{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.013681462140992, "eval_steps": 500, "global_step": 29111, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6e-06, "loss": 0.3428, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2e-05, "loss": 0.9194, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-05, "loss": 0.7425, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4e-05, "loss": 1.0582, "step": 4 }, { "epoch": 0.0, "learning_rate": 3e-05, "loss": 0.6779, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-05, "loss": 0.5529, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2e-05, "loss": 1.083, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.8e-05, "loss": 1.3991, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.4000000000000005e-05, "loss": 0.3849, "step": 9 }, { "epoch": 0.0, "learning_rate": 6e-05, "loss": 0.7452, "step": 10 }, { "epoch": 0.0, "learning_rate": 5.999999989902351e-05, "loss": 0.2711, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.999999959609404e-05, "loss": 0.5467, "step": 12 }, { "epoch": 0.0, "learning_rate": 5.9999999091211596e-05, "loss": 0.4036, "step": 13 }, { "epoch": 0.0, "learning_rate": 5.999999838437617e-05, "loss": 0.8199, "step": 14 }, { "epoch": 0.0, "learning_rate": 5.9999997475587777e-05, "loss": 0.7555, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.999999636484642e-05, "loss": 0.6472, "step": 16 }, { "epoch": 0.0, "learning_rate": 5.999999505215211e-05, "loss": 0.2303, "step": 17 }, { "epoch": 0.0, "learning_rate": 5.999999353750485e-05, "loss": 1.2105, "step": 18 }, { "epoch": 0.0, "learning_rate": 5.9999991820904654e-05, "loss": 0.4096, "step": 19 }, { "epoch": 0.0, "learning_rate": 5.9999989902351535e-05, "loss": 0.6439, "step": 20 }, { "epoch": 0.0, "learning_rate": 5.99999877818455e-05, "loss": 0.8879, "step": 21 }, { "epoch": 0.0, "learning_rate": 5.999998545938657e-05, "loss": 0.919, "step": 22 }, { "epoch": 0.0, "learning_rate": 5.999998293497475e-05, "loss": 0.7333, "step": 23 }, { "epoch": 0.0, "learning_rate": 5.999998020861007e-05, "loss": 0.5049, "step": 24 }, { "epoch": 0.0, "learning_rate": 5.999997728029254e-05, "loss": 0.4545, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.999997415002219e-05, "loss": 0.7481, "step": 26 }, { "epoch": 0.0, "learning_rate": 5.999997081779903e-05, "loss": 1.1394, "step": 27 }, { "epoch": 0.0, "learning_rate": 5.999996728362307e-05, "loss": 0.3066, "step": 28 }, { "epoch": 0.0, "learning_rate": 5.9999963547494376e-05, "loss": 0.3438, "step": 29 }, { "epoch": 0.0, "learning_rate": 5.999995960941293e-05, "loss": 0.8898, "step": 30 }, { "epoch": 0.0, "learning_rate": 5.9999955469378786e-05, "loss": 0.484, "step": 31 }, { "epoch": 0.0, "learning_rate": 5.999995112739195e-05, "loss": 0.5094, "step": 32 }, { "epoch": 0.0, "learning_rate": 5.999994658345247e-05, "loss": 0.7153, "step": 33 }, { "epoch": 0.0, "learning_rate": 5.999994183756036e-05, "loss": 0.8045, "step": 34 }, { "epoch": 0.0, "learning_rate": 5.999993688971567e-05, "loss": 0.7357, "step": 35 }, { "epoch": 0.0, "learning_rate": 5.999993173991842e-05, "loss": 0.3758, "step": 36 }, { "epoch": 0.0, "learning_rate": 5.999992638816865e-05, "loss": 0.3139, "step": 37 }, { "epoch": 0.0, "learning_rate": 5.9999920834466396e-05, "loss": 0.7652, "step": 38 }, { "epoch": 0.0, "learning_rate": 5.9999915078811695e-05, "loss": 0.5999, "step": 39 }, { "epoch": 0.0, "learning_rate": 5.999990912120458e-05, "loss": 0.7926, "step": 40 }, { "epoch": 0.0, "learning_rate": 5.9999902961645104e-05, "loss": 0.7055, "step": 41 }, { "epoch": 0.0, "learning_rate": 5.9999896600133296e-05, "loss": 0.8781, "step": 42 }, { "epoch": 0.0, "learning_rate": 5.99998900366692e-05, "loss": 0.712, "step": 43 }, { "epoch": 0.0, "learning_rate": 5.9999883271252863e-05, "loss": 0.5603, "step": 44 }, { "epoch": 0.0, "learning_rate": 5.999987630388435e-05, "loss": 0.0147, "step": 45 }, { "epoch": 0.0, "learning_rate": 5.999986913456367e-05, "loss": 0.6804, "step": 46 }, { "epoch": 0.0, "learning_rate": 5.9999861763290894e-05, "loss": 0.6314, "step": 47 }, { "epoch": 0.01, "learning_rate": 5.999985419006607e-05, "loss": 0.4011, "step": 48 }, { "epoch": 0.01, "learning_rate": 5.999984641488925e-05, "loss": 0.619, "step": 49 }, { "epoch": 0.01, "learning_rate": 5.9999838437760476e-05, "loss": 0.3415, "step": 50 }, { "epoch": 0.01, "learning_rate": 5.999983025867981e-05, "loss": 0.5493, "step": 51 }, { "epoch": 0.01, "learning_rate": 5.9999821877647305e-05, "loss": 1.3561, "step": 52 }, { "epoch": 0.01, "learning_rate": 5.999981329466303e-05, "loss": 0.9088, "step": 53 }, { "epoch": 0.01, "learning_rate": 5.999980450972702e-05, "loss": 0.5093, "step": 54 }, { "epoch": 0.01, "learning_rate": 5.9999795522839354e-05, "loss": 1.1372, "step": 55 }, { "epoch": 0.01, "learning_rate": 5.999978633400008e-05, "loss": 0.4116, "step": 56 }, { "epoch": 0.01, "learning_rate": 5.999977694320926e-05, "loss": 1.0208, "step": 57 }, { "epoch": 0.01, "learning_rate": 5.999976735046696e-05, "loss": 0.9734, "step": 58 }, { "epoch": 0.01, "learning_rate": 5.999975755577325e-05, "loss": 0.7556, "step": 59 }, { "epoch": 0.01, "learning_rate": 5.999974755912819e-05, "loss": 0.2636, "step": 60 }, { "epoch": 0.01, "learning_rate": 5.999973736053185e-05, "loss": 0.7675, "step": 61 }, { "epoch": 0.01, "learning_rate": 5.99997269599843e-05, "loss": 1.5695, "step": 62 }, { "epoch": 0.01, "learning_rate": 5.9999716357485605e-05, "loss": 0.3389, "step": 63 }, { "epoch": 0.01, "learning_rate": 5.999970555303584e-05, "loss": 1.0302, "step": 64 }, { "epoch": 0.01, "learning_rate": 5.999969454663508e-05, "loss": 0.7148, "step": 65 }, { "epoch": 0.01, "learning_rate": 5.9999683338283385e-05, "loss": 0.4888, "step": 66 }, { "epoch": 0.01, "learning_rate": 5.999967192798085e-05, "loss": 0.9079, "step": 67 }, { "epoch": 0.01, "learning_rate": 5.9999660315727545e-05, "loss": 1.1344, "step": 68 }, { "epoch": 0.01, "learning_rate": 5.999964850152354e-05, "loss": 1.4716, "step": 69 }, { "epoch": 0.01, "learning_rate": 5.999963648536892e-05, "loss": 0.7595, "step": 70 }, { "epoch": 0.01, "learning_rate": 5.999962426726377e-05, "loss": 0.6932, "step": 71 }, { "epoch": 0.01, "learning_rate": 5.999961184720816e-05, "loss": 0.5572, "step": 72 }, { "epoch": 0.01, "learning_rate": 5.999959922520219e-05, "loss": 0.6373, "step": 73 }, { "epoch": 0.01, "learning_rate": 5.999958640124594e-05, "loss": 0.6582, "step": 74 }, { "epoch": 0.01, "learning_rate": 5.999957337533949e-05, "loss": 0.2256, "step": 75 }, { "epoch": 0.01, "learning_rate": 5.9999560147482935e-05, "loss": 0.2942, "step": 76 }, { "epoch": 0.01, "learning_rate": 5.999954671767635e-05, "loss": 0.9425, "step": 77 }, { "epoch": 0.01, "learning_rate": 5.999953308591985e-05, "loss": 0.4731, "step": 78 }, { "epoch": 0.01, "learning_rate": 5.99995192522135e-05, "loss": 1.0924, "step": 79 }, { "epoch": 0.01, "learning_rate": 5.999950521655741e-05, "loss": 0.5215, "step": 80 }, { "epoch": 0.01, "learning_rate": 5.9999490978951674e-05, "loss": 0.7412, "step": 81 }, { "epoch": 0.01, "learning_rate": 5.999947653939638e-05, "loss": 0.2692, "step": 82 }, { "epoch": 0.01, "learning_rate": 5.999946189789164e-05, "loss": 0.8371, "step": 83 }, { "epoch": 0.01, "learning_rate": 5.999944705443753e-05, "loss": 0.4848, "step": 84 }, { "epoch": 0.01, "learning_rate": 5.999943200903417e-05, "loss": 0.916, "step": 85 }, { "epoch": 0.01, "learning_rate": 5.999941676168164e-05, "loss": 1.6293, "step": 86 }, { "epoch": 0.01, "learning_rate": 5.999940131238007e-05, "loss": 0.6239, "step": 87 }, { "epoch": 0.01, "learning_rate": 5.9999385661129537e-05, "loss": 0.7868, "step": 88 }, { "epoch": 0.01, "learning_rate": 5.999936980793017e-05, "loss": 0.8465, "step": 89 }, { "epoch": 0.01, "learning_rate": 5.999935375278206e-05, "loss": 0.4772, "step": 90 }, { "epoch": 0.01, "learning_rate": 5.9999337495685315e-05, "loss": 0.9464, "step": 91 }, { "epoch": 0.01, "learning_rate": 5.9999321036640054e-05, "loss": 0.9297, "step": 92 }, { "epoch": 0.01, "learning_rate": 5.9999304375646384e-05, "loss": 0.3442, "step": 93 }, { "epoch": 0.01, "learning_rate": 5.999928751270441e-05, "loss": 0.5771, "step": 94 }, { "epoch": 0.01, "learning_rate": 5.9999270447814255e-05, "loss": 0.1809, "step": 95 }, { "epoch": 0.01, "learning_rate": 5.999925318097604e-05, "loss": 0.4393, "step": 96 }, { "epoch": 0.01, "learning_rate": 5.9999235712189865e-05, "loss": 0.4695, "step": 97 }, { "epoch": 0.01, "learning_rate": 5.9999218041455855e-05, "loss": 0.1688, "step": 98 }, { "epoch": 0.01, "learning_rate": 5.999920016877413e-05, "loss": 0.4223, "step": 99 }, { "epoch": 0.01, "learning_rate": 5.9999182094144795e-05, "loss": 1.2611, "step": 100 }, { "epoch": 0.01, "learning_rate": 5.9999163817568e-05, "loss": 0.7294, "step": 101 }, { "epoch": 0.01, "learning_rate": 5.9999145339043846e-05, "loss": 0.572, "step": 102 }, { "epoch": 0.01, "learning_rate": 5.999912665857247e-05, "loss": 0.5104, "step": 103 }, { "epoch": 0.01, "learning_rate": 5.9999107776153995e-05, "loss": 1.2855, "step": 104 }, { "epoch": 0.01, "learning_rate": 5.999908869178854e-05, "loss": 0.7583, "step": 105 }, { "epoch": 0.01, "learning_rate": 5.999906940547624e-05, "loss": 0.5323, "step": 106 }, { "epoch": 0.01, "learning_rate": 5.999904991721723e-05, "loss": 0.8967, "step": 107 }, { "epoch": 0.01, "learning_rate": 5.999903022701162e-05, "loss": 0.7325, "step": 108 }, { "epoch": 0.01, "learning_rate": 5.9999010334859575e-05, "loss": 1.0264, "step": 109 }, { "epoch": 0.01, "learning_rate": 5.9998990240761206e-05, "loss": 0.6951, "step": 110 }, { "epoch": 0.01, "learning_rate": 5.999896994471665e-05, "loss": 0.9803, "step": 111 }, { "epoch": 0.01, "learning_rate": 5.999894944672606e-05, "loss": 1.0065, "step": 112 }, { "epoch": 0.01, "learning_rate": 5.999892874678954e-05, "loss": 1.156, "step": 113 }, { "epoch": 0.01, "learning_rate": 5.9998907844907265e-05, "loss": 1.3487, "step": 114 }, { "epoch": 0.01, "learning_rate": 5.999888674107935e-05, "loss": 0.9231, "step": 115 }, { "epoch": 0.01, "learning_rate": 5.999886543530597e-05, "loss": 0.4715, "step": 116 }, { "epoch": 0.01, "learning_rate": 5.999884392758723e-05, "loss": 1.2527, "step": 117 }, { "epoch": 0.01, "learning_rate": 5.9998822217923294e-05, "loss": 0.794, "step": 118 }, { "epoch": 0.01, "learning_rate": 5.999880030631431e-05, "loss": 0.5077, "step": 119 }, { "epoch": 0.01, "learning_rate": 5.999877819276042e-05, "loss": 0.7545, "step": 120 }, { "epoch": 0.01, "learning_rate": 5.9998755877261774e-05, "loss": 1.1326, "step": 121 }, { "epoch": 0.01, "learning_rate": 5.999873335981853e-05, "loss": 0.5292, "step": 122 }, { "epoch": 0.01, "learning_rate": 5.999871064043082e-05, "loss": 0.7974, "step": 123 }, { "epoch": 0.01, "learning_rate": 5.9998687719098815e-05, "loss": 0.2936, "step": 124 }, { "epoch": 0.01, "learning_rate": 5.9998664595822654e-05, "loss": 0.3067, "step": 125 }, { "epoch": 0.01, "learning_rate": 5.9998641270602514e-05, "loss": 1.1992, "step": 126 }, { "epoch": 0.01, "learning_rate": 5.999861774343854e-05, "loss": 0.5743, "step": 127 }, { "epoch": 0.01, "learning_rate": 5.9998594014330885e-05, "loss": 0.4679, "step": 128 }, { "epoch": 0.01, "learning_rate": 5.999857008327972e-05, "loss": 0.8141, "step": 129 }, { "epoch": 0.01, "learning_rate": 5.9998545950285194e-05, "loss": 1.6313, "step": 130 }, { "epoch": 0.01, "learning_rate": 5.999852161534748e-05, "loss": 0.7412, "step": 131 }, { "epoch": 0.01, "learning_rate": 5.999849707846674e-05, "loss": 0.3346, "step": 132 }, { "epoch": 0.01, "learning_rate": 5.9998472339643124e-05, "loss": 0.9776, "step": 133 }, { "epoch": 0.01, "learning_rate": 5.999844739887683e-05, "loss": 0.8834, "step": 134 }, { "epoch": 0.01, "learning_rate": 5.9998422256167996e-05, "loss": 0.5304, "step": 135 }, { "epoch": 0.01, "learning_rate": 5.99983969115168e-05, "loss": 0.7623, "step": 136 }, { "epoch": 0.01, "learning_rate": 5.9998371364923415e-05, "loss": 0.927, "step": 137 }, { "epoch": 0.01, "learning_rate": 5.999834561638802e-05, "loss": 0.9178, "step": 138 }, { "epoch": 0.01, "learning_rate": 5.999831966591078e-05, "loss": 0.3338, "step": 139 }, { "epoch": 0.01, "learning_rate": 5.999829351349187e-05, "loss": 0.695, "step": 140 }, { "epoch": 0.01, "learning_rate": 5.999826715913146e-05, "loss": 0.4511, "step": 141 }, { "epoch": 0.01, "learning_rate": 5.9998240602829744e-05, "loss": 0.4444, "step": 142 }, { "epoch": 0.01, "learning_rate": 5.9998213844586885e-05, "loss": 1.1885, "step": 143 }, { "epoch": 0.02, "learning_rate": 5.999818688440307e-05, "loss": 1.1767, "step": 144 }, { "epoch": 0.02, "learning_rate": 5.9998159722278484e-05, "loss": 1.2403, "step": 145 }, { "epoch": 0.02, "learning_rate": 5.999813235821331e-05, "loss": 1.2996, "step": 146 }, { "epoch": 0.02, "learning_rate": 5.999810479220772e-05, "loss": 0.8557, "step": 147 }, { "epoch": 0.02, "learning_rate": 5.9998077024261905e-05, "loss": 0.2506, "step": 148 }, { "epoch": 0.02, "learning_rate": 5.9998049054376064e-05, "loss": 0.9882, "step": 149 }, { "epoch": 0.02, "learning_rate": 5.999802088255037e-05, "loss": 1.0816, "step": 150 }, { "epoch": 0.02, "learning_rate": 5.999799250878502e-05, "loss": 0.1351, "step": 151 }, { "epoch": 0.02, "learning_rate": 5.99979639330802e-05, "loss": 0.9248, "step": 152 }, { "epoch": 0.02, "learning_rate": 5.999793515543611e-05, "loss": 1.0951, "step": 153 }, { "epoch": 0.02, "learning_rate": 5.999790617585294e-05, "loss": 0.6817, "step": 154 }, { "epoch": 0.02, "learning_rate": 5.9997876994330884e-05, "loss": 1.0633, "step": 155 }, { "epoch": 0.02, "learning_rate": 5.999784761087013e-05, "loss": 1.2909, "step": 156 }, { "epoch": 0.02, "learning_rate": 5.99978180254709e-05, "loss": 1.1981, "step": 157 }, { "epoch": 0.02, "learning_rate": 5.999778823813337e-05, "loss": 0.9717, "step": 158 }, { "epoch": 0.02, "learning_rate": 5.9997758248857746e-05, "loss": 0.5911, "step": 159 }, { "epoch": 0.02, "learning_rate": 5.999772805764424e-05, "loss": 0.8684, "step": 160 }, { "epoch": 0.02, "learning_rate": 5.999769766449304e-05, "loss": 0.4948, "step": 161 }, { "epoch": 0.02, "learning_rate": 5.999766706940436e-05, "loss": 0.5998, "step": 162 }, { "epoch": 0.02, "learning_rate": 5.999763627237841e-05, "loss": 0.7207, "step": 163 }, { "epoch": 0.02, "learning_rate": 5.999760527341538e-05, "loss": 0.5226, "step": 164 }, { "epoch": 0.02, "learning_rate": 5.999757407251549e-05, "loss": 0.9626, "step": 165 }, { "epoch": 0.02, "learning_rate": 5.9997542669678964e-05, "loss": 1.0989, "step": 166 }, { "epoch": 0.02, "learning_rate": 5.999751106490599e-05, "loss": 0.6218, "step": 167 }, { "epoch": 0.02, "learning_rate": 5.9997479258196796e-05, "loss": 0.9831, "step": 168 }, { "epoch": 0.02, "learning_rate": 5.999744724955159e-05, "loss": 0.4713, "step": 169 }, { "epoch": 0.02, "learning_rate": 5.999741503897059e-05, "loss": 0.515, "step": 170 }, { "epoch": 0.02, "learning_rate": 5.9997382626453994e-05, "loss": 0.8295, "step": 171 }, { "epoch": 0.02, "learning_rate": 5.9997350012002056e-05, "loss": 1.1058, "step": 172 }, { "epoch": 0.02, "learning_rate": 5.999731719561497e-05, "loss": 1.0303, "step": 173 }, { "epoch": 0.02, "learning_rate": 5.999728417729296e-05, "loss": 0.9091, "step": 174 }, { "epoch": 0.02, "learning_rate": 5.9997250957036254e-05, "loss": 1.0247, "step": 175 }, { "epoch": 0.02, "learning_rate": 5.999721753484508e-05, "loss": 1.4318, "step": 176 }, { "epoch": 0.02, "learning_rate": 5.999718391071965e-05, "loss": 0.3664, "step": 177 }, { "epoch": 0.02, "learning_rate": 5.9997150084660196e-05, "loss": 1.0752, "step": 178 }, { "epoch": 0.02, "learning_rate": 5.999711605666695e-05, "loss": 1.2664, "step": 179 }, { "epoch": 0.02, "learning_rate": 5.9997081826740135e-05, "loss": 0.906, "step": 180 }, { "epoch": 0.02, "learning_rate": 5.999704739487998e-05, "loss": 0.8174, "step": 181 }, { "epoch": 0.02, "learning_rate": 5.999701276108673e-05, "loss": 1.2334, "step": 182 }, { "epoch": 0.02, "learning_rate": 5.99969779253606e-05, "loss": 0.7416, "step": 183 }, { "epoch": 0.02, "learning_rate": 5.9996942887701835e-05, "loss": 0.5031, "step": 184 }, { "epoch": 0.02, "learning_rate": 5.9996907648110675e-05, "loss": 0.5598, "step": 185 }, { "epoch": 0.02, "learning_rate": 5.999687220658734e-05, "loss": 0.4041, "step": 186 }, { "epoch": 0.02, "learning_rate": 5.999683656313209e-05, "loss": 0.3481, "step": 187 }, { "epoch": 0.02, "learning_rate": 5.999680071774516e-05, "loss": 0.1282, "step": 188 }, { "epoch": 0.02, "learning_rate": 5.999676467042678e-05, "loss": 0.7511, "step": 189 }, { "epoch": 0.02, "learning_rate": 5.999672842117719e-05, "loss": 1.0231, "step": 190 }, { "epoch": 0.02, "learning_rate": 5.999669196999665e-05, "loss": 1.0167, "step": 191 }, { "epoch": 0.02, "learning_rate": 5.99966553168854e-05, "loss": 1.1536, "step": 192 }, { "epoch": 0.02, "learning_rate": 5.999661846184369e-05, "loss": 0.1816, "step": 193 }, { "epoch": 0.02, "learning_rate": 5.9996581404871756e-05, "loss": 0.9936, "step": 194 }, { "epoch": 0.02, "learning_rate": 5.999654414596986e-05, "loss": 0.8732, "step": 195 }, { "epoch": 0.02, "learning_rate": 5.9996506685138236e-05, "loss": 0.55, "step": 196 }, { "epoch": 0.02, "learning_rate": 5.9996469022377156e-05, "loss": 0.4215, "step": 197 }, { "epoch": 0.02, "learning_rate": 5.9996431157686864e-05, "loss": 1.5371, "step": 198 }, { "epoch": 0.02, "learning_rate": 5.999639309106762e-05, "loss": 1.0902, "step": 199 }, { "epoch": 0.02, "learning_rate": 5.999635482251967e-05, "loss": 0.3736, "step": 200 }, { "epoch": 0.02, "learning_rate": 5.999631635204328e-05, "loss": 0.6922, "step": 201 }, { "epoch": 0.02, "learning_rate": 5.999627767963871e-05, "loss": 0.4306, "step": 202 }, { "epoch": 0.02, "learning_rate": 5.999623880530621e-05, "loss": 1.3514, "step": 203 }, { "epoch": 0.02, "learning_rate": 5.999619972904605e-05, "loss": 0.5497, "step": 204 }, { "epoch": 0.02, "learning_rate": 5.9996160450858505e-05, "loss": 1.104, "step": 205 }, { "epoch": 0.02, "learning_rate": 5.999612097074381e-05, "loss": 0.5473, "step": 206 }, { "epoch": 0.02, "learning_rate": 5.999608128870226e-05, "loss": 0.8997, "step": 207 }, { "epoch": 0.02, "learning_rate": 5.99960414047341e-05, "loss": 0.4338, "step": 208 }, { "epoch": 0.02, "learning_rate": 5.9996001318839616e-05, "loss": 0.6747, "step": 209 }, { "epoch": 0.02, "learning_rate": 5.999596103101906e-05, "loss": 1.0087, "step": 210 }, { "epoch": 0.02, "learning_rate": 5.9995920541272716e-05, "loss": 1.013, "step": 211 }, { "epoch": 0.02, "learning_rate": 5.9995879849600855e-05, "loss": 0.495, "step": 212 }, { "epoch": 0.02, "learning_rate": 5.999583895600375e-05, "loss": 0.5989, "step": 213 }, { "epoch": 0.02, "learning_rate": 5.999579786048168e-05, "loss": 0.9684, "step": 214 }, { "epoch": 0.02, "learning_rate": 5.999575656303491e-05, "loss": 0.831, "step": 215 }, { "epoch": 0.02, "learning_rate": 5.999571506366373e-05, "loss": 0.6292, "step": 216 }, { "epoch": 0.02, "learning_rate": 5.9995673362368405e-05, "loss": 0.6174, "step": 217 }, { "epoch": 0.02, "learning_rate": 5.999563145914923e-05, "loss": 0.9829, "step": 218 }, { "epoch": 0.02, "learning_rate": 5.9995589354006485e-05, "loss": 0.9108, "step": 219 }, { "epoch": 0.02, "learning_rate": 5.999554704694045e-05, "loss": 0.4234, "step": 220 }, { "epoch": 0.02, "learning_rate": 5.9995504537951414e-05, "loss": 1.0879, "step": 221 }, { "epoch": 0.02, "learning_rate": 5.999546182703966e-05, "loss": 0.4105, "step": 222 }, { "epoch": 0.02, "learning_rate": 5.999541891420547e-05, "loss": 0.6773, "step": 223 }, { "epoch": 0.02, "learning_rate": 5.999537579944914e-05, "loss": 0.293, "step": 224 }, { "epoch": 0.02, "learning_rate": 5.999533248277096e-05, "loss": 0.3079, "step": 225 }, { "epoch": 0.02, "learning_rate": 5.999528896417122e-05, "loss": 0.5513, "step": 226 }, { "epoch": 0.02, "learning_rate": 5.999524524365021e-05, "loss": 1.221, "step": 227 }, { "epoch": 0.02, "learning_rate": 5.999520132120823e-05, "loss": 0.267, "step": 228 }, { "epoch": 0.02, "learning_rate": 5.999515719684558e-05, "loss": 0.7981, "step": 229 }, { "epoch": 0.02, "learning_rate": 5.999511287056254e-05, "loss": 0.6773, "step": 230 }, { "epoch": 0.02, "learning_rate": 5.999506834235942e-05, "loss": 1.264, "step": 231 }, { "epoch": 0.02, "learning_rate": 5.999502361223652e-05, "loss": 0.7642, "step": 232 }, { "epoch": 0.02, "learning_rate": 5.999497868019414e-05, "loss": 0.5259, "step": 233 }, { "epoch": 0.02, "learning_rate": 5.999493354623258e-05, "loss": 0.8714, "step": 234 }, { "epoch": 0.02, "learning_rate": 5.9994888210352146e-05, "loss": 0.5037, "step": 235 }, { "epoch": 0.02, "learning_rate": 5.999484267255314e-05, "loss": 0.767, "step": 236 }, { "epoch": 0.02, "learning_rate": 5.999479693283587e-05, "loss": 0.7262, "step": 237 }, { "epoch": 0.02, "learning_rate": 5.999475099120065e-05, "loss": 1.0712, "step": 238 }, { "epoch": 0.02, "learning_rate": 5.999470484764779e-05, "loss": 0.6277, "step": 239 }, { "epoch": 0.03, "learning_rate": 5.9994658502177586e-05, "loss": 1.4285, "step": 240 }, { "epoch": 0.03, "learning_rate": 5.9994611954790365e-05, "loss": 0.985, "step": 241 }, { "epoch": 0.03, "learning_rate": 5.9994565205486426e-05, "loss": 0.5596, "step": 242 }, { "epoch": 0.03, "learning_rate": 5.99945182542661e-05, "loss": 0.9541, "step": 243 }, { "epoch": 0.03, "learning_rate": 5.999447110112969e-05, "loss": 0.7538, "step": 244 }, { "epoch": 0.03, "learning_rate": 5.999442374607753e-05, "loss": 1.2015, "step": 245 }, { "epoch": 0.03, "learning_rate": 5.999437618910992e-05, "loss": 0.6754, "step": 246 }, { "epoch": 0.03, "learning_rate": 5.999432843022719e-05, "loss": 0.2748, "step": 247 }, { "epoch": 0.03, "learning_rate": 5.999428046942966e-05, "loss": 0.6822, "step": 248 }, { "epoch": 0.03, "learning_rate": 5.999423230671764e-05, "loss": 0.4171, "step": 249 }, { "epoch": 0.03, "learning_rate": 5.999418394209148e-05, "loss": 1.3801, "step": 250 }, { "epoch": 0.03, "learning_rate": 5.999413537555148e-05, "loss": 1.1393, "step": 251 }, { "epoch": 0.03, "learning_rate": 5.999408660709799e-05, "loss": 0.5944, "step": 252 }, { "epoch": 0.03, "learning_rate": 5.999403763673132e-05, "loss": 0.9604, "step": 253 }, { "epoch": 0.03, "learning_rate": 5.99939884644518e-05, "loss": 0.8931, "step": 254 }, { "epoch": 0.03, "learning_rate": 5.9993939090259785e-05, "loss": 0.3442, "step": 255 }, { "epoch": 0.03, "learning_rate": 5.9993889514155584e-05, "loss": 0.9036, "step": 256 }, { "epoch": 0.03, "learning_rate": 5.999383973613953e-05, "loss": 0.7131, "step": 257 }, { "epoch": 0.03, "learning_rate": 5.999378975621197e-05, "loss": 1.1595, "step": 258 }, { "epoch": 0.03, "learning_rate": 5.999373957437323e-05, "loss": 0.4031, "step": 259 }, { "epoch": 0.03, "learning_rate": 5.999368919062366e-05, "loss": 1.2396, "step": 260 }, { "epoch": 0.03, "learning_rate": 5.9993638604963594e-05, "loss": 1.046, "step": 261 }, { "epoch": 0.03, "learning_rate": 5.999358781739337e-05, "loss": 0.4309, "step": 262 }, { "epoch": 0.03, "learning_rate": 5.9993536827913326e-05, "loss": 0.4507, "step": 263 }, { "epoch": 0.03, "learning_rate": 5.999348563652381e-05, "loss": 1.4697, "step": 264 }, { "epoch": 0.03, "learning_rate": 5.9993434243225165e-05, "loss": 0.4806, "step": 265 }, { "epoch": 0.03, "learning_rate": 5.999338264801775e-05, "loss": 1.6621, "step": 266 }, { "epoch": 0.03, "learning_rate": 5.9993330850901885e-05, "loss": 0.352, "step": 267 }, { "epoch": 0.03, "learning_rate": 5.999327885187794e-05, "loss": 1.08, "step": 268 }, { "epoch": 0.03, "learning_rate": 5.999322665094627e-05, "loss": 1.2336, "step": 269 }, { "epoch": 0.03, "learning_rate": 5.999317424810721e-05, "loss": 0.9483, "step": 270 }, { "epoch": 0.03, "learning_rate": 5.9993121643361114e-05, "loss": 0.7047, "step": 271 }, { "epoch": 0.03, "learning_rate": 5.999306883670834e-05, "loss": 0.8348, "step": 272 }, { "epoch": 0.03, "learning_rate": 5.9993015828149256e-05, "loss": 0.4195, "step": 273 }, { "epoch": 0.03, "learning_rate": 5.999296261768419e-05, "loss": 0.9519, "step": 274 }, { "epoch": 0.03, "learning_rate": 5.999290920531353e-05, "loss": 0.5794, "step": 275 }, { "epoch": 0.03, "learning_rate": 5.999285559103762e-05, "loss": 0.6986, "step": 276 }, { "epoch": 0.03, "learning_rate": 5.9992801774856825e-05, "loss": 0.6854, "step": 277 }, { "epoch": 0.03, "learning_rate": 5.9992747756771505e-05, "loss": 0.5879, "step": 278 }, { "epoch": 0.03, "learning_rate": 5.999269353678202e-05, "loss": 0.2286, "step": 279 }, { "epoch": 0.03, "learning_rate": 5.999263911488875e-05, "loss": 0.9159, "step": 280 }, { "epoch": 0.03, "learning_rate": 5.9992584491092046e-05, "loss": 0.5316, "step": 281 }, { "epoch": 0.03, "learning_rate": 5.999252966539228e-05, "loss": 1.0982, "step": 282 }, { "epoch": 0.03, "learning_rate": 5.999247463778982e-05, "loss": 1.1945, "step": 283 }, { "epoch": 0.03, "learning_rate": 5.9992419408285036e-05, "loss": 0.5067, "step": 284 }, { "epoch": 0.03, "learning_rate": 5.999236397687831e-05, "loss": 0.4761, "step": 285 }, { "epoch": 0.03, "learning_rate": 5.9992308343570014e-05, "loss": 0.85, "step": 286 }, { "epoch": 0.03, "learning_rate": 5.9992252508360505e-05, "loss": 0.6546, "step": 287 }, { "epoch": 0.03, "learning_rate": 5.999219647125016e-05, "loss": 0.5184, "step": 288 }, { "epoch": 0.03, "learning_rate": 5.9992140232239386e-05, "loss": 1.0358, "step": 289 }, { "epoch": 0.03, "learning_rate": 5.999208379132853e-05, "loss": 1.1036, "step": 290 }, { "epoch": 0.03, "learning_rate": 5.999202714851799e-05, "loss": 1.2733, "step": 291 }, { "epoch": 0.03, "learning_rate": 5.9991970303808135e-05, "loss": 0.8015, "step": 292 }, { "epoch": 0.03, "learning_rate": 5.999191325719935e-05, "loss": 0.3264, "step": 293 }, { "epoch": 0.03, "learning_rate": 5.999185600869203e-05, "loss": 0.5159, "step": 294 }, { "epoch": 0.03, "learning_rate": 5.9991798558286556e-05, "loss": 0.3189, "step": 295 }, { "epoch": 0.03, "learning_rate": 5.999174090598331e-05, "loss": 0.3523, "step": 296 }, { "epoch": 0.03, "learning_rate": 5.999168305178268e-05, "loss": 1.0226, "step": 297 }, { "epoch": 0.03, "learning_rate": 5.999162499568505e-05, "loss": 0.8553, "step": 298 }, { "epoch": 0.03, "learning_rate": 5.999156673769082e-05, "loss": 0.4872, "step": 299 }, { "epoch": 0.03, "learning_rate": 5.999150827780039e-05, "loss": 1.1313, "step": 300 }, { "epoch": 0.03, "learning_rate": 5.9991449616014134e-05, "loss": 1.1378, "step": 301 }, { "epoch": 0.03, "learning_rate": 5.999139075233246e-05, "loss": 0.7351, "step": 302 }, { "epoch": 0.03, "learning_rate": 5.999133168675576e-05, "loss": 0.537, "step": 303 }, { "epoch": 0.03, "learning_rate": 5.999127241928443e-05, "loss": 0.6359, "step": 304 }, { "epoch": 0.03, "learning_rate": 5.9991212949918884e-05, "loss": 0.3419, "step": 305 }, { "epoch": 0.03, "learning_rate": 5.9991153278659495e-05, "loss": 0.891, "step": 306 }, { "epoch": 0.03, "learning_rate": 5.9991093405506686e-05, "loss": 1.0184, "step": 307 }, { "epoch": 0.03, "learning_rate": 5.999103333046086e-05, "loss": 0.3339, "step": 308 }, { "epoch": 0.03, "learning_rate": 5.99909730535224e-05, "loss": 1.8254, "step": 309 }, { "epoch": 0.03, "learning_rate": 5.999091257469174e-05, "loss": 1.2856, "step": 310 }, { "epoch": 0.03, "learning_rate": 5.999085189396926e-05, "loss": 1.1355, "step": 311 }, { "epoch": 0.03, "learning_rate": 5.9990791011355395e-05, "loss": 0.7054, "step": 312 }, { "epoch": 0.03, "learning_rate": 5.9990729926850536e-05, "loss": 0.7617, "step": 313 }, { "epoch": 0.03, "learning_rate": 5.9990668640455104e-05, "loss": 0.8972, "step": 314 }, { "epoch": 0.03, "learning_rate": 5.99906071521695e-05, "loss": 0.5876, "step": 315 }, { "epoch": 0.03, "learning_rate": 5.999054546199416e-05, "loss": 0.5443, "step": 316 }, { "epoch": 0.03, "learning_rate": 5.9990483569929474e-05, "loss": 1.2563, "step": 317 }, { "epoch": 0.03, "learning_rate": 5.999042147597588e-05, "loss": 0.4116, "step": 318 }, { "epoch": 0.03, "learning_rate": 5.999035918013377e-05, "loss": 0.6492, "step": 319 }, { "epoch": 0.03, "learning_rate": 5.999029668240359e-05, "loss": 0.3922, "step": 320 }, { "epoch": 0.03, "learning_rate": 5.999023398278575e-05, "loss": 1.03, "step": 321 }, { "epoch": 0.03, "learning_rate": 5.999017108128067e-05, "loss": 0.3946, "step": 322 }, { "epoch": 0.03, "learning_rate": 5.9990107977888776e-05, "loss": 0.8038, "step": 323 }, { "epoch": 0.03, "learning_rate": 5.999004467261049e-05, "loss": 0.4693, "step": 324 }, { "epoch": 0.03, "learning_rate": 5.9989981165446244e-05, "loss": 0.9836, "step": 325 }, { "epoch": 0.03, "learning_rate": 5.998991745639646e-05, "loss": 0.5109, "step": 326 }, { "epoch": 0.03, "learning_rate": 5.9989853545461575e-05, "loss": 1.0253, "step": 327 }, { "epoch": 0.03, "learning_rate": 5.9989789432641996e-05, "loss": 2.1097, "step": 328 }, { "epoch": 0.03, "learning_rate": 5.9989725117938185e-05, "loss": 1.2358, "step": 329 }, { "epoch": 0.03, "learning_rate": 5.998966060135056e-05, "loss": 1.1956, "step": 330 }, { "epoch": 0.03, "learning_rate": 5.9989595882879556e-05, "loss": 0.4013, "step": 331 }, { "epoch": 0.03, "learning_rate": 5.998953096252561e-05, "loss": 0.91, "step": 332 }, { "epoch": 0.03, "learning_rate": 5.998946584028916e-05, "loss": 0.6692, "step": 333 }, { "epoch": 0.03, "learning_rate": 5.998940051617064e-05, "loss": 0.5499, "step": 334 }, { "epoch": 0.03, "learning_rate": 5.9989334990170495e-05, "loss": 0.3446, "step": 335 }, { "epoch": 0.04, "learning_rate": 5.998926926228916e-05, "loss": 1.0602, "step": 336 }, { "epoch": 0.04, "learning_rate": 5.998920333252708e-05, "loss": 0.739, "step": 337 }, { "epoch": 0.04, "learning_rate": 5.9989137200884705e-05, "loss": 0.8836, "step": 338 }, { "epoch": 0.04, "learning_rate": 5.998907086736247e-05, "loss": 0.6315, "step": 339 }, { "epoch": 0.04, "learning_rate": 5.998900433196084e-05, "loss": 0.5393, "step": 340 }, { "epoch": 0.04, "learning_rate": 5.9988937594680235e-05, "loss": 0.7461, "step": 341 }, { "epoch": 0.04, "learning_rate": 5.9988870655521124e-05, "loss": 0.6756, "step": 342 }, { "epoch": 0.04, "learning_rate": 5.9988803514483955e-05, "loss": 0.621, "step": 343 }, { "epoch": 0.04, "learning_rate": 5.9988736171569175e-05, "loss": 1.0235, "step": 344 }, { "epoch": 0.04, "learning_rate": 5.998866862677724e-05, "loss": 0.344, "step": 345 }, { "epoch": 0.04, "learning_rate": 5.9988600880108605e-05, "loss": 0.2096, "step": 346 }, { "epoch": 0.04, "learning_rate": 5.998853293156372e-05, "loss": 1.0367, "step": 347 }, { "epoch": 0.04, "learning_rate": 5.998846478114305e-05, "loss": 1.008, "step": 348 }, { "epoch": 0.04, "learning_rate": 5.998839642884707e-05, "loss": 0.7346, "step": 349 }, { "epoch": 0.04, "learning_rate": 5.9988327874676196e-05, "loss": 0.3421, "step": 350 }, { "epoch": 0.04, "learning_rate": 5.998825911863093e-05, "loss": 1.105, "step": 351 }, { "epoch": 0.04, "learning_rate": 5.998819016071172e-05, "loss": 0.3417, "step": 352 }, { "epoch": 0.04, "learning_rate": 5.9988121000919024e-05, "loss": 1.9326, "step": 353 }, { "epoch": 0.04, "learning_rate": 5.998805163925332e-05, "loss": 0.3869, "step": 354 }, { "epoch": 0.04, "learning_rate": 5.998798207571507e-05, "loss": 0.4872, "step": 355 }, { "epoch": 0.04, "learning_rate": 5.9987912310304735e-05, "loss": 1.1009, "step": 356 }, { "epoch": 0.04, "learning_rate": 5.9987842343022794e-05, "loss": 0.6555, "step": 357 }, { "epoch": 0.04, "learning_rate": 5.998777217386972e-05, "loss": 0.3413, "step": 358 }, { "epoch": 0.04, "learning_rate": 5.9987701802845974e-05, "loss": 0.5421, "step": 359 }, { "epoch": 0.04, "learning_rate": 5.9987631229952036e-05, "loss": 1.9598, "step": 360 }, { "epoch": 0.04, "learning_rate": 5.998756045518838e-05, "loss": 1.0662, "step": 361 }, { "epoch": 0.04, "learning_rate": 5.9987489478555486e-05, "loss": 0.9753, "step": 362 }, { "epoch": 0.04, "learning_rate": 5.998741830005383e-05, "loss": 1.1983, "step": 363 }, { "epoch": 0.04, "learning_rate": 5.99873469196839e-05, "loss": 0.8003, "step": 364 }, { "epoch": 0.04, "learning_rate": 5.998727533744615e-05, "loss": 1.0921, "step": 365 }, { "epoch": 0.04, "learning_rate": 5.998720355334109e-05, "loss": 0.2308, "step": 366 }, { "epoch": 0.04, "learning_rate": 5.998713156736918e-05, "loss": 0.991, "step": 367 }, { "epoch": 0.04, "learning_rate": 5.998705937953093e-05, "loss": 0.0061, "step": 368 }, { "epoch": 0.04, "learning_rate": 5.998698698982681e-05, "loss": 0.3548, "step": 369 }, { "epoch": 0.04, "learning_rate": 5.99869143982573e-05, "loss": 1.1912, "step": 370 }, { "epoch": 0.04, "learning_rate": 5.99868416048229e-05, "loss": 0.1899, "step": 371 }, { "epoch": 0.04, "learning_rate": 5.998676860952411e-05, "loss": 0.3692, "step": 372 }, { "epoch": 0.04, "learning_rate": 5.998669541236141e-05, "loss": 0.8799, "step": 373 }, { "epoch": 0.04, "learning_rate": 5.998662201333528e-05, "loss": 0.6108, "step": 374 }, { "epoch": 0.04, "learning_rate": 5.998654841244624e-05, "loss": 0.5347, "step": 375 }, { "epoch": 0.04, "learning_rate": 5.9986474609694764e-05, "loss": 0.3824, "step": 376 }, { "epoch": 0.04, "learning_rate": 5.9986400605081366e-05, "loss": 0.3544, "step": 377 }, { "epoch": 0.04, "learning_rate": 5.998632639860652e-05, "loss": 1.4927, "step": 378 }, { "epoch": 0.04, "learning_rate": 5.998625199027075e-05, "loss": 0.5229, "step": 379 }, { "epoch": 0.04, "learning_rate": 5.9986177380074555e-05, "loss": 0.9284, "step": 380 }, { "epoch": 0.04, "learning_rate": 5.9986102568018414e-05, "loss": 0.3047, "step": 381 }, { "epoch": 0.04, "learning_rate": 5.9986027554102864e-05, "loss": 1.2392, "step": 382 }, { "epoch": 0.04, "learning_rate": 5.998595233832838e-05, "loss": 1.0289, "step": 383 }, { "epoch": 0.04, "learning_rate": 5.998587692069549e-05, "loss": 0.9172, "step": 384 }, { "epoch": 0.04, "learning_rate": 5.9985801301204686e-05, "loss": 0.6495, "step": 385 }, { "epoch": 0.04, "learning_rate": 5.998572547985648e-05, "loss": 0.4089, "step": 386 }, { "epoch": 0.04, "learning_rate": 5.9985649456651404e-05, "loss": 0.9602, "step": 387 }, { "epoch": 0.04, "learning_rate": 5.998557323158994e-05, "loss": 1.0755, "step": 388 }, { "epoch": 0.04, "learning_rate": 5.998549680467261e-05, "loss": 0.5977, "step": 389 }, { "epoch": 0.04, "learning_rate": 5.998542017589993e-05, "loss": 0.9036, "step": 390 }, { "epoch": 0.04, "learning_rate": 5.998534334527242e-05, "loss": 0.8767, "step": 391 }, { "epoch": 0.04, "learning_rate": 5.99852663127906e-05, "loss": 0.6399, "step": 392 }, { "epoch": 0.04, "learning_rate": 5.998518907845498e-05, "loss": 0.855, "step": 393 }, { "epoch": 0.04, "learning_rate": 5.998511164226608e-05, "loss": 0.7737, "step": 394 }, { "epoch": 0.04, "learning_rate": 5.998503400422443e-05, "loss": 0.4498, "step": 395 }, { "epoch": 0.04, "learning_rate": 5.998495616433054e-05, "loss": 0.3, "step": 396 }, { "epoch": 0.04, "learning_rate": 5.998487812258495e-05, "loss": 1.1668, "step": 397 }, { "epoch": 0.04, "learning_rate": 5.9984799878988166e-05, "loss": 0.6029, "step": 398 }, { "epoch": 0.04, "learning_rate": 5.9984721433540735e-05, "loss": 0.895, "step": 399 }, { "epoch": 0.04, "learning_rate": 5.998464278624317e-05, "loss": 1.0736, "step": 400 }, { "epoch": 0.04, "learning_rate": 5.998456393709601e-05, "loss": 0.4014, "step": 401 }, { "epoch": 0.04, "learning_rate": 5.9984484886099785e-05, "loss": 0.6383, "step": 402 }, { "epoch": 0.04, "learning_rate": 5.9984405633255006e-05, "loss": 0.6056, "step": 403 }, { "epoch": 0.04, "learning_rate": 5.9984326178562235e-05, "loss": 0.8859, "step": 404 }, { "epoch": 0.04, "learning_rate": 5.9984246522022e-05, "loss": 0.8923, "step": 405 }, { "epoch": 0.04, "learning_rate": 5.998416666363483e-05, "loss": 0.4865, "step": 406 }, { "epoch": 0.04, "learning_rate": 5.998408660340127e-05, "loss": 0.468, "step": 407 }, { "epoch": 0.04, "learning_rate": 5.998400634132184e-05, "loss": 0.6646, "step": 408 }, { "epoch": 0.04, "learning_rate": 5.998392587739711e-05, "loss": 0.8194, "step": 409 }, { "epoch": 0.04, "learning_rate": 5.99838452116276e-05, "loss": 0.5396, "step": 410 }, { "epoch": 0.04, "learning_rate": 5.998376434401386e-05, "loss": 1.0045, "step": 411 }, { "epoch": 0.04, "learning_rate": 5.998368327455643e-05, "loss": 0.7972, "step": 412 }, { "epoch": 0.04, "learning_rate": 5.998360200325587e-05, "loss": 0.9757, "step": 413 }, { "epoch": 0.04, "learning_rate": 5.9983520530112704e-05, "loss": 0.3121, "step": 414 }, { "epoch": 0.04, "learning_rate": 5.99834388551275e-05, "loss": 1.1613, "step": 415 }, { "epoch": 0.04, "learning_rate": 5.9983356978300806e-05, "loss": 1.1013, "step": 416 }, { "epoch": 0.04, "learning_rate": 5.998327489963316e-05, "loss": 0.2945, "step": 417 }, { "epoch": 0.04, "learning_rate": 5.9983192619125136e-05, "loss": 0.8757, "step": 418 }, { "epoch": 0.04, "learning_rate": 5.998311013677726e-05, "loss": 0.6347, "step": 419 }, { "epoch": 0.04, "learning_rate": 5.998302745259011e-05, "loss": 0.6058, "step": 420 }, { "epoch": 0.04, "learning_rate": 5.9982944566564236e-05, "loss": 0.3397, "step": 421 }, { "epoch": 0.04, "learning_rate": 5.9982861478700194e-05, "loss": 0.5895, "step": 422 }, { "epoch": 0.04, "learning_rate": 5.998277818899854e-05, "loss": 0.7061, "step": 423 }, { "epoch": 0.04, "learning_rate": 5.998269469745983e-05, "loss": 0.6204, "step": 424 }, { "epoch": 0.04, "learning_rate": 5.998261100408465e-05, "loss": 0.7977, "step": 425 }, { "epoch": 0.04, "learning_rate": 5.998252710887354e-05, "loss": 1.273, "step": 426 }, { "epoch": 0.04, "learning_rate": 5.998244301182708e-05, "loss": 0.4404, "step": 427 }, { "epoch": 0.04, "learning_rate": 5.9982358712945824e-05, "loss": 1.2518, "step": 428 }, { "epoch": 0.04, "learning_rate": 5.9982274212230345e-05, "loss": 0.9857, "step": 429 }, { "epoch": 0.04, "learning_rate": 5.998218950968121e-05, "loss": 0.3846, "step": 430 }, { "epoch": 0.05, "learning_rate": 5.9982104605298994e-05, "loss": 0.5455, "step": 431 }, { "epoch": 0.05, "learning_rate": 5.9982019499084255e-05, "loss": 0.6187, "step": 432 }, { "epoch": 0.05, "learning_rate": 5.998193419103759e-05, "loss": 1.4191, "step": 433 }, { "epoch": 0.05, "learning_rate": 5.998184868115955e-05, "loss": 1.1323, "step": 434 }, { "epoch": 0.05, "learning_rate": 5.998176296945072e-05, "loss": 0.5796, "step": 435 }, { "epoch": 0.05, "learning_rate": 5.998167705591169e-05, "loss": 0.7849, "step": 436 }, { "epoch": 0.05, "learning_rate": 5.9981590940543014e-05, "loss": 0.7984, "step": 437 }, { "epoch": 0.05, "learning_rate": 5.998150462334528e-05, "loss": 0.3404, "step": 438 }, { "epoch": 0.05, "learning_rate": 5.998141810431908e-05, "loss": 0.7405, "step": 439 }, { "epoch": 0.05, "learning_rate": 5.998133138346499e-05, "loss": 0.1319, "step": 440 }, { "epoch": 0.05, "learning_rate": 5.998124446078358e-05, "loss": 0.9613, "step": 441 }, { "epoch": 0.05, "learning_rate": 5.998115733627545e-05, "loss": 0.5717, "step": 442 }, { "epoch": 0.05, "learning_rate": 