiB)": 123.79,
+      "step": 56165,
+      "train_speed(iter/s)": 1.31511
+    },
+    {
+      "acc": 0.68725934,
+      "epoch": 1.4249112125824455,
+      "grad_norm": 2.75,
+      "learning_rate": 2.0952133490226106e-06,
+      "loss": 1.45846891,
+      "memory(GiB)": 123.79,
+      "step": 56170,
+      "train_speed(iter/s)": 1.315122
+    },
+    {
+      "acc": 0.67707357,
+      "epoch": 1.4250380517503805,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.0943599024301935e-06,
+      "loss": 1.42849436,
+      "memory(GiB)": 123.79,
+      "step": 56175,
+      "train_speed(iter/s)": 1.315134
+    },
+    {
+      "acc": 0.68437953,
+      "epoch": 1.4251648909183157,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.093506583640801e-06,
+      "loss": 1.3585165,
+      "memory(GiB)": 123.79,
+      "step": 56180,
+      "train_speed(iter/s)": 1.315147
+    },
+    {
+      "acc": 0.69184141,
+      "epoch": 1.4252917300862507,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.092653392691965e-06,
+      "loss": 1.34794121,
+      "memory(GiB)": 123.79,
+      "step": 56185,
+      "train_speed(iter/s)": 1.315159
+    },
+    {
+      "acc": 0.69325476,
+      "epoch": 1.4254185692541856,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.091800329621213e-06,
+      "loss": 1.38780041,
+      "memory(GiB)": 123.79,
+      "step": 56190,
+      "train_speed(iter/s)": 1.315171
+    },
+    {
+      "acc": 0.70022602,
+      "epoch": 1.4255454084221206,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.0909473944660667e-06,
+      "loss": 1.34633245,
+      "memory(GiB)": 123.79,
+      "step": 56195,
+      "train_speed(iter/s)": 1.315184
+    },
+    {
+      "acc": 0.68564615,
+      "epoch": 1.4256722475900558,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.0900945872640427e-06,
+      "loss": 1.43077793,
+      "memory(GiB)": 123.79,
+      "step": 56200,
+      "train_speed(iter/s)": 1.315197
+    },
+    {
+      "acc": 0.69291911,
+      "epoch": 1.4257990867579908,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.0892419080526484e-06,
+      "loss": 1.36464453,
+      "memory(GiB)": 123.79,
+      "step": 56205,
+      "train_speed(iter/s)": 1.315208
+    },
+    {
+      "acc": 0.69917717,
+      "epoch": 1.425925925925926,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.0883893568693935e-06,
+      "loss": 1.34609499,
+      "memory(GiB)": 123.79,
+      "step": 56210,
+      "train_speed(iter/s)": 1.31522
+    },
+    {
+      "acc": 0.69488373,
+      "epoch": 1.426052765093861,
+      "grad_norm": 3.0,
+      "learning_rate": 2.087536933751773e-06,
+      "loss": 1.38816471,
+      "memory(GiB)": 123.79,
+      "step": 56215,
+      "train_speed(iter/s)": 1.315233
+    },
+    {
+      "acc": 0.68935909,
+      "epoch": 1.426179604261796,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.0866846387372814e-06,
+      "loss": 1.3885994,
+      "memory(GiB)": 123.79,
+      "step": 56220,
+      "train_speed(iter/s)": 1.315245
+    },
+    {
+      "acc": 0.67313161,
+      "epoch": 1.426306443429731,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.085832471863404e-06,
+      "loss": 1.47421494,
+      "memory(GiB)": 123.79,
+      "step": 56225,
+      "train_speed(iter/s)": 1.315253
+    },
+    {
+      "acc": 0.67994127,
+      "epoch": 1.4264332825976662,
+      "grad_norm": 3.25,
+      "learning_rate": 2.08498043316763e-06,
+      "loss": 1.43615799,
+      "memory(GiB)": 123.79,
+      "step": 56230,
+      "train_speed(iter/s)": 1.315265
+    },
+    {
+      "acc": 0.69813461,
+      "epoch": 1.4265601217656012,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.0841285226874282e-06,
+      "loss": 1.38851061,
+      "memory(GiB)": 123.79,
+      "step": 56235,
+      "train_speed(iter/s)": 1.315277
+    },
+    {
+      "acc": 0.68076305,
+      "epoch": 1.4266869609335362,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.083276740460271e-06,
+      "loss": 1.39754324,
+      "memory(GiB)": 123.79,
+      "step": 56240,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.69179816,
+      "epoch": 1.4268138001014714,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.082425086523627e-06,
+      "loss": 1.32111702,
+      "memory(GiB)": 123.79,
+      "step": 56245,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.69082546,
+      "epoch": 1.4269406392694064,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.0815735609149556e-06,
+      "loss": 1.36515236,
+      "memory(GiB)": 123.79,
+      "step": 56250,
+      "train_speed(iter/s)": 1.315315
+    },
+    {
+      "acc": 0.70493507,
+      "epoch": 1.4270674784373414,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.080722163671705e-06,
+      "loss": 1.32584963,
+      "memory(GiB)": 123.79,
+      "step": 56255,
+      "train_speed(iter/s)": 1.315327
+    },
+    {
+      "acc": 0.69564757,
+      "epoch": 1.4271943176052764,
+      "grad_norm": 2.875,
+      "learning_rate": 2.079870894831329e-06,
+      "loss": 1.39469633,
+      "memory(GiB)": 123.79,
+      "step": 56260,
+      "train_speed(iter/s)": 1.31534
+    },
+    {
+      "acc": 0.68533816,
+      "epoch": 1.4273211567732116,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.0790197544312683e-06,
+      "loss": 1.43606968,
+      "memory(GiB)": 123.79,
+      "step": 56265,
+      "train_speed(iter/s)": 1.315353
+    },
+    {
+      "acc": 0.67453065,
+      "epoch": 1.4274479959411466,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.0781687425089624e-06,
+      "loss": 1.43138256,
+      "memory(GiB)": 123.79,
+      "step": 56270,
+      "train_speed(iter/s)": 1.315366
+    },
+    {
+      "acc": 0.68793783,
+      "epoch": 1.4275748351090818,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.0773178591018363e-06,
+      "loss": 1.44580402,
+      "memory(GiB)": 123.79,
+      "step": 56275,
+      "train_speed(iter/s)": 1.31538
+    },
+    {
+      "acc": 0.70265956,
+      "epoch": 1.4277016742770168,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.076467104247322e-06,
+      "loss": 1.30720501,
+      "memory(GiB)": 123.79,
+      "step": 56280,
+      "train_speed(iter/s)": 1.315393
+    },
+    {
+      "acc": 0.70627618,
+      "epoch": 1.4278285134449518,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.0756164779828365e-06,
+      "loss": 1.31342964,
+      "memory(GiB)": 123.79,
+      "step": 56285,
+      "train_speed(iter/s)": 1.315407
+    },
+    {
+      "acc": 0.6844264,
+      "epoch": 1.4279553526128868,
+      "grad_norm": 2.78125,
+      "learning_rate": 2.0747659803457943e-06,
+      "loss": 1.37620525,
+      "memory(GiB)": 123.79,
+      "step": 56290,
+      "train_speed(iter/s)": 1.315419
+    },
+    {
+      "acc": 0.70711613,
+      "epoch": 1.428082191780822,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.073915611373605e-06,
+      "loss": 1.33844366,
+      "memory(GiB)": 123.79,
+      "step": 56295,
+      "train_speed(iter/s)": 1.315433
+    },
+    {
+      "acc": 0.67919407,
+      "epoch": 1.428209030948757,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.0730653711036713e-06,
+      "loss": 1.39809074,
+      "memory(GiB)": 123.79,
+      "step": 56300,
+      "train_speed(iter/s)": 1.315446
+    },
+    {
+      "acc": 0.69662113,
+      "epoch": 1.4283358701166922,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.0722152595733903e-06,
+      "loss": 1.31231432,
+      "memory(GiB)": 123.79,
+      "step": 56305,
+      "train_speed(iter/s)": 1.31546
+    },
+    {
+      "acc": 0.69069309,
+      "epoch": 1.4284627092846272,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.0713652768201536e-06,
+      "loss": 1.33647575,
+      "memory(GiB)": 123.79,
+      "step": 56310,
+      "train_speed(iter/s)": 1.315473
+    },
+    {
+      "acc": 0.6859262,
+      "epoch": 1.4285895484525621,
+      "grad_norm": 3.984375,
+      "learning_rate": 2.0705154228813477e-06,
+      "loss": 1.40588608,
+      "memory(GiB)": 123.79,
+      "step": 56315,
+      "train_speed(iter/s)": 1.315486
+    },
+    {
+      "acc": 0.69828053,
+      "epoch": 1.4287163876204971,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.0696656977943524e-06,
+      "loss": 1.32580709,
+      "memory(GiB)": 123.79,
+      "step": 56320,
+      "train_speed(iter/s)": 1.315499
+    },
+    {
+      "acc": 0.67769527,
+      "epoch": 1.4288432267884323,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.068816101596543e-06,
+      "loss": 1.42083035,
+      "memory(GiB)": 123.79,
+      "step": 56325,
+      "train_speed(iter/s)": 1.315512
+    },
+    {
+      "acc": 0.698557,
+      "epoch": 1.4289700659563673,
+      "grad_norm": 2.75,
+      "learning_rate": 2.067966634325288e-06,
+      "loss": 1.33289394,
+      "memory(GiB)": 123.79,
+      "step": 56330,
+      "train_speed(iter/s)": 1.315525
+    },
+    {
+      "acc": 0.69514837,
+      "epoch": 1.4290969051243023,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.0671172960179513e-06,
+      "loss": 1.34271355,
+      "memory(GiB)": 123.79,
+      "step": 56335,
+      "train_speed(iter/s)": 1.315538
+    },
+    {
+      "acc": 0.6988657,
+      "epoch": 1.4292237442922375,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.06626808671189e-06,
+      "loss": 1.33208113,
+      "memory(GiB)": 123.79,
+      "step": 56340,
+      "train_speed(iter/s)": 1.315552
+    },
+    {
+      "acc": 0.68682871,
+      "epoch": 1.4293505834601725,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.065419006444455e-06,
+      "loss": 1.40413589,
+      "memory(GiB)": 123.79,
+      "step": 56345,
+      "train_speed(iter/s)": 1.315565
+    },
+    {
+      "acc": 0.68875155,
+      "epoch": 1.4294774226281075,
+      "grad_norm": 3.125,
+      "learning_rate": 2.0645700552529973e-06,
+      "loss": 1.38585386,
+      "memory(GiB)": 123.79,
+      "step": 56350,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.66994915,
+      "epoch": 1.4296042617960425,
+      "grad_norm": 3.0,
+      "learning_rate": 2.063721233174853e-06,
+      "loss": 1.47853823,
+      "memory(GiB)": 123.79,
+      "step": 56355,
+      "train_speed(iter/s)": 1.315591
+    },
+    {
+      "acc": 0.69769754,
+      "epoch": 1.4297311009639777,
+      "grad_norm": 3.78125,
+      "learning_rate": 2.0628725402473574e-06,
+      "loss": 1.30150547,
+      "memory(GiB)": 123.79,
+      "step": 56360,
+      "train_speed(iter/s)": 1.315603
+    },
+    {
+      "acc": 0.69646292,
+      "epoch": 1.4298579401319127,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.0620239765078404e-06,
+      "loss": 1.34693413,
+      "memory(GiB)": 123.79,
+      "step": 56365,
+      "train_speed(iter/s)": 1.315616
+    },
+    {
+      "acc": 0.68626776,
+      "epoch": 1.429984779299848,
+      "grad_norm": 2.875,
+      "learning_rate": 2.0611755419936285e-06,
+      "loss": 1.36744261,
+      "memory(GiB)": 123.79,
+      "step": 56370,
+      "train_speed(iter/s)": 1.315628
+    },
+    {
+      "acc": 0.69223332,
+      "epoch": 1.430111618467783,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.0603272367420357e-06,
+      "loss": 1.36039095,
+      "memory(GiB)": 123.79,
+      "step": 56375,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.69559431,
+      "epoch": 1.4302384576357179,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.0594790607903743e-06,
+      "loss": 1.33712673,
+      "memory(GiB)": 123.79,
+      "step": 56380,
+      "train_speed(iter/s)": 1.315654
+    },
+    {
+      "acc": 0.6931406,
+      "epoch": 1.4303652968036529,
+      "grad_norm": 2.875,
+      "learning_rate": 2.0586310141759534e-06,
+      "loss": 1.37845879,
+      "memory(GiB)": 123.79,
+      "step": 56385,
+      "train_speed(iter/s)": 1.315665
+    },
+    {
+      "acc": 0.68254747,
+      "epoch": 1.430492135971588,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.0577830969360753e-06,
+      "loss": 1.38873043,
+      "memory(GiB)": 123.79,
+      "step": 56390,
+      "train_speed(iter/s)": 1.315677
+    },
+    {
+      "acc": 0.70205979,
+      "epoch": 1.430618975139523,
+      "grad_norm": 3.921875,
+      "learning_rate": 2.0569353091080304e-06,
+      "loss": 1.34064789,
+      "memory(GiB)": 123.79,
+      "step": 56395,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.70278139,
+      "epoch": 1.430745814307458,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.056087650729109e-06,
+      "loss": 1.29965916,
+      "memory(GiB)": 123.79,
+      "step": 56400,
+      "train_speed(iter/s)": 1.315702
+    },
+    {
+      "acc": 0.68543429,
+      "epoch": 1.4308726534753933,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.0552401218365975e-06,
+      "loss": 1.41001015,
+      "memory(GiB)": 123.79,
+      "step": 56405,
+      "train_speed(iter/s)": 1.315715
+    },
+    {
+      "acc": 0.68284173,
+      "epoch": 1.4309994926433283,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.054392722467775e-06,
+      "loss": 1.3691309,
+      "memory(GiB)": 123.79,
+      "step": 56410,
+      "train_speed(iter/s)": 1.315727
+    },
+    {
+      "acc": 0.66762352,
+      "epoch": 1.4311263318112633,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.0535454526599086e-06,
+      "loss": 1.47854443,
+      "memory(GiB)": 123.79,
+      "step": 56415,
+      "train_speed(iter/s)": 1.31574
+    },
+    {
+      "acc": 0.69336329,
+      "epoch": 1.4312531709791982,
+      "grad_norm": 3.625,
+      "learning_rate": 2.0526983124502692e-06,
+      "loss": 1.35788031,
+      "memory(GiB)": 123.79,
+      "step": 56420,
+      "train_speed(iter/s)": 1.315751
+    },
+    {
+      "acc": 0.67710848,
+      "epoch": 1.4313800101471335,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.051851301876117e-06,
+      "loss": 1.43713436,
+      "memory(GiB)": 123.79,
+      "step": 56425,
+      "train_speed(iter/s)": 1.315764
+    },
+    {
+      "acc": 0.67997437,
+      "epoch": 1.4315068493150684,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.0510044209747078e-06,
+      "loss": 1.4465167,
+      "memory(GiB)": 123.79,
+      "step": 56430,
+      "train_speed(iter/s)": 1.315776
+    },
+    {
+      "acc": 0.68861885,
+      "epoch": 1.4316336884830037,
+      "grad_norm": 3.125,
+      "learning_rate": 2.05015766978329e-06,
+      "loss": 1.38084869,
+      "memory(GiB)": 123.79,
+      "step": 56435,
+      "train_speed(iter/s)": 1.315789
+    },
+    {
+      "acc": 0.67464466,
+      "epoch": 1.4317605276509386,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.049311048339108e-06,
+      "loss": 1.44504204,
+      "memory(GiB)": 123.79,
+      "step": 56440,
+      "train_speed(iter/s)": 1.315802
+    },
+    {
+      "acc": 0.68964686,
+      "epoch": 1.4318873668188736,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.0484645566793993e-06,
+      "loss": 1.40608673,
+      "memory(GiB)": 123.79,
+      "step": 56445,
+      "train_speed(iter/s)": 1.315814
+    },
+    {
+      "acc": 0.67502213,
+      "epoch": 1.4320142059868086,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.0476181948413975e-06,
+      "loss": 1.47962008,
+      "memory(GiB)": 123.79,
+      "step": 56450,
+      "train_speed(iter/s)": 1.315827
+    },
+    {
+      "acc": 0.68484941,
+      "epoch": 1.4321410451547438,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.0467719628623293e-06,
+      "loss": 1.3801589,
+      "memory(GiB)": 123.79,
+      "step": 56455,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.70481157,
+      "epoch": 1.4322678843226788,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.045925860779415e-06,
+      "loss": 1.32817392,
+      "memory(GiB)": 123.79,
+      "step": 56460,
+      "train_speed(iter/s)": 1.315852
+    },
+    {
+      "acc": 0.67882547,
+      "epoch": 1.432394723490614,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.0450798886298707e-06,
+      "loss": 1.39329853,
+      "memory(GiB)": 123.79,
+      "step": 56465,
+      "train_speed(iter/s)": 1.315864
+    },
+    {
+      "acc": 0.68963509,
+      "epoch": 1.432521562658549,
+      "grad_norm": 3.0,
+      "learning_rate": 2.044234046450905e-06,
+      "loss": 1.36542683,
+      "memory(GiB)": 123.79,
+      "step": 56470,
+      "train_speed(iter/s)": 1.315875
+    },
+    {
+      "acc": 0.69090862,
+      "epoch": 1.432648401826484,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.0433883342797233e-06,
+      "loss": 1.35337238,
+      "memory(GiB)": 123.79,
+      "step": 56475,
+      "train_speed(iter/s)": 1.315887
+    },
+    {
+      "acc": 0.70057287,
+      "epoch": 1.432775240994419,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.042542752153522e-06,
+      "loss": 1.3559351,
+      "memory(GiB)": 123.79,
+      "step": 56480,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.68819256,
+      "epoch": 1.4329020801623542,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.0416973001094953e-06,
+      "loss": 1.40177364,
+      "memory(GiB)": 123.79,
+      "step": 56485,
+      "train_speed(iter/s)": 1.315913
+    },
+    {
+      "acc": 0.6917563,
+      "epoch": 1.4330289193302892,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.0408519781848292e-06,
+      "loss": 1.38428564,
+      "memory(GiB)": 123.79,
+      "step": 56490,
+      "train_speed(iter/s)": 1.315925
+    },
+    {
+      "acc": 0.7012043,
+      "epoch": 1.4331557584982242,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.0400067864167044e-06,
+      "loss": 1.32241879,
+      "memory(GiB)": 123.79,
+      "step": 56495,
+      "train_speed(iter/s)": 1.315938
+    },
+    {
+      "acc": 0.70065584,
+      "epoch": 1.4332825976661594,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.0391617248422967e-06,
+      "loss": 1.3419899,
+      "memory(GiB)": 123.79,
+      "step": 56500,
+      "train_speed(iter/s)": 1.315951
+    },
+    {
+      "acc": 0.67479792,
+      "epoch": 1.4334094368340944,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.038316793498774e-06,
+      "loss": 1.46780357,
+      "memory(GiB)": 123.79,
+      "step": 56505,
+      "train_speed(iter/s)": 1.315963
+    },
+    {
+      "acc": 0.68784151,
+      "epoch": 1.4335362760020294,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.037471992423305e-06,
+      "loss": 1.38478851,
+      "memory(GiB)": 123.79,
+      "step": 56510,
+      "train_speed(iter/s)": 1.315977
+    },
+    {
+      "acc": 0.66595473,
+      "epoch": 1.4336631151699644,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.036627321653043e-06,
+      "loss": 1.45258751,
+      "memory(GiB)": 123.79,
+      "step": 56515,
+      "train_speed(iter/s)": 1.31599
+    },
+    {
+      "acc": 0.67895803,
+      "epoch": 1.4337899543378996,
+      "grad_norm": 3.625,
+      "learning_rate": 2.0357827812251403e-06,
+      "loss": 1.45129623,
+      "memory(GiB)": 123.79,
+      "step": 56520,
+      "train_speed(iter/s)": 1.316003
+    },
+    {
+      "acc": 0.6873631,
+      "epoch": 1.4339167935058346,
+      "grad_norm": 3.75,
+      "learning_rate": 2.0349383711767463e-06,
+      "loss": 1.39454432,
+      "memory(GiB)": 123.79,
+      "step": 56525,
+      "train_speed(iter/s)": 1.316016
+    },
+    {
+      "acc": 0.68704901,
+      "epoch": 1.4340436326737698,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.0340940915450026e-06,
+      "loss": 1.38195906,
+      "memory(GiB)": 123.79,
+      "step": 56530,
+      "train_speed(iter/s)": 1.31603
+    },
+    {
+      "acc": 0.68277631,
+      "epoch": 1.4341704718417048,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.033249942367041e-06,
+      "loss": 1.41304798,
+      "memory(GiB)": 123.79,
+      "step": 56535,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.69371047,
+      "epoch": 1.4342973110096398,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.032405923679991e-06,
+      "loss": 1.3344595,
+      "memory(GiB)": 123.79,
+      "step": 56540,
+      "train_speed(iter/s)": 1.316057
+    },
+    {
+      "acc": 0.68504848,
+      "epoch": 1.4344241501775747,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.0315620355209792e-06,
+      "loss": 1.4354557,
+      "memory(GiB)": 123.79,
+      "step": 56545,
+      "train_speed(iter/s)": 1.31607
+    },
+    {
+      "acc": 0.68748541,
+      "epoch": 1.43455098934551,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.030718277927124e-06,
+      "loss": 1.40265036,
+      "memory(GiB)": 123.79,
+      "step": 56550,
+      "train_speed(iter/s)": 1.316084
+    },
+    {
+      "acc": 0.69896355,
+      "epoch": 1.434677828513445,
+      "grad_norm": 3.125,
+      "learning_rate": 2.0298746509355326e-06,
+      "loss": 1.32718201,
+      "memory(GiB)": 123.79,
+      "step": 56555,
+      "train_speed(iter/s)": 1.316097
+    },
+    {
+      "acc": 0.69303346,
+      "epoch": 1.43480466768138,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.0290311545833166e-06,
+      "loss": 1.37437458,
+      "memory(GiB)": 123.79,
+      "step": 56560,
+      "train_speed(iter/s)": 1.31611
+    },
+    {
+      "acc": 0.701929,
+      "epoch": 1.4349315068493151,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.028187788907574e-06,
+      "loss": 1.39725838,
+      "memory(GiB)": 123.79,
+      "step": 56565,
+      "train_speed(iter/s)": 1.316123
+    },
+    {
+      "acc": 0.69769697,
+      "epoch": 1.4350583460172501,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.027344553945403e-06,
+      "loss": 1.40010452,
+      "memory(GiB)": 123.79,
+      "step": 56570,
+      "train_speed(iter/s)": 1.316134
+    },
+    {
+      "acc": 0.68795843,
+      "epoch": 1.4351851851851851,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.0265014497338868e-06,
+      "loss": 1.35107965,
+      "memory(GiB)": 123.79,
+      "step": 56575,
+      "train_speed(iter/s)": 1.316147
+    },
+    {
+      "acc": 0.68033509,
+      "epoch": 1.43531202435312,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.0256584763101145e-06,
+      "loss": 1.39008236,
+      "memory(GiB)": 123.79,
+      "step": 56580,
+      "train_speed(iter/s)": 1.31616
+    },
+    {
+      "acc": 0.68048062,
+      "epoch": 1.4354388635210553,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.024815633711162e-06,
+      "loss": 1.43385048,
+      "memory(GiB)": 123.79,
+      "step": 56585,
+      "train_speed(iter/s)": 1.316174
+    },
+    {
+      "acc": 0.69271193,
+      "epoch": 1.4355657026889903,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.0239729219741005e-06,
+      "loss": 1.34979515,
+      "memory(GiB)": 123.79,
+      "step": 56590,
+      "train_speed(iter/s)": 1.316186
+    },
+    {
+      "acc": 0.68689718,
+      "epoch": 1.4356925418569255,
+      "grad_norm": 3.703125,
+      "learning_rate": 2.0231303411359975e-06,
+      "loss": 1.36359854,
+      "memory(GiB)": 123.79,
+      "step": 56595,
+      "train_speed(iter/s)": 1.316199
+    },
+    {
+      "acc": 0.67811532,
+      "epoch": 1.4358193810248605,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.0222878912339127e-06,
+      "loss": 1.42331038,
+      "memory(GiB)": 123.79,
+      "step": 56600,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.69340229,
+      "epoch": 1.4359462201927955,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.021445572304901e-06,
+      "loss": 1.41115484,
+      "memory(GiB)": 123.79,
+      "step": 56605,
+      "train_speed(iter/s)": 1.316227
+    },
+    {
+      "acc": 0.69363289,
+      "epoch": 1.4360730593607305,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.0206033843860113e-06,
+      "loss": 1.34889069,
+      "memory(GiB)": 123.79,
+      "step": 56610,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.68590288,
+      "epoch": 1.4361998985286657,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.0197613275142868e-06,
+      "loss": 1.40086384,
+      "memory(GiB)": 123.79,
+      "step": 56615,
+      "train_speed(iter/s)": 1.316253
+    },
+    {
+      "acc": 0.69039731,
+      "epoch": 1.4363267376966007,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.018919401726765e-06,
+      "loss": 1.34919233,
+      "memory(GiB)": 123.79,
+      "step": 56620,
+      "train_speed(iter/s)": 1.316265
+    },
+    {
+      "acc": 0.69021645,
+      "epoch": 1.436453576864536,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.0180776070604773e-06,
+      "loss": 1.39609652,
+      "memory(GiB)": 123.79,
+      "step": 56625,
+      "train_speed(iter/s)": 1.316278
+    },
+    {
+      "acc": 0.69178438,
+      "epoch": 1.4365804160324709,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.0172359435524497e-06,
+      "loss": 1.42033138,
+      "memory(GiB)": 123.79,
+      "step": 56630,
+      "train_speed(iter/s)": 1.316292
+    },
+    {
+      "acc": 0.68785734,
+      "epoch": 1.4367072552004059,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.0163944112397027e-06,
+      "loss": 1.35907545,
+      "memory(GiB)": 123.79,
+      "step": 56635,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.68058376,
+      "epoch": 1.4368340943683409,
+      "grad_norm": 2.875,
+      "learning_rate": 2.01555301015925e-06,
+      "loss": 1.43751125,
+      "memory(GiB)": 123.79,
+      "step": 56640,
+      "train_speed(iter/s)": 1.316318
+    },
+    {
+      "acc": 0.68261042,
+      "epoch": 1.436960933536276,
+      "grad_norm": 2.625,
+      "learning_rate": 2.0147117403480994e-06,
+      "loss": 1.39809179,
+      "memory(GiB)": 123.79,
+      "step": 56645,
+      "train_speed(iter/s)": 1.316331
+    },
+    {
+      "acc": 0.68378,
+      "epoch": 1.437087772704211,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.0138706018432576e-06,
+      "loss": 1.35066376,
+      "memory(GiB)": 123.79,
+      "step": 56650,
+      "train_speed(iter/s)": 1.316345
+    },
+    {
+      "acc": 0.67546363,
+      "epoch": 1.437214611872146,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.0130295946817176e-06,
+      "loss": 1.42086163,
+      "memory(GiB)": 123.79,
+      "step": 56655,
+      "train_speed(iter/s)": 1.316358
+    },
+    {
+      "acc": 0.69671659,
+      "epoch": 1.4373414510400813,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.0121887189004713e-06,
+      "loss": 1.36788168,
+      "memory(GiB)": 123.79,
+      "step": 56660,
+      "train_speed(iter/s)": 1.316371
+    },
+    {
+      "acc": 0.69084625,
+      "epoch": 1.4374682902080163,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.0113479745365033e-06,
+      "loss": 1.33711491,
+      "memory(GiB)": 123.79,
+      "step": 56665,
+      "train_speed(iter/s)": 1.316384
+    },
+    {
+      "acc": 0.68301001,
+      "epoch": 1.4375951293759512,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.0105073616267984e-06,
+      "loss": 1.42443991,
+      "memory(GiB)": 123.79,
+      "step": 56670,
+      "train_speed(iter/s)": 1.316395
+    },
+    {
+      "acc": 0.69364576,
+      "epoch": 1.4377219685438862,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.0096668802083254e-06,
+      "loss": 1.41494036,
+      "memory(GiB)": 123.79,
+      "step": 56675,
+      "train_speed(iter/s)": 1.316408
+    },
+    {
+      "acc": 0.68823729,
+      "epoch": 1.4378488077118214,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.0088265303180516e-06,
+      "loss": 1.371766,
+      "memory(GiB)": 123.79,
+      "step": 56680,
+      "train_speed(iter/s)": 1.31642
+    },
+    {
+      "acc": 0.68078117,
+      "epoch": 1.4379756468797564,
+      "grad_norm": 3.75,
+      "learning_rate": 2.0079863119929434e-06,
+      "loss": 1.39833374,
+      "memory(GiB)": 123.79,
+      "step": 56685,
+      "train_speed(iter/s)": 1.316433
+    },
+    {
+      "acc": 0.68295393,
+      "epoch": 1.4381024860476916,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.0071462252699575e-06,
+      "loss": 1.45416632,
+      "memory(GiB)": 123.79,
+      "step": 56690,
+      "train_speed(iter/s)": 1.316446
+    },
+    {
+      "acc": 0.68118815,
+      "epoch": 1.4382293252156266,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.006306270186039e-06,
+      "loss": 1.37279587,
+      "memory(GiB)": 123.79,
+      "step": 56695,
+      "train_speed(iter/s)": 1.316459
+    },
+    {
+      "acc": 0.68964081,
+      "epoch": 1.4383561643835616,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.005466446778139e-06,
+      "loss": 1.36573095,
+      "memory(GiB)": 123.79,
+      "step": 56700,
+      "train_speed(iter/s)": 1.316473
+    },
+    {
+      "acc": 0.67946949,
+      "epoch": 1.4384830035514966,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.0046267550831935e-06,
+      "loss": 1.45852671,
+      "memory(GiB)": 123.79,
+      "step": 56705,
+      "train_speed(iter/s)": 1.316486
+    },
+    {
+      "acc": 0.68756881,
+      "epoch": 1.4386098427194318,
+      "grad_norm": 2.875,
+      "learning_rate": 2.003787195138139e-06,
+      "loss": 1.4250906,
+      "memory(GiB)": 123.79,
+      "step": 56710,
+      "train_speed(iter/s)": 1.316499
+    },
+    {
+      "acc": 0.69040422,
+      "epoch": 1.4387366818873668,
+      "grad_norm": 2.875,
+      "learning_rate": 2.002947766979897e-06,
+      "loss": 1.42117233,
+      "memory(GiB)": 123.79,
+      "step": 56715,
+      "train_speed(iter/s)": 1.316512
+    },
+    {
+      "acc": 0.69125805,
+      "epoch": 1.4388635210553018,
+      "grad_norm": 2.875,
+      "learning_rate": 2.0021084706453945e-06,
+      "loss": 1.34775639,
+      "memory(GiB)": 123.79,
+      "step": 56720,
+      "train_speed(iter/s)": 1.316525
+    },
+    {
+      "acc": 0.67574077,
+      "epoch": 1.438990360223237,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.0012693061715467e-06,
+      "loss": 1.36498308,
+      "memory(GiB)": 123.79,
+      "step": 56725,
+      "train_speed(iter/s)": 1.316537
+    },
+    {
+      "acc": 0.70298629,
+      "epoch": 1.439117199391172,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.000430273595263e-06,
+      "loss": 1.30545378,
+      "memory(GiB)": 123.79,
+      "step": 56730,
+      "train_speed(iter/s)": 1.316551
+    },
+    {
+      "acc": 0.67798681,
+      "epoch": 1.439244038559107,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.9995913729534477e-06,
+      "loss": 1.39635696,
+      "memory(GiB)": 123.79,
+      "step": 56735,
+      "train_speed(iter/s)": 1.316563
+    },
+    {
+      "acc": 0.67595954,
+      "epoch": 1.439370877727042,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.9987526042830003e-06,
+      "loss": 1.42818832,
+      "memory(GiB)": 123.79,
+      "step": 56740,
+      "train_speed(iter/s)": 1.316575
+    },
+    {
+      "acc": 0.69536891,
+      "epoch": 1.4394977168949772,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.9979139676208124e-06,
+      "loss": 1.343964,
+      "memory(GiB)": 123.79,
+      "step": 56745,
+      "train_speed(iter/s)": 1.316587
+    },
+    {
+      "acc": 0.68635058,
+      "epoch": 1.4396245560629122,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.9970754630037718e-06,
+      "loss": 1.3667551,
+      "memory(GiB)": 123.79,
+      "step": 56750,
+      "train_speed(iter/s)": 1.316599
+    },
+    {
+      "acc": 0.69635839,
+      "epoch": 1.4397513952308474,
+      "grad_norm": 3.0,
+      "learning_rate": 1.9962370904687596e-06,
+      "loss": 1.34213676,
+      "memory(GiB)": 123.79,
+      "step": 56755,
+      "train_speed(iter/s)": 1.316612
+    },
+    {
+      "acc": 0.68098869,
+      "epoch": 1.4398782343987824,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.9953988500526506e-06,
+      "loss": 1.43444071,
+      "memory(GiB)": 123.79,
+      "step": 56760,
+      "train_speed(iter/s)": 1.316623
+    },
+    {
+      "acc": 0.69219551,
+      "epoch": 1.4400050735667174,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.994560741792315e-06,
+      "loss": 1.35293846,
+      "memory(GiB)": 123.79,
+      "step": 56765,
+      "train_speed(iter/s)": 1.316636
+    },
+    {
+      "acc": 0.68179908,
+      "epoch": 1.4401319127346524,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.993722765724616e-06,
+      "loss": 1.38785172,
+      "memory(GiB)": 123.79,
+      "step": 56770,
+      "train_speed(iter/s)": 1.316648
+    },
+    {
+      "acc": 0.67779026,
+      "epoch": 1.4402587519025876,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.992884921886412e-06,
+      "loss": 1.42823792,
+      "memory(GiB)": 123.79,
+      "step": 56775,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.68696733,
+      "epoch": 1.4403855910705226,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.9920472103145555e-06,
+      "loss": 1.43358879,
+      "memory(GiB)": 123.79,
+      "step": 56780,
+      "train_speed(iter/s)": 1.316673
+    },
+    {
+      "acc": 0.68511839,
+      "epoch": 1.4405124302384578,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.99120963104589e-06,
+      "loss": 1.4093298,
+      "memory(GiB)": 123.79,
+      "step": 56785,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.67902727,
+      "epoch": 1.4406392694063928,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.990372184117262e-06,
+      "loss": 1.4150569,
+      "memory(GiB)": 123.79,
+      "step": 56790,
+      "train_speed(iter/s)": 1.316696
+    },
+    {
+      "acc": 0.68060074,
+      "epoch": 1.4407661085743277,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.9895348695655e-06,
+      "loss": 1.39152374,
+      "memory(GiB)": 123.79,
+      "step": 56795,
+      "train_speed(iter/s)": 1.316709
+    },
+    {
+      "acc": 0.70440121,
+      "epoch": 1.4408929477422627,
+      "grad_norm": 2.875,
+      "learning_rate": 1.9886976874274356e-06,
+      "loss": 1.34280148,
+      "memory(GiB)": 123.79,
+      "step": 56800,
+      "train_speed(iter/s)": 1.31672
+    },
+    {
+      "acc": 0.70200882,
+      "epoch": 1.441019786910198,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.9878606377398895e-06,
+      "loss": 1.30283194,
+      "memory(GiB)": 123.79,
+      "step": 56805,
+      "train_speed(iter/s)": 1.316733
+    },
+    {
+      "acc": 0.69288526,
+      "epoch": 1.441146626078133,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.9870237205396844e-06,
+      "loss": 1.37074175,
+      "memory(GiB)": 123.79,
+      "step": 56810,
+      "train_speed(iter/s)": 1.316745
+    },
+    {
+      "acc": 0.68684573,
+      "epoch": 1.441273465246068,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.986186935863626e-06,
+      "loss": 1.41196823,
+      "memory(GiB)": 123.79,
+      "step": 56815,
+      "train_speed(iter/s)": 1.316757
+    },
+    {
+      "acc": 0.6889132,
+      "epoch": 1.4414003044140031,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.9853502837485207e-06,
+      "loss": 1.37585649,
+      "memory(GiB)": 123.79,
+      "step": 56820,
+      "train_speed(iter/s)": 1.31677
+    },
+    {
+      "acc": 0.68282785,
+      "epoch": 1.4415271435819381,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.9845137642311707e-06,
+      "loss": 1.36629744,
+      "memory(GiB)": 123.79,
+      "step": 56825,
+      "train_speed(iter/s)": 1.316782
+    },
+    {
+      "acc": 0.69874005,
+      "epoch": 1.441653982749873,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.9836773773483704e-06,
+      "loss": 1.37961226,
+      "memory(GiB)": 123.79,
+      "step": 56830,
+      "train_speed(iter/s)": 1.316795
+    },
+    {
+      "acc": 0.68272839,
+      "epoch": 1.441780821917808,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.982841123136904e-06,
+      "loss": 1.42041531,
+      "memory(GiB)": 123.79,
+      "step": 56835,
+      "train_speed(iter/s)": 1.316808
+    },
+    {
+      "acc": 0.67549734,
+      "epoch": 1.4419076610857433,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.982005001633554e-06,
+      "loss": 1.40445099,
+      "memory(GiB)": 123.79,
+      "step": 56840,
+      "train_speed(iter/s)": 1.316821
+    },
+    {
+      "acc": 0.66338854,
+      "epoch": 1.4420345002536783,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.9811690128751002e-06,
+      "loss": 1.48631697,
+      "memory(GiB)": 123.79,
+      "step": 56845,
+      "train_speed(iter/s)": 1.316834
+    },
+    {
+      "acc": 0.68429189,
+      "epoch": 1.4421613394216135,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.980333156898313e-06,
+      "loss": 1.36965675,
+      "memory(GiB)": 123.79,
+      "step": 56850,
+      "train_speed(iter/s)": 1.316847
+    },
+    {
+      "acc": 0.67905774,
+      "epoch": 1.4422881785895485,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.979497433739952e-06,
+      "loss": 1.43855839,
+      "memory(GiB)": 123.79,
+      "step": 56855,
+      "train_speed(iter/s)": 1.31686
+    },
+    {
+      "acc": 0.68890128,
+      "epoch": 1.4424150177574835,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.9786618434367814e-06,
+      "loss": 1.33654995,
+      "memory(GiB)": 123.79,
+      "step": 56860,
+      "train_speed(iter/s)": 1.316873
+    },
+    {
+      "acc": 0.6880806,
+      "epoch": 1.4425418569254185,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.977826386025552e-06,
+      "loss": 1.38885231,
+      "memory(GiB)": 123.79,
+      "step": 56865,
+      "train_speed(iter/s)": 1.316886
+    },
+    {
+      "acc": 0.68436718,
+      "epoch": 1.4426686960933537,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.976991061543011e-06,
+      "loss": 1.39064016,
+      "memory(GiB)": 123.79,
+      "step": 56870,
+      "train_speed(iter/s)": 1.316899
+    },
+    {
+      "acc": 0.67241707,
+      "epoch": 1.4427955352612887,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.9761558700259e-06,
+      "loss": 1.42863197,
+      "memory(GiB)": 123.79,
+      "step": 56875,
+      "train_speed(iter/s)": 1.316913
+    },
+    {
+      "acc": 0.69514427,
+      "epoch": 1.4429223744292237,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.9753208115109546e-06,
+      "loss": 1.35109215,
+      "memory(GiB)": 123.79,
+      "step": 56880,
+      "train_speed(iter/s)": 1.316926
+    },
+    {
+      "acc": 0.68590479,
+      "epoch": 1.4430492135971589,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.9744858860349043e-06,
+      "loss": 1.38762722,
+      "memory(GiB)": 123.79,
+      "step": 56885,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.68659544,
+      "epoch": 1.4431760527650939,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.9736510936344723e-06,
+      "loss": 1.3865901,
+      "memory(GiB)": 123.79,
+      "step": 56890,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.67498264,
+      "epoch": 1.4433028919330289,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.9728164343463764e-06,
+      "loss": 1.47298164,
+      "memory(GiB)": 123.79,
+      "step": 56895,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.70337863,
+      "epoch": 1.4434297311009638,
+      "grad_norm": 4.125,
+      "learning_rate": 1.97198190820733e-06,
+      "loss": 1.37073345,
+      "memory(GiB)": 123.79,
+      "step": 56900,
+      "train_speed(iter/s)": 1.31698
+    },
+    {
+      "acc": 0.67462468,
+      "epoch": 1.443556570268899,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.9711475152540376e-06,
+      "loss": 1.42453804,
+      "memory(GiB)": 123.79,
+      "step": 56905,
+      "train_speed(iter/s)": 1.316992
+    },
+    {
+      "acc": 0.67922125,
+      "epoch": 1.443683409436834,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.9703132555232007e-06,
+      "loss": 1.45563059,
+      "memory(GiB)": 123.79,
+      "step": 56910,
+      "train_speed(iter/s)": 1.317004
+    },
+    {
+      "acc": 0.7043314,
+      "epoch": 1.4438102486047693,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.9694791290515135e-06,
+      "loss": 1.36247559,
+      "memory(GiB)": 123.79,
+      "step": 56915,
+      "train_speed(iter/s)": 1.317017
+    },
+    {
+      "acc": 0.68744335,
+      "epoch": 1.4439370877727042,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.968645135875665e-06,
+      "loss": 1.39996786,
+      "memory(GiB)": 123.79,
+      "step": 56920,
+      "train_speed(iter/s)": 1.31703
+    },
+    {
+      "acc": 0.6822753,
+      "epoch": 1.4440639269406392,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.967811276032335e-06,
+      "loss": 1.37250118,
+      "memory(GiB)": 123.79,
+      "step": 56925,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.68438368,
+      "epoch": 1.4441907661085742,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.966977549558206e-06,
+      "loss": 1.42994766,
+      "memory(GiB)": 123.79,
+      "step": 56930,
+      "train_speed(iter/s)": 1.317056
+    },
+    {
+      "acc": 0.69770203,
+      "epoch": 1.4443176052765094,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.966143956489945e-06,
+      "loss": 1.36827335,
+      "memory(GiB)": 123.79,
+      "step": 56935,
+      "train_speed(iter/s)": 1.317067
+    },
+    {
+      "acc": 0.70564895,
+      "epoch": 1.4444444444444444,
+      "grad_norm": 3.25,
+      "learning_rate": 1.965310496864217e-06,
+      "loss": 1.39932461,
+      "memory(GiB)": 123.79,
+      "step": 56940,
+      "train_speed(iter/s)": 1.317079
+    },
+    {
+      "acc": 0.69877801,
+      "epoch": 1.4445712836123796,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.9644771707176813e-06,
+      "loss": 1.36516962,
+      "memory(GiB)": 123.79,
+      "step": 56945,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.68554134,
+      "epoch": 1.4446981227803146,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.963643978086996e-06,
+      "loss": 1.38411674,
+      "memory(GiB)": 123.79,
+      "step": 56950,
+      "train_speed(iter/s)": 1.317104
+    },
+    {
+      "acc": 0.6880096,
+      "epoch": 1.4448249619482496,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.9628109190088023e-06,
+      "loss": 1.3762701,
+      "memory(GiB)": 123.79,
+      "step": 56955,
+      "train_speed(iter/s)": 1.317115
+    },
+    {
+      "acc": 0.68125029,
+      "epoch": 1.4449518011161846,
+      "grad_norm": 3.625,
+      "learning_rate": 1.961977993519743e-06,
+      "loss": 1.42890091,
+      "memory(GiB)": 123.79,
+      "step": 56960,
+      "train_speed(iter/s)": 1.317127
+    },
+    {
+      "acc": 0.69093924,
+      "epoch": 1.4450786402841198,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.9611452016564574e-06,
+      "loss": 1.3472662,
+      "memory(GiB)": 123.79,
+      "step": 56965,
+      "train_speed(iter/s)": 1.31714
+    },
+    {
+      "acc": 0.68049655,
+      "epoch": 1.4452054794520548,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.960312543455575e-06,
+      "loss": 1.4447134,
+      "memory(GiB)": 123.79,
+      "step": 56970,
+      "train_speed(iter/s)": 1.317152
+    },
+    {
+      "acc": 0.69243336,
+      "epoch": 1.4453323186199898,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.959480018953716e-06,
+      "loss": 1.40292511,
+      "memory(GiB)": 123.79,
+      "step": 56975,
+      "train_speed(iter/s)": 1.317164
+    },
+    {
+      "acc": 0.67275305,
+      "epoch": 1.445459157787925,
+      "grad_norm": 2.265625,
+      "learning_rate": 1.9586476281874994e-06,
+      "loss": 1.42718649,
+      "memory(GiB)": 123.79,
+      "step": 56980,
+      "train_speed(iter/s)": 1.317176
+    },
+    {
+      "acc": 0.6709754,
+      "epoch": 1.44558599695586,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.9578153711935403e-06,
+      "loss": 1.42617664,
+      "memory(GiB)": 123.79,
+      "step": 56985,
+      "train_speed(iter/s)": 1.317188
+    },
+    {
+      "acc": 0.67229891,
+      "epoch": 1.445712836123795,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.9569832480084456e-06,
+      "loss": 1.42912588,
+      "memory(GiB)": 123.79,
+      "step": 56990,
+      "train_speed(iter/s)": 1.3172
+    },
+    {
+      "acc": 0.67918305,
+      "epoch": 1.44583967529173,
+      "grad_norm": 4.375,
+      "learning_rate": 1.9561512586688096e-06,
+      "loss": 1.46116905,
+      "memory(GiB)": 123.79,
+      "step": 56995,
+      "train_speed(iter/s)": 1.317212
+    },
+    {
+      "acc": 0.68618793,
+      "epoch": 1.4459665144596652,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.9553194032112334e-06,
+      "loss": 1.36507778,
+      "memory(GiB)": 123.79,
+      "step": 57000,
+      "train_speed(iter/s)": 1.317225
+    },
+    {
+      "epoch": 1.4459665144596652,
+      "eval_acc": 0.6750290796626133,
+      "eval_loss": 1.3563456535339355,
+      "eval_runtime": 70.0135,
+      "eval_samples_per_second": 90.982,
+      "eval_steps_per_second": 22.753,
+      "step": 57000
+    },
+    {
+      "acc": 0.68794193,
+      "epoch": 1.4460933536276002,
+      "grad_norm": 5.625,
+      "learning_rate": 1.954487681672303e-06,
+      "loss": 1.42324524,
+      "memory(GiB)": 123.79,
+      "step": 57005,
+      "train_speed(iter/s)": 1.314726
+    },
+    {
+      "acc": 0.69964771,
+      "epoch": 1.4462201927955354,
+      "grad_norm": 4.625,
+      "learning_rate": 1.9536560940886033e-06,
+      "loss": 1.33212814,
+      "memory(GiB)": 123.79,
+      "step": 57010,
+      "train_speed(iter/s)": 1.314739
+    },
+    {
+      "acc": 0.66939087,
+      "epoch": 1.4463470319634704,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.9528246404967067e-06,
+      "loss": 1.45567398,
+      "memory(GiB)": 123.79,
+      "step": 57015,
+      "train_speed(iter/s)": 1.31475
+    },
+    {
+      "acc": 0.69179029,
+      "epoch": 1.4464738711314054,
+      "grad_norm": 3.375,
+      "learning_rate": 1.951993320933188e-06,
+      "loss": 1.40269165,
+      "memory(GiB)": 123.79,
+      "step": 57020,
+      "train_speed(iter/s)": 1.314762
+    },
+    {
+      "acc": 0.68126345,
+      "epoch": 1.4466007102993403,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.951162135434612e-06,
+      "loss": 1.45903664,
+      "memory(GiB)": 123.79,
+      "step": 57025,
+      "train_speed(iter/s)": 1.314774
+    },
+    {
+      "acc": 0.68826637,
+      "epoch": 1.4467275494672756,
+      "grad_norm": 3.25,
+      "learning_rate": 1.9503310840375374e-06,
+      "loss": 1.40639515,
+      "memory(GiB)": 123.79,
+      "step": 57030,
+      "train_speed(iter/s)": 1.314786
+    },
+    {
+      "acc": 0.68363829,
+      "epoch": 1.4468543886352105,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.949500166778517e-06,
+      "loss": 1.45758762,
+      "memory(GiB)": 123.79,
+      "step": 57035,
+      "train_speed(iter/s)": 1.314797
+    },
+    {
+      "acc": 0.68508596,
+      "epoch": 1.4469812278031455,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.948669383694099e-06,
+      "loss": 1.39431133,
+      "memory(GiB)": 123.79,
+      "step": 57040,
+      "train_speed(iter/s)": 1.314809
+    },
+    {
+      "acc": 0.69482508,
+      "epoch": 1.4471080669710807,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.947838734820825e-06,
+      "loss": 1.34236507,
+      "memory(GiB)": 123.79,
+      "step": 57045,
+      "train_speed(iter/s)": 1.314821
+    },
+    {
+      "acc": 0.69293823,
+      "epoch": 1.4472349061390157,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.94700822019523e-06,
+      "loss": 1.35296173,
+      "memory(GiB)": 123.79,
+      "step": 57050,
+      "train_speed(iter/s)": 1.314834
+    },
+    {
+      "acc": 0.70235276,
+      "epoch": 1.4473617453069507,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.9461778398538447e-06,
+      "loss": 1.32942448,
+      "memory(GiB)": 123.79,
+      "step": 57055,
+      "train_speed(iter/s)": 1.314845
+    },
+    {
+      "acc": 0.66412845,
+      "epoch": 1.4474885844748857,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.945347593833191e-06,
+      "loss": 1.53387232,
+      "memory(GiB)": 123.79,
+      "step": 57060,
+      "train_speed(iter/s)": 1.314858
+    },
+    {
+      "acc": 0.67160521,
+      "epoch": 1.447615423642821,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.9445174821697893e-06,
+      "loss": 1.44834929,
+      "memory(GiB)": 123.79,
+      "step": 57065,
+      "train_speed(iter/s)": 1.31487
+    },
+    {
+      "acc": 0.66813498,
+      "epoch": 1.447742262810756,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.94368750490015e-06,
+      "loss": 1.47262001,
+      "memory(GiB)": 123.79,
+      "step": 57070,
+      "train_speed(iter/s)": 1.314882
+    },
+    {
+      "acc": 0.70264115,
+      "epoch": 1.4478691019786911,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.94285766206078e-06,
+      "loss": 1.33286304,
+      "memory(GiB)": 123.79,
+      "step": 57075,
+      "train_speed(iter/s)": 1.314894
+    },
+    {
+      "acc": 0.68421803,
+      "epoch": 1.447995941146626,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.9420279536881794e-06,
+      "loss": 1.43054867,
+      "memory(GiB)": 123.79,
+      "step": 57080,
+      "train_speed(iter/s)": 1.314906
+    },
+    {
+      "acc": 0.68853998,
+      "epoch": 1.448122780314561,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.9411983798188398e-06,
+      "loss": 1.42471523,
+      "memory(GiB)": 123.79,
+      "step": 57085,
+      "train_speed(iter/s)": 1.314918
+    },
+    {
+      "acc": 0.67992749,
+      "epoch": 1.448249619482496,
+      "grad_norm": 3.25,
+      "learning_rate": 1.940368940489256e-06,
+      "loss": 1.41729727,
+      "memory(GiB)": 123.79,
+      "step": 57090,
+      "train_speed(iter/s)": 1.314931
+    },
+    {
+      "acc": 0.69322472,
+      "epoch": 1.4483764586504313,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.939539635735905e-06,
+      "loss": 1.36589737,
+      "memory(GiB)": 123.79,
+      "step": 57095,
+      "train_speed(iter/s)": 1.314943
+    },
+    {
+      "acc": 0.69992046,
+      "epoch": 1.4485032978183663,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.9387104655952625e-06,
+      "loss": 1.38805676,
+      "memory(GiB)": 123.79,
+      "step": 57100,
+      "train_speed(iter/s)": 1.31495
+    },
+    {
+      "acc": 0.69014854,
+      "epoch": 1.4486301369863015,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.9378814301038033e-06,
+      "loss": 1.36952744,
+      "memory(GiB)": 123.79,
+      "step": 57105,
+      "train_speed(iter/s)": 1.314957
+    },
+    {
+      "acc": 0.6879745,
+      "epoch": 1.4487569761542365,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.937052529297992e-06,
+      "loss": 1.36667652,
+      "memory(GiB)": 123.79,
+      "step": 57110,
+      "train_speed(iter/s)": 1.31497
+    },
+    {
+      "acc": 0.70201249,
+      "epoch": 1.4488838153221715,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.9362237632142838e-06,
+      "loss": 1.33841209,
+      "memory(GiB)": 123.79,
+      "step": 57115,
+      "train_speed(iter/s)": 1.314982
+    },
+    {
+      "acc": 0.69836845,
+      "epoch": 1.4490106544901065,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.9353951318891313e-06,
+      "loss": 1.3228076,
+      "memory(GiB)": 123.79,
+      "step": 57120,
+      "train_speed(iter/s)": 1.314994
+    },
+    {
+      "acc": 0.70003653,
+      "epoch": 1.4491374936580417,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.9345666353589855e-06,
+      "loss": 1.37789783,
+      "memory(GiB)": 123.79,
+      "step": 57125,
+      "train_speed(iter/s)": 1.315006
+    },
+    {
+      "acc": 0.69475837,
+      "epoch": 1.4492643328259767,
+      "grad_norm": 3.5,
+      "learning_rate": 1.9337382736602868e-06,
+      "loss": 1.36138868,
+      "memory(GiB)": 123.79,
+      "step": 57130,
+      "train_speed(iter/s)": 1.315018
+    },
+    {
+      "acc": 0.68786802,
+      "epoch": 1.4493911719939117,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.9329100468294646e-06,
+      "loss": 1.38458748,
+      "memory(GiB)": 123.79,
+      "step": 57135,
+      "train_speed(iter/s)": 1.315031
+    },
+    {
+      "acc": 0.69765806,
+      "epoch": 1.4495180111618469,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.9320819549029546e-06,
+      "loss": 1.33804016,
+      "memory(GiB)": 123.79,
+      "step": 57140,
+      "train_speed(iter/s)": 1.315043
+    },
+    {
+      "acc": 0.68845091,
+      "epoch": 1.4496448503297819,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.9312539979171774e-06,
+      "loss": 1.38652391,
+      "memory(GiB)": 123.79,
+      "step": 57145,
+      "train_speed(iter/s)": 1.315055
+    },
+    {
+      "acc": 0.67772913,
+      "epoch": 1.4497716894977168,
+      "grad_norm": 2.75,
