{
  "best_metric": 67.90186556586686,
  "best_model_checkpoint": "/home/jcanete/ft-data/all_results/mlqa/albeto_base_6/epochs_3_bs_16_lr_5e-5/checkpoint-9900",
  "epoch": 3.0,
  "global_step": 15387,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.06,
      "eval_exact_match": 32.0,
      "eval_f1": 54.977031783166986,
      "step": 300
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.838500032494963e-05,
      "loss": 2.8365,
      "step": 500
    },
    {
      "epoch": 0.12,
      "eval_exact_match": 33.8,
      "eval_f1": 57.59979011126237,
      "step": 600
    },
    {
      "epoch": 0.18,
      "eval_exact_match": 35.8,
      "eval_f1": 59.636197819947256,
      "step": 900
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.676025216091506e-05,
      "loss": 2.25,
      "step": 1000
    },
    {
      "epoch": 0.23,
      "eval_exact_match": 37.6,
      "eval_f1": 61.926773987976595,
      "step": 1200
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.5135503996880485e-05,
      "loss": 2.0708,
      "step": 1500
    },
    {
      "epoch": 0.29,
      "eval_exact_match": 35.0,
      "eval_f1": 60.867031634641314,
      "step": 1500
    },
    {
      "epoch": 0.35,
      "eval_exact_match": 36.2,
      "eval_f1": 62.98888755416137,
      "step": 1800
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.351075583284591e-05,
      "loss": 2.0327,
      "step": 2000
    },
    {
      "epoch": 0.41,
      "eval_exact_match": 38.2,
      "eval_f1": 63.263011526730395,
      "step": 2100
    },
    {
      "epoch": 0.47,
      "eval_exact_match": 39.6,
      "eval_f1": 64.57109992553288,
      "step": 2400
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.188600766881134e-05,
      "loss": 1.9952,
      "step": 2500
    },
    {
      "epoch": 0.53,
      "eval_exact_match": 38.8,
      "eval_f1": 64.11157926749733,
      "step": 2700
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.0261259504776764e-05,
      "loss": 2.0014,
      "step": 3000
    },
    {
      "epoch": 0.58,
      "eval_exact_match": 38.2,
      "eval_f1": 62.64970867760217,
      "step": 3000
    },
    {
      "epoch": 0.64,
      "eval_exact_match": 39.4,
      "eval_f1": 64.5734245071194,
      "step": 3300
    },
    {
      "epoch": 0.68,
      "learning_rate": 3.863651134074219e-05,
      "loss": 1.9401,
      "step": 3500
    },
    {
      "epoch": 0.7,
      "eval_exact_match": 39.4,
      "eval_f1": 63.6270145144117,
      "step": 3600
    },
    {
      "epoch": 0.76,
      "eval_exact_match": 39.4,
      "eval_f1": 64.21468207166183,
      "step": 3900
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.7011763176707616e-05,
      "loss": 1.9278,
      "step": 4000
    },
    {
      "epoch": 0.82,
      "eval_exact_match": 41.2,
      "eval_f1": 64.91136403427763,
      "step": 4200
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.5387015012673035e-05,
      "loss": 1.9042,
      "step": 4500
    },
    {
      "epoch": 0.88,
      "eval_exact_match": 40.6,
      "eval_f1": 64.2668781340533,
      "step": 4500
    },
    {
      "epoch": 0.94,
      "eval_exact_match": 42.0,
      "eval_f1": 65.84699832131481,
      "step": 4800
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.376226684863846e-05,
      "loss": 1.8852,
      "step": 5000
    },
    {
      "epoch": 0.99,
      "eval_exact_match": 42.8,
      "eval_f1": 66.14088379155932,
      "step": 5100
    },
    {
      "epoch": 1.05,
      "eval_exact_match": 41.2,
      "eval_f1": 65.89796053038226,
      "step": 5400
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.214076818093196e-05,
      "loss": 1.6107,
      "step": 5500
    },
    {
      "epoch": 1.11,
      "eval_exact_match": 39.2,
      "eval_f1": 64.49006562679027,
      "step": 5700
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.0516020016897384e-05,
      "loss": 1.5557,
      "step": 6000
    },
    {
      "epoch": 1.17,
      "eval_exact_match": 37.4,
      "eval_f1": 63.37951894001839,
      "step": 6000
    },
    {
      "epoch": 1.23,
      "eval_exact_match": 41.8,
      "eval_f1": 66.93944871042879,
      "step": 6300
    },
    {
      "epoch": 1.27,
      "learning_rate": 2.8891271852862807e-05,
      "loss": 1.6007,
      "step": 6500
    },
    {
      "epoch": 1.29,
      "eval_exact_match": 38.4,
      "eval_f1": 64.50298680440969,
      "step": 6600
    },
    {
      "epoch": 1.35,
      "eval_exact_match": 39.0,
      "eval_f1": 64.22154253592934,
      "step": 6900
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.7266523688828233e-05,
      "loss": 1.5494,
      "step": 7000
    },
    {
      "epoch": 1.4,
      "eval_exact_match": 38.8,
      "eval_f1": 64.59572952128524,
      "step": 7200
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.564177552479366e-05,
      "loss": 1.5433,
      "step": 7500
    },
    {
      "epoch": 1.46,
      "eval_exact_match": 40.4,
      "eval_f1": 66.29256222363962,
      "step": 7500
    },
    {
      "epoch": 1.52,
      "eval_exact_match": 40.4,
      "eval_f1": 65.90722682669741,
      "step": 7800