5.998107000994119e-05, "loss": 1.1161, "step": 443 }, { "epoch": 0.05, "learning_rate": 5.9980982481781374e-05, "loss": 0.5189, "step": 444 }, { "epoch": 0.05, "learning_rate": 5.998089475179661e-05, "loss": 0.4439, "step": 445 }, { "epoch": 0.05, "learning_rate": 5.998080681998747e-05, "loss": 0.6431, "step": 446 }, { "epoch": 0.05, "learning_rate": 5.9980718686354554e-05, "loss": 0.3385, "step": 447 }, { "epoch": 0.05, "learning_rate": 5.998063035089846e-05, "loss": 0.6684, "step": 448 }, { "epoch": 0.05, "learning_rate": 5.998054181361977e-05, "loss": 0.9218, "step": 449 }, { "epoch": 0.05, "learning_rate": 5.998045307451909e-05, "loss": 1.2758, "step": 450 }, { "epoch": 0.05, "learning_rate": 5.998036413359702e-05, "loss": 1.377, "step": 451 }, { "epoch": 0.05, "learning_rate": 5.998027499085414e-05, "loss": 0.5702, "step": 452 }, { "epoch": 0.05, "learning_rate": 5.9980185646291076e-05, "loss": 0.2099, "step": 453 }, { "epoch": 0.05, "learning_rate": 5.998009609990842e-05, "loss": 0.6466, "step": 454 }, { "epoch": 0.05, "learning_rate": 5.998000635170676e-05, "loss": 0.7193, "step": 455 }, { "epoch": 0.05, "learning_rate": 5.9979916401686715e-05, "loss": 1.03, "step": 456 }, { "epoch": 0.05, "learning_rate": 5.99798262498489e-05, "loss": 1.2962, "step": 457 }, { "epoch": 0.05, "learning_rate": 5.99797358961939e-05, "loss": 1.0681, "step": 458 }, { "epoch": 0.05, "learning_rate": 5.9979645340722324e-05, "loss": 1.2493, "step": 459 }, { "epoch": 0.05, "learning_rate": 5.99795545834348e-05, "loss": 0.8382, "step": 460 }, { "epoch": 0.05, "learning_rate": 5.997946362433193e-05, "loss": 1.5824, "step": 461 }, { "epoch": 0.05, "learning_rate": 5.997937246341432e-05, "loss": 0.6294, "step": 462 }, { "epoch": 0.05, "learning_rate": 5.9979281100682597e-05, "loss": 0.4715, "step": 463 }, { "epoch": 0.05, "learning_rate": 5.9979189536137364e-05, "loss": 0.8954, "step": 464 }, { "epoch": 0.05, "learning_rate": 5.997909776977924e-05, "loss": 1.5538, "step": 465 }, { "epoch": 0.05, "learning_rate": 5.9979005801608844e-05, "loss": 0.6262, "step": 466 }, { "epoch": 0.05, "learning_rate": 5.9978913631626795e-05, "loss": 0.8045, "step": 467 }, { "epoch": 0.05, "learning_rate": 5.997882125983372e-05, "loss": 1.2307, "step": 468 }, { "epoch": 0.05, "learning_rate": 5.9978728686230226e-05, "loss": 0.6168, "step": 469 }, { "epoch": 0.05, "learning_rate": 5.997863591081695e-05, "loss": 0.4689, "step": 470 }, { "epoch": 0.05, "learning_rate": 5.9978542933594514e-05, "loss": 0.739, "step": 471 }, { "epoch": 0.05, "learning_rate": 5.997844975456354e-05, "loss": 0.4124, "step": 472 }, { "epoch": 0.05, "learning_rate": 5.997835637372465e-05, "loss": 0.474, "step": 473 }, { "epoch": 0.05, "learning_rate": 5.997826279107849e-05, "loss": 0.639, "step": 474 }, { "epoch": 0.05, "learning_rate": 5.997816900662567e-05, "loss": 0.4526, "step": 475 }, { "epoch": 0.05, "learning_rate": 5.997807502036684e-05, "loss": 1.409, "step": 476 }, { "epoch": 0.05, "learning_rate": 5.997798083230262e-05, "loss": 1.0146, "step": 477 }, { "epoch": 0.05, "learning_rate": 5.997788644243364e-05, "loss": 0.71, "step": 478 }, { "epoch": 0.05, "learning_rate": 5.9977791850760545e-05, "loss": 1.0949, "step": 479 }, { "epoch": 0.05, "learning_rate": 5.997769705728397e-05, "loss": 0.9251, "step": 480 }, { "epoch": 0.05, "learning_rate": 5.997760206200455e-05, "loss": 1.0343, "step": 481 }, { "epoch": 0.05, "learning_rate": 5.9977506864922936e-05, "loss": 0.882, "step": 482 }, { "epoch": 0.05, "learning_rate": 5.997741146603976e-05, "loss": 1.0937, "step": 483 }, { "epoch": 0.05, "learning_rate": 5.997731586535565e-05, "loss": 0.8769, "step": 484 }, { "epoch": 0.05, "learning_rate": 5.9977220062871285e-05, "loss": 1.21, "step": 485 }, { "epoch": 0.05, "learning_rate": 5.9977124058587274e-05, "loss": 1.0219, "step": 486 }, { "epoch": 0.05, "learning_rate": 5.997702785250428e-05, "loss": 0.1869, "step": 487 }, { "epoch": 0.05, "learning_rate": 5.997693144462295e-05, "loss": 0.8565, "step": 488 }, { "epoch": 0.05, "learning_rate": 5.9976834834943935e-05, "loss": 0.2195, "step": 489 }, { "epoch": 0.05, "learning_rate": 5.9976738023467876e-05, "loss": 0.7519, "step": 490 }, { "epoch": 0.05, "learning_rate": 5.9976641010195434e-05, "loss": 0.2338, "step": 491 }, { "epoch": 0.05, "learning_rate": 5.997654379512725e-05, "loss": 0.5426, "step": 492 }, { "epoch": 0.05, "learning_rate": 5.9976446378264e-05, "loss": 0.2581, "step": 493 }, { "epoch": 0.05, "learning_rate": 5.997634875960632e-05, "loss": 0.9477, "step": 494 }, { "epoch": 0.05, "learning_rate": 5.9976250939154875e-05, "loss": 0.882, "step": 495 }, { "epoch": 0.05, "learning_rate": 5.9976152916910325e-05, "loss": 0.7402, "step": 496 }, { "epoch": 0.05, "learning_rate": 5.997605469287333e-05, "loss": 0.7216, "step": 497 }, { "epoch": 0.05, "learning_rate": 5.997595626704454e-05, "loss": 0.6904, "step": 498 }, { "epoch": 0.05, "learning_rate": 5.9975857639424635e-05, "loss": 1.6658, "step": 499 }, { "epoch": 0.05, "learning_rate": 5.997575881001426e-05, "loss": 0.7416, "step": 500 }, { "epoch": 0.05, "learning_rate": 5.99756597788141e-05, "loss": 0.3767, "step": 501 }, { "epoch": 0.05, "learning_rate": 5.9975560545824805e-05, "loss": 0.7365, "step": 502 }, { "epoch": 0.05, "learning_rate": 5.997546111104706e-05, "loss": 0.7862, "step": 503 }, { "epoch": 0.05, "learning_rate": 5.997536147448151e-05, "loss": 1.0097, "step": 504 }, { "epoch": 0.05, "learning_rate": 5.997526163612885e-05, "loss": 0.932, "step": 505 }, { "epoch": 0.05, "learning_rate": 5.997516159598974e-05, "loss": 0.8965, "step": 506 }, { "epoch": 0.05, "learning_rate": 5.9975061354064855e-05, "loss": 0.5128, "step": 507 }, { "epoch": 0.05, "learning_rate": 5.997496091035488e-05, "loss": 1.1341, "step": 508 }, { "epoch": 0.05, "learning_rate": 5.997486026486047e-05, "loss": 0.8977, "step": 509 }, { "epoch": 0.05, "learning_rate": 5.997475941758232e-05, "loss": 0.3935, "step": 510 }, { "epoch": 0.05, "learning_rate": 5.997465836852109e-05, "loss": 1.0232, "step": 511 }, { "epoch": 0.05, "learning_rate": 5.997455711767749e-05, "loss": 0.7775, "step": 512 }, { "epoch": 0.05, "learning_rate": 5.997445566505218e-05, "loss": 1.8293, "step": 513 }, { "epoch": 0.05, "learning_rate": 5.997435401064584e-05, "loss": 0.747, "step": 514 }, { "epoch": 0.05, "learning_rate": 5.997425215445918e-05, "loss": 0.9217, "step": 515 }, { "epoch": 0.05, "learning_rate": 5.997415009649285e-05, "loss": 0.4518, "step": 516 }, { "epoch": 0.05, "learning_rate": 5.997404783674757e-05, "loss": 1.0182, "step": 517 }, { "epoch": 0.05, "learning_rate": 5.9973945375224e-05, "loss": 0.4105, "step": 518 }, { "epoch": 0.05, "learning_rate": 5.997384271192285e-05, "loss": 0.9386, "step": 519 }, { "epoch": 0.05, "learning_rate": 5.99737398468448e-05, "loss": 1.1332, "step": 520 }, { "epoch": 0.05, "learning_rate": 5.997363677999055e-05, "loss": 0.4354, "step": 521 }, { "epoch": 0.05, "learning_rate": 5.997353351136079e-05, "loss": 1.01, "step": 522 }, { "epoch": 0.05, "learning_rate": 5.997343004095622e-05, "loss": 0.3975, "step": 523 }, { "epoch": 0.05, "learning_rate": 5.997332636877752e-05, "loss": 0.8681, "step": 524 }, { "epoch": 0.05, "learning_rate": 5.997322249482542e-05, "loss": 0.6739, "step": 525 }, { "epoch": 0.05, "learning_rate": 5.997311841910058e-05, "loss": 0.4276, "step": 526 }, { "epoch": 0.06, "learning_rate": 5.9973014141603724e-05, "loss": 1.083, "step": 527 }, { "epoch": 0.06, "learning_rate": 5.997290966233556e-05, "loss": 1.1757, "step": 528 }, { "epoch": 0.06, "learning_rate": 5.997280498129677e-05, "loss": 0.5469, "step": 529 }, { "epoch": 0.06, "learning_rate": 5.9972700098488074e-05, "loss": 0.9865, "step": 530 }, { "epoch": 0.06, "learning_rate": 5.9972595013910174e-05, "loss": 0.5466, "step": 531 }, { "epoch": 0.06, "learning_rate": 5.997248972756378e-05, "loss": 0.601, "step": 532 }, { "epoch": 0.06, "learning_rate": 5.99723842394496e-05, "loss": 0.606, "step": 533 }, { "epoch": 0.06, "learning_rate": 5.9972278549568334e-05, "loss": 0.3483, "step": 534 }, { "epoch": 0.06, "learning_rate": 5.9972172657920704e-05, "loss": 0.5548, "step": 535 }, { "epoch": 0.06, "learning_rate": 5.997206656450743e-05, "loss": 1.2073, "step": 536 }, { "epoch": 0.06, "learning_rate": 5.99719602693292e-05, "loss": 0.5423, "step": 537 }, { "epoch": 0.06, "learning_rate": 5.997185377238676e-05, "loss": 0.2945, "step": 538 }, { "epoch": 0.06, "learning_rate": 5.997174707368081e-05, "loss": 0.7932, "step": 539 }, { "epoch": 0.06, "learning_rate": 5.997164017321207e-05, "loss": 1.0955, "step": 540 }, { "epoch": 0.06, "learning_rate": 5.997153307098126e-05, "loss": 1.1067, "step": 541 }, { "epoch": 0.06, "learning_rate": 5.99714257669891e-05, "loss": 0.6928, "step": 542 }, { "epoch": 0.06, "learning_rate": 5.997131826123632e-05, "loss": 0.4297, "step": 543 }, { "epoch": 0.06, "learning_rate": 5.997121055372364e-05, "loss": 0.4857, "step": 544 }, { "epoch": 0.06, "learning_rate": 5.9971102644451775e-05, "loss": 0.4765, "step": 545 }, { "epoch": 0.06, "learning_rate": 5.997099453342146e-05, "loss": 0.4615, "step": 546 }, { "epoch": 0.06, "learning_rate": 5.997088622063343e-05, "loss": 1.1683, "step": 547 }, { "epoch": 0.06, "learning_rate": 5.99707777060884e-05, "loss": 1.1343, "step": 548 }, { "epoch": 0.06, "learning_rate": 5.997066898978712e-05, "loss": 0.286, "step": 549 }, { "epoch": 0.06, "learning_rate": 5.997056007173029e-05, "loss": 0.4312, "step": 550 }, { "epoch": 0.06, "learning_rate": 5.997045095191868e-05, "loss": 0.7993, "step": 551 }, { "epoch": 0.06, "learning_rate": 5.9970341630353e-05, "loss": 1.7869, "step": 552 }, { "epoch": 0.06, "learning_rate": 5.997023210703398e-05, "loss": 0.4431, "step": 553 }, { "epoch": 0.06, "learning_rate": 5.997012238196238e-05, "loss": 0.8978, "step": 554 }, { "epoch": 0.06, "learning_rate": 5.997001245513894e-05, "loss": 0.7929, "step": 555 }, { "epoch": 0.06, "learning_rate": 5.996990232656437e-05, "loss": 1.156, "step": 556 }, { "epoch": 0.06, "learning_rate": 5.9969791996239446e-05, "loss": 0.6943, "step": 557 }, { "epoch": 0.06, "learning_rate": 5.9969681464164884e-05, "loss": 0.4012, "step": 558 }, { "epoch": 0.06, "learning_rate": 5.996957073034144e-05, "loss": 0.6729, "step": 559 }, { "epoch": 0.06, "learning_rate": 5.996945979476986e-05, "loss": 0.9499, "step": 560 }, { "epoch": 0.06, "learning_rate": 5.996934865745088e-05, "loss": 0.6687, "step": 561 }, { "epoch": 0.06, "learning_rate": 5.9969237318385275e-05, "loss": 0.7176, "step": 562 }, { "epoch": 0.06, "learning_rate": 5.996912577757377e-05, "loss": 1.2134, "step": 563 }, { "epoch": 0.06, "learning_rate": 5.9969014035017114e-05, "loss": 1.4384, "step": 564 }, { "epoch": 0.06, "learning_rate": 5.996890209071607e-05, "loss": 0.8307, "step": 565 }, { "epoch": 0.06, "learning_rate": 5.9968789944671395e-05, "loss": 1.4706, "step": 566 }, { "epoch": 0.06, "learning_rate": 5.9968677596883834e-05, "loss": 1.2827, "step": 567 }, { "epoch": 0.06, "learning_rate": 5.9968565047354154e-05, "loss": 0.3983, "step": 568 }, { "epoch": 0.06, "learning_rate": 5.99684522960831e-05, "loss": 0.4231, "step": 569 }, { "epoch": 0.06, "learning_rate": 5.996833934307144e-05, "loss": 1.0138, "step": 570 }, { "epoch": 0.06, "learning_rate": 5.996822618831993e-05, "loss": 0.3762, "step": 571 }, { "epoch": 0.06, "learning_rate": 5.996811283182933e-05, "loss": 1.4293, "step": 572 }, { "epoch": 0.06, "learning_rate": 5.9967999273600414e-05, "loss": 0.3201, "step": 573 }, { "epoch": 0.06, "learning_rate": 5.996788551363393e-05, "loss": 1.0612, "step": 574 }, { "epoch": 0.06, "learning_rate": 5.9967771551930666e-05, "loss": 0.6201, "step": 575 }, { "epoch": 0.06, "learning_rate": 5.9967657388491364e-05, "loss": 0.4934, "step": 576 }, { "epoch": 0.06, "learning_rate": 5.996754302331681e-05, "loss": 0.8028, "step": 577 }, { "epoch": 0.06, "learning_rate": 5.996742845640777e-05, "loss": 0.5591, "step": 578 }, { "epoch": 0.06, "learning_rate": 5.9967313687765e-05, "loss": 1.1613, "step": 579 }, { "epoch": 0.06, "learning_rate": 5.9967198717389295e-05, "loss": 0.4789, "step": 580 }, { "epoch": 0.06, "learning_rate": 5.996708354528143e-05, "loss": 0.6552, "step": 581 }, { "epoch": 0.06, "learning_rate": 5.996696817144215e-05, "loss": 0.8929, "step": 582 }, { "epoch": 0.06, "learning_rate": 5.996685259587227e-05, "loss": 1.154, "step": 583 }, { "epoch": 0.06, "learning_rate": 5.996673681857253e-05, "loss": 0.5988, "step": 584 }, { "epoch": 0.06, "learning_rate": 5.9966620839543744e-05, "loss": 0.3541, "step": 585 }, { "epoch": 0.06, "learning_rate": 5.996650465878668e-05, "loss": 1.793, "step": 586 }, { "epoch": 0.06, "learning_rate": 5.9966388276302116e-05, "loss": 1.1191, "step": 587 }, { "epoch": 0.06, "learning_rate": 5.996627169209083e-05, "loss": 1.161, "step": 588 }, { "epoch": 0.06, "learning_rate": 5.996615490615363e-05, "loss": 0.5323, "step": 589 }, { "epoch": 0.06, "learning_rate": 5.996603791849127e-05, "loss": 0.4398, "step": 590 }, { "epoch": 0.06, "learning_rate": 5.9965920729104564e-05, "loss": 1.013, "step": 591 }, { "epoch": 0.06, "learning_rate": 5.9965803337994284e-05, "loss": 1.4305, "step": 592 }, { "epoch": 0.06, "learning_rate": 5.996568574516124e-05, "loss": 1.0351, "step": 593 }, { "epoch": 0.06, "learning_rate": 5.9965567950606195e-05, "loss": 0.7379, "step": 594 }, { "epoch": 0.06, "learning_rate": 5.9965449954329966e-05, "loss": 0.8741, "step": 595 }, { "epoch": 0.06, "learning_rate": 5.996533175633334e-05, "loss": 0.9616, "step": 596 }, { "epoch": 0.06, "learning_rate": 5.996521335661712e-05, "loss": 0.7693, "step": 597 }, { "epoch": 0.06, "learning_rate": 5.996509475518208e-05, "loss": 0.2797, "step": 598 }, { "epoch": 0.06, "learning_rate": 5.9964975952029035e-05, "loss": 0.6428, "step": 599 }, { "epoch": 0.06, "learning_rate": 5.996485694715879e-05, "loss": 1.2993, "step": 600 }, { "epoch": 0.06, "learning_rate": 5.996473774057215e-05, "loss": 0.2208, "step": 601 }, { "epoch": 0.06, "learning_rate": 5.9964618332269884e-05, "loss": 0.6142, "step": 602 }, { "epoch": 0.06, "learning_rate": 5.996449872225283e-05, "loss": 0.6081, "step": 603 }, { "epoch": 0.06, "learning_rate": 5.996437891052178e-05, "loss": 0.904, "step": 604 }, { "epoch": 0.06, "learning_rate": 5.996425889707754e-05, "loss": 0.3673, "step": 605 }, { "epoch": 0.06, "learning_rate": 5.9964138681920925e-05, "loss": 0.2397, "step": 606 }, { "epoch": 0.06, "learning_rate": 5.996401826505273e-05, "loss": 0.3421, "step": 607 }, { "epoch": 0.06, "learning_rate": 5.996389764647378e-05, "loss": 0.6316, "step": 608 }, { "epoch": 0.06, "learning_rate": 5.996377682618488e-05, "loss": 0.5886, "step": 609 }, { "epoch": 0.06, "learning_rate": 5.996365580418686e-05, "loss": 0.8632, "step": 610 }, { "epoch": 0.06, "learning_rate": 5.996353458048051e-05, "loss": 1.1772, "step": 611 }, { "epoch": 0.06, "learning_rate": 5.996341315506665e-05, "loss": 0.584, "step": 612 }, { "epoch": 0.06, "learning_rate": 5.996329152794611e-05, "loss": 1.2539, "step": 613 }, { "epoch": 0.06, "learning_rate": 5.99631696991197e-05, "loss": 0.4281, "step": 614 }, { "epoch": 0.06, "learning_rate": 5.996304766858825e-05, "loss": 1.1455, "step": 615 }, { "epoch": 0.06, "learning_rate": 5.996292543635256e-05, "loss": 0.2108, "step": 616 }, { "epoch": 0.06, "learning_rate": 5.996280300241348e-05, "loss": 0.9472, "step": 617 }, { "epoch": 0.06, "learning_rate": 5.996268036677181e-05, "loss": 0.8337, "step": 618 }, { "epoch": 0.06, "learning_rate": 5.996255752942839e-05, "loss": 0.6818, "step": 619 }, { "epoch": 0.06, "learning_rate": 5.996243449038405e-05, "loss": 0.235, "step": 620 }, { "epoch": 0.06, "learning_rate": 5.996231124963961e-05, "loss": 1.0229, "step": 621 }, { "epoch": 0.06, "learning_rate": 5.9962187807195894e-05, "loss": 0.7713, "step": 622 }, { "epoch": 0.07, "learning_rate": 5.996206416305374e-05, "loss": 0.9145, "step": 623 }, { "epoch": 0.07, "learning_rate": 5.996194031721398e-05, "loss": 0.6899, "step": 624 }, { "epoch": 0.07, "learning_rate": 5.9961816269677465e-05, "loss": 0.3646, "step": 625 }, { "epoch": 0.07, "learning_rate": 5.9961692020445e-05, "loss": 0.7239, "step": 626 }, { "epoch": 0.07, "learning_rate": 5.996156756951744e-05, "loss": 1.041, "step": 627 }, { "epoch": 0.07, "learning_rate": 5.9961442916895615e-05, "loss": 1.2714, "step": 628 }, { "epoch": 0.07, "learning_rate": 5.996131806258036e-05, "loss": 0.3631, "step": 629 }, { "epoch": 0.07, "learning_rate": 5.996119300657252e-05, "loss": 0.4657, "step": 630 }, { "epoch": 0.07, "learning_rate": 5.996106774887295e-05, "loss": 1.2122, "step": 631 }, { "epoch": 0.07, "learning_rate": 5.996094228948249e-05, "loss": 0.9052, "step": 632 }, { "epoch": 0.07, "learning_rate": 5.996081662840196e-05, "loss": 1.5324, "step": 633 }, { "epoch": 0.07, "learning_rate": 5.996069076563224e-05, "loss": 0.2405, "step": 634 }, { "epoch": 0.07, "learning_rate": 5.9960564701174136e-05, "loss": 0.561, "step": 635 }, { "epoch": 0.07, "learning_rate": 5.996043843502853e-05, "loss": 1.1966, "step": 636 }, { "epoch": 0.07, "learning_rate": 5.9960311967196266e-05, "loss": 1.3261, "step": 637 }, { "epoch": 0.07, "learning_rate": 5.996018529767819e-05, "loss": 1.2752, "step": 638 }, { "epoch": 0.07, "learning_rate": 5.996005842647517e-05, "loss": 1.3343, "step": 639 }, { "epoch": 0.07, "learning_rate": 5.995993135358803e-05, "loss": 1.2742, "step": 640 }, { "epoch": 0.07, "learning_rate": 5.995980407901765e-05, "loss": 1.195, "step": 641 }, { "epoch": 0.07, "learning_rate": 5.995967660276488e-05, "loss": 0.4826, "step": 642 }, { "epoch": 0.07, "learning_rate": 5.995954892483057e-05, "loss": 1.1263, "step": 643 }, { "epoch": 0.07, "learning_rate": 5.9959421045215586e-05, "loss": 0.6601, "step": 644 }, { "epoch": 0.07, "learning_rate": 5.9959292963920795e-05, "loss": 1.1473, "step": 645 }, { "epoch": 0.07, "learning_rate": 5.9959164680947054e-05, "loss": 0.8689, "step": 646 }, { "epoch": 0.07, "learning_rate": 5.995903619629523e-05, "loss": 0.6946, "step": 647 }, { "epoch": 0.07, "learning_rate": 5.995890750996617e-05, "loss": 1.3487, "step": 648 }, { "epoch": 0.07, "learning_rate": 5.995877862196077e-05, "loss": 0.4838, "step": 649 }, { "epoch": 0.07, "learning_rate": 5.9958649532279874e-05, "loss": 1.3281, "step": 650 }, { "epoch": 0.07, "learning_rate": 5.995852024092436e-05, "loss": 0.4087, "step": 651 }, { "epoch": 0.07, "learning_rate": 5.99583907478951e-05, "loss": 1.0071, "step": 652 }, { "epoch": 0.07, "learning_rate": 5.995826105319296e-05, "loss": 0.5601, "step": 653 }, { "epoch": 0.07, "learning_rate": 5.9958131156818815e-05, "loss": 0.3873, "step": 654 }, { "epoch": 0.07, "learning_rate": 5.995800105877354e-05, "loss": 1.108, "step": 655 }, { "epoch": 0.07, "learning_rate": 5.995787075905802e-05, "loss": 0.5899, "step": 656 }, { "epoch": 0.07, "learning_rate": 5.995774025767312e-05, "loss": 0.4346, "step": 657 }, { "epoch": 0.07, "learning_rate": 5.995760955461971e-05, "loss": 0.6697, "step": 658 }, { "epoch": 0.07, "learning_rate": 5.9957478649898703e-05, "loss": 0.574, "step": 659 }, { "epoch": 0.07, "learning_rate": 5.995734754351094e-05, "loss": 0.4991, "step": 660 }, { "epoch": 0.07, "learning_rate": 5.995721623545734e-05, "loss": 0.3646, "step": 661 }, { "epoch": 0.07, "learning_rate": 5.9957084725738756e-05, "loss": 0.7598, "step": 662 }, { "epoch": 0.07, "learning_rate": 5.995695301435609e-05, "loss": 0.2197, "step": 663 }, { "epoch": 0.07, "learning_rate": 5.995682110131023e-05, "loss": 0.3897, "step": 664 }, { "epoch": 0.07, "learning_rate": 5.995668898660206e-05, "loss": 0.3338, "step": 665 }, { "epoch": 0.07, "learning_rate": 5.9956556670232466e-05, "loss": 1.4492, "step": 666 }, { "epoch": 0.07, "learning_rate": 5.995642415220234e-05, "loss": 1.2185, "step": 667 }, { "epoch": 0.07, "learning_rate": 5.995629143251258e-05, "loss": 0.6433, "step": 668 }, { "epoch": 0.07, "learning_rate": 5.9956158511164074e-05, "loss": 0.6464, "step": 669 }, { "epoch": 0.07, "learning_rate": 5.9956025388157715e-05, "loss": 1.2896, "step": 670 }, { "epoch": 0.07, "learning_rate": 5.995589206349441e-05, "loss": 0.8154, "step": 671 }, { "epoch": 0.07, "learning_rate": 5.9955758537175045e-05, "loss": 0.4615, "step": 672 }, { "epoch": 0.07, "learning_rate": 5.995562480920052e-05, "loss": 0.7895, "step": 673 }, { "epoch": 0.07, "learning_rate": 5.9955490879571736e-05, "loss": 0.9129, "step": 674 }, { "epoch": 0.07, "learning_rate": 5.9955356748289605e-05, "loss": 0.8347, "step": 675 }, { "epoch": 0.07, "learning_rate": 5.995522241535501e-05, "loss": 0.9951, "step": 676 }, { "epoch": 0.07, "learning_rate": 5.9955087880768876e-05, "loss": 0.8384, "step": 677 }, { "epoch": 0.07, "learning_rate": 5.9954953144532094e-05, "loss": 1.2839, "step": 678 }, { "epoch": 0.07, "learning_rate": 5.995481820664558e-05, "loss": 1.2664, "step": 679 }, { "epoch": 0.07, "learning_rate": 5.9954683067110235e-05, "loss": 0.4286, "step": 680 }, { "epoch": 0.07, "learning_rate": 5.9954547725926974e-05, "loss": 0.7224, "step": 681 }, { "epoch": 0.07, "learning_rate": 5.99544121830967e-05, "loss": 0.9842, "step": 682 }, { "epoch": 0.07, "learning_rate": 5.995427643862035e-05, "loss": 0.6544, "step": 683 }, { "epoch": 0.07, "learning_rate": 5.99541404924988e-05, "loss": 0.3335, "step": 684 }, { "epoch": 0.07, "learning_rate": 5.995400434473299e-05, "loss": 0.8708, "step": 685 }, { "epoch": 0.07, "learning_rate": 5.9953867995323844e-05, "loss": 0.5911, "step": 686 }, { "epoch": 0.07, "learning_rate": 5.995373144427226e-05, "loss": 1.0953, "step": 687 }, { "epoch": 0.07, "learning_rate": 5.995359469157916e-05, "loss": 1.1812, "step": 688 }, { "epoch": 0.07, "learning_rate": 5.9953457737245467e-05, "loss": 0.8241, "step": 689 }, { "epoch": 0.07, "learning_rate": 5.9953320581272113e-05, "loss": 0.2465, "step": 690 }, { "epoch": 0.07, "learning_rate": 5.995318322366001e-05, "loss": 0.4496, "step": 691 }, { "epoch": 0.07, "learning_rate": 5.9953045664410086e-05, "loss": 0.6209, "step": 692 }, { "epoch": 0.07, "learning_rate": 5.995290790352327e-05, "loss": 0.2847, "step": 693 }, { "epoch": 0.07, "learning_rate": 5.995276994100048e-05, "loss": 1.2336, "step": 694 }, { "epoch": 0.07, "learning_rate": 5.995263177684265e-05, "loss": 0.2569, "step": 695 }, { "epoch": 0.07, "learning_rate": 5.995249341105072e-05, "loss": 1.2337, "step": 696 }, { "epoch": 0.07, "learning_rate": 5.995235484362561e-05, "loss": 0.383, "step": 697 }, { "epoch": 0.07, "learning_rate": 5.9952216074568256e-05, "loss": 0.2668, "step": 698 }, { "epoch": 0.07, "learning_rate": 5.995207710387959e-05, "loss": 0.0938, "step": 699 }, { "epoch": 0.07, "learning_rate": 5.995193793156055e-05, "loss": 0.5206, "step": 700 }, { "epoch": 0.07, "learning_rate": 5.995179855761207e-05, "loss": 0.7391, "step": 701 }, { "epoch": 0.07, "learning_rate": 5.995165898203509e-05, "loss": 1.0072, "step": 702 }, { "epoch": 0.07, "learning_rate": 5.9951519204830546e-05, "loss": 0.4841, "step": 703 }, { "epoch": 0.07, "learning_rate": 5.995137922599939e-05, "loss": 0.679, "step": 704 }, { "epoch": 0.07, "learning_rate": 5.995123904554256e-05, "loss": 0.3976, "step": 705 }, { "epoch": 0.07, "learning_rate": 5.995109866346099e-05, "loss": 1.0456, "step": 706 }, { "epoch": 0.07, "learning_rate": 5.9950958079755634e-05, "loss": 0.2242, "step": 707 }, { "epoch": 0.07, "learning_rate": 5.9950817294427435e-05, "loss": 0.6829, "step": 708 }, { "epoch": 0.07, "learning_rate": 5.995067630747735e-05, "loss": 0.8134, "step": 709 }, { "epoch": 0.07, "learning_rate": 5.9950535118906306e-05, "loss": 0.5542, "step": 710 }, { "epoch": 0.07, "learning_rate": 5.995039372871528e-05, "loss": 1.0436, "step": 711 }, { "epoch": 0.07, "learning_rate": 5.99502521369052e-05, "loss": 0.4129, "step": 712 }, { "epoch": 0.07, "learning_rate": 5.995011034347704e-05, "loss": 0.3315, "step": 713 }, { "epoch": 0.07, "learning_rate": 5.994996834843173e-05, "loss": 0.9642, "step": 714 }, { "epoch": 0.07, "learning_rate": 5.994982615177026e-05, "loss": 0.6063, "step": 715 }, { "epoch": 0.07, "learning_rate": 5.994968375349356e-05, "loss": 0.6911, "step": 716 }, { "epoch": 0.07, "learning_rate": 5.99495411536026e-05, "loss": 0.6698, "step": 717 }, { "epoch": 0.07, "learning_rate": 5.994939835209833e-05, "loss": 0.3433, "step": 718 }, { "epoch": 0.08, "learning_rate": 5.994925534898172e-05, "loss": 1.4361, "step": 719 }, { "epoch": 0.08, "learning_rate": 5.994911214425373e-05, "loss": 0.0853, "step": 720 }, { "epoch": 0.08, "learning_rate": 5.994896873791533e-05, "loss": 0.4882, "step": 721 }, { "epoch": 0.08, "learning_rate": 5.994882512996747e-05, "loss": 0.3186, "step": 722 }, { "epoch": 0.08, "learning_rate": 5.9948681320411136e-05, "loss": 0.8752, "step": 723 }, { "epoch": 0.08, "learning_rate": 5.994853730924728e-05, "loss": 0.5673, "step": 724 }, { "epoch": 0.08, "learning_rate": 5.994839309647688e-05, "loss": 1.2337, "step": 725 }, { "epoch": 0.08, "learning_rate": 5.994824868210091e-05, "loss": 0.9549, "step": 726 }, { "epoch": 0.08, "learning_rate": 5.994810406612033e-05, "loss": 0.6704, "step": 727 }, { "epoch": 0.08, "learning_rate": 5.994795924853613e-05, "loss": 0.622, "step": 728 }, { "epoch": 0.08, "learning_rate": 5.994781422934926e-05, "loss": 1.0037, "step": 729 }, { "epoch": 0.08, "learning_rate": 5.994766900856073e-05, "loss": 0.7989, "step": 730 }, { "epoch": 0.08, "learning_rate": 5.994752358617148e-05, "loss": 1.2352, "step": 731 }, { "epoch": 0.08, "learning_rate": 5.9947377962182535e-05, "loss": 0.3756, "step": 732 }, { "epoch": 0.08, "learning_rate": 5.9947232136594825e-05, "loss": 0.6912, "step": 733 }, { "epoch": 0.08, "learning_rate": 5.9947086109409364e-05, "loss": 0.523, "step": 734 }, { "epoch": 0.08, "learning_rate": 5.9946939880627135e-05, "loss": 1.4735, "step": 735 }, { "epoch": 0.08, "learning_rate": 5.99467934502491e-05, "loss": 0.7735, "step": 736 }, { "epoch": 0.08, "learning_rate": 5.9946646818276266e-05, "loss": 0.9752, "step": 737 }, { "epoch": 0.08, "learning_rate": 5.994649998470961e-05, "loss": 0.4977, "step": 738 }, { "epoch": 0.08, "learning_rate": 5.994635294955012e-05, "loss": 0.8706, "step": 739 }, { "epoch": 0.08, "learning_rate": 5.994620571279879e-05, "loss": 0.6615, "step": 740 }, { "epoch": 0.08, "learning_rate": 5.9946058274456615e-05, "loss": 1.0197, "step": 741 }, { "epoch": 0.08, "learning_rate": 5.994591063452458e-05, "loss": 0.4933, "step": 742 }, { "epoch": 0.08, "learning_rate": 5.9945762793003684e-05, "loss": 1.1404, "step": 743 }, { "epoch": 0.08, "learning_rate": 5.994561474989491e-05, "loss": 1.271, "step": 744 }, { "epoch": 0.08, "learning_rate": 5.9945466505199274e-05, "loss": 0.8533, "step": 745 }, { "epoch": 0.08, "learning_rate": 5.994531805891776e-05, "loss": 0.7387, "step": 746 }, { "epoch": 0.08, "learning_rate": 5.994516941105137e-05, "loss": 0.5743, "step": 747 }, { "epoch": 0.08, "learning_rate": 5.994502056160111e-05, "loss": 0.5607, "step": 748 }, { "epoch": 0.08, "learning_rate": 5.994487151056797e-05, "loss": 0.8548, "step": 749 }, { "epoch": 0.08, "learning_rate": 5.994472225795297e-05, "loss": 1.6797, "step": 750 }, { "epoch": 0.08, "learning_rate": 5.994457280375711e-05, "loss": 0.6031, "step": 751 }, { "epoch": 0.08, "learning_rate": 5.9944423147981383e-05, "loss": 1.0277, "step": 752 }, { "epoch": 0.08, "learning_rate": 5.994427329062682e-05, "loss": 0.9127, "step": 753 }, { "epoch": 0.08, "learning_rate": 5.9944123231694394e-05, "loss": 0.3272, "step": 754 }, { "epoch": 0.08, "learning_rate": 5.994397297118515e-05, "loss": 0.3215, "step": 755 }, { "epoch": 0.08, "learning_rate": 5.994382250910009e-05, "loss": 0.9165, "step": 756 }, { "epoch": 0.08, "learning_rate": 5.9943671845440214e-05, "loss": 0.9657, "step": 757 }, { "epoch": 0.08, "learning_rate": 5.994352098020655e-05, "loss": 1.2421, "step": 758 }, { "epoch": 0.08, "learning_rate": 5.994336991340011e-05, "loss": 0.3505, "step": 759 }, { "epoch": 0.08, "learning_rate": 5.99432186450219e-05, "loss": 0.5147, "step": 760 }, { "epoch": 0.08, "learning_rate": 5.994306717507296e-05, "loss": 0.5148, "step": 761 }, { "epoch": 0.08, "learning_rate": 5.99429155035543e-05, "loss": 0.3139, "step": 762 }, { "epoch": 0.08, "learning_rate": 5.994276363046692e-05, "loss": 1.2081, "step": 763 }, { "epoch": 0.08, "learning_rate": 5.994261155581188e-05, "loss": 0.4467, "step": 764 }, { "epoch": 0.08, "learning_rate": 5.994245927959018e-05, "loss": 1.1713, "step": 765 }, { "epoch": 0.08, "learning_rate": 5.994230680180285e-05, "loss": 0.3609, "step": 766 }, { "epoch": 0.08, "learning_rate": 5.9942154122450916e-05, "loss": 1.2669, "step": 767 }, { "epoch": 0.08, "learning_rate": 5.99420012415354e-05, "loss": 0.6601, "step": 768 }, { "epoch": 0.08, "learning_rate": 5.994184815905735e-05, "loss": 0.889, "step": 769 }, { "epoch": 0.08, "learning_rate": 5.994169487501778e-05, "loss": 0.3253, "step": 770 }, { "epoch": 0.08, "learning_rate": 5.994154138941772e-05, "loss": 0.2192, "step": 771 }, { "epoch": 0.08, "learning_rate": 5.9941387702258204e-05, "loss": 0.7537, "step": 772 }, { "epoch": 0.08, "learning_rate": 5.9941233813540294e-05, "loss": 0.441, "step": 773 }, { "epoch": 0.08, "learning_rate": 5.994107972326498e-05, "loss": 0.7112, "step": 774 }, { "epoch": 0.08, "learning_rate": 5.9940925431433334e-05, "loss": 0.9537, "step": 775 }, { "epoch": 0.08, "learning_rate": 5.9940770938046385e-05, "loss": 0.9248, "step": 776 }, { "epoch": 0.08, "learning_rate": 5.994061624310517e-05, "loss": 1.0155, "step": 777 }, { "epoch": 0.08, "learning_rate": 5.994046134661073e-05, "loss": 0.8101, "step": 778 }, { "epoch": 0.08, "learning_rate": 5.994030624856411e-05, "loss": 1.0182, "step": 779 }, { "epoch": 0.08, "learning_rate": 5.994015094896635e-05, "loss": 0.7217, "step": 780 }, { "epoch": 0.08, "learning_rate": 5.993999544781851e-05, "loss": 0.852, "step": 781 }, { "epoch": 0.08, "learning_rate": 5.9939839745121614e-05, "loss": 0.9143, "step": 782 }, { "epoch": 0.08, "learning_rate": 5.993968384087673e-05, "loss": 0.344, "step": 783 }, { "epoch": 0.08, "learning_rate": 5.99395277350849e-05, "loss": 0.5082, "step": 784 }, { "epoch": 0.08, "learning_rate": 5.993937142774717e-05, "loss": 0.4765, "step": 785 }, { "epoch": 0.08, "learning_rate": 5.99392149188646e-05, "loss": 1.3938, "step": 786 }, { "epoch": 0.08, "learning_rate": 5.993905820843824e-05, "loss": 1.1849, "step": 787 }, { "epoch": 0.08, "learning_rate": 5.993890129646915e-05, "loss": 1.0326, "step": 788 }, { "epoch": 0.08, "learning_rate": 5.9938744182958374e-05, "loss": 0.4335, "step": 789 }, { "epoch": 0.08, "learning_rate": 5.993858686790698e-05, "loss": 0.4254, "step": 790 }, { "epoch": 0.08, "learning_rate": 5.9938429351316023e-05, "loss": 0.2125, "step": 791 }, { "epoch": 0.08, "learning_rate": 5.9938271633186567e-05, "loss": 0.724, "step": 792 }, { "epoch": 0.08, "learning_rate": 5.993811371351967e-05, "loss": 0.681, "step": 793 }, { "epoch": 0.08, "learning_rate": 5.9937955592316396e-05, "loss": 0.176, "step": 794 }, { "epoch": 0.08, "learning_rate": 5.993779726957781e-05, "loss": 1.0405, "step": 795 }, { "epoch": 0.08, "learning_rate": 5.993763874530498e-05, "loss": 0.3473, "step": 796 }, { "epoch": 0.08, "learning_rate": 5.993748001949897e-05, "loss": 0.7596, "step": 797 }, { "epoch": 0.08, "learning_rate": 5.993732109216085e-05, "loss": 0.9848, "step": 798 }, { "epoch": 0.08, "learning_rate": 5.993716196329169e-05, "loss": 0.3917, "step": 799 }, { "epoch": 0.08, "learning_rate": 5.9937002632892547e-05, "loss": 1.0823, "step": 800 }, { "epoch": 0.08, "learning_rate": 5.993684310096452e-05, "loss": 0.9136, "step": 801 }, { "epoch": 0.08, "learning_rate": 5.993668336750867e-05, "loss": 0.9952, "step": 802 }, { "epoch": 0.08, "learning_rate": 5.9936523432526066e-05, "loss": 0.2182, "step": 803 }, { "epoch": 0.08, "learning_rate": 5.99363632960178e-05, "loss": 1.4427, "step": 804 }, { "epoch": 0.08, "learning_rate": 5.9936202957984935e-05, "loss": 0.7412, "step": 805 }, { "epoch": 0.08, "learning_rate": 5.993604241842855e-05, "loss": 0.7579, "step": 806 }, { "epoch": 0.08, "learning_rate": 5.993588167734974e-05, "loss": 1.4301, "step": 807 }, { "epoch": 0.08, "learning_rate": 5.993572073474958e-05, "loss": 0.4799, "step": 808 }, { "epoch": 0.08, "learning_rate": 5.993555959062915e-05, "loss": 0.3842, "step": 809 }, { "epoch": 0.08, "learning_rate": 5.993539824498953e-05, "loss": 0.9406, "step": 810 }, { "epoch": 0.08, "learning_rate": 5.993523669783183e-05, "loss": 0.8801, "step": 811 }, { "epoch": 0.08, "learning_rate": 5.993507494915711e-05, "loss": 0.2377, "step": 812 }, { "epoch": 0.08, "learning_rate": 5.9934912998966474e-05, "loss": 0.7201, "step": 813 }, { "epoch": 0.09, "learning_rate": 5.9934750847261e-05, "loss": 1.0119, "step": 814 }, { "epoch": 0.09, "learning_rate": 5.993458849404179e-05, "loss": 1.0173, "step": 815 }, { "epoch": 0.09, "learning_rate": 5.993442593930994e-05, "loss": 0.4815, "step": 816 }, { "epoch": 0.09, "learning_rate": 5.993426318306653e-05, "loss": 0.5186, "step": 817 }, { "epoch": 0.09, "learning_rate": 5.993410022531268e-05, "loss": 1.0873, "step": 818 }, { "epoch": 0.09, "learning_rate": 5.993393706604946e-05, "loss": 1.1352, "step": 819 }, { "epoch": 0.09, "learning_rate": 5.9933773705277984e-05, "loss": 0.9363, "step": 820 }, { "epoch": 0.09, "learning_rate": 5.993361014299934e-05, "loss": 1.0555, "step": 821 }, { "epoch": 0.09, "learning_rate": 5.993344637921464e-05, "loss": 1.0836, "step": 822 }, { "epoch": 0.09, "learning_rate": 5.993328241392498e-05, "loss": 0.96, "step": 823 }, { "epoch": 0.09, "learning_rate": 5.993311824713148e-05, "loss": 0.6277, "step": 824 }, { "epoch": 0.09, "learning_rate": 5.9932953878835214e-05, "loss": 1.3372, "step": 825 }, { "epoch": 0.09, "learning_rate": 5.9932789309037314e-05, "loss": 1.1603, "step": 826 }, { "epoch": 0.09, "learning_rate": 5.993262453773888e-05, "loss": 0.3044, "step": 827 }, { "epoch": 0.09, "learning_rate": 5.9932459564941015e-05, "loss": 0.0783, "step": 828 }, { "epoch": 0.09, "learning_rate": 5.993229439064485e-05, "loss": 0.4025, "step": 829 }, { "epoch": 0.09, "learning_rate": 5.993212901485148e-05, "loss": 1.4698, "step": 830 }, { "epoch": 0.09, "learning_rate": 5.9931963437562015e-05, "loss": 0.1157, "step": 831 }, { "epoch": 0.09, "learning_rate": 5.993179765877757e-05, "loss": 1.1024, "step": 832 }, { "epoch": 0.09, "learning_rate": 5.9931631678499275e-05, "loss": 0.5485, "step": 833 }, { "epoch": 0.09, "learning_rate": 5.993146549672824e-05, "loss": 1.2172, "step": 834 }, { "epoch": 0.09, "learning_rate": 5.9931299113465576e-05, "loss": 0.6279, "step": 835 }, { "epoch": 0.09, "learning_rate": 5.993113252871242e-05, "loss": 1.2368, "step": 836 }, { "epoch": 0.09, "learning_rate": 5.993096574246987e-05, "loss": 1.184, "step": 837 }, { "epoch": 0.09, "learning_rate": 5.993079875473907e-05, "loss": 1.3221, "step": 838 }, { "epoch": 0.09, "learning_rate": 5.993063156552114e-05, "loss": 0.9694, "step": 839 }, { "epoch": 0.09, "learning_rate": 5.993046417481719e-05, "loss": 0.3563, "step": 840 }, { "epoch": 0.09, "learning_rate": 5.9930296582628374e-05, "loss": 0.6108, "step": 841 }, { "epoch": 0.09, "learning_rate": 5.993012878895579e-05, "loss": 0.5728, "step": 842 }, { "epoch": 0.09, "learning_rate": 5.992996079380059e-05, "loss": 0.7932, "step": 843 }, { "epoch": 0.09, "learning_rate": 5.992979259716389e-05, "loss": 0.3722, "step": 844 }, { "epoch": 0.09, "learning_rate": 5.9929624199046836e-05, "loss": 1.1641, "step": 845 }, { "epoch": 0.09, "learning_rate": 5.9929455599450556e-05, "loss": 0.4653, "step": 846 }, { "epoch": 0.09, "learning_rate": 5.9929286798376174e-05, "loss": 0.3707, "step": 847 }, { "epoch": 0.09, "learning_rate": 5.992911779582485e-05, "loss": 1.1551, "step": 848 }, { "epoch": 0.09, "learning_rate": 5.9928948591797695e-05, "loss": 0.4885, "step": 849 }, { "epoch": 0.09, "learning_rate": 5.992877918629587e-05, "loss": 0.5234, "step": 850 }, { "epoch": 0.09, "learning_rate": 5.99286095793205e-05, "loss": 0.709, "step": 851 }, { "epoch": 0.09, "learning_rate": 5.992843977087273e-05, "loss": 1.2187, "step": 852 }, { "epoch": 0.09, "learning_rate": 5.992826976095372e-05, "loss": 0.5051, "step": 853 }, { "epoch": 0.09, "learning_rate": 5.992809954956459e-05, "loss": 0.507, "step": 854 }, { "epoch": 0.09, "learning_rate": 5.9927929136706495e-05, "loss": 0.4411, "step": 855 }, { "epoch": 0.09, "learning_rate": 5.992775852238059e-05, "loss": 0.4095, "step": 856 }, { "epoch": 0.09, "learning_rate": 5.992758770658801e-05, "loss": 0.5511, "step": 857 }, { "epoch": 0.09, "learning_rate": 5.992741668932992e-05, "loss": 0.8942, "step": 858 }, { "epoch": 0.09, "learning_rate": 5.9927245470607454e-05, "loss": 1.1359, "step": 859 }, { "epoch": 0.09, "learning_rate": 5.9927074050421784e-05, "loss": 0.4136, "step": 860 }, { "epoch": 0.09, "learning_rate": 5.992690242877405e-05, "loss": 1.1713, "step": 861 }, { "epoch": 0.09, "learning_rate": 5.992673060566542e-05, "loss": 0.2776, "step": 862 }, { "epoch": 0.09, "learning_rate": 5.992655858109703e-05, "loss": 1.2702, "step": 863 }, { "epoch": 0.09, "learning_rate": 5.992638635507005e-05, "loss": 1.1188, "step": 864 }, { "epoch": 0.09, "learning_rate": 5.992621392758564e-05, "loss": 0.3562, "step": 865 }, { "epoch": 0.09, "learning_rate": 5.992604129864496e-05, "loss": 1.1106, "step": 866 }, { "epoch": 0.09, "learning_rate": 5.992586846824917e-05, "loss": 1.2184, "step": 867 }, { "epoch": 0.09, "learning_rate": 5.9925695436399446e-05, "loss": 0.521, "step": 868 }, { "epoch": 0.09, "learning_rate": 5.992552220309693e-05, "loss": 1.0934, "step": 869 }, { "epoch": 0.09, "learning_rate": 5.992534876834281e-05, "loss": 1.2549, "step": 870 }, { "epoch": 0.09, "learning_rate": 5.992517513213824e-05, "loss": 0.5021, "step": 871 }, { "epoch": 0.09, "learning_rate": 5.9925001294484396e-05, "loss": 1.3307, "step": 872 }, { "epoch": 0.09, "learning_rate": 5.9924827255382446e-05, "loss": 1.1551, "step": 873 }, { "epoch": 0.09, "learning_rate": 5.992465301483355e-05, "loss": 0.5478, "step": 874 }, { "epoch": 0.09, "learning_rate": 5.99244785728389e-05, "loss": 1.3021, "step": 875 }, { "epoch": 0.09, "learning_rate": 5.9924303929399664e-05, "loss": 0.8585, "step": 876 }, { "epoch": 0.09, "learning_rate": 5.992412908451701e-05, "loss": 1.9794, "step": 877 }, { "epoch": 0.09, "learning_rate": 5.992395403819213e-05, "loss": 1.054, "step": 878 }, { "epoch": 0.09, "learning_rate": 5.992377879042618e-05, "loss": 1.0907, "step": 879 }, { "epoch": 0.09, "learning_rate": 5.9923603341220363e-05, "loss": 0.7989, "step": 880 }, { "epoch": 0.09, "learning_rate": 5.992342769057584e-05, "loss": 0.5464, "step": 881 }, { "epoch": 0.09, "learning_rate": 5.9923251838493814e-05, "loss": 0.6172, "step": 882 }, { "epoch": 0.09, "learning_rate": 5.992307578497546e-05, "loss": 0.5706, "step": 883 }, { "epoch": 0.09, "learning_rate": 5.992289953002195e-05, "loss": 