+      "learning_rate": 1.9304261759085525e-06,
+      "loss": 1.35769157,
+      "memory(GiB)": 123.79,
+      "step": 57150,
+      "train_speed(iter/s)": 1.315068
+    },
+    {
+      "acc": 0.6945919,
+      "epoch": 1.4498985286656518,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.929598488913485e-06,
+      "loss": 1.38397808,
+      "memory(GiB)": 123.79,
+      "step": 57155,
+      "train_speed(iter/s)": 1.31508
+    },
+    {
+      "acc": 0.69414439,
+      "epoch": 1.450025367833587,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.928770936968386e-06,
+      "loss": 1.33165092,
+      "memory(GiB)": 123.79,
+      "step": 57160,
+      "train_speed(iter/s)": 1.315092
+    },
+    {
+      "acc": 0.70022516,
+      "epoch": 1.450152207001522,
+      "grad_norm": 2.75,
+      "learning_rate": 1.927943520109653e-06,
+      "loss": 1.32276058,
+      "memory(GiB)": 123.79,
+      "step": 57165,
+      "train_speed(iter/s)": 1.315105
+    },
+    {
+      "acc": 0.68299737,
+      "epoch": 1.4502790461694572,
+      "grad_norm": 3.5,
+      "learning_rate": 1.9271162383736804e-06,
+      "loss": 1.4510478,
+      "memory(GiB)": 123.79,
+      "step": 57170,
+      "train_speed(iter/s)": 1.315118
+    },
+    {
+      "acc": 0.69441099,
+      "epoch": 1.4504058853373922,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.9262890917968547e-06,
+      "loss": 1.35344887,
+      "memory(GiB)": 123.79,
+      "step": 57175,
+      "train_speed(iter/s)": 1.31513
+    },
+    {
+      "acc": 0.67387266,
+      "epoch": 1.4505327245053272,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.925462080415558e-06,
+      "loss": 1.50443373,
+      "memory(GiB)": 123.79,
+      "step": 57180,
+      "train_speed(iter/s)": 1.315142
+    },
+    {
+      "acc": 0.67738886,
+      "epoch": 1.4506595636732622,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.924635204266166e-06,
+      "loss": 1.42984982,
+      "memory(GiB)": 123.79,
+      "step": 57185,
+      "train_speed(iter/s)": 1.315155
+    },
+    {
+      "acc": 0.69710956,
+      "epoch": 1.4507864028411974,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.923808463385048e-06,
+      "loss": 1.35394258,
+      "memory(GiB)": 123.79,
+      "step": 57190,
+      "train_speed(iter/s)": 1.315167
+    },
+    {
+      "acc": 0.68162036,
+      "epoch": 1.4509132420091324,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.922981857808568e-06,
+      "loss": 1.42086697,
+      "memory(GiB)": 123.79,
+      "step": 57195,
+      "train_speed(iter/s)": 1.31518
+    },
+    {
+      "acc": 0.68031702,
+      "epoch": 1.4510400811770674,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.9221553875730835e-06,
+      "loss": 1.43268404,
+      "memory(GiB)": 123.79,
+      "step": 57200,
+      "train_speed(iter/s)": 1.315192
+    },
+    {
+      "acc": 0.68975925,
+      "epoch": 1.4511669203450026,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.921329052714947e-06,
+      "loss": 1.39440813,
+      "memory(GiB)": 123.79,
+      "step": 57205,
+      "train_speed(iter/s)": 1.315205
+    },
+    {
+      "acc": 0.69489598,
+      "epoch": 1.4512937595129376,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.920502853270504e-06,
+      "loss": 1.37176228,
+      "memory(GiB)": 123.79,
+      "step": 57210,
+      "train_speed(iter/s)": 1.315218
+    },
+    {
+      "acc": 0.68603406,
+      "epoch": 1.4514205986808726,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.919676789276094e-06,
+      "loss": 1.42088013,
+      "memory(GiB)": 123.79,
+      "step": 57215,
+      "train_speed(iter/s)": 1.31523
+    },
+    {
+      "acc": 0.67656441,
+      "epoch": 1.4515474378488076,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.918850860768052e-06,
+      "loss": 1.46715508,
+      "memory(GiB)": 123.79,
+      "step": 57220,
+      "train_speed(iter/s)": 1.315242
+    },
+    {
+      "acc": 0.69128709,
+      "epoch": 1.4516742770167428,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.918025067782704e-06,
+      "loss": 1.36423655,
+      "memory(GiB)": 123.79,
+      "step": 57225,
+      "train_speed(iter/s)": 1.315255
+    },
+    {
+      "acc": 0.68562007,
+      "epoch": 1.4518011161846778,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.9171994103563766e-06,
+      "loss": 1.42567539,
+      "memory(GiB)": 123.79,
+      "step": 57230,
+      "train_speed(iter/s)": 1.315267
+    },
+    {
+      "acc": 0.67126589,
+      "epoch": 1.451927955352613,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.916373888525381e-06,
+      "loss": 1.49060497,
+      "memory(GiB)": 123.79,
+      "step": 57235,
+      "train_speed(iter/s)": 1.315279
+    },
+    {
+      "acc": 0.67765007,
+      "epoch": 1.452054794520548,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.9155485023260294e-06,
+      "loss": 1.45217943,
+      "memory(GiB)": 123.79,
+      "step": 57240,
+      "train_speed(iter/s)": 1.315292
+    },
+    {
+      "acc": 0.68461237,
+      "epoch": 1.452181633688483,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.914723251794624e-06,
+      "loss": 1.36785622,
+      "memory(GiB)": 123.79,
+      "step": 57245,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.68733668,
+      "epoch": 1.452308472856418,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.9138981369674688e-06,
+      "loss": 1.43235884,
+      "memory(GiB)": 123.79,
+      "step": 57250,
+      "train_speed(iter/s)": 1.315313
+    },
+    {
+      "acc": 0.67525229,
+      "epoch": 1.4524353120243532,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.9130731578808493e-06,
+      "loss": 1.42986021,
+      "memory(GiB)": 123.79,
+      "step": 57255,
+      "train_speed(iter/s)": 1.315325
+    },
+    {
+      "acc": 0.67946701,
+      "epoch": 1.4525621511922882,
+      "grad_norm": 3.0,
+      "learning_rate": 1.912248314571053e-06,
+      "loss": 1.42953463,
+      "memory(GiB)": 123.79,
+      "step": 57260,
+      "train_speed(iter/s)": 1.315337
+    },
+    {
+      "acc": 0.69849453,
+      "epoch": 1.4526889903602234,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.9114236070743638e-06,
+      "loss": 1.38655453,
+      "memory(GiB)": 123.79,
+      "step": 57265,
+      "train_speed(iter/s)": 1.315349
+    },
+    {
+      "acc": 0.69248729,
+      "epoch": 1.4528158295281584,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.910599035427055e-06,
+      "loss": 1.40347595,
+      "memory(GiB)": 123.79,
+      "step": 57270,
+      "train_speed(iter/s)": 1.315362
+    },
+    {
+      "acc": 0.686835,
+      "epoch": 1.4529426686960933,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.909774599665392e-06,
+      "loss": 1.35267067,
+      "memory(GiB)": 123.79,
+      "step": 57275,
+      "train_speed(iter/s)": 1.315374
+    },
+    {
+      "acc": 0.70180464,
+      "epoch": 1.4530695078640283,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.9089502998256382e-06,
+      "loss": 1.32218513,
+      "memory(GiB)": 123.79,
+      "step": 57280,
+      "train_speed(iter/s)": 1.315386
+    },
+    {
+      "acc": 0.68872976,
+      "epoch": 1.4531963470319635,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.9081261359440517e-06,
+      "loss": 1.36705122,
+      "memory(GiB)": 123.79,
+      "step": 57285,
+      "train_speed(iter/s)": 1.315399
+    },
+    {
+      "acc": 0.70146008,
+      "epoch": 1.4533231861998985,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.9073021080568837e-06,
+      "loss": 1.3604167,
+      "memory(GiB)": 123.79,
+      "step": 57290,
+      "train_speed(iter/s)": 1.315411
+    },
+    {
+      "acc": 0.69120836,
+      "epoch": 1.4534500253678335,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.9064782162003737e-06,
+      "loss": 1.34389801,
+      "memory(GiB)": 123.79,
+      "step": 57295,
+      "train_speed(iter/s)": 1.315424
+    },
+    {
+      "acc": 0.69099512,
+      "epoch": 1.4535768645357687,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.9056544604107646e-06,
+      "loss": 1.34862251,
+      "memory(GiB)": 123.79,
+      "step": 57300,
+      "train_speed(iter/s)": 1.315437
+    },
+    {
+      "acc": 0.68793211,
+      "epoch": 1.4537037037037037,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.9048308407242882e-06,
+      "loss": 1.36873665,
+      "memory(GiB)": 123.79,
+      "step": 57305,
+      "train_speed(iter/s)": 1.315449
+    },
+    {
+      "acc": 0.68205276,
+      "epoch": 1.4538305428716387,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.90400735717717e-06,
+      "loss": 1.40787268,
+      "memory(GiB)": 123.79,
+      "step": 57310,
+      "train_speed(iter/s)": 1.315462
+    },
+    {
+      "acc": 0.67832298,
+      "epoch": 1.4539573820395737,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.903184009805631e-06,
+      "loss": 1.37006731,
+      "memory(GiB)": 123.79,
+      "step": 57315,
+      "train_speed(iter/s)": 1.315475
+    },
+    {
+      "acc": 0.68783588,
+      "epoch": 1.454084221207509,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.9023607986458854e-06,
+      "loss": 1.47817593,
+      "memory(GiB)": 123.79,
+      "step": 57320,
+      "train_speed(iter/s)": 1.315487
+    },
+    {
+      "acc": 0.68986869,
+      "epoch": 1.454211060375444,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.901537723734142e-06,
+      "loss": 1.39967194,
+      "memory(GiB)": 123.79,
+      "step": 57325,
+      "train_speed(iter/s)": 1.3155
+    },
+    {
+      "acc": 0.69353933,
+      "epoch": 1.454337899543379,
+      "grad_norm": 2.75,
+      "learning_rate": 1.9007147851066031e-06,
+      "loss": 1.36269064,
+      "memory(GiB)": 123.79,
+      "step": 57330,
+      "train_speed(iter/s)": 1.315512
+    },
+    {
+      "acc": 0.70087299,
+      "epoch": 1.454464738711314,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.8998919827994654e-06,
+      "loss": 1.29457941,
+      "memory(GiB)": 123.79,
+      "step": 57335,
+      "train_speed(iter/s)": 1.315525
+    },
+    {
+      "acc": 0.6884655,
+      "epoch": 1.454591577879249,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.899069316848919e-06,
+      "loss": 1.44257908,
+      "memory(GiB)": 123.79,
+      "step": 57340,
+      "train_speed(iter/s)": 1.315538
+    },
+    {
+      "acc": 0.68611913,
+      "epoch": 1.454718417047184,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.8982467872911486e-06,
+      "loss": 1.47468605,
+      "memory(GiB)": 123.79,
+      "step": 57345,
+      "train_speed(iter/s)": 1.315551
+    },
+    {
+      "acc": 0.67890615,
+      "epoch": 1.4548452562151193,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.8974243941623332e-06,
+      "loss": 1.4044693,
+      "memory(GiB)": 123.79,
+      "step": 57350,
+      "train_speed(iter/s)": 1.315563
+    },
+    {
+      "acc": 0.67954712,
+      "epoch": 1.4549720953830543,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.896602137498645e-06,
+      "loss": 1.45514479,
+      "memory(GiB)": 123.79,
+      "step": 57355,
+      "train_speed(iter/s)": 1.315576
+    },
+    {
+      "acc": 0.68269591,
+      "epoch": 1.4550989345509893,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.89578001733625e-06,
+      "loss": 1.39449911,
+      "memory(GiB)": 123.79,
+      "step": 57360,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.69552183,
+      "epoch": 1.4552257737189245,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.8949580337113078e-06,
+      "loss": 1.35517063,
+      "memory(GiB)": 123.79,
+      "step": 57365,
+      "train_speed(iter/s)": 1.3156
+    },
+    {
+      "acc": 0.67332034,
+      "epoch": 1.4553526128868595,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8941361866599778e-06,
+      "loss": 1.4503581,
+      "memory(GiB)": 123.79,
+      "step": 57370,
+      "train_speed(iter/s)": 1.315613
+    },
+    {
+      "acc": 0.68073115,
+      "epoch": 1.4554794520547945,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.893314476218403e-06,
+      "loss": 1.42302551,
+      "memory(GiB)": 123.79,
+      "step": 57375,
+      "train_speed(iter/s)": 1.315626
+    },
+    {
+      "acc": 0.68324132,
+      "epoch": 1.4556062912227294,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.8924929024227279e-06,
+      "loss": 1.39651699,
+      "memory(GiB)": 123.79,
+      "step": 57380,
+      "train_speed(iter/s)": 1.315638
+    },
+    {
+      "acc": 0.67413263,
+      "epoch": 1.4557331303906647,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8916714653090874e-06,
+      "loss": 1.44993877,
+      "memory(GiB)": 123.79,
+      "step": 57385,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.68669882,
+      "epoch": 1.4558599695585996,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.8908501649136174e-06,
+      "loss": 1.41366625,
+      "memory(GiB)": 123.79,
+      "step": 57390,
+      "train_speed(iter/s)": 1.315664
+    },
+    {
+      "acc": 0.6968864,
+      "epoch": 1.4559868087265349,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.8900290012724358e-06,
+      "loss": 1.37081041,
+      "memory(GiB)": 123.79,
+      "step": 57395,
+      "train_speed(iter/s)": 1.315676
+    },
+    {
+      "acc": 0.69567995,
+      "epoch": 1.4561136478944698,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.889207974421663e-06,
+      "loss": 1.33034391,
+      "memory(GiB)": 123.79,
+      "step": 57400,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.70479898,
+      "epoch": 1.4562404870624048,
+      "grad_norm": 7.40625,
+      "learning_rate": 1.8883870843974134e-06,
+      "loss": 1.32671061,
+      "memory(GiB)": 123.79,
+      "step": 57405,
+      "train_speed(iter/s)": 1.315701
+    },
+    {
+      "acc": 0.6690753,
+      "epoch": 1.4563673262303398,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.887566331235794e-06,
+      "loss": 1.49780951,
+      "memory(GiB)": 123.79,
+      "step": 57410,
+      "train_speed(iter/s)": 1.315713
+    },
+    {
+      "acc": 0.66768632,
+      "epoch": 1.456494165398275,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.8867457149729013e-06,
+      "loss": 1.54160461,
+      "memory(GiB)": 123.79,
+      "step": 57415,
+      "train_speed(iter/s)": 1.315726
+    },
+    {
+      "acc": 0.68487382,
+      "epoch": 1.45662100456621,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.8859252356448305e-06,
+      "loss": 1.38348904,
+      "memory(GiB)": 123.79,
+      "step": 57420,
+      "train_speed(iter/s)": 1.315739
+    },
+    {
+      "acc": 0.70302081,
+      "epoch": 1.4567478437341452,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.8851048932876725e-06,
+      "loss": 1.30460777,
+      "memory(GiB)": 123.79,
+      "step": 57425,
+      "train_speed(iter/s)": 1.315752
+    },
+    {
+      "acc": 0.69793711,
+      "epoch": 1.4568746829020802,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.8842846879375104e-06,
+      "loss": 1.35926428,
+      "memory(GiB)": 123.79,
+      "step": 57430,
+      "train_speed(iter/s)": 1.315764
+    },
+    {
+      "acc": 0.69595108,
+      "epoch": 1.4570015220700152,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.8834646196304146e-06,
+      "loss": 1.37831106,
+      "memory(GiB)": 123.79,
+      "step": 57435,
+      "train_speed(iter/s)": 1.315777
+    },
+    {
+      "acc": 0.6865696,
+      "epoch": 1.4571283612379502,
+      "grad_norm": 3.625,
+      "learning_rate": 1.8826446884024612e-06,
+      "loss": 1.45076056,
+      "memory(GiB)": 123.79,
+      "step": 57440,
+      "train_speed(iter/s)": 1.31579
+    },
+    {
+      "acc": 0.69050136,
+      "epoch": 1.4572552004058854,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.8818248942897122e-06,
+      "loss": 1.35059633,
+      "memory(GiB)": 123.79,
+      "step": 57445,
+      "train_speed(iter/s)": 1.315802
+    },
+    {
+      "acc": 0.68570948,
+      "epoch": 1.4573820395738204,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.8810052373282277e-06,
+      "loss": 1.40898342,
+      "memory(GiB)": 123.79,
+      "step": 57450,
+      "train_speed(iter/s)": 1.315815
+    },
+    {
+      "acc": 0.68400316,
+      "epoch": 1.4575088787417554,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.880185717554055e-06,
+      "loss": 1.40821419,
+      "memory(GiB)": 123.79,
+      "step": 57455,
+      "train_speed(iter/s)": 1.315828
+    },
+    {
+      "acc": 0.67784252,
+      "epoch": 1.4576357179096906,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.879366335003245e-06,
+      "loss": 1.43079643,
+      "memory(GiB)": 123.79,
+      "step": 57460,
+      "train_speed(iter/s)": 1.31584
+    },
+    {
+      "acc": 0.68329268,
+      "epoch": 1.4577625570776256,
+      "grad_norm": 3.625,
+      "learning_rate": 1.8785470897118362e-06,
+      "loss": 1.45201387,
+      "memory(GiB)": 123.79,
+      "step": 57465,
+      "train_speed(iter/s)": 1.315853
+    },
+    {
+      "acc": 0.68759327,
+      "epoch": 1.4578893962455606,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.8777279817158627e-06,
+      "loss": 1.39294052,
+      "memory(GiB)": 123.79,
+      "step": 57470,
+      "train_speed(iter/s)": 1.315866
+    },
+    {
+      "acc": 0.69006543,
+      "epoch": 1.4580162354134956,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.8769090110513522e-06,
+      "loss": 1.39172935,
+      "memory(GiB)": 123.79,
+      "step": 57475,
+      "train_speed(iter/s)": 1.315879
+    },
+    {
+      "acc": 0.68086338,
+      "epoch": 1.4581430745814308,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8760901777543273e-06,
+      "loss": 1.45338249,
+      "memory(GiB)": 123.79,
+      "step": 57480,
+      "train_speed(iter/s)": 1.315891
+    },
+    {
+      "acc": 0.69535894,
+      "epoch": 1.4582699137493658,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.8752714818608036e-06,
+      "loss": 1.37339697,
+      "memory(GiB)": 123.79,
+      "step": 57485,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.69520831,
+      "epoch": 1.458396752917301,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.874452923406791e-06,
+      "loss": 1.37544565,
+      "memory(GiB)": 123.79,
+      "step": 57490,
+      "train_speed(iter/s)": 1.315916
+    },
+    {
+      "acc": 0.68040056,
+      "epoch": 1.458523592085236,
+      "grad_norm": 3.0,
+      "learning_rate": 1.8736345024282937e-06,
+      "loss": 1.42991867,
+      "memory(GiB)": 123.79,
+      "step": 57495,
+      "train_speed(iter/s)": 1.315929
+    },
+    {
+      "acc": 0.69230981,
+      "epoch": 1.458650431253171,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.8728162189613085e-06,
+      "loss": 1.34908371,
+      "memory(GiB)": 123.79,
+      "step": 57500,
+      "train_speed(iter/s)": 1.315942
+    },
+    {
+      "acc": 0.68684893,
+      "epoch": 1.458777270421106,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.8719980730418285e-06,
+      "loss": 1.41192169,
+      "memory(GiB)": 123.79,
+      "step": 57505,
+      "train_speed(iter/s)": 1.315954
+    },
+    {
+      "acc": 0.67966232,
+      "epoch": 1.4589041095890412,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.8711800647058388e-06,
+      "loss": 1.40708618,
+      "memory(GiB)": 123.79,
+      "step": 57510,
+      "train_speed(iter/s)": 1.315967
+    },
+    {
+      "acc": 0.67618809,
+      "epoch": 1.4590309487569761,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.8703621939893185e-06,
+      "loss": 1.47074375,
+      "memory(GiB)": 123.79,
+      "step": 57515,
+      "train_speed(iter/s)": 1.31598
+    },
+    {
+      "acc": 0.68443532,
+      "epoch": 1.4591577879249111,
+      "grad_norm": 3.75,
+      "learning_rate": 1.869544460928242e-06,
+      "loss": 1.39242973,
+      "memory(GiB)": 123.79,
+      "step": 57520,
+      "train_speed(iter/s)": 1.315992
+    },
+    {
+      "acc": 0.68361464,
+      "epoch": 1.4592846270928463,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.868726865558575e-06,
+      "loss": 1.36734304,
+      "memory(GiB)": 123.79,
+      "step": 57525,
+      "train_speed(iter/s)": 1.316004
+    },
+    {
+      "acc": 0.68469729,
+      "epoch": 1.4594114662607813,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.8679094079162835e-06,
+      "loss": 1.35315037,
+      "memory(GiB)": 123.79,
+      "step": 57530,
+      "train_speed(iter/s)": 1.316016
+    },
+    {
+      "acc": 0.68516493,
+      "epoch": 1.4595383054287163,
+      "grad_norm": 4.0,
+      "learning_rate": 1.867092088037319e-06,
+      "loss": 1.36746702,
+      "memory(GiB)": 123.79,
+      "step": 57535,
+      "train_speed(iter/s)": 1.316028
+    },
+    {
+      "acc": 0.68060274,
+      "epoch": 1.4596651445966513,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.8662749059576296e-06,
+      "loss": 1.45541162,
+      "memory(GiB)": 123.79,
+      "step": 57540,
+      "train_speed(iter/s)": 1.316041
+    },
+    {
+      "acc": 0.69139919,
+      "epoch": 1.4597919837645865,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.865457861713163e-06,
+      "loss": 1.41777334,
+      "memory(GiB)": 123.79,
+      "step": 57545,
+      "train_speed(iter/s)": 1.316053
+    },
+    {
+      "acc": 0.70135183,
+      "epoch": 1.4599188229325215,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.8646409553398558e-06,
+      "loss": 1.34324055,
+      "memory(GiB)": 123.79,
+      "step": 57550,
+      "train_speed(iter/s)": 1.316064
+    },
+    {
+      "acc": 0.69162951,
+      "epoch": 1.4600456621004567,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.8638241868736367e-06,
+      "loss": 1.36769714,
+      "memory(GiB)": 123.79,
+      "step": 57555,
+      "train_speed(iter/s)": 1.316076
+    },
+    {
+      "acc": 0.68650732,
+      "epoch": 1.4601725012683917,
+      "grad_norm": 3.25,
+      "learning_rate": 1.8630075563504297e-06,
+      "loss": 1.44228306,
+      "memory(GiB)": 123.79,
+      "step": 57560,
+      "train_speed(iter/s)": 1.316089
+    },
+    {
+      "acc": 0.68059049,
+      "epoch": 1.4602993404363267,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.8621910638061575e-06,
+      "loss": 1.40740604,
+      "memory(GiB)": 123.79,
+      "step": 57565,
+      "train_speed(iter/s)": 1.316101
+    },
+    {
+      "acc": 0.67546182,
+      "epoch": 1.4604261796042617,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.8613747092767336e-06,
+      "loss": 1.40688448,
+      "memory(GiB)": 123.79,
+      "step": 57570,
+      "train_speed(iter/s)": 1.316113
+    },
+    {
+      "acc": 0.69042635,
+      "epoch": 1.460553018772197,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.8605584927980596e-06,
+      "loss": 1.38448181,
+      "memory(GiB)": 123.79,
+      "step": 57575,
+      "train_speed(iter/s)": 1.316126
+    },
+    {
+      "acc": 0.68460388,
+      "epoch": 1.4606798579401319,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.859742414406041e-06,
+      "loss": 1.33330498,
+      "memory(GiB)": 123.79,
+      "step": 57580,
+      "train_speed(iter/s)": 1.316138
+    },
+    {
+      "acc": 0.70708551,
+      "epoch": 1.460806697108067,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.8589264741365714e-06,
+      "loss": 1.24759693,
+      "memory(GiB)": 123.79,
+      "step": 57585,
+      "train_speed(iter/s)": 1.316151
+    },
+    {
+      "acc": 0.67682123,
+      "epoch": 1.460933536276002,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.8581106720255414e-06,
+      "loss": 1.46971912,
+      "memory(GiB)": 123.79,
+      "step": 57590,
+      "train_speed(iter/s)": 1.316164
+    },
+    {
+      "acc": 0.68954511,
+      "epoch": 1.461060375443937,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.8572950081088282e-06,
+      "loss": 1.36725407,
+      "memory(GiB)": 123.79,
+      "step": 57595,
+      "train_speed(iter/s)": 1.316176
+    },
+    {
+      "acc": 0.6803226,
+      "epoch": 1.461187214611872,
+      "grad_norm": 3.375,
+      "learning_rate": 1.856479482422313e-06,
+      "loss": 1.3834568,
+      "memory(GiB)": 123.79,
+      "step": 57600,
+      "train_speed(iter/s)": 1.316189
+    },
+    {
+      "acc": 0.69411411,
+      "epoch": 1.4613140537798073,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.8556640950018651e-06,
+      "loss": 1.37719603,
+      "memory(GiB)": 123.79,
+      "step": 57605,
+      "train_speed(iter/s)": 1.316201
+    },
+    {
+      "acc": 0.7083962,
+      "epoch": 1.4614408929477423,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.8548488458833485e-06,
+      "loss": 1.34739714,
+      "memory(GiB)": 123.79,
+      "step": 57610,
+      "train_speed(iter/s)": 1.316212
+    },
+    {
+      "acc": 0.69408965,
+      "epoch": 1.4615677321156773,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.854033735102622e-06,
+      "loss": 1.39351273,
+      "memory(GiB)": 123.79,
+      "step": 57615,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.69085884,
+      "epoch": 1.4616945712836125,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.8532187626955377e-06,
+      "loss": 1.41985168,
+      "memory(GiB)": 123.79,
+      "step": 57620,
+      "train_speed(iter/s)": 1.316238
+    },
+    {
+      "acc": 0.6954896,
+      "epoch": 1.4618214104515475,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.8524039286979417e-06,
+      "loss": 1.4107645,
+      "memory(GiB)": 123.79,
+      "step": 57625,
+      "train_speed(iter/s)": 1.316251
+    },
+    {
+      "acc": 0.69435682,
+      "epoch": 1.4619482496194824,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.8515892331456736e-06,
+      "loss": 1.40683107,
+      "memory(GiB)": 123.79,
+      "step": 57630,
+      "train_speed(iter/s)": 1.316264
+    },
+    {
+      "acc": 0.71962419,
+      "epoch": 1.4620750887874174,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.850774676074568e-06,
+      "loss": 1.25426636,
+      "memory(GiB)": 123.79,
+      "step": 57635,
+      "train_speed(iter/s)": 1.316275
+    },
+    {
+      "acc": 0.67962246,
+      "epoch": 1.4622019279553526,
+      "grad_norm": 3.0,
+      "learning_rate": 1.8499602575204522e-06,
+      "loss": 1.43975782,
+      "memory(GiB)": 123.79,
+      "step": 57640,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.6809021,
+      "epoch": 1.4623287671232876,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.8491459775191484e-06,
+      "loss": 1.36690817,
+      "memory(GiB)": 123.79,
+      "step": 57645,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.6717144,
+      "epoch": 1.4624556062912228,
+      "grad_norm": 3.375,
+      "learning_rate": 1.8483318361064716e-06,
+      "loss": 1.42509775,
+      "memory(GiB)": 123.79,
+      "step": 57650,
+      "train_speed(iter/s)": 1.316313
+    },
+    {
+      "acc": 0.69436774,
+      "epoch": 1.4625824454591578,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.847517833318232e-06,
+      "loss": 1.36333447,
+      "memory(GiB)": 123.79,
+      "step": 57655,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.6997931,
+      "epoch": 1.4627092846270928,
+      "grad_norm": 3.125,
+      "learning_rate": 1.8467039691902334e-06,
+      "loss": 1.28080463,
+      "memory(GiB)": 123.79,
+      "step": 57660,
+      "train_speed(iter/s)": 1.316339
+    },
+    {
+      "acc": 0.69749928,
+      "epoch": 1.4628361237950278,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.8458902437582705e-06,
+      "loss": 1.34730196,
+      "memory(GiB)": 123.79,
+      "step": 57665,
+      "train_speed(iter/s)": 1.316352
+    },
+    {
+      "acc": 0.68741083,
+      "epoch": 1.462962962962963,
+      "grad_norm": 2.875,
+      "learning_rate": 1.8450766570581402e-06,
+      "loss": 1.43017292,
+      "memory(GiB)": 123.79,
+      "step": 57670,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.67970443,
+      "epoch": 1.463089802130898,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.8442632091256223e-06,
+      "loss": 1.41679535,
+      "memory(GiB)": 123.79,
+      "step": 57675,
+      "train_speed(iter/s)": 1.316378
+    },
+    {
+      "acc": 0.68606873,
+      "epoch": 1.463216641298833,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.8434498999964983e-06,
+      "loss": 1.39639149,
+      "memory(GiB)": 123.79,
+      "step": 57680,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.70454054,
+      "epoch": 1.4633434804667682,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.8426367297065384e-06,
+      "loss": 1.37764606,
+      "memory(GiB)": 123.79,
+      "step": 57685,
+      "train_speed(iter/s)": 1.316402
+    },
+    {
+      "acc": 0.67822232,
+      "epoch": 1.4634703196347032,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.841823698291516e-06,
+      "loss": 1.43696508,
+      "memory(GiB)": 123.79,
+      "step": 57690,
+      "train_speed(iter/s)": 1.316415
+    },
+    {
+      "acc": 0.6906909,
+      "epoch": 1.4635971588026382,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.8410108057871851e-06,
+      "loss": 1.3494277,
+      "memory(GiB)": 123.79,
+      "step": 57695,
+      "train_speed(iter/s)": 1.316428
+    },
+    {
+      "acc": 0.68901134,
+      "epoch": 1.4637239979705732,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.8401980522293017e-06,
+      "loss": 1.33846455,
+      "memory(GiB)": 123.79,
+      "step": 57700,
+      "train_speed(iter/s)": 1.316441
+    },
+    {
+      "acc": 0.67511935,
+      "epoch": 1.4638508371385084,
+      "grad_norm": 4.5,
+      "learning_rate": 1.839385437653617e-06,
+      "loss": 1.45054874,
+      "memory(GiB)": 123.79,
+      "step": 57705,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.68585954,
+      "epoch": 1.4639776763064434,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.8385729620958731e-06,
+      "loss": 1.44294872,
+      "memory(GiB)": 123.79,
+      "step": 57710,
+      "train_speed(iter/s)": 1.316466
+    },
+    {
+      "acc": 0.69343977,
+      "epoch": 1.4641045154743786,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.8377606255918024e-06,
+      "loss": 1.37515192,
+      "memory(GiB)": 123.79,
+      "step": 57715,
+      "train_speed(iter/s)": 1.316478
+    },
+    {
+      "acc": 0.67560668,
+      "epoch": 1.4642313546423136,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.8369484281771388e-06,
+      "loss": 1.42299175,
+      "memory(GiB)": 123.79,
+      "step": 57720,
+      "train_speed(iter/s)": 1.316491
+    },
+    {
+      "acc": 0.68842678,
+      "epoch": 1.4643581938102486,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.836136369887606e-06,
+      "loss": 1.36110783,
+      "memory(GiB)": 123.79,
+      "step": 57725,
+      "train_speed(iter/s)": 1.316503
+    },
+    {
+      "acc": 0.68565741,
+      "epoch": 1.4644850329781836,
+      "grad_norm": 6.125,
+      "learning_rate": 1.8353244507589225e-06,
+      "loss": 1.37649746,
+      "memory(GiB)": 123.79,
+      "step": 57730,
+      "train_speed(iter/s)": 1.316516
+    },
+    {
+      "acc": 0.68186159,
+      "epoch": 1.4646118721461188,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.8345126708267958e-06,
+      "loss": 1.39665298,
+      "memory(GiB)": 123.79,
+      "step": 57735,
+      "train_speed(iter/s)": 1.316528
+    },
+    {
+      "acc": 0.7008976,
+      "epoch": 1.4647387113140538,
+      "grad_norm": 3.5,
+      "learning_rate": 1.8337010301269364e-06,
+      "loss": 1.33029556,
+      "memory(GiB)": 123.79,
+      "step": 57740,
+      "train_speed(iter/s)": 1.316541
+    },
+    {
+      "acc": 0.68667598,
+      "epoch": 1.464865550481989,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.8328895286950422e-06,
+      "loss": 1.41362,
+      "memory(GiB)": 123.79,
+      "step": 57745,
+      "train_speed(iter/s)": 1.316554
+    },
+    {
+      "acc": 0.68754883,
+      "epoch": 1.464992389649924,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.8320781665668063e-06,
+      "loss": 1.39509993,
+      "memory(GiB)": 123.79,
+      "step": 57750,
+      "train_speed(iter/s)": 1.316567
+    },
+    {
+      "acc": 0.66924181,
+      "epoch": 1.465119228817859,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.8312669437779167e-06,
+      "loss": 1.48519182,
+      "memory(GiB)": 123.79,
+      "step": 57755,
+      "train_speed(iter/s)": 1.316579
+    },
+    {
+      "acc": 0.67615376,
+      "epoch": 1.465246067985794,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.8304558603640544e-06,
+      "loss": 1.43256893,
+      "memory(GiB)": 123.79,
+      "step": 57760,
+      "train_speed(iter/s)": 1.316591
+    },
+    {
+      "acc": 0.68608308,
+      "epoch": 1.4653729071537291,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.8296449163608942e-06,
+      "loss": 1.44096794,
+      "memory(GiB)": 123.79,
+      "step": 57765,
+      "train_speed(iter/s)": 1.316604
+    },
+    {
+      "acc": 0.70007453,
+      "epoch": 1.4654997463216641,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.8288341118041052e-06,
+      "loss": 1.27353477,
+      "memory(GiB)": 123.79,
+      "step": 57770,
+      "train_speed(iter/s)": 1.316617
+    },
+    {
+      "acc": 0.69266081,
+      "epoch": 1.4656265854895991,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.82802344672935e-06,
+      "loss": 1.41287918,
+      "memory(GiB)": 123.79,
+      "step": 57775,
+      "train_speed(iter/s)": 1.31663
+    },
+    {
+      "acc": 0.6766469,
+      "epoch": 1.4657534246575343,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.8272129211722855e-06,
+      "loss": 1.40640745,
+      "memory(GiB)": 123.79,
+      "step": 57780,
+      "train_speed(iter/s)": 1.316642
+    },
+    {
+      "acc": 0.68736601,
+      "epoch": 1.4658802638254693,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.8264025351685627e-06,
+      "loss": 1.42727547,
+      "memory(GiB)": 123.79,
+      "step": 57785,
+      "train_speed(iter/s)": 1.316655
+    },
+    {
+      "acc": 0.69248338,
+      "epoch": 1.4660071029934043,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.8255922887538251e-06,
+      "loss": 1.34697952,
+      "memory(GiB)": 123.79,
+      "step": 57790,
+      "train_speed(iter/s)": 1.316667
+    },
+    {
+      "acc": 0.68914852,
+      "epoch": 1.4661339421613393,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.8247821819637112e-06,
+      "loss": 1.36141891,
+      "memory(GiB)": 123.79,
+      "step": 57795,
+      "train_speed(iter/s)": 1.316681
+    },
+    {
+      "acc": 0.68717585,
+      "epoch": 1.4662607813292745,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.8239722148338534e-06,
+      "loss": 1.32898951,
+      "memory(GiB)": 123.79,
+      "step": 57800,
+      "train_speed(iter/s)": 1.316691
+    },
+    {
+      "acc": 0.67957911,
+      "epoch": 1.4663876204972095,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.823162387399876e-06,
+      "loss": 1.44245663,
+      "memory(GiB)": 123.79,
+      "step": 57805,
+      "train_speed(iter/s)": 1.316703
+    },
+    {
+      "acc": 0.70418954,
+      "epoch": 1.4665144596651447,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.822352699697404e-06,
+      "loss": 1.36191797,
+      "memory(GiB)": 123.79,
+      "step": 57810,
+      "train_speed(iter/s)": 1.316716
+    },
+    {
+      "acc": 0.69962554,
+      "epoch": 1.4666412988330797,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.8215431517620452e-06,
+      "loss": 1.38380985,
+      "memory(GiB)": 123.79,
+      "step": 57815,
+      "train_speed(iter/s)": 1.316728
+    },
+    {
+      "acc": 0.67448583,
+      "epoch": 1.4667681380010147,
+      "grad_norm": 3.25,
+      "learning_rate": 1.8207337436294097e-06,
+      "loss": 1.5256609,
+      "memory(GiB)": 123.79,
+      "step": 57820,
+      "train_speed(iter/s)": 1.316741
+    },
+    {
+      "acc": 0.6807044,
+      "epoch": 1.4668949771689497,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.819924475335097e-06,
+      "loss": 1.40160027,
+      "memory(GiB)": 123.79,
+      "step": 57825,
+      "train_speed(iter/s)": 1.316753
+    },
+    {
+      "acc": 0.68244801,
+      "epoch": 1.4670218163368849,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.8191153469147065e-06,
+      "loss": 1.37471809,
+      "memory(GiB)": 123.79,
+      "step": 57830,
+      "train_speed(iter/s)": 1.316765
+    },
+    {
+      "acc": 0.6745019,
+      "epoch": 1.4671486555048199,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.8183063584038236e-06,
+      "loss": 1.47597818,
+      "memory(GiB)": 123.79,
+      "step": 57835,
+      "train_speed(iter/s)": 1.316777
+    },
+    {
+      "acc": 0.68396235,
+      "epoch": 1.4672754946727549,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.8174975098380304e-06,
+      "loss": 1.41093187,
+      "memory(GiB)": 123.79,
+      "step": 57840,
+      "train_speed(iter/s)": 1.31679
+    },
+    {
+      "acc": 0.69183316,
+      "epoch": 1.46740233384069,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.8166888012529078e-06,
+      "loss": 1.38934898,
+      "memory(GiB)": 123.79,
+      "step": 57845,
+      "train_speed(iter/s)": 1.316803
+    },
+    {
+      "acc": 0.68538814,
+      "epoch": 1.467529173008625,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.8158802326840252e-06,
+      "loss": 1.36889715,
+      "memory(GiB)": 123.79,
+      "step": 57850,
+      "train_speed(iter/s)": 1.316815
+    },
+    {
+      "acc": 0.70408082,
+      "epoch": 1.46765601217656,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.8150718041669447e-06,
+      "loss": 1.31977215,
+      "memory(GiB)": 123.79,
+      "step": 57855,
+      "train_speed(iter/s)": 1.316826
+    },
+    {
+      "acc": 0.68763695,
+      "epoch": 1.467782851344495,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.814263515737224e-06,
+      "loss": 1.40299978,
+      "memory(GiB)": 123.79,
+      "step": 57860,
+      "train_speed(iter/s)": 1.316838
+    },
+    {
+      "acc": 0.68933802,
+      "epoch": 1.4679096905124303,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.813455367430419e-06,
+      "loss": 1.33243618,
+      "memory(GiB)": 123.79,
+      "step": 57865,
+      "train_speed(iter/s)": 1.316851
+    },
+    {
+      "acc": 0.68557944,
+      "epoch": 1.4680365296803652,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.812647359282076e-06,
+      "loss": 1.50047836,
+      "memory(GiB)": 123.79,
+      "step": 57870,
+      "train_speed(iter/s)": 1.316863
+    },
+    {
+      "acc": 0.66878166,
+      "epoch": 1.4681633688483005,
+      "grad_norm": 2.75,
+      "learning_rate": 1.8118394913277287e-06,
+      "loss": 1.39616308,
+      "memory(GiB)": 123.79,
+      "step": 57875,
+      "train_speed(iter/s)": 1.316875
+    },
+    {
+      "acc": 0.67681513,
+      "epoch": 1.4682902080162354,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.8110317636029162e-06,
+      "loss": 1.38618793,
+      "memory(GiB)": 123.79,
+      "step": 57880,
+      "train_speed(iter/s)": 1.316888
+    },
+    {
+      "acc": 0.68757963,
+      "epoch": 1.4684170471841704,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.810224176143165e-06,
+      "loss": 1.44136143,
+      "memory(GiB)": 123.79,
+      "step": 57885,
+      "train_speed(iter/s)": 1.3169
+    },
+    {
+      "acc": 0.68421049,
+      "epoch": 1.4685438863521054,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.8094167289839953e-06,
+      "loss": 1.31943922,
+      "memory(GiB)": 123.79,
+      "step": 57890,
+      "train_speed(iter/s)": 1.316912
+    },
+    {
+      "acc": 0.67638817,
+      "epoch": 1.4686707255200406,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.808609422160923e-06,
+      "loss": 1.45813751,
+      "memory(GiB)": 123.79,
+      "step": 57895,
+      "train_speed(iter/s)": 1.316924
+    },
+    {
+      "acc": 0.68537412,
+      "epoch": 1.4687975646879756,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.8078022557094571e-06,
+      "loss": 1.38470497,
+      "memory(GiB)": 123.79,
+      "step": 57900,
+      "train_speed(iter/s)": 1.316936
+    },
+    {
+      "acc": 0.68077154,
+      "epoch": 1.4689244038559108,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8069952296651e-06,
+      "loss": 1.43445673,
+      "memory(GiB)": 123.79,
+      "step": 57905,
+      "train_speed(iter/s)": 1.316949
+    },
+    {
+      "acc": 0.68379898,
+      "epoch": 1.4690512430238458,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.8061883440633481e-06,
+      "loss": 1.42066927,
+      "memory(GiB)": 123.79,
+      "step": 57910,
+      "train_speed(iter/s)": 1.316961
+    },
+    {
+      "acc": 0.69783087,
+      "epoch": 1.4691780821917808,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.8053815989396927e-06,
+      "loss": 1.37266579,
+      "memory(GiB)": 123.79,
+      "step": 57915,
+      "train_speed(iter/s)": 1.316974
+    },
+    {
+      "acc": 0.70556841,
+      "epoch": 1.4693049213597158,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.8045749943296171e-06,
+      "loss": 1.31972618,
+      "memory(GiB)": 123.79,
+      "step": 57920,
+      "train_speed(iter/s)": 1.316986
+    },
+    {
+      "acc": 0.68915157,
+      "epoch": 1.469431760527651,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.8037685302686003e-06,
+      "loss": 1.38504219,
+      "memory(GiB)": 123.79,
+      "step": 57925,
+      "train_speed(iter/s)": 1.316999
+    },
+    {
+      "acc": 0.68263254,
+      "epoch": 1.469558599695586,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.8029622067921133e-06,
+      "loss": 1.46189823,
+      "memory(GiB)": 123.79,
+      "step": 57930,
+      "train_speed(iter/s)": 1.317011
+    },
+    {
+      "acc": 0.70148377,
+      "epoch": 1.469685438863521,
+      "grad_norm": 3.375,
+      "learning_rate": 1.8021560239356223e-06,
+      "loss": 1.32665272,
+      "memory(GiB)": 123.79,
+      "step": 57935,
+      "train_speed(iter/s)": 1.317023
+    },
+    {
+      "acc": 0.69405031,
+      "epoch": 1.4698122780314562,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.8013499817345865e-06,
+      "loss": 1.36575069,
+      "memory(GiB)": 123.79,
+      "step": 57940,
+      "train_speed(iter/s)": 1.317036
+    },
+    {
+      "acc": 0.69294672,
+      "epoch": 1.4699391171993912,
+      "grad_norm": 3.25,
+      "learning_rate": 1.8005440802244595e-06,
+      "loss": 1.4002924,
+      "memory(GiB)": 123.79,
+      "step": 57945,
+      "train_speed(iter/s)": 1.317048
+    },
+    {
+      "acc": 0.68770881,
+      "epoch": 1.4700659563673262,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.7997383194406887e-06,
+      "loss": 1.34496813,
+      "memory(GiB)": 123.79,
+      "step": 57950,
+      "train_speed(iter/s)": 1.317061
+    },
+    {
+      "acc": 0.69033756,
+      "epoch": 1.4701927955352612,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.7989326994187146e-06,
+      "loss": 1.4089365,
+      "memory(GiB)": 123.79,
+      "step": 57955,
+      "train_speed(iter/s)": 1.317072
+    },
+    {
+      "acc": 0.69590316,
+      "epoch": 1.4703196347031964,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.798127220193972e-06,
+      "loss": 1.41595144,
+      "memory(GiB)": 123.79,
+      "step": 57960,
+      "train_speed(iter/s)": 1.317085
+    },
+    {
+      "acc": 0.67353096,
+      "epoch": 1.4704464738711314,
+      "grad_norm": 3.5,
+      "learning_rate": 1.7973218818018878e-06,
+      "loss": 1.41157885,
+      "memory(GiB)": 123.79,
+      "step": 57965,
+      "train_speed(iter/s)": 1.317098
+    },
+    {
+      "acc": 0.68679285,
+      "epoch": 1.4705733130390666,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7965166842778897e-06,
+      "loss": 1.38564892,
+      "memory(GiB)": 123.79,
+      "step": 57970,
+      "train_speed(iter/s)": 1.317111
+    },
+    {
+      "acc": 0.69984503,
+      "epoch": 1.4707001522070016,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.7957116276573888e-06,
+      "loss": 1.37687407,
+      "memory(GiB)": 123.79,
+      "step": 57975,
+      "train_speed(iter/s)": 1.317122
+    },
+    {
+      "acc": 0.69274898,
+      "epoch": 1.4708269913749366,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.7949067119757951e-06,
+      "loss": 1.32997503,
+      "memory(GiB)": 123.79,
+      "step": 57980,
+      "train_speed(iter/s)": 1.317135
+    },
+    {
+      "acc": 0.67153358,
+      "epoch": 1.4709538305428715,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.7941019372685154e-06,
+      "loss": 1.43820477,
+      "memory(GiB)": 123.79,
+      "step": 57985,
+      "train_speed(iter/s)": 1.317148
+    },
+    {
+      "acc": 0.68950977,
+      "epoch": 1.4710806697108068,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.7932973035709471e-06,
+      "loss": 1.32272882,
+      "memory(GiB)": 123.79,
+      "step": 57990,
+      "train_speed(iter/s)": 1.31716
+    },
+    {
+      "acc": 0.68680239,
+      "epoch": 1.4712075088787417,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.792492810918479e-06,
+      "loss": 1.38688784,
+      "memory(GiB)": 123.79,
+      "step": 57995,
+      "train_speed(iter/s)": 1.317171
+    },
+    {
+      "acc": 0.66852283,
+      "epoch": 1.4713343480466767,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.7916884593464957e-06,
+      "loss": 1.44040356,
+      "memory(GiB)": 123.79,
+      "step": 58000,
+      "train_speed(iter/s)": 1.317183
+    },
+    {
+      "epoch": 1.4713343480466767,
+      "eval_acc": 0.6750679221563014,
+      "eval_loss": 1.3563214540481567,
+      "eval_runtime": 69.749,
+      "eval_samples_per_second": 91.327,
+      "eval_steps_per_second": 22.839,
+      "step": 58000
+    },
+    {
+      "acc": 0.68598452,
+      "epoch": 1.471461187214612,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.79088424889038e-06,
+      "loss": 1.44533234,
+      "memory(GiB)": 123.79,
+      "step": 58005,
+      "train_speed(iter/s)": 1.31474
+    },
+    {
+      "acc": 0.6946382,
+      "epoch": 1.471588026382547,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.7900801795855043e-06,
+      "loss": 1.3399828,
+      "memory(GiB)": 123.79,
+      "step": 58010,
+      "train_speed(iter/s)": 1.314753
+    },
+    {
+      "acc": 0.6845654,
+      "epoch": 1.471714865550482,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.7892762514672303e-06,
+      "loss": 1.37822151,
+      "memory(GiB)": 123.79,
+      "step": 58015,
+      "train_speed(iter/s)": 1.314766
+    },
+    {
+      "acc": 0.68945179,
+      "epoch": 1.471841704718417,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.7884724645709228e-06,
+      "loss": 1.42351866,
+      "memory(GiB)": 123.79,
+      "step": 58020,
+      "train_speed(iter/s)": 1.314779
+    },
+    {
+      "acc": 0.68950748,
+      "epoch": 1.4719685438863521,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.7876688189319353e-06,
+      "loss": 1.41356831,
+      "memory(GiB)": 123.79,
+      "step": 58025,
+      "train_speed(iter/s)": 1.314792
+    },
+    {
+      "acc": 0.68923178,
+      "epoch": 1.472095383054287,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.7868653145856163e-06,
+      "loss": 1.34516535,
+      "memory(GiB)": 123.79,
+      "step": 58030,
+      "train_speed(iter/s)": 1.314805
+    },
+    {
+      "acc": 0.70743318,
+      "epoch": 1.4722222222222223,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.7860619515673034e-06,
+      "loss": 1.26716366,
+      "memory(GiB)": 123.79,
+      "step": 58035,
+      "train_speed(iter/s)": 1.314818
+    },
+    {
+      "acc": 0.69693494,
+      "epoch": 1.4723490613901573,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.785258729912337e-06,
+      "loss": 1.37037334,
+      "memory(GiB)": 123.79,
+      "step": 58040,
+      "train_speed(iter/s)": 1.314831
+    },
+    {
+      "acc": 0.66915712,
+      "epoch": 1.4724759005580923,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.784455649656044e-06,
+      "loss": 1.47697468,
+      "memory(GiB)": 123.79,
+      "step": 58045,
+      "train_speed(iter/s)": 1.314844
+    },
+    {
+      "acc": 0.69250164,
+      "epoch": 1.4726027397260273,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.7836527108337482e-06,
+      "loss": 1.3722415,
+      "memory(GiB)": 123.79,
+      "step": 58050,
+      "train_speed(iter/s)": 1.314857
+    },
+    {
+      "acc": 0.69483929,
+      "epoch": 1.4727295788939625,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.782849913480766e-06,
+      "loss": 1.38199196,
+      "memory(GiB)": 123.79,
+      "step": 58055,
+      "train_speed(iter/s)": 1.31487
+    },
+    {
+      "acc": 0.68606997,
+      "epoch": 1.4728564180618975,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.7820472576324078e-06,
+      "loss": 1.38468208,
+      "memory(GiB)": 123.79,
+      "step": 58060,
+      "train_speed(iter/s)": 1.314883
+    },
+    {
+      "acc": 0.70109854,
+      "epoch": 1.4729832572298327,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.7812447433239789e-06,
+      "loss": 1.35105219,
+      "memory(GiB)": 123.79,
+      "step": 58065,
+      "train_speed(iter/s)": 1.314895
+    },
+    {
+      "acc": 0.68299417,
+      "epoch": 1.4731100963977677,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.7804423705907764e-06,
+      "loss": 1.50381403,
+      "memory(GiB)": 123.79,
+      "step": 58070,
+      "train_speed(iter/s)": 1.314909
+    },
+    {
+      "acc": 0.68639526,
+      "epoch": 1.4732369355657027,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.779640139468093e-06,
+      "loss": 1.32613411,
+      "memory(GiB)": 123.79,
+      "step": 58075,
+      "train_speed(iter/s)": 1.31492
+    },
+    {
+      "acc": 0.68853064,
+      "epoch": 1.4733637747336377,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.778838049991214e-06,
+      "loss": 1.41448431,
+      "memory(GiB)": 123.79,
+      "step": 58080,
+      "train_speed(iter/s)": 1.314934
+    },
+    {
+      "acc": 0.67527161,
+      "epoch": 1.4734906139015729,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.778036102195419e-06,
+      "loss": 1.37413588,
+      "memory(GiB)": 123.79,
+      "step": 58085,
+      "train_speed(iter/s)": 1.314947
+    },
+    {
+      "acc": 0.68530946,
+      "epoch": 1.4736174530695079,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.7772342961159817e-06,
+      "loss": 1.40564499,
+      "memory(GiB)": 123.79,
+      "step": 58090,
+      "train_speed(iter/s)": 1.31496
+    },
+    {
+      "acc": 0.69515705,
+      "epoch": 1.4737442922374429,
+      "grad_norm": 3.375,
+      "learning_rate": 1.7764326317881681e-06,
+      "loss": 1.37018986,
+      "memory(GiB)": 123.79,
+      "step": 58095,
+      "train_speed(iter/s)": 1.314974
+    },
+    {
+      "acc": 0.69709206,
+      "epoch": 1.473871131405378,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.77563110924724e-06,
+      "loss": 1.39903164,
+      "memory(GiB)": 123.79,
+      "step": 58100,
+      "train_speed(iter/s)": 1.314987
+    },
+    {
+      "acc": 0.6868413,
+      "epoch": 1.473997970573313,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.7748297285284494e-06,
+      "loss": 1.40442848,
+      "memory(GiB)": 123.79,
+      "step": 58105,
+      "train_speed(iter/s)": 1.315
+    },
+    {
+      "acc": 0.69380207,
+      "epoch": 1.474124809741248,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.7740284896670507e-06,
+      "loss": 1.34978199,
+      "memory(GiB)": 123.79,
+      "step": 58110,
+      "train_speed(iter/s)": 1.315013
+    },
+    {
+      "acc": 0.69099131,
+      "epoch": 1.474251648909183,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.7732273926982796e-06,
+      "loss": 1.36701965,
+      "memory(GiB)": 123.79,
+      "step": 58115,
+      "train_speed(iter/s)": 1.315026
+    },
+    {
+      "acc": 0.69078631,
+      "epoch": 1.4743784880771182,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7724264376573747e-06,
+      "loss": 1.34959517,
+      "memory(GiB)": 123.79,
+      "step": 58120,
+      "train_speed(iter/s)": 1.315039
+    },
+    {
+      "acc": 0.67651935,
+      "epoch": 1.4745053272450532,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.7716256245795631e-06,
+      "loss": 1.44263515,
+      "memory(GiB)": 123.79,
+      "step": 58125,
+      "train_speed(iter/s)": 1.315052
+    },
+    {
+      "acc": 0.6825027,
+      "epoch": 1.4746321664129884,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.7708249535000737e-06,