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4017027360759085e-05,
      "loss": 1.5668,
      "step": 8000
    },
    {
      "epoch": 1.58,
      "eval_exact_match": 39.4,
      "eval_f1": 64.99196475468266,
      "step": 8100
    },
    {
      "epoch": 1.64,
      "eval_exact_match": 39.4,
      "eval_f1": 65.07368243702982,
      "step": 8400
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.239552869305258e-05,
      "loss": 1.5129,
      "step": 8500
    },
    {
      "epoch": 1.7,
      "eval_exact_match": 40.0,
      "eval_f1": 65.83329842597489,
      "step": 8700
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.0770780529018e-05,
      "loss": 1.5471,
      "step": 9000
    },
    {
      "epoch": 1.75,
      "eval_exact_match": 40.4,
      "eval_f1": 65.76191196746076,
      "step": 9000
    },
    {
      "epoch": 1.81,
      "eval_exact_match": 40.6,
      "eval_f1": 65.21251002406223,
      "step": 9300
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.914603236498343e-05,
      "loss": 1.5719,
      "step": 9500
    },
    {
      "epoch": 1.87,
      "eval_exact_match": 41.8,
      "eval_f1": 67.1975316426952,
      "step": 9600
    },
    {
      "epoch": 1.93,
      "eval_exact_match": 42.2,
      "eval_f1": 67.90186556586686,
      "step": 9900
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.7521284200948854e-05,
      "loss": 1.5268,
      "step": 10000
    },
    {
      "epoch": 1.99,
      "eval_exact_match": 41.4,
      "eval_f1": 67.21011959811429,
      "step": 10200
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5896536036914276e-05,
      "loss": 1.4043,
      "step": 10500
    },
    {
      "epoch": 2.05,
      "eval_exact_match": 39.4,
      "eval_f1": 65.42055520162825,
      "step": 10500
    },
    {
      "epoch": 2.11,
      "eval_exact_match": 40.6,
      "eval_f1": 65.79304588629275,
      "step": 10800
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.4271787872879704e-05,
      "loss": 1.2212,
      "step": 11000
    },
    {
      "epoch": 2.16,
      "eval_exact_match": 40.2,
      "eval_f1": 65.87095054507822,
      "step": 11100
    },
    {
      "epoch": 2.22,
      "eval_exact_match": 39.8,
      "eval_f1": 65.21785295961195,
      "step": 11400
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.2647039708845129e-05,
      "loss": 1.213,
      "step": 11500
    },
    {
      "epoch": 2.28,
      "eval_exact_match": 38.4,
      "eval_f1": 64.32482818227996,
      "step": 11700
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1025541041138624e-05,
      "loss": 1.1867,
      "step": 12000
    },
    {
      "epoch": 2.34,
      "eval_exact_match": 39.0,
      "eval_f1": 65.6968030914506,
      "step": 12000
    },
    {
      "epoch": 2.4,
      "eval_exact_match": 38.6,
      "eval_f1": 65.52251419255016,
      "step": 12300
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.40079287710405e-06,
      "loss": 1.1952,
      "step": 12500
    },
    {
      "epoch": 2.46,
      "eval_exact_match": 40.4,
      "eval_f1": 66.1063359741567,
      "step": 12600
    },
    {
      "epoch": 2.52,
      "eval_exact_match": 39.6,
      "eval_f1": 66.08011102694807,
      "step": 12900
    },
    {
      "epoch": 2.53,
      "learning_rate": 7.776044713069474e-06,
      "loss": 1.1853,
      "step": 13000
    },
    {
      "epoch": 2.57,
      "eval_exact_match": 40.4,
      "eval_f1": 65.58084028972286,
      "step": 13200
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.151296549034899e-06,
      "loss": 1.1973,
      "step": 13500
    },
    {
      "epoch": 2.63,
      "eval_exact_match": 40.0,
      "eval_f1": 65.89122550684388,
      "step": 13500
    },
    {
      "epoch": 2.69,
      "eval_exact_match": 41.4,
      "eval_f1": 66.5611335718005,
      "step": 13800
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.5265483850003255e-06,
      "loss": 1.2114,
      "step": 14000
    },
    {
      "epoch": 2.75,
      "eval_exact_match": 40.4,
      "eval_f1": 66.23708712914632,
      "step": 14100
    },
    {
      "epoch": 2.81,
      "eval_exact_match": 40.4,
      "eval_f1": 65.96705612594884,
      "step": 14400
    },
    {
      "epoch": 2.83,
      "learning_rate": 2.9018002209657504e-06,
      "loss": 1.1841,
      "step": 14500
    },
    {
      "epoch": 2.87,
      "eval_exact_match": 39.4,
      "eval_f1": 65.53914611143573,
      "step": 14700
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.2770520569311757e-06,
      "loss": 1.196,
      "step": 15000
    },
    {
      "epoch": 2.92,
      "eval_exact_match": 39.6,
      "eval_f1": 65.79040459151774,
      "step": 15000
    },
    {
      "epoch": 2.98,
      "eval_exact_match": 39.6,
      "eval_f1": 65.6171459547297,
      "step": 15300
    },
    {
      "epoch": 3.0,
      "step": 15387,
      "total_flos": 3762263988836640.0,
      "train_loss": 1.6093012457153615,
      "train_runtime": 660.22,
      "train_samples_per_second": 372.835,
      "train_steps_per_second": 23.306
    }
  ],
  "max_steps": 15387,
  "num_train_epochs": 3,
  "total_flos": 3762263988836640.0,
  "trial_name": null,
  "trial_params": null
}