0.6785, "step": 884 }, { "epoch": 0.09, "learning_rate": 5.992272307363448e-05, "loss": 0.3317, "step": 885 }, { "epoch": 0.09, "learning_rate": 5.9922546415814254e-05, "loss": 1.6285, "step": 886 }, { "epoch": 0.09, "learning_rate": 5.992236955656244e-05, "loss": 1.0494, "step": 887 }, { "epoch": 0.09, "learning_rate": 5.992219249588023e-05, "loss": 0.8356, "step": 888 }, { "epoch": 0.09, "learning_rate": 5.992201523376883e-05, "loss": 1.1417, "step": 889 }, { "epoch": 0.09, "learning_rate": 5.992183777022942e-05, "loss": 1.3191, "step": 890 }, { "epoch": 0.09, "learning_rate": 5.99216601052632e-05, "loss": 0.7606, "step": 891 }, { "epoch": 0.09, "learning_rate": 5.992148223887136e-05, "loss": 0.4033, "step": 892 }, { "epoch": 0.09, "learning_rate": 5.992130417105511e-05, "loss": 0.7853, "step": 893 }, { "epoch": 0.09, "learning_rate": 5.992112590181564e-05, "loss": 0.9212, "step": 894 }, { "epoch": 0.09, "learning_rate": 5.9920947431154156e-05, "loss": 0.7252, "step": 895 }, { "epoch": 0.09, "learning_rate": 5.9920768759071845e-05, "loss": 0.3186, "step": 896 }, { "epoch": 0.09, "learning_rate": 5.992058988556992e-05, "loss": 0.6578, "step": 897 }, { "epoch": 0.09, "learning_rate": 5.9920410810649586e-05, "loss": 0.1683, "step": 898 }, { "epoch": 0.09, "learning_rate": 5.992023153431205e-05, "loss": 0.8001, "step": 899 }, { "epoch": 0.09, "learning_rate": 5.992005205655852e-05, "loss": 0.8838, "step": 900 }, { "epoch": 0.09, "learning_rate": 5.9919872377390195e-05, "loss": 1.0171, "step": 901 }, { "epoch": 0.09, "learning_rate": 5.9919692496808296e-05, "loss": 0.4435, "step": 902 }, { "epoch": 0.09, "learning_rate": 5.9919512414814024e-05, "loss": 0.4638, "step": 903 }, { "epoch": 0.09, "learning_rate": 5.991933213140859e-05, "loss": 1.0116, "step": 904 }, { "epoch": 0.09, "learning_rate": 5.991915164659322e-05, "loss": 0.8014, "step": 905 }, { "epoch": 0.09, "learning_rate": 5.991897096036912e-05, "loss": 0.2861, "step": 906 }, { "epoch": 0.09, "learning_rate": 5.99187900727375e-05, "loss": 0.4029, "step": 907 }, { "epoch": 0.09, "learning_rate": 5.99186089836996e-05, "loss": 1.1404, "step": 908 }, { "epoch": 0.09, "learning_rate": 5.991842769325661e-05, "loss": 1.1428, "step": 909 }, { "epoch": 0.1, "learning_rate": 5.9918246201409775e-05, "loss": 0.6061, "step": 910 }, { "epoch": 0.1, "learning_rate": 5.9918064508160306e-05, "loss": 1.3707, "step": 911 }, { "epoch": 0.1, "learning_rate": 5.9917882613509415e-05, "loss": 1.0384, "step": 912 }, { "epoch": 0.1, "learning_rate": 5.9917700517458354e-05, "loss": 0.9177, "step": 913 }, { "epoch": 0.1, "learning_rate": 5.9917518220008324e-05, "loss": 0.7774, "step": 914 }, { "epoch": 0.1, "learning_rate": 5.991733572116057e-05, "loss": 1.2549, "step": 915 }, { "epoch": 0.1, "learning_rate": 5.99171530209163e-05, "loss": 1.1906, "step": 916 }, { "epoch": 0.1, "learning_rate": 5.991697011927677e-05, "loss": 0.9346, "step": 917 }, { "epoch": 0.1, "learning_rate": 5.9916787016243185e-05, "loss": 1.3892, "step": 918 }, { "epoch": 0.1, "learning_rate": 5.99166037118168e-05, "loss": 1.4803, "step": 919 }, { "epoch": 0.1, "learning_rate": 5.9916420205998835e-05, "loss": 1.2845, "step": 920 }, { "epoch": 0.1, "learning_rate": 5.991623649879052e-05, "loss": 0.3967, "step": 921 }, { "epoch": 0.1, "learning_rate": 5.991605259019311e-05, "loss": 0.6103, "step": 922 }, { "epoch": 0.1, "learning_rate": 5.991586848020783e-05, "loss": 0.4334, "step": 923 }, { "epoch": 0.1, "learning_rate": 5.991568416883593e-05, "loss": 1.1925, "step": 924 }, { "epoch": 0.1, "learning_rate": 5.9915499656078636e-05, "loss": 1.2406, "step": 925 }, { "epoch": 0.1, "learning_rate": 5.99153149419372e-05, "loss": 1.2489, "step": 926 }, { "epoch": 0.1, "learning_rate": 5.9915130026412866e-05, "loss": 0.1489, "step": 927 }, { "epoch": 0.1, "learning_rate": 5.991494490950688e-05, "loss": 0.7562, "step": 928 }, { "epoch": 0.1, "learning_rate": 5.991475959122047e-05, "loss": 1.0441, "step": 929 }, { "epoch": 0.1, "learning_rate": 5.991457407155492e-05, "loss": 1.1918, "step": 930 }, { "epoch": 0.1, "learning_rate": 5.991438835051144e-05, "loss": 1.1904, "step": 931 }, { "epoch": 0.1, "learning_rate": 5.99142024280913e-05, "loss": 1.4233, "step": 932 }, { "epoch": 0.1, "learning_rate": 5.991401630429575e-05, "loss": 0.8783, "step": 933 }, { "epoch": 0.1, "learning_rate": 5.9913829979126036e-05, "loss": 0.9895, "step": 934 }, { "epoch": 0.1, "learning_rate": 5.991364345258342e-05, "loss": 1.1522, "step": 935 }, { "epoch": 0.1, "learning_rate": 5.991345672466916e-05, "loss": 0.4582, "step": 936 }, { "epoch": 0.1, "learning_rate": 5.991326979538451e-05, "loss": 1.1119, "step": 937 }, { "epoch": 0.1, "learning_rate": 5.9913082664730727e-05, "loss": 0.5613, "step": 938 }, { "epoch": 0.1, "learning_rate": 5.9912895332709066e-05, "loss": 0.6899, "step": 939 }, { "epoch": 0.1, "learning_rate": 5.991270779932079e-05, "loss": 0.9858, "step": 940 }, { "epoch": 0.1, "learning_rate": 5.991252006456717e-05, "loss": 0.5131, "step": 941 }, { "epoch": 0.1, "learning_rate": 5.991233212844946e-05, "loss": 0.4744, "step": 942 }, { "epoch": 0.1, "learning_rate": 5.9912143990968934e-05, "loss": 0.3396, "step": 943 }, { "epoch": 0.1, "learning_rate": 5.991195565212685e-05, "loss": 0.4161, "step": 944 }, { "epoch": 0.1, "learning_rate": 5.991176711192448e-05, "loss": 1.9556, "step": 945 }, { "epoch": 0.1, "learning_rate": 5.9911578370363093e-05, "loss": 0.8502, "step": 946 }, { "epoch": 0.1, "learning_rate": 5.991138942744396e-05, "loss": 1.2824, "step": 947 }, { "epoch": 0.1, "learning_rate": 5.991120028316835e-05, "loss": 1.1232, "step": 948 }, { "epoch": 0.1, "learning_rate": 5.991101093753754e-05, "loss": 0.537, "step": 949 }, { "epoch": 0.1, "learning_rate": 5.99108213905528e-05, "loss": 0.4785, "step": 950 }, { "epoch": 0.1, "learning_rate": 5.991063164221541e-05, "loss": 0.8203, "step": 951 }, { "epoch": 0.1, "learning_rate": 5.991044169252665e-05, "loss": 0.7722, "step": 952 }, { "epoch": 0.1, "learning_rate": 5.9910251541487794e-05, "loss": 0.8056, "step": 953 }, { "epoch": 0.1, "learning_rate": 5.991006118910012e-05, "loss": 1.1747, "step": 954 }, { "epoch": 0.1, "learning_rate": 5.9909870635364914e-05, "loss": 0.3632, "step": 955 }, { "epoch": 0.1, "learning_rate": 5.9909679880283465e-05, "loss": 0.4468, "step": 956 }, { "epoch": 0.1, "learning_rate": 5.990948892385703e-05, "loss": 0.6376, "step": 957 }, { "epoch": 0.1, "learning_rate": 5.990929776608693e-05, "loss": 0.7961, "step": 958 }, { "epoch": 0.1, "learning_rate": 5.990910640697444e-05, "loss": 1.214, "step": 959 }, { "epoch": 0.1, "learning_rate": 5.990891484652083e-05, "loss": 0.5051, "step": 960 }, { "epoch": 0.1, "learning_rate": 5.9908723084727405e-05, "loss": 0.4279, "step": 961 }, { "epoch": 0.1, "learning_rate": 5.9908531121595465e-05, "loss": 0.2785, "step": 962 }, { "epoch": 0.1, "learning_rate": 5.990833895712628e-05, "loss": 0.5008, "step": 963 }, { "epoch": 0.1, "learning_rate": 5.990814659132116e-05, "loss": 0.3484, "step": 964 }, { "epoch": 0.1, "learning_rate": 5.990795402418139e-05, "loss": 1.1648, "step": 965 }, { "epoch": 0.1, "learning_rate": 5.990776125570828e-05, "loss": 0.64, "step": 966 }, { "epoch": 0.1, "learning_rate": 5.990756828590311e-05, "loss": 0.902, "step": 967 }, { "epoch": 0.1, "learning_rate": 5.99073751147672e-05, "loss": 1.42, "step": 968 }, { "epoch": 0.1, "learning_rate": 5.9907181742301825e-05, "loss": 1.1137, "step": 969 }, { "epoch": 0.1, "learning_rate": 5.990698816850831e-05, "loss": 1.2288, "step": 970 }, { "epoch": 0.1, "learning_rate": 5.990679439338795e-05, "loss": 0.626, "step": 971 }, { "epoch": 0.1, "learning_rate": 5.9906600416942034e-05, "loss": 1.1953, "step": 972 }, { "epoch": 0.1, "learning_rate": 5.990640623917189e-05, "loss": 0.7988, "step": 973 }, { "epoch": 0.1, "learning_rate": 5.990621186007882e-05, "loss": 1.1506, "step": 974 }, { "epoch": 0.1, "learning_rate": 5.990601727966413e-05, "loss": 1.1068, "step": 975 }, { "epoch": 0.1, "learning_rate": 5.990582249792912e-05, "loss": 1.2419, "step": 976 }, { "epoch": 0.1, "learning_rate": 5.990562751487512e-05, "loss": 1.5767, "step": 977 }, { "epoch": 0.1, "learning_rate": 5.990543233050343e-05, "loss": 1.4152, "step": 978 }, { "epoch": 0.1, "learning_rate": 5.990523694481536e-05, "loss": 1.0604, "step": 979 }, { "epoch": 0.1, "learning_rate": 5.990504135781224e-05, "loss": 0.5945, "step": 980 }, { "epoch": 0.1, "learning_rate": 5.990484556949538e-05, "loss": 1.0071, "step": 981 }, { "epoch": 0.1, "learning_rate": 5.9904649579866094e-05, "loss": 0.5025, "step": 982 }, { "epoch": 0.1, "learning_rate": 5.99044533889257e-05, "loss": 0.923, "step": 983 }, { "epoch": 0.1, "learning_rate": 5.990425699667553e-05, "loss": 0.407, "step": 984 }, { "epoch": 0.1, "learning_rate": 5.9904060403116897e-05, "loss": 1.3309, "step": 985 }, { "epoch": 0.1, "learning_rate": 5.990386360825112e-05, "loss": 1.0436, "step": 986 }, { "epoch": 0.1, "learning_rate": 5.9903666612079534e-05, "loss": 1.5201, "step": 987 }, { "epoch": 0.1, "learning_rate": 5.990346941460346e-05, "loss": 1.4066, "step": 988 }, { "epoch": 0.1, "learning_rate": 5.990327201582423e-05, "loss": 0.5552, "step": 989 }, { "epoch": 0.1, "learning_rate": 5.990307441574317e-05, "loss": 0.3512, "step": 990 }, { "epoch": 0.1, "learning_rate": 5.990287661436161e-05, "loss": 1.1168, "step": 991 }, { "epoch": 0.1, "learning_rate": 5.9902678611680884e-05, "loss": 0.9263, "step": 992 }, { "epoch": 0.1, "learning_rate": 5.990248040770231e-05, "loss": 0.6558, "step": 993 }, { "epoch": 0.1, "learning_rate": 5.9902282002427246e-05, "loss": 1.266, "step": 994 }, { "epoch": 0.1, "learning_rate": 5.990208339585701e-05, "loss": 0.6632, "step": 995 }, { "epoch": 0.1, "learning_rate": 5.990188458799295e-05, "loss": 0.352, "step": 996 }, { "epoch": 0.1, "learning_rate": 5.99016855788364e-05, "loss": 1.0033, "step": 997 }, { "epoch": 0.1, "learning_rate": 5.990148636838869e-05, "loss": 0.9699, "step": 998 }, { "epoch": 0.1, "learning_rate": 5.990128695665118e-05, "loss": 0.7187, "step": 999 }, { "epoch": 0.1, "learning_rate": 5.99010873436252e-05, "loss": 0.2894, "step": 1000 }, { "epoch": 0.1, "learning_rate": 5.9900887529312094e-05, "loss": 0.5079, "step": 1001 }, { "epoch": 0.1, "learning_rate": 5.9900687513713216e-05, "loss": 0.7694, "step": 1002 }, { "epoch": 0.1, "learning_rate": 5.99004872968299e-05, "loss": 0.7322, "step": 1003 }, { "epoch": 0.1, "learning_rate": 5.99002868786635e-05, "loss": 0.5312, "step": 1004 }, { "epoch": 0.1, "learning_rate": 5.990008625921536e-05, "loss": 1.2876, "step": 1005 }, { "epoch": 0.11, "learning_rate": 5.9899885438486844e-05, "loss": 0.77, "step": 1006 }, { "epoch": 0.11, "learning_rate": 5.989968441647929e-05, "loss": 0.7385, "step": 1007 }, { "epoch": 0.11, "learning_rate": 5.989948319319406e-05, "loss": 0.3989, "step": 1008 }, { "epoch": 0.11, "learning_rate": 5.98992817686325e-05, "loss": 0.667, "step": 1009 }, { "epoch": 0.11, "learning_rate": 5.989908014279598e-05, "loss": 1.7509, "step": 1010 }, { "epoch": 0.11, "learning_rate": 5.9898878315685835e-05, "loss": 0.8099, "step": 1011 }, { "epoch": 0.11, "learning_rate": 5.989867628730345e-05, "loss": 0.9024, "step": 1012 }, { "epoch": 0.11, "learning_rate": 5.989847405765016e-05, "loss": 1.4561, "step": 1013 }, { "epoch": 0.11, "learning_rate": 5.989827162672735e-05, "loss": 0.374, "step": 1014 }, { "epoch": 0.11, "learning_rate": 5.989806899453636e-05, "loss": 0.6439, "step": 1015 }, { "epoch": 0.11, "learning_rate": 5.989786616107858e-05, "loss": 0.451, "step": 1016 }, { "epoch": 0.11, "learning_rate": 5.989766312635535e-05, "loss": 1.221, "step": 1017 }, { "epoch": 0.11, "learning_rate": 5.989745989036805e-05, "loss": 0.6601, "step": 1018 }, { "epoch": 0.11, "learning_rate": 5.989725645311804e-05, "loss": 0.51, "step": 1019 }, { "epoch": 0.11, "learning_rate": 5.989705281460669e-05, "loss": 0.8948, "step": 1020 }, { "epoch": 0.11, "learning_rate": 5.989684897483539e-05, "loss": 0.6498, "step": 1021 }, { "epoch": 0.11, "learning_rate": 5.9896644933805495e-05, "loss": 0.8784, "step": 1022 }, { "epoch": 0.11, "learning_rate": 5.989644069151838e-05, "loss": 1.3535, "step": 1023 }, { "epoch": 0.11, "learning_rate": 5.9896236247975424e-05, "loss": 0.7215, "step": 1024 }, { "epoch": 0.11, "learning_rate": 5.989603160317799e-05, "loss": 1.1001, "step": 1025 }, { "epoch": 0.11, "learning_rate": 5.9895826757127476e-05, "loss": 1.6456, "step": 1026 }, { "epoch": 0.11, "learning_rate": 5.989562170982525e-05, "loss": 0.7488, "step": 1027 }, { "epoch": 0.11, "learning_rate": 5.98954164612727e-05, "loss": 1.191, "step": 1028 }, { "epoch": 0.11, "learning_rate": 5.98952110114712e-05, "loss": 0.5221, "step": 1029 }, { "epoch": 0.11, "learning_rate": 5.9895005360422116e-05, "loss": 1.3044, "step": 1030 }, { "epoch": 0.11, "learning_rate": 5.9894799508126864e-05, "loss": 1.052, "step": 1031 }, { "epoch": 0.11, "learning_rate": 5.9894593454586816e-05, "loss": 0.6772, "step": 1032 }, { "epoch": 0.11, "learning_rate": 5.989438719980337e-05, "loss": 0.363, "step": 1033 }, { "epoch": 0.11, "learning_rate": 5.989418074377789e-05, "loss": 1.0439, "step": 1034 }, { "epoch": 0.11, "learning_rate": 5.989397408651178e-05, "loss": 0.2482, "step": 1035 }, { "epoch": 0.11, "learning_rate": 5.989376722800644e-05, "loss": 1.1739, "step": 1036 }, { "epoch": 0.11, "learning_rate": 5.989356016826325e-05, "loss": 1.2435, "step": 1037 }, { "epoch": 0.11, "learning_rate": 5.98933529072836e-05, "loss": 0.5453, "step": 1038 }, { "epoch": 0.11, "learning_rate": 5.98931454450689e-05, "loss": 0.4943, "step": 1039 }, { "epoch": 0.11, "learning_rate": 5.9892937781620535e-05, "loss": 0.646, "step": 1040 }, { "epoch": 0.11, "learning_rate": 5.9892729916939914e-05, "loss": 1.1427, "step": 1041 }, { "epoch": 0.11, "learning_rate": 5.989252185102843e-05, "loss": 0.8821, "step": 1042 }, { "epoch": 0.11, "learning_rate": 5.989231358388747e-05, "loss": 0.6977, "step": 1043 }, { "epoch": 0.11, "learning_rate": 5.9892105115518464e-05, "loss": 1.1327, "step": 1044 }, { "epoch": 0.11, "learning_rate": 5.9891896445922796e-05, "loss": 1.2235, "step": 1045 }, { "epoch": 0.11, "learning_rate": 5.9891687575101875e-05, "loss": 1.2023, "step": 1046 }, { "epoch": 0.11, "learning_rate": 5.98914785030571e-05, "loss": 0.2978, "step": 1047 }, { "epoch": 0.11, "learning_rate": 5.98912692297899e-05, "loss": 0.5703, "step": 1048 }, { "epoch": 0.11, "learning_rate": 5.989105975530166e-05, "loss": 0.7893, "step": 1049 }, { "epoch": 0.11, "learning_rate": 5.98908500795938e-05, "loss": 1.1807, "step": 1050 }, { "epoch": 0.11, "learning_rate": 5.989064020266773e-05, "loss": 0.3702, "step": 1051 }, { "epoch": 0.11, "learning_rate": 5.989043012452486e-05, "loss": 0.3825, "step": 1052 }, { "epoch": 0.11, "learning_rate": 5.989021984516662e-05, "loss": 1.171, "step": 1053 }, { "epoch": 0.11, "learning_rate": 5.9890009364594404e-05, "loss": 0.7448, "step": 1054 }, { "epoch": 0.11, "learning_rate": 5.988979868280965e-05, "loss": 0.7291, "step": 1055 }, { "epoch": 0.11, "learning_rate": 5.9889587799813755e-05, "loss": 0.5429, "step": 1056 }, { "epoch": 0.11, "learning_rate": 5.988937671560815e-05, "loss": 1.1987, "step": 1057 }, { "epoch": 0.11, "learning_rate": 5.988916543019426e-05, "loss": 0.4178, "step": 1058 }, { "epoch": 0.11, "learning_rate": 5.9888953943573494e-05, "loss": 0.9107, "step": 1059 }, { "epoch": 0.11, "learning_rate": 5.988874225574729e-05, "loss": 0.3468, "step": 1060 }, { "epoch": 0.11, "learning_rate": 5.988853036671706e-05, "loss": 0.4314, "step": 1061 }, { "epoch": 0.11, "learning_rate": 5.988831827648425e-05, "loss": 0.4489, "step": 1062 }, { "epoch": 0.11, "learning_rate": 5.9888105985050265e-05, "loss": 0.2611, "step": 1063 }, { "epoch": 0.11, "learning_rate": 5.988789349241654e-05, "loss": 1.4867, "step": 1064 }, { "epoch": 0.11, "learning_rate": 5.988768079858452e-05, "loss": 0.9089, "step": 1065 }, { "epoch": 0.11, "learning_rate": 5.9887467903555614e-05, "loss": 1.5654, "step": 1066 }, { "epoch": 0.11, "learning_rate": 5.988725480733127e-05, "loss": 1.274, "step": 1067 }, { "epoch": 0.11, "learning_rate": 5.988704150991293e-05, "loss": 0.617, "step": 1068 }, { "epoch": 0.11, "learning_rate": 5.9886828011302e-05, "loss": 0.8763, "step": 1069 }, { "epoch": 0.11, "learning_rate": 5.988661431149996e-05, "loss": 0.6035, "step": 1070 }, { "epoch": 0.11, "learning_rate": 5.988640041050821e-05, "loss": 1.2007, "step": 1071 }, { "epoch": 0.11, "learning_rate": 5.988618630832821e-05, "loss": 0.431, "step": 1072 }, { "epoch": 0.11, "learning_rate": 5.9885972004961396e-05, "loss": 1.3719, "step": 1073 }, { "epoch": 0.11, "learning_rate": 5.9885757500409214e-05, "loss": 0.5055, "step": 1074 }, { "epoch": 0.11, "learning_rate": 5.98855427946731e-05, "loss": 0.4827, "step": 1075 }, { "epoch": 0.11, "learning_rate": 5.988532788775451e-05, "loss": 0.9831, "step": 1076 }, { "epoch": 0.11, "learning_rate": 5.988511277965488e-05, "loss": 0.9564, "step": 1077 }, { "epoch": 0.11, "learning_rate": 5.988489747037567e-05, "loss": 0.7683, "step": 1078 }, { "epoch": 0.11, "learning_rate": 5.988468195991831e-05, "loss": 0.976, "step": 1079 }, { "epoch": 0.11, "learning_rate": 5.9884466248284275e-05, "loss": 1.1275, "step": 1080 }, { "epoch": 0.11, "learning_rate": 5.9884250335475004e-05, "loss": 0.6424, "step": 1081 }, { "epoch": 0.11, "learning_rate": 5.988403422149195e-05, "loss": 0.9207, "step": 1082 }, { "epoch": 0.11, "learning_rate": 5.9883817906336574e-05, "loss": 1.1559, "step": 1083 }, { "epoch": 0.11, "learning_rate": 5.9883601390010317e-05, "loss": 0.3891, "step": 1084 }, { "epoch": 0.11, "learning_rate": 5.988338467251466e-05, "loss": 1.2534, "step": 1085 }, { "epoch": 0.11, "learning_rate": 5.988316775385104e-05, "loss": 0.5401, "step": 1086 }, { "epoch": 0.11, "learning_rate": 5.988295063402094e-05, "loss": 1.2274, "step": 1087 }, { "epoch": 0.11, "learning_rate": 5.9882733313025795e-05, "loss": 1.0672, "step": 1088 }, { "epoch": 0.11, "learning_rate": 5.988251579086708e-05, "loss": 1.2185, "step": 1089 }, { "epoch": 0.11, "learning_rate": 5.988229806754627e-05, "loss": 1.131, "step": 1090 }, { "epoch": 0.11, "learning_rate": 5.9882080143064826e-05, "loss": 0.6451, "step": 1091 }, { "epoch": 0.11, "learning_rate": 5.9881862017424196e-05, "loss": 0.2898, "step": 1092 }, { "epoch": 0.11, "learning_rate": 5.988164369062588e-05, "loss": 0.426, "step": 1093 }, { "epoch": 0.11, "learning_rate": 5.9881425162671316e-05, "loss": 0.797, "step": 1094 }, { "epoch": 0.11, "learning_rate": 5.9881206433561995e-05, "loss": 1.0782, "step": 1095 }, { "epoch": 0.11, "learning_rate": 5.988098750329938e-05, "loss": 0.5753, "step": 1096 }, { "epoch": 0.11, "learning_rate": 5.988076837188495e-05, "loss": 1.15, "step": 1097 }, { "epoch": 0.11, "learning_rate": 5.988054903932018e-05, "loss": 0.5467, "step": 1098 }, { "epoch": 0.11, "learning_rate": 5.9880329505606546e-05, "loss": 0.9434, "step": 1099 }, { "epoch": 0.11, "learning_rate": 5.988010977074552e-05, "loss": 1.2409, "step": 1100 }, { "epoch": 0.11, "learning_rate": 5.9879889834738595e-05, "loss": 1.3829, "step": 1101 }, { "epoch": 0.12, "learning_rate": 5.9879669697587235e-05, "loss": 0.8946, "step": 1102 }, { "epoch": 0.12, "learning_rate": 5.987944935929294e-05, "loss": 1.0878, "step": 1103 }, { "epoch": 0.12, "learning_rate": 5.987922881985718e-05, "loss": 1.0711, "step": 1104 }, { "epoch": 0.12, "learning_rate": 5.987900807928143e-05, "loss": 0.7089, "step": 1105 }, { "epoch": 0.12, "learning_rate": 5.98787871375672e-05, "loss": 0.4243, "step": 1106 }, { "epoch": 0.12, "learning_rate": 5.9878565994715964e-05, "loss": 0.6427, "step": 1107 }, { "epoch": 0.12, "learning_rate": 5.987834465072921e-05, "loss": 0.4402, "step": 1108 }, { "epoch": 0.12, "learning_rate": 5.987812310560844e-05, "loss": 0.3221, "step": 1109 }, { "epoch": 0.12, "learning_rate": 5.9877901359355125e-05, "loss": 0.8822, "step": 1110 }, { "epoch": 0.12, "learning_rate": 5.987767941197078e-05, "loss": 0.9588, "step": 1111 }, { "epoch": 0.12, "learning_rate": 5.987745726345688e-05, "loss": 0.6268, "step": 1112 }, { "epoch": 0.12, "learning_rate": 5.987723491381493e-05, "loss": 0.3289, "step": 1113 }, { "epoch": 0.12, "learning_rate": 5.987701236304642e-05, "loss": 0.3945, "step": 1114 }, { "epoch": 0.12, "learning_rate": 5.9876789611152864e-05, "loss": 1.2418, "step": 1115 }, { "epoch": 0.12, "learning_rate": 5.987656665813575e-05, "loss": 1.3719, "step": 1116 }, { "epoch": 0.12, "learning_rate": 5.987634350399657e-05, "loss": 1.194, "step": 1117 }, { "epoch": 0.12, "learning_rate": 5.987612014873684e-05, "loss": 0.8685, "step": 1118 }, { "epoch": 0.12, "learning_rate": 5.987589659235807e-05, "loss": 1.272, "step": 1119 }, { "epoch": 0.12, "learning_rate": 5.987567283486174e-05, "loss": 0.5285, "step": 1120 }, { "epoch": 0.12, "learning_rate": 5.987544887624938e-05, "loss": 0.9759, "step": 1121 }, { "epoch": 0.12, "learning_rate": 5.987522471652249e-05, "loss": 1.1382, "step": 1122 }, { "epoch": 0.12, "learning_rate": 5.9875000355682564e-05, "loss": 0.6739, "step": 1123 }, { "epoch": 0.12, "learning_rate": 5.987477579373114e-05, "loss": 0.8095, "step": 1124 }, { "epoch": 0.12, "learning_rate": 5.987455103066971e-05, "loss": 0.6214, "step": 1125 }, { "epoch": 0.12, "learning_rate": 5.987432606649979e-05, "loss": 0.3276, "step": 1126 }, { "epoch": 0.12, "learning_rate": 5.98741009012229e-05, "loss": 0.825, "step": 1127 }, { "epoch": 0.12, "learning_rate": 5.987387553484056e-05, "loss": 0.7025, "step": 1128 }, { "epoch": 0.12, "learning_rate": 5.9873649967354265e-05, "loss": 1.3311, "step": 1129 }, { "epoch": 0.12, "learning_rate": 5.987342419876556e-05, "loss": 0.8128, "step": 1130 }, { "epoch": 0.12, "learning_rate": 5.987319822907594e-05, "loss": 1.1424, "step": 1131 }, { "epoch": 0.12, "learning_rate": 5.987297205828695e-05, "loss": 0.7618, "step": 1132 }, { "epoch": 0.12, "learning_rate": 5.98727456864001e-05, "loss": 0.9041, "step": 1133 }, { "epoch": 0.12, "learning_rate": 5.9872519113416924e-05, "loss": 0.7709, "step": 1134 }, { "epoch": 0.12, "learning_rate": 5.9872292339338926e-05, "loss": 0.6789, "step": 1135 }, { "epoch": 0.12, "learning_rate": 5.987206536416765e-05, "loss": 0.3909, "step": 1136 }, { "epoch": 0.12, "learning_rate": 5.987183818790461e-05, "loss": 0.4987, "step": 1137 }, { "epoch": 0.12, "learning_rate": 5.987161081055136e-05, "loss": 0.3369, "step": 1138 }, { "epoch": 0.12, "learning_rate": 5.987138323210941e-05, "loss": 1.4733, "step": 1139 }, { "epoch": 0.12, "learning_rate": 5.9871155452580296e-05, "loss": 0.9952, "step": 1140 }, { "epoch": 0.12, "learning_rate": 5.987092747196555e-05, "loss": 0.8935, "step": 1141 }, { "epoch": 0.12, "learning_rate": 5.987069929026672e-05, "loss": 0.7702, "step": 1142 }, { "epoch": 0.12, "learning_rate": 5.987047090748532e-05, "loss": 0.6385, "step": 1143 }, { "epoch": 0.12, "learning_rate": 5.9870242323622906e-05, "loss": 0.4382, "step": 1144 }, { "epoch": 0.12, "learning_rate": 5.987001353868101e-05, "loss": 0.3251, "step": 1145 }, { "epoch": 0.12, "learning_rate": 5.9869784552661165e-05, "loss": 0.731, "step": 1146 }, { "epoch": 0.12, "learning_rate": 5.986955536556493e-05, "loss": 1.043, "step": 1147 }, { "epoch": 0.12, "learning_rate": 5.9869325977393826e-05, "loss": 0.8914, "step": 1148 }, { "epoch": 0.12, "learning_rate": 5.986909638814942e-05, "loss": 0.6814, "step": 1149 }, { "epoch": 0.12, "learning_rate": 5.986886659783324e-05, "loss": 0.2294, "step": 1150 }, { "epoch": 0.12, "learning_rate": 5.9868636606446835e-05, "loss": 1.1778, "step": 1151 }, { "epoch": 0.12, "learning_rate": 5.986840641399176e-05, "loss": 1.2454, "step": 1152 }, { "epoch": 0.12, "learning_rate": 5.9868176020469566e-05, "loss": 0.6138, "step": 1153 }, { "epoch": 0.12, "learning_rate": 5.986794542588179e-05, "loss": 0.5885, "step": 1154 }, { "epoch": 0.12, "learning_rate": 5.986771463023e-05, "loss": 1.0851, "step": 1155 }, { "epoch": 0.12, "learning_rate": 5.986748363351574e-05, "loss": 0.4423, "step": 1156 }, { "epoch": 0.12, "learning_rate": 5.986725243574057e-05, "loss": 1.1205, "step": 1157 }, { "epoch": 0.12, "learning_rate": 5.986702103690605e-05, "loss": 0.9115, "step": 1158 }, { "epoch": 0.12, "learning_rate": 5.986678943701373e-05, "loss": 0.7999, "step": 1159 }, { "epoch": 0.12, "learning_rate": 5.9866557636065164e-05, "loss": 1.352, "step": 1160 }, { "epoch": 0.12, "learning_rate": 5.986632563406193e-05, "loss": 0.4519, "step": 1161 }, { "epoch": 0.12, "learning_rate": 5.986609343100558e-05, "loss": 0.6265, "step": 1162 }, { "epoch": 0.12, "learning_rate": 5.986586102689768e-05, "loss": 0.4584, "step": 1163 }, { "epoch": 0.12, "learning_rate": 5.986562842173978e-05, "loss": 0.7435, "step": 1164 }, { "epoch": 0.12, "learning_rate": 5.986539561553346e-05, "loss": 0.8261, "step": 1165 }, { "epoch": 0.12, "learning_rate": 5.986516260828029e-05, "loss": 0.4568, "step": 1166 }, { "epoch": 0.12, "learning_rate": 5.986492939998183e-05, "loss": 0.8008, "step": 1167 }, { "epoch": 0.12, "learning_rate": 5.9864695990639654e-05, "loss": 1.3354, "step": 1168 }, { "epoch": 0.12, "learning_rate": 5.986446238025534e-05, "loss": 0.4447, "step": 1169 }, { "epoch": 0.12, "learning_rate": 5.986422856883044e-05, "loss": 0.4186, "step": 1170 }, { "epoch": 0.12, "learning_rate": 5.986399455636655e-05, "loss": 1.1338, "step": 1171 }, { "epoch": 0.12, "learning_rate": 5.986376034286524e-05, "loss": 0.4659, "step": 1172 }, { "epoch": 0.12, "learning_rate": 5.986352592832807e-05, "loss": 0.7383, "step": 1173 }, { "epoch": 0.12, "learning_rate": 5.986329131275663e-05, "loss": 0.6274, "step": 1174 }, { "epoch": 0.12, "learning_rate": 5.9863056496152516e-05, "loss": 0.9682, "step": 1175 }, { "epoch": 0.12, "learning_rate": 5.986282147851728e-05, "loss": 1.2361, "step": 1176 }, { "epoch": 0.12, "learning_rate": 5.9862586259852524e-05, "loss": 0.4617, "step": 1177 }, { "epoch": 0.12, "learning_rate": 5.986235084015982e-05, "loss": 0.2198, "step": 1178 }, { "epoch": 0.12, "learning_rate": 5.986211521944076e-05, "loss": 0.1934, "step": 1179 }, { "epoch": 0.12, "learning_rate": 5.9861879397696923e-05, "loss": 1.2329, "step": 1180 }, { "epoch": 0.12, "learning_rate": 5.986164337492991e-05, "loss": 0.6759, "step": 1181 }, { "epoch": 0.12, "learning_rate": 5.98614071511413e-05, "loss": 0.3979, "step": 1182 }, { "epoch": 0.12, "learning_rate": 5.9861170726332684e-05, "loss": 0.5472, "step": 1183 }, { "epoch": 0.12, "learning_rate": 5.9860934100505644e-05, "loss": 1.1118, "step": 1184 }, { "epoch": 0.12, "learning_rate": 5.986069727366179e-05, "loss": 0.6648, "step": 1185 }, { "epoch": 0.12, "learning_rate": 5.98604602458027e-05, "loss": 0.7801, "step": 1186 }, { "epoch": 0.12, "learning_rate": 5.9860223016929996e-05, "loss": 0.9906, "step": 1187 }, { "epoch": 0.12, "learning_rate": 5.985998558704524e-05, "loss": 0.5425, "step": 1188 }, { "epoch": 0.12, "learning_rate": 5.985974795615005e-05, "loss": 0.773, "step": 1189 }, { "epoch": 0.12, "learning_rate": 5.985951012424603e-05, "loss": 1.1927, "step": 1190 }, { "epoch": 0.12, "learning_rate": 5.985927209133477e-05, "loss": 1.2437, "step": 1191 }, { "epoch": 0.12, "learning_rate": 5.985903385741788e-05, "loss": 0.217, "step": 1192 }, { "epoch": 0.12, "learning_rate": 5.985879542249697e-05, "loss": 0.2344, "step": 1193 }, { "epoch": 0.12, "learning_rate": 5.9858556786573615e-05, "loss": 0.4038, "step": 1194 }, { "epoch": 0.12, "learning_rate": 5.985831794964945e-05, "loss": 1.336, "step": 1195 }, { "epoch": 0.12, "learning_rate": 5.9858078911726087e-05, "loss": 1.3634, "step": 1196 }, { "epoch": 0.13, "learning_rate": 5.985783967280511e-05, "loss": 2.0036, "step": 1197 }, { "epoch": 0.13, "learning_rate": 5.9857600232888146e-05, "loss": 0.8594, "step": 1198 }, { "epoch": 0.13, "learning_rate": 5.98573605919768e-05, "loss": 0.5549, "step": 1199 }, { "epoch": 0.13, "learning_rate": 5.9857120750072694e-05, "loss": 0.6745, "step": 1200 }, { "epoch": 0.13, "learning_rate": 5.985688070717743e-05, "loss": 0.9977, "step": 1201 }, { "epoch": 0.13, "learning_rate": 5.985664046329264e-05, "loss": 0.3641, "step": 1202 }, { "epoch": 0.13, "learning_rate": 5.985640001841992e-05, "loss": 0.694, "step": 1203 }, { "epoch": 0.13, "learning_rate": 5.9856159372560915e-05, "loss": 0.7303, "step": 1204 }, { "epoch": 0.13, "learning_rate": 5.985591852571723e-05, "loss": 1.0499, "step": 1205 }, { "epoch": 0.13, "learning_rate": 5.985567747789048e-05, "loss": 1.6533, "step": 1206 }, { "epoch": 0.13, "learning_rate": 5.985543622908229e-05, "loss": 0.7294, "step": 1207 }, { "epoch": 0.13, "learning_rate": 5.9855194779294306e-05, "loss": 1.0752, "step": 1208 }, { "epoch": 0.13, "learning_rate": 5.985495312852812e-05, "loss": 0.7946, "step": 1209 }, { "epoch": 0.13, "learning_rate": 5.985471127678538e-05, "loss": 0.4772, "step": 1210 }, { "epoch": 0.13, "learning_rate": 5.985446922406771e-05, "loss": 0.3128, "step": 1211 }, { "epoch": 0.13, "learning_rate": 5.985422697037674e-05, "loss": 0.9903, "step": 1212 }, { "epoch": 0.13, "learning_rate": 5.9853984515714095e-05, "loss": 0.64, "step": 1213 }, { "epoch": 0.13, "learning_rate": 5.985374186008142e-05, "loss": 1.3386, "step": 1214 }, { "epoch": 0.13, "learning_rate": 5.985349900348033e-05, "loss": 1.006, "step": 1215 }, { "epoch": 0.13, "learning_rate": 5.985325594591247e-05, "loss": 0.2699, "step": 1216 }, { "epoch": 0.13, "learning_rate": 5.985301268737949e-05, "loss": 0.8569, "step": 1217 }, { "epoch": 0.13, "learning_rate": 5.9852769227883e-05, "loss": 0.9329, "step": 1218 }, { "epoch": 0.13, "learning_rate": 5.985252556742466e-05, "loss": 0.692, "step": 1219 }, { "epoch": 0.13, "learning_rate": 5.98522817060061e-05, "loss": 0.3395, "step": 1220 }, { "epoch": 0.13, "learning_rate": 5.9852037643628966e-05, "loss": 0.6811, "step": 1221 }, { "epoch": 0.13, "learning_rate": 5.98517933802949e-05, "loss": 1.001, "step": 1222 }, { "epoch": 0.13, "learning_rate": 5.985154891600554e-05, "loss": 0.6864, "step": 1223 }, { "epoch": 0.13, "learning_rate": 5.9851304250762544e-05, "loss": 1.0706, "step": 1224 }, { "epoch": 0.13, "learning_rate": 5.985105938456754e-05, "loss": 0.3466, "step": 1225 }, { "epoch": 0.13, "learning_rate": 5.985081431742221e-05, "loss": 0.5576, "step": 1226 }, { "epoch": 0.13, "learning_rate": 5.9850569049328164e-05, "loss": 1.3914, "step": 1227 }, { "epoch": 0.13, "learning_rate": 5.9850323580287075e-05, "loss": 0.5729, "step": 1228 }, { "epoch": 0.13, "learning_rate": 5.98500779103006e-05, "loss": 0.7202, "step": 1229 }, { "epoch": 0.13, "learning_rate": 5.9849832039370374e-05, "loss": 0.4054, "step": 1230 }, { "epoch": 0.13, "learning_rate": 5.984958596749807e-05, "loss": 1.2889, "step": 1231 }, { "epoch": 0.13, "learning_rate": 5.984933969468534e-05, "loss": 0.4262, "step": 1232 }, { "epoch": 0.13, "learning_rate": 5.9849093220933837e-05, "loss": 1.1361, "step": 1233 }, { "epoch": 0.13, "learning_rate": 5.984884654624521e-05, "loss": 0.2944, "step": 1234 }, { "epoch": 0.13, "learning_rate": 5.984859967062115e-05, "loss": 0.666, "step": 1235 }, { "epoch": 0.13, "learning_rate": 5.98483525940633e-05, "loss": 1.4752, "step": 1236 }, { "epoch": 0.13, "learning_rate": 5.984810531657331e-05, "loss": 0.5622, "step": 1237 }, { "epoch": 0.13, "learning_rate": 5.984785783815287e-05, "loss": 0.5619, "step": 1238 }, { "epoch": 0.13, "learning_rate": 5.9847610158803634e-05, "loss": 0.9058, "step": 1239 }, { "epoch": 0.13, "learning_rate": 5.984736227852727e-05, "loss": 1.2311, "step": 1240 }, { "epoch": 0.13, "learning_rate": 5.984711419732545e-05, "loss": 0.3434, "step": 1241 }, { "epoch": 0.13, "learning_rate": 5.984686591519983e-05, "loss": 1.208, "step": 1242 }, { "epoch": 0.13, "learning_rate": 5.984661743215211e-05, "loss": 0.3896, "step": 1243 }, { "epoch": 0.13, "learning_rate": 5.984636874818393e-05, "loss": 0.6472, "step": 1244 }, { "epoch": 0.13, "learning_rate": 5.984611986329698e-05, "loss": 0.7189, "step": 1245 }, { "epoch": 0.13, "learning_rate": 5.984587077749294e-05, "loss": 0.5349, "step": 1246 }, { "epoch": 0.13, "learning_rate": 5.984562149077347e-05, "loss": 0.4701, "step": 1247 }, { "epoch": 0.13, "learning_rate": 5.9845372003140274e-05, "loss": 1.5125, "step": 1248 }, { "epoch": 0.13, "learning_rate": 5.984512231459501e-05, "loss": 0.3107, "step": 1249 }, { "epoch": 0.13, "learning_rate": 5.9844872425139365e-05, "loss": 1.0643, "step": 1250 }, { "epoch": 0.13, "learning_rate": 5.984462233477503e-05, "loss": 0.6261, "step": 1251 }, { "epoch": 0.13, "learning_rate": 5.984437204350367e-05, "loss": 0.2364, "step": 1252 }, { "epoch": 0.13, "learning_rate": 5.9844121551326985e-05, "loss": 0.3732, "step": 1253 }, { "epoch": 0.13, "learning_rate": 5.984387085824666e-05, "loss": 1.2139, "step": 1254 }, { "epoch": 0.13, "learning_rate": 5.984361996426437e-05, "loss": 0.7385, "step": 1255 }, { "epoch": 0.13, "learning_rate": 5.9843368869381826e-05, "loss": 0.508, "step": 1256 }, { "epoch": 0.13, "learning_rate": 5.984311757360069e-05, "loss": 0.2037, "step": 1257 }, { "epoch": 0.13, "learning_rate": 5.984286607692268e-05, "loss": 0.6255, "step": 1258 }, { "epoch": 0.13, "learning_rate": 5.984261437934948e-05, "loss": 0.9984, "step": 1259 }, { "epoch": 0.13, "learning_rate": 5.984236248088278e-05, "loss": 0.8153, "step": 1260 }, { "epoch": 0.13, "learning_rate": 5.9842110381524275e-05, "loss": 0.8252, "step": 1261 }, { "epoch": 0.13, "learning_rate": 5.984185808127567e-05, "loss": 0.8212, "step": 1262 }, { "epoch": 0.13, "learning_rate": 5.984160558013865e-05, "loss": 0.8612, "step": 1263 }, { "epoch": 0.13, "learning_rate": 5.984135287811493e-05, "loss": 1.1241, "step": 1264 }, { "epoch": 0.13, "learning_rate": 5.9841099975206206e-05, "loss": 0.4508, "step": 1265 }, { "epoch": 0.13, "learning_rate": 5.984084687141418e-05, "loss": 1.2533, "step": 1266 }, { "epoch": 0.13, "learning_rate": 5.984059356674055e-05, "loss": 1.0168, "step": 1267 }, { "epoch": 0.13, "learning_rate": 5.984034006118703e-05, "loss": 1.1412, "step": 1268 }, { "epoch": 0.13, "learning_rate": 5.9840086354755327e-05, "loss": 0.5245, "step": 1269 }, { "epoch": 0.13, "learning_rate": 5.9839832447447137e-05, "loss": 0.4206, "step": 1270 }, { "epoch": 0.13, "learning_rate": 5.9839578339264175e-05, "loss": 0.5666, "step": 1271 }, { "epoch": 0.13, "learning_rate": 5.9839324030208157e-05, "loss": 0.5934, "step": 1272 }, { "epoch": 0.13, "learning_rate": 5.9839069520280795e-05, "loss": 0.4695, "step": 1273 }, { "epoch": 0.13, "learning_rate": 5.98388148094838e-05, "loss": 0.2459, "step": 1274 }, { "epoch": 0.13, "learning_rate": 5.983855989781888e-05, "loss": 1.1728, "step": 1275 }, { "epoch": 0.13, "learning_rate": 5.983830478528776e-05, "loss": 0.4833, "step": 1276 }, { "epoch": 0.13, "learning_rate": 5.983804947189215e-05, "loss": 0.7261, "step": 1277 }, { "epoch": 0.13, "learning_rate": 5.983779395763377e-05, "loss": 0.9932, "step": 1278 }, { "epoch": 0.13, "learning_rate": 5.983753824251435e-05, "loss": 1.2804, "step": 1279 }, { "epoch": 0.13, "learning_rate": 5.98372823265356e-05, "loss": 0.6192, "step": 1280 }, { "epoch": 0.13, "learning_rate": 5.983702620969925e-05, "loss": 1.5474, "step": 1281 }, { "epoch": 0.13, "learning_rate": 5.983676989200702e-05, "loss": 0.8216, "step": 1282 }, { "epoch": 0.13, "learning_rate": 5.983651337346063e-05, "loss": 1.374, "step": 1283 }, { "epoch": 0.13, "learning_rate": 5.983625665406182e-05, "loss": 0.9345, "step": 1284 }, { "epoch": 0.13, "learning_rate": 5.9835999733812315e-05, "loss": 0.569, "step": 1285 }, { "epoch": 0.13, "learning_rate": 5.983574261271383e-05, "loss": 0.5786, "step": 1286 }, { "epoch": 0.13, "learning_rate": 5.9835485290768114e-05, "loss": 0.5933, "step": 1287 }, { "epoch": 0.13, "learning_rate": 5.9835227767976896e-05, "loss": 0.834, "step": 1288 }, { "epoch": 0.13, "learning_rate": 5.98349700443419e-05, "loss": 1.3663, "step": 1289 }, { "epoch": 0.13, "learning_rate": 5.983471211986486e-05, "loss": 0.8746, "step": 1290 }, { "epoch": 0.13, "learning_rate": 5.9834453994547526e-05, "loss": 0.8415, "step": 1291 }, { "epoch": 0.13, "learning_rate": 5.983419566839163e-05, "loss": 0.402, "step": 1292 }, { "epoch": 0.14, "learning_rate": 5.983393714139891e-05, "loss": 0.521, "step": 1293 }, { "epoch": 0.14, "learning_rate": 5.9833678413571096e-05, "loss": 0.4162, "step": 1294 }, { "epoch": 0.14, "learning_rate": 5.983341948490995e-05, "loss": 1.3137, "step": 1295 }, { "epoch": 0.14, "learning_rate": 5.98331603554172e-05, "loss": 0.3423, "step": 1296 }, { "epoch": 0.14, "learning_rate": 5.9832901025094595e-05, "loss": 1.0009, "step": 1297 }, { "epoch": 0.14, "learning_rate": 5.983264149394388e-05, "loss": 0.5933, "step": 1298 }, { "epoch": 0.14, "learning_rate": 5.9832381761966805e-05, "loss": 0.4302, "step": 1299 }, { "epoch": 0.14, "learning_rate": 5.983212182916511e-05, "loss": 1.4529, "step": 1300 }, { "epoch": 0.14, "learning_rate": 5.983186169554056e-05, "loss": 0.3852, "step": 1301 }, { "epoch": 0.14, "learning_rate": 5.9831601361094897e-05, "loss": 0.7691, "step": 1302 }, { "epoch": 0.14, "learning_rate": 5.983134082582986e-05, "loss": 0.5276, "step": 1303 }, { "epoch": 0.14, "learning_rate": 5.9831080089747236e-05, "loss": 0.3893, "step": 1304 }, { "epoch": 0.14, "learning_rate": 5.9830819152848744e-05, "loss": 1.169, "step": 1305 }, { "epoch": 0.14, "learning_rate": 5.983055801513617e-05, "loss": 0.7897, "step": 1306 }, { "epoch": 0.14, "learning_rate": 5.9830296676611244e-05, "loss": 1.4289, "step": 1307 }, { "epoch": 0.14, "learning_rate": 5.983003513727575e-05, "loss": 0.7541, "step": 1308 }, { "epoch": 0.14, "learning_rate": 5.9829773397131445e-05, "loss": 0.4067, "step": 1309 }, { "epoch": 0.14, "learning_rate": 5.9829511456180074e-05, "loss": 0.6682, "step": 1310 }, { "epoch": 0.14, "learning_rate": 5.982924931442342e-05, "loss": 1.1786, "step": 1311 }, { "epoch": 0.14, "learning_rate": 5.9828986971863236e-05, "loss": 1.4293, "step": 1312 }, { "epoch": 0.14, "learning_rate": 5.982872442850129e-05, "loss": 0.7073, "step": 1313 }, { "epoch": 0.14, "learning_rate": 5.982846168433935e-05, "loss": 0.7201, "step": 1314 }, { "epoch": 0.14, "learning_rate": 5.9828198739379185e-05, "loss": 1.48, "step": 1315 }, { "epoch": 0.14, "learning_rate": 5.9827935593622576e-05, "loss": 0.7836, "step": 1316 }, { "epoch": 0.14, "learning_rate": 5.982767224707127e-05, "loss": 1.2925, "step": 1317 }, { "epoch": 0.14, "learning_rate": 5.9827408699727065e-05, "loss": 0.7653, "step": 1318 }, { "epoch": 0.14, "learning_rate": 5.982714495159171e-05, "loss": 0.9173, "step": 1319 }, { "epoch": 0.14, "learning_rate": 5.9826881002667e-05, "loss": 1.0206, "step": 1320 }, { "epoch": 0.14, "learning_rate": 5.982661685295471e-05, "loss": 0.6956, "step": 1321 }, { "epoch": 0.14, "learning_rate": 5.982635250245661e-05, "loss": 1.2531, "step": 1322 }, { "epoch": 0.14, "learning_rate": 