+      "loss": 1.39643393,
+      "memory(GiB)": 123.79,
+      "step": 58130,
+      "train_speed(iter/s)": 1.315065
+    },
+    {
+      "acc": 0.67248883,
+      "epoch": 1.4747590055809234,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.7700244244541182e-06,
+      "loss": 1.41957846,
+      "memory(GiB)": 123.79,
+      "step": 58135,
+      "train_speed(iter/s)": 1.315078
+    },
+    {
+      "acc": 0.69745831,
+      "epoch": 1.4748858447488584,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.7692240374769081e-06,
+      "loss": 1.38119183,
+      "memory(GiB)": 123.79,
+      "step": 58140,
+      "train_speed(iter/s)": 1.31509
+    },
+    {
+      "acc": 0.68845491,
+      "epoch": 1.4750126839167934,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.7684237926036507e-06,
+      "loss": 1.43932228,
+      "memory(GiB)": 123.79,
+      "step": 58145,
+      "train_speed(iter/s)": 1.315102
+    },
+    {
+      "acc": 0.68016658,
+      "epoch": 1.4751395230847286,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.7676236898695442e-06,
+      "loss": 1.42827692,
+      "memory(GiB)": 123.79,
+      "step": 58150,
+      "train_speed(iter/s)": 1.315115
+    },
+    {
+      "acc": 0.69142284,
+      "epoch": 1.4752663622526636,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.7668237293097762e-06,
+      "loss": 1.42994614,
+      "memory(GiB)": 123.79,
+      "step": 58155,
+      "train_speed(iter/s)": 1.315128
+    },
+    {
+      "acc": 0.69782467,
+      "epoch": 1.4753932014205986,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.7660239109595374e-06,
+      "loss": 1.34804535,
+      "memory(GiB)": 123.79,
+      "step": 58160,
+      "train_speed(iter/s)": 1.315139
+    },
+    {
+      "acc": 0.66502857,
+      "epoch": 1.4755200405885338,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.7652242348540056e-06,
+      "loss": 1.47979298,
+      "memory(GiB)": 123.79,
+      "step": 58165,
+      "train_speed(iter/s)": 1.315151
+    },
+    {
+      "acc": 0.69332466,
+      "epoch": 1.4756468797564688,
+      "grad_norm": 3.0,
+      "learning_rate": 1.764424701028356e-06,
+      "loss": 1.31608877,
+      "memory(GiB)": 123.79,
+      "step": 58170,
+      "train_speed(iter/s)": 1.315163
+    },
+    {
+      "acc": 0.67261496,
+      "epoch": 1.4757737189244038,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.7636253095177507e-06,
+      "loss": 1.45930271,
+      "memory(GiB)": 123.79,
+      "step": 58175,
+      "train_speed(iter/s)": 1.315175
+    },
+    {
+      "acc": 0.67386856,
+      "epoch": 1.4759005580923388,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.762826060357355e-06,
+      "loss": 1.35927086,
+      "memory(GiB)": 123.79,
+      "step": 58180,
+      "train_speed(iter/s)": 1.315188
+    },
+    {
+      "acc": 0.69414244,
+      "epoch": 1.476027397260274,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.762026953582322e-06,
+      "loss": 1.3434886,
+      "memory(GiB)": 123.79,
+      "step": 58185,
+      "train_speed(iter/s)": 1.315201
+    },
+    {
+      "acc": 0.68011446,
+      "epoch": 1.476154236428209,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.7612279892278006e-06,
+      "loss": 1.3731102,
+      "memory(GiB)": 123.79,
+      "step": 58190,
+      "train_speed(iter/s)": 1.315215
+    },
+    {
+      "acc": 0.68809729,
+      "epoch": 1.4762810755961442,
+      "grad_norm": 3.5,
+      "learning_rate": 1.7604291673289314e-06,
+      "loss": 1.40485058,
+      "memory(GiB)": 123.79,
+      "step": 58195,
+      "train_speed(iter/s)": 1.315228
+    },
+    {
+      "acc": 0.69033098,
+      "epoch": 1.4764079147640792,
+      "grad_norm": 3.125,
+      "learning_rate": 1.759630487920852e-06,
+      "loss": 1.30835505,
+      "memory(GiB)": 123.79,
+      "step": 58200,
+      "train_speed(iter/s)": 1.31524
+    },
+    {
+      "acc": 0.69039221,
+      "epoch": 1.4765347539320142,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.7588319510386903e-06,
+      "loss": 1.36298656,
+      "memory(GiB)": 123.79,
+      "step": 58205,
+      "train_speed(iter/s)": 1.315253
+    },
+    {
+      "acc": 0.68752489,
+      "epoch": 1.4766615930999492,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.7580335567175704e-06,
+      "loss": 1.35670633,
+      "memory(GiB)": 123.79,
+      "step": 58210,
+      "train_speed(iter/s)": 1.315266
+    },
+    {
+      "acc": 0.66992989,
+      "epoch": 1.4767884322678844,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.7572353049926094e-06,
+      "loss": 1.50954065,
+      "memory(GiB)": 123.79,
+      "step": 58215,
+      "train_speed(iter/s)": 1.31528
+    },
+    {
+      "acc": 0.70145626,
+      "epoch": 1.4769152714358194,
+      "grad_norm": 2.75,
+      "learning_rate": 1.7564371958989173e-06,
+      "loss": 1.31519699,
+      "memory(GiB)": 123.79,
+      "step": 58220,
+      "train_speed(iter/s)": 1.315293
+    },
+    {
+      "acc": 0.69232578,
+      "epoch": 1.4770421106037546,
+      "grad_norm": 4.0,
+      "learning_rate": 1.7556392294715984e-06,
+      "loss": 1.3987607,
+      "memory(GiB)": 123.79,
+      "step": 58225,
+      "train_speed(iter/s)": 1.315307
+    },
+    {
+      "acc": 0.6749198,
+      "epoch": 1.4771689497716896,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.7548414057457518e-06,
+      "loss": 1.40326033,
+      "memory(GiB)": 123.79,
+      "step": 58230,
+      "train_speed(iter/s)": 1.315316
+    },
+    {
+      "acc": 0.68240991,
+      "epoch": 1.4772957889396245,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.7540437247564685e-06,
+      "loss": 1.39972153,
+      "memory(GiB)": 123.79,
+      "step": 58235,
+      "train_speed(iter/s)": 1.315329
+    },
+    {
+      "acc": 0.70451846,
+      "epoch": 1.4774226281075595,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.7532461865388345e-06,
+      "loss": 1.28882828,
+      "memory(GiB)": 123.79,
+      "step": 58240,
+      "train_speed(iter/s)": 1.315343
+    },
+    {
+      "acc": 0.68779707,
+      "epoch": 1.4775494672754947,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.752448791127927e-06,
+      "loss": 1.37972012,
+      "memory(GiB)": 123.79,
+      "step": 58245,
+      "train_speed(iter/s)": 1.315357
+    },
+    {
+      "acc": 0.69111476,
+      "epoch": 1.4776763064434297,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.7516515385588245e-06,
+      "loss": 1.41364059,
+      "memory(GiB)": 123.79,
+      "step": 58250,
+      "train_speed(iter/s)": 1.315371
+    },
+    {
+      "acc": 0.68119116,
+      "epoch": 1.4778031456113647,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.7508544288665885e-06,
+      "loss": 1.41164169,
+      "memory(GiB)": 123.79,
+      "step": 58255,
+      "train_speed(iter/s)": 1.315384
+    },
+    {
+      "acc": 0.6977838,
+      "epoch": 1.4779299847793,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.750057462086281e-06,
+      "loss": 1.3615695,
+      "memory(GiB)": 123.79,
+      "step": 58260,
+      "train_speed(iter/s)": 1.315398
+    },
+    {
+      "acc": 0.69503889,
+      "epoch": 1.478056823947235,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.7492606382529542e-06,
+      "loss": 1.3810667,
+      "memory(GiB)": 123.79,
+      "step": 58265,
+      "train_speed(iter/s)": 1.315412
+    },
+    {
+      "acc": 0.695609,
+      "epoch": 1.47818366311517,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.748463957401662e-06,
+      "loss": 1.38900547,
+      "memory(GiB)": 123.79,
+      "step": 58270,
+      "train_speed(iter/s)": 1.315425
+    },
+    {
+      "acc": 0.68731928,
+      "epoch": 1.478310502283105,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.7476674195674404e-06,
+      "loss": 1.38738899,
+      "memory(GiB)": 123.79,
+      "step": 58275,
+      "train_speed(iter/s)": 1.315439
+    },
+    {
+      "acc": 0.69408827,
+      "epoch": 1.47843734145104,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.7468710247853244e-06,
+      "loss": 1.41225786,
+      "memory(GiB)": 123.79,
+      "step": 58280,
+      "train_speed(iter/s)": 1.315452
+    },
+    {
+      "acc": 0.68857317,
+      "epoch": 1.478564180618975,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.7460747730903466e-06,
+      "loss": 1.35121822,
+      "memory(GiB)": 123.79,
+      "step": 58285,
+      "train_speed(iter/s)": 1.315464
+    },
+    {
+      "acc": 0.68231106,
+      "epoch": 1.4786910197869103,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.7452786645175297e-06,
+      "loss": 1.37984638,
+      "memory(GiB)": 123.79,
+      "step": 58290,
+      "train_speed(iter/s)": 1.315477
+    },
+    {
+      "acc": 0.67790961,
+      "epoch": 1.4788178589548453,
+      "grad_norm": 2.875,
+      "learning_rate": 1.7444826991018864e-06,
+      "loss": 1.44939327,
+      "memory(GiB)": 123.79,
+      "step": 58295,
+      "train_speed(iter/s)": 1.31549
+    },
+    {
+      "acc": 0.6825675,
+      "epoch": 1.4789446981227803,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7436868768784276e-06,
+      "loss": 1.37009087,
+      "memory(GiB)": 123.79,
+      "step": 58300,
+      "train_speed(iter/s)": 1.315503
+    },
+    {
+      "acc": 0.69646368,
+      "epoch": 1.4790715372907153,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.7428911978821594e-06,
+      "loss": 1.39452286,
+      "memory(GiB)": 123.79,
+      "step": 58305,
+      "train_speed(iter/s)": 1.315515
+    },
+    {
+      "acc": 0.67944126,
+      "epoch": 1.4791983764586505,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.7420956621480806e-06,
+      "loss": 1.41876822,
+      "memory(GiB)": 123.79,
+      "step": 58310,
+      "train_speed(iter/s)": 1.315528
+    },
+    {
+      "acc": 0.69551373,
+      "epoch": 1.4793252156265855,
+      "grad_norm": 2.75,
+      "learning_rate": 1.7413002697111765e-06,
+      "loss": 1.33698502,
+      "memory(GiB)": 123.79,
+      "step": 58315,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.69713287,
+      "epoch": 1.4794520547945205,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.7405050206064372e-06,
+      "loss": 1.3955492,
+      "memory(GiB)": 123.79,
+      "step": 58320,
+      "train_speed(iter/s)": 1.315553
+    },
+    {
+      "acc": 0.68271995,
+      "epoch": 1.4795788939624557,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.73970991486884e-06,
+      "loss": 1.40831032,
+      "memory(GiB)": 123.79,
+      "step": 58325,
+      "train_speed(iter/s)": 1.315566
+    },
+    {
+      "acc": 0.68104639,
+      "epoch": 1.4797057331303907,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.7389149525333565e-06,
+      "loss": 1.43635387,
+      "memory(GiB)": 123.79,
+      "step": 58330,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.66124468,
+      "epoch": 1.4798325722983257,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.7381201336349535e-06,
+      "loss": 1.3836153,
+      "memory(GiB)": 123.79,
+      "step": 58335,
+      "train_speed(iter/s)": 1.31559
+    },
+    {
+      "acc": 0.68624039,
+      "epoch": 1.4799594114662606,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.7373254582085896e-06,
+      "loss": 1.39270096,
+      "memory(GiB)": 123.79,
+      "step": 58340,
+      "train_speed(iter/s)": 1.315603
+    },
+    {
+      "acc": 0.68552904,
+      "epoch": 1.4800862506341959,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.7365309262892194e-06,
+      "loss": 1.34188175,
+      "memory(GiB)": 123.79,
+      "step": 58345,
+      "train_speed(iter/s)": 1.315615
+    },
+    {
+      "acc": 0.67305756,
+      "epoch": 1.4802130898021308,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.735736537911789e-06,
+      "loss": 1.45427074,
+      "memory(GiB)": 123.79,
+      "step": 58350,
+      "train_speed(iter/s)": 1.315628
+    },
+    {
+      "acc": 0.66791158,
+      "epoch": 1.480339928970066,
+      "grad_norm": 2.625,
+      "learning_rate": 1.7349422931112403e-06,
+      "loss": 1.44060583,
+      "memory(GiB)": 123.79,
+      "step": 58355,
+      "train_speed(iter/s)": 1.315641
+    },
+    {
+      "acc": 0.69470425,
+      "epoch": 1.480466768138001,
+      "grad_norm": 2.625,
+      "learning_rate": 1.7341481919225062e-06,
+      "loss": 1.39404135,
+      "memory(GiB)": 123.79,
+      "step": 58360,
+      "train_speed(iter/s)": 1.315654
+    },
+    {
+      "acc": 0.69160433,
+      "epoch": 1.480593607305936,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.733354234380516e-06,
+      "loss": 1.3506588,
+      "memory(GiB)": 123.79,
+      "step": 58365,
+      "train_speed(iter/s)": 1.315666
+    },
+    {
+      "acc": 0.68694754,
+      "epoch": 1.480720446473871,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.7325604205201912e-06,
+      "loss": 1.31518583,
+      "memory(GiB)": 123.79,
+      "step": 58370,
+      "train_speed(iter/s)": 1.315679
+    },
+    {
+      "acc": 0.6837018,
+      "epoch": 1.4808472856418062,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.7317667503764468e-06,
+      "loss": 1.3355444,
+      "memory(GiB)": 123.79,
+      "step": 58375,
+      "train_speed(iter/s)": 1.315691
+    },
+    {
+      "acc": 0.67436404,
+      "epoch": 1.4809741248097412,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.7309732239841926e-06,
+      "loss": 1.53784275,
+      "memory(GiB)": 123.79,
+      "step": 58380,
+      "train_speed(iter/s)": 1.315704
+    },
+    {
+      "acc": 0.68956265,
+      "epoch": 1.4811009639776764,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.730179841378331e-06,
+      "loss": 1.34985237,
+      "memory(GiB)": 123.79,
+      "step": 58385,
+      "train_speed(iter/s)": 1.315716
+    },
+    {
+      "acc": 0.68944983,
+      "epoch": 1.4812278031456114,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.7293866025937589e-06,
+      "loss": 1.37308807,
+      "memory(GiB)": 123.79,
+      "step": 58390,
+      "train_speed(iter/s)": 1.315729
+    },
+    {
+      "acc": 0.69536452,
+      "epoch": 1.4813546423135464,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.7285935076653659e-06,
+      "loss": 1.36332798,
+      "memory(GiB)": 123.79,
+      "step": 58395,
+      "train_speed(iter/s)": 1.315742
+    },
+    {
+      "acc": 0.70002193,
+      "epoch": 1.4814814814814814,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.7278005566280365e-06,
+      "loss": 1.34988594,
+      "memory(GiB)": 123.79,
+      "step": 58400,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.67621846,
+      "epoch": 1.4816083206494166,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.727007749516646e-06,
+      "loss": 1.46184912,
+      "memory(GiB)": 123.79,
+      "step": 58405,
+      "train_speed(iter/s)": 1.315767
+    },
+    {
+      "acc": 0.68149786,
+      "epoch": 1.4817351598173516,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.7262150863660709e-06,
+      "loss": 1.4609808,
+      "memory(GiB)": 123.79,
+      "step": 58410,
+      "train_speed(iter/s)": 1.31578
+    },
+    {
+      "acc": 0.6847681,
+      "epoch": 1.4818619989852866,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.7254225672111713e-06,
+      "loss": 1.39584541,
+      "memory(GiB)": 123.79,
+      "step": 58415,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.68047543,
+      "epoch": 1.4819888381532218,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.7246301920868052e-06,
+      "loss": 1.39603882,
+      "memory(GiB)": 123.79,
+      "step": 58420,
+      "train_speed(iter/s)": 1.315805
+    },
+    {
+      "acc": 0.69595737,
+      "epoch": 1.4821156773211568,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.723837961027829e-06,
+      "loss": 1.27588272,
+      "memory(GiB)": 123.79,
+      "step": 58425,
+      "train_speed(iter/s)": 1.315817
+    },
+    {
+      "acc": 0.68832731,
+      "epoch": 1.4822425164890918,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.723045874069087e-06,
+      "loss": 1.39518671,
+      "memory(GiB)": 123.79,
+      "step": 58430,
+      "train_speed(iter/s)": 1.315829
+    },
+    {
+      "acc": 0.68918791,
+      "epoch": 1.4823693556570268,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.7222539312454167e-06,
+      "loss": 1.41825247,
+      "memory(GiB)": 123.79,
+      "step": 58435,
+      "train_speed(iter/s)": 1.315842
+    },
+    {
+      "acc": 0.68175139,
+      "epoch": 1.482496194824962,
+      "grad_norm": 3.75,
+      "learning_rate": 1.7214621325916515e-06,
+      "loss": 1.4459794,
+      "memory(GiB)": 123.79,
+      "step": 58440,
+      "train_speed(iter/s)": 1.315855
+    },
+    {
+      "acc": 0.69667006,
+      "epoch": 1.482623033992897,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.7206704781426204e-06,
+      "loss": 1.33932648,
+      "memory(GiB)": 123.79,
+      "step": 58445,
+      "train_speed(iter/s)": 1.315867
+    },
+    {
+      "acc": 0.7052392,
+      "epoch": 1.4827498731608322,
+      "grad_norm": 3.25,
+      "learning_rate": 1.7198789679331445e-06,
+      "loss": 1.35031834,
+      "memory(GiB)": 123.79,
+      "step": 58450,
+      "train_speed(iter/s)": 1.31588
+    },
+    {
+      "acc": 0.69935436,
+      "epoch": 1.4828767123287672,
+      "grad_norm": 2.21875,
+      "learning_rate": 1.7190876019980329e-06,
+      "loss": 1.3200119,
+      "memory(GiB)": 123.79,
+      "step": 58455,
+      "train_speed(iter/s)": 1.315893
+    },
+    {
+      "acc": 0.68428907,
+      "epoch": 1.4830035514967022,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.7182963803720987e-06,
+      "loss": 1.37018623,
+      "memory(GiB)": 123.79,
+      "step": 58460,
+      "train_speed(iter/s)": 1.315905
+    },
+    {
+      "acc": 0.66703,
+      "epoch": 1.4831303906646371,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.7175053030901418e-06,
+      "loss": 1.45366631,
+      "memory(GiB)": 123.79,
+      "step": 58465,
+      "train_speed(iter/s)": 1.315919
+    },
+    {
+      "acc": 0.68941202,
+      "epoch": 1.4832572298325724,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.7167143701869582e-06,
+      "loss": 1.39382019,
+      "memory(GiB)": 123.79,
+      "step": 58470,
+      "train_speed(iter/s)": 1.315931
+    },
+    {
+      "acc": 0.68703032,
+      "epoch": 1.4833840690005073,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.7159235816973318e-06,
+      "loss": 1.3274786,
+      "memory(GiB)": 123.79,
+      "step": 58475,
+      "train_speed(iter/s)": 1.315945
+    },
+    {
+      "acc": 0.69361715,
+      "epoch": 1.4835109081684423,
+      "grad_norm": 2.75,
+      "learning_rate": 1.7151329376560506e-06,
+      "loss": 1.3993638,
+      "memory(GiB)": 123.79,
+      "step": 58480,
+      "train_speed(iter/s)": 1.315957
+    },
+    {
+      "acc": 0.67998428,
+      "epoch": 1.4836377473363775,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.7143424380978885e-06,
+      "loss": 1.38261452,
+      "memory(GiB)": 123.79,
+      "step": 58485,
+      "train_speed(iter/s)": 1.315969
+    },
+    {
+      "acc": 0.6827539,
+      "epoch": 1.4837645865043125,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.7135520830576157e-06,
+      "loss": 1.4588726,
+      "memory(GiB)": 123.79,
+      "step": 58490,
+      "train_speed(iter/s)": 1.315981
+    },
+    {
+      "acc": 0.68694377,
+      "epoch": 1.4838914256722475,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.712761872569995e-06,
+      "loss": 1.37346649,
+      "memory(GiB)": 123.79,
+      "step": 58495,
+      "train_speed(iter/s)": 1.315993
+    },
+    {
+      "acc": 0.6821198,
+      "epoch": 1.4840182648401825,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.7119718066697838e-06,
+      "loss": 1.42292271,
+      "memory(GiB)": 123.79,
+      "step": 58500,
+      "train_speed(iter/s)": 1.316005
+    },
+    {
+      "acc": 0.68047175,
+      "epoch": 1.4841451040081177,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.7111818853917323e-06,
+      "loss": 1.38929863,
+      "memory(GiB)": 123.79,
+      "step": 58505,
+      "train_speed(iter/s)": 1.316017
+    },
+    {
+      "acc": 0.68375831,
+      "epoch": 1.4842719431760527,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.710392108770585e-06,
+      "loss": 1.43962288,
+      "memory(GiB)": 123.79,
+      "step": 58510,
+      "train_speed(iter/s)": 1.316029
+    },
+    {
+      "acc": 0.68900018,
+      "epoch": 1.484398782343988,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.7096024768410796e-06,
+      "loss": 1.43275404,
+      "memory(GiB)": 123.79,
+      "step": 58515,
+      "train_speed(iter/s)": 1.31604
+    },
+    {
+      "acc": 0.66924582,
+      "epoch": 1.484525621511923,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.7088129896379484e-06,
+      "loss": 1.46445065,
+      "memory(GiB)": 123.79,
+      "step": 58520,
+      "train_speed(iter/s)": 1.316053
+    },
+    {
+      "acc": 0.66811295,
+      "epoch": 1.484652460679858,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7080236471959155e-06,
+      "loss": 1.4043786,
+      "memory(GiB)": 123.79,
+      "step": 58525,
+      "train_speed(iter/s)": 1.316064
+    },
+    {
+      "acc": 0.69301043,
+      "epoch": 1.4847792998477929,
+      "grad_norm": 3.25,
+      "learning_rate": 1.7072344495497007e-06,
+      "loss": 1.42271652,
+      "memory(GiB)": 123.79,
+      "step": 58530,
+      "train_speed(iter/s)": 1.316076
+    },
+    {
+      "acc": 0.71561279,
+      "epoch": 1.484906139015728,
+      "grad_norm": 3.125,
+      "learning_rate": 1.7064453967340155e-06,
+      "loss": 1.31203527,
+      "memory(GiB)": 123.79,
+      "step": 58535,
+      "train_speed(iter/s)": 1.316088
+    },
+    {
+      "acc": 0.67992544,
+      "epoch": 1.485032978183663,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.7056564887835664e-06,
+      "loss": 1.38778019,
+      "memory(GiB)": 123.79,
+      "step": 58540,
+      "train_speed(iter/s)": 1.316101
+    },
+    {
+      "acc": 0.66554432,
+      "epoch": 1.4851598173515983,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.704867725733052e-06,
+      "loss": 1.49258299,
+      "memory(GiB)": 123.79,
+      "step": 58545,
+      "train_speed(iter/s)": 1.316114
+    },
+    {
+      "acc": 0.68939514,
+      "epoch": 1.4852866565195333,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.7040791076171692e-06,
+      "loss": 1.35368767,
+      "memory(GiB)": 123.79,
+      "step": 58550,
+      "train_speed(iter/s)": 1.316126
+    },
+    {
+      "acc": 0.68568745,
+      "epoch": 1.4854134956874683,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.7032906344706017e-06,
+      "loss": 1.39501572,
+      "memory(GiB)": 123.79,
+      "step": 58555,
+      "train_speed(iter/s)": 1.316139
+    },
+    {
+      "acc": 0.67243981,
+      "epoch": 1.4855403348554033,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.7025023063280306e-06,
+      "loss": 1.43231297,
+      "memory(GiB)": 123.79,
+      "step": 58560,
+      "train_speed(iter/s)": 1.316151
+    },
+    {
+      "acc": 0.67555895,
+      "epoch": 1.4856671740233385,
+      "grad_norm": 3.125,
+      "learning_rate": 1.701714123224128e-06,
+      "loss": 1.44561234,
+      "memory(GiB)": 123.79,
+      "step": 58565,
+      "train_speed(iter/s)": 1.316163
+    },
+    {
+      "acc": 0.67643614,
+      "epoch": 1.4857940131912735,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.7009260851935684e-06,
+      "loss": 1.41789017,
+      "memory(GiB)": 123.79,
+      "step": 58570,
+      "train_speed(iter/s)": 1.316176
+    },
+    {
+      "acc": 0.66988959,
+      "epoch": 1.4859208523592085,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.7001381922710064e-06,
+      "loss": 1.49019051,
+      "memory(GiB)": 123.79,
+      "step": 58575,
+      "train_speed(iter/s)": 1.316188
+    },
+    {
+      "acc": 0.68065605,
+      "epoch": 1.4860476915271437,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.699350444491098e-06,
+      "loss": 1.36851797,
+      "memory(GiB)": 123.79,
+      "step": 58580,
+      "train_speed(iter/s)": 1.316199
+    },
+    {
+      "acc": 0.67409554,
+      "epoch": 1.4861745306950787,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.6985628418884953e-06,
+      "loss": 1.43477573,
+      "memory(GiB)": 123.79,
+      "step": 58585,
+      "train_speed(iter/s)": 1.316211
+    },
+    {
+      "acc": 0.69355106,
+      "epoch": 1.4863013698630136,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.6977753844978406e-06,
+      "loss": 1.39010172,
+      "memory(GiB)": 123.79,
+      "step": 58590,
+      "train_speed(iter/s)": 1.316223
+    },
+    {
+      "acc": 0.68358183,
+      "epoch": 1.4864282090309486,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.696988072353764e-06,
+      "loss": 1.3820364,
+      "memory(GiB)": 123.79,
+      "step": 58595,
+      "train_speed(iter/s)": 1.316235
+    },
+    {
+      "acc": 0.68277793,
+      "epoch": 1.4865550481988838,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6962009054909007e-06,
+      "loss": 1.40018129,
+      "memory(GiB)": 123.79,
+      "step": 58600,
+      "train_speed(iter/s)": 1.316248
+    },
+    {
+      "acc": 0.69380007,
+      "epoch": 1.4866818873668188,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.6954138839438723e-06,
+      "loss": 1.35845718,
+      "memory(GiB)": 123.79,
+      "step": 58605,
+      "train_speed(iter/s)": 1.316261
+    },
+    {
+      "acc": 0.68398008,
+      "epoch": 1.486808726534754,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.6946270077472966e-06,
+      "loss": 1.43675499,
+      "memory(GiB)": 123.79,
+      "step": 58610,
+      "train_speed(iter/s)": 1.316274
+    },
+    {
+      "acc": 0.69425774,
+      "epoch": 1.486935565702689,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.6938402769357787e-06,
+      "loss": 1.33486805,
+      "memory(GiB)": 123.79,
+      "step": 58615,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.68759184,
+      "epoch": 1.487062404870624,
+      "grad_norm": 5.96875,
+      "learning_rate": 1.6930536915439288e-06,
+      "loss": 1.39685535,
+      "memory(GiB)": 123.79,
+      "step": 58620,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.68913713,
+      "epoch": 1.487189244038559,
+      "grad_norm": 3.625,
+      "learning_rate": 1.6922672516063415e-06,
+      "loss": 1.39228497,
+      "memory(GiB)": 123.79,
+      "step": 58625,
+      "train_speed(iter/s)": 1.316314
+    },
+    {
+      "acc": 0.68271675,
+      "epoch": 1.4873160832064942,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.6914809571576086e-06,
+      "loss": 1.39343319,
+      "memory(GiB)": 123.79,
+      "step": 58630,
+      "train_speed(iter/s)": 1.316327
+    },
+    {
+      "acc": 0.67708774,
+      "epoch": 1.4874429223744292,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6906948082323149e-06,
+      "loss": 1.46628551,
+      "memory(GiB)": 123.79,
+      "step": 58635,
+      "train_speed(iter/s)": 1.31634
+    },
+    {
+      "acc": 0.6870338,
+      "epoch": 1.4875697615423642,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.689908804865038e-06,
+      "loss": 1.44293356,
+      "memory(GiB)": 123.79,
+      "step": 58640,
+      "train_speed(iter/s)": 1.316352
+    },
+    {
+      "acc": 0.68190174,
+      "epoch": 1.4876966007102994,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6891229470903509e-06,
+      "loss": 1.42391186,
+      "memory(GiB)": 123.79,
+      "step": 58645,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.68516903,
+      "epoch": 1.4878234398782344,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.6883372349428184e-06,
+      "loss": 1.43949814,
+      "memory(GiB)": 123.79,
+      "step": 58650,
+      "train_speed(iter/s)": 1.316379
+    },
+    {
+      "acc": 0.70407124,
+      "epoch": 1.4879502790461694,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.6875516684569999e-06,
+      "loss": 1.34570675,
+      "memory(GiB)": 123.79,
+      "step": 58655,
+      "train_speed(iter/s)": 1.316392
+    },
+    {
+      "acc": 0.6914443,
+      "epoch": 1.4880771182141044,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.686766247667448e-06,
+      "loss": 1.41428919,
+      "memory(GiB)": 123.79,
+      "step": 58660,
+      "train_speed(iter/s)": 1.316405
+    },
+    {
+      "acc": 0.68349743,
+      "epoch": 1.4882039573820396,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.685980972608709e-06,
+      "loss": 1.39280539,
+      "memory(GiB)": 123.79,
+      "step": 58665,
+      "train_speed(iter/s)": 1.316418
+    },
+    {
+      "acc": 0.68131104,
+      "epoch": 1.4883307965499746,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.6851958433153227e-06,
+      "loss": 1.36191444,
+      "memory(GiB)": 123.79,
+      "step": 58670,
+      "train_speed(iter/s)": 1.316432
+    },
+    {
+      "acc": 0.6812809,
+      "epoch": 1.4884576357179098,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.684410859821823e-06,
+      "loss": 1.4731741,
+      "memory(GiB)": 123.79,
+      "step": 58675,
+      "train_speed(iter/s)": 1.316444
+    },
+    {
+      "acc": 0.70332422,
+      "epoch": 1.4885844748858448,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.6836260221627364e-06,
+      "loss": 1.33429184,
+      "memory(GiB)": 123.79,
+      "step": 58680,
+      "train_speed(iter/s)": 1.316457
+    },
+    {
+      "acc": 0.68832707,
+      "epoch": 1.4887113140537798,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.682841330372582e-06,
+      "loss": 1.35774517,
+      "memory(GiB)": 123.79,
+      "step": 58685,
+      "train_speed(iter/s)": 1.31647
+    },
+    {
+      "acc": 0.69368639,
+      "epoch": 1.4888381532217148,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.6820567844858798e-06,
+      "loss": 1.31389723,
+      "memory(GiB)": 123.79,
+      "step": 58690,
+      "train_speed(iter/s)": 1.316484
+    },
+    {
+      "acc": 0.68360877,
+      "epoch": 1.48896499238965,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.681272384537132e-06,
+      "loss": 1.37670841,
+      "memory(GiB)": 123.79,
+      "step": 58695,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.69291019,
+      "epoch": 1.489091831557585,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.6804881305608423e-06,
+      "loss": 1.34366474,
+      "memory(GiB)": 123.79,
+      "step": 58700,
+      "train_speed(iter/s)": 1.316507
+    },
+    {
+      "acc": 0.67939425,
+      "epoch": 1.4892186707255202,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.679704022591503e-06,
+      "loss": 1.41897411,
+      "memory(GiB)": 123.79,
+      "step": 58705,
+      "train_speed(iter/s)": 1.31652
+    },
+    {
+      "acc": 0.68195486,
+      "epoch": 1.4893455098934552,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.678920060663608e-06,
+      "loss": 1.3921978,
+      "memory(GiB)": 123.79,
+      "step": 58710,
+      "train_speed(iter/s)": 1.316532
+    },
+    {
+      "acc": 0.68994083,
+      "epoch": 1.4894723490613901,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.6781362448116344e-06,
+      "loss": 1.35055075,
+      "memory(GiB)": 123.79,
+      "step": 58715,
+      "train_speed(iter/s)": 1.316545
+    },
+    {
+      "acc": 0.68696985,
+      "epoch": 1.4895991882293251,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.6773525750700586e-06,
+      "loss": 1.38973751,
+      "memory(GiB)": 123.79,
+      "step": 58720,
+      "train_speed(iter/s)": 1.316559
+    },
+    {
+      "acc": 0.69245625,
+      "epoch": 1.4897260273972603,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.676569051473353e-06,
+      "loss": 1.36175642,
+      "memory(GiB)": 123.79,
+      "step": 58725,
+      "train_speed(iter/s)": 1.316572
+    },
+    {
+      "acc": 0.68587198,
+      "epoch": 1.4898528665651953,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.6757856740559796e-06,
+      "loss": 1.3747241,
+      "memory(GiB)": 123.79,
+      "step": 58730,
+      "train_speed(iter/s)": 1.316586
+    },
+    {
+      "acc": 0.69633522,
+      "epoch": 1.4899797057331303,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.6750024428523926e-06,
+      "loss": 1.32790966,
+      "memory(GiB)": 123.79,
+      "step": 58735,
+      "train_speed(iter/s)": 1.3166
+    },
+    {
+      "acc": 0.69537711,
+      "epoch": 1.4901065449010655,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.6742193578970418e-06,
+      "loss": 1.40439415,
+      "memory(GiB)": 123.79,
+      "step": 58740,
+      "train_speed(iter/s)": 1.316613
+    },
+    {
+      "acc": 0.69179363,
+      "epoch": 1.4902333840690005,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.673436419224373e-06,
+      "loss": 1.40092869,
+      "memory(GiB)": 123.79,
+      "step": 58745,
+      "train_speed(iter/s)": 1.316626
+    },
+    {
+      "acc": 0.68317809,
+      "epoch": 1.4903602232369355,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.6726536268688248e-06,
+      "loss": 1.45522804,
+      "memory(GiB)": 123.79,
+      "step": 58750,
+      "train_speed(iter/s)": 1.31664
+    },
+    {
+      "acc": 0.6887392,
+      "epoch": 1.4904870624048705,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.671870980864822e-06,
+      "loss": 1.39484177,
+      "memory(GiB)": 123.79,
+      "step": 58755,
+      "train_speed(iter/s)": 1.316653
+    },
+    {
+      "acc": 0.69812574,
+      "epoch": 1.4906139015728057,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.6710884812467943e-06,
+      "loss": 1.29074726,
+      "memory(GiB)": 123.79,
+      "step": 58760,
+      "train_speed(iter/s)": 1.316665
+    },
+    {
+      "acc": 0.69223342,
+      "epoch": 1.4907407407407407,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.6703061280491579e-06,
+      "loss": 1.40786142,
+      "memory(GiB)": 123.79,
+      "step": 58765,
+      "train_speed(iter/s)": 1.316679
+    },
+    {
+      "acc": 0.68930879,
+      "epoch": 1.490867579908676,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.6695239213063237e-06,
+      "loss": 1.36336155,
+      "memory(GiB)": 123.79,
+      "step": 58770,
+      "train_speed(iter/s)": 1.316692
+    },
+    {
+      "acc": 0.69082289,
+      "epoch": 1.490994419076611,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.6687418610526972e-06,
+      "loss": 1.38249149,
+      "memory(GiB)": 123.79,
+      "step": 58775,
+      "train_speed(iter/s)": 1.316705
+    },
+    {
+      "acc": 0.69906492,
+      "epoch": 1.4911212582445459,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.6679599473226766e-06,
+      "loss": 1.32758007,
+      "memory(GiB)": 123.79,
+      "step": 58780,
+      "train_speed(iter/s)": 1.316718
+    },
+    {
+      "acc": 0.68299227,
+      "epoch": 1.4912480974124809,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.6671781801506536e-06,
+      "loss": 1.4442338,
+      "memory(GiB)": 123.79,
+      "step": 58785,
+      "train_speed(iter/s)": 1.316731
+    },
+    {
+      "acc": 0.68238907,
+      "epoch": 1.491374936580416,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.6663965595710147e-06,
+      "loss": 1.37768564,
+      "memory(GiB)": 123.79,
+      "step": 58790,
+      "train_speed(iter/s)": 1.316743
+    },
+    {
+      "acc": 0.70041637,
+      "epoch": 1.491501775748351,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.6656150856181386e-06,
+      "loss": 1.37371235,
+      "memory(GiB)": 123.79,
+      "step": 58795,
+      "train_speed(iter/s)": 1.316756
+    },
+    {
+      "acc": 0.67299309,
+      "epoch": 1.491628614916286,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.6648337583263974e-06,
+      "loss": 1.43401279,
+      "memory(GiB)": 123.79,
+      "step": 58800,
+      "train_speed(iter/s)": 1.316768
+    },
+    {
+      "acc": 0.68447123,
+      "epoch": 1.4917554540842213,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6640525777301586e-06,
+      "loss": 1.39521732,
+      "memory(GiB)": 123.79,
+      "step": 58805,
+      "train_speed(iter/s)": 1.316781
+    },
+    {
+      "acc": 0.69841371,
+      "epoch": 1.4918822932521563,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.663271543863781e-06,
+      "loss": 1.36435547,
+      "memory(GiB)": 123.79,
+      "step": 58810,
+      "train_speed(iter/s)": 1.316794
+    },
+    {
+      "acc": 0.68965197,
+      "epoch": 1.4920091324200913,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.6624906567616183e-06,
+      "loss": 1.41122961,
+      "memory(GiB)": 123.79,
+      "step": 58815,
+      "train_speed(iter/s)": 1.316807
+    },
+    {
+      "acc": 0.68628387,
+      "epoch": 1.4921359715880262,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.6617099164580175e-06,
+      "loss": 1.36296968,
+      "memory(GiB)": 123.79,
+      "step": 58820,
+      "train_speed(iter/s)": 1.31682
+    },
+    {
+      "acc": 0.6947525,
+      "epoch": 1.4922628107559615,
+      "grad_norm": 3.375,
+      "learning_rate": 1.660929322987319e-06,
+      "loss": 1.36562233,
+      "memory(GiB)": 123.79,
+      "step": 58825,
+      "train_speed(iter/s)": 1.316833
+    },
+    {
+      "acc": 0.67895613,
+      "epoch": 1.4923896499238964,
+      "grad_norm": 4.0,
+      "learning_rate": 1.660148876383857e-06,
+      "loss": 1.39221935,
+      "memory(GiB)": 123.79,
+      "step": 58830,
+      "train_speed(iter/s)": 1.316845
+    },
+    {
+      "acc": 0.68912597,
+      "epoch": 1.4925164890918317,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.6593685766819584e-06,
+      "loss": 1.34992313,
+      "memory(GiB)": 123.79,
+      "step": 58835,
+      "train_speed(iter/s)": 1.316858
+    },
+    {
+      "acc": 0.68410091,
+      "epoch": 1.4926433282597666,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.658588423915945e-06,
+      "loss": 1.34456577,
+      "memory(GiB)": 123.79,
+      "step": 58840,
+      "train_speed(iter/s)": 1.316872
+    },
+    {
+      "acc": 0.68963056,
+      "epoch": 1.4927701674277016,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.6578084181201293e-06,
+      "loss": 1.37675848,
+      "memory(GiB)": 123.79,
+      "step": 58845,
+      "train_speed(iter/s)": 1.316885
+    },
+    {
+      "acc": 0.68306322,
+      "epoch": 1.4928970065956366,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.6570285593288242e-06,
+      "loss": 1.42296953,
+      "memory(GiB)": 123.79,
+      "step": 58850,
+      "train_speed(iter/s)": 1.316898
+    },
+    {
+      "acc": 0.68548069,
+      "epoch": 1.4930238457635718,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.6562488475763267e-06,
+      "loss": 1.40037537,
+      "memory(GiB)": 123.79,
+      "step": 58855,
+      "train_speed(iter/s)": 1.316912
+    },
+    {
+      "acc": 0.67732773,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.6554692828969321e-06,
+      "loss": 1.40266171,
+      "memory(GiB)": 123.79,
+      "step": 58860,
+      "train_speed(iter/s)": 1.316925
+    },
+    {
+      "acc": 0.67059088,
+      "epoch": 1.493277524099442,
+      "grad_norm": 4.125,
+      "learning_rate": 1.6546898653249326e-06,
+      "loss": 1.39669209,
+      "memory(GiB)": 123.79,
+      "step": 58865,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.67628961,
+      "epoch": 1.493404363267377,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.65391059489461e-06,
+      "loss": 1.43634319,
+      "memory(GiB)": 123.79,
+      "step": 58870,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.67575178,
+      "epoch": 1.493531202435312,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.6531314716402369e-06,
+      "loss": 1.42028837,
+      "memory(GiB)": 123.79,
+      "step": 58875,
+      "train_speed(iter/s)": 1.316966
+    },
+    {
+      "acc": 0.68546343,
+      "epoch": 1.493658041603247,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.652352495596083e-06,
+      "loss": 1.38132429,
+      "memory(GiB)": 123.79,
+      "step": 58880,
+      "train_speed(iter/s)": 1.316979
+    },
+    {
+      "acc": 0.68650312,
+      "epoch": 1.4937848807711822,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.6515736667964144e-06,
+      "loss": 1.38802681,
+      "memory(GiB)": 123.79,
+      "step": 58885,
+      "train_speed(iter/s)": 1.316992
+    },
+    {
+      "acc": 0.68514538,
+      "epoch": 1.4939117199391172,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.6507949852754867e-06,
+      "loss": 1.45164318,
+      "memory(GiB)": 123.79,
+      "step": 58890,
+      "train_speed(iter/s)": 1.317006
+    },
+    {
+      "acc": 0.67816653,
+      "epoch": 1.4940385591070522,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.6500164510675453e-06,
+      "loss": 1.4109211,
+      "memory(GiB)": 123.79,
+      "step": 58895,
+      "train_speed(iter/s)": 1.317019
+    },
+    {
+      "acc": 0.70046329,
+      "epoch": 1.4941653982749874,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.649238064206839e-06,
+      "loss": 1.34214582,
+      "memory(GiB)": 123.79,
+      "step": 58900,
+      "train_speed(iter/s)": 1.317033
+    },
+    {
+      "acc": 0.68409429,
+      "epoch": 1.4942922374429224,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.6484598247276023e-06,
+      "loss": 1.40221853,
+      "memory(GiB)": 123.79,
+      "step": 58905,
+      "train_speed(iter/s)": 1.317046
+    },
+    {
+      "acc": 0.67015319,
+      "epoch": 1.4944190766108574,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.6476817326640682e-06,
+      "loss": 1.4081214,
+      "memory(GiB)": 123.79,
+      "step": 58910,
+      "train_speed(iter/s)": 1.317059
+    },
+    {
+      "acc": 0.69862866,
+      "epoch": 1.4945459157787924,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.646903788050455e-06,
+      "loss": 1.34596024,
+      "memory(GiB)": 123.79,
+      "step": 58915,
+      "train_speed(iter/s)": 1.317073
+    },
+    {
+      "acc": 0.68271179,
+      "epoch": 1.4946727549467276,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.6461259909209853e-06,
+      "loss": 1.37684307,
+      "memory(GiB)": 123.79,
+      "step": 58920,
+      "train_speed(iter/s)": 1.317086
+    },
+    {
+      "acc": 0.7002811,
+      "epoch": 1.4947995941146626,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6453483413098687e-06,
+      "loss": 1.3870326,
+      "memory(GiB)": 123.79,
+      "step": 58925,
+      "train_speed(iter/s)": 1.317099
+    },
+    {
+      "acc": 0.68254128,
+      "epoch": 1.4949264332825978,
+      "grad_norm": 3.0,
+      "learning_rate": 1.6445708392513093e-06,
+      "loss": 1.43946228,
+      "memory(GiB)": 123.79,
+      "step": 58930,
+      "train_speed(iter/s)": 1.317113
+    },
+    {
+      "acc": 0.68552866,
+      "epoch": 1.4950532724505328,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.643793484779505e-06,
+      "loss": 1.43767586,
+      "memory(GiB)": 123.79,
+      "step": 58935,
+      "train_speed(iter/s)": 1.317126
+    },
+    {
+      "acc": 0.69952383,
+      "epoch": 1.4951801116184678,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.6430162779286484e-06,
+      "loss": 1.35917826,
+      "memory(GiB)": 123.79,
+      "step": 58940,
+      "train_speed(iter/s)": 1.317139
+    },
+    {
+      "acc": 0.69578214,
+      "epoch": 1.4953069507864027,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.6422392187329233e-06,
+      "loss": 1.34357166,
+      "memory(GiB)": 123.79,
+      "step": 58945,
+      "train_speed(iter/s)": 1.31715
+    },
+    {
+      "acc": 0.67583141,
+      "epoch": 1.495433789954338,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.6414623072265085e-06,
+      "loss": 1.47315073,
+      "memory(GiB)": 123.79,
+      "step": 58950,
+      "train_speed(iter/s)": 1.317162
+    },
+    {
+      "acc": 0.69151449,
+      "epoch": 1.495560629122273,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.6406855434435765e-06,
+      "loss": 1.38067961,
+      "memory(GiB)": 123.79,
+      "step": 58955,
+      "train_speed(iter/s)": 1.317174
+    },
+    {
+      "acc": 0.67811661,
+      "epoch": 1.495687468290208,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.6399089274182922e-06,
+      "loss": 1.41409378,
+      "memory(GiB)": 123.79,
+      "step": 58960,
+      "train_speed(iter/s)": 1.317187
+    },
+    {
+      "acc": 0.69136658,
+      "epoch": 1.4958143074581431,
+      "grad_norm": 2.625,
+      "learning_rate": 1.6391324591848156e-06,
+      "loss": 1.37252655,
+      "memory(GiB)": 123.79,
+      "step": 58965,
+      "train_speed(iter/s)": 1.317199
+    },
+    {
+      "acc": 0.69265823,
+      "epoch": 1.4959411466260781,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6383561387772984e-06,
+      "loss": 1.39001446,
+      "memory(GiB)": 123.79,
+      "step": 58970,
+      "train_speed(iter/s)": 1.317211
+    },
+    {
+      "acc": 0.68418412,
+      "epoch": 1.4960679857940131,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.6375799662298868e-06,
+      "loss": 1.39513178,
+      "memory(GiB)": 123.79,
+      "step": 58975,
+      "train_speed(iter/s)": 1.317223
+    },
+    {
+      "acc": 0.67699761,
+      "epoch": 1.4961948249619481,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.6368039415767201e-06,
+      "loss": 1.43046818,
+      "memory(GiB)": 123.79,
+      "step": 58980,
+      "train_speed(iter/s)": 1.317236
+    },
+    {
+      "acc": 0.68122954,
+      "epoch": 1.4963216641298833,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.6360280648519305e-06,
+      "loss": 1.41242809,
+      "memory(GiB)": 123.79,
+      "step": 58985,
+      "train_speed(iter/s)": 1.317248
+    },
+    {
+      "acc": 0.69332924,
+      "epoch": 1.4964485032978183,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.6352523360896488e-06,
+      "loss": 1.34684429,
+      "memory(GiB)": 123.79,
+      "step": 58990,
+      "train_speed(iter/s)": 1.31726
+    },
+    {
+      "acc": 0.68095374,
+      "epoch": 1.4965753424657535,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.63447675532399e-06,
+      "loss": 1.40538177,
+      "memory(GiB)": 123.79,
+      "step": 58995,
+      "train_speed(iter/s)": 1.317273
+    },
+    {
+      "acc": 0.68961358,
+      "epoch": 1.4967021816336885,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.6337013225890698e-06,
+      "loss": 1.44947414,
+      "memory(GiB)": 123.79,
+      "step": 59000,
+      "train_speed(iter/s)": 1.317285
+    },
+    {
+      "epoch": 1.4967021816336885,
+      "eval_acc": 0.6750833756215321,
+      "eval_loss": 1.3563600778579712,
+      "eval_runtime": 69.8983,
+      "eval_samples_per_second": 91.132,
+      "eval_steps_per_second": 22.79,
+      "step": 59000
+    },
+    {
+      "acc": 0.692483,
+      "epoch": 1.4968290208016235,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.6329260379189932e-06,
+      "loss": 1.34462566,
+      "memory(GiB)": 123.79,
+      "step": 59005,
+      "train_speed(iter/s)": 1.314878
+    },
+    {
+      "acc": 0.67910762,
+      "epoch": 1.4969558599695585,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.6321509013478653e-06,
+      "loss": 1.36974325,
+      "memory(GiB)": 123.79,
+      "step": 59010,
+      "train_speed(iter/s)": 1.314889
+    },
+    {
+      "acc": 0.68610058,
+      "epoch": 1.4970826991374937,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.6313759129097757e-06,
+      "loss": 1.42271709,
+      "memory(GiB)": 123.79,
+      "step": 59015,
+      "train_speed(iter/s)": 1.3149
+    },
+    {
+      "acc": 0.68001909,
+      "epoch": 1.4972095383054287,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.6306010726388117e-06,
+      "loss": 1.4354991,
+      "memory(GiB)": 123.79,
+      "step": 59020,
+      "train_speed(iter/s)": 1.314911
+    },
+    {
+      "acc": 0.68600507,
+      "epoch": 1.497336377473364,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.6298263805690573e-06,
+      "loss": 1.45220499,
+      "memory(GiB)": 123.79,
+      "step": 59025,
+      "train_speed(iter/s)": 1.314922
+    },
+    {
+      "acc": 0.69003296,
+      "epoch": 1.4974632166412989,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.629051836734587e-06,
+      "loss": 1.36280298,
+      "memory(GiB)": 123.79,
+      "step": 59030,
+      "train_speed(iter/s)": 1.314934
+    },
+    {
+      "acc": 0.7056684,
+      "epoch": 1.4975900558092339,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.6282774411694641e-06,
+      "loss": 1.33694439,
+      "memory(GiB)": 123.79,
+      "step": 59035,
+      "train_speed(iter/s)": 1.314945
+    },
+    {
+      "acc": 0.68647966,
+      "epoch": 1.4977168949771689,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.6275031939077545e-06,
+      "loss": 1.39741211,
+      "memory(GiB)": 123.79,
+      "step": 59040,
+      "train_speed(iter/s)": 1.314956
+    },
+    {
+      "acc": 0.69763699,
+      "epoch": 1.497843734145104,
+      "grad_norm": 3.0,
+      "learning_rate": 1.6267290949835119e-06,
+      "loss": 1.32321987,
+      "memory(GiB)": 123.79,
+      "step": 59045,
+      "train_speed(iter/s)": 1.314967
+    },
+    {
+      "acc": 0.67891083,
+      "epoch": 1.497970573313039,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.6259551444307852e-06,
+      "loss": 1.37762108,
+      "memory(GiB)": 123.79,
+      "step": 59050,
+      "train_speed(iter/s)": 1.314977
+    },
+    {
+      "acc": 0.67676368,
+      "epoch": 1.498097412480974,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.6251813422836127e-06,
+      "loss": 1.40360546,
+      "memory(GiB)": 123.79,
+      "step": 59055,
+      "train_speed(iter/s)": 1.314988
+    },
+    {
+      "acc": 0.68880558,
+      "epoch": 1.4982242516489093,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.6244076885760334e-06,
+      "loss": 1.32181292,
+      "memory(GiB)": 123.79,
+      "step": 59060,
+      "train_speed(iter/s)": 1.314997
+    },
+    {
+      "acc": 0.68148184,
+      "epoch": 1.4983510908168443,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.6236341833420755e-06,
+      "loss": 1.45670853,
+      "memory(GiB)": 123.79,
+      "step": 59065,
+      "train_speed(iter/s)": 1.315008
+    },
+    {
+      "acc": 0.67714081,
+      "epoch": 1.4984779299847792,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.6228608266157596e-06,
+      "loss": 1.37586956,
+      "memory(GiB)": 123.79,
+      "step": 59070,
+      "train_speed(iter/s)": 1.31502
+    },
+    {
+      "acc": 0.69883394,
+      "epoch": 1.4986047691527142,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.6220876184311034e-06,
+      "loss": 1.34673223,
+      "memory(GiB)": 123.79,
+      "step": 59075,
+      "train_speed(iter/s)": 1.315032
+    },
+    {
+      "acc": 0.68173742,
+      "epoch": 1.4987316083206494,
+      "grad_norm": 5.125,
+      "learning_rate": 1.6213145588221146e-06,
+      "loss": 1.34843254,
+      "memory(GiB)": 123.79,
+      "step": 59080,
+      "train_speed(iter/s)": 1.315044
+    },
+    {
+      "acc": 0.69322042,
+      "epoch": 1.4988584474885844,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.620541647822796e-06,
+      "loss": 1.35766878,
+      "memory(GiB)": 123.79,
+      "step": 59085,
+      "train_speed(iter/s)": 1.315057
+    },
+    {
+      "acc": 0.68695712,
+      "epoch": 1.4989852866565196,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.6197688854671444e-06,
+      "loss": 1.41367359,
+      "memory(GiB)": 123.79,
+      "step": 59090,
+      "train_speed(iter/s)": 1.315069
+    },
+    {
+      "acc": 0.70238485,
+      "epoch": 1.4991121258244546,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.6189962717891484e-06,
+      "loss": 1.30352154,
+      "memory(GiB)": 123.79,
+      "step": 59095,
+      "train_speed(iter/s)": 1.315081
+    },
+    {
+      "acc": 0.67061653,
+      "epoch": 1.4992389649923896,
+      "grad_norm": 3.125,
+      "learning_rate": 1.6182238068227917e-06,
+      "loss": 1.4746253,
+      "memory(GiB)": 123.79,
+      "step": 59100,
+      "train_speed(iter/s)": 1.315093
+    },
+    {
+      "acc": 0.67765274,
+      "epoch": 1.4993658041603246,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.6174514906020505e-06,
+      "loss": 1.40982447,
+      "memory(GiB)": 123.79,
+      "step": 59105,
+      "train_speed(iter/s)": 1.315102
+    },
+    {
+      "acc": 0.69807606,
+      "epoch": 1.4994926433282598,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.6166793231608952e-06,
+      "loss": 1.32372217,
+      "memory(GiB)": 123.79,