5.982608795117449e-05, "loss": 0.698, "step": 1323 }, { "epoch": 0.14, "learning_rate": 5.982582319911012e-05, "loss": 0.9826, "step": 1324 }, { "epoch": 0.14, "learning_rate": 5.982555824626528e-05, "loss": 0.3365, "step": 1325 }, { "epoch": 0.14, "learning_rate": 5.982529309264177e-05, "loss": 0.7274, "step": 1326 }, { "epoch": 0.14, "learning_rate": 5.9825027738241355e-05, "loss": 0.7273, "step": 1327 }, { "epoch": 0.14, "learning_rate": 5.982476218306584e-05, "loss": 0.6551, "step": 1328 }, { "epoch": 0.14, "learning_rate": 5.9824496427117006e-05, "loss": 1.2101, "step": 1329 }, { "epoch": 0.14, "learning_rate": 5.982423047039664e-05, "loss": 0.7174, "step": 1330 }, { "epoch": 0.14, "learning_rate": 5.982396431290653e-05, "loss": 1.0655, "step": 1331 }, { "epoch": 0.14, "learning_rate": 5.9823697954648476e-05, "loss": 1.2041, "step": 1332 }, { "epoch": 0.14, "learning_rate": 5.982343139562427e-05, "loss": 1.0766, "step": 1333 }, { "epoch": 0.14, "learning_rate": 5.9823164635835685e-05, "loss": 1.1707, "step": 1334 }, { "epoch": 0.14, "learning_rate": 5.982289767528454e-05, "loss": 0.4054, "step": 1335 }, { "epoch": 0.14, "learning_rate": 5.982263051397263e-05, "loss": 1.0691, "step": 1336 }, { "epoch": 0.14, "learning_rate": 5.982236315190174e-05, "loss": 1.5927, "step": 1337 }, { "epoch": 0.14, "learning_rate": 5.982209558907369e-05, "loss": 0.331, "step": 1338 }, { "epoch": 0.14, "learning_rate": 5.982182782549027e-05, "loss": 1.1898, "step": 1339 }, { "epoch": 0.14, "learning_rate": 5.9821559861153276e-05, "loss": 0.7113, "step": 1340 }, { "epoch": 0.14, "learning_rate": 5.982129169606452e-05, "loss": 0.5259, "step": 1341 }, { "epoch": 0.14, "learning_rate": 5.9821023330225815e-05, "loss": 1.2389, "step": 1342 }, { "epoch": 0.14, "learning_rate": 5.982075476363895e-05, "loss": 1.5305, "step": 1343 }, { "epoch": 0.14, "learning_rate": 5.982048599630574e-05, "loss": 0.4522, "step": 1344 }, { "epoch": 0.14, "learning_rate": 5.9820217028227997e-05, "loss": 1.3086, "step": 1345 }, { "epoch": 0.14, "learning_rate": 5.981994785940754e-05, "loss": 1.1471, "step": 1346 }, { "epoch": 0.14, "learning_rate": 5.981967848984616e-05, "loss": 0.7488, "step": 1347 }, { "epoch": 0.14, "learning_rate": 5.9819408919545676e-05, "loss": 1.1559, "step": 1348 }, { "epoch": 0.14, "learning_rate": 5.9819139148507926e-05, "loss": 0.5817, "step": 1349 }, { "epoch": 0.14, "learning_rate": 5.98188691767347e-05, "loss": 1.1058, "step": 1350 }, { "epoch": 0.14, "learning_rate": 5.981859900422782e-05, "loss": 0.3626, "step": 1351 }, { "epoch": 0.14, "learning_rate": 5.981832863098911e-05, "loss": 1.178, "step": 1352 }, { "epoch": 0.14, "learning_rate": 5.9818058057020395e-05, "loss": 0.5928, "step": 1353 }, { "epoch": 0.14, "learning_rate": 5.981778728232348e-05, "loss": 1.223, "step": 1354 }, { "epoch": 0.14, "learning_rate": 5.981751630690021e-05, "loss": 1.1385, "step": 1355 }, { "epoch": 0.14, "learning_rate": 5.9817245130752394e-05, "loss": 0.7581, "step": 1356 }, { "epoch": 0.14, "learning_rate": 5.981697375388185e-05, "loss": 0.3809, "step": 1357 }, { "epoch": 0.14, "learning_rate": 5.9816702176290424e-05, "loss": 0.2603, "step": 1358 }, { "epoch": 0.14, "learning_rate": 5.981643039797993e-05, "loss": 1.0982, "step": 1359 }, { "epoch": 0.14, "learning_rate": 5.98161584189522e-05, "loss": 1.1443, "step": 1360 }, { "epoch": 0.14, "learning_rate": 5.981588623920909e-05, "loss": 0.5397, "step": 1361 }, { "epoch": 0.14, "learning_rate": 5.981561385875239e-05, "loss": 0.9112, "step": 1362 }, { "epoch": 0.14, "learning_rate": 5.9815341277583954e-05, "loss": 0.611, "step": 1363 }, { "epoch": 0.14, "learning_rate": 5.9815068495705624e-05, "loss": 0.7195, "step": 1364 }, { "epoch": 0.14, "learning_rate": 5.9814795513119225e-05, "loss": 1.2799, "step": 1365 }, { "epoch": 0.14, "learning_rate": 5.98145223298266e-05, "loss": 1.4077, "step": 1366 }, { "epoch": 0.14, "learning_rate": 5.981424894582959e-05, "loss": 0.4772, "step": 1367 }, { "epoch": 0.14, "learning_rate": 5.981397536113002e-05, "loss": 0.4487, "step": 1368 }, { "epoch": 0.14, "learning_rate": 5.981370157572976e-05, "loss": 0.3652, "step": 1369 }, { "epoch": 0.14, "learning_rate": 5.9813427589630625e-05, "loss": 0.9169, "step": 1370 }, { "epoch": 0.14, "learning_rate": 5.981315340283448e-05, "loss": 0.231, "step": 1371 }, { "epoch": 0.14, "learning_rate": 5.981287901534316e-05, "loss": 0.871, "step": 1372 }, { "epoch": 0.14, "learning_rate": 5.981260442715851e-05, "loss": 0.4194, "step": 1373 }, { "epoch": 0.14, "learning_rate": 5.9812329638282385e-05, "loss": 1.086, "step": 1374 }, { "epoch": 0.14, "learning_rate": 5.9812054648716626e-05, "loss": 0.8051, "step": 1375 }, { "epoch": 0.14, "learning_rate": 5.98117794584631e-05, "loss": 0.5963, "step": 1376 }, { "epoch": 0.14, "learning_rate": 5.981150406752364e-05, "loss": 0.6767, "step": 1377 }, { "epoch": 0.14, "learning_rate": 5.981122847590012e-05, "loss": 1.1542, "step": 1378 }, { "epoch": 0.14, "learning_rate": 5.9810952683594385e-05, "loss": 1.3777, "step": 1379 }, { "epoch": 0.14, "learning_rate": 5.981067669060829e-05, "loss": 1.0381, "step": 1380 }, { "epoch": 0.14, "learning_rate": 5.981040049694369e-05, "loss": 0.2294, "step": 1381 }, { "epoch": 0.14, "learning_rate": 5.981012410260246e-05, "loss": 0.9656, "step": 1382 }, { "epoch": 0.14, "learning_rate": 5.980984750758643e-05, "loss": 1.1624, "step": 1383 }, { "epoch": 0.14, "learning_rate": 5.98095707118975e-05, "loss": 0.2712, "step": 1384 }, { "epoch": 0.14, "learning_rate": 5.980929371553751e-05, "loss": 0.3935, "step": 1385 }, { "epoch": 0.14, "learning_rate": 5.9809016518508325e-05, "loss": 0.5584, "step": 1386 }, { "epoch": 0.14, "learning_rate": 5.980873912081182e-05, "loss": 1.3199, "step": 1387 }, { "epoch": 0.14, "learning_rate": 5.9808461522449856e-05, "loss": 0.5704, "step": 1388 }, { "epoch": 0.15, "learning_rate": 5.98081837234243e-05, "loss": 0.6953, "step": 1389 }, { "epoch": 0.15, "learning_rate": 5.980790572373703e-05, "loss": 1.1314, "step": 1390 }, { "epoch": 0.15, "learning_rate": 5.980762752338991e-05, "loss": 0.8206, "step": 1391 }, { "epoch": 0.15, "learning_rate": 5.9807349122384816e-05, "loss": 0.8191, "step": 1392 }, { "epoch": 0.15, "learning_rate": 5.980707052072363e-05, "loss": 1.1072, "step": 1393 }, { "epoch": 0.15, "learning_rate": 5.9806791718408216e-05, "loss": 0.8436, "step": 1394 }, { "epoch": 0.15, "learning_rate": 5.9806512715440447e-05, "loss": 1.2285, "step": 1395 }, { "epoch": 0.15, "learning_rate": 5.980623351182222e-05, "loss": 1.0717, "step": 1396 }, { "epoch": 0.15, "learning_rate": 5.9805954107555394e-05, "loss": 1.0274, "step": 1397 }, { "epoch": 0.15, "learning_rate": 5.9805674502641865e-05, "loss": 1.0197, "step": 1398 }, { "epoch": 0.15, "learning_rate": 5.98053946970835e-05, "loss": 0.5424, "step": 1399 }, { "epoch": 0.15, "learning_rate": 5.98051146908822e-05, "loss": 1.5222, "step": 1400 }, { "epoch": 0.15, "learning_rate": 5.980483448403984e-05, "loss": 1.616, "step": 1401 }, { "epoch": 0.15, "learning_rate": 5.980455407655831e-05, "loss": 0.8989, "step": 1402 }, { "epoch": 0.15, "learning_rate": 5.980427346843949e-05, "loss": 0.7539, "step": 1403 }, { "epoch": 0.15, "learning_rate": 5.980399265968528e-05, "loss": 0.7394, "step": 1404 }, { "epoch": 0.15, "learning_rate": 5.980371165029756e-05, "loss": 1.1468, "step": 1405 }, { "epoch": 0.15, "learning_rate": 5.980343044027824e-05, "loss": 0.6414, "step": 1406 }, { "epoch": 0.15, "learning_rate": 5.980314902962919e-05, "loss": 0.5358, "step": 1407 }, { "epoch": 0.15, "learning_rate": 5.98028674183523e-05, "loss": 0.7928, "step": 1408 }, { "epoch": 0.15, "learning_rate": 5.98025856064495e-05, "loss": 1.5331, "step": 1409 }, { "epoch": 0.15, "learning_rate": 5.980230359392267e-05, "loss": 1.1361, "step": 1410 }, { "epoch": 0.15, "learning_rate": 5.9802021380773685e-05, "loss": 0.5965, "step": 1411 }, { "epoch": 0.15, "learning_rate": 5.9801738967004476e-05, "loss": 0.2127, "step": 1412 }, { "epoch": 0.15, "learning_rate": 5.9801456352616936e-05, "loss": 0.508, "step": 1413 }, { "epoch": 0.15, "learning_rate": 5.980117353761296e-05, "loss": 0.7843, "step": 1414 }, { "epoch": 0.15, "learning_rate": 5.980089052199446e-05, "loss": 0.8532, "step": 1415 }, { "epoch": 0.15, "learning_rate": 5.980060730576334e-05, "loss": 0.6681, "step": 1416 }, { "epoch": 0.15, "learning_rate": 5.98003238889215e-05, "loss": 1.1472, "step": 1417 }, { "epoch": 0.15, "learning_rate": 5.980004027147085e-05, "loss": 0.5886, "step": 1418 }, { "epoch": 0.15, "learning_rate": 5.9799756453413304e-05, "loss": 0.8588, "step": 1419 }, { "epoch": 0.15, "learning_rate": 5.979947243475078e-05, "loss": 1.0718, "step": 1420 }, { "epoch": 0.15, "learning_rate": 5.9799188215485164e-05, "loss": 1.1616, "step": 1421 }, { "epoch": 0.15, "learning_rate": 5.9798903795618396e-05, "loss": 1.0331, "step": 1422 }, { "epoch": 0.15, "learning_rate": 5.979861917515237e-05, "loss": 0.4695, "step": 1423 }, { "epoch": 0.15, "learning_rate": 5.9798334354089024e-05, "loss": 0.9203, "step": 1424 }, { "epoch": 0.15, "learning_rate": 5.979804933243026e-05, "loss": 1.0782, "step": 1425 }, { "epoch": 0.15, "learning_rate": 5.9797764110177994e-05, "loss": 0.46, "step": 1426 }, { "epoch": 0.15, "learning_rate": 5.979747868733416e-05, "loss": 0.6677, "step": 1427 }, { "epoch": 0.15, "learning_rate": 5.9797193063900665e-05, "loss": 0.4221, "step": 1428 }, { "epoch": 0.15, "learning_rate": 5.979690723987945e-05, "loss": 0.7863, "step": 1429 }, { "epoch": 0.15, "learning_rate": 5.979662121527241e-05, "loss": 0.9599, "step": 1430 }, { "epoch": 0.15, "learning_rate": 5.97963349900815e-05, "loss": 1.1496, "step": 1431 }, { "epoch": 0.15, "learning_rate": 5.979604856430862e-05, "loss": 0.368, "step": 1432 }, { "epoch": 0.15, "learning_rate": 5.9795761937955735e-05, "loss": 1.1493, "step": 1433 }, { "epoch": 0.15, "learning_rate": 5.979547511102474e-05, "loss": 1.046, "step": 1434 }, { "epoch": 0.15, "learning_rate": 5.979518808351758e-05, "loss": 1.1624, "step": 1435 }, { "epoch": 0.15, "learning_rate": 5.979490085543618e-05, "loss": 0.4238, "step": 1436 }, { "epoch": 0.15, "learning_rate": 5.9794613426782484e-05, "loss": 0.6632, "step": 1437 }, { "epoch": 0.15, "learning_rate": 5.979432579755842e-05, "loss": 0.3699, "step": 1438 }, { "epoch": 0.15, "learning_rate": 5.979403796776593e-05, "loss": 1.3247, "step": 1439 }, { "epoch": 0.15, "learning_rate": 5.979374993740694e-05, "loss": 0.9463, "step": 1440 }, { "epoch": 0.15, "learning_rate": 5.9793461706483404e-05, "loss": 0.2914, "step": 1441 }, { "epoch": 0.15, "learning_rate": 5.979317327499725e-05, "loss": 0.7403, "step": 1442 }, { "epoch": 0.15, "learning_rate": 5.9792884642950426e-05, "loss": 0.9818, "step": 1443 }, { "epoch": 0.15, "learning_rate": 5.9792595810344874e-05, "loss": 0.4826, "step": 1444 }, { "epoch": 0.15, "learning_rate": 5.979230677718254e-05, "loss": 0.5136, "step": 1445 }, { "epoch": 0.15, "learning_rate": 5.979201754346537e-05, "loss": 0.6361, "step": 1446 }, { "epoch": 0.15, "learning_rate": 5.97917281091953e-05, "loss": 0.4955, "step": 1447 }, { "epoch": 0.15, "learning_rate": 5.979143847437429e-05, "loss": 1.4178, "step": 1448 }, { "epoch": 0.15, "learning_rate": 5.9791148639004294e-05, "loss": 0.2919, "step": 1449 }, { "epoch": 0.15, "learning_rate": 5.979085860308725e-05, "loss": 1.243, "step": 1450 }, { "epoch": 0.15, "learning_rate": 5.9790568366625115e-05, "loss": 0.6619, "step": 1451 }, { "epoch": 0.15, "learning_rate": 5.979027792961985e-05, "loss": 0.2704, "step": 1452 }, { "epoch": 0.15, "learning_rate": 5.978998729207339e-05, "loss": 0.4191, "step": 1453 }, { "epoch": 0.15, "learning_rate": 5.978969645398773e-05, "loss": 0.3443, "step": 1454 }, { "epoch": 0.15, "learning_rate": 5.9789405415364785e-05, "loss": 1.2468, "step": 1455 }, { "epoch": 0.15, "learning_rate": 5.9789114176206534e-05, "loss": 0.8373, "step": 1456 }, { "epoch": 0.15, "learning_rate": 5.978882273651494e-05, "loss": 0.197, "step": 1457 }, { "epoch": 0.15, "learning_rate": 5.978853109629196e-05, "loss": 0.8633, "step": 1458 }, { "epoch": 0.15, "learning_rate": 5.9788239255539565e-05, "loss": 0.4508, "step": 1459 }, { "epoch": 0.15, "learning_rate": 5.9787947214259714e-05, "loss": 0.9593, "step": 1460 }, { "epoch": 0.15, "learning_rate": 5.978765497245436e-05, "loss": 1.4033, "step": 1461 }, { "epoch": 0.15, "learning_rate": 5.978736253012549e-05, "loss": 0.5886, "step": 1462 }, { "epoch": 0.15, "learning_rate": 5.9787069887275067e-05, "loss": 0.7576, "step": 1463 }, { "epoch": 0.15, "learning_rate": 5.978677704390506e-05, "loss": 1.2092, "step": 1464 }, { "epoch": 0.15, "learning_rate": 5.978648400001743e-05, "loss": 1.2289, "step": 1465 }, { "epoch": 0.15, "learning_rate": 5.978619075561417e-05, "loss": 1.1411, "step": 1466 }, { "epoch": 0.15, "learning_rate": 5.978589731069723e-05, "loss": 1.3337, "step": 1467 }, { "epoch": 0.15, "learning_rate": 5.978560366526861e-05, "loss": 0.8942, "step": 1468 }, { "epoch": 0.15, "learning_rate": 5.978530981933028e-05, "loss": 0.5804, "step": 1469 }, { "epoch": 0.15, "learning_rate": 5.9785015772884204e-05, "loss": 0.8581, "step": 1470 }, { "epoch": 0.15, "learning_rate": 5.978472152593237e-05, "loss": 0.3049, "step": 1471 }, { "epoch": 0.15, "learning_rate": 5.978442707847676e-05, "loss": 1.3817, "step": 1472 }, { "epoch": 0.15, "learning_rate": 5.9784132430519366e-05, "loss": 0.472, "step": 1473 }, { "epoch": 0.15, "learning_rate": 5.9783837582062145e-05, "loss": 0.4696, "step": 1474 }, { "epoch": 0.15, "learning_rate": 5.978354253310711e-05, "loss": 0.5317, "step": 1475 }, { "epoch": 0.15, "learning_rate": 5.978324728365624e-05, "loss": 0.3223, "step": 1476 }, { "epoch": 0.15, "learning_rate": 5.9782951833711504e-05, "loss": 1.0923, "step": 1477 }, { "epoch": 0.15, "learning_rate": 5.9782656183274917e-05, "loss": 1.3893, "step": 1478 }, { "epoch": 0.15, "learning_rate": 5.978236033234844e-05, "loss": 0.4035, "step": 1479 }, { "epoch": 0.15, "learning_rate": 5.9782064280934096e-05, "loss": 0.5478, "step": 1480 }, { "epoch": 0.15, "learning_rate": 5.978176802903386e-05, "loss": 1.2457, "step": 1481 }, { "epoch": 0.15, "learning_rate": 5.9781471576649735e-05, "loss": 0.3713, "step": 1482 }, { "epoch": 0.15, "learning_rate": 5.9781174923783705e-05, "loss": 0.602, "step": 1483 }, { "epoch": 0.15, "learning_rate": 5.978087807043778e-05, "loss": 0.8071, "step": 1484 }, { "epoch": 0.16, "learning_rate": 5.978058101661395e-05, "loss": 0.2826, "step": 1485 }, { "epoch": 0.16, "learning_rate": 5.978028376231422e-05, "loss": 1.3517, "step": 1486 }, { "epoch": 0.16, "learning_rate": 5.977998630754058e-05, "loss": 0.5474, "step": 1487 }, { "epoch": 0.16, "learning_rate": 5.977968865229504e-05, "loss": 0.41, "step": 1488 }, { "epoch": 0.16, "learning_rate": 5.977939079657961e-05, "loss": 0.3781, "step": 1489 }, { "epoch": 0.16, "learning_rate": 5.977909274039628e-05, "loss": 1.1548, "step": 1490 }, { "epoch": 0.16, "learning_rate": 5.977879448374707e-05, "loss": 0.2953, "step": 1491 }, { "epoch": 0.16, "learning_rate": 5.9778496026633995e-05, "loss": 0.8904, "step": 1492 }, { "epoch": 0.16, "learning_rate": 5.977819736905903e-05, "loss": 1.0549, "step": 1493 }, { "epoch": 0.16, "learning_rate": 5.977789851102423e-05, "loss": 0.851, "step": 1494 }, { "epoch": 0.16, "learning_rate": 5.9777599452531566e-05, "loss": 0.4326, "step": 1495 }, { "epoch": 0.16, "learning_rate": 5.9777300193583086e-05, "loss": 0.7168, "step": 1496 }, { "epoch": 0.16, "learning_rate": 5.9777000734180776e-05, "loss": 0.4775, "step": 1497 }, { "epoch": 0.16, "learning_rate": 5.977670107432667e-05, "loss": 1.2521, "step": 1498 }, { "epoch": 0.16, "learning_rate": 5.977640121402278e-05, "loss": 1.1732, "step": 1499 }, { "epoch": 0.16, "learning_rate": 5.977610115327112e-05, "loss": 0.3223, "step": 1500 }, { "epoch": 0.16, "learning_rate": 5.977580089207372e-05, "loss": 0.3795, "step": 1501 }, { "epoch": 0.16, "learning_rate": 5.977550043043259e-05, "loss": 1.7981, "step": 1502 }, { "epoch": 0.16, "learning_rate": 5.977519976834976e-05, "loss": 1.064, "step": 1503 }, { "epoch": 0.16, "learning_rate": 5.9774898905827256e-05, "loss": 0.9727, "step": 1504 }, { "epoch": 0.16, "learning_rate": 5.977459784286709e-05, "loss": 1.0939, "step": 1505 }, { "epoch": 0.16, "learning_rate": 5.977429657947131e-05, "loss": 0.9203, "step": 1506 }, { "epoch": 0.16, "learning_rate": 5.977399511564192e-05, "loss": 0.3438, "step": 1507 }, { "epoch": 0.16, "learning_rate": 5.977369345138097e-05, "loss": 0.6396, "step": 1508 }, { "epoch": 0.16, "learning_rate": 5.977339158669048e-05, "loss": 0.7211, "step": 1509 }, { "epoch": 0.16, "learning_rate": 5.977308952157248e-05, "loss": 0.9261, "step": 1510 }, { "epoch": 0.16, "learning_rate": 5.977278725602902e-05, "loss": 0.8821, "step": 1511 }, { "epoch": 0.16, "learning_rate": 5.977248479006211e-05, "loss": 1.1679, "step": 1512 }, { "epoch": 0.16, "learning_rate": 5.9772182123673804e-05, "loss": 1.0774, "step": 1513 }, { "epoch": 0.16, "learning_rate": 5.977187925686613e-05, "loss": 0.471, "step": 1514 }, { "epoch": 0.16, "learning_rate": 5.977157618964113e-05, "loss": 0.255, "step": 1515 }, { "epoch": 0.16, "learning_rate": 5.977127292200085e-05, "loss": 1.0894, "step": 1516 }, { "epoch": 0.16, "learning_rate": 5.977096945394733e-05, "loss": 0.4679, "step": 1517 }, { "epoch": 0.16, "learning_rate": 5.97706657854826e-05, "loss": 0.689, "step": 1518 }, { "epoch": 0.16, "learning_rate": 5.977036191660872e-05, "loss": 1.0873, "step": 1519 }, { "epoch": 0.16, "learning_rate": 5.977005784732773e-05, "loss": 0.6783, "step": 1520 }, { "epoch": 0.16, "learning_rate": 5.976975357764167e-05, "loss": 0.6823, "step": 1521 }, { "epoch": 0.16, "learning_rate": 5.97694491075526e-05, "loss": 1.9225, "step": 1522 }, { "epoch": 0.16, "learning_rate": 5.976914443706256e-05, "loss": 1.219, "step": 1523 }, { "epoch": 0.16, "learning_rate": 5.976883956617361e-05, "loss": 1.1812, "step": 1524 }, { "epoch": 0.16, "learning_rate": 5.976853449488779e-05, "loss": 1.0911, "step": 1525 }, { "epoch": 0.16, "learning_rate": 5.976822922320717e-05, "loss": 0.922, "step": 1526 }, { "epoch": 0.16, "learning_rate": 5.976792375113379e-05, "loss": 0.6178, "step": 1527 }, { "epoch": 0.16, "learning_rate": 5.976761807866972e-05, "loss": 0.5219, "step": 1528 }, { "epoch": 0.16, "learning_rate": 5.9767312205817e-05, "loss": 0.6421, "step": 1529 }, { "epoch": 0.16, "learning_rate": 5.976700613257771e-05, "loss": 1.0532, "step": 1530 }, { "epoch": 0.16, "learning_rate": 5.976669985895388e-05, "loss": 0.5525, "step": 1531 }, { "epoch": 0.16, "learning_rate": 5.976639338494762e-05, "loss": 0.5042, "step": 1532 }, { "epoch": 0.16, "learning_rate": 5.976608671056094e-05, "loss": 0.1582, "step": 1533 }, { "epoch": 0.16, "learning_rate": 5.976577983579594e-05, "loss": 0.5363, "step": 1534 }, { "epoch": 0.16, "learning_rate": 5.9765472760654674e-05, "loss": 1.3239, "step": 1535 }, { "epoch": 0.16, "learning_rate": 5.976516548513921e-05, "loss": 0.5738, "step": 1536 }, { "epoch": 0.16, "learning_rate": 5.976485800925161e-05, "loss": 1.4471, "step": 1537 }, { "epoch": 0.16, "learning_rate": 5.9764550332993965e-05, "loss": 0.6933, "step": 1538 }, { "epoch": 0.16, "learning_rate": 5.9764242456368316e-05, "loss": 0.7901, "step": 1539 }, { "epoch": 0.16, "learning_rate": 5.976393437937676e-05, "loss": 0.5203, "step": 1540 }, { "epoch": 0.16, "learning_rate": 5.9763626102021354e-05, "loss": 0.5729, "step": 1541 }, { "epoch": 0.16, "learning_rate": 5.976331762430419e-05, "loss": 1.0259, "step": 1542 }, { "epoch": 0.16, "learning_rate": 5.9763008946227335e-05, "loss": 0.878, "step": 1543 }, { "epoch": 0.16, "learning_rate": 5.976270006779286e-05, "loss": 1.2225, "step": 1544 }, { "epoch": 0.16, "learning_rate": 5.9762390989002854e-05, "loss": 0.8482, "step": 1545 }, { "epoch": 0.16, "learning_rate": 5.976208170985941e-05, "loss": 0.4681, "step": 1546 }, { "epoch": 0.16, "learning_rate": 5.976177223036458e-05, "loss": 0.4437, "step": 1547 }, { "epoch": 0.16, "learning_rate": 5.976146255052047e-05, "loss": 1.2344, "step": 1548 }, { "epoch": 0.16, "learning_rate": 5.976115267032915e-05, "loss": 0.728, "step": 1549 }, { "epoch": 0.16, "learning_rate": 5.976084258979273e-05, "loss": 0.5192, "step": 1550 }, { "epoch": 0.16, "learning_rate": 5.976053230891326e-05, "loss": 0.5881, "step": 1551 }, { "epoch": 0.16, "learning_rate": 5.976022182769286e-05, "loss": 1.0447, "step": 1552 }, { "epoch": 0.16, "learning_rate": 5.9759911146133614e-05, "loss": 0.428, "step": 1553 }, { "epoch": 0.16, "learning_rate": 5.97596002642376e-05, "loss": 1.1781, "step": 1554 }, { "epoch": 0.16, "learning_rate": 5.975928918200693e-05, "loss": 0.5347, "step": 1555 }, { "epoch": 0.16, "learning_rate": 5.9758977899443686e-05, "loss": 0.653, "step": 1556 }, { "epoch": 0.16, "learning_rate": 5.975866641654996e-05, "loss": 0.7855, "step": 1557 }, { "epoch": 0.16, "learning_rate": 5.975835473332786e-05, "loss": 0.6045, "step": 1558 }, { "epoch": 0.16, "learning_rate": 5.9758042849779476e-05, "loss": 0.333, "step": 1559 }, { "epoch": 0.16, "learning_rate": 5.975773076590692e-05, "loss": 0.6349, "step": 1560 }, { "epoch": 0.16, "learning_rate": 5.9757418481712266e-05, "loss": 0.5512, "step": 1561 }, { "epoch": 0.16, "learning_rate": 5.975710599719764e-05, "loss": 0.4693, "step": 1562 }, { "epoch": 0.16, "learning_rate": 5.975679331236515e-05, "loss": 1.3529, "step": 1563 }, { "epoch": 0.16, "learning_rate": 5.975648042721688e-05, "loss": 0.6186, "step": 1564 }, { "epoch": 0.16, "learning_rate": 5.975616734175495e-05, "loss": 1.0524, "step": 1565 }, { "epoch": 0.16, "learning_rate": 5.975585405598146e-05, "loss": 1.1164, "step": 1566 }, { "epoch": 0.16, "learning_rate": 5.975554056989852e-05, "loss": 0.3714, "step": 1567 }, { "epoch": 0.16, "learning_rate": 5.975522688350825e-05, "loss": 1.1202, "step": 1568 }, { "epoch": 0.16, "learning_rate": 5.975491299681275e-05, "loss": 0.421, "step": 1569 }, { "epoch": 0.16, "learning_rate": 5.975459890981414e-05, "loss": 1.2126, "step": 1570 }, { "epoch": 0.16, "learning_rate": 5.975428462251453e-05, "loss": 0.7812, "step": 1571 }, { "epoch": 0.16, "learning_rate": 5.975397013491605e-05, "loss": 0.3223, "step": 1572 }, { "epoch": 0.16, "learning_rate": 5.9753655447020785e-05, "loss": 1.4873, "step": 1573 }, { "epoch": 0.16, "learning_rate": 5.9753340558830894e-05, "loss": 0.4398, "step": 1574 }, { "epoch": 0.16, "learning_rate": 5.975302547034846e-05, "loss": 1.1511, "step": 1575 }, { "epoch": 0.16, "learning_rate": 5.975271018157563e-05, "loss": 0.7895, "step": 1576 }, { "epoch": 0.16, "learning_rate": 5.9752394692514504e-05, "loss": 0.7781, "step": 1577 }, { "epoch": 0.16, "learning_rate": 5.975207900316723e-05, "loss": 0.8233, "step": 1578 }, { "epoch": 0.16, "learning_rate": 5.9751763113535915e-05, "loss": 0.5458, "step": 1579 }, { "epoch": 0.17, "learning_rate": 5.97514470236227e-05, "loss": 0.2077, "step": 1580 }, { "epoch": 0.17, "learning_rate": 5.97511307334297e-05, "loss": 1.0683, "step": 1581 }, { "epoch": 0.17, "learning_rate": 5.975081424295904e-05, "loss": 0.7868, "step": 1582 }, { "epoch": 0.17, "learning_rate": 5.975049755221288e-05, "loss": 0.4659, "step": 1583 }, { "epoch": 0.17, "learning_rate": 5.975018066119332e-05, "loss": 0.8735, "step": 1584 }, { "epoch": 0.17, "learning_rate": 5.9749863569902504e-05, "loss": 1.7026, "step": 1585 }, { "epoch": 0.17, "learning_rate": 5.974954627834257e-05, "loss": 1.1618, "step": 1586 }, { "epoch": 0.17, "learning_rate": 5.9749228786515645e-05, "loss": 0.8507, "step": 1587 }, { "epoch": 0.17, "learning_rate": 5.974891109442388e-05, "loss": 0.291, "step": 1588 }, { "epoch": 0.17, "learning_rate": 5.97485932020694e-05, "loss": 0.9365, "step": 1589 }, { "epoch": 0.17, "learning_rate": 5.974827510945436e-05, "loss": 1.0477, "step": 1590 }, { "epoch": 0.17, "learning_rate": 5.974795681658089e-05, "loss": 0.4357, "step": 1591 }, { "epoch": 0.17, "learning_rate": 5.974763832345113e-05, "loss": 0.2854, "step": 1592 }, { "epoch": 0.17, "learning_rate": 5.9747319630067235e-05, "loss": 0.5808, "step": 1593 }, { "epoch": 0.17, "learning_rate": 5.974700073643134e-05, "loss": 0.4915, "step": 1594 }, { "epoch": 0.17, "learning_rate": 5.97466816425456e-05, "loss": 0.7013, "step": 1595 }, { "epoch": 0.17, "learning_rate": 5.974636234841216e-05, "loss": 0.8173, "step": 1596 }, { "epoch": 0.17, "learning_rate": 5.974604285403316e-05, "loss": 0.4253, "step": 1597 }, { "epoch": 0.17, "learning_rate": 5.974572315941076e-05, "loss": 0.3929, "step": 1598 }, { "epoch": 0.17, "learning_rate": 5.9745403264547126e-05, "loss": 0.8833, "step": 1599 }, { "epoch": 0.17, "learning_rate": 5.974508316944438e-05, "loss": 0.2229, "step": 1600 }, { "epoch": 0.17, "learning_rate": 5.97447628741047e-05, "loss": 0.8667, "step": 1601 }, { "epoch": 0.17, "learning_rate": 5.974444237853025e-05, "loss": 1.2432, "step": 1602 }, { "epoch": 0.17, "learning_rate": 5.974412168272315e-05, "loss": 1.0144, "step": 1603 }, { "epoch": 0.17, "learning_rate": 5.97438007866856e-05, "loss": 1.0498, "step": 1604 }, { "epoch": 0.17, "learning_rate": 5.974347969041974e-05, "loss": 0.7871, "step": 1605 }, { "epoch": 0.17, "learning_rate": 5.974315839392773e-05, "loss": 0.718, "step": 1606 }, { "epoch": 0.17, "learning_rate": 5.974283689721173e-05, "loss": 0.8207, "step": 1607 }, { "epoch": 0.17, "learning_rate": 5.974251520027393e-05, "loss": 1.9797, "step": 1608 }, { "epoch": 0.17, "learning_rate": 5.9742193303116466e-05, "loss": 0.5761, "step": 1609 }, { "epoch": 0.17, "learning_rate": 5.97418712057415e-05, "loss": 0.3746, "step": 1610 }, { "epoch": 0.17, "learning_rate": 5.974154890815124e-05, "loss": 1.2391, "step": 1611 }, { "epoch": 0.17, "learning_rate": 5.974122641034783e-05, "loss": 0.7473, "step": 1612 }, { "epoch": 0.17, "learning_rate": 5.974090371233343e-05, "loss": 0.3672, "step": 1613 }, { "epoch": 0.17, "learning_rate": 5.9740580814110234e-05, "loss": 0.9086, "step": 1614 }, { "epoch": 0.17, "learning_rate": 5.97402577156804e-05, "loss": 0.8705, "step": 1615 }, { "epoch": 0.17, "learning_rate": 5.973993441704612e-05, "loss": 0.4229, "step": 1616 }, { "epoch": 0.17, "learning_rate": 5.973961091820955e-05, "loss": 1.1825, "step": 1617 }, { "epoch": 0.17, "learning_rate": 5.973928721917288e-05, "loss": 0.5196, "step": 1618 }, { "epoch": 0.17, "learning_rate": 5.973896331993829e-05, "loss": 0.9256, "step": 1619 }, { "epoch": 0.17, "learning_rate": 5.9738639220507956e-05, "loss": 1.0848, "step": 1620 }, { "epoch": 0.17, "learning_rate": 5.973831492088406e-05, "loss": 0.4907, "step": 1621 }, { "epoch": 0.17, "learning_rate": 5.973799042106879e-05, "loss": 0.9587, "step": 1622 }, { "epoch": 0.17, "learning_rate": 5.973766572106433e-05, "loss": 1.2298, "step": 1623 }, { "epoch": 0.17, "learning_rate": 5.973734082087285e-05, "loss": 1.1177, "step": 1624 }, { "epoch": 0.17, "learning_rate": 5.9737015720496554e-05, "loss": 1.2356, "step": 1625 }, { "epoch": 0.17, "learning_rate": 5.973669041993762e-05, "loss": 0.904, "step": 1626 }, { "epoch": 0.17, "learning_rate": 5.973636491919826e-05, "loss": 0.5165, "step": 1627 }, { "epoch": 0.17, "learning_rate": 5.973603921828064e-05, "loss": 0.3327, "step": 1628 }, { "epoch": 0.17, "learning_rate": 5.9735713317186964e-05, "loss": 1.0423, "step": 1629 }, { "epoch": 0.17, "learning_rate": 5.973538721591942e-05, "loss": 0.2924, "step": 1630 }, { "epoch": 0.17, "learning_rate": 5.97350609144802e-05, "loss": 1.5649, "step": 1631 }, { "epoch": 0.17, "learning_rate": 5.973473441287152e-05, "loss": 1.259, "step": 1632 }, { "epoch": 0.17, "learning_rate": 5.973440771109557e-05, "loss": 1.0439, "step": 1633 }, { "epoch": 0.17, "learning_rate": 5.9734080809154535e-05, "loss": 1.0087, "step": 1634 }, { "epoch": 0.17, "learning_rate": 5.973375370705062e-05, "loss": 0.7198, "step": 1635 }, { "epoch": 0.17, "learning_rate": 5.973342640478604e-05, "loss": 1.1978, "step": 1636 }, { "epoch": 0.17, "learning_rate": 5.973309890236299e-05, "loss": 0.3805, "step": 1637 }, { "epoch": 0.17, "learning_rate": 5.973277119978368e-05, "loss": 0.3001, "step": 1638 }, { "epoch": 0.17, "learning_rate": 5.9732443297050296e-05, "loss": 0.5086, "step": 1639 }, { "epoch": 0.17, "learning_rate": 5.973211519416507e-05, "loss": 0.5727, "step": 1640 }, { "epoch": 0.17, "learning_rate": 5.97317868911302e-05, "loss": 0.4515, "step": 1641 }, { "epoch": 0.17, "learning_rate": 5.97314583879479e-05, "loss": 0.299, "step": 1642 }, { "epoch": 0.17, "learning_rate": 5.973112968462038e-05, "loss": 0.2583, "step": 1643 }, { "epoch": 0.17, "learning_rate": 5.973080078114985e-05, "loss": 1.3761, "step": 1644 }, { "epoch": 0.17, "learning_rate": 5.9730471677538525e-05, "loss": 0.8374, "step": 1645 }, { "epoch": 0.17, "learning_rate": 5.9730142373788613e-05, "loss": 0.9351, "step": 1646 }, { "epoch": 0.17, "learning_rate": 5.972981286990235e-05, "loss": 0.5975, "step": 1647 }, { "epoch": 0.17, "learning_rate": 5.972948316588194e-05, "loss": 0.6283, "step": 1648 }, { "epoch": 0.17, "learning_rate": 5.9729153261729605e-05, "loss": 0.853, "step": 1649 }, { "epoch": 0.17, "learning_rate": 5.972882315744757e-05, "loss": 0.562, "step": 1650 }, { "epoch": 0.17, "learning_rate": 5.9728492853038046e-05, "loss": 0.7759, "step": 1651 }, { "epoch": 0.17, "learning_rate": 5.972816234850327e-05, "loss": 0.7603, "step": 1652 }, { "epoch": 0.17, "learning_rate": 5.9727831643845456e-05, "loss": 0.7458, "step": 1653 }, { "epoch": 0.17, "learning_rate": 5.9727500739066835e-05, "loss": 0.1512, "step": 1654 }, { "epoch": 0.17, "learning_rate": 5.972716963416964e-05, "loss": 0.9311, "step": 1655 }, { "epoch": 0.17, "learning_rate": 5.9726838329156085e-05, "loss": 1.7829, "step": 1656 }, { "epoch": 0.17, "learning_rate": 5.9726506824028415e-05, "loss": 1.0362, "step": 1657 }, { "epoch": 0.17, "learning_rate": 5.972617511878886e-05, "loss": 0.3984, "step": 1658 }, { "epoch": 0.17, "learning_rate": 5.972584321343965e-05, "loss": 1.3187, "step": 1659 }, { "epoch": 0.17, "learning_rate": 5.972551110798301e-05, "loss": 0.8627, "step": 1660 }, { "epoch": 0.17, "learning_rate": 5.97251788024212e-05, "loss": 0.8153, "step": 1661 }, { "epoch": 0.17, "learning_rate": 5.9724846296756435e-05, "loss": 0.5877, "step": 1662 }, { "epoch": 0.17, "learning_rate": 5.9724513590990957e-05, "loss": 0.3016, "step": 1663 }, { "epoch": 0.17, "learning_rate": 5.9724180685127e-05, "loss": 1.235, "step": 1664 }, { "epoch": 0.17, "learning_rate": 5.972384757916682e-05, "loss": 1.1945, "step": 1665 }, { "epoch": 0.17, "learning_rate": 5.972351427311266e-05, "loss": 1.0541, "step": 1666 }, { "epoch": 0.17, "learning_rate": 5.9723180766966754e-05, "loss": 0.7977, "step": 1667 }, { "epoch": 0.17, "learning_rate": 5.9722847060731346e-05, "loss": 1.1118, "step": 1668 }, { "epoch": 0.17, "learning_rate": 5.972251315440869e-05, "loss": 0.7747, "step": 1669 }, { "epoch": 0.17, "learning_rate": 5.972217904800102e-05, "loss": 0.9693, "step": 1670 }, { "epoch": 0.17, "learning_rate": 5.972184474151061e-05, "loss": 1.1743, "step": 1671 }, { "epoch": 0.17, "learning_rate": 5.972151023493969e-05, "loss": 1.3602, "step": 1672 }, { "epoch": 0.17, "learning_rate": 5.972117552829052e-05, "loss": 0.7116, "step": 1673 }, { "epoch": 0.17, "learning_rate": 5.9720840621565335e-05, "loss": 0.4143, "step": 1674 }, { "epoch": 0.17, "learning_rate": 5.972050551476642e-05, "loss": 0.8956, "step": 1675 }, { "epoch": 0.18, "learning_rate": 5.972017020789601e-05, "loss": 0.8287, "step": 1676 }, { "epoch": 0.18, "learning_rate": 5.971983470095637e-05, "loss": 0.8996, "step": 1677 }, { "epoch": 0.18, "learning_rate": 5.971949899394976e-05, "loss": 0.4245, "step": 1678 }, { "epoch": 0.18, "learning_rate": 5.971916308687843e-05, "loss": 0.7766, "step": 1679 }, { "epoch": 0.18, "learning_rate": 5.971882697974465e-05, "loss": 0.756, "step": 1680 }, { "epoch": 0.18, "learning_rate": 5.971849067255069e-05, "loss": 1.7485, "step": 1681 }, { "epoch": 0.18, "learning_rate": 5.971815416529879e-05, "loss": 0.8297, "step": 1682 }, { "epoch": 0.18, "learning_rate": 5.9717817457991236e-05, "loss": 0.6734, "step": 1683 }, { "epoch": 0.18, "learning_rate": 5.9717480550630285e-05, "loss": 0.6945, "step": 1684 }, { "epoch": 0.18, "learning_rate": 5.971714344321822e-05, "loss": 0.8739, "step": 1685 }, { "epoch": 0.18, "learning_rate": 5.9716806135757284e-05, "loss": 0.9211, "step": 1686 }, { "epoch": 0.18, "learning_rate": 5.9716468628249774e-05, "loss": 1.3445, "step": 1687 }, { "epoch": 0.18, "learning_rate": 5.9716130920697944e-05, "loss": 0.4089, "step": 1688 }, { "epoch": 0.18, "learning_rate": 5.971579301310407e-05, "loss": 0.3616, "step": 1689 }, { "epoch": 0.18, "learning_rate": 5.971545490547044e-05, "loss": 1.0587, "step": 1690 }, { "epoch": 0.18, "learning_rate": 5.971511659779931e-05, "loss": 0.3761, "step": 1691 }, { "epoch": 0.18, "learning_rate": 5.971477809009298e-05, "loss": 1.2383, "step": 1692 }, { "epoch": 0.18, "learning_rate": 5.97144393823537e-05, "loss": 1.0414, "step": 1693 }, { "epoch": 0.18, "learning_rate": 5.9714100474583776e-05, "loss": 0.707, "step": 1694 }, { "epoch": 0.18, "learning_rate": 5.9713761366785484e-05, "loss": 0.3756, "step": 1695 }, { "epoch": 0.18, "learning_rate": 5.97134220589611e-05, "loss": 1.4774, "step": 1696 }, { "epoch": 0.18, "learning_rate": 5.971308255111291e-05, "loss": 0.8555, "step": 1697 }, { "epoch": 0.18, "learning_rate": 5.9712742843243196e-05, "loss": 1.0747, "step": 1698 }, { "epoch": 0.18, "learning_rate": 5.971240293535426e-05, "loss": 0.6551, "step": 1699 }, { "epoch": 0.18, "learning_rate": 5.971206282744837e-05, "loss": 1.1192, "step": 1700 }, { "epoch": 0.18, "learning_rate": 5.9711722519527834e-05, "loss": 0.4741, "step": 1701 }, { "epoch": 0.18, "learning_rate": 5.971138201159492e-05, "loss": 1.599, "step": 1702 }, { "epoch": 0.18, "learning_rate": 5.9711041303651944e-05, "loss": 1.1942, "step": 1703 }, { "epoch": 0.18, "learning_rate": 5.971070039570119e-05, "loss": 1.4976, "step": 1704 }, { "epoch": 0.18, "learning_rate": 5.971035928774495e-05, "loss": 1.2224, "step": 1705 }, { "epoch": 0.18, "learning_rate": 5.971001797978553e-05, "loss": 1.2302, "step": 1706 }, { "epoch": 0.18, "learning_rate": 5.970967647182521e-05, "loss": 1.2249, "step": 1707 }, { "epoch": 0.18, "learning_rate": 5.9709334763866307e-05, "loss": 0.7659, "step": 1708 }, { "epoch": 0.18, "learning_rate": 5.970899285591111e-05, "loss": 0.7153, "step": 1709 }, { "epoch": 0.18, "learning_rate": 5.9708650747961924e-05, "loss": 0.5828, "step": 1710 }, { "epoch": 0.18, "learning_rate": 5.970830844002106e-05, "loss": 1.5408, "step": 1711 }, { "epoch": 0.18, "learning_rate": 5.9707965932090806e-05, "loss": 0.639, "step": 1712 }, { "epoch": 0.18, "learning_rate": 5.9707623224173484e-05, "loss": 1.2409, "step": 1713 }, { "epoch": 0.18, "learning_rate": 5.9707280316271385e-05, "loss": 0.4322, "step": 1714 }, { "epoch": 0.18, "learning_rate": 5.970693720838683e-05, "loss": 1.4228, "step": 1715 }, { "epoch": 0.18, "learning_rate": 5.9706593900522125e-05, "loss": 0.3335, "step": 1716 }, { "epoch": 0.18, "learning_rate": 5.9706250392679586e-05, "loss": 1.204, "step": 1717 }, { "epoch": 0.18, "learning_rate": 5.9705906684861513e-05, "loss": 0.5674, "step": 1718 }, { "epoch": 0.18, "learning_rate": 5.9705562777070224e-05, "loss": 1.0959, "step": 1719 }, { "epoch": 0.18, "learning_rate": 5.9705218669308035e-05, "loss": 1.3938, "step": 1720 }, { "epoch": 0.18, "learning_rate": 5.970487436157728e-05, "loss": 0.721, "step": 1721 }, { "epoch": 0.18, "learning_rate": 5.970452985388024e-05, "loss": 0.4895, "step": 1722 }, { "epoch": 0.18, "learning_rate": 5.970418514621927e-05, "loss": 0.3465, "step": 1723 }, { "epoch": 0.18, "learning_rate": 5.970384023859667e-05, "loss": 0.5106, "step": 1724 }, { "epoch": 0.18, "learning_rate": 5.9703495131014775e-05, "loss": 1.6342, "step": 1725 }, { "epoch": 0.18, "learning_rate": 5.970314982347589e-05, "loss": 1.1044, "step": 1726 }, { "epoch": 0.18, "learning_rate": 5.9702804315982366e-05, "loss": 0.456, "step": 1727 }, { "epoch": 0.18, "learning_rate": 5.97024586085365e-05, "loss": 0.3815, "step": 1728 }, { "epoch": 0.18, "learning_rate": 5.970211270114064e-05, "loss": 0.3734, "step": 1729 }, { "epoch": 0.18, "learning_rate": 5.97017665937971e-05, "loss": 0.5895, "step": 1730 }, { "epoch": 0.18, "learning_rate": 5.970142028650822e-05, "loss": 0.2096, "step": 1731 }, { "epoch": 0.18, "learning_rate": 5.970107377927633e-05, "loss": 1.3277, "step": 1732 }, { "epoch": 0.18, "learning_rate": 5.970072707210376e-05, "loss": 1.3938, "step": 1733 }, { "epoch": 0.18, "learning_rate": 5.9700380164992844e-05, "loss": 0.3548, "step": 1734 }, { "epoch": 0.18, "learning_rate": 5.9700033057945915e-05, "loss": 1.0703, "step": 1735 }, { "epoch": 0.18, "learning_rate": 5.969968575096532e-05, "loss": 1.5015, "step": 1736 }, { "epoch": 0.18, "learning_rate": 5.9699338244053384e-05, "loss": 0.6766, "step": 1737 }, { "epoch": 0.18, "learning_rate": 5.969899053721245e-05, "loss": 0.1925, "step": 1738 }, { "epoch": 0.18, "learning_rate": 5.969864263044487e-05, "loss": 1.3792, "step": 1739 }, { "epoch": 0.18, "learning_rate": 5.9698294523752964e-05, "loss": 1.3855, "step": 1740 }, { "epoch": 0.18, "learning_rate": 5.9697946217139094e-05, "loss": 0.5699, "step": 1741 }, { "epoch": 0.18, "learning_rate": 5.969759771060561e-05, "loss": 1.3341, "step": 1742 }, { "epoch": 0.18, "learning_rate": 5.969724900415483e-05, "loss": 0.7241, "step": 1743 }, { "epoch": 0.18, "learning_rate": 5.969690009778912e-05, "loss": 1.8137, "step": 1744 }, { "epoch": 0.18, "learning_rate": 5.969655099151084e-05, "loss": 0.6835, "step": 1745 }, { "epoch": 0.18, "learning_rate": 5.969620168532232e-05, "loss": 1.0058, "step": 1746 }, { "epoch": 0.18, "learning_rate": 5.969585217922591e-05, "loss": 1.0807, "step": 1747 }, { "epoch": 0.18, "learning_rate": 5.969550247322398e-05, "loss": 1.1109, "step": 1748 }, { "epoch": 0.18, "learning_rate": 5.9695152567318876e-05, "loss": 0.5191, "step": 1749 }, { "epoch": 0.18, "learning_rate": 5.969480246151295e-05, "loss": 0.7609, "step": 1750 }, { "epoch": 0.18, "learning_rate": 5.9694452155808564e-05, "loss": 0.9006, "step": 1751 }, { "epoch": 0.18, "learning_rate": 5.969410165020808e-05, "loss": 1.3093, "step": 1752 }, { "epoch": 0.18, "learning_rate": 5.969375094471384e-05, "loss": 0.454, "step": 1753 }, { "epoch": 0.18, "learning_rate": 5.969340003932823e-05, "loss": 1.1145, "step": 1754 }, { "epoch": 0.18, "learning_rate": 5.9693048934053586e-05, "loss": 1.1795, "step": 1755 }, { "epoch": 0.18, "learning_rate": 5.969269762889229e-05, "loss": 0.5097, "step": 1756 }, { "epoch": 0.18, "learning_rate": 5.9692346123846697e-05, "loss": 1.3648, "step": 1757 }, { "epoch": 0.18, "learning_rate": 5.9691994418919186e-05, "loss": 0.2325, "step": 1758 }, { "epoch": 0.18, "learning_rate": 5.969164251411211e-05, "loss": 0.9295, "step": 1759 }, { "epoch": 0.18, "learning_rate": 5.9691290409427844e-05, "loss": 0.484, "step": 1760 }, { "epoch": 0.18, "learning_rate": 5.9690938104868756e-05, "loss": 0.514, "step": 1761 }, { "epoch": 0.18, "learning_rate": 5.969058560043722e-05, "loss": 1.517, "step": 1762 }, { "epoch": 0.18, "learning_rate": 5.969023289613561e-05, "loss": 1.1225, "step": 1763 }, { "epoch": 0.18, "learning_rate": 5.96898799919663e-05, "loss": 0.7818, "step": 1764 }, { "epoch": 0.18, "learning_rate": 5.968952688793166e-05, "loss": 0.9452, "step": 1765 }, { "epoch": 0.18, "learning_rate": 5.968917358403408e-05, "loss": 1.1663, "step": 1766 }, { "epoch": 0.18, "learning_rate": 5.9688820080275926e-05, "loss": 0.9642, "step": 1767 }, { "epoch": 0.18, "learning_rate": 5.968846637665958e-05, "loss": 1.3435, "step": 1768 }, { "epoch": 0.18, "learning_rate": 5.968811247318743e-05, "loss": 0.3946, "step": 1769 }, { "epoch": 0.18, "learning_rate": 5.968775836986184e-05, "loss": 1.5311, "step": 1770 }, { "epoch": 0.18, "learning_rate": 5.968740406668523e-05, "loss": 1.2892, "step": 1771 }, { "epoch": 0.19, "learning_rate": 5.9687049563659945e-05, "loss": 1.1043, "step": 1772 }, { "epoch": 0.19, "learning_rate": 5.96866948607884e-05, "loss": 0.7045, "step": 1773 }, { "epoch": 0.19, "learning_rate": 5.9686339958072954e-05, "loss": 1.3444, "step": 1774 }, { "epoch": 0.19, "learning_rate": 5.968598485551603e-05, "loss": 0.6075, "step": 1775 }, { "epoch": 0.19, "learning_rate": 5.968562955312e-05, "loss": 0.7319, "step": 1776 }, { "epoch": 0.19, "learning_rate": 5.9685274050887244e-05, "loss": 1.1051, "step": 1777 }, { "epoch": 0.19, "learning_rate": 5.9684918348820186e-05, "loss": 0.2637, "step": 1778 }, { "epoch": 0.19, "learning_rate": 5.968456244692119e-05, "loss": 0.4795, "step": 1779 }, { "epoch": 0.19, "learning_rate": 5.9684206345192675e-05, "loss": 0.8716, "step": 1780 }, { "epoch": 0.19, "learning_rate": 5.968385004363702e-05, "loss": 0.4017, "step": 1781 }, { "epoch": 0.19, "learning_rate": 5.968349354225664e-05, "loss": 0.7364, "step": 1782 }, { "epoch": 0.19, "learning_rate": 5.968313684105392e-05, "loss": 1.306, "step": 1783 }, { "epoch": 0.19, "learning_rate": 5.9682779940031276e-05, "loss": 0.988, "step": 1784 }, { "epoch": 0.19, "learning_rate": 5.96824228391911e-05, "loss": 0.9943, "step": 1785 }, { "epoch": 0.19, "learning_rate": 5.9682065538535796e-05, "loss": 0.2861, "step": 1786 }, { "epoch": 0.19, "learning_rate": 5.9681708038067765e-05, "loss": 0.6755, "step": 1787 }, { "epoch": 0.19, "learning_rate": 5.968135033778944e-05, "loss": 1.195, "step": 1788 }, { "epoch": 0.19, "learning_rate": 5.968099243770319e-05, "loss": 0.4955, "step": 1789 }, { "epoch": 0.19, "learning_rate": 5.968063433781145e-05, "loss": 0.8253, "step": 1790 }, { "epoch": 0.19, "learning_rate": 5.968027603811663e-05, "loss": 0.4074, "step": 1791 }, { "epoch": 0.19, "learning_rate": 5.967991753862113e-05, "loss": 1.4068, "step": 1792 }, { "epoch": 0.19, "learning_rate": 5.967955883932737e-05, "loss": 0.6754, "step": 1793 }, { "epoch": 0.19, "learning_rate": 5.967919994023777e-05, "loss": 0.6863, "step": 1794 }, { "epoch": 0.19, "learning_rate": 5.967884084135473e-05, "loss": 0.7053, "step": 1795 }, { "epoch": 0.19, "learning_rate": 5.967848154268069e-05, "loss": 0.3614, "step": 1796 }, { "epoch": 0.19, "learning_rate": 5.967812204421805e-05, "loss": 0.3557, "step": 1797 }, { "epoch": 0.19, "learning_rate": 5.9677762345969234e-05, "loss": 1.0678, "step": 1798 }, { "epoch": 0.19, "learning_rate": 5.967740244793667e-05, "loss": 1.0938, "step": 1799 }, { "epoch": 0.19, "learning_rate": 5.967704235012277e-05, "loss": 0.7963, "step": 1800 }, { "epoch": 0.19, "learning_rate": 5.967668205252997e-05, "loss": 0.3787, "step": 1801 }, { "epoch": 0.19, "learning_rate": 5.9676321555160695e-05, "loss": 0.3361, "step": 1802 }, { "epoch": 0.19, "learning_rate": 5.9675960858017365e-05, "loss": 0.9153, "step": 1803 }, { "epoch": 0.19, "learning_rate": 5.9675599961102406e-05, "loss": 1.0598, "step": 1804 }, { "epoch": 0.19, "learning_rate": 5.967523886441825e-05, "loss": 1.0801, "step": 1805 }, { "epoch": 0.19, "learning_rate": 5.9674877567967335e-05, "loss": 0.7907, "step": 1806 }, { "epoch": 0.19, "learning_rate": 5.967451607175208e-05, "loss": 0.8268, "step": 1807 }, { "epoch": 0.19, "learning_rate": 5.967415437577493e-05, "loss": 0.1762, "step": 1808 }, { "epoch": 0.19, "learning_rate": 5.967379248003832e-05, "loss": 0.3241, "step": 1809 }, { "epoch": 0.19, "learning_rate": 5.967343038454468e-05, "loss": 0.3307, "step": 1810 }, { "epoch": 0.19, "learning_rate": 5.967306808929645e-05, "loss": 1.2425, "step": 1811 }, { "epoch": 0.19, "learning_rate": 5.9672705594296065e-05, "loss": 0.5995, "step": 1812 }, { "epoch": 0.19, "learning_rate": 5.967234289954598e-05, "loss": 0.8661, "step": 1813 }, { "epoch": 0.19, "learning_rate": 5.967198000504861e-05, "loss": 2.2168, "step": 1814 }, { "epoch": 0.19, "learning_rate": 5.9671616910806414e-05, "loss": 1.0555, "step": 1815 }, { "epoch": 0.19, "learning_rate": 5.967125361682185e-05, "loss": 1.0532, "step": 1816 }, { "epoch": 0.19, "learning_rate": 5.9670890123097344e-05, "loss": 0.324, "step": 1817 }, { "epoch": 0.19, "learning_rate": 5.967052642963534e-05, "loss": 0.5514, "step": 1818 }, { "epoch": 0.19, "learning_rate": 5.96701625364383e-05, "loss": 1.5591, "step": 1819 }, { "epoch": 0.19, "learning_rate": 5.966979844350867e-05, "loss": 1.3049, "step": 1820 }, { "epoch": 0.19, "learning_rate": 5.96694341508489e-05, "loss": 0.8917, "step": 1821 }, { "epoch": 0.19, "learning_rate": 5.966906965846143e-05, "loss": 0.6622, "step": 1822 }, { "epoch": 0.19, "learning_rate": 5.966870496634874e-05, "loss": 1.0785, "step": 1823 }, { "epoch": 0.19, "learning_rate": 5.966834007451325e-05, "loss": 0.8193, "step": 1824 }, { "epoch": 0.19, "learning_rate": 5.966797498295746e-05, "loss": 0.4305, "step": 1825 }, { "epoch": 0.19, "learning_rate": 5.9667609691683794e-05, "loss": 1.0939, "step": 1826 }, { "epoch": 0.19, "learning_rate": 5.9667244200694717e-05, "loss": 1.0219, "step": 1827 }, { "epoch": 0.19, "learning_rate": 5.96668785099927e-05, "loss": 0.6468, "step": 1828 }, { "epoch": 0.19, "learning_rate": 5.9666512619580194e-05, "loss": 0.5222, "step": 1829 }, { "epoch": 0.19, "learning_rate": 5.966614652945967e-05, "loss": 0.3804, "step": 1830 }, { "epoch": 0.19, "learning_rate": 5.966578023963359e-05, "loss": 0.5705, "step": 1831 }, { "epoch": 0.19, "learning_rate": 5.966541375010441e-05, "loss": 0.3191, "step": 1832 }, { "epoch": 0.19, "learning_rate": 5.966504706087462e-05, "loss": 0.7897, "step": 1833 }, { "epoch": 0.19, "learning_rate": 5.9664680171946655e-05, "loss": 0.907, "step": 1834 }, { "epoch": 0.19, "learning_rate": 5.966431308332302e-05, "loss": 1.1377, "step": 1835 }, { "epoch": 0.19, "learning_rate": 5.9663945795006164e-05, "loss": 0.7161, "step": 1836 }, { "epoch": 0.19, "learning_rate": 5.966357830699857e-05, "loss": 0.4177, "step": 1837 }, { "epoch": 0.19, "learning_rate": 5.966321061930271e-05, "loss": 0.9742, "step": 1838 }, { "epoch": 0.19, "learning_rate": 5.966284273192104e-05, "loss": 0.4797, "step": 1839 }, { "epoch": 0.19, "learning_rate": 5.966247464485607e-05, "loss": 1.0686, "step": 1840 }, { "epoch": 0.19, "learning_rate": 5.966210635811025e-05, "loss": 1.2232, "step": 1841 }, { "epoch": 0.19, "learning_rate": 5.966173787168608e-05, "loss": 0.5103, "step": 1842 }, { "epoch": 0.19, "learning_rate": 5.966136918558603e-05, "loss": 1.2935, "step": 1843 }, { "epoch": 0.19, "learning_rate": 5.966100029981258e-05, "loss": 1.2566, "step": 1844 }, { "epoch": 0.19, "learning_rate": 5.966063121436821e-05, "loss": 0.715, "step": 1845 }, { "epoch": 0.19, "learning_rate": 5.966026192925542e-05, "loss": 0.6583, "step": 1846 }, { "epoch": 0.19, "learning_rate": 5.965989244447668e-05, "loss": 0.5555, "step": 1847 }, { "epoch": 0.19, "learning_rate": 5.965952276003449e-05, "loss": 0.8796, "step": 1848 }, { "epoch": 0.19, "learning_rate": 5.9659152875931335e-05, "loss": 0.8468, "step": 1849 }, { "epoch": 0.19, "learning_rate": 5.965878279216969e-05, "loss": 0.4059, "step": 1850 }, { "epoch": 0.19, "learning_rate": 5.965841250875207e-05, "loss": 0.8333, "step": 1851 }, { "epoch": 0.19, "learning_rate": 5.9658042025680947e-05, "loss": 0.2274, "step": 1852 }, { "epoch": 0.19, "learning_rate": 5.965767134295883e-05, "loss": 0.5463, "step": 1853 }, { "epoch": 0.19, "learning_rate": 5.9657300460588215e-05, "loss": 1.4122, "step": 1854 }, { "epoch": 0.19, "learning_rate": 5.965692937857158e-05, "loss": 0.8301, "step": 1855 }, { "epoch": 0.19, "learning_rate": 5.9656558096911444e-05, "loss": 0.5918, "step": 1856 }, { "epoch": 0.19, "learning_rate": 5.96561866156103e-05, "loss": 0.8535, "step": 1857 }, { "epoch": 0.19, "learning_rate": 5.9655814934670646e-05, "loss": 1.0543, "step": 1858 }, { "epoch": 0.19, "learning_rate": 5.9655443054094976e-05, "loss": 0.2542, "step": 1859 }, { "epoch": 0.19, "learning_rate": 5.965507097388581e-05, "loss": 0.9193, "step": 1860 }, { "epoch": 0.19, "learning_rate": 5.965469869404565e-05, "loss": 0.4679, "step": 1861 }, { "epoch": 0.19, "learning_rate": 5.965432621457698e-05, "loss": 0.7942, "step": 1862 }, { "epoch": 0.19, "learning_rate": 5.965395353548234e-05, "loss": 0.7661, "step": 1863 }, { "epoch": 0.19, "learning_rate": 5.9653580656764214e-05, "loss": 0.4488, "step": 1864 }, { "epoch": 0.19, "learning_rate": 5.9653207578425124e-05, "loss": 0.446, "step": 1865 }, { "epoch": 0.19, "learning_rate": 5.9652834300467576e-05, "loss": 0.2976, "step": 1866 }, { "epoch": 0.19, "learning_rate": 5.9652460822894096e-05, "loss": 0.2865, "step": 1867 }, { "epoch": 0.2, "learning_rate": 5.965208714570718e-05, "loss": 1.299, "step": 1868 }, { "epoch": 0.2, "learning_rate": 5.965171326890936e-05, "loss": 0.0886, "step": 1869 }, { "epoch": 0.2, "learning_rate": 5.9651339192503134e-05, "loss": 0.4464, "step": 1870 }, { "epoch": 0.2, "learning_rate": 5.965096491649103e-05, "loss": 0.6519, "step": 1871 }, { "epoch": 0.2, "learning_rate": 5.965059044087557e-05, "loss": 1.2123, "step": 1872 }, { "epoch": 0.2, "learning_rate": 5.965021576565927e-05, "loss": 1.2297, "step": 1873 }, { "epoch": 0.2, "learning_rate": 5.9649840890844665e-05, "loss": 1.0086, "step": 1874 }, { "epoch": 0.2, "learning_rate": 5.964946581643426e-05, "loss": 0.795, "step": 1875 }, { "epoch": 0.2, "learning_rate": 5.9649090542430586e-05, "loss": 0.9411, "step": 1876 }, { "epoch": 0.2, "learning_rate": 5.9648715068836184e-05, "loss": 0.9203, "step": 1877 }, { "epoch": 0.2, "learning_rate": 5.9648339395653565e-05, "loss": 0.5508, "step": 1878 }, { "epoch": 0.2, "learning_rate": 5.964796352288526e-05, "loss": 1.3606, "step": 1879 }, { "epoch": 0.2, "learning_rate": 5.96475874505338e-05, "loss": 0.7264, "step": 1880 }, { "epoch": 0.2, "learning_rate": 5.964721117860172e-05, "loss": 0.9951, "step": 1881 }, { "epoch": 0.2, "learning_rate": 5.964683470709154e-05, "loss": 1.3432, "step": 1882 }, { "epoch": 0.2, "learning_rate": 5.964645803600582e-05, "loss": 0.9144, "step": 1883 }, { "epoch": 0.2, "learning_rate": 5.9646081165347085e-05, "loss": 0.3574, "step": 1884 }, { "epoch": 0.2, "learning_rate": 5.964570409511786e-05, "loss": 0.7935, "step": 1885 }, { "epoch": 0.2, "learning_rate": 5.9645326825320695e-05, "loss": 0.3645, "step": 1886 }, { "epoch": 0.2, "learning_rate": 5.964494935595812e-05, "loss": 1.1009, "step": 1887 }, { "epoch": 0.2, "learning_rate": 5.964457168703269e-05, "loss": 0.7529, "step": 1888 }, { "epoch": 0.2, "learning_rate": 5.964419381854693e-05, "loss": 0.9695, "step": 1889 }, { "epoch": 0.2, "learning_rate": 5.96438157505034e-05, "loss": 1.0352, "step": 1890 }, { "epoch": 0.2, "learning_rate": 5.9643437482904644e-05, "loss": 1.1654, "step": 1891 }, { "epoch": 0.2, "learning_rate": 5.964305901575319e-05, "loss": 0.6177, "step": 1892 }, { "epoch": 0.2, "learning_rate": 5.96426803490516e-05, "loss": 0.4051, "step": 1893 }, { "epoch": 0.2, "learning_rate": 5.964230148280243e-05, "loss": 0.288, "step": 1894 }, { "epoch": 0.2, "learning_rate": 5.964192241700822e-05, "loss": 1.1474, "step": 1895 }, { "epoch": 0.2, "learning_rate": 5.964154315167153e-05, "loss": 1.0338, "step": 1896 }, { "epoch": 0.2, "learning_rate": 5.96411636867949e-05, "loss": 0.5506, "step": 1897 }, { "epoch": 0.2, "learning_rate": 5.9640784022380883e-05, "loss": 0.8848, "step": 1898 }, { "epoch": 0.2, "learning_rate": 5.9640404158432055e-05, "loss": 0.5308, "step": 1899 }, { "epoch": 0.2, "learning_rate": 5.9640024094950954e-05, "loss": 0.6386, "step": 1900 }, { "epoch": 0.2, "learning_rate": 5.963964383194016e-05, "loss": 0.6058, "step": 1901 }, { "epoch": 0.2, "learning_rate": 5.9639263369402204e-05, "loss": 0.4288, "step": 1902 }, { "epoch": 0.2, "learning_rate": 5.963888270733967e-05, "loss": 0.1781, "step": 1903 }, { "epoch": 0.2, "learning_rate": 5.963850184575511e-05, "loss": 0.2039, "step": 1904 }, { "epoch": 0.2, "learning_rate": 5.963812078465109e-05, "loss": 1.3811, "step": 1905 }, { "epoch": 0.2, "learning_rate": 5.9637739524030174e-05, "loss": 0.9333, "step": 1906 }, { "epoch": 0.2, "learning_rate": 5.963735806389494e-05, "loss": 0.4603, "step": 1907 }, { "epoch": 0.2, "learning_rate": 5.963697640424794e-05, "loss": 0.8483, "step": 1908 }, { "epoch": 0.2, "learning_rate": 5.963659454509174e-05, "loss": 0.4911, "step": 1909 }, { "epoch": 0.2, "learning_rate": 5.963621248642893e-05, "loss": 0.8225, "step": 1910 }, { "epoch": 0.2, "learning_rate": 5.963583022826207e-05, "loss": 1.5098, "step": 1911 }, { "epoch": 0.2, "learning_rate": 5.963544777059373e-05, "loss": 0.7453, "step": 1912 }, { "epoch": 0.2, "learning_rate": 5.96350651134265e-05, "loss": 0.5219, "step": 1913 }, { "epoch": 0.2, "learning_rate": 5.963468225676294e-05, "loss": 1.3718, "step": 1914 }, { "epoch": 0.2, "learning_rate": 5.963429920060563e-05, "loss": 0.4002, "step": 1915 }, { "epoch": 0.2, "learning_rate": 5.963391594495716e-05, "loss": 1.433, "step": 1916 }, { "epoch": 0.2, "learning_rate": 5.963353248982008e-05, "loss": 0.5901, "step": 1917 }, { "epoch": 0.2, "learning_rate": 5.963314883519702e-05, "loss": 1.216, "step": 1918 }, { "epoch": 0.2, "learning_rate": 5.9632764981090525e-05, "loss": 1.2247, "step": 1919 }, { "epoch": 0.2, "learning_rate": 5.963238092750319e-05, "loss": 1.2762, "step": 1920 }, { "epoch": 0.2, "learning_rate": 5.96319966744376e-05, "loss": 1.1015, "step": 1921 }, { "epoch": 0.2, "learning_rate": 5.963161222189634e-05, "loss": 0.5929, "step": 1922 }, { "epoch": 0.2, "learning_rate": 5.9631227569882e-05, "loss": 1.0726, "step": 1923 }, { "epoch": 0.2, "learning_rate": 5.963084271839717e-05, "loss": 0.5378, "step": 1924 }, { "epoch": 0.2, "learning_rate": 5.963045766744444e-05, "loss": 1.2711, "step": 1925 }, { "epoch": 0.2, "learning_rate": 5.9630072417026406e-05, "loss": 0.3561, "step": 1926 }, { "epoch": 0.2, "learning_rate": 5.962968696714565e-05, "loss": 1.5209, "step": 1927 }, { "epoch": 0.2, "learning_rate": 5.962930131780478e-05, "loss": 0.1841, "step": 1928 }, { "epoch": 0.2, "learning_rate": 5.962891546900639e-05, "loss": 0.2783, "step": 1929 }, { "epoch": 0.2, "learning_rate": 5.9628529420753065e-05, "loss": 0.1413, "step": 1930 }, { "epoch": 0.2, "learning_rate": 5.9628143173047405e-05, "loss": 1.1861, "step": 1931 }, { "epoch": 0.2, "learning_rate": 5.9627756725892035e-05, "loss": 1.0775, "step": 1932 }, { "epoch": 0.2, "learning_rate": 5.962737007928953e-05, "loss": 0.9686, "step": 1933 }, { "epoch": 0.2, "learning_rate": 5.9626983233242506e-05, "loss": 0.8012, "step": 1934 }, { "epoch": 0.2, "learning_rate": 5.962659618775355e-05, "loss": 0.8776, "step": 1935 }, { "epoch": 0.2, "learning_rate": 5.9626208942825296e-05, "loss": 1.0231, "step": 1936 }, { "epoch": 0.2, "learning_rate": 5.9625821498460334e-05, "loss": 0.245, "step": 1937 }, { "epoch": 0.2, "learning_rate": 5.9625433854661275e-05, "loss": 1.2665, "step": 1938 }, { "epoch": 0.2, "learning_rate": 5.9625046011430715e-05, "loss": 0.5494, "step": 1939 }, { "epoch": 0.2, "learning_rate": 5.962465796877128e-05, "loss": 1.0269, "step": 1940 }, { "epoch": 0.2, "learning_rate": 5.962426972668559e-05, "loss": 1.7202, "step": 1941 }, { "epoch": 0.2, "learning_rate": 5.9623881285176245e-05, "loss": 0.8396, "step": 1942 }, { "epoch": 0.2, "learning_rate": 5.9623492644245854e-05, "loss": 0.6719, "step": 1943 }, { "epoch": 0.2, "learning_rate": 5.9623103803897046e-05, "loss": 0.5954, "step": 1944 }, { "epoch": 0.2, "learning_rate": 5.9622714764132436e-05, "loss": 0.3562, "step": 1945 }, { "epoch": 0.2, "learning_rate": 5.962232552495464e-05, "loss": 1.9204, "step": 1946 }, { "epoch": 0.2, "learning_rate": 5.962193608636628e-05, "loss": 0.3621, "step": 1947 }, { "epoch": 0.2, "learning_rate": 5.9621546448369976e-05, "loss": 1.4083, "step": 1948 }, { "epoch": 0.2, "learning_rate": 5.962115661096836e-05, "loss": 0.5891, "step": 1949 }, { "epoch": 0.2, "learning_rate": 5.962076657416404e-05, "loss": 1.0841, "step": 1950 }, { "epoch": 0.2, "learning_rate": 5.962037633795965e-05, "loss": 0.5047, "step": 1951 }, { "epoch": 0.2, "learning_rate": 5.961998590235782e-05, "loss": 0.6865, "step": 1952 }, { "epoch": 0.2, "learning_rate": 5.961959526736117e-05, "loss": 1.1682, "step": 1953 }, { "epoch": 0.2, "learning_rate": 5.961920443297235e-05, "loss": 0.5984, "step": 1954 }, { "epoch": 0.2, "learning_rate": 5.9618813399193965e-05, "loss": 0.4487, "step": 1955 }, { "epoch": 0.2, "learning_rate": 5.9618422166028654e-05, "loss": 0.4757, "step": 1956 }, { "epoch": 0.2, "learning_rate": 5.961803073347906e-05, "loss": 0.484, "step": 1957 }, { "epoch": 0.2, "learning_rate": 5.9617639101547815e-05, "loss": 0.9085, "step": 1958 }, { "epoch": 0.2, "learning_rate": 5.961724727023755e-05, "loss": 1.1666, "step": 1959 }, { "epoch": 0.2, "learning_rate": 5.961685523955091e-05, "loss": 1.2522, "step": 1960 }, { "epoch": 0.2, "learning_rate": 5.961646300949052e-05, "loss": 0.4727, "step": 1961 }, { "epoch": 0.2, "learning_rate": 5.961607058005904e-05, "loss": 1.2295, "step": 1962 }, { "epoch": 0.21, "learning_rate": 5.961567795125911e-05, "loss": 0.2913, "step": 1963 }, { "epoch": 0.21, "learning_rate": 5.961528512309336e-05, "loss": 0.7912, "step": 1964 }, { "epoch": 0.21, "learning_rate": 5.9614892095564435e-05, "loss": 0.964, "step": 1965 }, { "epoch": 0.21, "learning_rate": 5.961449886867499e-05, "loss": 0.3344, "step": 1966 }, { "epoch": 0.21, "learning_rate": 5.961410544242766e-05, "loss": 0.7124, "step": 1967 }, { "epoch": 0.21, "learning_rate": 5.961371181682511e-05, "loss": 0.1719, "step": 1968 }, { "epoch": 0.21, "learning_rate": 5.961331799186998e-05, "loss": 0.2466, "step": 1969 }, { "epoch": 0.21, "learning_rate": 5.961292396756492e-05, "loss": 0.9262, "step": 1970 }, { "epoch": 0.21, "learning_rate": 5.961252974391258e-05, "loss": 0.8153, "step": 1971 }, { "epoch": 0.21, "learning_rate": 5.9612135320915624e-05, "loss": 1.7372, "step": 1972 }, { "epoch": 0.21, "learning_rate": 5.96117406985767e-05, "loss": 0.9255, "step": 1973 }, { "epoch": 0.21, "learning_rate": 5.9611345876898475e-05, "loss": 0.2425, "step": 1974 }, { "epoch": 0.21, "learning_rate": 5.9610950855883587e-05, "loss": 0.3274, "step": 1975 }, { "epoch": 0.21, "learning_rate": 5.961055563553471e-05, "loss": 1.1695, "step": 1976 }, { "epoch": 0.21, "learning_rate": 5.96101602158545e-05, "loss": 0.7552, "step": 1977 }, { "epoch": 0.21, "learning_rate": 5.960976459684562e-05, "loss": 1.196, "step": 1978 }, { "epoch": 0.21, "learning_rate": 5.960936877851074e-05, "loss": 1.1628, "step": 1979 }, { "epoch": 0.21, "learning_rate": 5.960897276085251e-05, "loss": 1.3275, "step": 1980 }, { "epoch": 0.21, "learning_rate": 5.96085765438736e-05, "loss": 0.5518, "step": 1981 }, { "epoch": 0.21, "learning_rate": 5.960818012757669e-05, "loss": 0.3895, "step": 1982 }, { "epoch": 0.21, "learning_rate": 5.960778351196443e-05, "loss": 0.9717, "step": 1983 }, { "epoch": 0.21, "learning_rate": 5.96073866970395e-05, "loss": 1.1576, "step": 1984 }, { "epoch": 0.21, "learning_rate": 5.960698968280457e-05, "loss": 0.54, "step": 1985 }, { "epoch": 0.21, "learning_rate": 5.960659246926232e-05, "loss": 1.0495, "step": 1986 }, { "epoch": 0.21, "learning_rate": 5.960619505641541e-05, "loss": 0.416, "step": 1987 }, { "epoch": 0.21, "learning_rate": 5.960579744426653e-05, "loss": 1.2292, "step": 1988 }, { "epoch": 0.21, "learning_rate": 5.960539963281834e-05, "loss": 1.0268, "step": 1989 }, { "epoch": 0.21, "learning_rate": 5.960500162207353e-05, "loss": 0.7757, "step": 1990 }, { "epoch": 0.21, "learning_rate": 5.960460341203478e-05, "loss": 0.2704, "step": 1991 }, { "epoch": 0.21, "learning_rate": 5.960420500270476e-05, "loss": 1.0456, "step": 1992 }, { "epoch": 0.21, "learning_rate": 5.960380639408617e-05, "loss": 0.7641, "step": 1993 }, { "epoch": 0.21, "learning_rate": 5.960340758618167e-05, "loss": 0.9666, "step": 1994 }, { "epoch": 0.21, "learning_rate": 5.9603008578993966e-05, "loss": 1.1482, "step": 1995 }, { "epoch": 0.21, "learning_rate": 5.960260937252572e-05, "loss": 1.1578, "step": 1996 }, { "epoch": 0.21, "learning_rate": 5.960220996677965e-05, "loss": 1.075, "step": 1997 }, { "epoch": 0.21, "learning_rate": 5.960181036175843e-05, "loss": 0.4114, "step": 1998 }, { "epoch": 0.21, "learning_rate": 5.9601410557464735e-05, "loss": 0.4596, "step": 1999 }, { "epoch": 0.21, "learning_rate": 5.960101055390128e-05, "loss": 1.3615, "step": 2000 }, { "epoch": 0.21, "learning_rate": 5.960061035107074e-05, "loss": 0.9846, "step": 2001 }, { "epoch": 0.21, "learning_rate": 5.9600209948975826e-05, "loss": 0.3313, "step": 2002 }, { "epoch": 0.21, "learning_rate": 5.9599809347619214e-05, "loss": 0.6236, "step": 2003 }, { "epoch": 0.21, "learning_rate": 5.9599408547003624e-05, "loss": 0.7706, "step": 2004 }, { "epoch": 0.21, "learning_rate": 5.959900754713173e-05, "loss": 0.4009, "step": 2005 }, { "epoch": 0.21, "learning_rate": 5.959860634800625e-05, "loss": 1.0477, "step": 2006 }, { "epoch": 0.21, "learning_rate": 5.959820494962988e-05, "loss": 0.5823, "step": 2007 }, { "epoch": 0.21, "learning_rate": 5.959780335200532e-05, "loss": 1.0284, "step": 2008 }, { "epoch": 0.21, "learning_rate": 5.9597401555135266e-05, "loss": 0.37, "step": 2009 }, { "epoch": 0.21, "learning_rate": 5.959699955902243e-05, "loss": 1.2148, "step": 2010 }, { "epoch": 0.21, "learning_rate": 5.959659736366952e-05, "loss": 0.4441, "step": 2011 }, { "epoch": 0.21, "learning_rate": 5.959619496907925e-05, "loss": 0.582, "step": 2012 }, { "epoch": 0.21, "learning_rate": 5.959579237525431e-05, "loss": 1.1648, "step": 2013 }, { "epoch": 0.21, "learning_rate": 5.959538958219742e-05, "loss": 1.1637, "step": 2014 }, { "epoch": 0.21, "learning_rate": 5.95949865899113e-05, "loss": 0.3725, "step": 2015 }, { "epoch": 0.21, "learning_rate": 5.959458339839865e-05, "loss": 1.3701, "step": 2016 }, { "epoch": 0.21, "learning_rate": 5.959418000766219e-05, "loss": 1.5303, "step": 2017 }, { "epoch": 0.21, "learning_rate": 5.9593776417704635e-05, "loss": 1.2735, "step": 2018 }, { "epoch": 0.21, "learning_rate": 5.95933726285287e-05, "loss": 0.5132, "step": 2019 }, { "epoch": 0.21, "learning_rate": 5.959296864013711e-05, "loss": 0.7168, "step": 2020 }, { "epoch": 0.21, "learning_rate": 5.9592564452532574e-05, "loss": 0.3681, "step": 2021 }, { "epoch": 0.21, "learning_rate": 5.959216006571782e-05, "loss": 0.9829, "step": 2022 }, { "epoch": 0.21, "learning_rate": 5.959175547969556e-05, "loss": 0.7913, "step": 2023 }, { "epoch": 0.21, "learning_rate": 5.959135069446854e-05, "loss": 0.9391, "step": 2024 }, { "epoch": 0.21, "learning_rate": 5.959094571003946e-05, "loss": 0.6182, "step": 2025 }, { "epoch": 0.21, "learning_rate": 5.959054052641106e-05, "loss": 0.6741, "step": 2026 }, { "epoch": 0.21, "learning_rate": 5.959013514358607e-05, "loss": 0.388, "step": 2027 }, { "epoch": 0.21, "learning_rate": 5.9589729561567214e-05, "loss": 0.641, "step": 2028 }, { "epoch": 0.21, "learning_rate": 5.9589323780357216e-05, "loss": 0.8891, "step": 2029 }, { "epoch": 0.21, "learning_rate": 5.9588917799958815e-05, "loss": 0.9495, "step": 2030 }, { "epoch": 0.21, "learning_rate": 5.958851162037474e-05, "loss": 0.5533, "step": 2031 }, { "epoch": 0.21, "learning_rate": 5.958810524160774e-05, "loss": 1.7037, "step": 2032 }, { "epoch": 0.21, "learning_rate": 5.9587698663660526e-05, "loss": 1.286, "step": 2033 }, { "epoch": 0.21, "learning_rate": 5.9587291886535847e-05, "loss": 1.0749, "step": 2034 }, { "epoch": 0.21, "learning_rate": 5.958688491023645e-05, "loss": 0.6481, "step": 2035 }, { "epoch": 0.21, "learning_rate": 5.958647773476506e-05, "loss": 0.4417, "step": 2036 }, { "epoch": 0.21, "learning_rate": 5.958607036012443e-05, "loss": 1.0243, "step": 2037 }, { "epoch": 0.21, "learning_rate": 5.95856627863173e-05, "loss": 0.669, "step": 2038 }, { "epoch": 0.21, "learning_rate": 5.95852550133464e-05, "loss": 1.3056, "step": 2039 }, { "epoch": 0.21, "learning_rate": 5.9584847041214497e-05, "loss": 0.1797, "step": 2040 }, { "epoch": 0.21, "learning_rate": 5.958443886992431e-05, "loss": 0.4047, "step": 2041 }, { "epoch": 0.21, "learning_rate": 5.958403049947862e-05, "loss": 0.8831, "step": 2042 }, { "epoch": 0.21, "learning_rate": 5.9583621929880156e-05, "loss": 0.8707, "step": 2043 }, { "epoch": 0.21, "learning_rate": 5.9583213161131665e-05, "loss": 1.5142, "step": 2044 }, { "epoch": 0.21, "learning_rate": 5.95828041932359e-05, "loss": 0.5252, "step": 2045 }, { "epoch": 0.21, "learning_rate": 5.9582395026195626e-05, "loss": 0.8346, "step": 2046 }, { "epoch": 0.21, "learning_rate": 5.9581985660013594e-05, "loss": 0.476, "step": 2047 }, { "epoch": 0.21, "learning_rate": 5.958157609469255e-05, "loss": 0.3577, "step": 2048 }, { "epoch": 0.21, "learning_rate": 5.958116633023527e-05, "loss": 0.4357, "step": 2049 }, { "epoch": 0.21, "learning_rate": 5.958075636664448e-05, "loss": 0.5073, "step": 2050 }, { "epoch": 0.21, "learning_rate": 5.9580346203922974e-05, "loss": 0.2529, "step": 2051 }, { "epoch": 0.21, "learning_rate": 5.957993584207349e-05, "loss": 0.8726, "step": 2052 }, { "epoch": 0.21, "learning_rate": 5.95795252810988e-05, "loss": 1.6054, "step": 2053 }, { "epoch": 0.21, "learning_rate": 5.957911452100167e-05, "loss": 1.2258, "step": 2054 }, { "epoch": 0.21, "learning_rate": 5.9578703561784866e-05, "loss": 0.4865, "step": 2055 }, { "epoch": 0.21, "learning_rate": 5.957829240345114e-05, "loss": 0.9554, "step": 2056 }, { "epoch": 0.21, "learning_rate": 5.957788104600327e-05, "loss": 0.3803, "step": 2057 }, { "epoch": 0.21, "learning_rate": 5.957746948944404e-05, "loss": 1.2708, "step": 2058 }, { "epoch": 0.22, "learning_rate": 5.9577057733776204e-05, "loss": 1.0791, "step": 2059 }, { "epoch": 0.22, "learning_rate": 5.957664577900252e-05, "loss": 1.4993, "step": 2060 }, { "epoch": 0.22, "learning_rate": 5.957623362512579e-05, "loss": 1.3745, "step": 2061 }, { "epoch": 0.22, "learning_rate": 5.957582127214877e-05, "loss": 1.1055, "step": 2062 }, { "epoch": 0.22, "learning_rate": 5.957540872007424e-05, "loss": 1.5296, "step": 2063 }, { "epoch": 0.22, "learning_rate": 5.957499596890499e-05, "loss": 0.7764, "step": 2064 }, { "epoch": 0.22, "learning_rate": 5.957458301864377e-05, "loss": 1.0442, "step": 2065 }, { "epoch": 0.22, "learning_rate": 5.9574169869293385e-05, "loss": 0.7008, "step": 2066 }, { "epoch": 0.22, "learning_rate": 5.9573756520856614e-05, "loss": 1.1512, "step": 2067 }, { "epoch": 0.22, "learning_rate": 5.957334297333623e-05, "loss": 0.6655, "step": 2068 }, { "epoch": 0.22, "learning_rate": 5.957292922673502e-05, "loss": 0.8326, "step": 2069 }, { "epoch": 0.22, "learning_rate": 5.9572515281055764e-05, "loss": 0.5755, "step": 2070 }, { "epoch": 0.22, "learning_rate": 5.957210113630126e-05, "loss": 0.3314, "step": 2071 }, { "epoch": 0.22, "learning_rate": 5.9571686792474294e-05, "loss": 0.3173, "step": 2072 }, { "epoch": 0.22, "learning_rate": 5.957127224957765e-05, "loss": 0.7281, "step": 2073 }, { "epoch": 0.22, "learning_rate": 5.9570857507614124e-05, "loss": 0.2666, "step": 2074 }, { "epoch": 0.22, "learning_rate": 5.9570442566586494e-05, "loss": 1.4667, "step": 2075 }, { "epoch": 0.22, "learning_rate": 5.957002742649757e-05, "loss": 0.4396, "step": 2076 }, { "epoch": 0.22, "learning_rate": 5.956961208735014e-05, "loss": 1.3212, "step": 2077 }, { "epoch": 0.22, "learning_rate": 5.9569196549147e-05, "loss": 0.9646, "step": 2078 }, { "epoch": 0.22, "learning_rate": 5.956878081189094e-05, "loss": 0.6723, "step": 2079 }, { "epoch": 0.22, "learning_rate": 5.956836487558477e-05, "loss": 0.4778, "step": 2080 }, { "epoch": 0.22, "learning_rate": 5.95679487402313e-05, "loss": 0.6112, "step": 2081 }, { "epoch": 0.22, "learning_rate": 5.95675324058333e-05, "loss": 1.009, "step": 2082 }, { "epoch": 0.22, "learning_rate": 5.9567115872393604e-05, "loss": 0.7548, "step": 2083 }, { "epoch": 0.22, "learning_rate": 5.956669913991499e-05, "loss": 0.7939, "step": 2084 }, { "epoch": 0.22, "learning_rate": 5.9566282208400276e-05, "loss": 0.693, "step": 2085 }, { "epoch": 0.22, "learning_rate": 5.956586507785228e-05, "loss": 0.8715, "step": 2086 }, { "epoch": 0.22, "learning_rate": 5.956544774827378e-05, "loss": 0.5095, "step": 2087 }, { "epoch": 0.22, "learning_rate": 5.956503021966762e-05, "loss": 1.4779, "step": 2088 }, { "epoch": 0.22, "learning_rate": 5.956461249203658e-05, "loss": 0.8854, "step": 2089 }, { "epoch": 0.22, "learning_rate": 5.95641945653835e-05, "loss": 0.4817, "step": 2090 }, { "epoch": 0.22, "learning_rate": 5.956377643971117e-05, "loss": 0.6154, "step": 2091 }, { "epoch": 0.22, "learning_rate": 5.9563358115022424e-05, "loss": 1.0675, "step": 2092 }, { "epoch": 0.22, "learning_rate": 5.956293959132006e-05, "loss": 1.301, "step": 2093 }, { "epoch": 0.22, "learning_rate": 5.956252086860691e-05, "loss": 1.2832, "step": 2094 }, { "epoch": 0.22, "learning_rate": 5.9562101946885785e-05, "loss": 0.8815, "step": 2095 }, { "epoch": 0.22, "learning_rate": 5.95616828261595e-05, "loss": 1.2528, "step": 2096 }, { "epoch": 0.22, "learning_rate": 5.956126350643088e-05, "loss": 0.5294, "step": 2097 }, { "epoch": 0.22, "learning_rate": 5.956084398770276e-05, "loss": 2.3069, "step": 2098 }, { "epoch": 0.22, "learning_rate": 5.956042426997796e-05, "loss": 0.966, "step": 2099 }, { "epoch": 0.22, "learning_rate": 5.9560004353259295e-05, "loss": 0.7794, "step": 2100 }, { "epoch": 0.22, "learning_rate": 5.95595842375496e-05, "loss": 1.0558, "step": 2101 }, { "epoch": 0.22, "learning_rate": 5.955916392285169e-05, "loss": 0.5069, "step": 2102 }, { "epoch": 0.22, "learning_rate": 5.955874340916841e-05, "loss": 1.498, "step": 2103 }, { "epoch": 0.22, "learning_rate": 5.955832269650259e-05, "loss": 1.1242, "step": 2104 }, { "epoch": 0.22, "learning_rate": 5.955790178485705e-05, "loss": 0.7148, "step": 2105 }, { "epoch": 0.22, "learning_rate": 5.955748067423464e-05, "loss": 0.8965, "step": 2106 }, { "epoch": 0.22, "learning_rate": 5.9557059364638184e-05, "loss": 0.3493, "step": 2107 }, { "epoch": 0.22, "learning_rate": 5.955663785607051e-05, "loss": 3.8427, "step": 2108 }, { "epoch": 0.22, "learning_rate": 5.955621614853448e-05, "loss": 1.3349, "step": 2109 }, { "epoch": 0.22, "learning_rate": 5.955579424203291e-05, "loss": 1.1816, "step": 2110 }, { "epoch": 0.22, "learning_rate": 5.955537213656865e-05, "loss": 0.2385, "step": 2111 }, { "epoch": 0.22, "learning_rate": 5.955494983214454e-05, "loss": 0.6955, "step": 2112 }, { "epoch": 0.22, "learning_rate": 5.9554527328763425e-05, "loss": 1.3159, "step": 2113 }, { "epoch": 0.22, "learning_rate": 5.9554104626428146e-05, "loss": 1.4775, "step": 2114 }, { "epoch": 0.22, "learning_rate": 5.955368172514155e-05, "loss": 0.8633, "step": 2115 }, { "epoch": 0.22, "learning_rate": 5.955325862490648e-05, "loss": 0.4993, "step": 2116 }, { "epoch": 0.22, "learning_rate": 5.9552835325725794e-05, "loss": 1.7025, "step": 2117 }, { "epoch": 0.22, "learning_rate": 5.9552411827602336e-05, "loss": 1.132, "step": 2118 }, { "epoch": 0.22, "learning_rate": 5.9551988130538953e-05, "loss": 0.6955, "step": 2119 }, { "epoch": 0.22, "learning_rate": 5.95515642345385e-05, "loss": 0.8981, "step": 2120 }, { "epoch": 0.22, "learning_rate": 5.955114013960383e-05, "loss": 0.409, "step": 2121 }, { "epoch": 0.22, "learning_rate": 5.955071584573781e-05, "loss": 0.5566, "step": 2122 }, { "epoch": 0.22, "learning_rate": 5.955029135294326e-05, "loss": 0.955, "step": 2123 }, { "epoch": 0.22, "learning_rate": 5.954986666122309e-05, "loss": 1.0322, "step": 2124 }, { "epoch": 0.22, "learning_rate": 5.9549441770580114e-05, "loss": 0.6748, "step": 2125 }, { "epoch": 0.22, "learning_rate": 5.954901668101722e-05, "loss": 1.1897, "step": 2126 }, { "epoch": 0.22, "learning_rate": 5.954859139253725e-05, "loss": 0.4641, "step": 2127 }, { "epoch": 0.22, "learning_rate": 5.954816590514309e-05, "loss": 0.4766, "step": 2128 }, { "epoch": 0.22, "learning_rate": 5.954774021883758e-05, "loss": 0.6318, "step": 2129 }, { "epoch": 0.22, "learning_rate": 5.95473143336236e-05, "loss": 0.2728, "step": 2130 }, { "epoch": 0.22, "learning_rate": 5.9546888249504015e-05, "loss": 1.1278, "step": 2131 }, { "epoch": 0.22, "learning_rate": 5.954646196648169e-05, "loss": 1.5143, "step": 2132 }, { "epoch": 0.22, "learning_rate": 5.954603548455949e-05, "loss": 1.2229, "step": 2133 }, { "epoch": 0.22, "learning_rate": 5.9545608803740304e-05, "loss": 0.6684, "step": 2134 }, { "epoch": 0.22, "learning_rate": 5.954518192402699e-05, "loss": 0.6445, "step": 2135 }, { "epoch": 0.22, "learning_rate": 5.954475484542242e-05, "loss": 1.0448, "step": 2136 }, { "epoch": 0.22, "learning_rate": 5.954432756792947e-05, "loss": 1.8142, "step": 2137 }, { "epoch": 0.22, "learning_rate": 5.954390009155103e-05, "loss": 1.4679, "step": 2138 }, { "epoch": 0.22, "learning_rate": 5.954347241628996e-05, "loss": 1.2724, "step": 2139 }, { "epoch": 0.22, "learning_rate": 5.9543044542149146e-05, "loss": 1.2467, "step": 2140 }, { "epoch": 0.22, "learning_rate": 5.954261646913147e-05, "loss": 0.2342, "step": 2141 }, { "epoch": 0.22, "learning_rate": 5.9542188197239815e-05, "loss": 0.3442, "step": 2142 }, { "epoch": 0.22, "learning_rate": 5.9541759726477066e-05, "loss": 0.3381, "step": 2143 }, { "epoch": 0.22, "learning_rate": 5.954133105684609e-05, "loss": 0.2101, "step": 2144 }, { "epoch": 0.22, "learning_rate": 5.95409021883498e-05, "loss": 0.6154, "step": 2145 }, { "epoch": 0.22, "learning_rate": 5.954047312099106e-05, "loss": 0.7208, "step": 2146 }, { "epoch": 0.22, "learning_rate": 5.954004385477276e-05, "loss": 0.3371, "step": 2147 }, { "epoch": 0.22, "learning_rate": 5.95396143896978e-05, "loss": 0.3884, "step": 2148 }, { "epoch": 0.22, "learning_rate": 5.953918472576907e-05, "loss": 1.4382, "step": 2149 }, { "epoch": 0.22, "learning_rate": 5.953875486298947e-05, "loss": 1.2865, "step": 2150 }, { "epoch": 0.22, "learning_rate": 5.953832480136187e-05, "loss": 1.154, "step": 2151 }, { "epoch": 0.22, "learning_rate": 5.953789454088919e-05, "loss": 0.4753, "step": 2152 }, { "epoch": 0.22, "learning_rate": 5.9537464081574306e-05, "loss": 1.3756, "step": 2153 }, { "epoch": 0.22, "learning_rate": 5.953703342342013e-05, "loss": 0.5917, "step": 2154 }, { "epoch": 0.23, "learning_rate": 5.9536602566429554e-05, "loss": 0.117, "step": 2155 }, { "epoch": 0.23, "learning_rate": 5.953617151060548e-05, "loss": 1.366, "step": 2156 }, { "epoch": 0.23, "learning_rate": 5.953574025595081e-05, "loss": 0.6084, "step": 2157 }, { "epoch": 0.23, "learning_rate": 5.953530880246846e-05, "loss": 0.2377, "step": 2158 }, { "epoch": 0.23, "learning_rate": 5.9534877150161305e-05, "loss": 1.0246, "step": 2159 }, { "epoch": 0.23, "learning_rate": 5.953444529903228e-05, "loss": 0.2891, "step": 2160 }, { "epoch": 0.23, "learning_rate": 5.953401324908427e-05, "loss": 0.9977, "step": 2161 }, { "epoch": 0.23, "learning_rate": 5.9533581000320195e-05, "loss": 0.7804, "step": 2162 }, { "epoch": 0.23, "learning_rate": 5.953314855274296e-05, "loss": 0.7307, "step": 2163 }, { "epoch": 0.23, "learning_rate": 5.9532715906355484e-05, "loss": 1.0963, "step": 2164 }, { "epoch": 0.23, "learning_rate": 5.953228306116067e-05, "loss": 1.3649, "step": 2165 }, { "epoch": 0.23, "learning_rate": 5.9531850017161445e-05, "loss": 0.6751, "step": 2166 }, { "epoch": 0.23, "learning_rate": 5.9531416774360706e-05, "loss": 0.2114, "step": 2167 }, { "epoch": 0.23, "learning_rate": 5.953098333276139e-05, "loss": 1.06, "step": 2168 }, { "epoch": 0.23, "learning_rate": 5.95305496923664e-05, "loss": 0.712, "step": 2169 }, { "epoch": 0.23, "learning_rate": 5.953011585317865e-05, "loss": 0.8465, "step": 2170 }, { "epoch": 0.23, "learning_rate": 5.952968181520108e-05, "loss": 0.3946, "step": 2171 }, { "epoch": 0.23, "learning_rate": 5.95292475784366e-05, "loss": 1.4033, "step": 2172 }, { "epoch": 0.23, "learning_rate": 5.952881314288813e-05, "loss": 0.5484, "step": 2173 }, { "epoch": 0.23, "learning_rate": 5.9528378508558606e-05, "loss": 0.2928, "step": 2174 }, { "epoch": 0.23, "learning_rate": 5.9527943675450934e-05, "loss": 1.0854, "step": 2175 }, { "epoch": 0.23, "learning_rate": 5.952750864356807e-05, "loss": 1.2531, "step": 2176 }, { "epoch": 0.23, "learning_rate": 5.9527073412912916e-05, "loss": 1.4312, "step": 2177 }, { "epoch": 0.23, "learning_rate": 5.952663798348842e-05, "loss": 0.5237, "step": 2178 }, { "epoch": 0.23, "learning_rate": 5.95262023552975e-05, "loss": 1.1926, "step": 2179 }, { "epoch": 0.23, "learning_rate": 5.95257665283431e-05, "loss": 0.331, "step": 2180 }, { "epoch": 0.23, "learning_rate": 5.952533050262814e-05, "loss": 1.1466, "step": 2181 }, { "epoch": 0.23, "learning_rate": 5.952489427815557e-05, "loss": 1.2253, "step": 2182 }, { "epoch": 0.23, "learning_rate": 5.9524457854928316e-05, "loss": 0.5609, "step": 2183 }, { "epoch": 0.23, "learning_rate": 5.952402123294932e-05, "loss": 0.5352, "step": 2184 }, { "epoch": 0.23, "learning_rate": 5.952358441222153e-05, "loss": 0.4843, "step": 2185 }, { "epoch": 0.23, "learning_rate": 5.952314739274787e-05, "loss": 1.3914, "step": 2186 }, { "epoch": 0.23, "learning_rate": 5.952271017453129e-05, "loss": 0.2246, "step": 2187 }, { "epoch": 0.23, "learning_rate": 5.952227275757473e-05, "loss": 0.8889, "step": 2188 }, { "epoch": 0.23, "learning_rate": 5.9521835141881135e-05, "loss": 0.3666, "step": 2189 }, { "epoch": 0.23, "learning_rate": 5.9521397327453465e-05, "loss": 0.4368, "step": 2190 }, { "epoch": 0.23, "learning_rate": 5.952095931429465e-05, "loss": 0.6142, "step": 2191 }, { "epoch": 0.23, "learning_rate": 5.9520521102407636e-05, "loss": 1.2838, "step": 2192 }, { "epoch": 0.23, "learning_rate": 5.95200826917954e-05, "loss": 0.5639, "step": 2193 }, { "epoch": 0.23, "learning_rate": 5.951964408246086e-05, "loss": 1.1185, "step": 2194 }, { "epoch": 0.23, "learning_rate": 5.951920527440699e-05, "loss": 0.256, "step": 2195 }, { "epoch": 0.23, "learning_rate": 5.951876626763673e-05, "loss": 1.2391, "step": 2196 }, { "epoch": 