+      "step": 59110,
+      "train_speed(iter/s)": 1.315113
+    },
+    {
+      "acc": 0.67917662,
+      "epoch": 1.4996194824961948,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.615907304533288e-06,
+      "loss": 1.4267004,
+      "memory(GiB)": 123.79,
+      "step": 59115,
+      "train_speed(iter/s)": 1.315125
+    },
+    {
+      "acc": 0.68464026,
+      "epoch": 1.4997463216641298,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.6151354347531868e-06,
+      "loss": 1.42052155,
+      "memory(GiB)": 123.79,
+      "step": 59120,
+      "train_speed(iter/s)": 1.315137
+    },
+    {
+      "acc": 0.69962406,
+      "epoch": 1.499873160832065,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.61436371385454e-06,
+      "loss": 1.39396505,
+      "memory(GiB)": 123.79,
+      "step": 59125,
+      "train_speed(iter/s)": 1.315149
+    },
+    {
+      "acc": 0.6831418,
+      "epoch": 1.5,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.6135921418712959e-06,
+      "loss": 1.39157639,
+      "memory(GiB)": 123.79,
+      "step": 59130,
+      "train_speed(iter/s)": 1.315161
+    },
+    {
+      "acc": 0.69109926,
+      "epoch": 1.500126839167935,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.6128207188373867e-06,
+      "loss": 1.38178644,
+      "memory(GiB)": 123.79,
+      "step": 59135,
+      "train_speed(iter/s)": 1.315171
+    },
+    {
+      "acc": 0.6820715,
+      "epoch": 1.50025367833587,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.6120494447867451e-06,
+      "loss": 1.40134773,
+      "memory(GiB)": 123.79,
+      "step": 59140,
+      "train_speed(iter/s)": 1.315184
+    },
+    {
+      "acc": 0.69525747,
+      "epoch": 1.5003805175038052,
+      "grad_norm": 4.875,
+      "learning_rate": 1.6112783197532932e-06,
+      "loss": 1.33634033,
+      "memory(GiB)": 123.79,
+      "step": 59145,
+      "train_speed(iter/s)": 1.315196
+    },
+    {
+      "acc": 0.68155575,
+      "epoch": 1.5005073566717404,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.6105073437709545e-06,
+      "loss": 1.38231745,
+      "memory(GiB)": 123.79,
+      "step": 59150,
+      "train_speed(iter/s)": 1.315208
+    },
+    {
+      "acc": 0.66676526,
+      "epoch": 1.5006341958396754,
+      "grad_norm": 3.125,
+      "learning_rate": 1.6097365168736335e-06,
+      "loss": 1.44543486,
+      "memory(GiB)": 123.79,
+      "step": 59155,
+      "train_speed(iter/s)": 1.31522
+    },
+    {
+      "acc": 0.68036966,
+      "epoch": 1.5007610350076104,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.6089658390952351e-06,
+      "loss": 1.41171198,
+      "memory(GiB)": 123.79,
+      "step": 59160,
+      "train_speed(iter/s)": 1.315232
+    },
+    {
+      "acc": 0.69128561,
+      "epoch": 1.5008878741755454,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.6081953104696612e-06,
+      "loss": 1.39076843,
+      "memory(GiB)": 123.79,
+      "step": 59165,
+      "train_speed(iter/s)": 1.315244
+    },
+    {
+      "acc": 0.69712133,
+      "epoch": 1.5010147133434804,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6074249310308021e-06,
+      "loss": 1.39234676,
+      "memory(GiB)": 123.79,
+      "step": 59170,
+      "train_speed(iter/s)": 1.315255
+    },
+    {
+      "acc": 0.69528265,
+      "epoch": 1.5011415525114156,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.6066547008125399e-06,
+      "loss": 1.34526291,
+      "memory(GiB)": 123.79,
+      "step": 59175,
+      "train_speed(iter/s)": 1.315266
+    },
+    {
+      "acc": 0.68201833,
+      "epoch": 1.5012683916793506,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.6058846198487522e-06,
+      "loss": 1.3957406,
+      "memory(GiB)": 123.79,
+      "step": 59180,
+      "train_speed(iter/s)": 1.315278
+    },
+    {
+      "acc": 0.67600832,
+      "epoch": 1.5013952308472858,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6051146881733142e-06,
+      "loss": 1.36026287,
+      "memory(GiB)": 123.79,
+      "step": 59185,
+      "train_speed(iter/s)": 1.31529
+    },
+    {
+      "acc": 0.69162006,
+      "epoch": 1.5015220700152208,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.6043449058200916e-06,
+      "loss": 1.39760761,
+      "memory(GiB)": 123.79,
+      "step": 59190,
+      "train_speed(iter/s)": 1.315302
+    },
+    {
+      "acc": 0.68696918,
+      "epoch": 1.5016489091831557,
+      "grad_norm": 3.125,
+      "learning_rate": 1.6035752728229364e-06,
+      "loss": 1.37469244,
+      "memory(GiB)": 123.79,
+      "step": 59195,
+      "train_speed(iter/s)": 1.315314
+    },
+    {
+      "acc": 0.68599868,
+      "epoch": 1.5017757483510907,
+      "grad_norm": 2.75,
+      "learning_rate": 1.6028057892157067e-06,
+      "loss": 1.46141434,
+      "memory(GiB)": 123.79,
+      "step": 59200,
+      "train_speed(iter/s)": 1.315326
+    },
+    {
+      "acc": 0.6885498,
+      "epoch": 1.5019025875190257,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.602036455032246e-06,
+      "loss": 1.37296066,
+      "memory(GiB)": 123.79,
+      "step": 59205,
+      "train_speed(iter/s)": 1.315339
+    },
+    {
+      "acc": 0.68712139,
+      "epoch": 1.502029426686961,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.6012672703063925e-06,
+      "loss": 1.41904488,
+      "memory(GiB)": 123.79,
+      "step": 59210,
+      "train_speed(iter/s)": 1.315351
+    },
+    {
+      "acc": 0.68188696,
+      "epoch": 1.5021562658548961,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.600498235071979e-06,
+      "loss": 1.43358021,
+      "memory(GiB)": 123.79,
+      "step": 59215,
+      "train_speed(iter/s)": 1.315363
+    },
+    {
+      "acc": 0.66795111,
+      "epoch": 1.5022831050228311,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.5997293493628301e-06,
+      "loss": 1.44600449,
+      "memory(GiB)": 123.79,
+      "step": 59220,
+      "train_speed(iter/s)": 1.315376
+    },
+    {
+      "acc": 0.68693447,
+      "epoch": 1.5024099441907661,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.598960613212766e-06,
+      "loss": 1.39533262,
+      "memory(GiB)": 123.79,
+      "step": 59225,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.69323292,
+      "epoch": 1.5025367833587011,
+      "grad_norm": 3.125,
+      "learning_rate": 1.598192026655599e-06,
+      "loss": 1.41474171,
+      "memory(GiB)": 123.79,
+      "step": 59230,
+      "train_speed(iter/s)": 1.3154
+    },
+    {
+      "acc": 0.68015909,
+      "epoch": 1.502663622526636,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.5974235897251344e-06,
+      "loss": 1.38725967,
+      "memory(GiB)": 123.79,
+      "step": 59235,
+      "train_speed(iter/s)": 1.315412
+    },
+    {
+      "acc": 0.6764101,
+      "epoch": 1.5027904616945713,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.5966553024551717e-06,
+      "loss": 1.41960983,
+      "memory(GiB)": 123.79,
+      "step": 59240,
+      "train_speed(iter/s)": 1.315425
+    },
+    {
+      "acc": 0.6828083,
+      "epoch": 1.5029173008625063,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.5958871648795032e-06,
+      "loss": 1.41359615,
+      "memory(GiB)": 123.79,
+      "step": 59245,
+      "train_speed(iter/s)": 1.315437
+    },
+    {
+      "acc": 0.69029188,
+      "epoch": 1.5030441400304415,
+      "grad_norm": 3.0,
+      "learning_rate": 1.5951191770319164e-06,
+      "loss": 1.33567619,
+      "memory(GiB)": 123.79,
+      "step": 59250,
+      "train_speed(iter/s)": 1.31545
+    },
+    {
+      "acc": 0.68615713,
+      "epoch": 1.5031709791983765,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.594351338946189e-06,
+      "loss": 1.3638978,
+      "memory(GiB)": 123.79,
+      "step": 59255,
+      "train_speed(iter/s)": 1.315463
+    },
+    {
+      "acc": 0.69215288,
+      "epoch": 1.5032978183663115,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.5935836506560953e-06,
+      "loss": 1.37950268,
+      "memory(GiB)": 123.79,
+      "step": 59260,
+      "train_speed(iter/s)": 1.315476
+    },
+    {
+      "acc": 0.69085426,
+      "epoch": 1.5034246575342465,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.5928161121954012e-06,
+      "loss": 1.35364532,
+      "memory(GiB)": 123.79,
+      "step": 59265,
+      "train_speed(iter/s)": 1.315489
+    },
+    {
+      "acc": 0.69523544,
+      "epoch": 1.5035514967021817,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.592048723597866e-06,
+      "loss": 1.33664074,
+      "memory(GiB)": 123.79,
+      "step": 59270,
+      "train_speed(iter/s)": 1.315501
+    },
+    {
+      "acc": 0.70734725,
+      "epoch": 1.5036783358701167,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.591281484897244e-06,
+      "loss": 1.33132401,
+      "memory(GiB)": 123.79,
+      "step": 59275,
+      "train_speed(iter/s)": 1.315513
+    },
+    {
+      "acc": 0.69437294,
+      "epoch": 1.5038051750380519,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.5905143961272807e-06,
+      "loss": 1.39369001,
+      "memory(GiB)": 123.79,
+      "step": 59280,
+      "train_speed(iter/s)": 1.315525
+    },
+    {
+      "acc": 0.69709678,
+      "epoch": 1.5039320142059869,
+      "grad_norm": 3.625,
+      "learning_rate": 1.5897474573217153e-06,
+      "loss": 1.29143982,
+      "memory(GiB)": 123.79,
+      "step": 59285,
+      "train_speed(iter/s)": 1.315536
+    },
+    {
+      "acc": 0.68442068,
+      "epoch": 1.5040588533739219,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.588980668514285e-06,
+      "loss": 1.39644985,
+      "memory(GiB)": 123.79,
+      "step": 59290,
+      "train_speed(iter/s)": 1.315548
+    },
+    {
+      "acc": 0.68362885,
+      "epoch": 1.5041856925418569,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.5882140297387127e-06,
+      "loss": 1.36531172,
+      "memory(GiB)": 123.79,
+      "step": 59295,
+      "train_speed(iter/s)": 1.315558
+    },
+    {
+      "acc": 0.68599033,
+      "epoch": 1.5043125317097918,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.5874475410287189e-06,
+      "loss": 1.42504635,
+      "memory(GiB)": 123.79,
+      "step": 59300,
+      "train_speed(iter/s)": 1.31557
+    },
+    {
+      "acc": 0.68854713,
+      "epoch": 1.504439370877727,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.586681202418019e-06,
+      "loss": 1.39468565,
+      "memory(GiB)": 123.79,
+      "step": 59305,
+      "train_speed(iter/s)": 1.315581
+    },
+    {
+      "acc": 0.685606,
+      "epoch": 1.5045662100456623,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.5859150139403212e-06,
+      "loss": 1.45771446,
+      "memory(GiB)": 123.79,
+      "step": 59310,
+      "train_speed(iter/s)": 1.315592
+    },
+    {
+      "acc": 0.69314923,
+      "epoch": 1.5046930492135973,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.585148975629322e-06,
+      "loss": 1.36122284,
+      "memory(GiB)": 123.79,
+      "step": 59315,
+      "train_speed(iter/s)": 1.315603
+    },
+    {
+      "acc": 0.68621402,
+      "epoch": 1.5048198883815322,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.5843830875187155e-06,
+      "loss": 1.38169231,
+      "memory(GiB)": 123.79,
+      "step": 59320,
+      "train_speed(iter/s)": 1.315615
+    },
+    {
+      "acc": 0.68789868,
+      "epoch": 1.5049467275494672,
+      "grad_norm": 3.375,
+      "learning_rate": 1.583617349642192e-06,
+      "loss": 1.43120708,
+      "memory(GiB)": 123.79,
+      "step": 59325,
+      "train_speed(iter/s)": 1.315627
+    },
+    {
+      "acc": 0.68234777,
+      "epoch": 1.5050735667174022,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.5828517620334322e-06,
+      "loss": 1.41719913,
+      "memory(GiB)": 123.79,
+      "step": 59330,
+      "train_speed(iter/s)": 1.315639
+    },
+    {
+      "acc": 0.69751935,
+      "epoch": 1.5052004058853374,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.5820863247261054e-06,
+      "loss": 1.37203007,
+      "memory(GiB)": 123.79,
+      "step": 59335,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.69031105,
+      "epoch": 1.5053272450532724,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.5813210377538834e-06,
+      "loss": 1.39535332,
+      "memory(GiB)": 123.79,
+      "step": 59340,
+      "train_speed(iter/s)": 1.315663
+    },
+    {
+      "acc": 0.68517466,
+      "epoch": 1.5054540842212076,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.5805559011504252e-06,
+      "loss": 1.41304474,
+      "memory(GiB)": 123.79,
+      "step": 59345,
+      "train_speed(iter/s)": 1.315675
+    },
+    {
+      "acc": 0.69299469,
+      "epoch": 1.5055809233891426,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.5797909149493873e-06,
+      "loss": 1.29743433,
+      "memory(GiB)": 123.79,
+      "step": 59350,
+      "train_speed(iter/s)": 1.315687
+    },
+    {
+      "acc": 0.67083902,
+      "epoch": 1.5057077625570776,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.5790260791844114e-06,
+      "loss": 1.48566437,
+      "memory(GiB)": 123.79,
+      "step": 59355,
+      "train_speed(iter/s)": 1.315699
+    },
+    {
+      "acc": 0.69426889,
+      "epoch": 1.5058346017250126,
+      "grad_norm": 2.875,
+      "learning_rate": 1.5782613938891438e-06,
+      "loss": 1.37308445,
+      "memory(GiB)": 123.79,
+      "step": 59360,
+      "train_speed(iter/s)": 1.315711
+    },
+    {
+      "acc": 0.70313463,
+      "epoch": 1.5059614408929476,
+      "grad_norm": 4.25,
+      "learning_rate": 1.5774968590972172e-06,
+      "loss": 1.36396828,
+      "memory(GiB)": 123.79,
+      "step": 59365,
+      "train_speed(iter/s)": 1.315723
+    },
+    {
+      "acc": 0.69150896,
+      "epoch": 1.5060882800608828,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.5767324748422592e-06,
+      "loss": 1.39236889,
+      "memory(GiB)": 123.79,
+      "step": 59370,
+      "train_speed(iter/s)": 1.315734
+    },
+    {
+      "acc": 0.68706136,
+      "epoch": 1.506215119228818,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.5759682411578909e-06,
+      "loss": 1.36788483,
+      "memory(GiB)": 123.79,
+      "step": 59375,
+      "train_speed(iter/s)": 1.315745
+    },
+    {
+      "acc": 0.69587188,
+      "epoch": 1.506341958396753,
+      "grad_norm": 3.125,
+      "learning_rate": 1.575204158077726e-06,
+      "loss": 1.41396847,
+      "memory(GiB)": 123.79,
+      "step": 59380,
+      "train_speed(iter/s)": 1.315757
+    },
+    {
+      "acc": 0.69766855,
+      "epoch": 1.506468797564688,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.574440225635373e-06,
+      "loss": 1.37710953,
+      "memory(GiB)": 123.79,
+      "step": 59385,
+      "train_speed(iter/s)": 1.315769
+    },
+    {
+      "acc": 0.68945646,
+      "epoch": 1.506595636732623,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.5736764438644332e-06,
+      "loss": 1.39807711,
+      "memory(GiB)": 123.79,
+      "step": 59390,
+      "train_speed(iter/s)": 1.315779
+    },
+    {
+      "acc": 0.69119959,
+      "epoch": 1.506722475900558,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.5729128127985004e-06,
+      "loss": 1.33622589,
+      "memory(GiB)": 123.79,
+      "step": 59395,
+      "train_speed(iter/s)": 1.315791
+    },
+    {
+      "acc": 0.67488256,
+      "epoch": 1.5068493150684932,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.5721493324711633e-06,
+      "loss": 1.42634583,
+      "memory(GiB)": 123.79,
+      "step": 59400,
+      "train_speed(iter/s)": 1.315803
+    },
+    {
+      "acc": 0.68427439,
+      "epoch": 1.5069761542364282,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.5713860029160028e-06,
+      "loss": 1.39922714,
+      "memory(GiB)": 123.79,
+      "step": 59405,
+      "train_speed(iter/s)": 1.315813
+    },
+    {
+      "acc": 0.67688479,
+      "epoch": 1.5071029934043634,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.5706228241665932e-06,
+      "loss": 1.43908396,
+      "memory(GiB)": 123.79,
+      "step": 59410,
+      "train_speed(iter/s)": 1.315825
+    },
+    {
+      "acc": 0.69400358,
+      "epoch": 1.5072298325722984,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.5698597962565032e-06,
+      "loss": 1.39130802,
+      "memory(GiB)": 123.79,
+      "step": 59415,
+      "train_speed(iter/s)": 1.315836
+    },
+    {
+      "acc": 0.69946756,
+      "epoch": 1.5073566717402334,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.5690969192192933e-06,
+      "loss": 1.37493877,
+      "memory(GiB)": 123.79,
+      "step": 59420,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.69301577,
+      "epoch": 1.5074835109081683,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.5683341930885183e-06,
+      "loss": 1.37761326,
+      "memory(GiB)": 123.79,
+      "step": 59425,
+      "train_speed(iter/s)": 1.315858
+    },
+    {
+      "acc": 0.67453599,
+      "epoch": 1.5076103500761036,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.567571617897729e-06,
+      "loss": 1.44319973,
+      "memory(GiB)": 123.79,
+      "step": 59430,
+      "train_speed(iter/s)": 1.31587
+    },
+    {
+      "acc": 0.70175991,
+      "epoch": 1.5077371892440385,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.566809193680463e-06,
+      "loss": 1.34457951,
+      "memory(GiB)": 123.79,
+      "step": 59435,
+      "train_speed(iter/s)": 1.315881
+    },
+    {
+      "acc": 0.67892532,
+      "epoch": 1.5078640284119738,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.566046920470257e-06,
+      "loss": 1.46032858,
+      "memory(GiB)": 123.79,
+      "step": 59440,
+      "train_speed(iter/s)": 1.315892
+    },
+    {
+      "acc": 0.70120668,
+      "epoch": 1.5079908675799087,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.5652847983006376e-06,
+      "loss": 1.34290981,
+      "memory(GiB)": 123.79,
+      "step": 59445,
+      "train_speed(iter/s)": 1.315903
+    },
+    {
+      "acc": 0.69514284,
+      "epoch": 1.5081177067478437,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.564522827205131e-06,
+      "loss": 1.3516552,
+      "memory(GiB)": 123.79,
+      "step": 59450,
+      "train_speed(iter/s)": 1.315915
+    },
+    {
+      "acc": 0.68584666,
+      "epoch": 1.5082445459157787,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.5637610072172464e-06,
+      "loss": 1.3756815,
+      "memory(GiB)": 123.79,
+      "step": 59455,
+      "train_speed(iter/s)": 1.315926
+    },
+    {
+      "acc": 0.68379073,
+      "epoch": 1.5083713850837137,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.5629993383704933e-06,
+      "loss": 1.45334606,
+      "memory(GiB)": 123.79,
+      "step": 59460,
+      "train_speed(iter/s)": 1.315938
+    },
+    {
+      "acc": 0.69214253,
+      "epoch": 1.508498224251649,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.5622378206983764e-06,
+      "loss": 1.40193586,
+      "memory(GiB)": 123.79,
+      "step": 59465,
+      "train_speed(iter/s)": 1.315949
+    },
+    {
+      "acc": 0.6961133,
+      "epoch": 1.5086250634195841,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.5614764542343896e-06,
+      "loss": 1.36166906,
+      "memory(GiB)": 123.79,
+      "step": 59470,
+      "train_speed(iter/s)": 1.31596
+    },
+    {
+      "acc": 0.68462453,
+      "epoch": 1.5087519025875191,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5607152390120173e-06,
+      "loss": 1.34242134,
+      "memory(GiB)": 123.79,
+      "step": 59475,
+      "train_speed(iter/s)": 1.315972
+    },
+    {
+      "acc": 0.68149099,
+      "epoch": 1.5088787417554541,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.5599541750647457e-06,
+      "loss": 1.42025852,
+      "memory(GiB)": 123.79,
+      "step": 59480,
+      "train_speed(iter/s)": 1.315983
+    },
+    {
+      "acc": 0.68418994,
+      "epoch": 1.509005580923389,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.559193262426048e-06,
+      "loss": 1.3631691,
+      "memory(GiB)": 123.79,
+      "step": 59485,
+      "train_speed(iter/s)": 1.315994
+    },
+    {
+      "acc": 0.68939166,
+      "epoch": 1.509132420091324,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.5584325011293943e-06,
+      "loss": 1.38271122,
+      "memory(GiB)": 123.79,
+      "step": 59490,
+      "train_speed(iter/s)": 1.316005
+    },
+    {
+      "acc": 0.67711449,
+      "epoch": 1.5092592592592593,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.5576718912082417e-06,
+      "loss": 1.43066626,
+      "memory(GiB)": 123.79,
+      "step": 59495,
+      "train_speed(iter/s)": 1.316016
+    },
+    {
+      "acc": 0.69125566,
+      "epoch": 1.5093860984271943,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.5569114326960494e-06,
+      "loss": 1.40360947,
+      "memory(GiB)": 123.79,
+      "step": 59500,
+      "train_speed(iter/s)": 1.316027
+    },
+    {
+      "acc": 0.69036579,
+      "epoch": 1.5095129375951295,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.5561511256262651e-06,
+      "loss": 1.36027908,
+      "memory(GiB)": 123.79,
+      "step": 59505,
+      "train_speed(iter/s)": 1.316039
+    },
+    {
+      "acc": 0.68807626,
+      "epoch": 1.5096397767630645,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.55539097003233e-06,
+      "loss": 1.39431953,
+      "memory(GiB)": 123.79,
+      "step": 59510,
+      "train_speed(iter/s)": 1.31605
+    },
+    {
+      "acc": 0.69668655,
+      "epoch": 1.5097666159309995,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.5546309659476788e-06,
+      "loss": 1.37217445,
+      "memory(GiB)": 123.79,
+      "step": 59515,
+      "train_speed(iter/s)": 1.316062
+    },
+    {
+      "acc": 0.70222573,
+      "epoch": 1.5098934550989345,
+      "grad_norm": 2.875,
+      "learning_rate": 1.55387111340574e-06,
+      "loss": 1.37251081,
+      "memory(GiB)": 123.79,
+      "step": 59520,
+      "train_speed(iter/s)": 1.316073
+    },
+    {
+      "acc": 0.69505405,
+      "epoch": 1.5100202942668695,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.553111412439936e-06,
+      "loss": 1.36986427,
+      "memory(GiB)": 123.79,
+      "step": 59525,
+      "train_speed(iter/s)": 1.316085
+    },
+    {
+      "acc": 0.68632669,
+      "epoch": 1.5101471334348047,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.5523518630836809e-06,
+      "loss": 1.44103966,
+      "memory(GiB)": 123.79,
+      "step": 59530,
+      "train_speed(iter/s)": 1.316096
+    },
+    {
+      "acc": 0.70297308,
+      "epoch": 1.5102739726027399,
+      "grad_norm": 3.75,
+      "learning_rate": 1.551592465370384e-06,
+      "loss": 1.31091366,
+      "memory(GiB)": 123.79,
+      "step": 59535,
+      "train_speed(iter/s)": 1.316107
+    },
+    {
+      "acc": 0.66630349,
+      "epoch": 1.5104008117706749,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.5508332193334457e-06,
+      "loss": 1.38704395,
+      "memory(GiB)": 123.79,
+      "step": 59540,
+      "train_speed(iter/s)": 1.316119
+    },
+    {
+      "acc": 0.6808311,
+      "epoch": 1.5105276509386099,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.5500741250062628e-06,
+      "loss": 1.48263655,
+      "memory(GiB)": 123.79,
+      "step": 59545,
+      "train_speed(iter/s)": 1.316131
+    },
+    {
+      "acc": 0.71275215,
+      "epoch": 1.5106544901065448,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.549315182422222e-06,
+      "loss": 1.34635963,
+      "memory(GiB)": 123.79,
+      "step": 59550,
+      "train_speed(iter/s)": 1.316143
+    },
+    {
+      "acc": 0.69755535,
+      "epoch": 1.5107813292744798,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.5485563916147062e-06,
+      "loss": 1.40320587,
+      "memory(GiB)": 123.79,
+      "step": 59555,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.66435008,
+      "epoch": 1.510908168442415,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.5477977526170895e-06,
+      "loss": 1.39361706,
+      "memory(GiB)": 123.79,
+      "step": 59560,
+      "train_speed(iter/s)": 1.316166
+    },
+    {
+      "acc": 0.6834794,
+      "epoch": 1.51103500761035,
+      "grad_norm": 3.125,
+      "learning_rate": 1.5470392654627392e-06,
+      "loss": 1.42930489,
+      "memory(GiB)": 123.79,
+      "step": 59565,
+      "train_speed(iter/s)": 1.316178
+    },
+    {
+      "acc": 0.69158049,
+      "epoch": 1.5111618467782852,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.5462809301850212e-06,
+      "loss": 1.38263769,
+      "memory(GiB)": 123.79,
+      "step": 59570,
+      "train_speed(iter/s)": 1.316189
+    },
+    {
+      "acc": 0.68776646,
+      "epoch": 1.5112886859462202,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.5455227468172862e-06,
+      "loss": 1.42415009,
+      "memory(GiB)": 123.79,
+      "step": 59575,
+      "train_speed(iter/s)": 1.316201
+    },
+    {
+      "acc": 0.67986631,
+      "epoch": 1.5114155251141552,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.5447647153928842e-06,
+      "loss": 1.43905525,
+      "memory(GiB)": 123.79,
+      "step": 59580,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.68927693,
+      "epoch": 1.5115423642820902,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.5440068359451548e-06,
+      "loss": 1.4094635,
+      "memory(GiB)": 123.79,
+      "step": 59585,
+      "train_speed(iter/s)": 1.316224
+    },
+    {
+      "acc": 0.69380655,
+      "epoch": 1.5116692034500254,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.5432491085074381e-06,
+      "loss": 1.38116112,
+      "memory(GiB)": 123.79,
+      "step": 59590,
+      "train_speed(iter/s)": 1.316237
+    },
+    {
+      "acc": 0.66857557,
+      "epoch": 1.5117960426179604,
+      "grad_norm": 2.75,
+      "learning_rate": 1.5424915331130568e-06,
+      "loss": 1.44176464,
+      "memory(GiB)": 123.79,
+      "step": 59595,
+      "train_speed(iter/s)": 1.316246
+    },
+    {
+      "acc": 0.70199904,
+      "epoch": 1.5119228817858956,
+      "grad_norm": 4.125,
+      "learning_rate": 1.5417341097953332e-06,
+      "loss": 1.28317509,
+      "memory(GiB)": 123.79,
+      "step": 59600,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.70445304,
+      "epoch": 1.5120497209538306,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.540976838587585e-06,
+      "loss": 1.39034986,
+      "memory(GiB)": 123.79,
+      "step": 59605,
+      "train_speed(iter/s)": 1.316269
+    },
+    {
+      "acc": 0.68655128,
+      "epoch": 1.5121765601217656,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.5402197195231205e-06,
+      "loss": 1.39206753,
+      "memory(GiB)": 123.79,
+      "step": 59610,
+      "train_speed(iter/s)": 1.316281
+    },
+    {
+      "acc": 0.69615297,
+      "epoch": 1.5123033992897006,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.5394627526352379e-06,
+      "loss": 1.37897892,
+      "memory(GiB)": 123.79,
+      "step": 59615,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.69054193,
+      "epoch": 1.5124302384576356,
+      "grad_norm": 3.125,
+      "learning_rate": 1.5387059379572322e-06,
+      "loss": 1.38111868,
+      "memory(GiB)": 123.79,
+      "step": 59620,
+      "train_speed(iter/s)": 1.316304
+    },
+    {
+      "acc": 0.71182137,
+      "epoch": 1.5125570776255708,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.537949275522394e-06,
+      "loss": 1.3426692,
+      "memory(GiB)": 123.79,
+      "step": 59625,
+      "train_speed(iter/s)": 1.316316
+    },
+    {
+      "acc": 0.67898402,
+      "epoch": 1.512683916793506,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.5371927653640056e-06,
+      "loss": 1.4274724,
+      "memory(GiB)": 123.79,
+      "step": 59630,
+      "train_speed(iter/s)": 1.316328
+    },
+    {
+      "acc": 0.69489241,
+      "epoch": 1.512810755961441,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.5364364075153366e-06,
+      "loss": 1.36836872,
+      "memory(GiB)": 123.79,
+      "step": 59635,
+      "train_speed(iter/s)": 1.316339
+    },
+    {
+      "acc": 0.6952569,
+      "epoch": 1.512937595129376,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.5356802020096595e-06,
+      "loss": 1.3611042,
+      "memory(GiB)": 123.79,
+      "step": 59640,
+      "train_speed(iter/s)": 1.31635
+    },
+    {
+      "acc": 0.69355812,
+      "epoch": 1.513064434297311,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5349241488802346e-06,
+      "loss": 1.37238321,
+      "memory(GiB)": 123.79,
+      "step": 59645,
+      "train_speed(iter/s)": 1.31636
+    },
+    {
+      "acc": 0.68892803,
+      "epoch": 1.513191273465246,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.5341682481603155e-06,
+      "loss": 1.41282654,
+      "memory(GiB)": 123.79,
+      "step": 59650,
+      "train_speed(iter/s)": 1.316371
+    },
+    {
+      "acc": 0.69297428,
+      "epoch": 1.5133181126331812,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.5334124998831512e-06,
+      "loss": 1.37389021,
+      "memory(GiB)": 123.79,
+      "step": 59655,
+      "train_speed(iter/s)": 1.316381
+    },
+    {
+      "acc": 0.66590204,
+      "epoch": 1.5134449518011162,
+      "grad_norm": 2.75,
+      "learning_rate": 1.532656904081982e-06,
+      "loss": 1.436724,
+      "memory(GiB)": 123.79,
+      "step": 59660,
+      "train_speed(iter/s)": 1.316392
+    },
+    {
+      "acc": 0.6972784,
+      "epoch": 1.5135717909690514,
+      "grad_norm": 3.375,
+      "learning_rate": 1.5319014607900428e-06,
+      "loss": 1.35213909,
+      "memory(GiB)": 123.79,
+      "step": 59665,
+      "train_speed(iter/s)": 1.316403
+    },
+    {
+      "acc": 0.6943522,
+      "epoch": 1.5136986301369864,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.5311461700405617e-06,
+      "loss": 1.39647226,
+      "memory(GiB)": 123.79,
+      "step": 59670,
+      "train_speed(iter/s)": 1.316414
+    },
+    {
+      "acc": 0.68594089,
+      "epoch": 1.5138254693049213,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.5303910318667586e-06,
+      "loss": 1.41557236,
+      "memory(GiB)": 123.79,
+      "step": 59675,
+      "train_speed(iter/s)": 1.316426
+    },
+    {
+      "acc": 0.68271275,
+      "epoch": 1.5139523084728563,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.529636046301849e-06,
+      "loss": 1.41027861,
+      "memory(GiB)": 123.79,
+      "step": 59680,
+      "train_speed(iter/s)": 1.316437
+    },
+    {
+      "acc": 0.66928701,
+      "epoch": 1.5140791476407913,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.5288812133790405e-06,
+      "loss": 1.4521452,
+      "memory(GiB)": 123.79,
+      "step": 59685,
+      "train_speed(iter/s)": 1.316448
+    },
+    {
+      "acc": 0.67957306,
+      "epoch": 1.5142059868087265,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5281265331315332e-06,
+      "loss": 1.42527084,
+      "memory(GiB)": 123.79,
+      "step": 59690,
+      "train_speed(iter/s)": 1.31646
+    },
+    {
+      "acc": 0.69314113,
+      "epoch": 1.5143328259766617,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.5273720055925217e-06,
+      "loss": 1.37631216,
+      "memory(GiB)": 123.79,
+      "step": 59695,
+      "train_speed(iter/s)": 1.316471
+    },
+    {
+      "acc": 0.68899927,
+      "epoch": 1.5144596651445967,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.5266176307951936e-06,
+      "loss": 1.40612774,
+      "memory(GiB)": 123.79,
+      "step": 59700,
+      "train_speed(iter/s)": 1.316483
+    },
+    {
+      "acc": 0.66557183,
+      "epoch": 1.5145865043125317,
+      "grad_norm": 2.875,
+      "learning_rate": 1.5258634087727298e-06,
+      "loss": 1.36963024,
+      "memory(GiB)": 123.79,
+      "step": 59705,
+      "train_speed(iter/s)": 1.316494
+    },
+    {
+      "acc": 0.69170818,
+      "epoch": 1.5147133434804667,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.5251093395583045e-06,
+      "loss": 1.38345785,
+      "memory(GiB)": 123.79,
+      "step": 59710,
+      "train_speed(iter/s)": 1.316506
+    },
+    {
+      "acc": 0.6900219,
+      "epoch": 1.5148401826484017,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.5243554231850843e-06,
+      "loss": 1.38325777,
+      "memory(GiB)": 123.79,
+      "step": 59715,
+      "train_speed(iter/s)": 1.316517
+    },
+    {
+      "acc": 0.6852335,
+      "epoch": 1.514967021816337,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.5236016596862302e-06,
+      "loss": 1.37488422,
+      "memory(GiB)": 123.79,
+      "step": 59720,
+      "train_speed(iter/s)": 1.316528
+    },
+    {
+      "acc": 0.68069105,
+      "epoch": 1.515093860984272,
+      "grad_norm": 2.625,
+      "learning_rate": 1.5228480490948943e-06,
+      "loss": 1.35252295,
+      "memory(GiB)": 123.79,
+      "step": 59725,
+      "train_speed(iter/s)": 1.316539
+    },
+    {
+      "acc": 0.6930861,
+      "epoch": 1.5152207001522071,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.5220945914442292e-06,
+      "loss": 1.39354353,
+      "memory(GiB)": 123.79,
+      "step": 59730,
+      "train_speed(iter/s)": 1.31655
+    },
+    {
+      "acc": 0.67998133,
+      "epoch": 1.515347539320142,
+      "grad_norm": 2.875,
+      "learning_rate": 1.52134128676737e-06,
+      "loss": 1.34987717,
+      "memory(GiB)": 123.79,
+      "step": 59735,
+      "train_speed(iter/s)": 1.316561
+    },
+    {
+      "acc": 0.68938432,
+      "epoch": 1.515474378488077,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.5205881350974504e-06,
+      "loss": 1.47221498,
+      "memory(GiB)": 123.79,
+      "step": 59740,
+      "train_speed(iter/s)": 1.316573
+    },
+    {
+      "acc": 0.68048105,
+      "epoch": 1.515601217656012,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.5198351364676012e-06,
+      "loss": 1.39034081,
+      "memory(GiB)": 123.79,
+      "step": 59745,
+      "train_speed(iter/s)": 1.316584
+    },
+    {
+      "acc": 0.68841734,
+      "epoch": 1.5157280568239473,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.5190822909109415e-06,
+      "loss": 1.41445274,
+      "memory(GiB)": 123.79,
+      "step": 59750,
+      "train_speed(iter/s)": 1.316593
+    },
+    {
+      "acc": 0.6831975,
+      "epoch": 1.5158548959918823,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.5183295984605824e-06,
+      "loss": 1.42631617,
+      "memory(GiB)": 123.79,
+      "step": 59755,
+      "train_speed(iter/s)": 1.316605
+    },
+    {
+      "acc": 0.69646044,
+      "epoch": 1.5159817351598175,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.5175770591496303e-06,
+      "loss": 1.35653229,
+      "memory(GiB)": 123.79,
+      "step": 59760,
+      "train_speed(iter/s)": 1.316617
+    },
+    {
+      "acc": 0.68704662,
+      "epoch": 1.5161085743277525,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.5168246730111892e-06,
+      "loss": 1.41484776,
+      "memory(GiB)": 123.79,
+      "step": 59765,
+      "train_speed(iter/s)": 1.316626
+    },
+    {
+      "acc": 0.70621004,
+      "epoch": 1.5162354134956875,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.5160724400783511e-06,
+      "loss": 1.39579544,
+      "memory(GiB)": 123.79,
+      "step": 59770,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.69199705,
+      "epoch": 1.5163622526636225,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.5153203603841992e-06,
+      "loss": 1.39201593,
+      "memory(GiB)": 123.79,
+      "step": 59775,
+      "train_speed(iter/s)": 1.316649
+    },
+    {
+      "acc": 0.68527203,
+      "epoch": 1.5164890918315574,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.5145684339618172e-06,
+      "loss": 1.45688286,
+      "memory(GiB)": 123.79,
+      "step": 59780,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.7029727,
+      "epoch": 1.5166159309994927,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.5138166608442768e-06,
+      "loss": 1.3049017,
+      "memory(GiB)": 123.79,
+      "step": 59785,
+      "train_speed(iter/s)": 1.316672
+    },
+    {
+      "acc": 0.68568468,
+      "epoch": 1.5167427701674279,
+      "grad_norm": 3.375,
+      "learning_rate": 1.5130650410646452e-06,
+      "loss": 1.35889235,
+      "memory(GiB)": 123.79,
+      "step": 59790,
+      "train_speed(iter/s)": 1.316683
+    },
+    {
+      "acc": 0.6968935,
+      "epoch": 1.5168696093353629,
+      "grad_norm": 2.4375,
+      "learning_rate": 1.5123135746559792e-06,
+      "loss": 1.37486362,
+      "memory(GiB)": 123.79,
+      "step": 59795,
+      "train_speed(iter/s)": 1.316695
+    },
+    {
+      "acc": 0.70289893,
+      "epoch": 1.5169964485032978,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.5115622616513343e-06,
+      "loss": 1.336656,
+      "memory(GiB)": 123.79,
+      "step": 59800,
+      "train_speed(iter/s)": 1.316706
+    },
+    {
+      "acc": 0.67356095,
+      "epoch": 1.5171232876712328,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5108111020837564e-06,
+      "loss": 1.51547451,
+      "memory(GiB)": 123.79,
+      "step": 59805,
+      "train_speed(iter/s)": 1.316718
+    },
+    {
+      "acc": 0.68595734,
+      "epoch": 1.5172501268391678,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.5100600959862838e-06,
+      "loss": 1.35793552,
+      "memory(GiB)": 123.79,
+      "step": 59810,
+      "train_speed(iter/s)": 1.316729
+    },
+    {
+      "acc": 0.68236384,
+      "epoch": 1.517376966007103,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.5093092433919497e-06,
+      "loss": 1.39960957,
+      "memory(GiB)": 123.79,
+      "step": 59815,
+      "train_speed(iter/s)": 1.31674
+    },
+    {
+      "acc": 0.68897882,
+      "epoch": 1.517503805175038,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.5085585443337803e-06,
+      "loss": 1.33236942,
+      "memory(GiB)": 123.79,
+      "step": 59820,
+      "train_speed(iter/s)": 1.316752
+    },
+    {
+      "acc": 0.68728156,
+      "epoch": 1.5176306443429732,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.507807998844794e-06,
+      "loss": 1.44444122,
+      "memory(GiB)": 123.79,
+      "step": 59825,
+      "train_speed(iter/s)": 1.316764
+    },
+    {
+      "acc": 0.6875174,
+      "epoch": 1.5177574835109082,
+      "grad_norm": 3.125,
+      "learning_rate": 1.5070576069580039e-06,
+      "loss": 1.36129961,
+      "memory(GiB)": 123.79,
+      "step": 59830,
+      "train_speed(iter/s)": 1.316775
+    },
+    {
+      "acc": 0.69530382,
+      "epoch": 1.5178843226788432,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.5063073687064144e-06,
+      "loss": 1.40940447,
+      "memory(GiB)": 123.79,
+      "step": 59835,
+      "train_speed(iter/s)": 1.316787
+    },
+    {
+      "acc": 0.70221558,
+      "epoch": 1.5180111618467782,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.5055572841230253e-06,
+      "loss": 1.36850891,
+      "memory(GiB)": 123.79,
+      "step": 59840,
+      "train_speed(iter/s)": 1.316798
+    },
+    {
+      "acc": 0.69390678,
+      "epoch": 1.5181380010147132,
+      "grad_norm": 3.375,
+      "learning_rate": 1.5048073532408287e-06,
+      "loss": 1.38234558,
+      "memory(GiB)": 123.79,
+      "step": 59845,
+      "train_speed(iter/s)": 1.31681
+    },
+    {
+      "acc": 0.6983098,
+      "epoch": 1.5182648401826484,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.5040575760928094e-06,
+      "loss": 1.36247826,
+      "memory(GiB)": 123.79,
+      "step": 59850,
+      "train_speed(iter/s)": 1.316821
+    },
+    {
+      "acc": 0.68086214,
+      "epoch": 1.5183916793505836,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.5033079527119466e-06,
+      "loss": 1.41778965,
+      "memory(GiB)": 123.79,
+      "step": 59855,
+      "train_speed(iter/s)": 1.316832
+    },
+    {
+      "acc": 0.69277468,
+      "epoch": 1.5185185185185186,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.5025584831312112e-06,
+      "loss": 1.35513105,
+      "memory(GiB)": 123.79,
+      "step": 59860,
+      "train_speed(iter/s)": 1.316843
+    },
+    {
+      "acc": 0.67953906,
+      "epoch": 1.5186453576864536,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.5018091673835667e-06,
+      "loss": 1.41724319,
+      "memory(GiB)": 123.79,
+      "step": 59865,
+      "train_speed(iter/s)": 1.316855
+    },
+    {
+      "acc": 0.69056921,
+      "epoch": 1.5187721968543886,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.501060005501977e-06,
+      "loss": 1.36217127,
+      "memory(GiB)": 123.79,
+      "step": 59870,
+      "train_speed(iter/s)": 1.316866
+    },
+    {
+      "acc": 0.69195709,
+      "epoch": 1.5188990360223236,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.500310997519388e-06,
+      "loss": 1.42303696,
+      "memory(GiB)": 123.79,
+      "step": 59875,
+      "train_speed(iter/s)": 1.316878
+    },
+    {
+      "acc": 0.69456863,
+      "epoch": 1.5190258751902588,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.4995621434687468e-06,
+      "loss": 1.4010211,
+      "memory(GiB)": 123.79,
+      "step": 59880,
+      "train_speed(iter/s)": 1.316889
+    },
+    {
+      "acc": 0.67569337,
+      "epoch": 1.5191527143581938,
+      "grad_norm": 3.0,
+      "learning_rate": 1.4988134433829892e-06,
+      "loss": 1.45952911,
+      "memory(GiB)": 123.79,
+      "step": 59885,
+      "train_speed(iter/s)": 1.3169
+    },
+    {
+      "acc": 0.6751843,
+      "epoch": 1.519279553526129,
+      "grad_norm": 3.25,
+      "learning_rate": 1.4980648972950507e-06,
+      "loss": 1.40359135,
+      "memory(GiB)": 123.79,
+      "step": 59890,
+      "train_speed(iter/s)": 1.316911
+    },
+    {
+      "acc": 0.68744721,
+      "epoch": 1.519406392694064,
+      "grad_norm": 3.5,
+      "learning_rate": 1.4973165052378518e-06,
+      "loss": 1.36158247,
+      "memory(GiB)": 123.79,
+      "step": 59895,
+      "train_speed(iter/s)": 1.316923
+    },
+    {
+      "acc": 0.68301411,
+      "epoch": 1.519533231861999,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.49656826724431e-06,
+      "loss": 1.42266359,
+      "memory(GiB)": 123.79,
+      "step": 59900,
+      "train_speed(iter/s)": 1.316932
+    },
+    {
+      "acc": 0.68398399,
+      "epoch": 1.519660071029934,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.4958201833473386e-06,
+      "loss": 1.34757442,
+      "memory(GiB)": 123.79,
+      "step": 59905,
+      "train_speed(iter/s)": 1.316943
+    },
+    {
+      "acc": 0.66683292,
+      "epoch": 1.5197869101978692,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.4950722535798423e-06,
+      "loss": 1.44577503,
+      "memory(GiB)": 123.79,
+      "step": 59910,
+      "train_speed(iter/s)": 1.316953
+    },
+    {
+      "acc": 0.68027229,
+      "epoch": 1.5199137493658041,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.4943244779747134e-06,
+      "loss": 1.42649775,
+      "memory(GiB)": 123.79,
+      "step": 59915,
+      "train_speed(iter/s)": 1.316965
+    },
+    {
+      "acc": 0.67708311,
+      "epoch": 1.5200405885337394,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.4935768565648478e-06,
+      "loss": 1.38965435,
+      "memory(GiB)": 123.79,
+      "step": 59920,
+      "train_speed(iter/s)": 1.316976
+    },
+    {
+      "acc": 0.68321972,
+      "epoch": 1.5201674277016743,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.4928293893831265e-06,
+      "loss": 1.45541992,
+      "memory(GiB)": 123.79,
+      "step": 59925,
+      "train_speed(iter/s)": 1.316987
+    },
+    {
+      "acc": 0.68563914,
+      "epoch": 1.5202942668696093,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.4920820764624288e-06,
+      "loss": 1.38560543,
+      "memory(GiB)": 123.79,
+      "step": 59930,
+      "train_speed(iter/s)": 1.316999
+    },
+    {
+      "acc": 0.68073196,
+      "epoch": 1.5204211060375443,
+      "grad_norm": 3.25,
+      "learning_rate": 1.4913349178356202e-06,
+      "loss": 1.42352295,
+      "memory(GiB)": 123.79,
+      "step": 59935,
+      "train_speed(iter/s)": 1.31701
+    },
+    {
+      "acc": 0.6849391,
+      "epoch": 1.5205479452054793,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.4905879135355684e-06,
+      "loss": 1.44403362,
+      "memory(GiB)": 123.79,
+      "step": 59940,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.68176808,
+      "epoch": 1.5206747843734145,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.4898410635951282e-06,
+      "loss": 1.37012415,
+      "memory(GiB)": 123.79,
+      "step": 59945,
+      "train_speed(iter/s)": 1.317032
+    },
+    {
+      "acc": 0.69504042,
+      "epoch": 1.5208016235413497,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.4890943680471503e-06,
+      "loss": 1.34693918,
+      "memory(GiB)": 123.79,
+      "step": 59950,
+      "train_speed(iter/s)": 1.317043
+    },
+    {
+      "acc": 0.68372893,
+      "epoch": 1.5209284627092847,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.4883478269244766e-06,
+      "loss": 1.38862696,
+      "memory(GiB)": 123.79,
+      "step": 59955,
+      "train_speed(iter/s)": 1.317055
+    },
+    {
+      "acc": 0.68716621,
+      "epoch": 1.5210553018772197,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.4876014402599443e-06,
+      "loss": 1.44541578,
+      "memory(GiB)": 123.79,
+      "step": 59960,
+      "train_speed(iter/s)": 1.317066
+    },
+    {
+      "acc": 0.67696171,
+      "epoch": 1.5211821410451547,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.4868552080863824e-06,
+      "loss": 1.43513279,
+      "memory(GiB)": 123.79,
+      "step": 59965,
+      "train_speed(iter/s)": 1.317077
+    },
+    {
+      "acc": 0.70115662,
+      "epoch": 1.5213089802130897,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.4861091304366139e-06,
+      "loss": 1.33164997,
+      "memory(GiB)": 123.79,
+      "step": 59970,
+      "train_speed(iter/s)": 1.317087
+    },
+    {
+      "acc": 0.70415792,
+      "epoch": 1.521435819381025,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.4853632073434533e-06,
+      "loss": 1.27971764,
+      "memory(GiB)": 123.79,
+      "step": 59975,
+      "train_speed(iter/s)": 1.317098
+    },
+    {
+      "acc": 0.68986773,
+      "epoch": 1.52156265854896,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.484617438839711e-06,
+      "loss": 1.44485607,
+      "memory(GiB)": 123.79,
+      "step": 59980,
+      "train_speed(iter/s)": 1.317109
+    },
+    {
+      "acc": 0.68835821,
+      "epoch": 1.521689497716895,
+      "grad_norm": 4.625,
+      "learning_rate": 1.483871824958189e-06,
+      "loss": 1.35937529,
+      "memory(GiB)": 123.79,
+      "step": 59985,
+      "train_speed(iter/s)": 1.31712
+    },
+    {
+      "acc": 0.6847867,
+      "epoch": 1.52181633688483,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.483126365731682e-06,
+      "loss": 1.38055964,
+      "memory(GiB)": 123.79,
+      "step": 59990,
+      "train_speed(iter/s)": 1.317131
+    },
+    {
+      "acc": 0.69605312,
+      "epoch": 1.521943176052765,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.4823810611929795e-06,
+      "loss": 1.32878246,
+      "memory(GiB)": 123.79,
+      "step": 59995,
+      "train_speed(iter/s)": 1.317143
+    },
+    {
+      "acc": 0.70013566,
+      "epoch": 1.5220700152207,
+      "grad_norm": 2.875,
+      "learning_rate": 1.481635911374863e-06,
+      "loss": 1.30873928,
+      "memory(GiB)": 123.79,
+      "step": 60000,
+      "train_speed(iter/s)": 1.317153
+    },
+    {
+      "epoch": 1.5220700152207,
+      "eval_acc": 0.6750800343317525,
+      "eval_loss": 1.3563423156738281,
+      "eval_runtime": 70.5812,
+      "eval_samples_per_second": 90.251,
+      "eval_steps_per_second": 22.57,
+      "step": 60000
+    },
+    {
+      "acc": 0.68689504,
+      "epoch": 1.522196854388635,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.480890916310106e-06,
+      "loss": 1.41500225,
+      "memory(GiB)": 123.79,
+      "step": 60005,
+      "train_speed(iter/s)": 1.314767
+    },
+    {
+      "acc": 0.69558449,
+      "epoch": 1.5223236935565703,
+      "grad_norm": 3.5,
+      "learning_rate": 1.4801460760314811e-06,
+      "loss": 1.33207388,
+      "memory(GiB)": 123.79,
+      "step": 60010,
+      "train_speed(iter/s)": 1.314778
+    },
+    {
+      "acc": 0.67748837,
+      "epoch": 1.5224505327245055,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.4794013905717453e-06,
+      "loss": 1.44382362,
+      "memory(GiB)": 123.79,
+      "step": 60015,
+      "train_speed(iter/s)": 1.314791
+    },
+    {
+      "acc": 0.68361111,
+      "epoch": 1.5225773718924405,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.4786568599636548e-06,
+      "loss": 1.39567318,
+      "memory(GiB)": 123.79,
+      "step": 60020,
+      "train_speed(iter/s)": 1.314803
+    },
+    {
+      "acc": 0.68799911,
+      "epoch": 1.5227042110603755,
+      "grad_norm": 2.875,
+      "learning_rate": 1.4779124842399556e-06,
+      "loss": 1.43346558,
+      "memory(GiB)": 123.79,
+      "step": 60025,
+      "train_speed(iter/s)": 1.314815
+    },
+    {
+      "acc": 0.6878551,
+      "epoch": 1.5228310502283104,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.4771682634333933e-06,
+      "loss": 1.35807829,
+      "memory(GiB)": 123.79,
+      "step": 60030,
+      "train_speed(iter/s)": 1.314826
+    },
+    {
+      "acc": 0.69295492,
+      "epoch": 1.5229578893962454,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.4764241975766975e-06,
+      "loss": 1.33781252,
+      "memory(GiB)": 123.79,
+      "step": 60035,
+      "train_speed(iter/s)": 1.314838
+    },
+    {
+      "acc": 0.69479527,
+      "epoch": 1.5230847285641806,
+      "grad_norm": 3.375,
+      "learning_rate": 1.475680286702596e-06,
+      "loss": 1.34368811,
+      "memory(GiB)": 123.79,
+      "step": 60040,
+      "train_speed(iter/s)": 1.31485
+    },
+    {
+      "acc": 0.69085183,
+      "epoch": 1.5232115677321156,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.474936530843812e-06,
+      "loss": 1.33167801,
+      "memory(GiB)": 123.79,
+      "step": 60045,
+      "train_speed(iter/s)": 1.314862
+    },
+    {
+      "acc": 0.68983593,
+      "epoch": 1.5233384069000508,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.4741929300330588e-06,
+      "loss": 1.36394634,
+      "memory(GiB)": 123.79,
+      "step": 60050,
+      "train_speed(iter/s)": 1.314874
+    },
+    {
+      "acc": 0.68790121,
+      "epoch": 1.5234652460679858,
+      "grad_norm": 3.25,
+      "learning_rate": 1.4734494843030405e-06,
+      "loss": 1.38480186,
+      "memory(GiB)": 123.79,
+      "step": 60055,
+      "train_speed(iter/s)": 1.314886
+    },
+    {
+      "acc": 0.68490982,
+      "epoch": 1.5235920852359208,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.4727061936864573e-06,
+      "loss": 1.413904,
+      "memory(GiB)": 123.79,
+      "step": 60060,
+      "train_speed(iter/s)": 1.314898
+    },
+    {
+      "acc": 0.67209473,
+      "epoch": 1.5237189244038558,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.4719630582160056e-06,
+      "loss": 1.46754742,
+      "memory(GiB)": 123.79,
+      "step": 60065,
+      "train_speed(iter/s)": 1.31491
+    },
+    {
+      "acc": 0.68611379,
+      "epoch": 1.523845763571791,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.4712200779243718e-06,
+      "loss": 1.39669294,
+      "memory(GiB)": 123.79,
+      "step": 60070,
+      "train_speed(iter/s)": 1.314922
+    },
+    {
+      "acc": 0.68718643,
+      "epoch": 1.523972602739726,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.4704772528442308e-06,
+      "loss": 1.42622395,
+      "memory(GiB)": 123.79,
+      "step": 60075,
+      "train_speed(iter/s)": 1.314933
+    },
+    {
+      "acc": 0.69981804,
+      "epoch": 1.5240994419076612,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.46973458300826e-06,
+      "loss": 1.28443203,
+      "memory(GiB)": 123.79,
+      "step": 60080,
+      "train_speed(iter/s)": 1.314945
+    },
+    {
+      "acc": 0.69228921,
+      "epoch": 1.5242262810755962,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.4689920684491232e-06,
+      "loss": 1.38885479,
+      "memory(GiB)": 123.79,
+      "step": 60085,
+      "train_speed(iter/s)": 1.314957
+    },
+    {
+      "acc": 0.69386129,
+      "epoch": 1.5243531202435312,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.4682497091994807e-06,
+      "loss": 1.39833097,
+      "memory(GiB)": 123.79,
+      "step": 60090,