0.23, "learning_rate": 5.9518327062153044e-05, "loss": 0.437, "step": 2197 }, { "epoch": 0.23, "learning_rate": 5.951788765795889e-05, "loss": 0.4153, "step": 2198 }, { "epoch": 0.23, "learning_rate": 5.951744805505722e-05, "loss": 0.3477, "step": 2199 }, { "epoch": 0.23, "learning_rate": 5.9517008253450996e-05, "loss": 0.3265, "step": 2200 }, { "epoch": 0.23, "learning_rate": 5.951656825314318e-05, "loss": 0.6041, "step": 2201 }, { "epoch": 0.23, "learning_rate": 5.9516128054136736e-05, "loss": 1.1067, "step": 2202 }, { "epoch": 0.23, "learning_rate": 5.951568765643463e-05, "loss": 1.3096, "step": 2203 }, { "epoch": 0.23, "learning_rate": 5.951524706003981e-05, "loss": 0.9146, "step": 2204 }, { "epoch": 0.23, "learning_rate": 5.951480626495526e-05, "loss": 1.1682, "step": 2205 }, { "epoch": 0.23, "learning_rate": 5.9514365271183925e-05, "loss": 0.6626, "step": 2206 }, { "epoch": 0.23, "learning_rate": 5.9513924078728804e-05, "loss": 1.0744, "step": 2207 }, { "epoch": 0.23, "learning_rate": 5.951348268759285e-05, "loss": 0.7508, "step": 2208 }, { "epoch": 0.23, "learning_rate": 5.951304109777903e-05, "loss": 0.7076, "step": 2209 }, { "epoch": 0.23, "learning_rate": 5.951259930929033e-05, "loss": 0.4733, "step": 2210 }, { "epoch": 0.23, "learning_rate": 5.951215732212971e-05, "loss": 1.171, "step": 2211 }, { "epoch": 0.23, "learning_rate": 5.951171513630016e-05, "loss": 0.4551, "step": 2212 }, { "epoch": 0.23, "learning_rate": 5.951127275180464e-05, "loss": 0.2547, "step": 2213 }, { "epoch": 0.23, "learning_rate": 5.9510830168646145e-05, "loss": 1.2617, "step": 2214 }, { "epoch": 0.23, "learning_rate": 5.951038738682764e-05, "loss": 1.4125, "step": 2215 }, { "epoch": 0.23, "learning_rate": 5.9509944406352115e-05, "loss": 0.7564, "step": 2216 }, { "epoch": 0.23, "learning_rate": 5.950950122722255e-05, "loss": 1.0767, "step": 2217 }, { "epoch": 0.23, "learning_rate": 5.950905784944193e-05, "loss": 0.3592, "step": 2218 }, { "epoch": 0.23, "learning_rate": 5.9508614273013227e-05, "loss": 0.7049, "step": 2219 }, { "epoch": 0.23, "learning_rate": 5.950817049793944e-05, "loss": 1.0159, "step": 2220 }, { "epoch": 0.23, "learning_rate": 5.950772652422356e-05, "loss": 0.9651, "step": 2221 }, { "epoch": 0.23, "learning_rate": 5.950728235186856e-05, "loss": 0.8767, "step": 2222 }, { "epoch": 0.23, "learning_rate": 5.9506837980877453e-05, "loss": 0.7385, "step": 2223 }, { "epoch": 0.23, "learning_rate": 5.95063934112532e-05, "loss": 0.4957, "step": 2224 }, { "epoch": 0.23, "learning_rate": 5.950594864299883e-05, "loss": 0.5574, "step": 2225 }, { "epoch": 0.23, "learning_rate": 5.95055036761173e-05, "loss": 0.8105, "step": 2226 }, { "epoch": 0.23, "learning_rate": 5.950505851061163e-05, "loss": 1.0311, "step": 2227 }, { "epoch": 0.23, "learning_rate": 5.9504613146484816e-05, "loss": 0.4101, "step": 2228 }, { "epoch": 0.23, "learning_rate": 5.950416758373984e-05, "loss": 0.6107, "step": 2229 }, { "epoch": 0.23, "learning_rate": 5.9503721822379716e-05, "loss": 0.6081, "step": 2230 }, { "epoch": 0.23, "learning_rate": 5.950327586240744e-05, "loss": 0.713, "step": 2231 }, { "epoch": 0.23, "learning_rate": 5.950282970382602e-05, "loss": 1.129, "step": 2232 }, { "epoch": 0.23, "learning_rate": 5.950238334663844e-05, "loss": 0.4747, "step": 2233 }, { "epoch": 0.23, "learning_rate": 5.9501936790847726e-05, "loss": 0.9176, "step": 2234 }, { "epoch": 0.23, "learning_rate": 5.950149003645687e-05, "loss": 1.1786, "step": 2235 }, { "epoch": 0.23, "learning_rate": 5.950104308346889e-05, "loss": 0.7764, "step": 2236 }, { "epoch": 0.23, "learning_rate": 5.95005959318868e-05, "loss": 1.0083, "step": 2237 }, { "epoch": 0.23, "learning_rate": 5.9500148581713586e-05, "loss": 1.6475, "step": 2238 }, { "epoch": 0.23, "learning_rate": 5.9499701032952284e-05, "loss": 0.4699, "step": 2239 }, { "epoch": 0.23, "learning_rate": 5.9499253285605896e-05, "loss": 0.879, "step": 2240 }, { "epoch": 0.23, "learning_rate": 5.949880533967744e-05, "loss": 1.1908, "step": 2241 }, { "epoch": 0.23, "learning_rate": 5.949835719516992e-05, "loss": 1.0879, "step": 2242 }, { "epoch": 0.23, "learning_rate": 5.949790885208636e-05, "loss": 0.538, "step": 2243 }, { "epoch": 0.23, "learning_rate": 5.949746031042979e-05, "loss": 0.6309, "step": 2244 }, { "epoch": 0.23, "learning_rate": 5.949701157020321e-05, "loss": 0.9691, "step": 2245 }, { "epoch": 0.23, "learning_rate": 5.949656263140965e-05, "loss": 0.5161, "step": 2246 }, { "epoch": 0.23, "learning_rate": 5.949611349405215e-05, "loss": 1.1541, "step": 2247 }, { "epoch": 0.23, "learning_rate": 5.9495664158133694e-05, "loss": 1.3233, "step": 2248 }, { "epoch": 0.23, "learning_rate": 5.949521462365734e-05, "loss": 1.4514, "step": 2249 }, { "epoch": 0.23, "learning_rate": 5.9494764890626094e-05, "loss": 1.3323, "step": 2250 }, { "epoch": 0.24, "learning_rate": 5.9494314959042994e-05, "loss": 0.9415, "step": 2251 }, { "epoch": 0.24, "learning_rate": 5.949386482891107e-05, "loss": 0.6239, "step": 2252 }, { "epoch": 0.24, "learning_rate": 5.9493414500233346e-05, "loss": 0.7657, "step": 2253 }, { "epoch": 0.24, "learning_rate": 5.9492963973012855e-05, "loss": 0.443, "step": 2254 }, { "epoch": 0.24, "learning_rate": 5.949251324725263e-05, "loss": 0.706, "step": 2255 }, { "epoch": 0.24, "learning_rate": 5.949206232295572e-05, "loss": 0.7652, "step": 2256 }, { "epoch": 0.24, "learning_rate": 5.9491611200125135e-05, "loss": 0.624, "step": 2257 }, { "epoch": 0.24, "learning_rate": 5.949115987876393e-05, "loss": 0.7601, "step": 2258 }, { "epoch": 0.24, "learning_rate": 5.949070835887513e-05, "loss": 0.544, "step": 2259 }, { "epoch": 0.24, "learning_rate": 5.949025664046179e-05, "loss": 1.0484, "step": 2260 }, { "epoch": 0.24, "learning_rate": 5.9489804723526936e-05, "loss": 0.3094, "step": 2261 }, { "epoch": 0.24, "learning_rate": 5.948935260807362e-05, "loss": 0.265, "step": 2262 }, { "epoch": 0.24, "learning_rate": 5.948890029410488e-05, "loss": 1.1151, "step": 2263 }, { "epoch": 0.24, "learning_rate": 5.948844778162377e-05, "loss": 1.4296, "step": 2264 }, { "epoch": 0.24, "learning_rate": 5.9487995070633315e-05, "loss": 0.9292, "step": 2265 }, { "epoch": 0.24, "learning_rate": 5.948754216113659e-05, "loss": 0.9861, "step": 2266 }, { "epoch": 0.24, "learning_rate": 5.9487089053136625e-05, "loss": 1.1632, "step": 2267 }, { "epoch": 0.24, "learning_rate": 5.948663574663648e-05, "loss": 0.6824, "step": 2268 }, { "epoch": 0.24, "learning_rate": 5.94861822416392e-05, "loss": 1.2681, "step": 2269 }, { "epoch": 0.24, "learning_rate": 5.9485728538147836e-05, "loss": 0.535, "step": 2270 }, { "epoch": 0.24, "learning_rate": 5.948527463616545e-05, "loss": 1.05, "step": 2271 }, { "epoch": 0.24, "learning_rate": 5.9484820535695096e-05, "loss": 0.4362, "step": 2272 }, { "epoch": 0.24, "learning_rate": 5.948436623673983e-05, "loss": 1.3459, "step": 2273 }, { "epoch": 0.24, "learning_rate": 5.948391173930271e-05, "loss": 0.7544, "step": 2274 }, { "epoch": 0.24, "learning_rate": 5.9483457043386786e-05, "loss": 1.1762, "step": 2275 }, { "epoch": 0.24, "learning_rate": 5.948300214899513e-05, "loss": 0.8736, "step": 2276 }, { "epoch": 0.24, "learning_rate": 5.9482547056130814e-05, "loss": 0.2496, "step": 2277 }, { "epoch": 0.24, "learning_rate": 5.9482091764796876e-05, "loss": 0.2819, "step": 2278 }, { "epoch": 0.24, "learning_rate": 5.94816362749964e-05, "loss": 1.2048, "step": 2279 }, { "epoch": 0.24, "learning_rate": 5.9481180586732446e-05, "loss": 1.0063, "step": 2280 }, { "epoch": 0.24, "learning_rate": 5.9480724700008086e-05, "loss": 0.9979, "step": 2281 }, { "epoch": 0.24, "learning_rate": 5.948026861482637e-05, "loss": 1.1897, "step": 2282 }, { "epoch": 0.24, "learning_rate": 5.94798123311904e-05, "loss": 1.2511, "step": 2283 }, { "epoch": 0.24, "learning_rate": 5.9479355849103227e-05, "loss": 0.6354, "step": 2284 }, { "epoch": 0.24, "learning_rate": 5.9478899168567926e-05, "loss": 0.8931, "step": 2285 }, { "epoch": 0.24, "learning_rate": 5.947844228958758e-05, "loss": 0.4732, "step": 2286 }, { "epoch": 0.24, "learning_rate": 5.947798521216525e-05, "loss": 1.1158, "step": 2287 }, { "epoch": 0.24, "learning_rate": 5.947752793630403e-05, "loss": 1.1441, "step": 2288 }, { "epoch": 0.24, "learning_rate": 5.947707046200698e-05, "loss": 0.7991, "step": 2289 }, { "epoch": 0.24, "learning_rate": 5.947661278927719e-05, "loss": 1.4804, "step": 2290 }, { "epoch": 0.24, "learning_rate": 5.947615491811774e-05, "loss": 1.4707, "step": 2291 }, { "epoch": 0.24, "learning_rate": 5.947569684853172e-05, "loss": 0.2531, "step": 2292 }, { "epoch": 0.24, "learning_rate": 5.947523858052219e-05, "loss": 1.0407, "step": 2293 }, { "epoch": 0.24, "learning_rate": 5.947478011409226e-05, "loss": 1.2114, "step": 2294 }, { "epoch": 0.24, "learning_rate": 5.947432144924501e-05, "loss": 1.2823, "step": 2295 }, { "epoch": 0.24, "learning_rate": 5.947386258598352e-05, "loss": 0.7946, "step": 2296 }, { "epoch": 0.24, "learning_rate": 5.947340352431089e-05, "loss": 0.915, "step": 2297 }, { "epoch": 0.24, "learning_rate": 5.9472944264230184e-05, "loss": 0.6693, "step": 2298 }, { "epoch": 0.24, "learning_rate": 5.947248480574453e-05, "loss": 0.7276, "step": 2299 }, { "epoch": 0.24, "learning_rate": 5.9472025148857005e-05, "loss": 0.9363, "step": 2300 }, { "epoch": 0.24, "learning_rate": 5.94715652935707e-05, "loss": 0.6293, "step": 2301 }, { "epoch": 0.24, "learning_rate": 5.947110523988871e-05, "loss": 1.4878, "step": 2302 }, { "epoch": 0.24, "learning_rate": 5.947064498781414e-05, "loss": 0.7111, "step": 2303 }, { "epoch": 0.24, "learning_rate": 5.947018453735007e-05, "loss": 1.0444, "step": 2304 }, { "epoch": 0.24, "learning_rate": 5.9469723888499635e-05, "loss": 0.2699, "step": 2305 }, { "epoch": 0.24, "learning_rate": 5.94692630412659e-05, "loss": 1.0238, "step": 2306 }, { "epoch": 0.24, "learning_rate": 5.946880199565199e-05, "loss": 0.2044, "step": 2307 }, { "epoch": 0.24, "learning_rate": 5.9468340751660985e-05, "loss": 0.6276, "step": 2308 }, { "epoch": 0.24, "learning_rate": 5.9467879309296015e-05, "loss": 1.0323, "step": 2309 }, { "epoch": 0.24, "learning_rate": 5.946741766856018e-05, "loss": 1.4021, "step": 2310 }, { "epoch": 0.24, "learning_rate": 5.9466955829456575e-05, "loss": 0.5083, "step": 2311 }, { "epoch": 0.24, "learning_rate": 5.946649379198833e-05, "loss": 1.453, "step": 2312 }, { "epoch": 0.24, "learning_rate": 5.9466031556158534e-05, "loss": 0.9906, "step": 2313 }, { "epoch": 0.24, "learning_rate": 5.946556912197031e-05, "loss": 0.5653, "step": 2314 }, { "epoch": 0.24, "learning_rate": 5.946510648942677e-05, "loss": 0.5232, "step": 2315 }, { "epoch": 0.24, "learning_rate": 5.946464365853103e-05, "loss": 1.3394, "step": 2316 }, { "epoch": 0.24, "learning_rate": 5.9464180629286205e-05, "loss": 0.8687, "step": 2317 }, { "epoch": 0.24, "learning_rate": 5.94637174016954e-05, "loss": 0.734, "step": 2318 }, { "epoch": 0.24, "learning_rate": 5.946325397576175e-05, "loss": 0.7426, "step": 2319 }, { "epoch": 0.24, "learning_rate": 5.946279035148837e-05, "loss": 1.0083, "step": 2320 }, { "epoch": 0.24, "learning_rate": 5.946232652887837e-05, "loss": 0.6107, "step": 2321 }, { "epoch": 0.24, "learning_rate": 5.946186250793489e-05, "loss": 0.4496, "step": 2322 }, { "epoch": 0.24, "learning_rate": 5.9461398288661037e-05, "loss": 1.4258, "step": 2323 }, { "epoch": 0.24, "learning_rate": 5.9460933871059954e-05, "loss": 0.5721, "step": 2324 }, { "epoch": 0.24, "learning_rate": 5.946046925513476e-05, "loss": 0.6354, "step": 2325 }, { "epoch": 0.24, "learning_rate": 5.946000444088857e-05, "loss": 1.4412, "step": 2326 }, { "epoch": 0.24, "learning_rate": 5.9459539428324526e-05, "loss": 0.524, "step": 2327 }, { "epoch": 0.24, "learning_rate": 5.945907421744575e-05, "loss": 0.7372, "step": 2328 }, { "epoch": 0.24, "learning_rate": 5.945860880825539e-05, "loss": 1.1486, "step": 2329 }, { "epoch": 0.24, "learning_rate": 5.945814320075657e-05, "loss": 1.683, "step": 2330 }, { "epoch": 0.24, "learning_rate": 5.945767739495241e-05, "loss": 1.3173, "step": 2331 }, { "epoch": 0.24, "learning_rate": 5.945721139084606e-05, "loss": 1.3063, "step": 2332 }, { "epoch": 0.24, "learning_rate": 5.945674518844066e-05, "loss": 0.3046, "step": 2333 }, { "epoch": 0.24, "learning_rate": 5.945627878773934e-05, "loss": 0.3135, "step": 2334 }, { "epoch": 0.24, "learning_rate": 5.945581218874525e-05, "loss": 1.1092, "step": 2335 }, { "epoch": 0.24, "learning_rate": 5.9455345391461524e-05, "loss": 1.2765, "step": 2336 }, { "epoch": 0.24, "learning_rate": 5.9454878395891306e-05, "loss": 0.7703, "step": 2337 }, { "epoch": 0.24, "learning_rate": 5.9454411202037735e-05, "loss": 1.396, "step": 2338 }, { "epoch": 0.24, "learning_rate": 5.945394380990395e-05, "loss": 0.1971, "step": 2339 }, { "epoch": 0.24, "learning_rate": 5.945347621949312e-05, "loss": 0.4835, "step": 2340 }, { "epoch": 0.24, "learning_rate": 5.945300843080838e-05, "loss": 0.7981, "step": 2341 }, { "epoch": 0.24, "learning_rate": 5.945254044385288e-05, "loss": 0.929, "step": 2342 }, { "epoch": 0.24, "learning_rate": 5.945207225862976e-05, "loss": 0.7683, "step": 2343 }, { "epoch": 0.24, "learning_rate": 5.945160387514219e-05, "loss": 0.2817, "step": 2344 }, { "epoch": 0.24, "learning_rate": 5.945113529339331e-05, "loss": 0.7934, "step": 2345 }, { "epoch": 0.25, "learning_rate": 5.9450666513386266e-05, "loss": 0.844, "step": 2346 }, { "epoch": 0.25, "learning_rate": 5.945019753512424e-05, "loss": 0.4864, "step": 2347 }, { "epoch": 0.25, "learning_rate": 5.9449728358610376e-05, "loss": 0.3522, "step": 2348 }, { "epoch": 0.25, "learning_rate": 5.944925898384782e-05, "loss": 0.9566, "step": 2349 }, { "epoch": 0.25, "learning_rate": 5.9448789410839754e-05, "loss": 0.6402, "step": 2350 }, { "epoch": 0.25, "learning_rate": 5.9448319639589323e-05, "loss": 0.5176, "step": 2351 }, { "epoch": 0.25, "learning_rate": 5.9447849670099695e-05, "loss": 0.584, "step": 2352 }, { "epoch": 0.25, "learning_rate": 5.9447379502374033e-05, "loss": 1.2202, "step": 2353 }, { "epoch": 0.25, "learning_rate": 5.94469091364155e-05, "loss": 0.9766, "step": 2354 }, { "epoch": 0.25, "learning_rate": 5.944643857222727e-05, "loss": 1.5476, "step": 2355 }, { "epoch": 0.25, "learning_rate": 5.9445967809812506e-05, "loss": 0.2287, "step": 2356 }, { "epoch": 0.25, "learning_rate": 5.944549684917437e-05, "loss": 0.7215, "step": 2357 }, { "epoch": 0.25, "learning_rate": 5.944502569031605e-05, "loss": 1.059, "step": 2358 }, { "epoch": 0.25, "learning_rate": 5.9444554333240696e-05, "loss": 1.0977, "step": 2359 }, { "epoch": 0.25, "learning_rate": 5.9444082777951494e-05, "loss": 0.6434, "step": 2360 }, { "epoch": 0.25, "learning_rate": 5.9443611024451616e-05, "loss": 0.27, "step": 2361 }, { "epoch": 0.25, "learning_rate": 5.944313907274425e-05, "loss": 0.1639, "step": 2362 }, { "epoch": 0.25, "learning_rate": 5.944266692283254e-05, "loss": 0.6303, "step": 2363 }, { "epoch": 0.25, "learning_rate": 5.944219457471969e-05, "loss": 0.7131, "step": 2364 }, { "epoch": 0.25, "learning_rate": 5.944172202840889e-05, "loss": 0.9789, "step": 2365 }, { "epoch": 0.25, "learning_rate": 5.9441249283903296e-05, "loss": 1.0975, "step": 2366 }, { "epoch": 0.25, "learning_rate": 5.94407763412061e-05, "loss": 0.6038, "step": 2367 }, { "epoch": 0.25, "learning_rate": 5.944030320032049e-05, "loss": 0.2874, "step": 2368 }, { "epoch": 0.25, "learning_rate": 5.943982986124964e-05, "loss": 1.4613, "step": 2369 }, { "epoch": 0.25, "learning_rate": 5.9439356323996755e-05, "loss": 0.1652, "step": 2370 }, { "epoch": 0.25, "learning_rate": 5.9438882588565e-05, "loss": 1.4449, "step": 2371 }, { "epoch": 0.25, "learning_rate": 5.943840865495758e-05, "loss": 0.7492, "step": 2372 }, { "epoch": 0.25, "learning_rate": 5.943793452317769e-05, "loss": 0.9536, "step": 2373 }, { "epoch": 0.25, "learning_rate": 5.9437460193228495e-05, "loss": 1.5481, "step": 2374 }, { "epoch": 0.25, "learning_rate": 5.943698566511322e-05, "loss": 0.5109, "step": 2375 }, { "epoch": 0.25, "learning_rate": 5.9436510938835035e-05, "loss": 1.3623, "step": 2376 }, { "epoch": 0.25, "learning_rate": 5.943603601439716e-05, "loss": 0.7812, "step": 2377 }, { "epoch": 0.25, "learning_rate": 5.943556089180276e-05, "loss": 0.3914, "step": 2378 }, { "epoch": 0.25, "learning_rate": 5.943508557105506e-05, "loss": 2.1447, "step": 2379 }, { "epoch": 0.25, "learning_rate": 5.9434610052157256e-05, "loss": 1.1347, "step": 2380 }, { "epoch": 0.25, "learning_rate": 5.943413433511254e-05, "loss": 1.3108, "step": 2381 }, { "epoch": 0.25, "learning_rate": 5.943365841992412e-05, "loss": 0.426, "step": 2382 }, { "epoch": 0.25, "learning_rate": 5.94331823065952e-05, "loss": 0.9664, "step": 2383 }, { "epoch": 0.25, "learning_rate": 5.943270599512898e-05, "loss": 0.4482, "step": 2384 }, { "epoch": 0.25, "learning_rate": 5.943222948552867e-05, "loss": 0.9426, "step": 2385 }, { "epoch": 0.25, "learning_rate": 5.9431752777797475e-05, "loss": 0.8223, "step": 2386 }, { "epoch": 0.25, "learning_rate": 5.943127587193861e-05, "loss": 1.0236, "step": 2387 }, { "epoch": 0.25, "learning_rate": 5.943079876795528e-05, "loss": 0.3806, "step": 2388 }, { "epoch": 0.25, "learning_rate": 5.94303214658507e-05, "loss": 0.4643, "step": 2389 }, { "epoch": 0.25, "learning_rate": 5.9429843965628086e-05, "loss": 0.3629, "step": 2390 }, { "epoch": 0.25, "learning_rate": 5.942936626729065e-05, "loss": 0.8003, "step": 2391 }, { "epoch": 0.25, "learning_rate": 5.942888837084159e-05, "loss": 0.951, "step": 2392 }, { "epoch": 0.25, "learning_rate": 5.942841027628416e-05, "loss": 1.7588, "step": 2393 }, { "epoch": 0.25, "learning_rate": 5.942793198362155e-05, "loss": 1.0363, "step": 2394 }, { "epoch": 0.25, "learning_rate": 5.942745349285699e-05, "loss": 0.1096, "step": 2395 }, { "epoch": 0.25, "learning_rate": 5.942697480399369e-05, "loss": 0.4748, "step": 2396 }, { "epoch": 0.25, "learning_rate": 5.942649591703489e-05, "loss": 0.6735, "step": 2397 }, { "epoch": 0.25, "learning_rate": 5.94260168319838e-05, "loss": 1.4609, "step": 2398 }, { "epoch": 0.25, "learning_rate": 5.942553754884365e-05, "loss": 0.8257, "step": 2399 }, { "epoch": 0.25, "learning_rate": 5.9425058067617664e-05, "loss": 1.1025, "step": 2400 }, { "epoch": 0.25, "learning_rate": 5.9424578388309075e-05, "loss": 0.6678, "step": 2401 }, { "epoch": 0.25, "learning_rate": 5.942409851092111e-05, "loss": 1.32, "step": 2402 }, { "epoch": 0.25, "learning_rate": 5.9423618435456996e-05, "loss": 1.5019, "step": 2403 }, { "epoch": 0.25, "learning_rate": 5.942313816191997e-05, "loss": 0.451, "step": 2404 }, { "epoch": 0.25, "learning_rate": 5.942265769031326e-05, "loss": 1.5848, "step": 2405 }, { "epoch": 0.25, "learning_rate": 5.9422177020640105e-05, "loss": 0.6211, "step": 2406 }, { "epoch": 0.25, "learning_rate": 5.942169615290374e-05, "loss": 0.3865, "step": 2407 }, { "epoch": 0.25, "learning_rate": 5.94212150871074e-05, "loss": 1.168, "step": 2408 }, { "epoch": 0.25, "learning_rate": 5.942073382325432e-05, "loss": 0.3664, "step": 2409 }, { "epoch": 0.25, "learning_rate": 5.9420252361347745e-05, "loss": 1.0292, "step": 2410 }, { "epoch": 0.25, "learning_rate": 5.941977070139091e-05, "loss": 0.9881, "step": 2411 }, { "epoch": 0.25, "learning_rate": 5.9419288843387076e-05, "loss": 0.6843, "step": 2412 }, { "epoch": 0.25, "learning_rate": 5.941880678733946e-05, "loss": 1.1419, "step": 2413 }, { "epoch": 0.25, "learning_rate": 5.9418324533251326e-05, "loss": 0.5654, "step": 2414 }, { "epoch": 0.25, "learning_rate": 5.9417842081125916e-05, "loss": 0.1629, "step": 2415 }, { "epoch": 0.25, "learning_rate": 5.941735943096647e-05, "loss": 1.2661, "step": 2416 }, { "epoch": 0.25, "learning_rate": 5.941687658277625e-05, "loss": 0.7565, "step": 2417 }, { "epoch": 0.25, "learning_rate": 5.94163935365585e-05, "loss": 1.0842, "step": 2418 }, { "epoch": 0.25, "learning_rate": 5.9415910292316466e-05, "loss": 1.3038, "step": 2419 }, { "epoch": 0.25, "learning_rate": 5.941542685005342e-05, "loss": 0.6296, "step": 2420 }, { "epoch": 0.25, "learning_rate": 5.9414943209772585e-05, "loss": 0.8953, "step": 2421 }, { "epoch": 0.25, "learning_rate": 5.941445937147724e-05, "loss": 0.6164, "step": 2422 }, { "epoch": 0.25, "learning_rate": 5.941397533517064e-05, "loss": 0.3991, "step": 2423 }, { "epoch": 0.25, "learning_rate": 5.941349110085604e-05, "loss": 1.2221, "step": 2424 }, { "epoch": 0.25, "learning_rate": 5.94130066685367e-05, "loss": 1.182, "step": 2425 }, { "epoch": 0.25, "learning_rate": 5.9412522038215885e-05, "loss": 0.4735, "step": 2426 }, { "epoch": 0.25, "learning_rate": 5.941203720989685e-05, "loss": 0.3509, "step": 2427 }, { "epoch": 0.25, "learning_rate": 5.9411552183582866e-05, "loss": 1.5815, "step": 2428 }, { "epoch": 0.25, "learning_rate": 5.9411066959277184e-05, "loss": 1.4229, "step": 2429 }, { "epoch": 0.25, "learning_rate": 5.941058153698309e-05, "loss": 0.7642, "step": 2430 }, { "epoch": 0.25, "learning_rate": 5.941009591670384e-05, "loss": 0.3748, "step": 2431 }, { "epoch": 0.25, "learning_rate": 5.9409610098442706e-05, "loss": 0.2362, "step": 2432 }, { "epoch": 0.25, "learning_rate": 5.940912408220296e-05, "loss": 1.3857, "step": 2433 }, { "epoch": 0.25, "learning_rate": 5.940863786798787e-05, "loss": 1.2678, "step": 2434 }, { "epoch": 0.25, "learning_rate": 5.9408151455800705e-05, "loss": 1.7366, "step": 2435 }, { "epoch": 0.25, "learning_rate": 5.940766484564476e-05, "loss": 1.1844, "step": 2436 }, { "epoch": 0.25, "learning_rate": 5.9407178037523284e-05, "loss": 0.9611, "step": 2437 }, { "epoch": 0.25, "learning_rate": 5.940669103143956e-05, "loss": 1.2574, "step": 2438 }, { "epoch": 0.25, "learning_rate": 5.9406203827396896e-05, "loss": 0.7657, "step": 2439 }, { "epoch": 0.25, "learning_rate": 5.940571642539853e-05, "loss": 1.722, "step": 2440 }, { "epoch": 0.25, "learning_rate": 5.940522882544777e-05, "loss": 0.4804, "step": 2441 }, { "epoch": 0.26, "learning_rate": 5.9404741027547886e-05, "loss": 0.3569, "step": 2442 }, { "epoch": 0.26, "learning_rate": 5.940425303170217e-05, "loss": 0.7239, "step": 2443 }, { "epoch": 0.26, "learning_rate": 5.940376483791389e-05, "loss": 1.3671, "step": 2444 }, { "epoch": 0.26, "learning_rate": 5.940327644618635e-05, "loss": 0.2636, "step": 2445 }, { "epoch": 0.26, "learning_rate": 5.940278785652284e-05, "loss": 0.7056, "step": 2446 }, { "epoch": 0.26, "learning_rate": 5.940229906892664e-05, "loss": 0.9751, "step": 2447 }, { "epoch": 0.26, "learning_rate": 5.940181008340103e-05, "loss": 0.9621, "step": 2448 }, { "epoch": 0.26, "learning_rate": 5.940132089994933e-05, "loss": 0.8322, "step": 2449 }, { "epoch": 0.26, "learning_rate": 5.940083151857481e-05, "loss": 0.3321, "step": 2450 }, { "epoch": 0.26, "learning_rate": 5.940034193928077e-05, "loss": 0.6299, "step": 2451 }, { "epoch": 0.26, "learning_rate": 5.9399852162070514e-05, "loss": 0.7999, "step": 2452 }, { "epoch": 0.26, "learning_rate": 5.939936218694732e-05, "loss": 0.6572, "step": 2453 }, { "epoch": 0.26, "learning_rate": 5.939887201391451e-05, "loss": 0.3988, "step": 2454 }, { "epoch": 0.26, "learning_rate": 5.939838164297537e-05, "loss": 0.2372, "step": 2455 }, { "epoch": 0.26, "learning_rate": 5.939789107413321e-05, "loss": 0.7626, "step": 2456 }, { "epoch": 0.26, "learning_rate": 5.939740030739131e-05, "loss": 0.4466, "step": 2457 }, { "epoch": 0.26, "learning_rate": 5.9396909342753e-05, "loss": 1.2644, "step": 2458 }, { "epoch": 0.26, "learning_rate": 5.939641818022157e-05, "loss": 0.6405, "step": 2459 }, { "epoch": 0.26, "learning_rate": 5.939592681980033e-05, "loss": 0.926, "step": 2460 }, { "epoch": 0.26, "learning_rate": 5.9395435261492596e-05, "loss": 1.0053, "step": 2461 }, { "epoch": 0.26, "learning_rate": 5.9394943505301666e-05, "loss": 0.691, "step": 2462 }, { "epoch": 0.26, "learning_rate": 5.9394451551230856e-05, "loss": 1.3618, "step": 2463 }, { "epoch": 0.26, "learning_rate": 5.939395939928347e-05, "loss": 0.4342, "step": 2464 }, { "epoch": 0.26, "learning_rate": 5.939346704946283e-05, "loss": 0.4194, "step": 2465 }, { "epoch": 0.26, "learning_rate": 5.939297450177225e-05, "loss": 0.6301, "step": 2466 }, { "epoch": 0.26, "learning_rate": 5.939248175621504e-05, "loss": 0.4475, "step": 2467 }, { "epoch": 0.26, "learning_rate": 5.9391988812794524e-05, "loss": 1.3916, "step": 2468 }, { "epoch": 0.26, "learning_rate": 5.9391495671514014e-05, "loss": 0.814, "step": 2469 }, { "epoch": 0.26, "learning_rate": 5.939100233237683e-05, "loss": 0.9706, "step": 2470 }, { "epoch": 0.26, "learning_rate": 5.939050879538631e-05, "loss": 0.782, "step": 2471 }, { "epoch": 0.26, "learning_rate": 5.9390015060545746e-05, "loss": 0.8421, "step": 2472 }, { "epoch": 0.26, "learning_rate": 5.9389521127858473e-05, "loss": 0.4413, "step": 2473 }, { "epoch": 0.26, "learning_rate": 5.938902699732784e-05, "loss": 0.6026, "step": 2474 }, { "epoch": 0.26, "learning_rate": 5.9388532668957135e-05, "loss": 1.3496, "step": 2475 }, { "epoch": 0.26, "learning_rate": 5.938803814274971e-05, "loss": 0.4738, "step": 2476 }, { "epoch": 0.26, "learning_rate": 5.938754341870889e-05, "loss": 0.3504, "step": 2477 }, { "epoch": 0.26, "learning_rate": 5.9387048496838e-05, "loss": 0.6569, "step": 2478 }, { "epoch": 0.26, "learning_rate": 5.938655337714038e-05, "loss": 0.5622, "step": 2479 }, { "epoch": 0.26, "learning_rate": 5.938605805961935e-05, "loss": 1.5717, "step": 2480 }, { "epoch": 0.26, "learning_rate": 5.938556254427827e-05, "loss": 1.1482, "step": 2481 }, { "epoch": 0.26, "learning_rate": 5.9385066831120446e-05, "loss": 1.0061, "step": 2482 }, { "epoch": 0.26, "learning_rate": 5.938457092014923e-05, "loss": 0.4089, "step": 2483 }, { "epoch": 0.26, "learning_rate": 5.9384074811367956e-05, "loss": 0.7128, "step": 2484 }, { "epoch": 0.26, "learning_rate": 5.938357850477997e-05, "loss": 1.7269, "step": 2485 }, { "epoch": 0.26, "learning_rate": 5.938308200038861e-05, "loss": 1.3817, "step": 2486 }, { "epoch": 0.26, "learning_rate": 5.938258529819722e-05, "loss": 0.5303, "step": 2487 }, { "epoch": 0.26, "learning_rate": 5.938208839820913e-05, "loss": 1.447, "step": 2488 }, { "epoch": 0.26, "learning_rate": 5.9381591300427695e-05, "loss": 0.8087, "step": 2489 }, { "epoch": 0.26, "learning_rate": 5.938109400485627e-05, "loss": 0.7273, "step": 2490 }, { "epoch": 0.26, "learning_rate": 5.93805965114982e-05, "loss": 1.0258, "step": 2491 }, { "epoch": 0.26, "learning_rate": 5.938009882035682e-05, "loss": 0.8635, "step": 2492 }, { "epoch": 0.26, "learning_rate": 5.937960093143549e-05, "loss": 0.163, "step": 2493 }, { "epoch": 0.26, "learning_rate": 5.937910284473756e-05, "loss": 1.4456, "step": 2494 }, { "epoch": 0.26, "learning_rate": 5.9378604560266384e-05, "loss": 0.5398, "step": 2495 }, { "epoch": 0.26, "learning_rate": 5.937810607802532e-05, "loss": 1.6963, "step": 2496 }, { "epoch": 0.26, "learning_rate": 5.937760739801771e-05, "loss": 0.3557, "step": 2497 }, { "epoch": 0.26, "learning_rate": 5.937710852024693e-05, "loss": 1.1284, "step": 2498 }, { "epoch": 0.26, "learning_rate": 5.937660944471632e-05, "loss": 1.0145, "step": 2499 }, { "epoch": 0.26, "learning_rate": 5.9376110171429266e-05, "loss": 1.0856, "step": 2500 }, { "epoch": 0.26, "learning_rate": 5.9375610700389096e-05, "loss": 0.2365, "step": 2501 }, { "epoch": 0.26, "learning_rate": 5.93751110315992e-05, "loss": 0.5598, "step": 2502 }, { "epoch": 0.26, "learning_rate": 5.937461116506292e-05, "loss": 0.8969, "step": 2503 }, { "epoch": 0.26, "learning_rate": 5.937411110078363e-05, "loss": 0.5588, "step": 2504 }, { "epoch": 0.26, "learning_rate": 5.9373610838764704e-05, "loss": 0.4789, "step": 2505 }, { "epoch": 0.26, "learning_rate": 5.93731103790095e-05, "loss": 0.8494, "step": 2506 }, { "epoch": 0.26, "learning_rate": 5.937260972152139e-05, "loss": 0.732, "step": 2507 }, { "epoch": 0.26, "learning_rate": 5.937210886630375e-05, "loss": 0.3265, "step": 2508 }, { "epoch": 0.26, "learning_rate": 5.937160781335994e-05, "loss": 1.1373, "step": 2509 }, { "epoch": 0.26, "learning_rate": 5.937110656269333e-05, "loss": 1.0949, "step": 2510 }, { "epoch": 0.26, "learning_rate": 5.937060511430731e-05, "loss": 0.8844, "step": 2511 }, { "epoch": 0.26, "learning_rate": 5.9370103468205245e-05, "loss": 0.6912, "step": 2512 }, { "epoch": 0.26, "learning_rate": 5.936960162439053e-05, "loss": 1.2181, "step": 2513 }, { "epoch": 0.26, "learning_rate": 5.9369099582866516e-05, "loss": 0.6284, "step": 2514 }, { "epoch": 0.26, "learning_rate": 5.93685973436366e-05, "loss": 1.3012, "step": 2515 }, { "epoch": 0.26, "learning_rate": 5.936809490670415e-05, "loss": 1.5047, "step": 2516 }, { "epoch": 0.26, "learning_rate": 5.936759227207257e-05, "loss": 0.7484, "step": 2517 }, { "epoch": 0.26, "learning_rate": 5.9367089439745225e-05, "loss": 0.8328, "step": 2518 }, { "epoch": 0.26, "learning_rate": 5.93665864097255e-05, "loss": 0.6068, "step": 2519 }, { "epoch": 0.26, "learning_rate": 5.936608318201678e-05, "loss": 1.4214, "step": 2520 }, { "epoch": 0.26, "learning_rate": 5.936557975662248e-05, "loss": 0.4492, "step": 2521 }, { "epoch": 0.26, "learning_rate": 5.936507613354596e-05, "loss": 0.3957, "step": 2522 }, { "epoch": 0.26, "learning_rate": 5.936457231279062e-05, "loss": 0.5441, "step": 2523 }, { "epoch": 0.26, "learning_rate": 5.936406829435984e-05, "loss": 0.898, "step": 2524 }, { "epoch": 0.26, "learning_rate": 5.9363564078257025e-05, "loss": 0.3645, "step": 2525 }, { "epoch": 0.26, "learning_rate": 5.936305966448558e-05, "loss": 0.4549, "step": 2526 }, { "epoch": 0.26, "learning_rate": 5.936255505304888e-05, "loss": 1.0317, "step": 2527 }, { "epoch": 0.26, "learning_rate": 5.9362050243950315e-05, "loss": 0.7594, "step": 2528 }, { "epoch": 0.26, "learning_rate": 5.9361545237193315e-05, "loss": 1.0476, "step": 2529 }, { "epoch": 0.26, "learning_rate": 5.9361040032781265e-05, "loss": 1.3252, "step": 2530 }, { "epoch": 0.26, "learning_rate": 5.9360534630717546e-05, "loss": 0.9272, "step": 2531 }, { "epoch": 0.26, "learning_rate": 5.9360029031005594e-05, "loss": 1.073, "step": 2532 }, { "epoch": 0.26, "learning_rate": 5.9359523233648796e-05, "loss": 0.5463, "step": 2533 }, { "epoch": 0.26, "learning_rate": 5.935901723865055e-05, "loss": 1.3266, "step": 2534 }, { "epoch": 0.26, "learning_rate": 5.935851104601427e-05, "loss": 1.0634, "step": 2535 }, { "epoch": 0.26, "learning_rate": 5.9358004655743364e-05, "loss": 0.4737, "step": 2536 }, { "epoch": 0.26, "learning_rate": 5.935749806784124e-05, "loss": 0.8927, "step": 2537 }, { "epoch": 0.27, "learning_rate": 5.935699128231131e-05, "loss": 1.4821, "step": 2538 }, { "epoch": 0.27, "learning_rate": 5.935648429915698e-05, "loss": 0.5721, "step": 2539 }, { "epoch": 0.27, "learning_rate": 5.935597711838167e-05, "loss": 1.0461, "step": 2540 }, { "epoch": 0.27, "learning_rate": 5.9355469739988786e-05, "loss": 0.8125, "step": 2541 }, { "epoch": 0.27, "learning_rate": 5.9354962163981756e-05, "loss": 0.8475, "step": 2542 }, { "epoch": 0.27, "learning_rate": 5.935445439036398e-05, "loss": 1.5665, "step": 2543 }, { "epoch": 0.27, "learning_rate": 5.93539464191389e-05, "loss": 0.6302, "step": 2544 }, { "epoch": 0.27, "learning_rate": 5.935343825030991e-05, "loss": 0.7478, "step": 2545 }, { "epoch": 0.27, "learning_rate": 5.935292988388044e-05, "loss": 1.1032, "step": 2546 }, { "epoch": 0.27, "learning_rate": 5.935242131985392e-05, "loss": 0.4404, "step": 2547 }, { "epoch": 0.27, "learning_rate": 5.935191255823377e-05, "loss": 0.7717, "step": 2548 }, { "epoch": 0.27, "learning_rate": 5.93514035990234e-05, "loss": 0.2387, "step": 2549 }, { "epoch": 0.27, "learning_rate": 5.935089444222625e-05, "loss": 0.7637, "step": 2550 }, { "epoch": 0.27, "learning_rate": 5.935038508784576e-05, "loss": 1.3851, "step": 2551 }, { "epoch": 0.27, "learning_rate": 5.934987553588534e-05, "loss": 0.5588, "step": 2552 }, { "epoch": 0.27, "learning_rate": 5.934936578634842e-05, "loss": 0.4468, "step": 2553 }, { "epoch": 0.27, "learning_rate": 5.934885583923843e-05, "loss": 0.3118, "step": 2554 }, { "epoch": 0.27, "learning_rate": 5.934834569455882e-05, "loss": 0.8087, "step": 2555 }, { "epoch": 0.27, "learning_rate": 5.934783535231302e-05, "loss": 1.7367, "step": 2556 }, { "epoch": 0.27, "learning_rate": 5.9347324812504455e-05, "loss": 0.7655, "step": 2557 }, { "epoch": 0.27, "learning_rate": 5.934681407513656e-05, "loss": 1.0987, "step": 2558 }, { "epoch": 0.27, "learning_rate": 5.934630314021278e-05, "loss": 1.3155, "step": 2559 }, { "epoch": 0.27, "learning_rate": 5.934579200773656e-05, "loss": 1.1531, "step": 2560 }, { "epoch": 0.27, "learning_rate": 5.9345280677711326e-05, "loss": 0.3394, "step": 2561 }, { "epoch": 0.27, "learning_rate": 5.934476915014053e-05, "loss": 0.9591, "step": 2562 }, { "epoch": 0.27, "learning_rate": 5.9344257425027624e-05, "loss": 0.3065, "step": 2563 }, { "epoch": 0.27, "learning_rate": 5.934374550237603e-05, "loss": 0.4265, "step": 2564 }, { "epoch": 0.27, "learning_rate": 5.9343233382189216e-05, "loss": 0.7292, "step": 2565 }, { "epoch": 0.27, "learning_rate": 5.934272106447062e-05, "loss": 0.7156, "step": 2566 }, { "epoch": 0.27, "learning_rate": 5.934220854922369e-05, "loss": 0.8719, "step": 2567 }, { "epoch": 0.27, "learning_rate": 5.9341695836451874e-05, "loss": 0.4372, "step": 2568 }, { "epoch": 0.27, "learning_rate": 5.9341182926158636e-05, "loss": 0.9865, "step": 2569 }, { "epoch": 0.27, "learning_rate": 5.9340669818347405e-05, "loss": 0.4374, "step": 2570 }, { "epoch": 0.27, "learning_rate": 5.934015651302167e-05, "loss": 0.4535, "step": 2571 }, { "epoch": 0.27, "learning_rate": 5.9339643010184854e-05, "loss": 1.0484, "step": 2572 }, { "epoch": 0.27, "learning_rate": 5.9339129309840426e-05, "loss": 0.5489, "step": 2573 }, { "epoch": 0.27, "learning_rate": 5.933861541199184e-05, "loss": 0.8986, "step": 2574 }, { "epoch": 0.27, "learning_rate": 5.933810131664257e-05, "loss": 1.125, "step": 2575 }, { "epoch": 0.27, "learning_rate": 5.9337587023796066e-05, "loss": 0.8883, "step": 2576 }, { "epoch": 0.27, "learning_rate": 5.933707253345578e-05, "loss": 0.1541, "step": 2577 }, { "epoch": 0.27, "learning_rate": 5.933655784562519e-05, "loss": 0.6773, "step": 2578 }, { "epoch": 0.27, "learning_rate": 5.9336042960307765e-05, "loss": 0.4451, "step": 2579 }, { "epoch": 0.27, "learning_rate": 5.933552787750696e-05, "loss": 0.8857, "step": 2580 }, { "epoch": 0.27, "learning_rate": 5.9335012597226234e-05, "loss": 1.3809, "step": 2581 }, { "epoch": 0.27, "learning_rate": 5.933449711946908e-05, "loss": 0.7226, "step": 2582 }, { "epoch": 0.27, "learning_rate": 5.933398144423895e-05, "loss": 0.9354, "step": 2583 }, { "epoch": 0.27, "learning_rate": 5.9333465571539325e-05, "loss": 0.8022, "step": 2584 }, { "epoch": 0.27, "learning_rate": 5.9332949501373665e-05, "loss": 1.3622, "step": 2585 }, { "epoch": 0.27, "learning_rate": 5.933243323374546e-05, "loss": 0.5013, "step": 2586 }, { "epoch": 0.27, "learning_rate": 5.933191676865818e-05, "loss": 0.9714, "step": 2587 }, { "epoch": 0.27, "learning_rate": 5.93314001061153e-05, "loss": 0.596, "step": 2588 }, { "epoch": 0.27, "learning_rate": 5.933088324612028e-05, "loss": 1.305, "step": 2589 }, { "epoch": 0.27, "learning_rate": 5.933036618867663e-05, "loss": 1.1218, "step": 2590 }, { "epoch": 0.27, "learning_rate": 5.932984893378782e-05, "loss": 1.6575, "step": 2591 }, { "epoch": 0.27, "learning_rate": 5.932933148145733e-05, "loss": 1.0962, "step": 2592 }, { "epoch": 0.27, "learning_rate": 5.9328813831688636e-05, "loss": 0.9458, "step": 2593 }, { "epoch": 0.27, "learning_rate": 5.932829598448523e-05, "loss": 0.5359, "step": 2594 }, { "epoch": 0.27, "learning_rate": 5.93277779398506e-05, "loss": 1.1678, "step": 2595 }, { "epoch": 0.27, "learning_rate": 5.9327259697788234e-05, "loss": 0.8146, "step": 2596 }, { "epoch": 0.27, "learning_rate": 5.9326741258301614e-05, "loss": 1.0887, "step": 2597 }, { "epoch": 0.27, "learning_rate": 5.9326222621394234e-05, "loss": 1.0693, "step": 2598 }, { "epoch": 0.27, "learning_rate": 5.932570378706958e-05, "loss": 0.642, "step": 2599 }, { "epoch": 0.27, "learning_rate": 5.932518475533116e-05, "loss": 0.9722, "step": 2600 }, { "epoch": 0.27, "learning_rate": 5.932466552618245e-05, "loss": 1.2252, "step": 2601 }, { "epoch": 0.27, "learning_rate": 5.932414609962696e-05, "loss": 0.5287, "step": 2602 }, { "epoch": 0.27, "learning_rate": 5.932362647566818e-05, "loss": 2.2017, "step": 2603 }, { "epoch": 0.27, "learning_rate": 5.93231066543096e-05, "loss": 1.4378, "step": 2604 }, { "epoch": 0.27, "learning_rate": 5.932258663555473e-05, "loss": 1.2446, "step": 2605 }, { "epoch": 0.27, "learning_rate": 5.932206641940707e-05, "loss": 0.4927, "step": 2606 }, { "epoch": 0.27, "learning_rate": 5.932154600587012e-05, "loss": 0.471, "step": 2607 }, { "epoch": 0.27, "learning_rate": 5.932102539494738e-05, "loss": 1.3291, "step": 2608 }, { "epoch": 0.27, "learning_rate": 5.932050458664236e-05, "loss": 0.8707, "step": 2609 }, { "epoch": 0.27, "learning_rate": 5.931998358095856e-05, "loss": 1.3362, "step": 2610 }, { "epoch": 0.27, "learning_rate": 5.931946237789949e-05, "loss": 0.6123, "step": 2611 }, { "epoch": 0.27, "learning_rate": 5.931894097746867e-05, "loss": 1.538, "step": 2612 }, { "epoch": 0.27, "learning_rate": 5.9318419379669586e-05, "loss": 1.3753, "step": 2613 }, { "epoch": 0.27, "learning_rate": 5.931789758450578e-05, "loss": 0.3828, "step": 2614 }, { "epoch": 0.27, "learning_rate": 5.9317375591980736e-05, "loss": 1.0363, "step": 2615 }, { "epoch": 0.27, "learning_rate": 5.931685340209797e-05, "loss": 1.2527, "step": 2616 }, { "epoch": 0.27, "learning_rate": 5.931633101486102e-05, "loss": 0.8679, "step": 2617 }, { "epoch": 0.27, "learning_rate": 5.9315808430273385e-05, "loss": 1.1277, "step": 2618 }, { "epoch": 0.27, "learning_rate": 5.9315285648338584e-05, "loss": 1.0704, "step": 2619 }, { "epoch": 0.27, "learning_rate": 5.9314762669060144e-05, "loss": 0.9855, "step": 2620 }, { "epoch": 0.27, "learning_rate": 5.931423949244158e-05, "loss": 0.4528, "step": 2621 }, { "epoch": 0.27, "learning_rate": 5.9313716118486416e-05, "loss": 0.4909, "step": 2622 }, { "epoch": 0.27, "learning_rate": 5.931319254719817e-05, "loss": 0.2792, "step": 2623 }, { "epoch": 0.27, "learning_rate": 5.931266877858037e-05, "loss": 1.2182, "step": 2624 }, { "epoch": 0.27, "learning_rate": 5.931214481263654e-05, "loss": 1.5253, "step": 2625 }, { "epoch": 0.27, "learning_rate": 5.9311620649370224e-05, "loss": 1.1284, "step": 2626 }, { "epoch": 0.27, "learning_rate": 5.9311096288784924e-05, "loss": 0.3393, "step": 2627 }, { "epoch": 0.27, "learning_rate": 5.9310571730884186e-05, "loss": 0.524, "step": 2628 }, { "epoch": 0.27, "learning_rate": 5.9310046975671535e-05, "loss": 0.4817, "step": 2629 }, { "epoch": 0.27, "learning_rate": 5.9309522023150514e-05, "loss": 0.2893, "step": 2630 }, { "epoch": 0.27, "learning_rate": 5.930899687332463e-05, "loss": 0.3368, "step": 2631 }, { "epoch": 0.27, "learning_rate": 5.930847152619745e-05, "loss": 0.3209, "step": 2632 }, { "epoch": 0.27, "learning_rate": 