+      "train_speed(iter/s)": 1.314968
+    },
+    {
+      "acc": 0.68609843,
+      "epoch": 1.5244799594114662,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.467507505291984e-06,
+      "loss": 1.38575592,
+      "memory(GiB)": 123.79,
+      "step": 60095,
+      "train_speed(iter/s)": 1.31498
+    },
+    {
+      "acc": 0.66802759,
+      "epoch": 1.5246067985794012,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.4667654567592781e-06,
+      "loss": 1.4846796,
+      "memory(GiB)": 123.79,
+      "step": 60100,
+      "train_speed(iter/s)": 1.314992
+    },
+    {
+      "acc": 0.67649384,
+      "epoch": 1.5247336377473364,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.4660235636340025e-06,
+      "loss": 1.42287617,
+      "memory(GiB)": 123.79,
+      "step": 60105,
+      "train_speed(iter/s)": 1.315004
+    },
+    {
+      "acc": 0.68066044,
+      "epoch": 1.5248604769152716,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.465281825948789e-06,
+      "loss": 1.39613342,
+      "memory(GiB)": 123.79,
+      "step": 60110,
+      "train_speed(iter/s)": 1.315016
+    },
+    {
+      "acc": 0.67506514,
+      "epoch": 1.5249873160832066,
+      "grad_norm": 3.125,
+      "learning_rate": 1.464540243736262e-06,
+      "loss": 1.40884781,
+      "memory(GiB)": 123.79,
+      "step": 60115,
+      "train_speed(iter/s)": 1.315028
+    },
+    {
+      "acc": 0.6897264,
+      "epoch": 1.5251141552511416,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.4637988170290396e-06,
+      "loss": 1.34911728,
+      "memory(GiB)": 123.79,
+      "step": 60120,
+      "train_speed(iter/s)": 1.315041
+    },
+    {
+      "acc": 0.70129747,
+      "epoch": 1.5252409944190766,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.4630575458597334e-06,
+      "loss": 1.33636971,
+      "memory(GiB)": 123.79,
+      "step": 60125,
+      "train_speed(iter/s)": 1.315052
+    },
+    {
+      "acc": 0.69516287,
+      "epoch": 1.5253678335870116,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.4623164302609472e-06,
+      "loss": 1.32229166,
+      "memory(GiB)": 123.79,
+      "step": 60130,
+      "train_speed(iter/s)": 1.315064
+    },
+    {
+      "acc": 0.68172784,
+      "epoch": 1.5254946727549468,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.4615754702652796e-06,
+      "loss": 1.47824879,
+      "memory(GiB)": 123.79,
+      "step": 60135,
+      "train_speed(iter/s)": 1.315076
+    },
+    {
+      "acc": 0.70265999,
+      "epoch": 1.5256215119228818,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.4608346659053208e-06,
+      "loss": 1.33249168,
+      "memory(GiB)": 123.79,
+      "step": 60140,
+      "train_speed(iter/s)": 1.315089
+    },
+    {
+      "acc": 0.68259926,
+      "epoch": 1.525748351090817,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.4600940172136541e-06,
+      "loss": 1.38693829,
+      "memory(GiB)": 123.79,
+      "step": 60145,
+      "train_speed(iter/s)": 1.315101
+    },
+    {
+      "acc": 0.69304504,
+      "epoch": 1.525875190258752,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.4593535242228575e-06,
+      "loss": 1.35162764,
+      "memory(GiB)": 123.79,
+      "step": 60150,
+      "train_speed(iter/s)": 1.315113
+    },
+    {
+      "acc": 0.6727066,
+      "epoch": 1.526002029426687,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.4586131869655001e-06,
+      "loss": 1.38538322,
+      "memory(GiB)": 123.79,
+      "step": 60155,
+      "train_speed(iter/s)": 1.315125
+    },
+    {
+      "acc": 0.67614622,
+      "epoch": 1.526128868594622,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.4578730054741462e-06,
+      "loss": 1.42164326,
+      "memory(GiB)": 123.79,
+      "step": 60160,
+      "train_speed(iter/s)": 1.315135
+    },
+    {
+      "acc": 0.68320904,
+      "epoch": 1.526255707762557,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.4571329797813511e-06,
+      "loss": 1.3552804,
+      "memory(GiB)": 123.79,
+      "step": 60165,
+      "train_speed(iter/s)": 1.315145
+    },
+    {
+      "acc": 0.69965334,
+      "epoch": 1.5263825469304921,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.4563931099196678e-06,
+      "loss": 1.32168093,
+      "memory(GiB)": 123.79,
+      "step": 60170,
+      "train_speed(iter/s)": 1.315157
+    },
+    {
+      "acc": 0.69480386,
+      "epoch": 1.5265093860984273,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.455653395921635e-06,
+      "loss": 1.34941492,
+      "memory(GiB)": 123.79,
+      "step": 60175,
+      "train_speed(iter/s)": 1.315168
+    },
+    {
+      "acc": 0.70041857,
+      "epoch": 1.5266362252663623,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.4549138378197891e-06,
+      "loss": 1.40105553,
+      "memory(GiB)": 123.79,
+      "step": 60180,
+      "train_speed(iter/s)": 1.31518
+    },
+    {
+      "acc": 0.67841539,
+      "epoch": 1.5267630644342973,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.4541744356466615e-06,
+      "loss": 1.41744347,
+      "memory(GiB)": 123.79,
+      "step": 60185,
+      "train_speed(iter/s)": 1.315192
+    },
+    {
+      "acc": 0.68779879,
+      "epoch": 1.5268899036022323,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.4534351894347748e-06,
+      "loss": 1.32413645,
+      "memory(GiB)": 123.79,
+      "step": 60190,
+      "train_speed(iter/s)": 1.315204
+    },
+    {
+      "acc": 0.69605069,
+      "epoch": 1.5270167427701673,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.4526960992166412e-06,
+      "loss": 1.37662334,
+      "memory(GiB)": 123.79,
+      "step": 60195,
+      "train_speed(iter/s)": 1.315216
+    },
+    {
+      "acc": 0.68807421,
+      "epoch": 1.5271435819381025,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.4519571650247687e-06,
+      "loss": 1.35834141,
+      "memory(GiB)": 123.79,
+      "step": 60200,
+      "train_speed(iter/s)": 1.315228
+    },
+    {
+      "acc": 0.69538517,
+      "epoch": 1.5272704211060375,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.4512183868916629e-06,
+      "loss": 1.38267498,
+      "memory(GiB)": 123.79,
+      "step": 60205,
+      "train_speed(iter/s)": 1.31524
+    },
+    {
+      "acc": 0.69124599,
+      "epoch": 1.5273972602739727,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.4504797648498186e-06,
+      "loss": 1.39040527,
+      "memory(GiB)": 123.79,
+      "step": 60210,
+      "train_speed(iter/s)": 1.315252
+    },
+    {
+      "acc": 0.68126092,
+      "epoch": 1.5275240994419077,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.4497412989317184e-06,
+      "loss": 1.36953869,
+      "memory(GiB)": 123.79,
+      "step": 60215,
+      "train_speed(iter/s)": 1.315265
+    },
+    {
+      "acc": 0.68471594,
+      "epoch": 1.5276509386098427,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.4490029891698476e-06,
+      "loss": 1.46019154,
+      "memory(GiB)": 123.79,
+      "step": 60220,
+      "train_speed(iter/s)": 1.315275
+    },
+    {
+      "acc": 0.68818164,
+      "epoch": 1.5277777777777777,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.44826483559668e-06,
+      "loss": 1.32849827,
+      "memory(GiB)": 123.79,
+      "step": 60225,
+      "train_speed(iter/s)": 1.315287
+    },
+    {
+      "acc": 0.69709249,
+      "epoch": 1.527904616945713,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.4475268382446833e-06,
+      "loss": 1.32652798,
+      "memory(GiB)": 123.79,
+      "step": 60230,
+      "train_speed(iter/s)": 1.3153
+    },
+    {
+      "acc": 0.6909029,
+      "epoch": 1.5280314561136479,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.4467889971463144e-06,
+      "loss": 1.39048862,
+      "memory(GiB)": 123.79,
+      "step": 60235,
+      "train_speed(iter/s)": 1.315308
+    },
+    {
+      "acc": 0.69262629,
+      "epoch": 1.528158295281583,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.4460513123340308e-06,
+      "loss": 1.38022976,
+      "memory(GiB)": 123.79,
+      "step": 60240,
+      "train_speed(iter/s)": 1.31532
+    },
+    {
+      "acc": 0.69649143,
+      "epoch": 1.528285134449518,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.4453137838402775e-06,
+      "loss": 1.38695679,
+      "memory(GiB)": 123.79,
+      "step": 60245,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.69700961,
+      "epoch": 1.528411973617453,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.4445764116974948e-06,
+      "loss": 1.33357887,
+      "memory(GiB)": 123.79,
+      "step": 60250,
+      "train_speed(iter/s)": 1.315345
+    },
+    {
+      "acc": 0.69446411,
+      "epoch": 1.528538812785388,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.4438391959381149e-06,
+      "loss": 1.35135231,
+      "memory(GiB)": 123.79,
+      "step": 60255,
+      "train_speed(iter/s)": 1.315358
+    },
+    {
+      "acc": 0.69097629,
+      "epoch": 1.528665651953323,
+      "grad_norm": 3.375,
+      "learning_rate": 1.4431021365945647e-06,
+      "loss": 1.36735725,
+      "memory(GiB)": 123.79,
+      "step": 60260,
+      "train_speed(iter/s)": 1.31537
+    },
+    {
+      "acc": 0.67833033,
+      "epoch": 1.5287924911212583,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.4423652336992627e-06,
+      "loss": 1.43105431,
+      "memory(GiB)": 123.79,
+      "step": 60265,
+      "train_speed(iter/s)": 1.315383
+    },
+    {
+      "acc": 0.69525223,
+      "epoch": 1.5289193302891935,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.4416284872846215e-06,
+      "loss": 1.32834024,
+      "memory(GiB)": 123.79,
+      "step": 60270,
+      "train_speed(iter/s)": 1.315396
+    },
+    {
+      "acc": 0.70589991,
+      "epoch": 1.5290461694571285,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.440891897383046e-06,
+      "loss": 1.27086992,
+      "memory(GiB)": 123.79,
+      "step": 60275,
+      "train_speed(iter/s)": 1.315407
+    },
+    {
+      "acc": 0.70819683,
+      "epoch": 1.5291730086250634,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.4401554640269354e-06,
+      "loss": 1.25214653,
+      "memory(GiB)": 123.79,
+      "step": 60280,
+      "train_speed(iter/s)": 1.315419
+    },
+    {
+      "acc": 0.67424078,
+      "epoch": 1.5292998477929984,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.4394191872486812e-06,
+      "loss": 1.41462784,
+      "memory(GiB)": 123.79,
+      "step": 60285,
+      "train_speed(iter/s)": 1.31543
+    },
+    {
+      "acc": 0.66987314,
+      "epoch": 1.5294266869609334,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.4386830670806684e-06,
+      "loss": 1.44824934,
+      "memory(GiB)": 123.79,
+      "step": 60290,
+      "train_speed(iter/s)": 1.315443
+    },
+    {
+      "acc": 0.69208527,
+      "epoch": 1.5295535261288686,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.4379471035552738e-06,
+      "loss": 1.36250315,
+      "memory(GiB)": 123.79,
+      "step": 60295,
+      "train_speed(iter/s)": 1.315455
+    },
+    {
+      "acc": 0.68616428,
+      "epoch": 1.5296803652968036,
+      "grad_norm": 3.375,
+      "learning_rate": 1.437211296704869e-06,
+      "loss": 1.42319384,
+      "memory(GiB)": 123.79,
+      "step": 60300,
+      "train_speed(iter/s)": 1.315467
+    },
+    {
+      "acc": 0.68213873,
+      "epoch": 1.5298072044647388,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.4364756465618167e-06,
+      "loss": 1.4368948,
+      "memory(GiB)": 123.79,
+      "step": 60305,
+      "train_speed(iter/s)": 1.31548
+    },
+    {
+      "acc": 0.70520821,
+      "epoch": 1.5299340436326738,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.4357401531584792e-06,
+      "loss": 1.33923168,
+      "memory(GiB)": 123.79,
+      "step": 60310,
+      "train_speed(iter/s)": 1.315492
+    },
+    {
+      "acc": 0.69035373,
+      "epoch": 1.5300608828006088,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.4350048165272006e-06,
+      "loss": 1.31778154,
+      "memory(GiB)": 123.79,
+      "step": 60315,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.67854719,
+      "epoch": 1.5301877219685438,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.4342696367003272e-06,
+      "loss": 1.45062237,
+      "memory(GiB)": 123.79,
+      "step": 60320,
+      "train_speed(iter/s)": 1.315516
+    },
+    {
+      "acc": 0.68284163,
+      "epoch": 1.5303145611364788,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.433534613710193e-06,
+      "loss": 1.34423275,
+      "memory(GiB)": 123.79,
+      "step": 60325,
+      "train_speed(iter/s)": 1.315528
+    },
+    {
+      "acc": 0.69025669,
+      "epoch": 1.530441400304414,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.4327997475891331e-06,
+      "loss": 1.38679504,
+      "memory(GiB)": 123.79,
+      "step": 60330,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.6918591,
+      "epoch": 1.5305682394723492,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.432065038369465e-06,
+      "loss": 1.36154118,
+      "memory(GiB)": 123.79,
+      "step": 60335,
+      "train_speed(iter/s)": 1.315552
+    },
+    {
+      "acc": 0.69688616,
+      "epoch": 1.5306950786402842,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.4313304860835048e-06,
+      "loss": 1.32206287,
+      "memory(GiB)": 123.79,
+      "step": 60340,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.68268251,
+      "epoch": 1.5308219178082192,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.4305960907635641e-06,
+      "loss": 1.41350918,
+      "memory(GiB)": 123.79,
+      "step": 60345,
+      "train_speed(iter/s)": 1.315576
+    },
+    {
+      "acc": 0.70273809,
+      "epoch": 1.5309487569761542,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.4298618524419455e-06,
+      "loss": 1.38586578,
+      "memory(GiB)": 123.79,
+      "step": 60350,
+      "train_speed(iter/s)": 1.315588
+    },
+    {
+      "acc": 0.6825398,
+      "epoch": 1.5310755961440892,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.4291277711509388e-06,
+      "loss": 1.42826834,
+      "memory(GiB)": 123.79,
+      "step": 60355,
+      "train_speed(iter/s)": 1.315601
+    },
+    {
+      "acc": 0.69782662,
+      "epoch": 1.5312024353120244,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.428393846922837e-06,
+      "loss": 1.26499157,
+      "memory(GiB)": 123.79,
+      "step": 60360,
+      "train_speed(iter/s)": 1.315613
+    },
+    {
+      "acc": 0.70468192,
+      "epoch": 1.5313292744799594,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.4276600797899199e-06,
+      "loss": 1.30048704,
+      "memory(GiB)": 123.79,
+      "step": 60365,
+      "train_speed(iter/s)": 1.315626
+    },
+    {
+      "acc": 0.69026308,
+      "epoch": 1.5314561136478946,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.426926469784463e-06,
+      "loss": 1.34053612,
+      "memory(GiB)": 123.79,
+      "step": 60370,
+      "train_speed(iter/s)": 1.315639
+    },
+    {
+      "acc": 0.68795033,
+      "epoch": 1.5315829528158296,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.42619301693873e-06,
+      "loss": 1.44581928,
+      "memory(GiB)": 123.79,
+      "step": 60375,
+      "train_speed(iter/s)": 1.315651
+    },
+    {
+      "acc": 0.69610157,
+      "epoch": 1.5317097919837646,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.4254597212849858e-06,
+      "loss": 1.35201521,
+      "memory(GiB)": 123.79,
+      "step": 60380,
+      "train_speed(iter/s)": 1.315664
+    },
+    {
+      "acc": 0.67619267,
+      "epoch": 1.5318366311516995,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.4247265828554819e-06,
+      "loss": 1.40320482,
+      "memory(GiB)": 123.79,
+      "step": 60385,
+      "train_speed(iter/s)": 1.315676
+    },
+    {
+      "acc": 0.68543882,
+      "epoch": 1.5319634703196348,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.423993601682465e-06,
+      "loss": 1.42660885,
+      "memory(GiB)": 123.79,
+      "step": 60390,
+      "train_speed(iter/s)": 1.315689
+    },
+    {
+      "acc": 0.68464241,
+      "epoch": 1.5320903094875697,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.423260777798176e-06,
+      "loss": 1.37195015,
+      "memory(GiB)": 123.79,
+      "step": 60395,
+      "train_speed(iter/s)": 1.315701
+    },
+    {
+      "acc": 0.67912002,
+      "epoch": 1.532217148655505,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.4225281112348466e-06,
+      "loss": 1.42778988,
+      "memory(GiB)": 123.79,
+      "step": 60400,
+      "train_speed(iter/s)": 1.315714
+    },
+    {
+      "acc": 0.6648922,
+      "epoch": 1.53234398782344,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.421795602024703e-06,
+      "loss": 1.46493416,
+      "memory(GiB)": 123.79,
+      "step": 60405,
+      "train_speed(iter/s)": 1.315726
+    },
+    {
+      "acc": 0.68750477,
+      "epoch": 1.532470826991375,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.4210632501999643e-06,
+      "loss": 1.40066433,
+      "memory(GiB)": 123.79,
+      "step": 60410,
+      "train_speed(iter/s)": 1.315738
+    },
+    {
+      "acc": 0.68067131,
+      "epoch": 1.53259766615931,
+      "grad_norm": 2.875,
+      "learning_rate": 1.4203310557928428e-06,
+      "loss": 1.37995234,
+      "memory(GiB)": 123.79,
+      "step": 60415,
+      "train_speed(iter/s)": 1.31575
+    },
+    {
+      "acc": 0.66779642,
+      "epoch": 1.532724505327245,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.4195990188355435e-06,
+      "loss": 1.51357174,
+      "memory(GiB)": 123.79,
+      "step": 60420,
+      "train_speed(iter/s)": 1.315762
+    },
+    {
+      "acc": 0.69032803,
+      "epoch": 1.5328513444951801,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.418867139360265e-06,
+      "loss": 1.36384449,
+      "memory(GiB)": 123.79,
+      "step": 60425,
+      "train_speed(iter/s)": 1.315774
+    },
+    {
+      "acc": 0.69752083,
+      "epoch": 1.5329781836631153,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.418135417399198e-06,
+      "loss": 1.37184172,
+      "memory(GiB)": 123.79,
+      "step": 60430,
+      "train_speed(iter/s)": 1.315787
+    },
+    {
+      "acc": 0.67823071,
+      "epoch": 1.5331050228310503,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.4174038529845273e-06,
+      "loss": 1.46667385,
+      "memory(GiB)": 123.79,
+      "step": 60435,
+      "train_speed(iter/s)": 1.315799
+    },
+    {
+      "acc": 0.67631769,
+      "epoch": 1.5332318619989853,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.4166724461484304e-06,
+      "loss": 1.43673382,
+      "memory(GiB)": 123.79,
+      "step": 60440,
+      "train_speed(iter/s)": 1.315811
+    },
+    {
+      "acc": 0.67750731,
+      "epoch": 1.5333587011669203,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.4159411969230758e-06,
+      "loss": 1.4316618,
+      "memory(GiB)": 123.79,
+      "step": 60445,
+      "train_speed(iter/s)": 1.315824
+    },
+    {
+      "acc": 0.68644824,
+      "epoch": 1.5334855403348553,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.4152101053406325e-06,
+      "loss": 1.35435734,
+      "memory(GiB)": 123.79,
+      "step": 60450,
+      "train_speed(iter/s)": 1.315835
+    },
+    {
+      "acc": 0.69382401,
+      "epoch": 1.5336123795027905,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.4144791714332517e-06,
+      "loss": 1.33952599,
+      "memory(GiB)": 123.79,
+      "step": 60455,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.68494592,
+      "epoch": 1.5337392186707255,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.4137483952330855e-06,
+      "loss": 1.37931156,
+      "memory(GiB)": 123.79,
+      "step": 60460,
+      "train_speed(iter/s)": 1.315858
+    },
+    {
+      "acc": 0.69348421,
+      "epoch": 1.5338660578386607,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.4130177767722753e-06,
+      "loss": 1.36857729,
+      "memory(GiB)": 123.79,
+      "step": 60465,
+      "train_speed(iter/s)": 1.31587
+    },
+    {
+      "acc": 0.69252291,
+      "epoch": 1.5339928970065957,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.4122873160829603e-06,
+      "loss": 1.38959255,
+      "memory(GiB)": 123.79,
+      "step": 60470,
+      "train_speed(iter/s)": 1.315881
+    },
+    {
+      "acc": 0.67482538,
+      "epoch": 1.5341197361745307,
+      "grad_norm": 2.625,
+      "learning_rate": 1.4115570131972655e-06,
+      "loss": 1.39670811,
+      "memory(GiB)": 123.79,
+      "step": 60475,
+      "train_speed(iter/s)": 1.315894
+    },
+    {
+      "acc": 0.69766941,
+      "epoch": 1.5342465753424657,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.4108268681473136e-06,
+      "loss": 1.39925823,
+      "memory(GiB)": 123.79,
+      "step": 60480,
+      "train_speed(iter/s)": 1.315906
+    },
+    {
+      "acc": 0.69831266,
+      "epoch": 1.5343734145104007,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.4100968809652215e-06,
+      "loss": 1.41892776,
+      "memory(GiB)": 123.79,
+      "step": 60485,
+      "train_speed(iter/s)": 1.315917
+    },
+    {
+      "acc": 0.68570571,
+      "epoch": 1.5345002536783359,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.4093670516830982e-06,
+      "loss": 1.39850912,
+      "memory(GiB)": 123.79,
+      "step": 60490,
+      "train_speed(iter/s)": 1.315929
+    },
+    {
+      "acc": 0.68304954,
+      "epoch": 1.534627092846271,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.4086373803330417e-06,
+      "loss": 1.39331875,
+      "memory(GiB)": 123.79,
+      "step": 60495,
+      "train_speed(iter/s)": 1.315941
+    },
+    {
+      "acc": 0.68172569,
+      "epoch": 1.534753932014206,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.4079078669471457e-06,
+      "loss": 1.40964289,
+      "memory(GiB)": 123.79,
+      "step": 60500,
+      "train_speed(iter/s)": 1.315953
+    },
+    {
+      "acc": 0.69530039,
+      "epoch": 1.534880771182141,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.4071785115575005e-06,
+      "loss": 1.33987713,
+      "memory(GiB)": 123.79,
+      "step": 60505,
+      "train_speed(iter/s)": 1.315964
+    },
+    {
+      "acc": 0.7024086,
+      "epoch": 1.535007610350076,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.4064493141961872e-06,
+      "loss": 1.33931713,
+      "memory(GiB)": 123.79,
+      "step": 60510,
+      "train_speed(iter/s)": 1.315975
+    },
+    {
+      "acc": 0.70403929,
+      "epoch": 1.535134449518011,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.4057202748952736e-06,
+      "loss": 1.36959238,
+      "memory(GiB)": 123.79,
+      "step": 60515,
+      "train_speed(iter/s)": 1.315988
+    },
+    {
+      "acc": 0.69216385,
+      "epoch": 1.5352612886859462,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.4049913936868314e-06,
+      "loss": 1.36021595,
+      "memory(GiB)": 123.79,
+      "step": 60520,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.67885294,
+      "epoch": 1.5353881278538812,
+      "grad_norm": 3.125,
+      "learning_rate": 1.4042626706029184e-06,
+      "loss": 1.41545601,
+      "memory(GiB)": 123.79,
+      "step": 60525,
+      "train_speed(iter/s)": 1.316012
+    },
+    {
+      "acc": 0.6752387,
+      "epoch": 1.5355149670218164,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.4035341056755864e-06,
+      "loss": 1.44262199,
+      "memory(GiB)": 123.79,
+      "step": 60530,
+      "train_speed(iter/s)": 1.316021
+    },
+    {
+      "acc": 0.70366883,
+      "epoch": 1.5356418061897514,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.402805698936882e-06,
+      "loss": 1.35478401,
+      "memory(GiB)": 123.79,
+      "step": 60535,
+      "train_speed(iter/s)": 1.316033
+    },
+    {
+      "acc": 0.68660016,
+      "epoch": 1.5357686453576864,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.4020774504188428e-06,
+      "loss": 1.41784477,
+      "memory(GiB)": 123.79,
+      "step": 60540,
+      "train_speed(iter/s)": 1.316045
+    },
+    {
+      "acc": 0.68723269,
+      "epoch": 1.5358954845256214,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.4013493601535016e-06,
+      "loss": 1.40217848,
+      "memory(GiB)": 123.79,
+      "step": 60545,
+      "train_speed(iter/s)": 1.316057
+    },
+    {
+      "acc": 0.67253528,
+      "epoch": 1.5360223236935566,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.400621428172882e-06,
+      "loss": 1.47498674,
+      "memory(GiB)": 123.79,
+      "step": 60550,
+      "train_speed(iter/s)": 1.316069
+    },
+    {
+      "acc": 0.68828249,
+      "epoch": 1.5361491628614916,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.399893654509002e-06,
+      "loss": 1.43716612,
+      "memory(GiB)": 123.79,
+      "step": 60555,
+      "train_speed(iter/s)": 1.316082
+    },
+    {
+      "acc": 0.68559828,
+      "epoch": 1.5362760020294268,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.3991660391938721e-06,
+      "loss": 1.39757919,
+      "memory(GiB)": 123.79,
+      "step": 60560,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.68739171,
+      "epoch": 1.5364028411973618,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.398438582259497e-06,
+      "loss": 1.38774977,
+      "memory(GiB)": 123.79,
+      "step": 60565,
+      "train_speed(iter/s)": 1.316106
+    },
+    {
+      "acc": 0.68087592,
+      "epoch": 1.5365296803652968,
+      "grad_norm": 3.375,
+      "learning_rate": 1.3977112837378726e-06,
+      "loss": 1.39539757,
+      "memory(GiB)": 123.79,
+      "step": 60570,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.68353167,
+      "epoch": 1.5366565195332318,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.3969841436609888e-06,
+      "loss": 1.41406631,
+      "memory(GiB)": 123.79,
+      "step": 60575,
+      "train_speed(iter/s)": 1.31613
+    },
+    {
+      "acc": 0.69270225,
+      "epoch": 1.5367833587011668,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.396257162060829e-06,
+      "loss": 1.38443871,
+      "memory(GiB)": 123.79,
+      "step": 60580,
+      "train_speed(iter/s)": 1.316142
+    },
+    {
+      "acc": 0.67186127,
+      "epoch": 1.536910197869102,
+      "grad_norm": 2.5,
+      "learning_rate": 1.395530338969367e-06,
+      "loss": 1.45400457,
+      "memory(GiB)": 123.79,
+      "step": 60585,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.68115158,
+      "epoch": 1.5370370370370372,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3948036744185767e-06,
+      "loss": 1.3749835,
+      "memory(GiB)": 123.79,
+      "step": 60590,
+      "train_speed(iter/s)": 1.316167
+    },
+    {
+      "acc": 0.68103118,
+      "epoch": 1.5371638762049722,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.3940771684404153e-06,
+      "loss": 1.3473526,
+      "memory(GiB)": 123.79,
+      "step": 60595,
+      "train_speed(iter/s)": 1.316178
+    },
+    {
+      "acc": 0.70046349,
+      "epoch": 1.5372907153729072,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.393350821066839e-06,
+      "loss": 1.31873245,
+      "memory(GiB)": 123.79,
+      "step": 60600,
+      "train_speed(iter/s)": 1.316189
+    },
+    {
+      "acc": 0.70240016,
+      "epoch": 1.5374175545408422,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.3926246323297948e-06,
+      "loss": 1.36561565,
+      "memory(GiB)": 123.79,
+      "step": 60605,
+      "train_speed(iter/s)": 1.316201
+    },
+    {
+      "acc": 0.6906477,
+      "epoch": 1.5375443937087772,
+      "grad_norm": 3.125,
+      "learning_rate": 1.3918986022612285e-06,
+      "loss": 1.33306589,
+      "memory(GiB)": 123.79,
+      "step": 60610,
+      "train_speed(iter/s)": 1.316212
+    },
+    {
+      "acc": 0.69416723,
+      "epoch": 1.5376712328767124,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.3911727308930684e-06,
+      "loss": 1.39760151,
+      "memory(GiB)": 123.79,
+      "step": 60615,
+      "train_speed(iter/s)": 1.316223
+    },
+    {
+      "acc": 0.68113294,
+      "epoch": 1.5377980720446474,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.3904470182572428e-06,
+      "loss": 1.39719543,
+      "memory(GiB)": 123.79,
+      "step": 60620,
+      "train_speed(iter/s)": 1.316233
+    },
+    {
+      "acc": 0.68368883,
+      "epoch": 1.5379249112125826,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3897214643856744e-06,
+      "loss": 1.48485394,
+      "memory(GiB)": 123.79,
+      "step": 60625,
+      "train_speed(iter/s)": 1.316245
+    },
+    {
+      "acc": 0.70603485,
+      "epoch": 1.5380517503805176,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.388996069310276e-06,
+      "loss": 1.32493849,
+      "memory(GiB)": 123.79,
+      "step": 60630,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.69122877,
+      "epoch": 1.5381785895484525,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.3882708330629514e-06,
+      "loss": 1.44173374,
+      "memory(GiB)": 123.79,
+      "step": 60635,
+      "train_speed(iter/s)": 1.316269
+    },
+    {
+      "acc": 0.67867255,
+      "epoch": 1.5383054287163875,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.3875457556755989e-06,
+      "loss": 1.41877861,
+      "memory(GiB)": 123.79,
+      "step": 60640,
+      "train_speed(iter/s)": 1.316282
+    },
+    {
+      "acc": 0.69232669,
+      "epoch": 1.5384322678843225,
+      "grad_norm": 3.5,
+      "learning_rate": 1.386820837180114e-06,
+      "loss": 1.36615181,
+      "memory(GiB)": 123.79,
+      "step": 60645,
+      "train_speed(iter/s)": 1.316293
+    },
+    {
+      "acc": 0.6967247,
+      "epoch": 1.5385591070522577,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.386096077608382e-06,
+      "loss": 1.35442009,
+      "memory(GiB)": 123.79,
+      "step": 60650,
+      "train_speed(iter/s)": 1.316305
+    },
+    {
+      "acc": 0.68881617,
+      "epoch": 1.538685946220193,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.385371476992276e-06,
+      "loss": 1.37913704,
+      "memory(GiB)": 123.79,
+      "step": 60655,
+      "train_speed(iter/s)": 1.316317
+    },
+    {
+      "acc": 0.67576571,
+      "epoch": 1.538812785388128,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.3846470353636726e-06,
+      "loss": 1.42362738,
+      "memory(GiB)": 123.79,
+      "step": 60660,
+      "train_speed(iter/s)": 1.316329
+    },
+    {
+      "acc": 0.67922015,
+      "epoch": 1.538939624556063,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.3839227527544336e-06,
+      "loss": 1.3941762,
+      "memory(GiB)": 123.79,
+      "step": 60665,
+      "train_speed(iter/s)": 1.316341
+    },
+    {
+      "acc": 0.69460278,
+      "epoch": 1.539066463723998,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.3831986291964184e-06,
+      "loss": 1.3403162,
+      "memory(GiB)": 123.79,
+      "step": 60670,
+      "train_speed(iter/s)": 1.316353
+    },
+    {
+      "acc": 0.68670821,
+      "epoch": 1.539193302891933,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.382474664721472e-06,
+      "loss": 1.40875387,
+      "memory(GiB)": 123.79,
+      "step": 60675,
+      "train_speed(iter/s)": 1.316365
+    },
+    {
+      "acc": 0.69635472,
+      "epoch": 1.5393201420598681,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.3817508593614425e-06,
+      "loss": 1.29803696,
+      "memory(GiB)": 123.79,
+      "step": 60680,
+      "train_speed(iter/s)": 1.316377
+    },
+    {
+      "acc": 0.68857336,
+      "epoch": 1.539446981227803,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.381027213148165e-06,
+      "loss": 1.41365318,
+      "memory(GiB)": 123.79,
+      "step": 60685,
+      "train_speed(iter/s)": 1.316389
+    },
+    {
+      "acc": 0.67870107,
+      "epoch": 1.5395738203957383,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.3803037261134678e-06,
+      "loss": 1.39755697,
+      "memory(GiB)": 123.79,
+      "step": 60690,
+      "train_speed(iter/s)": 1.316401
+    },
+    {
+      "acc": 0.69232283,
+      "epoch": 1.5397006595636733,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.3795803982891736e-06,
+      "loss": 1.40467472,
+      "memory(GiB)": 123.79,
+      "step": 60695,
+      "train_speed(iter/s)": 1.316413
+    },
+    {
+      "acc": 0.68650784,
+      "epoch": 1.5398274987316083,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.3788572297070974e-06,
+      "loss": 1.35410423,
+      "memory(GiB)": 123.79,
+      "step": 60700,
+      "train_speed(iter/s)": 1.316424
+    },
+    {
+      "acc": 0.69176073,
+      "epoch": 1.5399543378995433,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.3781342203990478e-06,
+      "loss": 1.38200502,
+      "memory(GiB)": 123.79,
+      "step": 60705,
+      "train_speed(iter/s)": 1.316436
+    },
+    {
+      "acc": 0.68819685,
+      "epoch": 1.5400811770674785,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3774113703968255e-06,
+      "loss": 1.41750259,
+      "memory(GiB)": 123.79,
+      "step": 60710,
+      "train_speed(iter/s)": 1.316448
+    },
+    {
+      "acc": 0.68622561,
+      "epoch": 1.5402080162354135,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3766886797322248e-06,
+      "loss": 1.36343098,
+      "memory(GiB)": 123.79,
+      "step": 60715,
+      "train_speed(iter/s)": 1.316461
+    },
+    {
+      "acc": 0.70364857,
+      "epoch": 1.5403348554033487,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.3759661484370324e-06,
+      "loss": 1.31567202,
+      "memory(GiB)": 123.79,
+      "step": 60720,
+      "train_speed(iter/s)": 1.316472
+    },
+    {
+      "acc": 0.70058794,
+      "epoch": 1.5404616945712837,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.3752437765430294e-06,
+      "loss": 1.34688988,
+      "memory(GiB)": 123.79,
+      "step": 60725,
+      "train_speed(iter/s)": 1.316485
+    },
+    {
+      "acc": 0.6875783,
+      "epoch": 1.5405885337392187,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.3745215640819886e-06,
+      "loss": 1.38466902,
+      "memory(GiB)": 123.79,
+      "step": 60730,
+      "train_speed(iter/s)": 1.316497
+    },
+    {
+      "acc": 0.69906049,
+      "epoch": 1.5407153729071537,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.373799511085676e-06,
+      "loss": 1.31634445,
+      "memory(GiB)": 123.79,
+      "step": 60735,
+      "train_speed(iter/s)": 1.316509
+    },
+    {
+      "acc": 0.68260632,
+      "epoch": 1.5408422120750886,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.3730776175858506e-06,
+      "loss": 1.41242619,
+      "memory(GiB)": 123.79,
+      "step": 60740,
+      "train_speed(iter/s)": 1.316521
+    },
+    {
+      "acc": 0.69865789,
+      "epoch": 1.5409690512430239,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3723558836142631e-06,
+      "loss": 1.35108261,
+      "memory(GiB)": 123.79,
+      "step": 60745,
+      "train_speed(iter/s)": 1.316532
+    },
+    {
+      "acc": 0.69428077,
+      "epoch": 1.541095890410959,
+      "grad_norm": 2.625,
+      "learning_rate": 1.371634309202663e-06,
+      "loss": 1.35442038,
+      "memory(GiB)": 123.79,
+      "step": 60750,
+      "train_speed(iter/s)": 1.316545
+    },
+    {
+      "acc": 0.68189831,
+      "epoch": 1.541222729578894,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.3709128943827842e-06,
+      "loss": 1.39483414,
+      "memory(GiB)": 123.79,
+      "step": 60755,
+      "train_speed(iter/s)": 1.316557
+    },
+    {
+      "acc": 0.66727476,
+      "epoch": 1.541349568746829,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3701916391863573e-06,
+      "loss": 1.43362617,
+      "memory(GiB)": 123.79,
+      "step": 60760,
+      "train_speed(iter/s)": 1.316569
+    },
+    {
+      "acc": 0.66921172,
+      "epoch": 1.541476407914764,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.3694705436451093e-06,
+      "loss": 1.41640701,
+      "memory(GiB)": 123.79,
+      "step": 60765,
+      "train_speed(iter/s)": 1.31658
+    },
+    {
+      "acc": 0.69797583,
+      "epoch": 1.541603247082699,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.368749607790758e-06,
+      "loss": 1.31542873,
+      "memory(GiB)": 123.79,
+      "step": 60770,
+      "train_speed(iter/s)": 1.316592
+    },
+    {
+      "acc": 0.69576721,
+      "epoch": 1.5417300862506342,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3680288316550095e-06,
+      "loss": 1.33888826,
+      "memory(GiB)": 123.79,
+      "step": 60775,
+      "train_speed(iter/s)": 1.316604
+    },
+    {
+      "acc": 0.68500681,
+      "epoch": 1.5418569254185692,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.3673082152695672e-06,
+      "loss": 1.45639725,
+      "memory(GiB)": 123.79,
+      "step": 60780,
+      "train_speed(iter/s)": 1.316616
+    },
+    {
+      "acc": 0.66131339,
+      "epoch": 1.5419837645865044,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.3665877586661296e-06,
+      "loss": 1.46303825,
+      "memory(GiB)": 123.79,
+      "step": 60785,
+      "train_speed(iter/s)": 1.316626
+    },
+    {
+      "acc": 0.69247894,
+      "epoch": 1.5421106037544394,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.3658674618763862e-06,
+      "loss": 1.35329857,
+      "memory(GiB)": 123.79,
+      "step": 60790,
+      "train_speed(iter/s)": 1.316638
+    },
+    {
+      "acc": 0.67161827,
+      "epoch": 1.5422374429223744,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.365147324932014e-06,
+      "loss": 1.38054981,
+      "memory(GiB)": 123.79,
+      "step": 60795,
+      "train_speed(iter/s)": 1.31665
+    },
+    {
+      "acc": 0.67686625,
+      "epoch": 1.5423642820903094,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.3644273478646925e-06,
+      "loss": 1.46049805,
+      "memory(GiB)": 123.79,
+      "step": 60800,
+      "train_speed(iter/s)": 1.316661
+    },
+    {
+      "acc": 0.68343782,
+      "epoch": 1.5424911212582444,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3637075307060877e-06,
+      "loss": 1.44286966,
+      "memory(GiB)": 123.79,
+      "step": 60805,
+      "train_speed(iter/s)": 1.316673
+    },
+    {
+      "acc": 0.68853502,
+      "epoch": 1.5426179604261796,
+      "grad_norm": 2.75,
+      "learning_rate": 1.362987873487862e-06,
+      "loss": 1.37967396,
+      "memory(GiB)": 123.79,
+      "step": 60810,
+      "train_speed(iter/s)": 1.316685
+    },
+    {
+      "acc": 0.6835609,
+      "epoch": 1.5427447995941148,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.362268376241665e-06,
+      "loss": 1.37237644,
+      "memory(GiB)": 123.79,
+      "step": 60815,
+      "train_speed(iter/s)": 1.316696
+    },
+    {
+      "acc": 0.65865116,
+      "epoch": 1.5428716387620498,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.3615490389991476e-06,
+      "loss": 1.50986357,
+      "memory(GiB)": 123.79,
+      "step": 60820,
+      "train_speed(iter/s)": 1.316707
+    },
+    {
+      "acc": 0.69082031,
+      "epoch": 1.5429984779299848,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.360829861791948e-06,
+      "loss": 1.35431461,
+      "memory(GiB)": 123.79,
+      "step": 60825,
+      "train_speed(iter/s)": 1.316719
+    },
+    {
+      "acc": 0.69332314,
+      "epoch": 1.5431253170979198,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.3601108446516985e-06,
+      "loss": 1.3969635,
+      "memory(GiB)": 123.79,
+      "step": 60830,
+      "train_speed(iter/s)": 1.31673
+    },
+    {
+      "acc": 0.69588442,
+      "epoch": 1.5432521562658548,
+      "grad_norm": 3.125,
+      "learning_rate": 1.3593919876100254e-06,
+      "loss": 1.33480663,
+      "memory(GiB)": 123.79,
+      "step": 60835,
+      "train_speed(iter/s)": 1.316742
+    },
+    {
+      "acc": 0.68700647,
+      "epoch": 1.54337899543379,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.3586732906985467e-06,
+      "loss": 1.4176115,
+      "memory(GiB)": 123.79,
+      "step": 60840,
+      "train_speed(iter/s)": 1.316754
+    },
+    {
+      "acc": 0.69125953,
+      "epoch": 1.543505834601725,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.357954753948874e-06,
+      "loss": 1.36817894,
+      "memory(GiB)": 123.79,
+      "step": 60845,
+      "train_speed(iter/s)": 1.316765
+    },
+    {
+      "acc": 0.70800724,
+      "epoch": 1.5436326737696602,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3572363773926117e-06,
+      "loss": 1.3203928,
+      "memory(GiB)": 123.79,
+      "step": 60850,
+      "train_speed(iter/s)": 1.316776
+    },
+    {
+      "acc": 0.69432755,
+      "epoch": 1.5437595129375952,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.3565181610613571e-06,
+      "loss": 1.38334904,
+      "memory(GiB)": 123.79,
+      "step": 60855,
+      "train_speed(iter/s)": 1.316786
+    },
+    {
+      "acc": 0.68640499,
+      "epoch": 1.5438863521055302,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.3558001049867008e-06,
+      "loss": 1.37675819,
+      "memory(GiB)": 123.79,
+      "step": 60860,
+      "train_speed(iter/s)": 1.316798
+    },
+    {
+      "acc": 0.69554605,
+      "epoch": 1.5440131912734651,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.3550822092002264e-06,
+      "loss": 1.38478136,
+      "memory(GiB)": 123.79,
+      "step": 60865,
+      "train_speed(iter/s)": 1.316809
+    },
+    {
+      "acc": 0.69823532,
+      "epoch": 1.5441400304414004,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.3543644737335099e-06,
+      "loss": 1.29302645,
+      "memory(GiB)": 123.79,
+      "step": 60870,
+      "train_speed(iter/s)": 1.31682
+    },
+    {
+      "acc": 0.6721066,
+      "epoch": 1.5442668696093353,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.35364689861812e-06,
+      "loss": 1.42852058,
+      "memory(GiB)": 123.79,
+      "step": 60875,
+      "train_speed(iter/s)": 1.316832
+    },
+    {
+      "acc": 0.69746323,
+      "epoch": 1.5443937087772706,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.3529294838856194e-06,
+      "loss": 1.35454388,
+      "memory(GiB)": 123.79,
+      "step": 60880,
+      "train_speed(iter/s)": 1.316844
+    },
+    {
+      "acc": 0.6831193,
+      "epoch": 1.5445205479452055,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.3522122295675616e-06,
+      "loss": 1.37759323,
+      "memory(GiB)": 123.79,
+      "step": 60885,
+      "train_speed(iter/s)": 1.316856
+    },
+    {
+      "acc": 0.6883769,
+      "epoch": 1.5446473871131405,
+      "grad_norm": 3.75,
+      "learning_rate": 1.351495135695499e-06,
+      "loss": 1.39741764,
+      "memory(GiB)": 123.79,
+      "step": 60890,
+      "train_speed(iter/s)": 1.316869
+    },
+    {
+      "acc": 0.68878422,
+      "epoch": 1.5447742262810755,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.3507782023009692e-06,
+      "loss": 1.384093,
+      "memory(GiB)": 123.79,
+      "step": 60895,
+      "train_speed(iter/s)": 1.31688
+    },
+    {
+      "acc": 0.68789711,
+      "epoch": 1.5449010654490105,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.3500614294155056e-06,
+      "loss": 1.36393738,
+      "memory(GiB)": 123.79,
+      "step": 60900,
+      "train_speed(iter/s)": 1.316891
+    },
+    {
+      "acc": 0.69794188,
+      "epoch": 1.5450279046169457,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.3493448170706347e-06,
+      "loss": 1.39649973,
+      "memory(GiB)": 123.79,
+      "step": 60905,
+      "train_speed(iter/s)": 1.316903
+    },
+    {
+      "acc": 0.68364215,
+      "epoch": 1.545154743784881,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.348628365297881e-06,
+      "loss": 1.38727341,
+      "memory(GiB)": 123.79,
+      "step": 60910,
+      "train_speed(iter/s)": 1.316915
+    },
+    {
+      "acc": 0.68756309,
+      "epoch": 1.545281582952816,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3479120741287526e-06,
+      "loss": 1.37121792,
+      "memory(GiB)": 123.79,
+      "step": 60915,
+      "train_speed(iter/s)": 1.316927
+    },
+    {
+      "acc": 0.68782845,
+      "epoch": 1.545408422120751,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.3471959435947552e-06,
+      "loss": 1.44777908,
+      "memory(GiB)": 123.79,
+      "step": 60920,
+      "train_speed(iter/s)": 1.316939
+    },
+    {
+      "acc": 0.68214655,
+      "epoch": 1.545535261288686,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.3464799737273898e-06,
+      "loss": 1.41853104,
+      "memory(GiB)": 123.79,
+      "step": 60925,
+      "train_speed(iter/s)": 1.31695
+    },
+    {
+      "acc": 0.68412695,
+      "epoch": 1.545662100456621,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.3457641645581487e-06,
+      "loss": 1.39436407,
+      "memory(GiB)": 123.79,
+      "step": 60930,
+      "train_speed(iter/s)": 1.316962
+    },
+    {
+      "acc": 0.66972809,
+      "epoch": 1.545788939624556,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.3450485161185133e-06,
+      "loss": 1.43191605,
+      "memory(GiB)": 123.79,
+      "step": 60935,
+      "train_speed(iter/s)": 1.316974
+    },
+    {
+      "acc": 0.68632779,
+      "epoch": 1.545915778792491,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.344333028439961e-06,
+      "loss": 1.36359329,
+      "memory(GiB)": 123.79,
+      "step": 60940,
+      "train_speed(iter/s)": 1.316986
+    },
+    {
+      "acc": 0.70063248,
+      "epoch": 1.5460426179604263,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3436177015539647e-06,
+      "loss": 1.35181618,
+      "memory(GiB)": 123.79,
+      "step": 60945,
+      "train_speed(iter/s)": 1.316997
+    },
+    {
+      "acc": 0.70561323,
+      "epoch": 1.5461694571283613,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.3429025354919877e-06,
+      "loss": 1.33555775,
+      "memory(GiB)": 123.79,
+      "step": 60950,
+      "train_speed(iter/s)": 1.317009
+    },
+    {
+      "acc": 0.67629137,
+      "epoch": 1.5462962962962963,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3421875302854826e-06,
+      "loss": 1.41849775,
+      "memory(GiB)": 123.79,
+      "step": 60955,
+      "train_speed(iter/s)": 1.317021
+    },
+    {
+      "acc": 0.67701187,
+      "epoch": 1.5464231354642313,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.3414726859659016e-06,
+      "loss": 1.39013119,
+      "memory(GiB)": 123.79,
+      "step": 60960,
+      "train_speed(iter/s)": 1.317033
+    },
+    {
+      "acc": 0.66904764,
+      "epoch": 1.5465499746321663,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.3407580025646866e-06,
+      "loss": 1.44728823,
+      "memory(GiB)": 123.79,
+      "step": 60965,
+      "train_speed(iter/s)": 1.317045
+    },
+    {
+      "acc": 0.68257551,
+      "epoch": 1.5466768138001015,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.3400434801132716e-06,
+      "loss": 1.42390537,
+      "memory(GiB)": 123.79,
+      "step": 60970,
+      "train_speed(iter/s)": 1.317057
+    },
+    {
+      "acc": 0.68243279,
+      "epoch": 1.5468036529680367,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.3393291186430852e-06,
+      "loss": 1.40342293,
+      "memory(GiB)": 123.79,
+      "step": 60975,
+      "train_speed(iter/s)": 1.317068
+    },
+    {
+      "acc": 0.67510004,
+      "epoch": 1.5469304921359717,
+      "grad_norm": 6.25,
+      "learning_rate": 1.338614918185548e-06,
+      "loss": 1.43470373,
+      "memory(GiB)": 123.79,
+      "step": 60980,
+      "train_speed(iter/s)": 1.31708
+    },
+    {
+      "acc": 0.69404154,
+      "epoch": 1.5470573313039067,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.3379008787720732e-06,
+      "loss": 1.385711,
+      "memory(GiB)": 123.79,
+      "step": 60985,
+      "train_speed(iter/s)": 1.317091
+    },
+    {
+      "acc": 0.69928551,
+      "epoch": 1.5471841704718416,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.3371870004340681e-06,
+      "loss": 1.40311985,
+      "memory(GiB)": 123.79,
+      "step": 60990,
+      "train_speed(iter/s)": 1.317103
+    },
+    {
+      "acc": 0.69820647,
+      "epoch": 1.5473110096397766,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.3364732832029315e-06,
+      "loss": 1.35926037,
+      "memory(GiB)": 123.79,
+      "step": 60995,
+      "train_speed(iter/s)": 1.317115
+    },
+    {
+      "acc": 0.69369326,
+      "epoch": 1.5474378488077118,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.335759727110057e-06,
+      "loss": 1.34865913,
+      "memory(GiB)": 123.79,
+      "step": 61000,
+      "train_speed(iter/s)": 1.317126
+    },
+    {
+      "epoch": 1.5474378488077118,
+      "eval_acc": 0.6750954877969833,
+      "eval_loss": 1.3562579154968262,
+      "eval_runtime": 70.1245,
+      "eval_samples_per_second": 90.838,
+      "eval_steps_per_second": 22.717,
+      "step": 61000
+    },
+    {
+      "acc": 0.68403788,
+      "epoch": 1.5475646879756468,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.335046332186829e-06,
+      "loss": 1.43646593,
+      "memory(GiB)": 123.79,
+      "step": 61005,
+      "train_speed(iter/s)": 1.314793
+    },
+    {
+      "acc": 0.69234982,
+      "epoch": 1.547691527143582,
+      "grad_norm": 3.25,
+      "learning_rate": 1.3343330984646262e-06,
+      "loss": 1.34765835,
+      "memory(GiB)": 123.79,
+      "step": 61010,
+      "train_speed(iter/s)": 1.314804
+    },
+    {
+      "acc": 0.70378704,
+      "epoch": 1.547818366311517,
+      "grad_norm": 2.75,
+      "learning_rate": 1.33362002597482e-06,
+      "loss": 1.38828716,
+      "memory(GiB)": 123.79,
+      "step": 61015,
+      "train_speed(iter/s)": 1.314816
+    },
+    {
+      "acc": 0.68200235,
+      "epoch": 1.547945205479452,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3329071147487743e-06,
+      "loss": 1.40856361,
+      "memory(GiB)": 123.79,
+      "step": 61020,
+      "train_speed(iter/s)": 1.314828
+    },
+    {
+      "acc": 0.68930206,
+      "epoch": 1.548072044647387,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3321943648178442e-06,
+      "loss": 1.34202213,
+      "memory(GiB)": 123.79,
+      "step": 61025,
+      "train_speed(iter/s)": 1.314693
+    },
+    {
+      "acc": 0.6830204,
+      "epoch": 1.5481988838153222,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.3314817762133848e-06,
+      "loss": 1.41982441,
+      "memory(GiB)": 123.79,
+      "step": 61030,
+      "train_speed(iter/s)": 1.314704
+    },
+    {
+      "acc": 0.69252415,
+      "epoch": 1.5483257229832572,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.330769348966734e-06,
+      "loss": 1.42234859,
+      "memory(GiB)": 123.79,
+      "step": 61035,
+      "train_speed(iter/s)": 1.314716
+    },
+    {
+      "acc": 0.69202833,
+      "epoch": 1.5484525621511924,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3300570831092292e-06,
+      "loss": 1.38016491,
+      "memory(GiB)": 123.79,
+      "step": 61040,
+      "train_speed(iter/s)": 1.314727
+    },
+    {
+      "acc": 0.69759765,
+      "epoch": 1.5485794013191274,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.3293449786721973e-06,
+      "loss": 1.30055561,
+      "memory(GiB)": 123.79,
+      "step": 61045,
+      "train_speed(iter/s)": 1.314738
+    },
+    {
+      "acc": 0.68279853,
+      "epoch": 1.5487062404870624,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3286330356869648e-06,
+      "loss": 1.42690935,
+      "memory(GiB)": 123.79,
+      "step": 61050,
+      "train_speed(iter/s)": 1.314749
+    },
+    {
+      "acc": 0.67251987,
+      "epoch": 1.5488330796549974,
+      "grad_norm": 3.375,
+      "learning_rate": 1.3279212541848413e-06,
+      "loss": 1.43316994,
+      "memory(GiB)": 123.79,
+      "step": 61055,
+      "train_speed(iter/s)": 1.31476
+    },
+    {
+      "acc": 0.67727175,
+      "epoch": 1.5489599188229324,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.3272096341971342e-06,
+      "loss": 1.40444841,
+      "memory(GiB)": 123.79,
+      "step": 61060,
+      "train_speed(iter/s)": 1.314772
+    },
+    {
+      "acc": 0.68380771,
+      "epoch": 1.5490867579908676,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.326498175755147e-06,
+      "loss": 1.44330044,
+      "memory(GiB)": 123.79,
+      "step": 61065,
+      "train_speed(iter/s)": 1.314636
+    },
+    {
+      "acc": 0.69062757,
+      "epoch": 1.5492135971588028,
+      "grad_norm": 4.375,
+      "learning_rate": 1.3257868788901722e-06,
+      "loss": 1.41743116,
+      "memory(GiB)": 123.79,
+      "step": 61070,
+      "train_speed(iter/s)": 1.314646
+    },
+    {
+      "acc": 0.68940001,
+      "epoch": 1.5493404363267378,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3250757436334932e-06,
+      "loss": 1.35631428,
+      "memory(GiB)": 123.79,
+      "step": 61075,
+      "train_speed(iter/s)": 1.314657
+    },
+    {
+      "acc": 0.68899283,
+      "epoch": 1.5494672754946728,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.3243647700163887e-06,
+      "loss": 1.34472208,
+      "memory(GiB)": 123.79,
+      "step": 61080,
+      "train_speed(iter/s)": 1.31467
+    },
+    {
+      "acc": 0.68681054,
+      "epoch": 1.5495941146626078,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.323653958070134e-06,
+      "loss": 1.40588427,
+      "memory(GiB)": 123.79,
+      "step": 61085,
+      "train_speed(iter/s)": 1.314681
+    },
+    {
+      "acc": 0.68823657,
+      "epoch": 1.5497209538305428,
+      "grad_norm": 3.75,
+      "learning_rate": 1.3229433078259928e-06,
+      "loss": 1.36990786,
+      "memory(GiB)": 123.79,
+      "step": 61090,
+      "train_speed(iter/s)": 1.314693
+    },
+    {
+      "acc": 0.68314652,
+      "epoch": 1.549847792998478,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.3222328193152195e-06,
+      "loss": 1.34861364,
+      "memory(GiB)": 123.79,
+      "step": 61095,
+      "train_speed(iter/s)": 1.314703
+    },
+    {
+      "acc": 0.69967694,
+      "epoch": 1.549974632166413,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.3215224925690683e-06,
+      "loss": 1.41521063,
+      "memory(GiB)": 123.79,
+      "step": 61100,
+      "train_speed(iter/s)": 1.314715
+    },
+    {
+      "acc": 0.71058102,
+      "epoch": 1.5501014713343482,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.3208123276187807e-06,
+      "loss": 1.31173306,
+      "memory(GiB)": 123.79,
+      "step": 61105,
+      "train_speed(iter/s)": 1.314723
+    },
+    {
+      "acc": 0.68196573,
+      "epoch": 1.5502283105022832,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.3201023244955952e-06,
+      "loss": 1.41337643,
+      "memory(GiB)": 123.79,
+      "step": 61110,
+      "train_speed(iter/s)": 1.314734
+    },
+    {
+      "acc": 0.6995482,
+      "epoch": 1.5503551496702181,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.319392483230736e-06,
+      "loss": 1.36041336,
+      "memory(GiB)": 123.79,
+      "step": 61115,
+      "train_speed(iter/s)": 1.314746
+    },
+    {
+      "acc": 0.6773037,
+      "epoch": 1.5504819888381531,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.318682803855429e-06,
+      "loss": 1.46533747,
+      "memory(GiB)": 123.79,
+      "step": 61120,
+      "train_speed(iter/s)": 1.314758
+    },
+    {
+      "acc": 0.68901405,
+      "epoch": 1.5506088280060881,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.3179732864008888e-06,
+      "loss": 1.41198788,
+      "memory(GiB)": 123.79,
+      "step": 61125,
+      "train_speed(iter/s)": 1.31477
+    },
+    {
+      "acc": 0.68291626,
+      "epoch": 1.5507356671740233,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.3172639308983226e-06,
+      "loss": 1.37107105,
+      "memory(GiB)": 123.79,
+      "step": 61130,
+      "train_speed(iter/s)": 1.314782
+    },
+    {
+      "acc": 0.69171982,
+      "epoch": 1.5508625063419585,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.3165547373789306e-06,
+      "loss": 1.35550508,
+      "memory(GiB)": 123.79,
+      "step": 61135,
+      "train_speed(iter/s)": 1.314794
+    },
+    {
+      "acc": 0.68221655,
+      "epoch": 1.5509893455098935,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.3158457058739066e-06,
+      "loss": 1.39664688,
+      "memory(GiB)": 123.79,
+      "step": 61140,
+      "train_speed(iter/s)": 1.314804
+    },
+    {
+      "acc": 0.68721566,
+      "epoch": 1.5511161846778285,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.3151368364144373e-06,
+      "loss": 1.45044756,
+      "memory(GiB)": 123.79,
+      "step": 61145,
+      "train_speed(iter/s)": 1.314816
+    },
+    {
+      "acc": 0.68738241,
+      "epoch": 1.5512430238457635,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.3144281290317012e-06,
+      "loss": 1.3901062,
+      "memory(GiB)": 123.79,
+      "step": 61150,
+      "train_speed(iter/s)": 1.314828
+    },
+    {
+      "acc": 0.68540936,
+      "epoch": 1.5513698630136985,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.3137195837568716e-06,
+      "loss": 1.32636223,
+      "memory(GiB)": 123.79,
+      "step": 61155,
+      "train_speed(iter/s)": 1.31484
+    },
+    {
+      "acc": 0.69074211,
+      "epoch": 1.5514967021816337,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.313011200621112e-06,
+      "loss": 1.3690918,
+      "memory(GiB)": 123.79,
+      "step": 61160,
+      "train_speed(iter/s)": 1.314853
+    },
+    {
+      "acc": 0.68953896,
+      "epoch": 1.5516235413495687,
+      "grad_norm": 2.390625,
+      "learning_rate": 1.312302979655582e-06,
+      "loss": 1.36859608,
+      "memory(GiB)": 123.79,
+      "step": 61165,
+      "train_speed(iter/s)": 1.314866
+    },
+    {
+      "acc": 0.68365393,
+      "epoch": 1.551750380517504,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.3115949208914302e-06,
+      "loss": 1.37258015,
+      "memory(GiB)": 123.79,
+      "step": 61170,
+      "train_speed(iter/s)": 1.314878
+    },
+    {
+      "acc": 0.68699965,
+      "epoch": 1.551877219685439,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.3108870243598022e-06,
+      "loss": 1.36818523,
+      "memory(GiB)": 123.79,
+      "step": 61175,
+      "train_speed(iter/s)": 1.31489
+    },
+    {
+      "acc": 0.68280296,
+      "epoch": 1.552004058853374,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.310179290091833e-06,
+      "loss": 1.42988014,
+      "memory(GiB)": 123.79,
+      "step": 61180,
+      "train_speed(iter/s)": 1.314903
+    },
+    {
+      "acc": 0.70891237,
+      "epoch": 1.5521308980213089,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3094717181186518e-06,
+      "loss": 1.35815372,
+      "memory(GiB)": 123.79,
+      "step": 61185,
+      "train_speed(iter/s)": 1.314915
+    },
+    {
+      "acc": 0.68528719,
+      "epoch": 1.552257737189244,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.3087643084713836e-06,
+      "loss": 1.40271988,
+      "memory(GiB)": 123.79,
+      "step": 61190,
+      "train_speed(iter/s)": 1.314927
+    },
+    {
+      "acc": 0.69638929,
+      "epoch": 1.552384576357179,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.30805706118114e-06,
+      "loss": 1.36729813,
+      "memory(GiB)": 123.79,
+      "step": 61195,
+      "train_speed(iter/s)": 1.31494
+    },
+    {
+      "acc": 0.6838068,
+      "epoch": 1.5525114155251143,
+      "grad_norm": 3.5,
+      "learning_rate": 1.3073499762790287e-06,
+      "loss": 1.41148405,
+      "memory(GiB)": 123.79,
+      "step": 61200,
+      "train_speed(iter/s)": 1.314952
+    },
+    {
+      "acc": 0.69247351,
+      "epoch": 1.5526382546930493,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.306643053796154e-06,
+      "loss": 1.38352766,
+      "memory(GiB)": 123.79,
+      "step": 61205,
+      "train_speed(iter/s)": 1.314965
+    },
+    {
+      "acc": 0.69567947,
+      "epoch": 1.5527650938609843,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.3059362937636084e-06,
+      "loss": 1.40714769,
+      "memory(GiB)": 123.79,
+      "step": 61210,
+      "train_speed(iter/s)": 1.314978
+    },
+    {
+      "acc": 0.70161033,
+      "epoch": 1.5528919330289193,
+      "grad_norm": 3.25,
+      "learning_rate": 1.3052296962124756e-06,
+      "loss": 1.30638237,
+      "memory(GiB)": 123.79,
+      "step": 61215,
+      "train_speed(iter/s)": 1.314989
+    },
+    {
+      "acc": 0.67648444,
+      "epoch": 1.5530187721968542,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.3045232611738357e-06,
+      "loss": 1.43010321,
+      "memory(GiB)": 123.79,
+      "step": 61220,
+      "train_speed(iter/s)": 1.315001
+    },
+    {
+      "acc": 0.68273582,
+      "epoch": 1.5531456113647895,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3038169886787632e-06,
+      "loss": 1.40489492,
+      "memory(GiB)": 123.79,
+      "step": 61225,
+      "train_speed(iter/s)": 1.315013
+    },
+    {
+      "acc": 0.71136184,
+      "epoch": 1.5532724505327247,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.3031108787583235e-06,
+      "loss": 1.32762489,
+      "memory(GiB)": 123.79,
+      "step": 61230,
+      "train_speed(iter/s)": 1.315025
+    },
+    {
+      "acc": 0.67805586,
+      "epoch": 1.5533992897006597,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.3024049314435694e-06,
+      "loss": 1.38393373,
+      "memory(GiB)": 123.79,
+      "step": 61235,
+      "train_speed(iter/s)": 1.315036
+    },
+    {
+      "acc": 0.67650003,
+      "epoch": 1.5535261288685946,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.301699146765557e-06,
+      "loss": 1.40756493,
+      "memory(GiB)": 123.79,
+      "step": 61240,
+      "train_speed(iter/s)": 1.315048
+    },
+    {
+      "acc": 0.70192719,
+      "epoch": 1.5536529680365296,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.3009935247553274e-06,
+      "loss": 1.389639,
+      "memory(GiB)": 123.79,
+      "step": 61245,
+      "train_speed(iter/s)": 1.31506
+    },
+    {
+      "acc": 0.67412519,
+      "epoch": 1.5537798072044646,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3002880654439192e-06,
+      "loss": 1.44991989,
+      "memory(GiB)": 123.79,
+      "step": 61250,
+      "train_speed(iter/s)": 1.315073
+    },
+    {
+      "acc": 0.675772,
+      "epoch": 1.5539066463723998,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.2995827688623568e-06,
+      "loss": 1.45188913,
+      "memory(GiB)": 123.79,
+      "step": 61255,
+      "train_speed(iter/s)": 1.315085
+    },
+    {
+      "acc": 0.69584169,
+      "epoch": 1.5540334855403348,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.298877635041667e-06,
+      "loss": 1.32171078,
+      "memory(GiB)": 123.79,
+      "step": 61260,
+      "train_speed(iter/s)": 1.315097
+    },
+    {
+      "acc": 0.67112861,
+      "epoch": 1.55416032470827,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.2981726640128633e-06,
+      "loss": 1.43237057,
+      "memory(GiB)": 123.79,
+      "step": 61265,
+      "train_speed(iter/s)": 1.31511
+    },
+    {
+      "acc": 0.69702263,
+      "epoch": 1.554287163876205,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.297467855806953e-06,
+      "loss": 1.32280846,
+      "memory(GiB)": 123.79,
+      "step": 61270,
+      "train_speed(iter/s)": 1.315122
+    },
+    {
+      "acc": 0.68550692,
+      "epoch": 1.55441400304414,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.2967632104549371e-06,
+      "loss": 1.40776272,
+      "memory(GiB)": 123.79,
+      "step": 61275,
+      "train_speed(iter/s)": 1.315135
+    },
+    {
+      "acc": 0.68628335,
+      "epoch": 1.554540842212075,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.296058727987809e-06,
+      "loss": 1.37993593,
+      "memory(GiB)": 123.79,
+      "step": 61280,
+      "train_speed(iter/s)": 1.315147
+    },
+    {
+      "acc": 0.68553424,
+      "epoch": 1.55466768138001,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.295354408436555e-06,
+      "loss": 1.45037403,
+      "memory(GiB)": 123.79,
+      "step": 61285,
+      "train_speed(iter/s)": 1.31516
+    },
+    {
+      "acc": 0.69571867,
+      "epoch": 1.5547945205479452,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.294650251832154e-06,
+      "loss": 1.36079607,
+      "memory(GiB)": 123.79,
+      "step": 61290,
+      "train_speed(iter/s)": 1.315172
+    },
+    {
+      "acc": 0.68325148,
+      "epoch": 1.5549213597158804,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.2939462582055784e-06,
+      "loss": 1.36695766,
+      "memory(GiB)": 123.79,
+      "step": 61295,
+      "train_speed(iter/s)": 1.315184
+    },
+    {
+      "acc": 0.69590845,
+      "epoch": 1.5550481988838154,
+      "grad_norm": 3.0,
+      "learning_rate": 1.2932424275877926e-06,
+      "loss": 1.34241486,
+      "memory(GiB)": 123.79,
+      "step": 61300,
+      "train_speed(iter/s)": 1.315197
+    },
+    {
+      "acc": 0.68916683,
+      "epoch": 1.5551750380517504,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.2925387600097543e-06,
+      "loss": 1.37341394,
+      "memory(GiB)": 123.79,
+      "step": 61305,
+      "train_speed(iter/s)": 1.31521
+    },
+    {
+      "acc": 0.68023915,
+      "epoch": 1.5553018772196854,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.291835255502414e-06,
+      "loss": 1.35712595,
+      "memory(GiB)": 123.79,
+      "step": 61310,
+      "train_speed(iter/s)": 1.315222
+    },
+    {
+      "acc": 0.67570772,
+      "epoch": 1.5554287163876204,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2911319140967148e-06,
+      "loss": 1.38876343,
+      "memory(GiB)": 123.79,
+      "step": 61315,
+      "train_speed(iter/s)": 1.315234
+    },
+    {
+      "acc": 0.68611865,
+      "epoch": 1.5555555555555556,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.290428735823593e-06,
+      "loss": 1.41211529,
+      "memory(GiB)": 123.79,
+      "step": 61320,
+      "train_speed(iter/s)": 1.315247
+    },
+    {
+      "acc": 0.67348614,
+      "epoch": 1.5556823947234906,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2897257207139758e-06,
+      "loss": 1.46037035,
+      "memory(GiB)": 123.79,
+      "step": 61325,
+      "train_speed(iter/s)": 1.315259
+    },
+    {
+      "acc": 0.69116149,
+      "epoch": 1.5558092338914258,
+      "grad_norm": 3.0,
+      "learning_rate": 1.28902286879879e-06,
+      "loss": 1.35616903,
+      "memory(GiB)": 123.79,
+      "step": 61330,
+      "train_speed(iter/s)": 1.315272
+    },
+    {
+      "acc": 0.674508,
+      "epoch": 1.5559360730593608,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.2883201801089445e-06,
+      "loss": 1.4247961,
+      "memory(GiB)": 123.79,
+      "step": 61335,
+      "train_speed(iter/s)": 1.315284
+    },
+    {
+      "acc": 0.66922302,
+      "epoch": 1.5560629122272958,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.2876176546753494e-06,
+      "loss": 1.45676498,
+      "memory(GiB)": 123.79,
+      "step": 61340,
+      "train_speed(iter/s)": 1.315297
+    },
+    {
+      "acc": 0.68606925,
+      "epoch": 1.5561897513952307,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.286915292528903e-06,
+      "loss": 1.43588524,
+      "memory(GiB)": 123.79,
+      "step": 61345,
+      "train_speed(iter/s)": 1.315309
+    },
+    {
+      "acc": 0.68981185,
+      "epoch": 1.556316590563166,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.286213093700503e-06,
+      "loss": 1.37024374,
+      "memory(GiB)": 123.79,
+      "step": 61350,
+      "train_speed(iter/s)": 1.315321
+    },
+    {
+      "acc": 0.69421911,
+      "epoch": 1.556443429731101,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.28551105822103e-06,
+      "loss": 1.35217724,
+      "memory(GiB)": 123.79,
+      "step": 61355,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.69101038,
+      "epoch": 1.5565702688990362,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.2848091861213636e-06,
+      "loss": 1.37251329,
+      "memory(GiB)": 123.79,
+      "step": 61360,
+      "train_speed(iter/s)": 1.315346
+    },
+    {
+      "acc": 0.67027588,
+      "epoch": 1.5566971080669711,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.2841074774323775e-06,
+      "loss": 1.48669033,
+      "memory(GiB)": 123.79,
+      "step": 61365,
+      "train_speed(iter/s)": 1.315358
+    },
+    {
+      "acc": 0.6753067,
+      "epoch": 1.5568239472349061,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.2834059321849363e-06,
+      "loss": 1.42558517,
+      "memory(GiB)": 123.79,
+      "step": 61370,
+      "train_speed(iter/s)": 1.31537
+    },
+    {
+      "acc": 0.70902457,
+      "epoch": 1.5569507864028411,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.2827045504098928e-06,
+      "loss": 1.37919683,
+      "memory(GiB)": 123.79,
+      "step": 61375,
+      "train_speed(iter/s)": 1.315382
+    },
+    {
+      "acc": 0.68793559,
+      "epoch": 1.5570776255707761,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.2820033321381009e-06,
+      "loss": 1.36889267,
+      "memory(GiB)": 123.79,
+      "step": 61380,
+      "train_speed(iter/s)": 1.315393
+    },
+    {
+      "acc": 0.70275888,
+      "epoch": 1.5572044647387113,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.2813022774004024e-06,
+      "loss": 1.32633057,
+      "memory(GiB)": 123.79,
+      "step": 61385,
+      "train_speed(iter/s)": 1.315405
+    },
+    {
+      "acc": 0.69363747,
+      "epoch": 1.5573313039066465,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.280601386227634e-06,
+      "loss": 1.40708103,
+      "memory(GiB)": 123.79,
+      "step": 61390,
+      "train_speed(iter/s)": 1.315417
+    },
+    {
+      "acc": 0.68439279,
+      "epoch": 1.5574581430745815,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.279900658650619e-06,
+      "loss": 1.41965446,
+      "memory(GiB)": 123.79,
+      "step": 61395,
+      "train_speed(iter/s)": 1.315429
+    },
+    {
+      "acc": 0.6747911,
+      "epoch": 1.5575849822425165,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.2792000947001842e-06,
+      "loss": 1.43047495,
+      "memory(GiB)": 123.79,
+      "step": 61400,
+      "train_speed(iter/s)": 1.315441
+    },
+    {
+      "acc": 0.69073739,
+      "epoch": 1.5577118214104515,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.2784996944071415e-06,
+      "loss": 1.38692093,
+      "memory(GiB)": 123.79,
+      "step": 61405,
+      "train_speed(iter/s)": 1.315454
+    },
+    {
+      "acc": 0.68613753,
+      "epoch": 1.5578386605783865,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.2777994578022972e-06,
+      "loss": 1.43375435,
+      "memory(GiB)": 123.79,
+      "step": 61410,
+      "train_speed(iter/s)": 1.315466
+    },
+    {
+      "acc": 0.68476596,
+      "epoch": 1.5579654997463217,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.2770993849164514e-06,
+      "loss": 1.42668591,
+      "memory(GiB)": 123.79,
+      "step": 61415,
+      "train_speed(iter/s)": 1.315478
+    },
+    {
+      "acc": 0.66656027,
+      "epoch": 1.5580923389142567,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.276399475780396e-06,
+      "loss": 1.51126671,
+      "memory(GiB)": 123.79,
+      "step": 61420,
+      "train_speed(iter/s)": 1.315491
+    },
+    {
+      "acc": 0.69086628,
+      "epoch": 1.558219178082192,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.2756997304249164e-06,
+      "loss": 1.41039791,
+      "memory(GiB)": 123.79,
+      "step": 61425,
+      "train_speed(iter/s)": 1.315504
+    },
+    {
+      "acc": 0.69260855,
+      "epoch": 1.558346017250127,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.2750001488807906e-06,
+      "loss": 1.34077339,
+      "memory(GiB)": 123.79,
+      "step": 61430,
+      "train_speed(iter/s)": 1.315516
+    },
+    {
+      "acc": 0.69731398,
+      "epoch": 1.5584728564180619,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.2743007311787892e-06,
+      "loss": 1.4092205,
+      "memory(GiB)": 123.79,
+      "step": 61435,
+      "train_speed(iter/s)": 1.315528
+    },
+    {
+      "acc": 0.69402409,
+      "epoch": 1.5585996955859969,
+      "grad_norm": 3.375,
+      "learning_rate": 1.2736014773496757e-06,
+      "loss": 1.33651142,
+      "memory(GiB)": 123.79,
+      "step": 61440,
+      "train_speed(iter/s)": 1.315541
+    },
+    {
+      "acc": 0.69668808,
+      "epoch": 1.5587265347539319,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.2729023874242064e-06,
+      "loss": 1.30267906,
+      "memory(GiB)": 123.79,
+      "step": 61445,
+      "train_speed(iter/s)": 1.315553
+    },
+    {
+      "acc": 0.67070975,
+      "epoch": 1.558853373921867,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.2722034614331303e-06,
+      "loss": 1.49753561,
+      "memory(GiB)": 123.79,
+      "step": 61450,
+      "train_speed(iter/s)": 1.315565
+    },
+    {
+      "acc": 0.672925,
+      "epoch": 1.5589802130898023,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.2715046994071889e-06,
+      "loss": 1.43763628,
+      "memory(GiB)": 123.79,
+      "step": 61455,
+      "train_speed(iter/s)": 1.315578
+    },
+    {
+      "acc": 0.67800369,
+      "epoch": 1.5591070522577373,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2708061013771179e-06,
+      "loss": 1.39759359,
+      "memory(GiB)": 123.79,
+      "step": 61460,
+      "train_speed(iter/s)": 1.31559
+    },
+    {
+      "acc": 0.68950367,
+      "epoch": 1.5592338914256723,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2701076673736428e-06,
+      "loss": 1.34615631,
+      "memory(GiB)": 123.79,
+      "step": 61465,
+      "train_speed(iter/s)": 1.315602
+    },
+    {
+      "acc": 0.68108625,
+      "epoch": 1.5593607305936072,
+      "grad_norm": 3.125,
+      "learning_rate": 1.269409397427488e-06,
+      "loss": 1.40873203,
+      "memory(GiB)": 123.79,
+      "step": 61470,
+      "train_speed(iter/s)": 1.315614
+    },
+    {
+      "acc": 0.6888587,
+      "epoch": 1.5594875697615422,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.2687112915693622e-06,
+      "loss": 1.33907909,
+      "memory(GiB)": 123.79,
+      "step": 61475,
+      "train_speed(iter/s)": 1.315626
+    },
+    {
+      "acc": 0.67166815,
+      "epoch": 1.5596144089294774,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.2680133498299729e-06,
+      "loss": 1.46527004,
+      "memory(GiB)": 123.79,
+      "step": 61480,
+      "train_speed(iter/s)": 1.315638
+    },
+    {
+      "acc": 0.68372464,
+      "epoch": 1.5597412480974124,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2673155722400177e-06,
+      "loss": 1.38644857,
+      "memory(GiB)": 123.79,
+      "step": 61485,
+      "train_speed(iter/s)": 1.315649
+    },
+    {
+      "acc": 0.6810287,
+      "epoch": 1.5598680872653476,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2666179588301908e-06,
+      "loss": 1.47898855,
+      "memory(GiB)": 123.79,
+      "step": 61490,
+      "train_speed(iter/s)": 1.315661
+    },
+    {
+      "acc": 0.69384875,
+      "epoch": 1.5599949264332826,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.2659205096311738e-06,
+      "loss": 1.37034664,
+      "memory(GiB)": 123.79,
+      "step": 61495,
+      "train_speed(iter/s)": 1.315674
+    },
+    {
+      "acc": 0.67693381,
+      "epoch": 1.5601217656012176,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.2652232246736423e-06,
+      "loss": 1.38674755,
+      "memory(GiB)": 123.79,
+      "step": 61500,
+      "train_speed(iter/s)": 1.315686
+    },
+    {
+      "acc": 0.69492741,
+      "epoch": 1.5602486047691526,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2645261039882694e-06,
+      "loss": 1.44975128,
+      "memory(GiB)": 123.79,
+      "step": 61505,
+      "train_speed(iter/s)": 1.315697
+    },
+    {
+      "acc": 0.67765694,
+      "epoch": 1.5603754439370878,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.263829147605718e-06,
+      "loss": 1.41589565,
+      "memory(GiB)": 123.79,
+      "step": 61510,
+      "train_speed(iter/s)": 1.315708
+    },
+    {
+      "acc": 0.69323287,
+      "epoch": 1.5605022831050228,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.26313235555664e-06,
+      "loss": 1.40820608,
+      "memory(GiB)": 123.79,
+      "step": 61515,
+      "train_speed(iter/s)": 1.31572
+    },
+    {
+      "acc": 0.70413375,
+      "epoch": 1.560629122272958,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.2624357278716832e-06,
+      "loss": 1.34282703,
+      "memory(GiB)": 123.79,
+      "step": 61520,
+      "train_speed(iter/s)": 1.315733
+    },
+    {
+      "acc": 0.67251682,
+      "epoch": 1.560755961440893,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.2617392645814913e-06,
+      "loss": 1.37273293,
+      "memory(GiB)": 123.79,
+      "step": 61525,
+      "train_speed(iter/s)": 1.315745
+    },
+    {
+      "acc": 0.67997522,
+      "epoch": 1.560882800608828,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.2610429657166983e-06,
+      "loss": 1.47229862,
+      "memory(GiB)": 123.79,
+      "step": 61530,
+      "train_speed(iter/s)": 1.315757
+    },
+    {
+      "acc": 0.70328679,
+      "epoch": 1.561009639776763,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.2603468313079265e-06,
+      "loss": 1.31386824,
+      "memory(GiB)": 123.79,
+      "step": 61535,
+      "train_speed(iter/s)": 1.315768
+    },
+    {
+      "acc": 0.6960535,
+      "epoch": 1.561136478944698,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.2596508613857982e-06,
+      "loss": 1.33943043,
+      "memory(GiB)": 123.79,
+      "step": 61540,
+      "train_speed(iter/s)": 1.31578
+    },
+    {
+      "acc": 0.67977457,
+      "epoch": 1.5612633181126332,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.258955055980925e-06,
+      "loss": 1.36414557,
+      "memory(GiB)": 123.79,
+      "step": 61545,
+      "train_speed(iter/s)": 1.315792
+    },
+    {
+      "acc": 0.68356562,
+      "epoch": 1.5613901572805684,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.258259415123911e-06,
+      "loss": 1.38391075,
+      "memory(GiB)": 123.79,
+      "step": 61550,
+      "train_speed(iter/s)": 1.315804
+    },
+    {
+      "acc": 0.70062356,
+      "epoch": 1.5615169964485034,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2575639388453532e-06,
+      "loss": 1.2990694,
+      "memory(GiB)": 123.79,
+      "step": 61555,
+      "train_speed(iter/s)": 1.315816
+    },
+    {
+      "acc": 0.69281988,
+      "epoch": 1.5616438356164384,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.2568686271758423e-06,
+      "loss": 1.34530077,
+      "memory(GiB)": 123.79,
+      "step": 61560,
+      "train_speed(iter/s)": 1.315828
+    },
+    {
+      "acc": 0.68168955,
+      "epoch": 1.5617706747843734,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2561734801459612e-06,
+      "loss": 1.4308898,
+      "memory(GiB)": 123.79,
+      "step": 61565,
+      "train_speed(iter/s)": 1.315839
+    },
+    {
+      "acc": 0.69640822,
+      "epoch": 1.5618975139523084,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.2554784977862856e-06,
+      "loss": 1.37560854,
+      "memory(GiB)": 123.79,
+      "step": 61570,
+      "train_speed(iter/s)": 1.315851
+    },
+    {
+      "acc": 0.68605652,
+      "epoch": 1.5620243531202436,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.2547836801273833e-06,
+      "loss": 1.4015317,
+      "memory(GiB)": 123.79,
+      "step": 61575,
+      "train_speed(iter/s)": 1.315863
+    },
+    {
+      "acc": 0.68168411,
+      "epoch": 1.5621511922881786,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.2540890271998162e-06,
+      "loss": 1.42549829,
+      "memory(GiB)": 123.79,
+      "step": 61580,
+      "train_speed(iter/s)": 1.315875
+    },
+    {
+      "acc": 0.67510223,
+      "epoch": 1.5622780314561138,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.2533945390341379e-06,
+      "loss": 1.46564636,
+      "memory(GiB)": 123.79,
+      "step": 61585,
+      "train_speed(iter/s)": 1.315885
+    },
+    {
+      "acc": 0.68378386,
+      "epoch": 1.5624048706240488,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.2527002156608946e-06,
+      "loss": 1.39191093,
+      "memory(GiB)": 123.79,
+      "step": 61590,
+      "train_speed(iter/s)": 1.315897
+    },
+    {
+      "acc": 0.69002323,
+      "epoch": 1.5625317097919837,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.2520060571106275e-06,
+      "loss": 1.36828995,
+      "memory(GiB)": 123.79,
+      "step": 61595,
+      "train_speed(iter/s)": 1.315909
+    },
+    {
+      "acc": 0.67872238,
+      "epoch": 1.5626585489599187,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.2513120634138665e-06,
+      "loss": 1.46081161,
+      "memory(GiB)": 123.79,
+      "step": 61600,
+      "train_speed(iter/s)": 1.315773
+    },
+    {
+      "acc": 0.69009714,
+      "epoch": 1.5627853881278537,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.250618234601138e-06,
+      "loss": 1.35208511,
+      "memory(GiB)": 123.79,
+      "step": 61605,
+      "train_speed(iter/s)": 1.315785
+    },
+    {
+      "acc": 0.69107285,
+      "epoch": 1.562912227295789,
+      "grad_norm": 3.625,
+      "learning_rate": 1.2499245707029595e-06,
+      "loss": 1.34494963,
+      "memory(GiB)": 123.79,
+      "step": 61610,
+      "train_speed(iter/s)": 1.315796
+    },
+    {
+      "acc": 0.68198338,
+      "epoch": 1.5630390664637241,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.2492310717498412e-06,
+      "loss": 1.39510832,
+      "memory(GiB)": 123.79,
+      "step": 61615,
+      "train_speed(iter/s)": 1.315808
+    },
+    {
+      "acc": 0.70581098,
+      "epoch": 1.5631659056316591,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.2485377377722863e-06,
+      "loss": 1.39163055,
+      "memory(GiB)": 123.79,
+      "step": 61620,
+      "train_speed(iter/s)": 1.31582
+    },
+    {
+      "acc": 0.70258904,
+      "epoch": 1.5632927447995941,
+      "grad_norm": 5.8125,
+      "learning_rate": 1.2478445688007894e-06,
+      "loss": 1.36478109,
+      "memory(GiB)": 123.79,
+      "step": 61625,
+      "train_speed(iter/s)": 1.315833
+    },
+    {
+      "acc": 0.69323606,
+      "epoch": 1.5634195839675291,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.2471515648658434e-06,
+      "loss": 1.33040609,
+      "memory(GiB)": 123.79,
+      "step": 61630,
+      "train_speed(iter/s)": 1.315845
+    },
+    {
+      "acc": 0.69695783,
+      "epoch": 1.563546423135464,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.2464587259979254e-06,
+      "loss": 1.39331713,
+      "memory(GiB)": 123.79,
+      "step": 61635,
+      "train_speed(iter/s)": 1.315857
+    },
+    {
+      "acc": 0.68258929,
+      "epoch": 1.5636732623033993,
+      "grad_norm": 5.0,
+      "learning_rate": 1.2457660522275095e-06,
+      "loss": 1.39813347,
+      "memory(GiB)": 123.79,
+      "step": 61640,
+      "train_speed(iter/s)": 1.315869
+    },
+    {
+      "acc": 0.69321122,
+      "epoch": 1.5638001014713343,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.2450735435850654e-06,
+      "loss": 1.36430426,
+      "memory(GiB)": 123.79,
+      "step": 61645,
+      "train_speed(iter/s)": 1.315881
+    },
+    {
+      "acc": 0.68299646,
+      "epoch": 1.5639269406392695,
+      "grad_norm": 3.5,
+      "learning_rate": 1.244381200101053e-06,
+      "loss": 1.43121071,
+      "memory(GiB)": 123.79,
+      "step": 61650,
+      "train_speed(iter/s)": 1.315893
+    },
+    {
+      "acc": 0.68388906,
+      "epoch": 1.5640537798072045,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2436890218059217e-06,
+      "loss": 1.42835426,
+      "memory(GiB)": 123.79,
+      "step": 61655,
+      "train_speed(iter/s)": 1.315904
+    },
+    {
+      "acc": 0.68930569,
+      "epoch": 1.5641806189751395,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.2429970087301163e-06,
+      "loss": 1.35427513,
+      "memory(GiB)": 123.79,
+      "step": 61660,
+      "train_speed(iter/s)": 1.315916
+    },
+    {
+      "acc": 0.67977509,
+      "epoch": 1.5643074581430745,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.2423051609040777e-06,
+      "loss": 1.44750452,
+      "memory(GiB)": 123.79,
+      "step": 61665,
+      "train_speed(iter/s)": 1.315928
+    },
+    {
+      "acc": 0.67713938,
+      "epoch": 1.5644342973110097,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.2416134783582368e-06,
+      "loss": 1.4182128,
+      "memory(GiB)": 123.79,
+      "step": 61670,
+      "train_speed(iter/s)": 1.315939
+    },
+    {
+      "acc": 0.68381662,
+      "epoch": 1.5645611364789447,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.2409219611230116e-06,
+      "loss": 1.41154995,
+      "memory(GiB)": 123.79,
+      "step": 61675,
+      "train_speed(iter/s)": 1.315951
+    },
+    {
+      "acc": 0.69612131,
+      "epoch": 1.56468797564688,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2402306092288236e-06,
+      "loss": 1.34820976,
+      "memory(GiB)": 123.79,
+      "step": 61680,
+      "train_speed(iter/s)": 1.315964
+    },
+    {
+      "acc": 0.69564252,
+      "epoch": 1.5648148148148149,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.2395394227060793e-06,
+      "loss": 1.41899548,
+      "memory(GiB)": 123.79,
+      "step": 61685,
+      "train_speed(iter/s)": 1.315976
+    },
+    {
+      "acc": 0.67754097,
+      "epoch": 1.5649416539827499,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.238848401585182e-06,
+      "loss": 1.40724335,
+      "memory(GiB)": 123.79,
+      "step": 61690,
+      "train_speed(iter/s)": 1.315988
+    },
+    {
+      "acc": 0.67971234,
+      "epoch": 1.5650684931506849,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.2381575458965218e-06,
+      "loss": 1.36165924,
+      "memory(GiB)": 123.79,
+      "step": 61695,
+      "train_speed(iter/s)": 1.316
+    },
+    {
+      "acc": 0.68393731,
+      "epoch": 1.5651953323186198,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.2374668556704888e-06,
+      "loss": 1.39424963,
+      "memory(GiB)": 123.79,
+      "step": 61700,
+      "train_speed(iter/s)": 1.316012
+    },
+    {
+      "acc": 0.69531984,
+      "epoch": 1.565322171486555,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.2367763309374625e-06,
+      "loss": 1.34736376,
+      "memory(GiB)": 123.79,
+      "step": 61705,
+      "train_speed(iter/s)": 1.316024
+    },
+    {
+      "acc": 0.69165711,
+      "epoch": 1.5654490106544903,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.2360859717278145e-06,
+      "loss": 1.37240152,
+      "memory(GiB)": 123.79,
+      "step": 61710,
+      "train_speed(iter/s)": 1.316035
+    },
+    {
+      "acc": 0.67532835,
+      "epoch": 1.5655758498224253,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2353957780719106e-06,
+      "loss": 1.37645302,
+      "memory(GiB)": 123.79,
+      "step": 61715,
+      "train_speed(iter/s)": 1.316046
+    },
+    {
+      "acc": 0.68211827,
+      "epoch": 1.5657026889903602,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.2347057500001075e-06,
+      "loss": 1.47502365,
+      "memory(GiB)": 123.79,
+      "step": 61720,
+      "train_speed(iter/s)": 1.316058
+    },
+    {
+      "acc": 0.68647432,
+      "epoch": 1.5658295281582952,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2340158875427566e-06,
+      "loss": 1.39127369,
+      "memory(GiB)": 123.79,
+      "step": 61725,
+      "train_speed(iter/s)": 1.316069
+    },
+    {
+      "acc": 0.68395548,
+      "epoch": 1.5659563673262302,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.2333261907302013e-06,
+      "loss": 1.37475185,
+      "memory(GiB)": 123.79,
+      "step": 61730,
+      "train_speed(iter/s)": 1.316081
+    },
+    {
+      "acc": 0.70202713,
+      "epoch": 1.5660832064941654,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.2326366595927763e-06,
+      "loss": 1.35226421,
+      "memory(GiB)": 123.79,
+      "step": 61735,
+      "train_speed(iter/s)": 1.316093
+    },
+    {
+      "acc": 0.68491802,
+      "epoch": 1.5662100456621004,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.2319472941608118e-06,
+      "loss": 1.42308006,
+      "memory(GiB)": 123.79,
+      "step": 61740,
+      "train_speed(iter/s)": 1.316104
+    },
+    {
+      "acc": 0.68491621,
+      "epoch": 1.5663368848300356,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.231258094464628e-06,
+      "loss": 1.41449509,
+      "memory(GiB)": 123.79,
+      "step": 61745,
+      "train_speed(iter/s)": 1.316115
+    },
+    {
+      "acc": 0.68554058,
+      "epoch": 1.5664637239979706,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.23056906053454e-06,
+      "loss": 1.41339874,
+      "memory(GiB)": 123.79,
+      "step": 61750,
+      "train_speed(iter/s)": 1.316127
+    },
+    {
+      "acc": 0.67834358,
+      "epoch": 1.5665905631659056,
+      "grad_norm": 3.125,
+      "learning_rate": 1.2298801924008535e-06,
+      "loss": 1.43450832,
+      "memory(GiB)": 123.79,
+      "step": 61755,
+      "train_speed(iter/s)": 1.316139
+    },
+    {
+      "acc": 0.67976818,
+      "epoch": 1.5667174023338406,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.2291914900938685e-06,
+      "loss": 1.39567966,
+      "memory(GiB)": 123.79,
+      "step": 61760,
+      "train_speed(iter/s)": 1.316151
+    },
+    {
+      "acc": 0.68358278,
+      "epoch": 1.5668442415017756,
+      "grad_norm": 3.625,
+      "learning_rate": 1.2285029536438759e-06,
+      "loss": 1.39834919,
+      "memory(GiB)": 123.79,
+      "step": 61765,
+      "train_speed(iter/s)": 1.316163
+    },
+    {
+      "acc": 0.67618256,
+      "epoch": 1.5669710806697108,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.227814583081165e-06,
+      "loss": 1.41695461,
+      "memory(GiB)": 123.79,
+      "step": 61770,
+      "train_speed(iter/s)": 1.316174
+    },
+    {
+      "acc": 0.69286847,
+      "epoch": 1.567097919837646,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.2271263784360088e-06,
+      "loss": 1.36288433,
+      "memory(GiB)": 123.79,
+      "step": 61775,
+      "train_speed(iter/s)": 1.316186
+    },
+    {
+      "acc": 0.68895264,
+      "epoch": 1.567224759005581,
+      "grad_norm": 3.25,
+      "learning_rate": 1.2264383397386787e-06,
+      "loss": 1.43013411,
+      "memory(GiB)": 123.79,
+      "step": 61780,
+      "train_speed(iter/s)": 1.316198
+    },
+    {
+      "acc": 0.69182329,
+      "epoch": 1.567351598173516,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.225750467019437e-06,
+      "loss": 1.42511425,
+      "memory(GiB)": 123.79,
+      "step": 61785,
+      "train_speed(iter/s)": 1.316209
+    },
+    {
+      "acc": 0.68086443,
+      "epoch": 1.567478437341451,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.2250627603085435e-06,
+      "loss": 1.44101295,
+      "memory(GiB)": 123.79,
+      "step": 61790,
+      "train_speed(iter/s)": 1.316221
+    },
+    {
+      "acc": 0.68902264,
+      "epoch": 1.567605276509386,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.2243752196362423e-06,
+      "loss": 1.40020103,
+      "memory(GiB)": 123.79,
+      "step": 61795,
+      "train_speed(iter/s)": 1.316234
+    },
+    {
+      "acc": 0.68147359,
+      "epoch": 1.5677321156773212,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.2236878450327743e-06,
+      "loss": 1.4253294,
+      "memory(GiB)": 123.79,
+      "step": 61800,
+      "train_speed(iter/s)": 1.316245
+    },
+    {
+      "acc": 0.68250179,
+      "epoch": 1.5678589548452562,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.223000636528377e-06,
+      "loss": 1.42537041,
+      "memory(GiB)": 123.79,
+      "step": 61805,
+      "train_speed(iter/s)": 1.316257
+    },
+    {
+      "acc": 0.70313106,
+      "epoch": 1.5679857940131914,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.2223135941532754e-06,
+      "loss": 1.31244221,
+      "memory(GiB)": 123.79,
+      "step": 61810,
+      "train_speed(iter/s)": 1.316269
+    },
+    {
+      "acc": 0.68654633,
+      "epoch": 1.5681126331811264,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2216267179376857e-06,
+      "loss": 1.4048521,
+      "memory(GiB)": 123.79,
+      "step": 61815,
+      "train_speed(iter/s)": 1.31628
+    },
+    {
+      "acc": 0.69207096,
+      "epoch": 1.5682394723490614,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.2209400079118233e-06,
+      "loss": 1.40155153,
+      "memory(GiB)": 123.79,
+      "step": 61820,
+      "train_speed(iter/s)": 1.316292
+    },
+    {
+      "acc": 0.69281254,
+      "epoch": 1.5683663115169963,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.2202534641058916e-06,
+      "loss": 1.37735939,
+      "memory(GiB)": 123.79,
+      "step": 61825,
+      "train_speed(iter/s)": 1.316303
+    },
+    {
+      "acc": 0.70624461,
+      "epoch": 1.5684931506849316,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.2195670865500896e-06,
+      "loss": 1.26519871,
+      "memory(GiB)": 123.79,
+      "step": 61830,
+      "train_speed(iter/s)": 1.316314
+    },
+    {
+      "acc": 0.6897337,
+      "epoch": 1.5686199898528665,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.2188808752746022e-06,
+      "loss": 1.44145555,
+      "memory(GiB)": 123.79,
+      "step": 61835,
+      "train_speed(iter/s)": 1.316326
+    },
+    {
+      "acc": 0.70409374,
+      "epoch": 1.5687468290208018,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2181948303096176e-06,
+      "loss": 1.37430477,
+      "memory(GiB)": 123.79,
+      "step": 61840,
+      "train_speed(iter/s)": 1.316337
+    },
+    {
+      "acc": 0.68492899,
+      "epoch": 1.5688736681887367,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.2175089516853083e-06,
+      "loss": 1.35085354,
+      "memory(GiB)": 123.79,
+      "step": 61845,
+      "train_speed(iter/s)": 1.316349
+    },
+    {
+      "acc": 0.69685073,
+      "epoch": 1.5690005073566717,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.216823239431843e-06,
+      "loss": 1.35108795,
+      "memory(GiB)": 123.79,
+      "step": 61850,
+      "train_speed(iter/s)": 1.316361
+    },
+    {
+      "acc": 0.69750566,
+      "epoch": 1.5691273465246067,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.2161376935793827e-06,
+      "loss": 1.3638895,
+      "memory(GiB)": 123.79,
+      "step": 61855,
+      "train_speed(iter/s)": 1.316372
+    },
+    {
+      "acc": 0.68360748,
+      "epoch": 1.5692541856925417,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.21545231415808e-06,
+      "loss": 1.40115337,
+      "memory(GiB)": 123.79,
+      "step": 61860,
+      "train_speed(iter/s)": 1.316384
+    },
+    {
+      "acc": 0.67966785,
+      "epoch": 1.569381024860477,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.2147671011980816e-06,
+      "loss": 1.43618841,
+      "memory(GiB)": 123.79,
+      "step": 61865,
+      "train_speed(iter/s)": 1.316396
+    },
+    {
+      "acc": 0.69251494,
+      "epoch": 1.5695078640284121,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.2140820547295256e-06,
+      "loss": 1.35303593,
+      "memory(GiB)": 123.79,
+      "step": 61870,
+      "train_speed(iter/s)": 1.316407
+    },
+    {
+      "acc": 0.68055568,
+      "epoch": 1.5696347031963471,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2133971747825435e-06,
+      "loss": 1.3345973,
+      "memory(GiB)": 123.79,
+      "step": 61875,
+      "train_speed(iter/s)": 1.316419
+    },
+    {
+      "acc": 0.69084549,
+      "epoch": 1.5697615423642821,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.2127124613872603e-06,
+      "loss": 1.40256319,
+      "memory(GiB)": 123.79,
+      "step": 61880,
+      "train_speed(iter/s)": 1.31643
+    },
+    {
+      "acc": 0.6896277,
+      "epoch": 1.569888381532217,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.2120279145737918e-06,
+      "loss": 1.39016399,
+      "memory(GiB)": 123.79,
+      "step": 61885,
+      "train_speed(iter/s)": 1.316441
+    },
+    {
+      "acc": 0.69884634,
+      "epoch": 1.570015220700152,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.2113435343722474e-06,
+      "loss": 1.31149139,
+      "memory(GiB)": 123.79,
+      "step": 61890,
+      "train_speed(iter/s)": 1.316453
+    },
+    {
+      "acc": 0.68534636,
+      "epoch": 1.5701420598680873,
+      "grad_norm": 3.0,
+      "learning_rate": 1.21065932081273e-06,
+      "loss": 1.38768959,
+      "memory(GiB)": 123.79,
+      "step": 61895,
+      "train_speed(iter/s)": 1.316465
+    },
+    {
+      "acc": 0.68945746,
+      "epoch": 1.5702688990360223,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2099752739253334e-06,
+      "loss": 1.40294361,
+      "memory(GiB)": 123.79,
+      "step": 61900,
+      "train_speed(iter/s)": 1.316476
+    },
+    {
+      "acc": 0.6968883,
+      "epoch": 1.5703957382039575,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.209291393740144e-06,
+      "loss": 1.37745762,
+      "memory(GiB)": 123.79,
+      "step": 61905,
+      "train_speed(iter/s)": 1.316488
+    },
+    {
+      "acc": 0.67653093,
+      "epoch": 1.5705225773718925,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.2086076802872472e-06,
+      "loss": 1.4185195,
+      "memory(GiB)": 123.79,
+      "step": 61910,
+      "train_speed(iter/s)": 1.316499
+    },
+    {
+      "acc": 0.67847643,
+      "epoch": 1.5706494165398275,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.2079241335967096e-06,
+      "loss": 1.44789524,
+      "memory(GiB)": 123.79,
+      "step": 61915,
+      "train_speed(iter/s)": 1.316511
+    },
+    {
+      "acc": 0.67455535,
+      "epoch": 1.5707762557077625,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.207240753698599e-06,
+      "loss": 1.44528046,
+      "memory(GiB)": 123.79,
+      "step": 61920,
+      "train_speed(iter/s)": 1.316523
+    },
+    {
+      "acc": 0.70739989,
+      "epoch": 1.5709030948756975,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.2065575406229723e-06,
+      "loss": 1.25943413,
+      "memory(GiB)": 123.79,
+      "step": 61925,
+      "train_speed(iter/s)": 1.316534
+    },
+    {
+      "acc": 0.70419283,
+      "epoch": 1.5710299340436327,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.2058744943998847e-06,
+      "loss": 1.35051003,
+      "memory(GiB)": 123.79,
+      "step": 61930,
+      "train_speed(iter/s)": 1.316546
+    },
+    {
+      "acc": 0.6886426,
+      "epoch": 1.5711567732115679,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.2051916150593746e-06,
+      "loss": 1.38735847,
+      "memory(GiB)": 123.79,
+      "step": 61935,
+      "train_speed(iter/s)": 1.316558
+    },
+    {
+      "acc": 0.67733622,
+      "epoch": 1.5712836123795029,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.2045089026314783e-06,
+      "loss": 1.40344114,
+      "memory(GiB)": 123.79,
+      "step": 61940,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.68057799,
+      "epoch": 1.5714104515474379,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.2038263571462278e-06,
+      "loss": 1.38813667,
+      "memory(GiB)": 123.79,
+      "step": 61945,
+      "train_speed(iter/s)": 1.31658
+    },
+    {
+      "acc": 0.68716421,
+      "epoch": 1.5715372907153728,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.203143978633644e-06,
+      "loss": 1.39982281,
+      "memory(GiB)": 123.79,
+      "step": 61950,
+      "train_speed(iter/s)": 1.316592
+    },
+    {
+      "acc": 0.68061781,
+      "epoch": 1.5716641298833078,
+      "grad_norm": 3.375,
+      "learning_rate": 1.2024617671237388e-06,
+      "loss": 1.34242067,
+      "memory(GiB)": 123.79,
+      "step": 61955,
+      "train_speed(iter/s)": 1.316604
+    },
+    {
+      "acc": 0.67988949,
+      "epoch": 1.571790969051243,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.2017797226465178e-06,
+      "loss": 1.41022224,
+      "memory(GiB)": 123.79,
+      "step": 61960,
+      "train_speed(iter/s)": 1.316615
+    },
+    {
+      "acc": 0.66930819,
+      "epoch": 1.571917808219178,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2010978452319843e-06,
+      "loss": 1.41858444,
+      "memory(GiB)": 123.79,
+      "step": 61965,
+      "train_speed(iter/s)": 1.316627
+    },
+    {
+      "acc": 0.68158588,
+      "epoch": 1.5720446473871132,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.2004161349101295e-06,
+      "loss": 1.38036022,
+      "memory(GiB)": 123.79,
+      "step": 61970,
+      "train_speed(iter/s)": 1.316639
+    },
+    {
+      "acc": 0.69626064,
+      "epoch": 1.5721714865550482,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.1997345917109348e-06,
+      "loss": 1.36785345,
+      "memory(GiB)": 123.79,
+      "step": 61975,
+      "train_speed(iter/s)": 1.31665
+    },
+    {
+      "acc": 0.68995399,
+      "epoch": 1.5722983257229832,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.1990532156643808e-06,
+      "loss": 1.42261267,
+      "memory(GiB)": 123.79,
+      "step": 61980,
+      "train_speed(iter/s)": 1.316662
+    },
+    {
+      "acc": 0.68698101,
+      "epoch": 1.5724251648909182,
+      "grad_norm": 3.25,
+      "learning_rate": 1.198372006800436e-06,
+      "loss": 1.39728584,
+      "memory(GiB)": 123.79,
+      "step": 61985,
+      "train_speed(iter/s)": 1.316674
+    },
+    {
+      "acc": 0.67887945,
+      "epoch": 1.5725520040588534,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.1976909651490637e-06,
+      "loss": 1.39321747,
+      "memory(GiB)": 123.79,
+      "step": 61990,
+      "train_speed(iter/s)": 1.316686
+    },
+    {
+      "acc": 0.69777288,
+      "epoch": 1.5726788432267884,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.1970100907402188e-06,
+      "loss": 1.35838566,
+      "memory(GiB)": 123.79,
+      "step": 61995,
+      "train_speed(iter/s)": 1.316698
+    },
+    {
+      "acc": 0.698316,
+      "epoch": 1.5728056823947236,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.1963293836038492e-06,
+      "loss": 1.39908829,
+      "memory(GiB)": 123.79,
+      "step": 62000,
+      "train_speed(iter/s)": 1.31671
+    },
+    {
+      "epoch": 1.5728056823947236,
+      "eval_acc": 0.6751004997316526,
+      "eval_loss": 1.3563133478164673,
+      "eval_runtime": 69.7623,
+      "eval_samples_per_second": 91.31,
+      "eval_steps_per_second": 22.835,
+      "step": 62000
+    },
+    {
+      "acc": 0.68593898,
+      "epoch": 1.5729325215626586,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.195648843769896e-06,
+      "loss": 1.30189915,
+      "memory(GiB)": 123.79,
+      "step": 62005,
+      "train_speed(iter/s)": 1.314425
+    },
+    {
+      "acc": 0.68845072,
+      "epoch": 1.5730593607305936,
+      "grad_norm": 3.75,
+      "learning_rate": 1.1949684712682912e-06,
+      "loss": 1.43780661,
+      "memory(GiB)": 123.79,
+      "step": 62010,
+      "train_speed(iter/s)": 1.314437
+    },
+    {
+      "acc": 0.69117804,
+      "epoch": 1.5731861998985286,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.1942882661289618e-06,
+      "loss": 1.34747334,
+      "memory(GiB)": 123.79,
+      "step": 62015,
+      "train_speed(iter/s)": 1.314449
+    },
+    {
+      "acc": 0.69438252,
+      "epoch": 1.5733130390664636,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.1936082283818252e-06,
+      "loss": 1.367449,
+      "memory(GiB)": 123.79,
+      "step": 62020,
+      "train_speed(iter/s)": 1.31446
+    },
+    {
+      "acc": 0.70801945,
+      "epoch": 1.5734398782343988,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.1929283580567936e-06,
+      "loss": 1.31774588,
+      "memory(GiB)": 123.79,
+      "step": 62025,
+      "train_speed(iter/s)": 1.314472
+    },
+    {
+      "acc": 0.70197735,
+      "epoch": 1.573566717402334,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.1922486551837697e-06,
+      "loss": 1.37631464,
+      "memory(GiB)": 123.79,
+      "step": 62030,
+      "train_speed(iter/s)": 1.314484
+    },
+    {
+      "acc": 0.68783383,
+      "epoch": 1.573693556570269,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.1915691197926505e-06,
+      "loss": 1.45134087,
+      "memory(GiB)": 123.79,
+      "step": 62035,
+      "train_speed(iter/s)": 1.314495
+    },
+    {
+      "acc": 0.68672037,
+      "epoch": 1.573820395738204,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.1908897519133244e-06,
+      "loss": 1.36705341,
+      "memory(GiB)": 123.79,
+      "step": 62040,
+      "train_speed(iter/s)": 1.314506
+    },
+    {
+      "acc": 0.69497929,
+      "epoch": 1.573947234906139,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.190210551575674e-06,
+      "loss": 1.36550446,
+      "memory(GiB)": 123.79,
+      "step": 62045,
+      "train_speed(iter/s)": 1.314517
+    },
+    {
+      "acc": 0.69410014,
+      "epoch": 1.574074074074074,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.189531518809573e-06,