5.9307945981772496e-05, "loss": 0.3961, "step": 2633 }, { "epoch": 0.28, "learning_rate": 5.930742024005331e-05, "loss": 1.2003, "step": 2634 }, { "epoch": 0.28, "learning_rate": 5.930689430104342e-05, "loss": 1.0457, "step": 2635 }, { "epoch": 0.28, "learning_rate": 5.930636816474638e-05, "loss": 0.5862, "step": 2636 }, { "epoch": 0.28, "learning_rate": 5.9305841831165726e-05, "loss": 0.7107, "step": 2637 }, { "epoch": 0.28, "learning_rate": 5.9305315300305005e-05, "loss": 1.0804, "step": 2638 }, { "epoch": 0.28, "learning_rate": 5.9304788572167756e-05, "loss": 0.4147, "step": 2639 }, { "epoch": 0.28, "learning_rate": 5.9304261646757523e-05, "loss": 0.3859, "step": 2640 }, { "epoch": 0.28, "learning_rate": 5.930373452407786e-05, "loss": 0.6715, "step": 2641 }, { "epoch": 0.28, "learning_rate": 5.9303207204132315e-05, "loss": 0.4899, "step": 2642 }, { "epoch": 0.28, "learning_rate": 5.9302679686924435e-05, "loss": 0.6024, "step": 2643 }, { "epoch": 0.28, "learning_rate": 5.930215197245777e-05, "loss": 1.3914, "step": 2644 }, { "epoch": 0.28, "learning_rate": 5.930162406073588e-05, "loss": 0.6193, "step": 2645 }, { "epoch": 0.28, "learning_rate": 5.93010959517623e-05, "loss": 0.9948, "step": 2646 }, { "epoch": 0.28, "learning_rate": 5.930056764554061e-05, "loss": 0.4151, "step": 2647 }, { "epoch": 0.28, "learning_rate": 5.930003914207435e-05, "loss": 1.4833, "step": 2648 }, { "epoch": 0.28, "learning_rate": 5.929951044136709e-05, "loss": 1.618, "step": 2649 }, { "epoch": 0.28, "learning_rate": 5.929898154342237e-05, "loss": 0.2565, "step": 2650 }, { "epoch": 0.28, "learning_rate": 5.929845244824376e-05, "loss": 0.5999, "step": 2651 }, { "epoch": 0.28, "learning_rate": 5.929792315583483e-05, "loss": 1.0457, "step": 2652 }, { "epoch": 0.28, "learning_rate": 5.929739366619914e-05, "loss": 0.3861, "step": 2653 }, { "epoch": 0.28, "learning_rate": 5.929686397934024e-05, "loss": 1.3846, "step": 2654 }, { "epoch": 0.28, "learning_rate": 5.9296334095261706e-05, "loss": 1.0072, "step": 2655 }, { "epoch": 0.28, "learning_rate": 5.929580401396712e-05, "loss": 0.8833, "step": 2656 }, { "epoch": 0.28, "learning_rate": 5.929527373546003e-05, "loss": 0.4797, "step": 2657 }, { "epoch": 0.28, "learning_rate": 5.9294743259744006e-05, "loss": 0.8734, "step": 2658 }, { "epoch": 0.28, "learning_rate": 5.929421258682262e-05, "loss": 0.4785, "step": 2659 }, { "epoch": 0.28, "learning_rate": 5.929368171669947e-05, "loss": 1.4404, "step": 2660 }, { "epoch": 0.28, "learning_rate": 5.92931506493781e-05, "loss": 0.6345, "step": 2661 }, { "epoch": 0.28, "learning_rate": 5.9292619384862085e-05, "loss": 1.1783, "step": 2662 }, { "epoch": 0.28, "learning_rate": 5.929208792315502e-05, "loss": 0.4316, "step": 2663 }, { "epoch": 0.28, "learning_rate": 5.929155626426046e-05, "loss": 1.3434, "step": 2664 }, { "epoch": 0.28, "learning_rate": 5.929102440818201e-05, "loss": 1.3378, "step": 2665 }, { "epoch": 0.28, "learning_rate": 5.929049235492323e-05, "loss": 1.2747, "step": 2666 }, { "epoch": 0.28, "learning_rate": 5.9289960104487715e-05, "loss": 0.406, "step": 2667 }, { "epoch": 0.28, "learning_rate": 5.9289427656879034e-05, "loss": 1.0988, "step": 2668 }, { "epoch": 0.28, "learning_rate": 5.9288895012100794e-05, "loss": 0.5087, "step": 2669 }, { "epoch": 0.28, "learning_rate": 5.928836217015655e-05, "loss": 1.0425, "step": 2670 }, { "epoch": 0.28, "learning_rate": 5.928782913104991e-05, "loss": 1.5172, "step": 2671 }, { "epoch": 0.28, "learning_rate": 5.9287295894784454e-05, "loss": 1.3561, "step": 2672 }, { "epoch": 0.28, "learning_rate": 5.928676246136378e-05, "loss": 0.4556, "step": 2673 }, { "epoch": 0.28, "learning_rate": 5.9286228830791475e-05, "loss": 0.8976, "step": 2674 }, { "epoch": 0.28, "learning_rate": 5.9285695003071125e-05, "loss": 1.5591, "step": 2675 }, { "epoch": 0.28, "learning_rate": 5.928516097820633e-05, "loss": 0.7666, "step": 2676 }, { "epoch": 0.28, "learning_rate": 5.928462675620069e-05, "loss": 1.5987, "step": 2677 }, { "epoch": 0.28, "learning_rate": 5.928409233705779e-05, "loss": 1.0594, "step": 2678 }, { "epoch": 0.28, "learning_rate": 5.928355772078123e-05, "loss": 0.3165, "step": 2679 }, { "epoch": 0.28, "learning_rate": 5.928302290737461e-05, "loss": 1.1122, "step": 2680 }, { "epoch": 0.28, "learning_rate": 5.9282487896841525e-05, "loss": 1.1447, "step": 2681 }, { "epoch": 0.28, "learning_rate": 5.92819526891856e-05, "loss": 1.9509, "step": 2682 }, { "epoch": 0.28, "learning_rate": 5.9281417284410405e-05, "loss": 0.7571, "step": 2683 }, { "epoch": 0.28, "learning_rate": 5.928088168251957e-05, "loss": 0.8399, "step": 2684 }, { "epoch": 0.28, "learning_rate": 5.928034588351669e-05, "loss": 2.0131, "step": 2685 }, { "epoch": 0.28, "learning_rate": 5.9279809887405376e-05, "loss": 1.3577, "step": 2686 }, { "epoch": 0.28, "learning_rate": 5.9279273694189226e-05, "loss": 0.3802, "step": 2687 }, { "epoch": 0.28, "learning_rate": 5.927873730387185e-05, "loss": 1.363, "step": 2688 }, { "epoch": 0.28, "learning_rate": 5.927820071645688e-05, "loss": 0.6278, "step": 2689 }, { "epoch": 0.28, "learning_rate": 5.927766393194791e-05, "loss": 0.5664, "step": 2690 }, { "epoch": 0.28, "learning_rate": 5.927712695034855e-05, "loss": 0.6498, "step": 2691 }, { "epoch": 0.28, "learning_rate": 5.927658977166242e-05, "loss": 0.22, "step": 2692 }, { "epoch": 0.28, "learning_rate": 5.9276052395893145e-05, "loss": 1.4021, "step": 2693 }, { "epoch": 0.28, "learning_rate": 5.9275514823044334e-05, "loss": 1.2245, "step": 2694 }, { "epoch": 0.28, "learning_rate": 5.9274977053119606e-05, "loss": 0.9353, "step": 2695 }, { "epoch": 0.28, "learning_rate": 5.927443908612258e-05, "loss": 0.7509, "step": 2696 }, { "epoch": 0.28, "learning_rate": 5.9273900922056887e-05, "loss": 1.176, "step": 2697 }, { "epoch": 0.28, "learning_rate": 5.927336256092614e-05, "loss": 1.554, "step": 2698 }, { "epoch": 0.28, "learning_rate": 5.927282400273396e-05, "loss": 0.7096, "step": 2699 }, { "epoch": 0.28, "learning_rate": 5.927228524748398e-05, "loss": 1.4928, "step": 2700 }, { "epoch": 0.28, "learning_rate": 5.927174629517983e-05, "loss": 0.3737, "step": 2701 }, { "epoch": 0.28, "learning_rate": 5.9271207145825126e-05, "loss": 1.368, "step": 2702 }, { "epoch": 0.28, "learning_rate": 5.927066779942351e-05, "loss": 0.9825, "step": 2703 }, { "epoch": 0.28, "learning_rate": 5.9270128255978605e-05, "loss": 0.8878, "step": 2704 }, { "epoch": 0.28, "learning_rate": 5.9269588515494046e-05, "loss": 0.4844, "step": 2705 }, { "epoch": 0.28, "learning_rate": 5.926904857797346e-05, "loss": 0.7186, "step": 2706 }, { "epoch": 0.28, "learning_rate": 5.926850844342049e-05, "loss": 1.0444, "step": 2707 }, { "epoch": 0.28, "learning_rate": 5.9267968111838776e-05, "loss": 0.433, "step": 2708 }, { "epoch": 0.28, "learning_rate": 5.926742758323195e-05, "loss": 1.0749, "step": 2709 }, { "epoch": 0.28, "learning_rate": 5.9266886857603636e-05, "loss": 1.165, "step": 2710 }, { "epoch": 0.28, "learning_rate": 5.92663459349575e-05, "loss": 0.4987, "step": 2711 }, { "epoch": 0.28, "learning_rate": 5.9265804815297166e-05, "loss": 1.3191, "step": 2712 }, { "epoch": 0.28, "learning_rate": 5.926526349862629e-05, "loss": 0.3793, "step": 2713 }, { "epoch": 0.28, "learning_rate": 5.92647219849485e-05, "loss": 0.4512, "step": 2714 }, { "epoch": 0.28, "learning_rate": 5.9264180274267445e-05, "loss": 0.4635, "step": 2715 }, { "epoch": 0.28, "learning_rate": 5.9263638366586776e-05, "loss": 1.1169, "step": 2716 }, { "epoch": 0.28, "learning_rate": 5.9263096261910144e-05, "loss": 1.0108, "step": 2717 }, { "epoch": 0.28, "learning_rate": 5.92625539602412e-05, "loss": 0.6084, "step": 2718 }, { "epoch": 0.28, "learning_rate": 5.926201146158358e-05, "loss": 0.4117, "step": 2719 }, { "epoch": 0.28, "learning_rate": 5.9261468765940955e-05, "loss": 1.4335, "step": 2720 }, { "epoch": 0.28, "learning_rate": 5.9260925873316965e-05, "loss": 0.3926, "step": 2721 }, { "epoch": 0.28, "learning_rate": 5.926038278371526e-05, "loss": 0.2936, "step": 2722 }, { "epoch": 0.28, "learning_rate": 5.925983949713952e-05, "loss": 1.2136, "step": 2723 }, { "epoch": 0.28, "learning_rate": 5.925929601359338e-05, "loss": 1.1828, "step": 2724 }, { "epoch": 0.28, "learning_rate": 5.9258752333080505e-05, "loss": 0.8575, "step": 2725 }, { "epoch": 0.28, "learning_rate": 5.925820845560455e-05, "loss": 1.5144, "step": 2726 }, { "epoch": 0.28, "learning_rate": 5.9257664381169185e-05, "loss": 1.1017, "step": 2727 }, { "epoch": 0.28, "learning_rate": 5.925712010977807e-05, "loss": 0.9514, "step": 2728 }, { "epoch": 0.29, "learning_rate": 5.925657564143487e-05, "loss": 1.3756, "step": 2729 }, { "epoch": 0.29, "learning_rate": 5.925603097614325e-05, "loss": 0.6636, "step": 2730 }, { "epoch": 0.29, "learning_rate": 5.9255486113906874e-05, "loss": 0.1571, "step": 2731 }, { "epoch": 0.29, "learning_rate": 5.925494105472941e-05, "loss": 1.155, "step": 2732 }, { "epoch": 0.29, "learning_rate": 5.925439579861452e-05, "loss": 0.3547, "step": 2733 }, { "epoch": 0.29, "learning_rate": 5.925385034556589e-05, "loss": 1.2675, "step": 2734 }, { "epoch": 0.29, "learning_rate": 5.9253304695587185e-05, "loss": 0.9908, "step": 2735 }, { "epoch": 0.29, "learning_rate": 5.925275884868207e-05, "loss": 0.3143, "step": 2736 }, { "epoch": 0.29, "learning_rate": 5.925221280485424e-05, "loss": 0.4503, "step": 2737 }, { "epoch": 0.29, "learning_rate": 5.925166656410735e-05, "loss": 0.9399, "step": 2738 }, { "epoch": 0.29, "learning_rate": 5.925112012644508e-05, "loss": 1.061, "step": 2739 }, { "epoch": 0.29, "learning_rate": 5.925057349187112e-05, "loss": 0.8009, "step": 2740 }, { "epoch": 0.29, "learning_rate": 5.9250026660389144e-05, "loss": 0.9961, "step": 2741 }, { "epoch": 0.29, "learning_rate": 5.924947963200283e-05, "loss": 0.1827, "step": 2742 }, { "epoch": 0.29, "learning_rate": 5.924893240671587e-05, "loss": 0.7969, "step": 2743 }, { "epoch": 0.29, "learning_rate": 5.9248384984531924e-05, "loss": 1.0558, "step": 2744 }, { "epoch": 0.29, "learning_rate": 5.924783736545471e-05, "loss": 1.5229, "step": 2745 }, { "epoch": 0.29, "learning_rate": 5.924728954948789e-05, "loss": 0.1764, "step": 2746 }, { "epoch": 0.29, "learning_rate": 5.9246741536635164e-05, "loss": 1.0094, "step": 2747 }, { "epoch": 0.29, "learning_rate": 5.924619332690021e-05, "loss": 0.5965, "step": 2748 }, { "epoch": 0.29, "learning_rate": 5.9245644920286736e-05, "loss": 1.0913, "step": 2749 }, { "epoch": 0.29, "learning_rate": 5.924509631679842e-05, "loss": 0.3368, "step": 2750 }, { "epoch": 0.29, "learning_rate": 5.9244547516438956e-05, "loss": 0.2505, "step": 2751 }, { "epoch": 0.29, "learning_rate": 5.9243998519212045e-05, "loss": 0.3146, "step": 2752 }, { "epoch": 0.29, "learning_rate": 5.9243449325121376e-05, "loss": 1.0226, "step": 2753 }, { "epoch": 0.29, "learning_rate": 5.924289993417065e-05, "loss": 0.9015, "step": 2754 }, { "epoch": 0.29, "learning_rate": 5.924235034636357e-05, "loss": 0.2823, "step": 2755 }, { "epoch": 0.29, "learning_rate": 5.924180056170381e-05, "loss": 1.3014, "step": 2756 }, { "epoch": 0.29, "learning_rate": 5.924125058019511e-05, "loss": 0.5031, "step": 2757 }, { "epoch": 0.29, "learning_rate": 5.924070040184114e-05, "loss": 0.7172, "step": 2758 }, { "epoch": 0.29, "learning_rate": 5.9240150026645616e-05, "loss": 1.0156, "step": 2759 }, { "epoch": 0.29, "learning_rate": 5.923959945461226e-05, "loss": 0.7945, "step": 2760 }, { "epoch": 0.29, "learning_rate": 5.9239048685744746e-05, "loss": 0.6136, "step": 2761 }, { "epoch": 0.29, "learning_rate": 5.923849772004681e-05, "loss": 0.2929, "step": 2762 }, { "epoch": 0.29, "learning_rate": 5.923794655752214e-05, "loss": 1.1684, "step": 2763 }, { "epoch": 0.29, "learning_rate": 5.9237395198174453e-05, "loss": 0.5391, "step": 2764 }, { "epoch": 0.29, "learning_rate": 5.923684364200747e-05, "loss": 1.8468, "step": 2765 }, { "epoch": 0.29, "learning_rate": 5.9236291889024886e-05, "loss": 0.9262, "step": 2766 }, { "epoch": 0.29, "learning_rate": 5.9235739939230437e-05, "loss": 0.933, "step": 2767 }, { "epoch": 0.29, "learning_rate": 5.923518779262782e-05, "loss": 0.2285, "step": 2768 }, { "epoch": 0.29, "learning_rate": 5.923463544922077e-05, "loss": 1.9963, "step": 2769 }, { "epoch": 0.29, "learning_rate": 5.9234082909012984e-05, "loss": 0.4567, "step": 2770 }, { "epoch": 0.29, "learning_rate": 5.92335301720082e-05, "loss": 0.3799, "step": 2771 }, { "epoch": 0.29, "learning_rate": 5.923297723821012e-05, "loss": 0.3455, "step": 2772 }, { "epoch": 0.29, "learning_rate": 5.923242410762248e-05, "loss": 0.7612, "step": 2773 }, { "epoch": 0.29, "learning_rate": 5.923187078024901e-05, "loss": 0.9375, "step": 2774 }, { "epoch": 0.29, "learning_rate": 5.9231317256093415e-05, "loss": 1.3596, "step": 2775 }, { "epoch": 0.29, "learning_rate": 5.9230763535159435e-05, "loss": 0.3814, "step": 2776 }, { "epoch": 0.29, "learning_rate": 5.92302096174508e-05, "loss": 0.5135, "step": 2777 }, { "epoch": 0.29, "learning_rate": 5.9229655502971224e-05, "loss": 1.0617, "step": 2778 }, { "epoch": 0.29, "learning_rate": 5.922910119172445e-05, "loss": 1.0885, "step": 2779 }, { "epoch": 0.29, "learning_rate": 5.922854668371421e-05, "loss": 1.2433, "step": 2780 }, { "epoch": 0.29, "learning_rate": 5.9227991978944226e-05, "loss": 0.1814, "step": 2781 }, { "epoch": 0.29, "learning_rate": 5.9227437077418245e-05, "loss": 1.0337, "step": 2782 }, { "epoch": 0.29, "learning_rate": 5.9226881979139995e-05, "loss": 1.4093, "step": 2783 }, { "epoch": 0.29, "learning_rate": 5.922632668411322e-05, "loss": 0.1327, "step": 2784 }, { "epoch": 0.29, "learning_rate": 5.922577119234164e-05, "loss": 0.9838, "step": 2785 }, { "epoch": 0.29, "learning_rate": 5.922521550382901e-05, "loss": 0.3115, "step": 2786 }, { "epoch": 0.29, "learning_rate": 5.922465961857907e-05, "loss": 1.575, "step": 2787 }, { "epoch": 0.29, "learning_rate": 5.922410353659556e-05, "loss": 0.5902, "step": 2788 }, { "epoch": 0.29, "learning_rate": 5.922354725788222e-05, "loss": 1.1626, "step": 2789 }, { "epoch": 0.29, "learning_rate": 5.922299078244279e-05, "loss": 0.6804, "step": 2790 }, { "epoch": 0.29, "learning_rate": 5.922243411028104e-05, "loss": 0.5145, "step": 2791 }, { "epoch": 0.29, "learning_rate": 5.92218772414007e-05, "loss": 0.865, "step": 2792 }, { "epoch": 0.29, "learning_rate": 5.9221320175805506e-05, "loss": 1.0072, "step": 2793 }, { "epoch": 0.29, "learning_rate": 5.9220762913499226e-05, "loss": 1.6327, "step": 2794 }, { "epoch": 0.29, "learning_rate": 5.922020545448562e-05, "loss": 1.2285, "step": 2795 }, { "epoch": 0.29, "learning_rate": 5.9219647798768413e-05, "loss": 1.4209, "step": 2796 }, { "epoch": 0.29, "learning_rate": 5.921908994635138e-05, "loss": 0.4416, "step": 2797 }, { "epoch": 0.29, "learning_rate": 5.921853189723827e-05, "loss": 1.4759, "step": 2798 }, { "epoch": 0.29, "learning_rate": 5.921797365143284e-05, "loss": 0.3907, "step": 2799 }, { "epoch": 0.29, "learning_rate": 5.9217415208938845e-05, "loss": 0.5157, "step": 2800 }, { "epoch": 0.29, "learning_rate": 5.921685656976005e-05, "loss": 0.4241, "step": 2801 }, { "epoch": 0.29, "learning_rate": 5.921629773390021e-05, "loss": 0.886, "step": 2802 }, { "epoch": 0.29, "learning_rate": 5.9215738701363094e-05, "loss": 1.0885, "step": 2803 }, { "epoch": 0.29, "learning_rate": 5.921517947215246e-05, "loss": 1.1068, "step": 2804 }, { "epoch": 0.29, "learning_rate": 5.921462004627207e-05, "loss": 0.7593, "step": 2805 }, { "epoch": 0.29, "learning_rate": 5.92140604237257e-05, "loss": 0.1878, "step": 2806 }, { "epoch": 0.29, "learning_rate": 5.92135006045171e-05, "loss": 0.6355, "step": 2807 }, { "epoch": 0.29, "learning_rate": 5.921294058865005e-05, "loss": 0.7202, "step": 2808 }, { "epoch": 0.29, "learning_rate": 5.9212380376128336e-05, "loss": 0.9624, "step": 2809 }, { "epoch": 0.29, "learning_rate": 5.9211819966955696e-05, "loss": 1.0834, "step": 2810 }, { "epoch": 0.29, "learning_rate": 5.921125936113593e-05, "loss": 0.9718, "step": 2811 }, { "epoch": 0.29, "learning_rate": 5.92106985586728e-05, "loss": 0.0845, "step": 2812 }, { "epoch": 0.29, "learning_rate": 5.921013755957007e-05, "loss": 1.5132, "step": 2813 }, { "epoch": 0.29, "learning_rate": 5.9209576363831534e-05, "loss": 1.0119, "step": 2814 }, { "epoch": 0.29, "learning_rate": 5.920901497146097e-05, "loss": 0.5317, "step": 2815 }, { "epoch": 0.29, "learning_rate": 5.920845338246215e-05, "loss": 0.8008, "step": 2816 }, { "epoch": 0.29, "learning_rate": 5.920789159683886e-05, "loss": 0.3769, "step": 2817 }, { "epoch": 0.29, "learning_rate": 5.920732961459486e-05, "loss": 1.1968, "step": 2818 }, { "epoch": 0.29, "learning_rate": 5.920676743573397e-05, "loss": 1.3696, "step": 2819 }, { "epoch": 0.29, "learning_rate": 5.920620506025994e-05, "loss": 0.3714, "step": 2820 }, { "epoch": 0.29, "learning_rate": 5.9205642488176586e-05, "loss": 1.0457, "step": 2821 }, { "epoch": 0.29, "learning_rate": 5.920507971948767e-05, "loss": 0.8528, "step": 2822 }, { "epoch": 0.29, "learning_rate": 5.9204516754197e-05, "loss": 0.974, "step": 2823 }, { "epoch": 0.29, "learning_rate": 5.920395359230834e-05, "loss": 1.2514, "step": 2824 }, { "epoch": 0.3, "learning_rate": 5.920339023382552e-05, "loss": 0.5525, "step": 2825 }, { "epoch": 0.3, "learning_rate": 5.92028266787523e-05, "loss": 0.9678, "step": 2826 }, { "epoch": 0.3, "learning_rate": 5.920226292709248e-05, "loss": 0.5675, "step": 2827 }, { "epoch": 0.3, "learning_rate": 5.920169897884986e-05, "loss": 1.0024, "step": 2828 }, { "epoch": 0.3, "learning_rate": 5.920113483402824e-05, "loss": 0.5912, "step": 2829 }, { "epoch": 0.3, "learning_rate": 5.920057049263141e-05, "loss": 0.8177, "step": 2830 }, { "epoch": 0.3, "learning_rate": 5.920000595466317e-05, "loss": 1.0191, "step": 2831 }, { "epoch": 0.3, "learning_rate": 5.919944122012732e-05, "loss": 0.7078, "step": 2832 }, { "epoch": 0.3, "learning_rate": 5.919887628902767e-05, "loss": 0.2313, "step": 2833 }, { "epoch": 0.3, "learning_rate": 5.919831116136801e-05, "loss": 0.684, "step": 2834 }, { "epoch": 0.3, "learning_rate": 5.919774583715216e-05, "loss": 0.4811, "step": 2835 }, { "epoch": 0.3, "learning_rate": 5.919718031638391e-05, "loss": 1.5135, "step": 2836 }, { "epoch": 0.3, "learning_rate": 5.9196614599067074e-05, "loss": 1.2611, "step": 2837 }, { "epoch": 0.3, "learning_rate": 5.9196048685205465e-05, "loss": 0.8136, "step": 2838 }, { "epoch": 0.3, "learning_rate": 5.9195482574802884e-05, "loss": 0.8076, "step": 2839 }, { "epoch": 0.3, "learning_rate": 5.9194916267863146e-05, "loss": 0.5588, "step": 2840 }, { "epoch": 0.3, "learning_rate": 5.919434976439006e-05, "loss": 0.4689, "step": 2841 }, { "epoch": 0.3, "learning_rate": 5.919378306438744e-05, "loss": 0.6898, "step": 2842 }, { "epoch": 0.3, "learning_rate": 5.919321616785912e-05, "loss": 1.1822, "step": 2843 }, { "epoch": 0.3, "learning_rate": 5.9192649074808886e-05, "loss": 1.3, "step": 2844 }, { "epoch": 0.3, "learning_rate": 5.9192081785240566e-05, "loss": 0.5203, "step": 2845 }, { "epoch": 0.3, "learning_rate": 5.9191514299157985e-05, "loss": 0.4292, "step": 2846 }, { "epoch": 0.3, "learning_rate": 5.919094661656497e-05, "loss": 1.0435, "step": 2847 }, { "epoch": 0.3, "learning_rate": 5.919037873746532e-05, "loss": 0.3858, "step": 2848 }, { "epoch": 0.3, "learning_rate": 5.9189810661862866e-05, "loss": 0.8147, "step": 2849 }, { "epoch": 0.3, "learning_rate": 5.918924238976145e-05, "loss": 0.8846, "step": 2850 }, { "epoch": 0.3, "learning_rate": 5.918867392116488e-05, "loss": 1.4726, "step": 2851 }, { "epoch": 0.3, "learning_rate": 5.918810525607698e-05, "loss": 0.7187, "step": 2852 }, { "epoch": 0.3, "learning_rate": 5.918753639450159e-05, "loss": 0.9964, "step": 2853 }, { "epoch": 0.3, "learning_rate": 5.918696733644254e-05, "loss": 1.0452, "step": 2854 }, { "epoch": 0.3, "learning_rate": 5.9186398081903655e-05, "loss": 0.9971, "step": 2855 }, { "epoch": 0.3, "learning_rate": 5.9185828630888755e-05, "loss": 0.219, "step": 2856 }, { "epoch": 0.3, "learning_rate": 5.918525898340169e-05, "loss": 0.4714, "step": 2857 }, { "epoch": 0.3, "learning_rate": 5.9184689139446295e-05, "loss": 0.4649, "step": 2858 }, { "epoch": 0.3, "learning_rate": 5.91841190990264e-05, "loss": 0.4831, "step": 2859 }, { "epoch": 0.3, "learning_rate": 5.918354886214584e-05, "loss": 1.1006, "step": 2860 }, { "epoch": 0.3, "learning_rate": 5.918297842880846e-05, "loss": 0.88, "step": 2861 }, { "epoch": 0.3, "learning_rate": 5.91824077990181e-05, "loss": 1.2113, "step": 2862 }, { "epoch": 0.3, "learning_rate": 5.9181836972778595e-05, "loss": 0.8923, "step": 2863 }, { "epoch": 0.3, "learning_rate": 5.918126595009379e-05, "loss": 1.3334, "step": 2864 }, { "epoch": 0.3, "learning_rate": 5.918069473096754e-05, "loss": 0.9472, "step": 2865 }, { "epoch": 0.3, "learning_rate": 5.918012331540367e-05, "loss": 1.6245, "step": 2866 }, { "epoch": 0.3, "learning_rate": 5.917955170340605e-05, "loss": 0.7324, "step": 2867 }, { "epoch": 0.3, "learning_rate": 5.91789798949785e-05, "loss": 0.8754, "step": 2868 }, { "epoch": 0.3, "learning_rate": 5.9178407890124896e-05, "loss": 0.9468, "step": 2869 }, { "epoch": 0.3, "learning_rate": 5.9177835688849074e-05, "loss": 1.0209, "step": 2870 }, { "epoch": 0.3, "learning_rate": 5.917726329115488e-05, "loss": 1.2166, "step": 2871 }, { "epoch": 0.3, "learning_rate": 5.91766906970462e-05, "loss": 0.1996, "step": 2872 }, { "epoch": 0.3, "learning_rate": 5.917611790652684e-05, "loss": 0.6335, "step": 2873 }, { "epoch": 0.3, "learning_rate": 5.9175544919600696e-05, "loss": 0.7249, "step": 2874 }, { "epoch": 0.3, "learning_rate": 5.91749717362716e-05, "loss": 1.3134, "step": 2875 }, { "epoch": 0.3, "learning_rate": 5.917439835654343e-05, "loss": 0.2361, "step": 2876 }, { "epoch": 0.3, "learning_rate": 5.917382478042003e-05, "loss": 1.4289, "step": 2877 }, { "epoch": 0.3, "learning_rate": 5.917325100790527e-05, "loss": 0.5901, "step": 2878 }, { "epoch": 0.3, "learning_rate": 5.917267703900301e-05, "loss": 0.836, "step": 2879 }, { "epoch": 0.3, "learning_rate": 5.9172102873717115e-05, "loss": 0.0416, "step": 2880 }, { "epoch": 0.3, "learning_rate": 5.917152851205145e-05, "loss": 0.6917, "step": 2881 }, { "epoch": 0.3, "learning_rate": 5.917095395400988e-05, "loss": 0.6784, "step": 2882 }, { "epoch": 0.3, "learning_rate": 5.917037919959627e-05, "loss": 1.4727, "step": 2883 }, { "epoch": 0.3, "learning_rate": 5.91698042488145e-05, "loss": 0.9626, "step": 2884 }, { "epoch": 0.3, "learning_rate": 5.9169229101668425e-05, "loss": 0.9842, "step": 2885 }, { "epoch": 0.3, "learning_rate": 5.9168653758161925e-05, "loss": 0.8004, "step": 2886 }, { "epoch": 0.3, "learning_rate": 5.9168078218298876e-05, "loss": 1.2499, "step": 2887 }, { "epoch": 0.3, "learning_rate": 5.916750248208315e-05, "loss": 0.2957, "step": 2888 }, { "epoch": 0.3, "learning_rate": 5.916692654951862e-05, "loss": 0.9792, "step": 2889 }, { "epoch": 0.3, "learning_rate": 5.916635042060917e-05, "loss": 0.7331, "step": 2890 }, { "epoch": 0.3, "learning_rate": 5.916577409535866e-05, "loss": 0.8728, "step": 2891 }, { "epoch": 0.3, "learning_rate": 5.9165197573771e-05, "loss": 1.1234, "step": 2892 }, { "epoch": 0.3, "learning_rate": 5.916462085585004e-05, "loss": 0.5277, "step": 2893 }, { "epoch": 0.3, "learning_rate": 5.9164043941599686e-05, "loss": 0.8045, "step": 2894 }, { "epoch": 0.3, "learning_rate": 5.9163466831023805e-05, "loss": 0.3069, "step": 2895 }, { "epoch": 0.3, "learning_rate": 5.9162889524126294e-05, "loss": 1.1534, "step": 2896 }, { "epoch": 0.3, "learning_rate": 5.9162312020911035e-05, "loss": 0.1646, "step": 2897 }, { "epoch": 0.3, "learning_rate": 5.916173432138191e-05, "loss": 0.6574, "step": 2898 }, { "epoch": 0.3, "learning_rate": 5.916115642554282e-05, "loss": 1.2794, "step": 2899 }, { "epoch": 0.3, "learning_rate": 5.916057833339764e-05, "loss": 1.0739, "step": 2900 }, { "epoch": 0.3, "learning_rate": 5.916000004495027e-05, "loss": 0.7572, "step": 2901 }, { "epoch": 0.3, "learning_rate": 5.915942156020461e-05, "loss": 0.6383, "step": 2902 }, { "epoch": 0.3, "learning_rate": 5.915884287916455e-05, "loss": 0.5827, "step": 2903 }, { "epoch": 0.3, "learning_rate": 5.915826400183397e-05, "loss": 0.4375, "step": 2904 }, { "epoch": 0.3, "learning_rate": 5.915768492821679e-05, "loss": 0.6895, "step": 2905 }, { "epoch": 0.3, "learning_rate": 5.915710565831688e-05, "loss": 0.8306, "step": 2906 }, { "epoch": 0.3, "learning_rate": 5.915652619213817e-05, "loss": 0.4006, "step": 2907 }, { "epoch": 0.3, "learning_rate": 5.9155946529684545e-05, "loss": 1.4098, "step": 2908 }, { "epoch": 0.3, "learning_rate": 5.915536667095992e-05, "loss": 0.9709, "step": 2909 }, { "epoch": 0.3, "learning_rate": 5.915478661596818e-05, "loss": 0.7235, "step": 2910 }, { "epoch": 0.3, "learning_rate": 5.915420636471324e-05, "loss": 0.721, "step": 2911 }, { "epoch": 0.3, "learning_rate": 5.9153625917199e-05, "loss": 0.7076, "step": 2912 }, { "epoch": 0.3, "learning_rate": 5.9153045273429375e-05, "loss": 0.6616, "step": 2913 }, { "epoch": 0.3, "learning_rate": 5.915246443340827e-05, "loss": 0.9401, "step": 2914 }, { "epoch": 0.3, "learning_rate": 5.9151883397139595e-05, "loss": 0.5446, "step": 2915 }, { "epoch": 0.3, "learning_rate": 5.915130216462726e-05, "loss": 1.3456, "step": 2916 }, { "epoch": 0.3, "learning_rate": 5.9150720735875185e-05, "loss": 1.2127, "step": 2917 }, { "epoch": 0.3, "learning_rate": 5.915013911088727e-05, "loss": 1.4383, "step": 2918 }, { "epoch": 0.3, "learning_rate": 5.914955728966745e-05, "loss": 1.2771, "step": 2919 }, { "epoch": 0.3, "learning_rate": 5.914897527221962e-05, "loss": 1.0181, "step": 2920 }, { "epoch": 0.31, "learning_rate": 5.914839305854771e-05, "loss": 1.2496, "step": 2921 }, { "epoch": 0.31, "learning_rate": 5.9147810648655646e-05, "loss": 0.769, "step": 2922 }, { "epoch": 0.31, "learning_rate": 5.914722804254734e-05, "loss": 1.3349, "step": 2923 }, { "epoch": 0.31, "learning_rate": 5.9146645240226706e-05, "loss": 1.064, "step": 2924 }, { "epoch": 0.31, "learning_rate": 5.914606224169768e-05, "loss": 1.2358, "step": 2925 }, { "epoch": 0.31, "learning_rate": 5.914547904696418e-05, "loss": 1.42, "step": 2926 }, { "epoch": 0.31, "learning_rate": 5.9144895656030145e-05, "loss": 0.4582, "step": 2927 }, { "epoch": 0.31, "learning_rate": 5.9144312068899484e-05, "loss": 0.8668, "step": 2928 }, { "epoch": 0.31, "learning_rate": 5.914372828557613e-05, "loss": 0.5006, "step": 2929 }, { "epoch": 0.31, "learning_rate": 5.9143144306064026e-05, "loss": 0.7759, "step": 2930 }, { "epoch": 0.31, "learning_rate": 5.9142560130367094e-05, "loss": 1.2044, "step": 2931 }, { "epoch": 0.31, "learning_rate": 5.9141975758489255e-05, "loss": 0.5771, "step": 2932 }, { "epoch": 0.31, "learning_rate": 5.914139119043447e-05, "loss": 0.6923, "step": 2933 }, { "epoch": 0.31, "learning_rate": 5.914080642620664e-05, "loss": 0.9146, "step": 2934 }, { "epoch": 0.31, "learning_rate": 5.9140221465809734e-05, "loss": 1.2536, "step": 2935 }, { "epoch": 0.31, "learning_rate": 5.913963630924766e-05, "loss": 1.0025, "step": 2936 }, { "epoch": 0.31, "learning_rate": 5.9139050956524384e-05, "loss": 0.5066, "step": 2937 }, { "epoch": 0.31, "learning_rate": 5.9138465407643836e-05, "loss": 0.3384, "step": 2938 }, { "epoch": 0.31, "learning_rate": 5.913787966260995e-05, "loss": 1.5101, "step": 2939 }, { "epoch": 0.31, "learning_rate": 5.9137293721426684e-05, "loss": 0.628, "step": 2940 }, { "epoch": 0.31, "learning_rate": 5.913670758409797e-05, "loss": 0.3855, "step": 2941 }, { "epoch": 0.31, "learning_rate": 5.9136121250627754e-05, "loss": 0.9008, "step": 2942 }, { "epoch": 0.31, "learning_rate": 5.913553472101999e-05, "loss": 0.8351, "step": 2943 }, { "epoch": 0.31, "learning_rate": 5.913494799527862e-05, "loss": 1.5316, "step": 2944 }, { "epoch": 0.31, "learning_rate": 5.91343610734076e-05, "loss": 0.5994, "step": 2945 }, { "epoch": 0.31, "learning_rate": 5.913377395541088e-05, "loss": 0.3664, "step": 2946 }, { "epoch": 0.31, "learning_rate": 5.9133186641292405e-05, "loss": 0.8473, "step": 2947 }, { "epoch": 0.31, "learning_rate": 5.9132599131056144e-05, "loss": 1.2503, "step": 2948 }, { "epoch": 0.31, "learning_rate": 5.913201142470604e-05, "loss": 0.7683, "step": 2949 }, { "epoch": 0.31, "learning_rate": 5.913142352224604e-05, "loss": 0.5704, "step": 2950 }, { "epoch": 0.31, "learning_rate": 5.9130835423680116e-05, "loss": 0.653, "step": 2951 }, { "epoch": 0.31, "learning_rate": 5.913024712901223e-05, "loss": 0.2336, "step": 2952 }, { "epoch": 0.31, "learning_rate": 5.912965863824634e-05, "loss": 0.4518, "step": 2953 }, { "epoch": 0.31, "learning_rate": 5.91290699513864e-05, "loss": 0.3145, "step": 2954 }, { "epoch": 0.31, "learning_rate": 5.9128481068436375e-05, "loss": 0.4775, "step": 2955 }, { "epoch": 0.31, "learning_rate": 5.9127891989400235e-05, "loss": 1.2128, "step": 2956 }, { "epoch": 0.31, "learning_rate": 5.912730271428193e-05, "loss": 0.7766, "step": 2957 }, { "epoch": 0.31, "learning_rate": 5.912671324308545e-05, "loss": 1.5242, "step": 2958 }, { "epoch": 0.31, "learning_rate": 5.912612357581475e-05, "loss": 0.3102, "step": 2959 }, { "epoch": 0.31, "learning_rate": 5.91255337124738e-05, "loss": 1.0712, "step": 2960 }, { "epoch": 0.31, "learning_rate": 5.9124943653066574e-05, "loss": 0.7684, "step": 2961 }, { "epoch": 0.31, "learning_rate": 5.912435339759704e-05, "loss": 0.4754, "step": 2962 }, { "epoch": 0.31, "learning_rate": 5.912376294606917e-05, "loss": 0.9636, "step": 2963 }, { "epoch": 0.31, "learning_rate": 5.912317229848694e-05, "loss": 0.786, "step": 2964 }, { "epoch": 0.31, "learning_rate": 5.912258145485434e-05, "loss": 0.3696, "step": 2965 }, { "epoch": 0.31, "learning_rate": 5.912199041517533e-05, "loss": 0.4997, "step": 2966 }, { "epoch": 0.31, "learning_rate": 5.912139917945389e-05, "loss": 0.3374, "step": 2967 }, { "epoch": 0.31, "learning_rate": 5.9120807747694005e-05, "loss": 1.2106, "step": 2968 }, { "epoch": 0.31, "learning_rate": 5.912021611989966e-05, "loss": 1.0574, "step": 2969 }, { "epoch": 0.31, "learning_rate": 5.911962429607483e-05, "loss": 1.2109, "step": 2970 }, { "epoch": 0.31, "learning_rate": 5.9119032276223504e-05, "loss": 0.4241, "step": 2971 }, { "epoch": 0.31, "learning_rate": 5.911844006034967e-05, "loss": 1.3319, "step": 2972 }, { "epoch": 0.31, "learning_rate": 5.911784764845731e-05, "loss": 0.6483, "step": 2973 }, { "epoch": 0.31, "learning_rate": 5.911725504055041e-05, "loss": 1.128, "step": 2974 }, { "epoch": 0.31, "learning_rate": 5.9116662236632964e-05, "loss": 0.3242, "step": 2975 }, { "epoch": 0.31, "learning_rate": 5.911606923670896e-05, "loss": 0.1435, "step": 2976 }, { "epoch": 0.31, "learning_rate": 5.9115476040782395e-05, "loss": 1.1566, "step": 2977 }, { "epoch": 0.31, "learning_rate": 5.9114882648857246e-05, "loss": 1.1364, "step": 2978 }, { "epoch": 0.31, "learning_rate": 5.911428906093752e-05, "loss": 1.0829, "step": 2979 }, { "epoch": 0.31, "learning_rate": 5.911369527702723e-05, "loss": 1.1602, "step": 2980 }, { "epoch": 0.31, "learning_rate": 5.911310129713034e-05, "loss": 0.5225, "step": 2981 }, { "epoch": 0.31, "learning_rate": 5.911250712125086e-05, "loss": 1.3044, "step": 2982 }, { "epoch": 0.31, "learning_rate": 5.9111912749392815e-05, "loss": 0.3714, "step": 2983 }, { "epoch": 0.31, "learning_rate": 5.9111318181560164e-05, "loss": 1.216, "step": 2984 }, { "epoch": 0.31, "learning_rate": 5.911072341775695e-05, "loss": 0.9589, "step": 2985 }, { "epoch": 0.31, "learning_rate": 5.9110128457987135e-05, "loss": 0.675, "step": 2986 }, { "epoch": 0.31, "learning_rate": 5.910953330225476e-05, "loss": 1.3943, "step": 2987 }, { "epoch": 0.31, "learning_rate": 5.910893795056381e-05, "loss": 0.5483, "step": 2988 }, { "epoch": 0.31, "learning_rate": 5.9108342402918306e-05, "loss": 1.0341, "step": 2989 }, { "epoch": 0.31, "learning_rate": 5.910774665932225e-05, "loss": 1.0912, "step": 2990 }, { "epoch": 0.31, "learning_rate": 5.9107150719779656e-05, "loss": 1.2974, "step": 2991 }, { "epoch": 0.31, "learning_rate": 5.910655458429454e-05, "loss": 0.6461, "step": 2992 }, { "epoch": 0.31, "learning_rate": 5.910595825287089e-05, "loss": 0.538, "step": 2993 }, { "epoch": 0.31, "learning_rate": 5.9105361725512754e-05, "loss": 1.0217, "step": 2994 }, { "epoch": 0.31, "learning_rate": 5.9104765002224134e-05, "loss": 0.8746, "step": 2995 }, { "epoch": 0.31, "learning_rate": 5.910416808300905e-05, "loss": 0.5942, "step": 2996 }, { "epoch": 0.31, "learning_rate": 5.91035709678715e-05, "loss": 1.4787, "step": 2997 }, { "epoch": 0.31, "learning_rate": 5.910297365681552e-05, "loss": 0.1738, "step": 2998 }, { "epoch": 0.31, "learning_rate": 5.910237614984514e-05, "loss": 1.2781, "step": 2999 }, { "epoch": 0.31, "learning_rate": 5.910177844696438e-05, "loss": 1.2833, "step": 3000 }, { "epoch": 0.31, "learning_rate": 5.910118054817725e-05, "loss": 1.1111, "step": 3001 }, { "epoch": 0.31, "learning_rate": 5.9100582453487784e-05, "loss": 0.4985, "step": 3002 }, { "epoch": 0.31, "learning_rate": 5.9099984162900004e-05, "loss": 0.792, "step": 3003 }, { "epoch": 0.31, "learning_rate": 5.909938567641794e-05, "loss": 1.0741, "step": 3004 }, { "epoch": 0.31, "learning_rate": 5.909878699404562e-05, "loss": 0.6228, "step": 3005 }, { "epoch": 0.31, "learning_rate": 5.909818811578707e-05, "loss": 0.2893, "step": 3006 }, { "epoch": 0.31, "learning_rate": 5.9097589041646324e-05, "loss": 1.1549, "step": 3007 }, { "epoch": 0.31, "learning_rate": 5.909698977162743e-05, "loss": 0.6111, "step": 3008 }, { "epoch": 0.31, "learning_rate": 5.909639030573439e-05, "loss": 1.4957, "step": 3009 }, { "epoch": 0.31, "learning_rate": 5.909579064397128e-05, "loss": 0.948, "step": 3010 }, { "epoch": 0.31, "learning_rate": 5.90951907863421e-05, "loss": 1.1236, "step": 3011 }, { "epoch": 0.31, "learning_rate": 5.909459073285091e-05, "loss": 0.2939, "step": 3012 }, { "epoch": 0.31, "learning_rate": 5.909399048350174e-05, "loss": 0.7846, "step": 3013 }, { "epoch": 0.31, "learning_rate": 5.9093390038298634e-05, "loss": 0.6199, "step": 3014 }, { "epoch": 0.31, "learning_rate": 5.909278939724563e-05, "loss": 1.7211, "step": 3015 }, { "epoch": 0.31, "learning_rate": 5.909218856034678e-05, "loss": 0.4846, "step": 3016 }, { "epoch": 0.32, "learning_rate": 5.909158752760613e-05, "loss": 1.2358, "step": 3017 }, { "epoch": 0.32, "learning_rate": 5.909098629902771e-05, "loss": 0.4176, "step": 3018 }, { "epoch": 0.32, "learning_rate": 5.909038487461558e-05, "loss": 1.3902, "step": 3019 }, { "epoch": 0.32, "learning_rate": 5.908978325437379e-05, "loss": 0.2878, "step": 3020 }, { "epoch": 0.32, "learning_rate": 5.908918143830638e-05, "loss": 1.4452, "step": 3021 }, { "epoch": 0.32, "learning_rate": 5.9088579426417404e-05, "loss": 0.4539, "step": 3022 }, { "epoch": 0.32, "learning_rate": 5.908797721871092e-05, "loss": 0.4502, "step": 3023 }, { "epoch": 0.32, "learning_rate": 5.908737481519098e-05, "loss": 0.609, "step": 3024 }, { "epoch": 0.32, "learning_rate": 5.908677221586164e-05, "loss": 1.2031, "step": 3025 }, { "epoch": 0.32, "learning_rate": 5.9086169420726956e-05, "loss": 1.2693, "step": 3026 }, { "epoch": 0.32, "learning_rate": 5.908556642979098e-05, "loss": 0.8862, "step": 3027 }, { "epoch": 0.32, "learning_rate": 5.908496324305778e-05, "loss": 1.2461, "step": 3028 }, { "epoch": 0.32, "learning_rate": 5.9084359860531414e-05, "loss": 0.126, "step": 3029 }, { "epoch": 0.32, "learning_rate": 5.908375628221594e-05, "loss": 1.8978, "step": 3030 }, { "epoch": 0.32, "learning_rate": 5.9083152508115427e-05, "loss": 0.4646, "step": 3031 }, { "epoch": 0.32, "learning_rate": 5.908254853823394e-05, "loss": 1.7056, "step": 3032 }, { "epoch": 0.32, "learning_rate": 5.908194437257553e-05, "loss": 0.4172, "step": 3033 }, { "epoch": 0.32, "learning_rate": 5.908134001114428e-05, "loss": 0.7873, "step": 3034 }, { "epoch": 0.32, "learning_rate": 5.908073545394425e-05, "loss": 1.1387, "step": 3035 }, { "epoch": 0.32, "learning_rate": 5.9080130700979516e-05, "loss": 0.9538, "step": 3036 }, { "epoch": 0.32, "learning_rate": 5.9079525752254146e-05, "loss": 0.79, "step": 3037 }, { "epoch": 0.32, "learning_rate": 5.907892060777222e-05, "loss": 1.5696, "step": 3038 }, { "epoch": 0.32, "learning_rate": 5.907831526753779e-05, "loss": 0.7002, "step": 3039 }, { "epoch": 0.32, "learning_rate": 5.9077709731554957e-05, "loss": 1.2868, "step": 3040 }, { "epoch": 0.32, "learning_rate": 5.907710399982779e-05, "loss": 0.5385, "step": 3041 }, { "epoch": 0.32, "learning_rate": 5.9076498072360355e-05, "loss": 1.5694, "step": 3042 }, { "epoch": 0.32, "learning_rate": 5.907589194915674e-05, "loss": 0.9141, "step": 3043 }, { "epoch": 0.32, "learning_rate": 5.907528563022101e-05, "loss": 1.0863, "step": 3044 }, { "epoch": 0.32, "learning_rate": 5.907467911555728e-05, "loss": 0.3399, "step": 3045 }, { "epoch": 0.32, "learning_rate": 5.9074072405169605e-05, "loss": 1.0623, "step": 3046 }, { "epoch": 0.32, "learning_rate": 5.9073465499062075e-05, "loss": 1.2232, "step": 3047 }, { "epoch": 0.32, "learning_rate": 5.907285839723878e-05, "loss": 1.0678, "step": 3048 }, { "epoch": 0.32, "learning_rate": 5.907225109970381e-05, "loss": 0.525, "step": 3049 }, { "epoch": 0.32, "learning_rate": 5.9071643606461234e-05, "loss": 0.7459, "step": 3050 }, { "epoch": 0.32, "learning_rate": 5.907103591751517e-05, "loss": 0.3722, "step": 3051 }, { "epoch": 0.32, "learning_rate": 5.907042803286969e-05, "loss": 0.4873, "step": 3052 }, { "epoch": 0.32, "learning_rate": 5.906981995252889e-05, "loss": 0.3504, "step": 3053 }, { "epoch": 0.32, "learning_rate": 5.906921167649686e-05, "loss": 0.8613, "step": 3054 }, { "epoch": 0.32, "learning_rate": 5.90686032047777e-05, "loss": 1.2359, "step": 3055 }, { "epoch": 0.32, "learning_rate": 5.906799453737551e-05, "loss": 0.3414, "step": 3056 }, { "epoch": 0.32, "learning_rate": 5.9067385674294376e-05, "loss": 1.0665, "step": 3057 }, { "epoch": 0.32, "learning_rate": 5.906677661553841e-05, "loss": 0.3992, "step": 3058 }, { "epoch": 0.32, "learning_rate": 5.90661673611117e-05, "loss": 0.9134, "step": 3059 }, { "epoch": 0.32, "learning_rate": 5.9065557911018346e-05, "loss": 1.1944, "step": 3060 }, { "epoch": 0.32, "learning_rate": 5.906494826526247e-05, "loss": 0.7734, "step": 3061 }, { "epoch": 0.32, "learning_rate": 5.906433842384816e-05, "loss": 1.2075, "step": 3062 }, { "epoch": 0.32, "learning_rate": 5.9063728386779515e-05, "loss": 0.3147, "step": 3063 }, { "epoch": 0.32, "learning_rate": 5.906311815406066e-05, "loss": 0.4216, "step": 3064 }, { "epoch": 0.32, "learning_rate": 5.9062507725695686e-05, "loss": 0.4878, "step": 3065 }, { "epoch": 0.32, "learning_rate": 5.906189710168872e-05, "loss": 1.2849, "step": 3066 }, { "epoch": 0.32, "learning_rate": 5.906128628204386e-05, "loss": 1.1784, "step": 3067 }, { "epoch": 0.32, "learning_rate": 5.906067526676521e-05, "loss": 0.7731, "step": 3068 }, { "epoch": 0.32, "learning_rate": 5.9060064055856905e-05, "loss": 0.5661, "step": 3069 }, { "epoch": 0.32, "learning_rate": 