+      "loss": 1.31002655,
+      "memory(GiB)": 123.79,
+      "step": 62050,
+      "train_speed(iter/s)": 1.314528
+    },
+    {
+      "acc": 0.66912956,
+      "epoch": 1.5742009132420092,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.188852653644888e-06,
+      "loss": 1.51076355,
+      "memory(GiB)": 123.79,
+      "step": 62055,
+      "train_speed(iter/s)": 1.31454
+    },
+    {
+      "acc": 0.68529129,
+      "epoch": 1.5743277524099442,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.1881739561114792e-06,
+      "loss": 1.34807768,
+      "memory(GiB)": 123.79,
+      "step": 62060,
+      "train_speed(iter/s)": 1.314551
+    },
+    {
+      "acc": 0.68868604,
+      "epoch": 1.5744545915778794,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.1874954262391968e-06,
+      "loss": 1.40896873,
+      "memory(GiB)": 123.79,
+      "step": 62065,
+      "train_speed(iter/s)": 1.314562
+    },
+    {
+      "acc": 0.68053474,
+      "epoch": 1.5745814307458144,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.1868170640578901e-06,
+      "loss": 1.46921358,
+      "memory(GiB)": 123.79,
+      "step": 62070,
+      "train_speed(iter/s)": 1.314573
+    },
+    {
+      "acc": 0.68990312,
+      "epoch": 1.5747082699137493,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.1861388695973918e-06,
+      "loss": 1.34958363,
+      "memory(GiB)": 123.79,
+      "step": 62075,
+      "train_speed(iter/s)": 1.314584
+    },
+    {
+      "acc": 0.67478867,
+      "epoch": 1.5748351090816843,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1854608428875332e-06,
+      "loss": 1.41727438,
+      "memory(GiB)": 123.79,
+      "step": 62080,
+      "train_speed(iter/s)": 1.314595
+    },
+    {
+      "acc": 0.69362121,
+      "epoch": 1.5749619482496193,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.1847829839581377e-06,
+      "loss": 1.38673658,
+      "memory(GiB)": 123.79,
+      "step": 62085,
+      "train_speed(iter/s)": 1.314607
+    },
+    {
+      "acc": 0.69425678,
+      "epoch": 1.5750887874175545,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.1841052928390223e-06,
+      "loss": 1.33953333,
+      "memory(GiB)": 123.79,
+      "step": 62090,
+      "train_speed(iter/s)": 1.314619
+    },
+    {
+      "acc": 0.69345231,
+      "epoch": 1.5752156265854897,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.183427769559991e-06,
+      "loss": 1.32655897,
+      "memory(GiB)": 123.79,
+      "step": 62095,
+      "train_speed(iter/s)": 1.31463
+    },
+    {
+      "acc": 0.68646555,
+      "epoch": 1.5753424657534247,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.1827504141508456e-06,
+      "loss": 1.34461813,
+      "memory(GiB)": 123.79,
+      "step": 62100,
+      "train_speed(iter/s)": 1.314642
+    },
+    {
+      "acc": 0.67735171,
+      "epoch": 1.5754693049213597,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.1820732266413803e-06,
+      "loss": 1.43803396,
+      "memory(GiB)": 123.79,
+      "step": 62105,
+      "train_speed(iter/s)": 1.314653
+    },
+    {
+      "acc": 0.69771109,
+      "epoch": 1.5755961440892947,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.181396207061382e-06,
+      "loss": 1.37127562,
+      "memory(GiB)": 123.79,
+      "step": 62110,
+      "train_speed(iter/s)": 1.314664
+    },
+    {
+      "acc": 0.69467926,
+      "epoch": 1.5757229832572297,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1807193554406248e-06,
+      "loss": 1.3841197,
+      "memory(GiB)": 123.79,
+      "step": 62115,
+      "train_speed(iter/s)": 1.314675
+    },
+    {
+      "acc": 0.67932897,
+      "epoch": 1.575849822425165,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.1800426718088837e-06,
+      "loss": 1.42656822,
+      "memory(GiB)": 123.79,
+      "step": 62120,
+      "train_speed(iter/s)": 1.314687
+    },
+    {
+      "acc": 0.69063568,
+      "epoch": 1.5759766615931,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.1793661561959201e-06,
+      "loss": 1.41093063,
+      "memory(GiB)": 123.79,
+      "step": 62125,
+      "train_speed(iter/s)": 1.314699
+    },
+    {
+      "acc": 0.69234076,
+      "epoch": 1.5761035007610351,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.178689808631493e-06,
+      "loss": 1.32720995,
+      "memory(GiB)": 123.79,
+      "step": 62130,
+      "train_speed(iter/s)": 1.314711
+    },
+    {
+      "acc": 0.69378719,
+      "epoch": 1.57623033992897,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.178013629145346e-06,
+      "loss": 1.31017036,
+      "memory(GiB)": 123.79,
+      "step": 62135,
+      "train_speed(iter/s)": 1.314723
+    },
+    {
+      "acc": 0.67497058,
+      "epoch": 1.576357179096905,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.1773376177672246e-06,
+      "loss": 1.41007671,
+      "memory(GiB)": 123.79,
+      "step": 62140,
+      "train_speed(iter/s)": 1.314734
+    },
+    {
+      "acc": 0.6996047,
+      "epoch": 1.57648401826484,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.176661774526862e-06,
+      "loss": 1.35107899,
+      "memory(GiB)": 123.79,
+      "step": 62145,
+      "train_speed(iter/s)": 1.314746
+    },
+    {
+      "acc": 0.67825155,
+      "epoch": 1.5766108574327753,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1759860994539846e-06,
+      "loss": 1.40731478,
+      "memory(GiB)": 123.79,
+      "step": 62150,
+      "train_speed(iter/s)": 1.314759
+    },
+    {
+      "acc": 0.70584478,
+      "epoch": 1.5767376966007103,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.1753105925783114e-06,
+      "loss": 1.37982054,
+      "memory(GiB)": 123.79,
+      "step": 62155,
+      "train_speed(iter/s)": 1.314771
+    },
+    {
+      "acc": 0.68133125,
+      "epoch": 1.5768645357686455,
+      "grad_norm": 2.875,
+      "learning_rate": 1.174635253929554e-06,
+      "loss": 1.41840973,
+      "memory(GiB)": 123.79,
+      "step": 62160,
+      "train_speed(iter/s)": 1.314783
+    },
+    {
+      "acc": 0.68146935,
+      "epoch": 1.5769913749365805,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.1739600835374177e-06,
+      "loss": 1.39988251,
+      "memory(GiB)": 123.79,
+      "step": 62165,
+      "train_speed(iter/s)": 1.314795
+    },
+    {
+      "acc": 0.69312963,
+      "epoch": 1.5771182141045155,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.173285081431599e-06,
+      "loss": 1.34448395,
+      "memory(GiB)": 123.79,
+      "step": 62170,
+      "train_speed(iter/s)": 1.314807
+    },
+    {
+      "acc": 0.68303294,
+      "epoch": 1.5772450532724505,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.1726102476417871e-06,
+      "loss": 1.40340776,
+      "memory(GiB)": 123.79,
+      "step": 62175,
+      "train_speed(iter/s)": 1.314817
+    },
+    {
+      "acc": 0.69445314,
+      "epoch": 1.5773718924403854,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.1719355821976647e-06,
+      "loss": 1.37126408,
+      "memory(GiB)": 123.79,
+      "step": 62180,
+      "train_speed(iter/s)": 1.314828
+    },
+    {
+      "acc": 0.67870598,
+      "epoch": 1.5774987316083207,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.1712610851289069e-06,
+      "loss": 1.45320606,
+      "memory(GiB)": 123.79,
+      "step": 62185,
+      "train_speed(iter/s)": 1.31484
+    },
+    {
+      "acc": 0.67507992,
+      "epoch": 1.5776255707762559,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.1705867564651802e-06,
+      "loss": 1.41309023,
+      "memory(GiB)": 123.79,
+      "step": 62190,
+      "train_speed(iter/s)": 1.314852
+    },
+    {
+      "acc": 0.68486972,
+      "epoch": 1.5777524099441909,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.1699125962361451e-06,
+      "loss": 1.45169086,
+      "memory(GiB)": 123.79,
+      "step": 62195,
+      "train_speed(iter/s)": 1.314864
+    },
+    {
+      "acc": 0.70151348,
+      "epoch": 1.5778792491121258,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.1692386044714543e-06,
+      "loss": 1.3086462,
+      "memory(GiB)": 123.79,
+      "step": 62200,
+      "train_speed(iter/s)": 1.314876
+    },
+    {
+      "acc": 0.66801476,
+      "epoch": 1.5780060882800608,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.1685647812007512e-06,
+      "loss": 1.45443916,
+      "memory(GiB)": 123.79,
+      "step": 62205,
+      "train_speed(iter/s)": 1.314888
+    },
+    {
+      "acc": 0.67790279,
+      "epoch": 1.5781329274479958,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.167891126453678e-06,
+      "loss": 1.37209396,
+      "memory(GiB)": 123.79,
+      "step": 62210,
+      "train_speed(iter/s)": 1.314897
+    },
+    {
+      "acc": 0.68012934,
+      "epoch": 1.578259766615931,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.1672176402598607e-06,
+      "loss": 1.42654839,
+      "memory(GiB)": 123.79,
+      "step": 62215,
+      "train_speed(iter/s)": 1.314909
+    },
+    {
+      "acc": 0.69011784,
+      "epoch": 1.578386605783866,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.1665443226489236e-06,
+      "loss": 1.30818672,
+      "memory(GiB)": 123.79,
+      "step": 62220,
+      "train_speed(iter/s)": 1.314921
+    },
+    {
+      "acc": 0.68518162,
+      "epoch": 1.5785134449518012,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.1658711736504814e-06,
+      "loss": 1.35462284,
+      "memory(GiB)": 123.79,
+      "step": 62225,
+      "train_speed(iter/s)": 1.314932
+    },
+    {
+      "acc": 0.689849,
+      "epoch": 1.5786402841197362,
+      "grad_norm": 2.875,
+      "learning_rate": 1.165198193294146e-06,
+      "loss": 1.39044733,
+      "memory(GiB)": 123.79,
+      "step": 62230,
+      "train_speed(iter/s)": 1.314944
+    },
+    {
+      "acc": 0.69569302,
+      "epoch": 1.5787671232876712,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.1645253816095131e-06,
+      "loss": 1.32079926,
+      "memory(GiB)": 123.79,
+      "step": 62235,
+      "train_speed(iter/s)": 1.314951
+    },
+    {
+      "acc": 0.69294314,
+      "epoch": 1.5788939624556062,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.1638527386261772e-06,
+      "loss": 1.43309937,
+      "memory(GiB)": 123.79,
+      "step": 62240,
+      "train_speed(iter/s)": 1.314963
+    },
+    {
+      "acc": 0.69151344,
+      "epoch": 1.5790208016235412,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.163180264373726e-06,
+      "loss": 1.33571911,
+      "memory(GiB)": 123.79,
+      "step": 62245,
+      "train_speed(iter/s)": 1.314975
+    },
+    {
+      "acc": 0.68730235,
+      "epoch": 1.5791476407914764,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.162507958881739e-06,
+      "loss": 1.42435665,
+      "memory(GiB)": 123.79,
+      "step": 62250,
+      "train_speed(iter/s)": 1.314987
+    },
+    {
+      "acc": 0.71120286,
+      "epoch": 1.5792744799594116,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.161835822179782e-06,
+      "loss": 1.25417519,
+      "memory(GiB)": 123.79,
+      "step": 62255,
+      "train_speed(iter/s)": 1.314998
+    },
+    {
+      "acc": 0.67513208,
+      "epoch": 1.5794013191273466,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.1611638542974229e-06,
+      "loss": 1.44387789,
+      "memory(GiB)": 123.79,
+      "step": 62260,
+      "train_speed(iter/s)": 1.315011
+    },
+    {
+      "acc": 0.69706678,
+      "epoch": 1.5795281582952816,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.160492055264217e-06,
+      "loss": 1.41253605,
+      "memory(GiB)": 123.79,
+      "step": 62265,
+      "train_speed(iter/s)": 1.315023
+    },
+    {
+      "acc": 0.69151196,
+      "epoch": 1.5796549974632166,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.159820425109714e-06,
+      "loss": 1.36376228,
+      "memory(GiB)": 123.79,
+      "step": 62270,
+      "train_speed(iter/s)": 1.315035
+    },
+    {
+      "acc": 0.68008652,
+      "epoch": 1.5797818366311516,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1591489638634513e-06,
+      "loss": 1.41490917,
+      "memory(GiB)": 123.79,
+      "step": 62275,
+      "train_speed(iter/s)": 1.315046
+    },
+    {
+      "acc": 0.69721885,
+      "epoch": 1.5799086757990868,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.1584776715549662e-06,
+      "loss": 1.39947138,
+      "memory(GiB)": 123.79,
+      "step": 62280,
+      "train_speed(iter/s)": 1.315058
+    },
+    {
+      "acc": 0.68870168,
+      "epoch": 1.5800355149670218,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.1578065482137845e-06,
+      "loss": 1.36329947,
+      "memory(GiB)": 123.79,
+      "step": 62285,
+      "train_speed(iter/s)": 1.31507
+    },
+    {
+      "acc": 0.66747179,
+      "epoch": 1.580162354134957,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.157135593869425e-06,
+      "loss": 1.47235394,
+      "memory(GiB)": 123.79,
+      "step": 62290,
+      "train_speed(iter/s)": 1.315081
+    },
+    {
+      "acc": 0.68206611,
+      "epoch": 1.580289193302892,
+      "grad_norm": 2.875,
+      "learning_rate": 1.156464808551399e-06,
+      "loss": 1.39836493,
+      "memory(GiB)": 123.79,
+      "step": 62295,
+      "train_speed(iter/s)": 1.315093
+    },
+    {
+      "acc": 0.68980474,
+      "epoch": 1.580416032470827,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.1557941922892113e-06,
+      "loss": 1.38992004,
+      "memory(GiB)": 123.79,
+      "step": 62300,
+      "train_speed(iter/s)": 1.315105
+    },
+    {
+      "acc": 0.689991,
+      "epoch": 1.580542871638762,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.155123745112358e-06,
+      "loss": 1.32980661,
+      "memory(GiB)": 123.79,
+      "step": 62305,
+      "train_speed(iter/s)": 1.315117
+    },
+    {
+      "acc": 0.71645923,
+      "epoch": 1.5806697108066972,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.1544534670503282e-06,
+      "loss": 1.27398033,
+      "memory(GiB)": 123.79,
+      "step": 62310,
+      "train_speed(iter/s)": 1.315127
+    },
+    {
+      "acc": 0.68340631,
+      "epoch": 1.5807965499746321,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.1537833581326048e-06,
+      "loss": 1.34704561,
+      "memory(GiB)": 123.79,
+      "step": 62315,
+      "train_speed(iter/s)": 1.315139
+    },
+    {
+      "acc": 0.68766832,
+      "epoch": 1.5809233891425674,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.1531134183886606e-06,
+      "loss": 1.37496424,
+      "memory(GiB)": 123.79,
+      "step": 62320,
+      "train_speed(iter/s)": 1.31515
+    },
+    {
+      "acc": 0.69341903,
+      "epoch": 1.5810502283105023,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.1524436478479638e-06,
+      "loss": 1.3877882,
+      "memory(GiB)": 123.79,
+      "step": 62325,
+      "train_speed(iter/s)": 1.315162
+    },
+    {
+      "acc": 0.68430386,
+      "epoch": 1.5811770674784373,
+      "grad_norm": 4.125,
+      "learning_rate": 1.1517740465399736e-06,
+      "loss": 1.40886316,
+      "memory(GiB)": 123.79,
+      "step": 62330,
+      "train_speed(iter/s)": 1.315174
+    },
+    {
+      "acc": 0.71264896,
+      "epoch": 1.5813039066463723,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.1511046144941417e-06,
+      "loss": 1.34481831,
+      "memory(GiB)": 123.79,
+      "step": 62335,
+      "train_speed(iter/s)": 1.315186
+    },
+    {
+      "acc": 0.69192038,
+      "epoch": 1.5814307458143073,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.150435351739913e-06,
+      "loss": 1.36615219,
+      "memory(GiB)": 123.79,
+      "step": 62340,
+      "train_speed(iter/s)": 1.315197
+    },
+    {
+      "acc": 0.69657488,
+      "epoch": 1.5815575849822425,
+      "grad_norm": 2.625,
+      "learning_rate": 1.1497662583067231e-06,
+      "loss": 1.33458004,
+      "memory(GiB)": 123.79,
+      "step": 62345,
+      "train_speed(iter/s)": 1.315208
+    },
+    {
+      "acc": 0.70212607,
+      "epoch": 1.5816844241501777,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.1490973342240063e-06,
+      "loss": 1.34034824,
+      "memory(GiB)": 123.79,
+      "step": 62350,
+      "train_speed(iter/s)": 1.315221
+    },
+    {
+      "acc": 0.67562456,
+      "epoch": 1.5818112633181127,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.1484285795211803e-06,
+      "loss": 1.45913429,
+      "memory(GiB)": 123.79,
+      "step": 62355,
+      "train_speed(iter/s)": 1.315232
+    },
+    {
+      "acc": 0.69651051,
+      "epoch": 1.5819381024860477,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.1477599942276613e-06,
+      "loss": 1.39455013,
+      "memory(GiB)": 123.79,
+      "step": 62360,
+      "train_speed(iter/s)": 1.315243
+    },
+    {
+      "acc": 0.68970704,
+      "epoch": 1.5820649416539827,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.1470915783728547e-06,
+      "loss": 1.33006096,
+      "memory(GiB)": 123.79,
+      "step": 62365,
+      "train_speed(iter/s)": 1.315255
+    },
+    {
+      "acc": 0.69218144,
+      "epoch": 1.5821917808219177,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1464233319861661e-06,
+      "loss": 1.35390224,
+      "memory(GiB)": 123.79,
+      "step": 62370,
+      "train_speed(iter/s)": 1.315266
+    },
+    {
+      "acc": 0.68600149,
+      "epoch": 1.582318619989853,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.145755255096983e-06,
+      "loss": 1.33250141,
+      "memory(GiB)": 123.79,
+      "step": 62375,
+      "train_speed(iter/s)": 1.315278
+    },
+    {
+      "acc": 0.67467146,
+      "epoch": 1.582445459157788,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.1450873477346901e-06,
+      "loss": 1.37951279,
+      "memory(GiB)": 123.79,
+      "step": 62380,
+      "train_speed(iter/s)": 1.315289
+    },
+    {
+      "acc": 0.69193039,
+      "epoch": 1.582572298325723,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1444196099286682e-06,
+      "loss": 1.34549713,
+      "memory(GiB)": 123.79,
+      "step": 62385,
+      "train_speed(iter/s)": 1.315301
+    },
+    {
+      "acc": 0.68031917,
+      "epoch": 1.582699137493658,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.143752041708287e-06,
+      "loss": 1.426929,
+      "memory(GiB)": 123.79,
+      "step": 62390,
+      "train_speed(iter/s)": 1.315312
+    },
+    {
+      "acc": 0.69282112,
+      "epoch": 1.582825976661593,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.1430846431029062e-06,
+      "loss": 1.35458775,
+      "memory(GiB)": 123.79,
+      "step": 62395,
+      "train_speed(iter/s)": 1.315323
+    },
+    {
+      "acc": 0.6780961,
+      "epoch": 1.582952815829528,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.1424174141418815e-06,
+      "loss": 1.47878036,
+      "memory(GiB)": 123.79,
+      "step": 62400,
+      "train_speed(iter/s)": 1.315334
+    },
+    {
+      "acc": 0.68696404,
+      "epoch": 1.583079654997463,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.1417503548545634e-06,
+      "loss": 1.38900681,
+      "memory(GiB)": 123.79,
+      "step": 62405,
+      "train_speed(iter/s)": 1.315346
+    },
+    {
+      "acc": 0.66838274,
+      "epoch": 1.5832064941653983,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.1410834652702918e-06,
+      "loss": 1.44508896,
+      "memory(GiB)": 123.79,
+      "step": 62410,
+      "train_speed(iter/s)": 1.315358
+    },
+    {
+      "acc": 0.69252405,
+      "epoch": 1.5833333333333335,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.1404167454183957e-06,
+      "loss": 1.35928917,
+      "memory(GiB)": 123.79,
+      "step": 62415,
+      "train_speed(iter/s)": 1.315369
+    },
+    {
+      "acc": 0.67409191,
+      "epoch": 1.5834601725012685,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.1397501953282042e-06,
+      "loss": 1.43169498,
+      "memory(GiB)": 123.79,
+      "step": 62420,
+      "train_speed(iter/s)": 1.315379
+    },
+    {
+      "acc": 0.67564745,
+      "epoch": 1.5835870116692035,
+      "grad_norm": 4.625,
+      "learning_rate": 1.1390838150290334e-06,
+      "loss": 1.40899305,
+      "memory(GiB)": 123.79,
+      "step": 62425,
+      "train_speed(iter/s)": 1.315391
+    },
+    {
+      "acc": 0.70047884,
+      "epoch": 1.5837138508371384,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.1384176045501944e-06,
+      "loss": 1.35657902,
+      "memory(GiB)": 123.79,
+      "step": 62430,
+      "train_speed(iter/s)": 1.315402
+    },
+    {
+      "acc": 0.67861872,
+      "epoch": 1.5838406900050734,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.1377515639209902e-06,
+      "loss": 1.39508762,
+      "memory(GiB)": 123.79,
+      "step": 62435,
+      "train_speed(iter/s)": 1.315413
+    },
+    {
+      "acc": 0.69425945,
+      "epoch": 1.5839675291730086,
+      "grad_norm": 4.0,
+      "learning_rate": 1.1370856931707159e-06,
+      "loss": 1.42933044,
+      "memory(GiB)": 123.79,
+      "step": 62440,
+      "train_speed(iter/s)": 1.315425
+    },
+    {
+      "acc": 0.68754272,
+      "epoch": 1.5840943683409436,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.1364199923286589e-06,
+      "loss": 1.38375912,
+      "memory(GiB)": 123.79,
+      "step": 62445,
+      "train_speed(iter/s)": 1.315436
+    },
+    {
+      "acc": 0.69928198,
+      "epoch": 1.5842212075088788,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.1357544614241012e-06,
+      "loss": 1.38822336,
+      "memory(GiB)": 123.79,
+      "step": 62450,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.6951086,
+      "epoch": 1.5843480466768138,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.135089100486314e-06,
+      "loss": 1.3663641,
+      "memory(GiB)": 123.79,
+      "step": 62455,
+      "train_speed(iter/s)": 1.31546
+    },
+    {
+      "acc": 0.69069519,
+      "epoch": 1.5844748858447488,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1344239095445642e-06,
+      "loss": 1.36037006,
+      "memory(GiB)": 123.79,
+      "step": 62460,
+      "train_speed(iter/s)": 1.315471
+    },
+    {
+      "acc": 0.69630632,
+      "epoch": 1.5846017250126838,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.1337588886281092e-06,
+      "loss": 1.30006809,
+      "memory(GiB)": 123.79,
+      "step": 62465,
+      "train_speed(iter/s)": 1.315482
+    },
+    {
+      "acc": 0.69015446,
+      "epoch": 1.584728564180619,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1330940377662002e-06,
+      "loss": 1.39260635,
+      "memory(GiB)": 123.79,
+      "step": 62470,
+      "train_speed(iter/s)": 1.315494
+    },
+    {
+      "acc": 0.68604507,
+      "epoch": 1.584855403348554,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.1324293569880795e-06,
+      "loss": 1.41350946,
+      "memory(GiB)": 123.79,
+      "step": 62475,
+      "train_speed(iter/s)": 1.315506
+    },
+    {
+      "acc": 0.68309331,
+      "epoch": 1.5849822425164892,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.1317648463229835e-06,
+      "loss": 1.35319605,
+      "memory(GiB)": 123.79,
+      "step": 62480,
+      "train_speed(iter/s)": 1.315517
+    },
+    {
+      "acc": 0.67717028,
+      "epoch": 1.5851090816844242,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.1311005058001396e-06,
+      "loss": 1.42114429,
+      "memory(GiB)": 123.79,
+      "step": 62485,
+      "train_speed(iter/s)": 1.315529
+    },
+    {
+      "acc": 0.68229718,
+      "epoch": 1.5852359208523592,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.130436335448769e-06,
+      "loss": 1.41376591,
+      "memory(GiB)": 123.79,
+      "step": 62490,
+      "train_speed(iter/s)": 1.31554
+    },
+    {
+      "acc": 0.69375305,
+      "epoch": 1.5853627600202942,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.129772335298085e-06,
+      "loss": 1.38877296,
+      "memory(GiB)": 123.79,
+      "step": 62495,
+      "train_speed(iter/s)": 1.315551
+    },
+    {
+      "acc": 0.68031425,
+      "epoch": 1.5854895991882292,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.1291085053772926e-06,
+      "loss": 1.43150702,
+      "memory(GiB)": 123.79,
+      "step": 62500,
+      "train_speed(iter/s)": 1.315564
+    },
+    {
+      "acc": 0.6863699,
+      "epoch": 1.5856164383561644,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.1284448457155893e-06,
+      "loss": 1.44015808,
+      "memory(GiB)": 123.79,
+      "step": 62505,
+      "train_speed(iter/s)": 1.315576
+    },
+    {
+      "acc": 0.68679132,
+      "epoch": 1.5857432775240996,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.1277813563421697e-06,
+      "loss": 1.36411171,
+      "memory(GiB)": 123.79,
+      "step": 62510,
+      "train_speed(iter/s)": 1.315587
+    },
+    {
+      "acc": 0.69395924,
+      "epoch": 1.5858701166920346,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.127118037286213e-06,
+      "loss": 1.35111074,
+      "memory(GiB)": 123.79,
+      "step": 62515,
+      "train_speed(iter/s)": 1.3156
+    },
+    {
+      "acc": 0.68433456,
+      "epoch": 1.5859969558599696,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1264548885768944e-06,
+      "loss": 1.46368542,
+      "memory(GiB)": 123.79,
+      "step": 62520,
+      "train_speed(iter/s)": 1.315612
+    },
+    {
+      "acc": 0.68501611,
+      "epoch": 1.5861237950279046,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.1257919102433856e-06,
+      "loss": 1.3995245,
+      "memory(GiB)": 123.79,
+      "step": 62525,
+      "train_speed(iter/s)": 1.315625
+    },
+    {
+      "acc": 0.68452749,
+      "epoch": 1.5862506341958396,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.125129102314847e-06,
+      "loss": 1.38435354,
+      "memory(GiB)": 123.79,
+      "step": 62530,
+      "train_speed(iter/s)": 1.315637
+    },
+    {
+      "acc": 0.67036858,
+      "epoch": 1.5863774733637748,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.124466464820429e-06,
+      "loss": 1.43446198,
+      "memory(GiB)": 123.79,
+      "step": 62535,
+      "train_speed(iter/s)": 1.315649
+    },
+    {
+      "acc": 0.68972445,
+      "epoch": 1.5865043125317098,
+      "grad_norm": 3.75,
+      "learning_rate": 1.123803997789278e-06,
+      "loss": 1.34509106,
+      "memory(GiB)": 123.79,
+      "step": 62540,
+      "train_speed(iter/s)": 1.315661
+    },
+    {
+      "acc": 0.68448391,
+      "epoch": 1.586631151699645,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1231417012505342e-06,
+      "loss": 1.41880054,
+      "memory(GiB)": 123.79,
+      "step": 62545,
+      "train_speed(iter/s)": 1.315673
+    },
+    {
+      "acc": 0.6877111,
+      "epoch": 1.58675799086758,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.1224795752333283e-06,
+      "loss": 1.39137363,
+      "memory(GiB)": 123.79,
+      "step": 62550,
+      "train_speed(iter/s)": 1.315685
+    },
+    {
+      "acc": 0.68503132,
+      "epoch": 1.586884830035515,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.1218176197667802e-06,
+      "loss": 1.44367695,
+      "memory(GiB)": 123.79,
+      "step": 62555,
+      "train_speed(iter/s)": 1.315697
+    },
+    {
+      "acc": 0.67875695,
+      "epoch": 1.58701166920345,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.1211558348800095e-06,
+      "loss": 1.39569654,
+      "memory(GiB)": 123.79,
+      "step": 62560,
+      "train_speed(iter/s)": 1.315709
+    },
+    {
+      "acc": 0.68856964,
+      "epoch": 1.587138508371385,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.120494220602123e-06,
+      "loss": 1.36336994,
+      "memory(GiB)": 123.79,
+      "step": 62565,
+      "train_speed(iter/s)": 1.315721
+    },
+    {
+      "acc": 0.70927343,
+      "epoch": 1.5872653475393201,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.1198327769622224e-06,
+      "loss": 1.3587574,
+      "memory(GiB)": 123.79,
+      "step": 62570,
+      "train_speed(iter/s)": 1.315735
+    },
+    {
+      "acc": 0.6889421,
+      "epoch": 1.5873921867072553,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.1191715039893975e-06,
+      "loss": 1.34782696,
+      "memory(GiB)": 123.79,
+      "step": 62575,
+      "train_speed(iter/s)": 1.315747
+    },
+    {
+      "acc": 0.68946042,
+      "epoch": 1.5875190258751903,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.1185104017127379e-06,
+      "loss": 1.50531492,
+      "memory(GiB)": 123.79,
+      "step": 62580,
+      "train_speed(iter/s)": 1.31576
+    },
+    {
+      "acc": 0.69387541,
+      "epoch": 1.5876458650431253,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.1178494701613202e-06,
+      "loss": 1.43622065,
+      "memory(GiB)": 123.79,
+      "step": 62585,
+      "train_speed(iter/s)": 1.315772
+    },
+    {
+      "acc": 0.6978169,
+      "epoch": 1.5877727042110603,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.1171887093642158e-06,
+      "loss": 1.34866848,
+      "memory(GiB)": 123.79,
+      "step": 62590,
+      "train_speed(iter/s)": 1.315783
+    },
+    {
+      "acc": 0.71039639,
+      "epoch": 1.5878995433789953,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.1165281193504873e-06,
+      "loss": 1.32605877,
+      "memory(GiB)": 123.79,
+      "step": 62595,
+      "train_speed(iter/s)": 1.315645
+    },
+    {
+      "acc": 0.68328714,
+      "epoch": 1.5880263825469305,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.1158677001491902e-06,
+      "loss": 1.38937492,
+      "memory(GiB)": 123.79,
+      "step": 62600,
+      "train_speed(iter/s)": 1.315658
+    },
+    {
+      "acc": 0.67470651,
+      "epoch": 1.5881532217148655,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.1152074517893735e-06,
+      "loss": 1.4965292,
+      "memory(GiB)": 123.79,
+      "step": 62605,
+      "train_speed(iter/s)": 1.31567
+    },
+    {
+      "acc": 0.67875943,
+      "epoch": 1.5882800608828007,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.1145473743000773e-06,
+      "loss": 1.4532074,
+      "memory(GiB)": 123.79,
+      "step": 62610,
+      "train_speed(iter/s)": 1.315682
+    },
+    {
+      "acc": 0.69513245,
+      "epoch": 1.5884069000507357,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.1138874677103345e-06,
+      "loss": 1.35410271,
+      "memory(GiB)": 123.79,
+      "step": 62615,
+      "train_speed(iter/s)": 1.315694
+    },
+    {
+      "acc": 0.69697952,
+      "epoch": 1.5885337392186707,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1132277320491713e-06,
+      "loss": 1.33949032,
+      "memory(GiB)": 123.79,
+      "step": 62620,
+      "train_speed(iter/s)": 1.315706
+    },
+    {
+      "acc": 0.68117356,
+      "epoch": 1.5886605783866057,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.1125681673456062e-06,
+      "loss": 1.37334366,
+      "memory(GiB)": 123.79,
+      "step": 62625,
+      "train_speed(iter/s)": 1.315716
+    },
+    {
+      "acc": 0.71002727,
+      "epoch": 1.588787417554541,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.1119087736286489e-06,
+      "loss": 1.32919388,
+      "memory(GiB)": 123.79,
+      "step": 62630,
+      "train_speed(iter/s)": 1.315727
+    },
+    {
+      "acc": 0.67079401,
+      "epoch": 1.5889142567224759,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.1112495509273025e-06,
+      "loss": 1.44136887,
+      "memory(GiB)": 123.79,
+      "step": 62635,
+      "train_speed(iter/s)": 1.315739
+    },
+    {
+      "acc": 0.6858418,
+      "epoch": 1.589041095890411,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.110590499270563e-06,
+      "loss": 1.37656765,
+      "memory(GiB)": 123.79,
+      "step": 62640,
+      "train_speed(iter/s)": 1.31575
+    },
+    {
+      "acc": 0.70084429,
+      "epoch": 1.589167935058346,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.109931618687417e-06,
+      "loss": 1.35344667,
+      "memory(GiB)": 123.79,
+      "step": 62645,
+      "train_speed(iter/s)": 1.315761
+    },
+    {
+      "acc": 0.67376385,
+      "epoch": 1.589294774226281,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.1092729092068495e-06,
+      "loss": 1.43702993,
+      "memory(GiB)": 123.79,
+      "step": 62650,
+      "train_speed(iter/s)": 1.315772
+    },
+    {
+      "acc": 0.68952203,
+      "epoch": 1.589421613394216,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.1086143708578285e-06,
+      "loss": 1.4427825,
+      "memory(GiB)": 123.79,
+      "step": 62655,
+      "train_speed(iter/s)": 1.315784
+    },
+    {
+      "acc": 0.68993859,
+      "epoch": 1.589548452562151,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.107956003669321e-06,
+      "loss": 1.3478673,
+      "memory(GiB)": 123.79,
+      "step": 62660,
+      "train_speed(iter/s)": 1.315796
+    },
+    {
+      "acc": 0.69399228,
+      "epoch": 1.5896752917300863,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.107297807670284e-06,
+      "loss": 1.38326283,
+      "memory(GiB)": 123.79,
+      "step": 62665,
+      "train_speed(iter/s)": 1.315807
+    },
+    {
+      "acc": 0.69628611,
+      "epoch": 1.5898021308980215,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.106639782889672e-06,
+      "loss": 1.38869553,
+      "memory(GiB)": 123.79,
+      "step": 62670,
+      "train_speed(iter/s)": 1.315819
+    },
+    {
+      "acc": 0.69398241,
+      "epoch": 1.5899289700659565,
+      "grad_norm": 3.125,
+      "learning_rate": 1.1059819293564233e-06,
+      "loss": 1.35163975,
+      "memory(GiB)": 123.79,
+      "step": 62675,
+      "train_speed(iter/s)": 1.31583
+    },
+    {
+      "acc": 0.68688154,
+      "epoch": 1.5900558092338914,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.105324247099474e-06,
+      "loss": 1.39908314,
+      "memory(GiB)": 123.79,
+      "step": 62680,
+      "train_speed(iter/s)": 1.315842
+    },
+    {
+      "acc": 0.68214903,
+      "epoch": 1.5901826484018264,
+      "grad_norm": 2.75,
+      "learning_rate": 1.1046667361477537e-06,
+      "loss": 1.32299261,
+      "memory(GiB)": 123.79,
+      "step": 62685,
+      "train_speed(iter/s)": 1.315853
+    },
+    {
+      "acc": 0.69857244,
+      "epoch": 1.5903094875697614,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.1040093965301835e-06,
+      "loss": 1.3757349,
+      "memory(GiB)": 123.79,
+      "step": 62690,
+      "train_speed(iter/s)": 1.315865
+    },
+    {
+      "acc": 0.68239374,
+      "epoch": 1.5904363267376966,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.1033522282756716e-06,
+      "loss": 1.39094915,
+      "memory(GiB)": 123.79,
+      "step": 62695,
+      "train_speed(iter/s)": 1.315877
+    },
+    {
+      "acc": 0.68553867,
+      "epoch": 1.5905631659056316,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.1026952314131268e-06,
+      "loss": 1.37664967,
+      "memory(GiB)": 123.79,
+      "step": 62700,
+      "train_speed(iter/s)": 1.315889
+    },
+    {
+      "acc": 0.70620012,
+      "epoch": 1.5906900050735668,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.1020384059714463e-06,
+      "loss": 1.32556391,
+      "memory(GiB)": 123.79,
+      "step": 62705,
+      "train_speed(iter/s)": 1.3159
+    },
+    {
+      "acc": 0.70364523,
+      "epoch": 1.5908168442415018,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.1013817519795211e-06,
+      "loss": 1.30131073,
+      "memory(GiB)": 123.79,
+      "step": 62710,
+      "train_speed(iter/s)": 1.315912
+    },
+    {
+      "acc": 0.68971825,
+      "epoch": 1.5909436834094368,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.1007252694662302e-06,
+      "loss": 1.39856968,
+      "memory(GiB)": 123.79,
+      "step": 62715,
+      "train_speed(iter/s)": 1.315924
+    },
+    {
+      "acc": 0.7031702,
+      "epoch": 1.5910705225773718,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.1000689584604519e-06,
+      "loss": 1.29111786,
+      "memory(GiB)": 123.79,
+      "step": 62720,
+      "train_speed(iter/s)": 1.315937
+    },
+    {
+      "acc": 0.67992282,
+      "epoch": 1.5911973617453068,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.099412818991053e-06,
+      "loss": 1.41511812,
+      "memory(GiB)": 123.79,
+      "step": 62725,
+      "train_speed(iter/s)": 1.315949
+    },
+    {
+      "acc": 0.69237313,
+      "epoch": 1.591324200913242,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.098756851086893e-06,
+      "loss": 1.37045755,
+      "memory(GiB)": 123.79,
+      "step": 62730,
+      "train_speed(iter/s)": 1.315961
+    },
+    {
+      "acc": 0.67574091,
+      "epoch": 1.5914510400811772,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.0981010547768244e-06,
+      "loss": 1.44686813,
+      "memory(GiB)": 123.79,
+      "step": 62735,
+      "train_speed(iter/s)": 1.315973
+    },
+    {
+      "acc": 0.68185978,
+      "epoch": 1.5915778792491122,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.0974454300896924e-06,
+      "loss": 1.41785192,
+      "memory(GiB)": 123.79,
+      "step": 62740,
+      "train_speed(iter/s)": 1.315985
+    },
+    {
+      "acc": 0.68287764,
+      "epoch": 1.5917047184170472,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.0967899770543344e-06,
+      "loss": 1.45210772,
+      "memory(GiB)": 123.79,
+      "step": 62745,
+      "train_speed(iter/s)": 1.315996
+    },
+    {
+      "acc": 0.68870239,
+      "epoch": 1.5918315575849822,
+      "grad_norm": 3.375,
+      "learning_rate": 1.0961346956995795e-06,
+      "loss": 1.40947809,
+      "memory(GiB)": 123.79,
+      "step": 62750,
+      "train_speed(iter/s)": 1.316008
+    },
+    {
+      "acc": 0.69743814,
+      "epoch": 1.5919583967529172,
+      "grad_norm": 3.5625,
+      "learning_rate": 1.0954795860542495e-06,
+      "loss": 1.32984676,
+      "memory(GiB)": 123.79,
+      "step": 62755,
+      "train_speed(iter/s)": 1.31602
+    },
+    {
+      "acc": 0.67741523,
+      "epoch": 1.5920852359208524,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.0948246481471603e-06,
+      "loss": 1.44198246,
+      "memory(GiB)": 123.79,
+      "step": 62760,
+      "train_speed(iter/s)": 1.316033
+    },
+    {
+      "acc": 0.69214406,
+      "epoch": 1.5922120750887874,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.0941698820071183e-06,
+      "loss": 1.33252172,
+      "memory(GiB)": 123.79,
+      "step": 62765,
+      "train_speed(iter/s)": 1.316045
+    },
+    {
+      "acc": 0.68870821,
+      "epoch": 1.5923389142567226,
+      "grad_norm": 3.0,
+      "learning_rate": 1.0935152876629234e-06,
+      "loss": 1.27925167,
+      "memory(GiB)": 123.79,
+      "step": 62770,
+      "train_speed(iter/s)": 1.316057
+    },
+    {
+      "acc": 0.69505472,
+      "epoch": 1.5924657534246576,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0928608651433675e-06,
+      "loss": 1.34312363,
+      "memory(GiB)": 123.79,
+      "step": 62775,
+      "train_speed(iter/s)": 1.316069
+    },
+    {
+      "acc": 0.67982378,
+      "epoch": 1.5925925925925926,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.0922066144772342e-06,
+      "loss": 1.43172798,
+      "memory(GiB)": 123.79,
+      "step": 62780,
+      "train_speed(iter/s)": 1.316082
+    },
+    {
+      "acc": 0.68853664,
+      "epoch": 1.5927194317605275,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.0915525356933004e-06,
+      "loss": 1.45793724,
+      "memory(GiB)": 123.79,
+      "step": 62785,
+      "train_speed(iter/s)": 1.316094
+    },
+    {
+      "acc": 0.70019226,
+      "epoch": 1.5928462709284628,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.0908986288203382e-06,
+      "loss": 1.370327,
+      "memory(GiB)": 123.79,
+      "step": 62790,
+      "train_speed(iter/s)": 1.316106
+    },
+    {
+      "acc": 0.699965,
+      "epoch": 1.5929731100963977,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.0902448938871064e-06,
+      "loss": 1.32768211,
+      "memory(GiB)": 123.79,
+      "step": 62795,
+      "train_speed(iter/s)": 1.316118
+    },
+    {
+      "acc": 0.68997869,
+      "epoch": 1.593099949264333,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.0895913309223594e-06,
+      "loss": 1.359624,
+      "memory(GiB)": 123.79,
+      "step": 62800,
+      "train_speed(iter/s)": 1.316131
+    },
+    {
+      "acc": 0.6979609,
+      "epoch": 1.593226788432268,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.0889379399548432e-06,
+      "loss": 1.34988861,
+      "memory(GiB)": 123.79,
+      "step": 62805,
+      "train_speed(iter/s)": 1.316143
+    },
+    {
+      "acc": 0.67898383,
+      "epoch": 1.593353627600203,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.0882847210133007e-06,
+      "loss": 1.40707541,
+      "memory(GiB)": 123.79,
+      "step": 62810,
+      "train_speed(iter/s)": 1.316154
+    },
+    {
+      "acc": 0.6874702,
+      "epoch": 1.593480466768138,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0876316741264598e-06,
+      "loss": 1.38645315,
+      "memory(GiB)": 123.79,
+      "step": 62815,
+      "train_speed(iter/s)": 1.316166
+    },
+    {
+      "acc": 0.68639231,
+      "epoch": 1.593607305936073,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0869787993230435e-06,
+      "loss": 1.45328112,
+      "memory(GiB)": 123.79,
+      "step": 62820,
+      "train_speed(iter/s)": 1.316178
+    },
+    {
+      "acc": 0.70075607,
+      "epoch": 1.5937341451040081,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.0863260966317713e-06,
+      "loss": 1.35246105,
+      "memory(GiB)": 123.79,
+      "step": 62825,
+      "train_speed(iter/s)": 1.316191
+    },
+    {
+      "acc": 0.69373732,
+      "epoch": 1.5938609842719433,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.0856735660813523e-06,
+      "loss": 1.33758659,
+      "memory(GiB)": 123.79,
+      "step": 62830,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.6849175,
+      "epoch": 1.5939878234398783,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.0850212077004845e-06,
+      "loss": 1.40292425,
+      "memory(GiB)": 123.79,
+      "step": 62835,
+      "train_speed(iter/s)": 1.316215
+    },
+    {
+      "acc": 0.69535184,
+      "epoch": 1.5941146626078133,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.084369021517862e-06,
+      "loss": 1.37474232,
+      "memory(GiB)": 123.79,
+      "step": 62840,
+      "train_speed(iter/s)": 1.316228
+    },
+    {
+      "acc": 0.70150223,
+      "epoch": 1.5942415017757483,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.0837170075621733e-06,
+      "loss": 1.37147141,
+      "memory(GiB)": 123.79,
+      "step": 62845,
+      "train_speed(iter/s)": 1.316239
+    },
+    {
+      "acc": 0.69417253,
+      "epoch": 1.5943683409436833,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.0830651658620966e-06,
+      "loss": 1.36756506,
+      "memory(GiB)": 123.79,
+      "step": 62850,
+      "train_speed(iter/s)": 1.316252
+    },
+    {
+      "acc": 0.7118824,
+      "epoch": 1.5944951801116185,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.0824134964462996e-06,
+      "loss": 1.28820019,
+      "memory(GiB)": 123.79,
+      "step": 62855,
+      "train_speed(iter/s)": 1.316263
+    },
+    {
+      "acc": 0.69134636,
+      "epoch": 1.5946220192795535,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.0817619993434486e-06,
+      "loss": 1.34760571,
+      "memory(GiB)": 123.79,
+      "step": 62860,
+      "train_speed(iter/s)": 1.316276
+    },
+    {
+      "acc": 0.67452221,
+      "epoch": 1.5947488584474887,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.0811106745821987e-06,
+      "loss": 1.45005379,
+      "memory(GiB)": 123.79,
+      "step": 62865,
+      "train_speed(iter/s)": 1.316289
+    },
+    {
+      "acc": 0.68806686,
+      "epoch": 1.5948756976154237,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.0804595221911978e-06,
+      "loss": 1.39351463,
+      "memory(GiB)": 123.79,
+      "step": 62870,
+      "train_speed(iter/s)": 1.316301
+    },
+    {
+      "acc": 0.69165616,
+      "epoch": 1.5950025367833587,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0798085421990867e-06,
+      "loss": 1.38191023,
+      "memory(GiB)": 123.79,
+      "step": 62875,
+      "train_speed(iter/s)": 1.316311
+    },
+    {
+      "acc": 0.69729166,
+      "epoch": 1.5951293759512937,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.079157734634499e-06,
+      "loss": 1.35709829,
+      "memory(GiB)": 123.79,
+      "step": 62880,
+      "train_speed(iter/s)": 1.316324
+    },
+    {
+      "acc": 0.69320416,
+      "epoch": 1.5952562151192287,
+      "grad_norm": 2.296875,
+      "learning_rate": 1.0785070995260582e-06,
+      "loss": 1.40835342,
+      "memory(GiB)": 123.79,
+      "step": 62885,
+      "train_speed(iter/s)": 1.316337
+    },
+    {
+      "acc": 0.69944334,
+      "epoch": 1.5953830542871639,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.0778566369023841e-06,
+      "loss": 1.34427681,
+      "memory(GiB)": 123.79,
+      "step": 62890,
+      "train_speed(iter/s)": 1.316346
+    },
+    {
+      "acc": 0.68838425,
+      "epoch": 1.595509893455099,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.0772063467920863e-06,
+      "loss": 1.3736907,
+      "memory(GiB)": 123.79,
+      "step": 62895,
+      "train_speed(iter/s)": 1.316359
+    },
+    {
+      "acc": 0.68638802,
+      "epoch": 1.595636732623034,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.076556229223767e-06,
+      "loss": 1.41057386,
+      "memory(GiB)": 123.79,
+      "step": 62900,
+      "train_speed(iter/s)": 1.316371
+    },
+    {
+      "acc": 0.70578041,
+      "epoch": 1.595763571790969,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.0759062842260221e-06,
+      "loss": 1.34204235,
+      "memory(GiB)": 123.79,
+      "step": 62905,
+      "train_speed(iter/s)": 1.316383
+    },
+    {
+      "acc": 0.69198508,
+      "epoch": 1.595890410958904,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.0752565118274383e-06,
+      "loss": 1.41084976,
+      "memory(GiB)": 123.79,
+      "step": 62910,
+      "train_speed(iter/s)": 1.316396
+    },
+    {
+      "acc": 0.69262905,
+      "epoch": 1.596017250126839,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.0746069120565961e-06,
+      "loss": 1.37808552,
+      "memory(GiB)": 123.79,
+      "step": 62915,
+      "train_speed(iter/s)": 1.316408
+    },
+    {
+      "acc": 0.69886231,
+      "epoch": 1.5961440892947742,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.073957484942067e-06,
+      "loss": 1.3493783,
+      "memory(GiB)": 123.79,
+      "step": 62920,
+      "train_speed(iter/s)": 1.31642
+    },
+    {
+      "acc": 0.67995224,
+      "epoch": 1.5962709284627092,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0733082305124166e-06,
+      "loss": 1.37627411,
+      "memory(GiB)": 123.79,
+      "step": 62925,
+      "train_speed(iter/s)": 1.316432
+    },
+    {
+      "acc": 0.67750978,
+      "epoch": 1.5963977676306444,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.0726591487962018e-06,
+      "loss": 1.41139717,
+      "memory(GiB)": 123.79,
+      "step": 62930,
+      "train_speed(iter/s)": 1.316445
+    },
+    {
+      "acc": 0.68398175,
+      "epoch": 1.5965246067985794,
+      "grad_norm": 3.125,
+      "learning_rate": 1.0720102398219716e-06,
+      "loss": 1.38393517,
+      "memory(GiB)": 123.79,
+      "step": 62935,
+      "train_speed(iter/s)": 1.316458
+    },
+    {
+      "acc": 0.69692259,
+      "epoch": 1.5966514459665144,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.0713615036182684e-06,
+      "loss": 1.34411163,
+      "memory(GiB)": 123.79,
+      "step": 62940,
+      "train_speed(iter/s)": 1.31647
+    },
+    {
+      "acc": 0.70196242,
+      "epoch": 1.5967782851344494,
+      "grad_norm": 3.875,
+      "learning_rate": 1.0707129402136252e-06,
+      "loss": 1.37152109,
+      "memory(GiB)": 123.79,
+      "step": 62945,
+      "train_speed(iter/s)": 1.316483
+    },
+    {
+      "acc": 0.6935092,
+      "epoch": 1.5969051243023846,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0700645496365725e-06,
+      "loss": 1.35906734,
+      "memory(GiB)": 123.79,
+      "step": 62950,
+      "train_speed(iter/s)": 1.316495
+    },
+    {
+      "acc": 0.68451724,
+      "epoch": 1.5970319634703196,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.0694163319156254e-06,
+      "loss": 1.36171017,
+      "memory(GiB)": 123.79,
+      "step": 62955,
+      "train_speed(iter/s)": 1.316508
+    },
+    {
+      "acc": 0.6928226,
+      "epoch": 1.5971588026382548,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0687682870792953e-06,
+      "loss": 1.34972687,
+      "memory(GiB)": 123.79,
+      "step": 62960,
+      "train_speed(iter/s)": 1.316519
+    },
+    {
+      "acc": 0.68494329,
+      "epoch": 1.5972856418061898,
+      "grad_norm": 4.375,
+      "learning_rate": 1.0681204151560891e-06,
+      "loss": 1.39025726,
+      "memory(GiB)": 123.79,
+      "step": 62965,
+      "train_speed(iter/s)": 1.316532
+    },
+    {
+      "acc": 0.67207851,
+      "epoch": 1.5974124809741248,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.067472716174503e-06,
+      "loss": 1.43155413,
+      "memory(GiB)": 123.79,
+      "step": 62970,
+      "train_speed(iter/s)": 1.316545
+    },
+    {
+      "acc": 0.68204622,
+      "epoch": 1.5975393201420598,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0668251901630227e-06,
+      "loss": 1.41021061,
+      "memory(GiB)": 123.79,
+      "step": 62975,
+      "train_speed(iter/s)": 1.316557
+    },
+    {
+      "acc": 0.70298891,
+      "epoch": 1.5976661593099948,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0661778371501303e-06,
+      "loss": 1.36244545,
+      "memory(GiB)": 123.79,
+      "step": 62980,
+      "train_speed(iter/s)": 1.31657
+    },
+    {
+      "acc": 0.70144424,
+      "epoch": 1.59779299847793,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.0655306571643004e-06,
+      "loss": 1.39980602,
+      "memory(GiB)": 123.79,
+      "step": 62985,
+      "train_speed(iter/s)": 1.316582
+    },
+    {
+      "acc": 0.66880083,
+      "epoch": 1.5979198376458652,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0648836502339998e-06,
+      "loss": 1.45861683,
+      "memory(GiB)": 123.79,
+      "step": 62990,
+      "train_speed(iter/s)": 1.316595
+    },
+    {
+      "acc": 0.68844995,
+      "epoch": 1.5980466768138002,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0642368163876832e-06,
+      "loss": 1.37614613,
+      "memory(GiB)": 123.79,
+      "step": 62995,
+      "train_speed(iter/s)": 1.316607
+    },
+    {
+      "acc": 0.6854516,
+      "epoch": 1.5981735159817352,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.0635901556538042e-06,
+      "loss": 1.37183161,
+      "memory(GiB)": 123.79,
+      "step": 63000,
+      "train_speed(iter/s)": 1.316619
+    },
+    {
+      "epoch": 1.5981735159817352,
+      "eval_acc": 0.675049127401291,
+      "eval_loss": 1.3562912940979004,
+      "eval_runtime": 70.1104,
+      "eval_samples_per_second": 90.857,
+      "eval_steps_per_second": 22.721,
+      "step": 63000
+    },
+    {
+      "acc": 0.69399638,
+      "epoch": 1.5983003551496702,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.0629436680608051e-06,
+      "loss": 1.35314007,
+      "memory(GiB)": 123.79,
+      "step": 63005,
+      "train_speed(iter/s)": 1.314362
+    },
+    {
+      "acc": 0.69543734,
+      "epoch": 1.5984271943176052,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.0622973536371223e-06,
+      "loss": 1.33714294,
+      "memory(GiB)": 123.79,
+      "step": 63010,
+      "train_speed(iter/s)": 1.314375
+    },
+    {
+      "acc": 0.69831676,
+      "epoch": 1.5985540334855404,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.06165121241118e-06,
+      "loss": 1.31710672,
+      "memory(GiB)": 123.79,
+      "step": 63015,
+      "train_speed(iter/s)": 1.314387
+    },
+    {
+      "acc": 0.68831415,
+      "epoch": 1.5986808726534754,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.0610052444114023e-06,
+      "loss": 1.37291994,
+      "memory(GiB)": 123.79,
+      "step": 63020,
+      "train_speed(iter/s)": 1.314398
+    },
+    {
+      "acc": 0.69025435,
+      "epoch": 1.5988077118214106,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0603594496662001e-06,
+      "loss": 1.40817251,
+      "memory(GiB)": 123.79,
+      "step": 63025,
+      "train_speed(iter/s)": 1.314411
+    },
+    {
+      "acc": 0.6871223,
+      "epoch": 1.5989345509893456,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.0597138282039786e-06,
+      "loss": 1.33762579,
+      "memory(GiB)": 123.79,
+      "step": 63030,
+      "train_speed(iter/s)": 1.314424
+    },
+    {
+      "acc": 0.69298854,
+      "epoch": 1.5990613901572805,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.0590683800531348e-06,
+      "loss": 1.3340126,
+      "memory(GiB)": 123.79,
+      "step": 63035,
+      "train_speed(iter/s)": 1.314436
+    },
+    {
+      "acc": 0.69668407,
+      "epoch": 1.5991882293252155,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.058423105242059e-06,
+      "loss": 1.36562729,
+      "memory(GiB)": 123.79,
+      "step": 63040,
+      "train_speed(iter/s)": 1.314449
+    },
+    {
+      "acc": 0.69210863,
+      "epoch": 1.5993150684931505,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.057778003799133e-06,
+      "loss": 1.38609467,
+      "memory(GiB)": 123.79,
+      "step": 63045,
+      "train_speed(iter/s)": 1.314461
+    },
+    {
+      "acc": 0.68323584,
+      "epoch": 1.5994419076610857,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0571330757527309e-06,
+      "loss": 1.43359814,
+      "memory(GiB)": 123.79,
+      "step": 63050,
+      "train_speed(iter/s)": 1.314472
+    },
+    {
+      "acc": 0.67907944,
+      "epoch": 1.599568746829021,
+      "grad_norm": 3.25,
+      "learning_rate": 1.0564883211312199e-06,
+      "loss": 1.46517735,
+      "memory(GiB)": 123.79,
+      "step": 63055,
+      "train_speed(iter/s)": 1.314485
+    },
+    {
+      "acc": 0.69542751,
+      "epoch": 1.599695585996956,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.0558437399629584e-06,
+      "loss": 1.36132841,
+      "memory(GiB)": 123.79,
+      "step": 63060,
+      "train_speed(iter/s)": 1.314496
+    },
+    {
+      "acc": 0.69854221,
+      "epoch": 1.599822425164891,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.0551993322762994e-06,
+      "loss": 1.3837327,
+      "memory(GiB)": 123.79,
+      "step": 63065,
+      "train_speed(iter/s)": 1.314508
+    },
+    {
+      "acc": 0.68600039,
+      "epoch": 1.599949264332826,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.0545550980995857e-06,
+      "loss": 1.34637356,
+      "memory(GiB)": 123.79,
+      "step": 63070,
+      "train_speed(iter/s)": 1.314519
+    },
+    {
+      "acc": 0.69885035,
+      "epoch": 1.600076103500761,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.0539110374611538e-06,
+      "loss": 1.35662441,
+      "memory(GiB)": 123.79,
+      "step": 63075,
+      "train_speed(iter/s)": 1.31453
+    },
+    {
+      "acc": 0.70179477,
+      "epoch": 1.6002029426686961,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.0532671503893328e-06,
+      "loss": 1.28037605,
+      "memory(GiB)": 123.79,
+      "step": 63080,
+      "train_speed(iter/s)": 1.314542
+    },
+    {
+      "acc": 0.68250284,
+      "epoch": 1.600329781836631,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.052623436912442e-06,
+      "loss": 1.34388113,
+      "memory(GiB)": 123.79,
+      "step": 63085,
+      "train_speed(iter/s)": 1.314554
+    },
+    {
+      "acc": 0.68405399,
+      "epoch": 1.6004566210045663,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0519798970587992e-06,
+      "loss": 1.37268353,
+      "memory(GiB)": 123.79,
+      "step": 63090,
+      "train_speed(iter/s)": 1.314566
+    },
+    {
+      "acc": 0.67814178,
+      "epoch": 1.6005834601725013,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.0513365308567054e-06,
+      "loss": 1.42199278,
+      "memory(GiB)": 123.79,
+      "step": 63095,
+      "train_speed(iter/s)": 1.314579
+    },
+    {
+      "acc": 0.678019,
+      "epoch": 1.6007102993404363,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.0506933383344602e-06,
+      "loss": 1.43305044,
+      "memory(GiB)": 123.79,
+      "step": 63100,
+      "train_speed(iter/s)": 1.314591
+    },
+    {
+      "acc": 0.69100919,
+      "epoch": 1.6008371385083713,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.0500503195203537e-06,
+      "loss": 1.41755877,
+      "memory(GiB)": 123.79,
+      "step": 63105,
+      "train_speed(iter/s)": 1.314604
+    },
+    {
+      "acc": 0.6916172,
+      "epoch": 1.6009639776763065,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.049407474442672e-06,
+      "loss": 1.40880327,
+      "memory(GiB)": 123.79,
+      "step": 63110,
+      "train_speed(iter/s)": 1.314612
+    },
+    {
+      "acc": 0.67352796,
+      "epoch": 1.6010908168442415,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.048764803129686e-06,
+      "loss": 1.44865637,
+      "memory(GiB)": 123.79,
+      "step": 63115,
+      "train_speed(iter/s)": 1.314624
+    },
+    {
+      "acc": 0.705127,
+      "epoch": 1.6012176560121767,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.0481223056096635e-06,
+      "loss": 1.33452778,
+      "memory(GiB)": 123.79,
+      "step": 63120,
+      "train_speed(iter/s)": 1.314637
+    },
+    {
+      "acc": 0.68358793,
+      "epoch": 1.6013444951801117,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.0474799819108677e-06,
+      "loss": 1.40403976,
+      "memory(GiB)": 123.79,
+      "step": 63125,
+      "train_speed(iter/s)": 1.314649
+    },
+    {
+      "acc": 0.69333591,
+      "epoch": 1.6014713343480467,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.0468378320615502e-06,
+      "loss": 1.35484114,
+      "memory(GiB)": 123.79,
+      "step": 63130,
+      "train_speed(iter/s)": 1.314661
+    },
+    {
+      "acc": 0.69123969,
+      "epoch": 1.6015981735159817,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.0461958560899516e-06,
+      "loss": 1.38058701,
+      "memory(GiB)": 123.79,
+      "step": 63135,
+      "train_speed(iter/s)": 1.314673
+    },
+    {
+      "acc": 0.70114431,
+      "epoch": 1.6017250126839166,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.045554054024313e-06,
+      "loss": 1.30428581,
+      "memory(GiB)": 123.79,
+      "step": 63140,
+      "train_speed(iter/s)": 1.314686
+    },
+    {
+      "acc": 0.68290715,
+      "epoch": 1.6018518518518519,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0449124258928627e-06,
+      "loss": 1.40037479,
+      "memory(GiB)": 123.79,
+      "step": 63145,
+      "train_speed(iter/s)": 1.314698
+    },
+    {
+      "acc": 0.6913434,
+      "epoch": 1.601978691019787,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.044270971723823e-06,
+      "loss": 1.35726795,
+      "memory(GiB)": 123.79,
+      "step": 63150,
+      "train_speed(iter/s)": 1.314711
+    },
+    {
+      "acc": 0.69256701,
+      "epoch": 1.602105530187722,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.0436296915454048e-06,
+      "loss": 1.37519588,
+      "memory(GiB)": 123.79,
+      "step": 63155,
+      "train_speed(iter/s)": 1.314723
+    },
+    {
+      "acc": 0.67227659,
+      "epoch": 1.602232369355657,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.042988585385818e-06,
+      "loss": 1.42720413,
+      "memory(GiB)": 123.79,
+      "step": 63160,
+      "train_speed(iter/s)": 1.314736
+    },
+    {
+      "acc": 0.69470186,
+      "epoch": 1.602359208523592,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.0423476532732602e-06,
+      "loss": 1.34980087,
+      "memory(GiB)": 123.79,
+      "step": 63165,
+      "train_speed(iter/s)": 1.314748
+    },
+    {
+      "acc": 0.6889493,
+      "epoch": 1.602486047691527,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.0417068952359216e-06,
+      "loss": 1.37598505,
+      "memory(GiB)": 123.79,
+      "step": 63170,
+      "train_speed(iter/s)": 1.314761
+    },
+    {
+      "acc": 0.70504742,
+      "epoch": 1.6026128868594622,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.0410663113019869e-06,
+      "loss": 1.32769566,
+      "memory(GiB)": 123.79,
+      "step": 63175,
+      "train_speed(iter/s)": 1.314773
+    },
+    {
+      "acc": 0.70302505,
+      "epoch": 1.6027397260273972,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.040425901499631e-06,
+      "loss": 1.37470016,
+      "memory(GiB)": 123.79,
+      "step": 63180,
+      "train_speed(iter/s)": 1.314785
+    },
+    {
+      "acc": 0.70523911,
+      "epoch": 1.6028665651953324,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.0397856658570216e-06,
+      "loss": 1.31782646,
+      "memory(GiB)": 123.79,
+      "step": 63185,
+      "train_speed(iter/s)": 1.314796
+    },
+    {
+      "acc": 0.68698235,
+      "epoch": 1.6029934043632674,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.03914560440232e-06,
+      "loss": 1.39270477,
+      "memory(GiB)": 123.79,
+      "step": 63190,
+      "train_speed(iter/s)": 1.314808
+    },
+    {
+      "acc": 0.67917061,
+      "epoch": 1.6031202435312024,
+      "grad_norm": 3.25,
+      "learning_rate": 1.0385057171636791e-06,
+      "loss": 1.42988262,
+      "memory(GiB)": 123.79,
+      "step": 63195,
+      "train_speed(iter/s)": 1.31482
+    },
+    {
+      "acc": 0.70233765,
+      "epoch": 1.6032470826991374,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0378660041692435e-06,
+      "loss": 1.37191982,
+      "memory(GiB)": 123.79,
+      "step": 63200,
+      "train_speed(iter/s)": 1.314832
+    },
+    {
+      "acc": 0.6978169,
+      "epoch": 1.6033739218670724,
+      "grad_norm": 2.75,
+      "learning_rate": 1.0372264654471504e-06,
+      "loss": 1.32927914,
+      "memory(GiB)": 123.79,
+      "step": 63205,
+      "train_speed(iter/s)": 1.314844
+    },
+    {
+      "acc": 0.6953846,
+      "epoch": 1.6035007610350076,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0365871010255303e-06,
+      "loss": 1.40241585,
+      "memory(GiB)": 123.79,
+      "step": 63210,
+      "train_speed(iter/s)": 1.314855
+    },
+    {
+      "acc": 0.68315582,
+      "epoch": 1.6036276002029428,
+      "grad_norm": 2.625,
+      "learning_rate": 1.0359479109325043e-06,
+      "loss": 1.38997383,
+      "memory(GiB)": 123.79,
+      "step": 63215,
+      "train_speed(iter/s)": 1.314867
+    },
+    {
+      "acc": 0.6944562,
+      "epoch": 1.6037544393708778,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0353088951961877e-06,
+      "loss": 1.3589447,
+      "memory(GiB)": 123.79,
+      "step": 63220,
+      "train_speed(iter/s)": 1.31488
+    },
+    {
+      "acc": 0.69652042,
+      "epoch": 1.6038812785388128,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.0346700538446853e-06,
+      "loss": 1.32842474,
+      "memory(GiB)": 123.79,
+      "step": 63225,
+      "train_speed(iter/s)": 1.314891
+    },
+    {
+      "acc": 0.6896759,
+      "epoch": 1.6040081177067478,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.034031386906101e-06,
+      "loss": 1.36677122,
+      "memory(GiB)": 123.79,
+      "step": 63230,
+      "train_speed(iter/s)": 1.314903
+    },
+    {
+      "acc": 0.68890638,
+      "epoch": 1.6041349568746828,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.0333928944085214e-06,
+      "loss": 1.36941261,
+      "memory(GiB)": 123.79,
+      "step": 63235,
+      "train_speed(iter/s)": 1.314916
+    },
+    {
+      "acc": 0.68943143,
+      "epoch": 1.604261796042618,
+      "grad_norm": 3.375,
+      "learning_rate": 1.0327545763800322e-06,
+      "loss": 1.39205055,
+      "memory(GiB)": 123.79,
+      "step": 63240,
+      "train_speed(iter/s)": 1.314928
+    },
+    {
+      "acc": 0.69504032,
+      "epoch": 1.604388635210553,
+      "grad_norm": 7.0,
+      "learning_rate": 1.0321164328487077e-06,
+      "loss": 1.32199373,
+      "memory(GiB)": 123.79,
+      "step": 63245,
+      "train_speed(iter/s)": 1.31494
+    },
+    {
+      "acc": 0.67495785,
+      "epoch": 1.6045154743784882,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.0314784638426207e-06,
+      "loss": 1.41151619,
+      "memory(GiB)": 123.79,
+      "step": 63250,
+      "train_speed(iter/s)": 1.314953
+    },
+    {
+      "acc": 0.67820568,
+      "epoch": 1.6046423135464232,
+      "grad_norm": 3.0,
+      "learning_rate": 1.030840669389827e-06,
+      "loss": 1.39955502,
+      "memory(GiB)": 123.79,
+      "step": 63255,
+      "train_speed(iter/s)": 1.314965
+    },
+    {
+      "acc": 0.68686638,
+      "epoch": 1.6047691527143582,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.0302030495183812e-06,
+      "loss": 1.40070267,
+      "memory(GiB)": 123.79,
+      "step": 63260,
+      "train_speed(iter/s)": 1.314978
+    },
+    {
+      "acc": 0.6919054,
+      "epoch": 1.6048959918822931,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.0295656042563302e-06,
+      "loss": 1.37289543,
+      "memory(GiB)": 123.79,
+      "step": 63265,
+      "train_speed(iter/s)": 1.314991
+    },
+    {
+      "acc": 0.68469429,
+      "epoch": 1.6050228310502284,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.0289283336317119e-06,
+      "loss": 1.37597351,
+      "memory(GiB)": 123.79,
+      "step": 63270,
+      "train_speed(iter/s)": 1.315004
+    },
+    {
+      "acc": 0.69268055,
+      "epoch": 1.6051496702181633,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.0282912376725535e-06,
+      "loss": 1.40437603,
+      "memory(GiB)": 123.79,
+      "step": 63275,
+      "train_speed(iter/s)": 1.315016
+    },
+    {
+      "acc": 0.67951961,
+      "epoch": 1.6052765093860986,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.0276543164068776e-06,
+      "loss": 1.42002468,
+      "memory(GiB)": 123.79,
+      "step": 63280,
+      "train_speed(iter/s)": 1.315028
+    },
+    {
+      "acc": 0.70060935,
+      "epoch": 1.6054033485540335,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.0270175698627015e-06,
+      "loss": 1.33427162,
+      "memory(GiB)": 123.79,
+      "step": 63285,
+      "train_speed(iter/s)": 1.315041
+    },
+    {
+      "acc": 0.68885684,
+      "epoch": 1.6055301877219685,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.0263809980680323e-06,
+      "loss": 1.36552124,
+      "memory(GiB)": 123.79,
+      "step": 63290,
+      "train_speed(iter/s)": 1.315054
+    },
+    {
+      "acc": 0.68309183,
+      "epoch": 1.6056570268899035,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.0257446010508648e-06,
+      "loss": 1.33998308,
+      "memory(GiB)": 123.79,
+      "step": 63295,
+      "train_speed(iter/s)": 1.315066
+    },
+    {
+      "acc": 0.67747679,
+      "epoch": 1.6057838660578385,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.0251083788391952e-06,
+      "loss": 1.53177242,
+      "memory(GiB)": 123.79,
+      "step": 63300,
+      "train_speed(iter/s)": 1.315079
+    },
+    {
+      "acc": 0.68044233,
+      "epoch": 1.6059107052257737,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.0244723314610055e-06,
+      "loss": 1.39869757,
+      "memory(GiB)": 123.79,
+      "step": 63305,
+      "train_speed(iter/s)": 1.315091
+    },
+    {
+      "acc": 0.68313541,
+      "epoch": 1.606037544393709,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.0238364589442723e-06,
+      "loss": 1.46806946,
+      "memory(GiB)": 123.79,
+      "step": 63310,
+      "train_speed(iter/s)": 1.315103
+    },
+    {
+      "acc": 0.6861517,
+      "epoch": 1.606164383561644,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0232007613169637e-06,
+      "loss": 1.40463781,
+      "memory(GiB)": 123.79,
+      "step": 63315,
+      "train_speed(iter/s)": 1.315115
+    },
+    {
+      "acc": 0.69285884,
+      "epoch": 1.606291222729579,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.0225652386070406e-06,
+      "loss": 1.30930443,
+      "memory(GiB)": 123.79,
+      "step": 63320,
+      "train_speed(iter/s)": 1.315127
+    },
+    {
+      "acc": 0.69250674,
+      "epoch": 1.606418061897514,
+      "grad_norm": 3.625,
+      "learning_rate": 1.0219298908424568e-06,
+      "loss": 1.37228432,
+      "memory(GiB)": 123.79,
+      "step": 63325,
+      "train_speed(iter/s)": 1.315139
+    },
+    {
+      "acc": 0.68521371,
+      "epoch": 1.606544901065449,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.0212947180511567e-06,
+      "loss": 1.38495407,
+      "memory(GiB)": 123.79,
+      "step": 63330,
+      "train_speed(iter/s)": 1.315152
+    },
+    {
+      "acc": 0.68521881,
+      "epoch": 1.606671740233384,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.020659720261079e-06,
+      "loss": 1.42822361,
+      "memory(GiB)": 123.79,
+      "step": 63335,
+      "train_speed(iter/s)": 1.315164
+    },
+    {
+      "acc": 0.69249921,
+      "epoch": 1.606798579401319,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.020024897500153e-06,
+      "loss": 1.37884445,
+      "memory(GiB)": 123.79,
+      "step": 63340,
+      "train_speed(iter/s)": 1.315177
+    },
+    {
+      "acc": 0.69277091,
+      "epoch": 1.6069254185692543,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.0193902497963014e-06,
+      "loss": 1.33750725,
+      "memory(GiB)": 123.79,
+      "step": 63345,
+      "train_speed(iter/s)": 1.315189
+    },
+    {
+      "acc": 0.69784079,
+      "epoch": 1.6070522577371893,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.0187557771774387e-06,
+      "loss": 1.40491047,
+      "memory(GiB)": 123.79,
+      "step": 63350,
+      "train_speed(iter/s)": 1.315202
+    },
+    {
+      "acc": 0.68778963,
+      "epoch": 1.6071790969051243,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0181214796714717e-06,
+      "loss": 1.43299732,
+      "memory(GiB)": 123.79,
+      "step": 63355,
+      "train_speed(iter/s)": 1.315214
+    },
+    {
+      "acc": 0.68181667,
+      "epoch": 1.6073059360730593,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.0174873573062998e-06,
+      "loss": 1.47646217,
+      "memory(GiB)": 123.79,
+      "step": 63360,
+      "train_speed(iter/s)": 1.315226
+    },
+    {
+      "acc": 0.69510527,
+      "epoch": 1.6074327752409943,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.0168534101098148e-06,
+      "loss": 1.35817862,
+      "memory(GiB)": 123.79,
+      "step": 63365,
+      "train_speed(iter/s)": 1.315239
+    },
+    {
+      "acc": 0.690203,
+      "epoch": 1.6075596144089295,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.0162196381099004e-06,
+      "loss": 1.39120083,
+      "memory(GiB)": 123.79,
+      "step": 63370,
+      "train_speed(iter/s)": 1.315251
+    },
+    {
+      "acc": 0.70102196,
+      "epoch": 1.6076864535768647,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.0155860413344327e-06,
+      "loss": 1.37272196,
+      "memory(GiB)": 123.79,
+      "step": 63375,
+      "train_speed(iter/s)": 1.315263
+    },
+    {
+      "acc": 0.69357357,
+      "epoch": 1.6078132927447997,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.0149526198112797e-06,
+      "loss": 1.33984108,
+      "memory(GiB)": 123.79,
+      "step": 63380,
+      "train_speed(iter/s)": 1.315275
+    },
+    {
+      "acc": 0.70144081,
+      "epoch": 1.6079401319127347,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.0143193735683016e-06,
+      "loss": 1.37739334,
+      "memory(GiB)": 123.79,
+      "step": 63385,
+      "train_speed(iter/s)": 1.315287
+    },
+    {
+      "acc": 0.6896451,
+      "epoch": 1.6080669710806696,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.0136863026333543e-06,
+      "loss": 1.40554466,
+      "memory(GiB)": 123.79,
+      "step": 63390,
+      "train_speed(iter/s)": 1.315299
+    },
+    {
+      "acc": 0.67998152,
+      "epoch": 1.6081938102486046,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.0130534070342802e-06,
+      "loss": 1.43170033,
+      "memory(GiB)": 123.79,
+      "step": 63395,
+      "train_speed(iter/s)": 1.315311
+    },
+    {
+      "acc": 0.69971499,
+      "epoch": 1.6083206494165398,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.0124206867989157e-06,
+      "loss": 1.40553532,
+      "memory(GiB)": 123.79,
+      "step": 63400,
+      "train_speed(iter/s)": 1.315322
+    },
+    {
+      "acc": 0.68520517,
+      "epoch": 1.6084474885844748,
+      "grad_norm": 2.875,
+      "learning_rate": 1.0117881419550945e-06,
+      "loss": 1.43418846,
+      "memory(GiB)": 123.79,
+      "step": 63405,
+      "train_speed(iter/s)": 1.315333
+    },
+    {
+      "acc": 0.69418793,
+      "epoch": 1.60857432775241,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.0111557725306382e-06,
+      "loss": 1.38260574,
+      "memory(GiB)": 123.79,
+      "step": 63410,
+      "train_speed(iter/s)": 1.315343
+    },
+    {
+      "acc": 0.6707551,
+      "epoch": 1.608701166920345,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0105235785533585e-06,
+      "loss": 1.47531443,
+      "memory(GiB)": 123.79,
+      "step": 63415,
+      "train_speed(iter/s)": 1.315355
+    },
+    {
+      "acc": 0.67943125,
+      "epoch": 1.60882800608828,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.0098915600510623e-06,
+      "loss": 1.44060249,
+      "memory(GiB)": 123.79,
+      "step": 63420,
+      "train_speed(iter/s)": 1.315367
+    },
+    {
+      "acc": 0.69192615,
+      "epoch": 1.608954845256215,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.0092597170515512e-06,
+      "loss": 1.34550114,
+      "memory(GiB)": 123.79,
+      "step": 63425,
+      "train_speed(iter/s)": 1.315378
+    },
+    {
+      "acc": 0.68090034,
+      "epoch": 1.6090816844241502,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.0086280495826161e-06,
+      "loss": 1.44075689,
+      "memory(GiB)": 123.79,
+      "step": 63430,
+      "train_speed(iter/s)": 1.315389
+    },
+    {
+      "acc": 0.69307852,
+      "epoch": 1.6092085235920852,
+      "grad_norm": 2.75,
+      "learning_rate": 1.0079965576720375e-06,
+      "loss": 1.34482946,
+      "memory(GiB)": 123.79,
+      "step": 63435,
+      "train_speed(iter/s)": 1.315401
+    },
+    {
+      "acc": 0.69305124,
+      "epoch": 1.6093353627600204,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.0073652413475936e-06,
+      "loss": 1.36501179,
+      "memory(GiB)": 123.79,
+      "step": 63440,
+      "train_speed(iter/s)": 1.315413
+    },
+    {
+      "acc": 0.70703211,
+      "epoch": 1.6094622019279554,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.0067341006370535e-06,
+      "loss": 1.35129805,
+      "memory(GiB)": 123.79,
+      "step": 63445,
+      "train_speed(iter/s)": 1.315425
+    },
+    {
+      "acc": 0.67713814,
+      "epoch": 1.6095890410958904,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.0061031355681766e-06,
+      "loss": 1.34448061,
+      "memory(GiB)": 123.79,
+      "step": 63450,
+      "train_speed(iter/s)": 1.315437
+    },
+    {
+      "acc": 0.68351893,
+      "epoch": 1.6097158802638254,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.0054723461687133e-06,
+      "loss": 1.43168077,
+      "memory(GiB)": 123.79,
+      "step": 63455,
+      "train_speed(iter/s)": 1.315448
+    },
+    {
+      "acc": 0.70073948,
+      "epoch": 1.6098427194317604,
+      "grad_norm": 2.625,
+      "learning_rate": 1.0048417324664118e-06,
+      "loss": 1.33550987,
+      "memory(GiB)": 123.79,
+      "step": 63460,
+      "train_speed(iter/s)": 1.31546
+    },
+    {
+      "acc": 0.68993564,
+      "epoch": 1.6099695585996956,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.0042112944890075e-06,
+      "loss": 1.40241547,
+      "memory(GiB)": 123.79,
+      "step": 63465,
+      "train_speed(iter/s)": 1.315472
+    },
+    {
+      "acc": 0.69851704,
+      "epoch": 1.6100963977676308,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.003581032264231e-06,
+      "loss": 1.3292861,
+      "memory(GiB)": 123.79,
+      "step": 63470,
+      "train_speed(iter/s)": 1.315484
+    },
+    {
+      "acc": 0.69365683,
+      "epoch": 1.6102232369355658,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.0029509458198027e-06,
+      "loss": 1.33441792,
+      "memory(GiB)": 123.79,
+      "step": 63475,
+      "train_speed(iter/s)": 1.315496
+    },
+    {
+      "acc": 0.69525099,
+      "epoch": 1.6103500761035008,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.0023210351834378e-06,
+      "loss": 1.40994511,
+      "memory(GiB)": 123.79,
+      "step": 63480,
+      "train_speed(iter/s)": 1.315509
+    },
+    {
+      "acc": 0.69066858,
+      "epoch": 1.6104769152714358,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.001691300382842e-06,
+      "loss": 1.40779877,
+      "memory(GiB)": 123.79,
+      "step": 63485,
+      "train_speed(iter/s)": 1.315519
+    },
+    {
+      "acc": 0.68844919,
+      "epoch": 1.6106037544393708,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.001061741445714e-06,
+      "loss": 1.35239811,
+      "memory(GiB)": 123.79,
+      "step": 63490,
+      "train_speed(iter/s)": 1.31553
+    },
+    {
+      "acc": 0.67826705,
+      "epoch": 1.610730593607306,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.000432358399745e-06,
+      "loss": 1.40198841,
+      "memory(GiB)": 123.79,
+      "step": 63495,
+      "train_speed(iter/s)": 1.315541
+    },
+    {
+      "acc": 0.70269642,
+      "epoch": 1.610857432775241,
+      "grad_norm": 3.5,
+      "learning_rate": 9.99803151272617e-07,
+      "loss": 1.32973948,
+      "memory(GiB)": 123.79,
+      "step": 63500,
+      "train_speed(iter/s)": 1.315553
+    },
+    {
+      "acc": 0.69364462,
+      "epoch": 1.6109842719431762,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.991741200920062e-07,
+      "loss": 1.42210884,
+      "memory(GiB)": 123.79,
+      "step": 63505,
+      "train_speed(iter/s)": 1.315565
+    },
+    {
+      "acc": 0.69861889,
+      "epoch": 1.6111111111111112,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.985452648855803e-07,
+      "loss": 1.35764856,
+      "memory(GiB)": 123.79,
+      "step": 63510,
+      "train_speed(iter/s)": 1.315577
+    },
+    {
+      "acc": 0.67544088,
+      "epoch": 1.6112379502790461,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.979165856809985e-07,
+      "loss": 1.39980507,
+      "memory(GiB)": 123.79,
+      "step": 63515,
+      "train_speed(iter/s)": 1.315589
+    },
+    {
+      "acc": 0.68084183,
+      "epoch": 1.6113647894469811,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.972880825059134e-07,
+      "loss": 1.44752483,
+      "memory(GiB)": 123.79,
+      "step": 63520,
+      "train_speed(iter/s)": 1.315601
+    },
+    {
+      "acc": 0.68887892,
+      "epoch": 1.6114916286149161,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.966597553879681e-07,
+      "loss": 1.34129601,
+      "memory(GiB)": 123.79,
+      "step": 63525,
+      "train_speed(iter/s)": 1.315613
+    },
+    {
+      "acc": 0.6916677,
+      "epoch": 1.6116184677828513,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.96031604354803e-07,
+      "loss": 1.45772266,
+      "memory(GiB)": 123.79,
+      "step": 63530,
+      "train_speed(iter/s)": 1.315625
+    },
+    {
+      "acc": 0.68696733,
+      "epoch": 1.6117453069507865,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.954036294340425e-07,
+      "loss": 1.45103579,
+      "memory(GiB)": 123.79,
+      "step": 63535,
+      "train_speed(iter/s)": 1.315638
+    },
+    {
+      "acc": 0.69177408,
+      "epoch": 1.6118721461187215,
+      "grad_norm": 3.875,
+      "learning_rate": 9.947758306533101e-07,
+      "loss": 1.33466473,
+      "memory(GiB)": 123.79,
+      "step": 63540,
+      "train_speed(iter/s)": 1.31565
+    },
+    {
+      "acc": 0.70044217,
+      "epoch": 1.6119989852866565,
+      "grad_norm": 2.984375,
+      "learning_rate": 9.941482080402177e-07,
+      "loss": 1.37543297,
+      "memory(GiB)": 123.79,
+      "step": 63545,
+      "train_speed(iter/s)": 1.315661
+    },
+    {
+      "acc": 0.68879099,
+      "epoch": 1.6121258244545915,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.935207616223741e-07,
+      "loss": 1.37473793,
+      "memory(GiB)": 123.79,
+      "step": 63550,
+      "train_speed(iter/s)": 1.315673
+    },
+    {
+      "acc": 0.69638491,
+      "epoch": 1.6122526636225265,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.928934914273735e-07,
+      "loss": 1.35512524,
+      "memory(GiB)": 123.79,
+      "step": 63555,
+      "train_speed(iter/s)": 1.315685
+    },
+    {
+      "acc": 0.68975501,
+      "epoch": 1.6123795027904617,
+      "grad_norm": 3.0,
+      "learning_rate": 9.922663974828066e-07,
+      "loss": 1.39748383,
+      "memory(GiB)": 123.79,
+      "step": 63560,
+      "train_speed(iter/s)": 1.315697
+    },
+    {
+      "acc": 0.68238468,
+      "epoch": 1.6125063419583967,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.916394798162582e-07,
+      "loss": 1.4028162,
+      "memory(GiB)": 123.79,
+      "step": 63565,
+      "train_speed(iter/s)": 1.315708
+    },
+    {
+      "acc": 0.68370867,
+      "epoch": 1.612633181126332,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.91012738455303e-07,
+      "loss": 1.38783302,
+      "memory(GiB)": 123.79,
+      "step": 63570,
+      "train_speed(iter/s)": 1.31572
+    },
+    {
+      "acc": 0.69785008,
+      "epoch": 1.612760020294267,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.903861734275032e-07,
+      "loss": 1.31210518,
+      "memory(GiB)": 123.79,
+      "step": 63575,
+      "train_speed(iter/s)": 1.315731
+    },
+    {
+      "acc": 0.7039072,
+      "epoch": 1.612886859462202,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.897597847604228e-07,
+      "loss": 1.36420517,
+      "memory(GiB)": 123.79,
+      "step": 63580,
+      "train_speed(iter/s)": 1.315742
+    },
+    {
+      "acc": 0.6814497,
+      "epoch": 1.6130136986301369,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.89133572481612e-07,
+      "loss": 1.43110676,
+      "memory(GiB)": 123.79,
+      "step": 63585,
+      "train_speed(iter/s)": 1.315754
+    },
+    {
+      "acc": 0.68511362,
+      "epoch": 1.613140537798072,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.885075366186148e-07,
+      "loss": 1.40105047,
+      "memory(GiB)": 123.79,
+      "step": 63590,
+      "train_speed(iter/s)": 1.315766
+    },
+    {
+      "acc": 0.69952908,
+      "epoch": 1.613267376966007,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.87881677198963e-07,
+      "loss": 1.32750511,
+      "memory(GiB)": 123.79,
+      "step": 63595,
+      "train_speed(iter/s)": 1.315777
+    },
+    {
+      "acc": 0.67563653,
+      "epoch": 1.6133942161339423,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.872559942501897e-07,
+      "loss": 1.5031765,
+      "memory(GiB)": 123.79,
+      "step": 63600,
+      "train_speed(iter/s)": 1.315789
+    },
+    {
+      "acc": 0.66383352,
+      "epoch": 1.6135210553018773,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.866304877998134e-07,
+      "loss": 1.4424798,
+      "memory(GiB)": 123.79,
+      "step": 63605,
+      "train_speed(iter/s)": 1.315801
+    },
+    {
+      "acc": 0.69350452,
+      "epoch": 1.6136478944698123,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.860051578753466e-07,
+      "loss": 1.43087454,
+      "memory(GiB)": 123.79,
+      "step": 63610,
+      "train_speed(iter/s)": 1.315812
+    },
+    {
+      "acc": 0.68320365,
+      "epoch": 1.6137747336377473,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.853800045042938e-07,
+      "loss": 1.40421371,
+      "memory(GiB)": 123.79,
+      "step": 63615,
+      "train_speed(iter/s)": 1.315824
+    },
+    {
+      "acc": 0.68412657,
+      "epoch": 1.6139015728056822,
+      "grad_norm": 2.625,
+      "learning_rate": 9.847550277141526e-07,
+      "loss": 1.4095253,
+      "memory(GiB)": 123.79,
+      "step": 63620,
+      "train_speed(iter/s)": 1.315835
+    },
+    {
+      "acc": 0.6902473,
+      "epoch": 1.6140284119736175,
+      "grad_norm": 3.375,
+      "learning_rate": 9.841302275324128e-07,
+      "loss": 1.36906719,
+      "memory(GiB)": 123.79,
+      "step": 63625,
+      "train_speed(iter/s)": 1.315847
+    },
+    {
+      "acc": 0.6819582,
+      "epoch": 1.6141552511415527,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.835056039865542e-07,
+      "loss": 1.40986099,
+      "memory(GiB)": 123.79,
+      "step": 63630,
+      "train_speed(iter/s)": 1.315859
+    },
+    {
+      "acc": 0.68194556,
+      "epoch": 1.6142820903094877,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.82881157104052e-07,
+      "loss": 1.44816151,
+      "memory(GiB)": 123.79,
+      "step": 63635,
+      "train_speed(iter/s)": 1.315871
+    },
+    {
+      "acc": 0.67247024,
+      "epoch": 1.6144089294774226,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.822568869123712e-07,
+      "loss": 1.47790632,
+      "memory(GiB)": 123.79,
+      "step": 63640,
+      "train_speed(iter/s)": 1.315883
+    },
+    {
+      "acc": 0.67836485,
+      "epoch": 1.6145357686453576,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.816327934389707e-07,
+      "loss": 1.39369774,
+      "memory(GiB)": 123.79,
+      "step": 63645,
+      "train_speed(iter/s)": 1.315895
+    },
+    {
+      "acc": 0.70425396,
+      "epoch": 1.6146626078132926,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.810088767113008e-07,
+      "loss": 1.32606297,
+      "memory(GiB)": 123.79,
+      "step": 63650,
+      "train_speed(iter/s)": 1.315907
+    },
+    {
+      "acc": 0.68315635,
+      "epoch": 1.6147894469812278,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.80385136756804e-07,
+      "loss": 1.43304272,
+      "memory(GiB)": 123.79,
+      "step": 63655,
+      "train_speed(iter/s)": 1.31592
+    },
+    {
+      "acc": 0.68405714,
+      "epoch": 1.6149162861491628,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.797615736029148e-07,
+      "loss": 1.40207357,
+      "memory(GiB)": 123.79,
+      "step": 63660,
+      "train_speed(iter/s)": 1.315932
+    },
+    {
+      "acc": 0.69038372,
+      "epoch": 1.615043125317098,
+      "grad_norm": 2.75,
+      "learning_rate": 9.791381872770594e-07,
+      "loss": 1.38212147,
+      "memory(GiB)": 123.79,
+      "step": 63665,
+      "train_speed(iter/s)": 1.315945
+    },
+    {
+      "acc": 0.68561897,
+      "epoch": 1.615169964485033,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.785149778066615e-07,
+      "loss": 1.42091608,
+      "memory(GiB)": 123.79,
+      "step": 63670,
+      "train_speed(iter/s)": 1.315957
+    },
+    {
+      "acc": 0.69424248,
+      "epoch": 1.615296803652968,
+      "grad_norm": 3.375,
+      "learning_rate": 9.778919452191277e-07,
+      "loss": 1.36161995,
+      "memory(GiB)": 123.79,
+      "step": 63675,
+      "train_speed(iter/s)": 1.315969
+    },
+    {
+      "acc": 0.7104744,
+      "epoch": 1.615423642820903,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.77269089541864e-07,
+      "loss": 1.3163826,
+      "memory(GiB)": 123.79,
+      "step": 63680,
+      "train_speed(iter/s)": 1.315982
+    },
+    {
+      "acc": 0.68603239,
+      "epoch": 1.615550481988838,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.766464108022644e-07,
+      "loss": 1.43273144,
+      "memory(GiB)": 123.79,
+      "step": 63685,
+      "train_speed(iter/s)": 1.315994
+    },
+    {
+      "acc": 0.69820395,
+      "epoch": 1.6156773211567732,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.760239090277213e-07,
+      "loss": 1.36387081,
+      "memory(GiB)": 123.79,
+      "step": 63690,
+      "train_speed(iter/s)": 1.316006
+    },
+    {
+      "acc": 0.7006834,
+      "epoch": 1.6158041603247084,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.75401584245611e-07,
+      "loss": 1.36688824,
+      "memory(GiB)": 123.79,
+      "step": 63695,
+      "train_speed(iter/s)": 1.316018
+    },
+    {
+      "acc": 0.66652727,
+      "epoch": 1.6159309994926434,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.747794364833063e-07,
+      "loss": 1.49878254,
+      "memory(GiB)": 123.79,
+      "step": 63700,
+      "train_speed(iter/s)": 1.31603
+    },
+    {
+      "acc": 0.67548132,
+      "epoch": 1.6160578386605784,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.741574657681747e-07,
+      "loss": 1.43865681,
+      "memory(GiB)": 123.79,
+      "step": 63705,
+      "train_speed(iter/s)": 1.316043
+    },
+    {
+      "acc": 0.67584615,
+      "epoch": 1.6161846778285134,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.735356721275734e-07,
+      "loss": 1.39755096,
+      "memory(GiB)": 123.79,
+      "step": 63710,
+      "train_speed(iter/s)": 1.316055
+    },
+    {
+      "acc": 0.68857217,
+      "epoch": 1.6163115169964484,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.729140555888483e-07,
+      "loss": 1.29655466,
+      "memory(GiB)": 123.79,
+      "step": 63715,
+      "train_speed(iter/s)": 1.316068
+    },
+    {
+      "acc": 0.69414015,
+      "epoch": 1.6164383561643836,
+      "grad_norm": 3.625,
+      "learning_rate": 9.722926161793417e-07,
+      "loss": 1.35395727,
+      "memory(GiB)": 123.79,
+      "step": 63720,
+      "train_speed(iter/s)": 1.31608
+    },
+    {
+      "acc": 0.6847105,
+      "epoch": 1.6165651953323186,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.716713539263895e-07,
+      "loss": 1.43237906,
+      "memory(GiB)": 123.79,
+      "step": 63725,
+      "train_speed(iter/s)": 1.316092
+    },
+    {
+      "acc": 0.68849301,
+      "epoch": 1.6166920345002538,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.710502688573175e-07,
+      "loss": 1.40714436,
+      "memory(GiB)": 123.79,
+      "step": 63730,
+      "train_speed(iter/s)": 1.316104
+    },
+    {
+      "acc": 0.68222761,
+      "epoch": 1.6168188736681888,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.704293609994403e-07,
+      "loss": 1.41091518,
+      "memory(GiB)": 123.79,
+      "step": 63735,
+      "train_speed(iter/s)": 1.316115
+    },
+    {
+      "acc": 0.69284267,
+      "epoch": 1.6169457128361238,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.69808630380072e-07,
+      "loss": 1.42506695,
+      "memory(GiB)": 123.79,
+      "step": 63740,
+      "train_speed(iter/s)": 1.316128
+    },
+    {
+      "acc": 0.69510899,
+      "epoch": 1.6170725520040587,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.691880770265132e-07,
+      "loss": 1.32833748,
+      "memory(GiB)": 123.79,
+      "step": 63745,
+      "train_speed(iter/s)": 1.316141
+    },
+    {
+      "acc": 0.68956213,
+      "epoch": 1.617199391171994,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.685677009660587e-07,
+      "loss": 1.42150879,
+      "memory(GiB)": 123.79,
+      "step": 63750,
+      "train_speed(iter/s)": 1.316153
+    },
+    {
+      "acc": 0.67727718,
+      "epoch": 1.617326230339929,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.679475022259965e-07,
+      "loss": 1.47047873,
+      "memory(GiB)": 123.79,
+      "step": 63755,
+      "train_speed(iter/s)": 1.316166
+    },
+    {
+      "acc": 0.6925107,
+      "epoch": 1.6174530695078642,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.673274808336047e-07,
+      "loss": 1.40965958,
+      "memory(GiB)": 123.79,
+      "step": 63760,
+      "train_speed(iter/s)": 1.316179
+    },
+    {
+      "acc": 0.67467442,
+      "epoch": 1.6175799086757991,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.66707636816155e-07,
+      "loss": 1.49770136,
+      "memory(GiB)": 123.79,
+      "step": 63765,
+      "train_speed(iter/s)": 1.31619
+    },
+    {
+      "acc": 0.67912712,
+      "epoch": 1.6177067478437341,
+      "grad_norm": 3.421875,
+      "learning_rate": 9.660879702009106e-07,
+      "loss": 1.48965778,
+      "memory(GiB)": 123.79,
+      "step": 63770,
+      "train_speed(iter/s)": 1.316203
+    },
+    {
+      "acc": 0.67214699,
+      "epoch": 1.6178335870116691,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.654684810151276e-07,
+      "loss": 1.47250271,
+      "memory(GiB)": 123.79,
+      "step": 63775,
+      "train_speed(iter/s)": 1.316213
+    },
+    {
+      "acc": 0.68728371,
+      "epoch": 1.6179604261796041,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.648491692860534e-07,
+      "loss": 1.38605042,
+      "memory(GiB)": 123.79,
+      "step": 63780,
+      "train_speed(iter/s)": 1.316226
+    },
+    {
+      "acc": 0.69613371,
+      "epoch": 1.6180872653475393,
+      "grad_norm": 3.25,
+      "learning_rate": 9.642300350409289e-07,
+      "loss": 1.3931448,
+      "memory(GiB)": 123.79,
+      "step": 63785,
+      "train_speed(iter/s)": 1.316238
+    },
+    {
+      "acc": 0.67802792,
+      "epoch": 1.6182141045154745,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.636110783069852e-07,
+      "loss": 1.48018799,
+      "memory(GiB)": 123.79,
+      "step": 63790,
+      "train_speed(iter/s)": 1.31625
+    },
+    {
+      "acc": 0.68119307,
+      "epoch": 1.6183409436834095,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.629922991114482e-07,
+      "loss": 1.46517639,
+      "memory(GiB)": 123.79,
+      "step": 63795,
+      "train_speed(iter/s)": 1.316262
+    },
+    {
+      "acc": 0.68185596,
+      "epoch": 1.6184677828513445,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.623736974815334e-07,
+      "loss": 1.39399471,
+      "memory(GiB)": 123.79,
+      "step": 63800,
+      "train_speed(iter/s)": 1.316275
+    },
+    {
+      "acc": 0.70379438,
+      "epoch": 1.6185946220192795,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.617552734444502e-07,
+      "loss": 1.29916134,
+      "memory(GiB)": 123.79,
+      "step": 63805,
+      "train_speed(iter/s)": 1.316287
+    },
+    {
+      "acc": 0.69903331,
+      "epoch": 1.6187214611872145,
+      "grad_norm": 2.5625,
+      "learning_rate": 9.611370270273996e-07,
+      "loss": 1.30386801,
+      "memory(GiB)": 123.79,
+      "step": 63810,
+      "train_speed(iter/s)": 1.3163
+    },
+    {
+      "acc": 0.68196583,
+      "epoch": 1.6188483003551497,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.605189582575741e-07,
+      "loss": 1.34863338,
+      "memory(GiB)": 123.79,
+      "step": 63815,
+      "train_speed(iter/s)": 1.316312
+    },
+    {
+      "acc": 0.68543997,
+      "epoch": 1.6189751395230847,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.599010671621605e-07,
+      "loss": 1.41550274,
+      "memory(GiB)": 123.79,
+      "step": 63820,
+      "train_speed(iter/s)": 1.316324
+    },
+    {
+      "acc": 0.68744493,
+      "epoch": 1.61910197869102,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.592833537683344e-07,
+      "loss": 1.41393089,
+      "memory(GiB)": 123.79,
+      "step": 63825,
+      "train_speed(iter/s)": 1.316337
+    },
+    {
+      "acc": 0.6725709,
+      "epoch": 1.619228817858955,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.586658181032693e-07,
+      "loss": 1.4490921,
+      "memory(GiB)": 123.79,
+      "step": 63830,
+      "train_speed(iter/s)": 1.316349
+    },
+    {
+      "acc": 0.68531666,
+      "epoch": 1.6193556570268899,
+      "grad_norm": 2.625,
+      "learning_rate": 9.580484601941237e-07,
+      "loss": 1.3333046,
+      "memory(GiB)": 123.79,
+      "step": 63835,
+      "train_speed(iter/s)": 1.316362
+    },
+    {
+      "acc": 0.681394,
+      "epoch": 1.6194824961948249,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.574312800680514e-07,
+      "loss": 1.44983807,
+      "memory(GiB)": 123.79,
+      "step": 63840,
+      "train_speed(iter/s)": 1.316375
+    },
+    {
+      "acc": 0.68640728,
+      "epoch": 1.6196093353627599,
+      "grad_norm": 3.5,
+      "learning_rate": 9.56814277752201e-07,
+      "loss": 1.38446407,
+      "memory(GiB)": 123.79,
+      "step": 63845,
+      "train_speed(iter/s)": 1.316387
+    },
+    {
+      "acc": 0.6735353,
+      "epoch": 1.619736174530695,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.561974532737124e-07,
+      "loss": 1.46848793,
+      "memory(GiB)": 123.79,
+      "step": 63850,
+      "train_speed(iter/s)": 1.3164
+    },
+    {
+      "acc": 0.67263708,
+      "epoch": 1.6198630136986303,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.555808066597123e-07,
+      "loss": 1.3476141,
+      "memory(GiB)": 123.79,
+      "step": 63855,
+      "train_speed(iter/s)": 1.316413
+    },
+    {
+      "acc": 0.69550962,
+      "epoch": 1.6199898528665653,
+      "grad_norm": 3.5,
+      "learning_rate": 9.549643379373236e-07,
+      "loss": 1.3652277,
+      "memory(GiB)": 123.79,
+      "step": 63860,
+      "train_speed(iter/s)": 1.316425
+    },
+    {
+      "acc": 0.68447294,
+      "epoch": 1.6201166920345003,
+      "grad_norm": 3.5,
+      "learning_rate": 9.543480471336652e-07,
+      "loss": 1.38811474,
+      "memory(GiB)": 123.79,
+      "step": 63865,
+      "train_speed(iter/s)": 1.316437
+    },
+    {
+      "acc": 0.71180315,
+      "epoch": 1.6202435312024352,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.537319342758434e-07,
+      "loss": 1.29254942,
+      "memory(GiB)": 123.79,
+      "step": 63870,
+      "train_speed(iter/s)": 1.31645
+    },
+    {
+      "acc": 0.67774124,
+      "epoch": 1.6203703703703702,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.531159993909533e-07,
+      "loss": 1.37674942,
+      "memory(GiB)": 123.79,
+      "step": 63875,
+      "train_speed(iter/s)": 1.316462
+    },
+    {
+      "acc": 0.70510559,
+      "epoch": 1.6204972095383054,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.525002425060914e-07,
+      "loss": 1.27190542,
+      "memory(GiB)": 123.79,
+      "step": 63880,
+      "train_speed(iter/s)": 1.316474
+    },
+    {
+      "acc": 0.68657274,
+      "epoch": 1.6206240487062404,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.518846636483392e-07,
+      "loss": 1.40576019,
+      "memory(GiB)": 123.79,
+      "step": 63885,
+      "train_speed(iter/s)": 1.316486
+    },
+    {
+      "acc": 0.66840734,
+      "epoch": 1.6207508878741756,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.512692628447745e-07,
+      "loss": 1.46900101,
+      "memory(GiB)": 123.79,
+      "step": 63890,
+      "train_speed(iter/s)": 1.316498
+    },
+    {
+      "acc": 0.68989134,
+      "epoch": 1.6208777270421106,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.506540401224612e-07,
+      "loss": 1.3565239,
+      "memory(GiB)": 123.79,
+      "step": 63895,
+      "train_speed(iter/s)": 1.31651
+    },
+    {
+      "acc": 0.69092541,
+      "epoch": 1.6210045662100456,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.500389955084638e-07,
+      "loss": 1.37884312,
+      "memory(GiB)": 123.79,
+      "step": 63900,
+      "train_speed(iter/s)": 1.316523
+    },
+    {
+      "acc": 0.69110394,
+      "epoch": 1.6211314053779806,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.494241290298334e-07,
+      "loss": 1.38187733,
+      "memory(GiB)": 123.79,
+      "step": 63905,
+      "train_speed(iter/s)": 1.316535
+    },
+    {
+      "acc": 0.6810441,
+      "epoch": 1.6212582445459158,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.488094407136139e-07,
+      "loss": 1.39420538,
+      "memory(GiB)": 123.79,
+      "step": 63910,
+      "train_speed(iter/s)": 1.316548
+    },
+    {
+      "acc": 0.68828049,
+      "epoch": 1.6213850837138508,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.481949305868421e-07,
+      "loss": 1.40504837,
+      "memory(GiB)": 123.79,
+      "step": 63915,
+      "train_speed(iter/s)": 1.31656
+    },
+    {
+      "acc": 0.67036009,
+      "epoch": 1.621511922881786,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.475805986765479e-07,
+      "loss": 1.48103762,
+      "memory(GiB)": 123.79,
+      "step": 63920,
+      "train_speed(iter/s)": 1.316573
+    },
+    {
+      "acc": 0.6861505,
+      "epoch": 1.621638762049721,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.469664450097515e-07,
+      "loss": 1.35582733,
+      "memory(GiB)": 123.79,
+      "step": 63925,
+      "train_speed(iter/s)": 1.316585
+    },
+    {
+      "acc": 0.69516406,
+      "epoch": 1.621765601217656,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.463524696134663e-07,
+      "loss": 1.40649433,
+      "memory(GiB)": 123.79,
+      "step": 63930,
+      "train_speed(iter/s)": 1.316598
+    },
+    {
+      "acc": 0.69028687,
+      "epoch": 1.621892440385591,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.457386725146978e-07,
+      "loss": 1.35761824,
+      "memory(GiB)": 123.79,
+      "step": 63935,
+      "train_speed(iter/s)": 1.316609
+    },
+    {
+      "acc": 0.67965317,
+      "epoch": 1.622019279553526,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.451250537404433e-07,
+      "loss": 1.42714663,
+      "memory(GiB)": 123.79,
+      "step": 63940,
+      "train_speed(iter/s)": 1.31662
+    },
+    {
+      "acc": 0.68199396,
+      "epoch": 1.6221461187214612,
+      "grad_norm": 2.75,
+      "learning_rate": 9.44511613317693e-07,
+      "loss": 1.39931946,
+      "memory(GiB)": 123.79,
+      "step": 63945,
+      "train_speed(iter/s)": 1.316632
+    },
+    {
+      "acc": 0.68531199,
+      "epoch": 1.6222729578893964,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.438983512734279e-07,
+      "loss": 1.39378357,
+      "memory(GiB)": 123.79,
+      "step": 63950,
+      "train_speed(iter/s)": 1.316644
+    },
+    {
+      "acc": 0.69970617,
+      "epoch": 1.6223997970573314,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.432852676346233e-07,
+      "loss": 1.35362549,
+      "memory(GiB)": 123.79,
+      "step": 63955,
+      "train_speed(iter/s)": 1.316656
+    },
+    {
+      "acc": 0.681604,
+      "epoch": 1.6225266362252664,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.426723624282436e-07,
+      "loss": 1.41467876,
+      "memory(GiB)": 123.79,
+      "step": 63960,
+      "train_speed(iter/s)": 1.316668
+    },
+    {
+      "acc": 0.69203553,
+      "epoch": 1.6226534753932014,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.420596356812473e-07,
+      "loss": 1.38363457,
+      "memory(GiB)": 123.79,
+      "step": 63965,
+      "train_speed(iter/s)": 1.31668
+    },
+    {
+      "acc": 0.69327559,
+      "epoch": 1.6227803145611364,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.414470874205883e-07,
+      "loss": 1.3102807,
+      "memory(GiB)": 123.79,
+      "step": 63970,
+      "train_speed(iter/s)": 1.316692
+    },
+    {
+      "acc": 0.68723869,
+      "epoch": 1.6229071537290716,
+      "grad_norm": 3.875,
+      "learning_rate": 9.408347176732053e-07,
+      "loss": 1.39461098,
+      "memory(GiB)": 123.79,
+      "step": 63975,
+      "train_speed(iter/s)": 1.316704
+    },
+    {
+      "acc": 0.70662842,
+      "epoch": 1.6230339928970066,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.402225264660336e-07,
+      "loss": 1.29774914,
+      "memory(GiB)": 123.79,
+      "step": 63980,
+      "train_speed(iter/s)": 1.316717
+    },
+    {
+      "acc": 0.70343313,
+      "epoch": 1.6231608320649418,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.396105138259997e-07,
+      "loss": 1.32893,
+      "memory(GiB)": 123.79,
+      "step": 63985,
+      "train_speed(iter/s)": 1.316729
+    },
+    {
+      "acc": 0.69149971,
+      "epoch": 1.6232876712328768,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.389986797800271e-07,
+      "loss": 1.35515289,
+      "memory(GiB)": 123.79,
+      "step": 63990,
+      "train_speed(iter/s)": 1.316742
+    },
+    {
+      "acc": 0.68718462,
+      "epoch": 1.6234145104008117,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.383870243550214e-07,
+      "loss": 1.38487062,
+      "memory(GiB)": 123.79,
+      "step": 63995,
+      "train_speed(iter/s)": 1.316754
+    },
+    {
+      "acc": 0.68943152,
+      "epoch": 1.6235413495687467,
+      "grad_norm": 2.75,
+      "learning_rate": 9.377755475778877e-07,
+      "loss": 1.37657585,
+      "memory(GiB)": 123.79,
+      "step": 64000,
+      "train_speed(iter/s)": 1.316767
+    },
+    {
+      "epoch": 1.6235413495687467,
+      "eval_acc": 0.6750808696541974,
+      "eval_loss": 1.356166124343872,
+      "eval_runtime": 69.7129,
+      "eval_samples_per_second": 91.375,
+      "eval_steps_per_second": 22.851,
+      "step": 64000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 78840,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.616016151355326e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}