5.9059452649323045e-05, "loss": 1.4721, "step": 3070 }, { "epoch": 0.32, "learning_rate": 5.905884104716775e-05, "loss": 0.4964, "step": 3071 }, { "epoch": 0.32, "learning_rate": 5.905822924939513e-05, "loss": 0.7425, "step": 3072 }, { "epoch": 0.32, "learning_rate": 5.905761725600933e-05, "loss": 0.3463, "step": 3073 }, { "epoch": 0.32, "learning_rate": 5.905700506701443e-05, "loss": 0.4642, "step": 3074 }, { "epoch": 0.32, "learning_rate": 5.905639268241457e-05, "loss": 0.5104, "step": 3075 }, { "epoch": 0.32, "learning_rate": 5.905578010221388e-05, "loss": 0.5875, "step": 3076 }, { "epoch": 0.32, "learning_rate": 5.905516732641647e-05, "loss": 0.8804, "step": 3077 }, { "epoch": 0.32, "learning_rate": 5.9054554355026483e-05, "loss": 1.1665, "step": 3078 }, { "epoch": 0.32, "learning_rate": 5.905394118804802e-05, "loss": 0.6617, "step": 3079 }, { "epoch": 0.32, "learning_rate": 5.9053327825485235e-05, "loss": 0.7391, "step": 3080 }, { "epoch": 0.32, "learning_rate": 5.905271426734224e-05, "loss": 0.9001, "step": 3081 }, { "epoch": 0.32, "learning_rate": 5.905210051362318e-05, "loss": 1.4088, "step": 3082 }, { "epoch": 0.32, "learning_rate": 5.905148656433217e-05, "loss": 0.3745, "step": 3083 }, { "epoch": 0.32, "learning_rate": 5.9050872419473346e-05, "loss": 1.3592, "step": 3084 }, { "epoch": 0.32, "learning_rate": 5.905025807905085e-05, "loss": 1.0724, "step": 3085 }, { "epoch": 0.32, "learning_rate": 5.9049643543068813e-05, "loss": 0.55, "step": 3086 }, { "epoch": 0.32, "learning_rate": 5.9049028811531364e-05, "loss": 0.7246, "step": 3087 }, { "epoch": 0.32, "learning_rate": 5.904841388444266e-05, "loss": 0.8326, "step": 3088 }, { "epoch": 0.32, "learning_rate": 5.904779876180684e-05, "loss": 1.2657, "step": 3089 }, { "epoch": 0.32, "learning_rate": 5.904718344362802e-05, "loss": 0.6925, "step": 3090 }, { "epoch": 0.32, "learning_rate": 5.904656792991036e-05, "loss": 0.7523, "step": 3091 }, { "epoch": 0.32, "learning_rate": 5.9045952220658004e-05, "loss": 0.3374, "step": 3092 }, { "epoch": 0.32, "learning_rate": 5.904533631587509e-05, "loss": 0.1718, "step": 3093 }, { "epoch": 0.32, "learning_rate": 5.9044720215565774e-05, "loss": 1.4006, "step": 3094 }, { "epoch": 0.32, "learning_rate": 5.904410391973419e-05, "loss": 0.7046, "step": 3095 }, { "epoch": 0.32, "learning_rate": 5.90434874283845e-05, "loss": 0.3002, "step": 3096 }, { "epoch": 0.32, "learning_rate": 5.9042870741520846e-05, "loss": 0.9917, "step": 3097 }, { "epoch": 0.32, "learning_rate": 5.904225385914738e-05, "loss": 1.5052, "step": 3098 }, { "epoch": 0.32, "learning_rate": 5.904163678126827e-05, "loss": 1.0369, "step": 3099 }, { "epoch": 0.32, "learning_rate": 5.904101950788764e-05, "loss": 0.2433, "step": 3100 }, { "epoch": 0.32, "learning_rate": 5.904040203900967e-05, "loss": 1.2046, "step": 3101 }, { "epoch": 0.32, "learning_rate": 5.90397843746385e-05, "loss": 1.4253, "step": 3102 }, { "epoch": 0.32, "learning_rate": 5.9039166514778306e-05, "loss": 1.2282, "step": 3103 }, { "epoch": 0.32, "learning_rate": 5.903854845943323e-05, "loss": 1.0217, "step": 3104 }, { "epoch": 0.32, "learning_rate": 5.9037930208607445e-05, "loss": 1.5121, "step": 3105 }, { "epoch": 0.32, "learning_rate": 5.903731176230511e-05, "loss": 1.816, "step": 3106 }, { "epoch": 0.32, "learning_rate": 5.903669312053039e-05, "loss": 0.5746, "step": 3107 }, { "epoch": 0.32, "learning_rate": 5.903607428328743e-05, "loss": 0.998, "step": 3108 }, { "epoch": 0.32, "learning_rate": 5.9035455250580424e-05, "loss": 1.4222, "step": 3109 }, { "epoch": 0.32, "learning_rate": 5.9034836022413526e-05, "loss": 0.3649, "step": 3110 }, { "epoch": 0.32, "learning_rate": 5.9034216598790896e-05, "loss": 0.8924, "step": 3111 }, { "epoch": 0.33, "learning_rate": 5.903359697971672e-05, "loss": 1.6622, "step": 3112 }, { "epoch": 0.33, "learning_rate": 5.903297716519517e-05, "loss": 1.0832, "step": 3113 }, { "epoch": 0.33, "learning_rate": 5.9032357155230405e-05, "loss": 1.2719, "step": 3114 }, { "epoch": 0.33, "learning_rate": 5.90317369498266e-05, "loss": 0.4842, "step": 3115 }, { "epoch": 0.33, "learning_rate": 5.9031116548987933e-05, "loss": 0.7151, "step": 3116 }, { "epoch": 0.33, "learning_rate": 5.903049595271859e-05, "loss": 0.3802, "step": 3117 }, { "epoch": 0.33, "learning_rate": 5.902987516102274e-05, "loss": 0.8907, "step": 3118 }, { "epoch": 0.33, "learning_rate": 5.902925417390456e-05, "loss": 0.5113, "step": 3119 }, { "epoch": 0.33, "learning_rate": 5.9028632991368236e-05, "loss": 0.6802, "step": 3120 }, { "epoch": 0.33, "learning_rate": 5.902801161341795e-05, "loss": 0.1609, "step": 3121 }, { "epoch": 0.33, "learning_rate": 5.902739004005787e-05, "loss": 1.4009, "step": 3122 }, { "epoch": 0.33, "learning_rate": 5.90267682712922e-05, "loss": 0.9526, "step": 3123 }, { "epoch": 0.33, "learning_rate": 5.902614630712513e-05, "loss": 1.0785, "step": 3124 }, { "epoch": 0.33, "learning_rate": 5.9025524147560815e-05, "loss": 1.2418, "step": 3125 }, { "epoch": 0.33, "learning_rate": 5.9024901792603465e-05, "loss": 0.9317, "step": 3126 }, { "epoch": 0.33, "learning_rate": 5.902427924225728e-05, "loss": 1.272, "step": 3127 }, { "epoch": 0.33, "learning_rate": 5.902365649652643e-05, "loss": 1.3795, "step": 3128 }, { "epoch": 0.33, "learning_rate": 5.902303355541511e-05, "loss": 0.4157, "step": 3129 }, { "epoch": 0.33, "learning_rate": 5.902241041892753e-05, "loss": 0.3311, "step": 3130 }, { "epoch": 0.33, "learning_rate": 5.902178708706787e-05, "loss": 0.3166, "step": 3131 }, { "epoch": 0.33, "learning_rate": 5.902116355984032e-05, "loss": 0.4908, "step": 3132 }, { "epoch": 0.33, "learning_rate": 5.90205398372491e-05, "loss": 0.5479, "step": 3133 }, { "epoch": 0.33, "learning_rate": 5.901991591929839e-05, "loss": 0.4503, "step": 3134 }, { "epoch": 0.33, "learning_rate": 5.90192918059924e-05, "loss": 1.4929, "step": 3135 }, { "epoch": 0.33, "learning_rate": 5.901866749733533e-05, "loss": 1.3756, "step": 3136 }, { "epoch": 0.33, "learning_rate": 5.9018042993331365e-05, "loss": 1.1196, "step": 3137 }, { "epoch": 0.33, "learning_rate": 5.901741829398474e-05, "loss": 1.0663, "step": 3138 }, { "epoch": 0.33, "learning_rate": 5.9016793399299634e-05, "loss": 0.6229, "step": 3139 }, { "epoch": 0.33, "learning_rate": 5.9016168309280276e-05, "loss": 0.8544, "step": 3140 }, { "epoch": 0.33, "learning_rate": 5.901554302393086e-05, "loss": 0.6946, "step": 3141 }, { "epoch": 0.33, "learning_rate": 5.901491754325559e-05, "loss": 1.0296, "step": 3142 }, { "epoch": 0.33, "learning_rate": 5.901429186725869e-05, "loss": 1.1251, "step": 3143 }, { "epoch": 0.33, "learning_rate": 5.9013665995944366e-05, "loss": 0.4885, "step": 3144 }, { "epoch": 0.33, "learning_rate": 5.9013039929316823e-05, "loss": 0.4332, "step": 3145 }, { "epoch": 0.33, "learning_rate": 5.901241366738029e-05, "loss": 0.8887, "step": 3146 }, { "epoch": 0.33, "learning_rate": 5.901178721013898e-05, "loss": 1.1765, "step": 3147 }, { "epoch": 0.33, "learning_rate": 5.901116055759709e-05, "loss": 0.7011, "step": 3148 }, { "epoch": 0.33, "learning_rate": 5.901053370975887e-05, "loss": 0.9077, "step": 3149 }, { "epoch": 0.33, "learning_rate": 5.900990666662853e-05, "loss": 1.7989, "step": 3150 }, { "epoch": 0.33, "learning_rate": 5.9009279428210276e-05, "loss": 0.9017, "step": 3151 }, { "epoch": 0.33, "learning_rate": 5.9008651994508345e-05, "loss": 0.5645, "step": 3152 }, { "epoch": 0.33, "learning_rate": 5.900802436552695e-05, "loss": 0.8016, "step": 3153 }, { "epoch": 0.33, "learning_rate": 5.900739654127033e-05, "loss": 1.0648, "step": 3154 }, { "epoch": 0.33, "learning_rate": 5.90067685217427e-05, "loss": 0.7071, "step": 3155 }, { "epoch": 0.33, "learning_rate": 5.900614030694829e-05, "loss": 1.2785, "step": 3156 }, { "epoch": 0.33, "learning_rate": 5.900551189689134e-05, "loss": 1.4186, "step": 3157 }, { "epoch": 0.33, "learning_rate": 5.900488329157606e-05, "loss": 0.197, "step": 3158 }, { "epoch": 0.33, "learning_rate": 5.900425449100669e-05, "loss": 1.0949, "step": 3159 }, { "epoch": 0.33, "learning_rate": 5.900362549518747e-05, "loss": 0.3654, "step": 3160 }, { "epoch": 0.33, "learning_rate": 5.900299630412263e-05, "loss": 0.2915, "step": 3161 }, { "epoch": 0.33, "learning_rate": 5.900236691781641e-05, "loss": 1.3981, "step": 3162 }, { "epoch": 0.33, "learning_rate": 5.900173733627304e-05, "loss": 1.0237, "step": 3163 }, { "epoch": 0.33, "learning_rate": 5.900110755949676e-05, "loss": 1.3183, "step": 3164 }, { "epoch": 0.33, "learning_rate": 5.90004775874918e-05, "loss": 0.5599, "step": 3165 }, { "epoch": 0.33, "learning_rate": 5.8999847420262416e-05, "loss": 0.291, "step": 3166 }, { "epoch": 0.33, "learning_rate": 5.899921705781284e-05, "loss": 0.9378, "step": 3167 }, { "epoch": 0.33, "learning_rate": 5.899858650014733e-05, "loss": 0.4804, "step": 3168 }, { "epoch": 0.33, "learning_rate": 5.8997955747270115e-05, "loss": 0.8358, "step": 3169 }, { "epoch": 0.33, "learning_rate": 5.899732479918545e-05, "loss": 0.8183, "step": 3170 }, { "epoch": 0.33, "learning_rate": 5.899669365589758e-05, "loss": 1.3802, "step": 3171 }, { "epoch": 0.33, "learning_rate": 5.899606231741075e-05, "loss": 0.2908, "step": 3172 }, { "epoch": 0.33, "learning_rate": 5.899543078372921e-05, "loss": 0.5625, "step": 3173 }, { "epoch": 0.33, "learning_rate": 5.899479905485722e-05, "loss": 1.4148, "step": 3174 }, { "epoch": 0.33, "learning_rate": 5.899416713079903e-05, "loss": 0.8789, "step": 3175 }, { "epoch": 0.33, "learning_rate": 5.899353501155888e-05, "loss": 1.1353, "step": 3176 }, { "epoch": 0.33, "learning_rate": 5.899290269714104e-05, "loss": 0.1499, "step": 3177 }, { "epoch": 0.33, "learning_rate": 5.899227018754976e-05, "loss": 0.7084, "step": 3178 }, { "epoch": 0.33, "learning_rate": 5.899163748278931e-05, "loss": 0.6197, "step": 3179 }, { "epoch": 0.33, "learning_rate": 5.899100458286394e-05, "loss": 0.6247, "step": 3180 }, { "epoch": 0.33, "learning_rate": 5.8990371487777906e-05, "loss": 1.2729, "step": 3181 }, { "epoch": 0.33, "learning_rate": 5.8989738197535476e-05, "loss": 0.3425, "step": 3182 }, { "epoch": 0.33, "learning_rate": 5.898910471214091e-05, "loss": 0.2602, "step": 3183 }, { "epoch": 0.33, "learning_rate": 5.8988471031598476e-05, "loss": 0.955, "step": 3184 }, { "epoch": 0.33, "learning_rate": 5.898783715591243e-05, "loss": 1.4905, "step": 3185 }, { "epoch": 0.33, "learning_rate": 5.898720308508706e-05, "loss": 0.3694, "step": 3186 }, { "epoch": 0.33, "learning_rate": 5.8986568819126614e-05, "loss": 1.4703, "step": 3187 }, { "epoch": 0.33, "learning_rate": 5.898593435803537e-05, "loss": 1.4115, "step": 3188 }, { "epoch": 0.33, "learning_rate": 5.8985299701817595e-05, "loss": 1.2417, "step": 3189 }, { "epoch": 0.33, "learning_rate": 5.898466485047757e-05, "loss": 0.2826, "step": 3190 }, { "epoch": 0.33, "learning_rate": 5.898402980401955e-05, "loss": 0.6273, "step": 3191 }, { "epoch": 0.33, "learning_rate": 5.898339456244784e-05, "loss": 0.5095, "step": 3192 }, { "epoch": 0.33, "learning_rate": 5.8982759125766695e-05, "loss": 1.2442, "step": 3193 }, { "epoch": 0.33, "learning_rate": 5.8982123493980394e-05, "loss": 0.8035, "step": 3194 }, { "epoch": 0.33, "learning_rate": 5.8981487667093214e-05, "loss": 1.0367, "step": 3195 }, { "epoch": 0.33, "learning_rate": 5.898085164510945e-05, "loss": 1.2756, "step": 3196 }, { "epoch": 0.33, "learning_rate": 5.8980215428033364e-05, "loss": 1.1299, "step": 3197 }, { "epoch": 0.33, "learning_rate": 5.897957901586925e-05, "loss": 0.6437, "step": 3198 }, { "epoch": 0.33, "learning_rate": 5.89789424086214e-05, "loss": 0.3768, "step": 3199 }, { "epoch": 0.33, "learning_rate": 5.8978305606294086e-05, "loss": 0.4493, "step": 3200 }, { "epoch": 0.33, "learning_rate": 5.89776686088916e-05, "loss": 0.6407, "step": 3201 }, { "epoch": 0.33, "learning_rate": 5.897703141641823e-05, "loss": 0.9958, "step": 3202 }, { "epoch": 0.33, "learning_rate": 5.897639402887825e-05, "loss": 0.8523, "step": 3203 }, { "epoch": 0.33, "learning_rate": 5.8975756446275976e-05, "loss": 0.5904, "step": 3204 }, { "epoch": 0.33, "learning_rate": 5.8975118668615696e-05, "loss": 0.7624, "step": 3205 }, { "epoch": 0.33, "learning_rate": 5.897448069590169e-05, "loss": 1.2484, "step": 3206 }, { "epoch": 0.33, "learning_rate": 5.897384252813826e-05, "loss": 0.6678, "step": 3207 }, { "epoch": 0.34, "learning_rate": 5.89732041653297e-05, "loss": 0.5833, "step": 3208 }, { "epoch": 0.34, "learning_rate": 5.8972565607480316e-05, "loss": 1.246, "step": 3209 }, { "epoch": 0.34, "learning_rate": 5.89719268545944e-05, "loss": 1.1013, "step": 3210 }, { "epoch": 0.34, "learning_rate": 5.897128790667624e-05, "loss": 0.4496, "step": 3211 }, { "epoch": 0.34, "learning_rate": 5.897064876373015e-05, "loss": 0.8163, "step": 3212 }, { "epoch": 0.34, "learning_rate": 5.897000942576043e-05, "loss": 0.6311, "step": 3213 }, { "epoch": 0.34, "learning_rate": 5.896936989277139e-05, "loss": 0.5009, "step": 3214 }, { "epoch": 0.34, "learning_rate": 5.8968730164767335e-05, "loss": 1.0158, "step": 3215 }, { "epoch": 0.34, "learning_rate": 5.8968090241752556e-05, "loss": 0.5685, "step": 3216 }, { "epoch": 0.34, "learning_rate": 5.8967450123731385e-05, "loss": 1.1398, "step": 3217 }, { "epoch": 0.34, "learning_rate": 5.8966809810708106e-05, "loss": 0.3486, "step": 3218 }, { "epoch": 0.34, "learning_rate": 5.896616930268705e-05, "loss": 0.8588, "step": 3219 }, { "epoch": 0.34, "learning_rate": 5.8965528599672506e-05, "loss": 0.4733, "step": 3220 }, { "epoch": 0.34, "learning_rate": 5.896488770166881e-05, "loss": 1.2738, "step": 3221 }, { "epoch": 0.34, "learning_rate": 5.8964246608680274e-05, "loss": 1.1556, "step": 3222 }, { "epoch": 0.34, "learning_rate": 5.8963605320711197e-05, "loss": 0.9426, "step": 3223 }, { "epoch": 0.34, "learning_rate": 5.896296383776591e-05, "loss": 0.6964, "step": 3224 }, { "epoch": 0.34, "learning_rate": 5.896232215984873e-05, "loss": 0.106, "step": 3225 }, { "epoch": 0.34, "learning_rate": 5.896168028696396e-05, "loss": 1.4974, "step": 3226 }, { "epoch": 0.34, "learning_rate": 5.896103821911595e-05, "loss": 1.5217, "step": 3227 }, { "epoch": 0.34, "learning_rate": 5.8960395956309e-05, "loss": 0.9322, "step": 3228 }, { "epoch": 0.34, "learning_rate": 5.895975349854744e-05, "loss": 1.2315, "step": 3229 }, { "epoch": 0.34, "learning_rate": 5.895911084583559e-05, "loss": 1.4521, "step": 3230 }, { "epoch": 0.34, "learning_rate": 5.8958467998177795e-05, "loss": 0.8835, "step": 3231 }, { "epoch": 0.34, "learning_rate": 5.895782495557837e-05, "loss": 0.9027, "step": 3232 }, { "epoch": 0.34, "learning_rate": 5.8957181718041634e-05, "loss": 0.9058, "step": 3233 }, { "epoch": 0.34, "learning_rate": 5.895653828557193e-05, "loss": 1.5095, "step": 3234 }, { "epoch": 0.34, "learning_rate": 5.895589465817358e-05, "loss": 1.2235, "step": 3235 }, { "epoch": 0.34, "learning_rate": 5.8955250835850924e-05, "loss": 0.6201, "step": 3236 }, { "epoch": 0.34, "learning_rate": 5.89546068186083e-05, "loss": 0.5779, "step": 3237 }, { "epoch": 0.34, "learning_rate": 5.895396260645003e-05, "loss": 0.49, "step": 3238 }, { "epoch": 0.34, "learning_rate": 5.895331819938047e-05, "loss": 0.7917, "step": 3239 }, { "epoch": 0.34, "learning_rate": 5.895267359740393e-05, "loss": 0.9655, "step": 3240 }, { "epoch": 0.34, "learning_rate": 5.8952028800524784e-05, "loss": 0.6282, "step": 3241 }, { "epoch": 0.34, "learning_rate": 5.895138380874734e-05, "loss": 1.1606, "step": 3242 }, { "epoch": 0.34, "learning_rate": 5.895073862207597e-05, "loss": 1.0759, "step": 3243 }, { "epoch": 0.34, "learning_rate": 5.895009324051499e-05, "loss": 0.5267, "step": 3244 }, { "epoch": 0.34, "learning_rate": 5.894944766406875e-05, "loss": 1.8436, "step": 3245 }, { "epoch": 0.34, "learning_rate": 5.8948801892741606e-05, "loss": 0.5987, "step": 3246 }, { "epoch": 0.34, "learning_rate": 5.894815592653791e-05, "loss": 1.3062, "step": 3247 }, { "epoch": 0.34, "learning_rate": 5.8947509765461996e-05, "loss": 2.0358, "step": 3248 }, { "epoch": 0.34, "learning_rate": 5.894686340951822e-05, "loss": 0.6827, "step": 3249 }, { "epoch": 0.34, "learning_rate": 5.8946216858710935e-05, "loss": 0.8306, "step": 3250 }, { "epoch": 0.34, "learning_rate": 5.894557011304449e-05, "loss": 0.9665, "step": 3251 }, { "epoch": 0.34, "learning_rate": 5.894492317252323e-05, "loss": 0.9745, "step": 3252 }, { "epoch": 0.34, "learning_rate": 5.894427603715153e-05, "loss": 1.261, "step": 3253 }, { "epoch": 0.34, "learning_rate": 5.894362870693373e-05, "loss": 0.4268, "step": 3254 }, { "epoch": 0.34, "learning_rate": 5.89429811818742e-05, "loss": 1.4456, "step": 3255 }, { "epoch": 0.34, "learning_rate": 5.894233346197729e-05, "loss": 0.948, "step": 3256 }, { "epoch": 0.34, "learning_rate": 5.8941685547247364e-05, "loss": 0.2878, "step": 3257 }, { "epoch": 0.34, "learning_rate": 5.894103743768878e-05, "loss": 1.4433, "step": 3258 }, { "epoch": 0.34, "learning_rate": 5.894038913330591e-05, "loss": 0.8034, "step": 3259 }, { "epoch": 0.34, "learning_rate": 5.8939740634103105e-05, "loss": 0.537, "step": 3260 }, { "epoch": 0.34, "learning_rate": 5.8939091940084736e-05, "loss": 1.9609, "step": 3261 }, { "epoch": 0.34, "learning_rate": 5.893844305125517e-05, "loss": 0.8726, "step": 3262 }, { "epoch": 0.34, "learning_rate": 5.893779396761879e-05, "loss": 0.8658, "step": 3263 }, { "epoch": 0.34, "learning_rate": 5.8937144689179944e-05, "loss": 1.4476, "step": 3264 }, { "epoch": 0.34, "learning_rate": 5.8936495215943014e-05, "loss": 0.8107, "step": 3265 }, { "epoch": 0.34, "learning_rate": 5.893584554791236e-05, "loss": 0.7094, "step": 3266 }, { "epoch": 0.34, "learning_rate": 5.893519568509237e-05, "loss": 0.4042, "step": 3267 }, { "epoch": 0.34, "learning_rate": 5.893454562748741e-05, "loss": 1.4032, "step": 3268 }, { "epoch": 0.34, "learning_rate": 5.893389537510187e-05, "loss": 0.802, "step": 3269 }, { "epoch": 0.34, "learning_rate": 5.89332449279401e-05, "loss": 0.3713, "step": 3270 }, { "epoch": 0.34, "learning_rate": 5.89325942860065e-05, "loss": 0.995, "step": 3271 }, { "epoch": 0.34, "learning_rate": 5.8931943449305444e-05, "loss": 0.9281, "step": 3272 }, { "epoch": 0.34, "learning_rate": 5.893129241784132e-05, "loss": 0.3501, "step": 3273 }, { "epoch": 0.34, "learning_rate": 5.8930641191618504e-05, "loss": 1.264, "step": 3274 }, { "epoch": 0.34, "learning_rate": 5.892998977064139e-05, "loss": 0.1579, "step": 3275 }, { "epoch": 0.34, "learning_rate": 5.892933815491433e-05, "loss": 0.5185, "step": 3276 }, { "epoch": 0.34, "learning_rate": 5.892868634444175e-05, "loss": 0.9144, "step": 3277 }, { "epoch": 0.34, "learning_rate": 5.8928034339228024e-05, "loss": 0.4468, "step": 3278 }, { "epoch": 0.34, "learning_rate": 5.892738213927754e-05, "loss": 0.5566, "step": 3279 }, { "epoch": 0.34, "learning_rate": 5.892672974459467e-05, "loss": 0.6829, "step": 3280 }, { "epoch": 0.34, "learning_rate": 5.8926077155183844e-05, "loss": 1.2994, "step": 3281 }, { "epoch": 0.34, "learning_rate": 5.8925424371049426e-05, "loss": 1.154, "step": 3282 }, { "epoch": 0.34, "learning_rate": 5.892477139219583e-05, "loss": 0.9047, "step": 3283 }, { "epoch": 0.34, "learning_rate": 5.892411821862742e-05, "loss": 0.673, "step": 3284 }, { "epoch": 0.34, "learning_rate": 5.8923464850348635e-05, "loss": 0.2722, "step": 3285 }, { "epoch": 0.34, "learning_rate": 5.8922811287363836e-05, "loss": 0.2253, "step": 3286 }, { "epoch": 0.34, "learning_rate": 5.892215752967745e-05, "loss": 0.5722, "step": 3287 }, { "epoch": 0.34, "learning_rate": 5.892150357729386e-05, "loss": 0.4133, "step": 3288 }, { "epoch": 0.34, "learning_rate": 5.8920849430217474e-05, "loss": 0.7409, "step": 3289 }, { "epoch": 0.34, "learning_rate": 5.89201950884527e-05, "loss": 0.7015, "step": 3290 }, { "epoch": 0.34, "learning_rate": 5.891954055200394e-05, "loss": 0.4904, "step": 3291 }, { "epoch": 0.34, "learning_rate": 5.8918885820875596e-05, "loss": 0.8352, "step": 3292 }, { "epoch": 0.34, "learning_rate": 5.891823089507209e-05, "loss": 0.8884, "step": 3293 }, { "epoch": 0.34, "learning_rate": 5.8917575774597805e-05, "loss": 1.3569, "step": 3294 }, { "epoch": 0.34, "learning_rate": 5.891692045945717e-05, "loss": 0.4125, "step": 3295 }, { "epoch": 0.34, "learning_rate": 5.8916264949654606e-05, "loss": 0.9294, "step": 3296 }, { "epoch": 0.34, "learning_rate": 5.8915609245194503e-05, "loss": 0.5244, "step": 3297 }, { "epoch": 0.34, "learning_rate": 5.8914953346081276e-05, "loss": 1.1003, "step": 3298 }, { "epoch": 0.34, "learning_rate": 5.891429725231936e-05, "loss": 0.9937, "step": 3299 }, { "epoch": 0.34, "learning_rate": 5.8913640963913154e-05, "loss": 0.4072, "step": 3300 }, { "epoch": 0.34, "learning_rate": 5.891298448086708e-05, "loss": 1.0991, "step": 3301 }, { "epoch": 0.34, "learning_rate": 5.891232780318557e-05, "loss": 0.5925, "step": 3302 }, { "epoch": 0.34, "learning_rate": 5.8911670930873026e-05, "loss": 1.1767, "step": 3303 }, { "epoch": 0.35, "learning_rate": 5.891101386393388e-05, "loss": 0.6278, "step": 3304 }, { "epoch": 0.35, "learning_rate": 5.891035660237255e-05, "loss": 1.2188, "step": 3305 }, { "epoch": 0.35, "learning_rate": 5.890969914619347e-05, "loss": 0.538, "step": 3306 }, { "epoch": 0.35, "learning_rate": 5.8909041495401056e-05, "loss": 0.3016, "step": 3307 }, { "epoch": 0.35, "learning_rate": 5.890838364999974e-05, "loss": 0.2985, "step": 3308 }, { "epoch": 0.35, "learning_rate": 5.8907725609993955e-05, "loss": 0.7129, "step": 3309 }, { "epoch": 0.35, "learning_rate": 5.8907067375388116e-05, "loss": 1.1264, "step": 3310 }, { "epoch": 0.35, "learning_rate": 5.890640894618666e-05, "loss": 1.7222, "step": 3311 }, { "epoch": 0.35, "learning_rate": 5.890575032239404e-05, "loss": 1.3263, "step": 3312 }, { "epoch": 0.35, "learning_rate": 5.890509150401465e-05, "loss": 0.4389, "step": 3313 }, { "epoch": 0.35, "learning_rate": 5.890443249105296e-05, "loss": 1.7677, "step": 3314 }, { "epoch": 0.35, "learning_rate": 5.890377328351339e-05, "loss": 0.3828, "step": 3315 }, { "epoch": 0.35, "learning_rate": 5.8903113881400385e-05, "loss": 1.0684, "step": 3316 }, { "epoch": 0.35, "learning_rate": 5.890245428471838e-05, "loss": 0.9156, "step": 3317 }, { "epoch": 0.35, "learning_rate": 5.8901794493471816e-05, "loss": 0.1505, "step": 3318 }, { "epoch": 0.35, "learning_rate": 5.8901134507665135e-05, "loss": 1.334, "step": 3319 }, { "epoch": 0.35, "learning_rate": 5.890047432730277e-05, "loss": 2.2085, "step": 3320 }, { "epoch": 0.35, "learning_rate": 5.889981395238918e-05, "loss": 0.4468, "step": 3321 }, { "epoch": 0.35, "learning_rate": 5.889915338292881e-05, "loss": 1.0567, "step": 3322 }, { "epoch": 0.35, "learning_rate": 5.889849261892609e-05, "loss": 0.5261, "step": 3323 }, { "epoch": 0.35, "learning_rate": 5.8897831660385485e-05, "loss": 1.1609, "step": 3324 }, { "epoch": 0.35, "learning_rate": 5.889717050731144e-05, "loss": 0.7204, "step": 3325 }, { "epoch": 0.35, "learning_rate": 5.88965091597084e-05, "loss": 0.6538, "step": 3326 }, { "epoch": 0.35, "learning_rate": 5.8895847617580824e-05, "loss": 0.871, "step": 3327 }, { "epoch": 0.35, "learning_rate": 5.889518588093316e-05, "loss": 0.5662, "step": 3328 }, { "epoch": 0.35, "learning_rate": 5.889452394976987e-05, "loss": 0.3053, "step": 3329 }, { "epoch": 0.35, "learning_rate": 5.8893861824095406e-05, "loss": 0.3234, "step": 3330 }, { "epoch": 0.35, "learning_rate": 5.889319950391422e-05, "loss": 1.0134, "step": 3331 }, { "epoch": 0.35, "learning_rate": 5.8892536989230776e-05, "loss": 0.6822, "step": 3332 }, { "epoch": 0.35, "learning_rate": 5.889187428004954e-05, "loss": 0.7215, "step": 3333 }, { "epoch": 0.35, "learning_rate": 5.889121137637497e-05, "loss": 1.3502, "step": 3334 }, { "epoch": 0.35, "learning_rate": 5.8890548278211515e-05, "loss": 0.9484, "step": 3335 }, { "epoch": 0.35, "learning_rate": 5.888988498556365e-05, "loss": 1.4835, "step": 3336 }, { "epoch": 0.35, "learning_rate": 5.888922149843584e-05, "loss": 1.0606, "step": 3337 }, { "epoch": 0.35, "learning_rate": 5.8888557816832556e-05, "loss": 0.6576, "step": 3338 }, { "epoch": 0.35, "learning_rate": 5.888789394075825e-05, "loss": 0.7796, "step": 3339 }, { "epoch": 0.35, "learning_rate": 5.8887229870217405e-05, "loss": 1.1352, "step": 3340 }, { "epoch": 0.35, "learning_rate": 5.888656560521449e-05, "loss": 0.5715, "step": 3341 }, { "epoch": 0.35, "learning_rate": 5.8885901145753975e-05, "loss": 1.1031, "step": 3342 }, { "epoch": 0.35, "learning_rate": 5.888523649184033e-05, "loss": 1.2893, "step": 3343 }, { "epoch": 0.35, "learning_rate": 5.888457164347803e-05, "loss": 1.0766, "step": 3344 }, { "epoch": 0.35, "learning_rate": 5.8883906600671555e-05, "loss": 1.0034, "step": 3345 }, { "epoch": 0.35, "learning_rate": 5.888324136342538e-05, "loss": 1.3097, "step": 3346 }, { "epoch": 0.35, "learning_rate": 5.888257593174398e-05, "loss": 1.2451, "step": 3347 }, { "epoch": 0.35, "learning_rate": 5.888191030563183e-05, "loss": 0.8991, "step": 3348 }, { "epoch": 0.35, "learning_rate": 5.8881244485093426e-05, "loss": 1.0746, "step": 3349 }, { "epoch": 0.35, "learning_rate": 5.888057847013324e-05, "loss": 0.6149, "step": 3350 }, { "epoch": 0.35, "learning_rate": 5.8879912260755764e-05, "loss": 0.517, "step": 3351 }, { "epoch": 0.35, "learning_rate": 5.8879245856965466e-05, "loss": 0.5133, "step": 3352 }, { "epoch": 0.35, "learning_rate": 5.887857925876685e-05, "loss": 0.9245, "step": 3353 }, { "epoch": 0.35, "learning_rate": 5.8877912466164386e-05, "loss": 1.5995, "step": 3354 }, { "epoch": 0.35, "learning_rate": 5.8877245479162583e-05, "loss": 0.4811, "step": 3355 }, { "epoch": 0.35, "learning_rate": 5.887657829776591e-05, "loss": 0.8562, "step": 3356 }, { "epoch": 0.35, "learning_rate": 5.8875910921978874e-05, "loss": 0.7802, "step": 3357 }, { "epoch": 0.35, "learning_rate": 5.887524335180596e-05, "loss": 2.2732, "step": 3358 }, { "epoch": 0.35, "learning_rate": 5.887457558725167e-05, "loss": 0.8788, "step": 3359 }, { "epoch": 0.35, "learning_rate": 5.887390762832049e-05, "loss": 1.3352, "step": 3360 }, { "epoch": 0.35, "learning_rate": 5.8873239475016914e-05, "loss": 0.5969, "step": 3361 }, { "epoch": 0.35, "learning_rate": 5.887257112734545e-05, "loss": 1.2856, "step": 3362 }, { "epoch": 0.35, "learning_rate": 5.8871902585310595e-05, "loss": 0.6603, "step": 3363 }, { "epoch": 0.35, "learning_rate": 5.887123384891684e-05, "loss": 1.2613, "step": 3364 }, { "epoch": 0.35, "learning_rate": 5.8870564918168705e-05, "loss": 1.5896, "step": 3365 }, { "epoch": 0.35, "learning_rate": 5.886989579307068e-05, "loss": 0.505, "step": 3366 }, { "epoch": 0.35, "learning_rate": 5.886922647362726e-05, "loss": 1.1908, "step": 3367 }, { "epoch": 0.35, "learning_rate": 5.886855695984297e-05, "loss": 1.0408, "step": 3368 }, { "epoch": 0.35, "learning_rate": 5.886788725172231e-05, "loss": 0.3496, "step": 3369 }, { "epoch": 0.35, "learning_rate": 5.886721734926979e-05, "loss": 1.443, "step": 3370 }, { "epoch": 0.35, "learning_rate": 5.8866547252489915e-05, "loss": 0.496, "step": 3371 }, { "epoch": 0.35, "learning_rate": 5.8865876961387195e-05, "loss": 0.3264, "step": 3372 }, { "epoch": 0.35, "learning_rate": 5.886520647596615e-05, "loss": 1.0881, "step": 3373 }, { "epoch": 0.35, "learning_rate": 5.886453579623128e-05, "loss": 0.2959, "step": 3374 }, { "epoch": 0.35, "learning_rate": 5.886386492218712e-05, "loss": 1.8262, "step": 3375 }, { "epoch": 0.35, "learning_rate": 5.886319385383817e-05, "loss": 0.5972, "step": 3376 }, { "epoch": 0.35, "learning_rate": 5.886252259118896e-05, "loss": 0.3708, "step": 3377 }, { "epoch": 0.35, "learning_rate": 5.8861851134243994e-05, "loss": 0.4222, "step": 3378 }, { "epoch": 0.35, "learning_rate": 5.886117948300779e-05, "loss": 0.3252, "step": 3379 }, { "epoch": 0.35, "learning_rate": 5.886050763748489e-05, "loss": 1.3197, "step": 3380 }, { "epoch": 0.35, "learning_rate": 5.88598355976798e-05, "loss": 1.1579, "step": 3381 }, { "epoch": 0.35, "learning_rate": 5.885916336359706e-05, "loss": 1.1611, "step": 3382 }, { "epoch": 0.35, "learning_rate": 5.8858490935241175e-05, "loss": 1.6174, "step": 3383 }, { "epoch": 0.35, "learning_rate": 5.885781831261668e-05, "loss": 1.2944, "step": 3384 }, { "epoch": 0.35, "learning_rate": 5.8857145495728105e-05, "loss": 0.8827, "step": 3385 }, { "epoch": 0.35, "learning_rate": 5.885647248457999e-05, "loss": 0.856, "step": 3386 }, { "epoch": 0.35, "learning_rate": 5.8855799279176846e-05, "loss": 1.0292, "step": 3387 }, { "epoch": 0.35, "learning_rate": 5.885512587952321e-05, "loss": 1.2303, "step": 3388 }, { "epoch": 0.35, "learning_rate": 5.885445228562362e-05, "loss": 1.4048, "step": 3389 }, { "epoch": 0.35, "learning_rate": 5.885377849748261e-05, "loss": 0.3024, "step": 3390 }, { "epoch": 0.35, "learning_rate": 5.885310451510472e-05, "loss": 1.1845, "step": 3391 }, { "epoch": 0.35, "learning_rate": 5.885243033849447e-05, "loss": 1.3295, "step": 3392 }, { "epoch": 0.35, "learning_rate": 5.885175596765642e-05, "loss": 0.4238, "step": 3393 }, { "epoch": 0.35, "learning_rate": 5.885108140259509e-05, "loss": 0.7544, "step": 3394 }, { "epoch": 0.35, "learning_rate": 5.8850406643315036e-05, "loss": 0.6365, "step": 3395 }, { "epoch": 0.35, "learning_rate": 5.884973168982079e-05, "loss": 0.2266, "step": 3396 }, { "epoch": 0.35, "learning_rate": 5.884905654211691e-05, "loss": 0.9828, "step": 3397 }, { "epoch": 0.35, "learning_rate": 5.8848381200207925e-05, "loss": 0.2368, "step": 3398 }, { "epoch": 0.35, "learning_rate": 5.884770566409839e-05, "loss": 0.6654, "step": 3399 }, { "epoch": 0.36, "learning_rate": 5.8847029933792845e-05, "loss": 0.5435, "step": 3400 }, { "epoch": 0.36, "learning_rate": 5.8846354009295855e-05, "loss": 0.9925, "step": 3401 }, { "epoch": 0.36, "learning_rate": 5.8845677890611954e-05, "loss": 0.911, "step": 3402 }, { "epoch": 0.36, "learning_rate": 5.8845001577745696e-05, "loss": 0.7394, "step": 3403 }, { "epoch": 0.36, "learning_rate": 5.884432507070164e-05, "loss": 0.76, "step": 3404 }, { "epoch": 0.36, "learning_rate": 5.884364836948433e-05, "loss": 1.2769, "step": 3405 }, { "epoch": 0.36, "learning_rate": 5.884297147409833e-05, "loss": 1.2659, "step": 3406 }, { "epoch": 0.36, "learning_rate": 5.88422943845482e-05, "loss": 1.1917, "step": 3407 }, { "epoch": 0.36, "learning_rate": 5.8841617100838496e-05, "loss": 1.4178, "step": 3408 }, { "epoch": 0.36, "learning_rate": 5.8840939622973774e-05, "loss": 0.7475, "step": 3409 }, { "epoch": 0.36, "learning_rate": 5.8840261950958584e-05, "loss": 0.9315, "step": 3410 }, { "epoch": 0.36, "learning_rate": 5.8839584084797506e-05, "loss": 0.5419, "step": 3411 }, { "epoch": 0.36, "learning_rate": 5.8838906024495095e-05, "loss": 1.1897, "step": 3412 }, { "epoch": 0.36, "learning_rate": 5.883822777005591e-05, "loss": 1.0303, "step": 3413 }, { "epoch": 0.36, "learning_rate": 5.883754932148453e-05, "loss": 0.7363, "step": 3414 }, { "epoch": 0.36, "learning_rate": 5.8836870678785516e-05, "loss": 0.4347, "step": 3415 }, { "epoch": 0.36, "learning_rate": 5.8836191841963434e-05, "loss": 0.7859, "step": 3416 }, { "epoch": 0.36, "learning_rate": 5.883551281102285e-05, "loss": 1.4208, "step": 3417 }, { "epoch": 0.36, "learning_rate": 5.883483358596835e-05, "loss": 1.1082, "step": 3418 }, { "epoch": 0.36, "learning_rate": 5.883415416680449e-05, "loss": 0.6, "step": 3419 }, { "epoch": 0.36, "learning_rate": 5.8833474553535855e-05, "loss": 0.3913, "step": 3420 }, { "epoch": 0.36, "learning_rate": 5.883279474616702e-05, "loss": 0.7495, "step": 3421 }, { "epoch": 0.36, "learning_rate": 5.883211474470255e-05, "loss": 0.7404, "step": 3422 }, { "epoch": 0.36, "learning_rate": 5.883143454914703e-05, "loss": 0.4927, "step": 3423 }, { "epoch": 0.36, "learning_rate": 5.8830754159505045e-05, "loss": 0.9264, "step": 3424 }, { "epoch": 0.36, "learning_rate": 5.883007357578116e-05, "loss": 1.2255, "step": 3425 }, { "epoch": 0.36, "learning_rate": 5.8829392797979976e-05, "loss": 1.6919, "step": 3426 }, { "epoch": 0.36, "learning_rate": 5.8828711826106054e-05, "loss": 0.8331, "step": 3427 }, { "epoch": 0.36, "learning_rate": 5.882803066016399e-05, "loss": 0.5975, "step": 3428 }, { "epoch": 0.36, "learning_rate": 5.8827349300158375e-05, "loss": 0.3601, "step": 3429 }, { "epoch": 0.36, "learning_rate": 5.882666774609379e-05, "loss": 0.6919, "step": 3430 }, { "epoch": 0.36, "learning_rate": 5.8825985997974816e-05, "loss": 1.1871, "step": 3431 }, { "epoch": 0.36, "learning_rate": 5.882530405580606e-05, "loss": 0.5786, "step": 3432 }, { "epoch": 0.36, "learning_rate": 5.882462191959209e-05, "loss": 0.4388, "step": 3433 }, { "epoch": 0.36, "learning_rate": 5.882393958933751e-05, "loss": 0.6991, "step": 3434 }, { "epoch": 0.36, "learning_rate": 5.882325706504692e-05, "loss": 0.8499, "step": 3435 }, { "epoch": 0.36, "learning_rate": 5.882257434672491e-05, "loss": 0.4115, "step": 3436 }, { "epoch": 0.36, "learning_rate": 5.882189143437607e-05, "loss": 0.8203, "step": 3437 }, { "epoch": 0.36, "learning_rate": 5.8821208328005e-05, "loss": 0.5751, "step": 3438 }, { "epoch": 0.36, "learning_rate": 5.8820525027616295e-05, "loss": 1.3199, "step": 3439 }, { "epoch": 0.36, "learning_rate": 5.881984153321456e-05, "loss": 1.6074, "step": 3440 }, { "epoch": 0.36, "learning_rate": 5.88191578448044e-05, "loss": 0.4893, "step": 3441 }, { "epoch": 0.36, "learning_rate": 5.881847396239041e-05, "loss": 0.4721, "step": 3442 }, { "epoch": 0.36, "learning_rate": 5.881778988597719e-05, "loss": 0.4626, "step": 3443 }, { "epoch": 0.36, "learning_rate": 5.881710561556936e-05, "loss": 0.6995, "step": 3444 }, { "epoch": 0.36, "learning_rate": 5.881642115117152e-05, "loss": 0.3903, "step": 3445 }, { "epoch": 0.36, "learning_rate": 5.881573649278827e-05, "loss": 0.3807, "step": 3446 }, { "epoch": 0.36, "learning_rate": 5.881505164042422e-05, "loss": 1.027, "step": 3447 }, { "epoch": 0.36, "learning_rate": 5.8814366594084e-05, "loss": 1.6018, "step": 3448 }, { "epoch": 0.36, "learning_rate": 5.881368135377219e-05, "loss": 0.7836, "step": 3449 }, { "epoch": 0.36, "learning_rate": 5.881299591949342e-05, "loss": 1.392, "step": 3450 }, { "epoch": 0.36, "learning_rate": 5.881231029125231e-05, "loss": 0.4853, "step": 3451 }, { "epoch": 0.36, "learning_rate": 5.881162446905347e-05, "loss": 1.2804, "step": 3452 }, { "epoch": 0.36, "learning_rate": 5.881093845290151e-05, "loss": 1.0398, "step": 3453 }, { "epoch": 0.36, "learning_rate": 5.881025224280105e-05, "loss": 1.3552, "step": 3454 }, { "epoch": 0.36, "learning_rate": 5.880956583875672e-05, "loss": 0.4513, "step": 3455 }, { "epoch": 0.36, "learning_rate": 5.8808879240773135e-05, "loss": 0.4294, "step": 3456 }, { "epoch": 0.36, "learning_rate": 5.880819244885491e-05, "loss": 1.2594, "step": 3457 }, { "epoch": 0.36, "learning_rate": 5.880750546300668e-05, "loss": 0.3935, "step": 3458 }, { "epoch": 0.36, "learning_rate": 5.880681828323306e-05, "loss": 1.3444, "step": 3459 }, { "epoch": 0.36, "learning_rate": 5.8806130909538674e-05, "loss": 0.8695, "step": 3460 }, { "epoch": 0.36, "learning_rate": 5.880544334192816e-05, "loss": 0.468, "step": 3461 }, { "epoch": 0.36, "learning_rate": 5.8804755580406146e-05, "loss": 1.2628, "step": 3462 }, { "epoch": 0.36, "learning_rate": 5.880406762497726e-05, "loss": 0.6812, "step": 3463 }, { "epoch": 0.36, "learning_rate": 5.8803379475646116e-05, "loss": 1.1483, "step": 3464 }, { "epoch": 0.36, "learning_rate": 5.880269113241737e-05, "loss": 0.7297, "step": 3465 }, { "epoch": 0.36, "learning_rate": 5.8802002595295646e-05, "loss": 0.8184, "step": 3466 }, { "epoch": 0.36, "learning_rate": 5.880131386428558e-05, "loss": 0.5217, "step": 3467 }, { "epoch": 0.36, "learning_rate": 5.88006249393918e-05, "loss": 1.0894, "step": 3468 }, { "epoch": 0.36, "learning_rate": 5.8799935820618965e-05, "loss": 0.816, "step": 3469 }, { "epoch": 0.36, "learning_rate": 5.879924650797169e-05, "loss": 0.511, "step": 3470 }, { "epoch": 0.36, "learning_rate": 5.8798557001454625e-05, "loss": 1.1378, "step": 3471 }, { "epoch": 0.36, "learning_rate": 5.879786730107242e-05, "loss": 0.1665, "step": 3472 }, { "epoch": 0.36, "learning_rate": 5.879717740682971e-05, "loss": 1.0672, "step": 3473 }, { "epoch": 0.36, "learning_rate": 5.879648731873114e-05, "loss": 1.3273, "step": 3474 }, { "epoch": 0.36, "learning_rate": 5.879579703678136e-05, "loss": 0.8614, "step": 3475 }, { "epoch": 0.36, "learning_rate": 5.879510656098501e-05, "loss": 1.0003, "step": 3476 }, { "epoch": 0.36, "learning_rate": 5.8794415891346725e-05, "loss": 0.8458, "step": 3477 }, { "epoch": 0.36, "learning_rate": 5.879372502787117e-05, "loss": 0.2535, "step": 3478 }, { "epoch": 0.36, "learning_rate": 5.879303397056301e-05, "loss": 0.8565, "step": 3479 }, { "epoch": 0.36, "learning_rate": 5.879234271942688e-05, "loss": 0.8309, "step": 3480 }, { "epoch": 0.36, "learning_rate": 5.879165127446743e-05, "loss": 0.5229, "step": 3481 }, { "epoch": 0.36, "learning_rate": 5.8790959635689324e-05, "loss": 1.4202, "step": 3482 }, { "epoch": 0.36, "learning_rate": 5.879026780309722e-05, "loss": 0.4121, "step": 3483 }, { "epoch": 0.36, "learning_rate": 5.8789575776695753e-05, "loss": 0.8243, "step": 3484 }, { "epoch": 0.36, "learning_rate": 5.878888355648962e-05, "loss": 0.4501, "step": 3485 }, { "epoch": 0.36, "learning_rate": 5.878819114248344e-05, "loss": 0.8939, "step": 3486 }, { "epoch": 0.36, "learning_rate": 5.8787498534681905e-05, "loss": 1.0085, "step": 3487 }, { "epoch": 0.36, "learning_rate": 5.8786805733089667e-05, "loss": 0.4334, "step": 3488 }, { "epoch": 0.36, "learning_rate": 5.8786112737711374e-05, "loss": 1.1565, "step": 3489 }, { "epoch": 0.36, "learning_rate": 5.8785419548551716e-05, "loss": 1.3126, "step": 3490 }, { "epoch": 0.36, "learning_rate": 5.878472616561535e-05, "loss": 1.1533, "step": 3491 }, { "epoch": 0.36, "learning_rate": 5.8784032588906933e-05, "loss": 0.6015, "step": 3492 }, { "epoch": 0.36, "learning_rate": 5.878333881843115e-05, "loss": 0.4899, "step": 3493 }, { "epoch": 0.36, "learning_rate": 5.878264485419267e-05, "loss": 0.884, "step": 3494 }, { "epoch": 0.37, "learning_rate": 5.878195069619615e-05, "loss": 1.1936, "step": 3495 }, { "epoch": 0.37, "learning_rate": 5.878125634444628e-05, "loss": 0.9954, "step": 3496 }, { "epoch": 0.37, "learning_rate": 5.8780561798947725e-05, "loss": 0.51, "step": 3497 }, { "epoch": 0.37, "learning_rate": 5.8779867059705156e-05, "loss": 1.1695, "step": 3498 }, { "epoch": 0.37, "learning_rate": 5.8779172126723256e-05, "loss": 0.7435, "step": 3499 }, { "epoch": 0.37, "learning_rate": 5.877847700000671e-05, "loss": 0.3975, "step": 3500 }, { "epoch": 0.37, "learning_rate": 5.877778167956019e-05, "loss": 1.1363, "step": 3501 }, { "epoch": 0.37, "learning_rate": 5.877708616538837e-05, "loss": 1.3342, "step": 3502 }, { "epoch": 0.37, "learning_rate": 5.8776390457495954e-05, "loss": 0.468, "step": 3503 }, { "epoch": 0.37, "learning_rate": 5.8775694555887595e-05, "loss": 0.6654, "step": 3504 }, { "epoch": 0.37, "learning_rate": 5.8774998460568e-05, "loss": 1.0588, "step": 3505 }, { "epoch": 0.37, "learning_rate": 5.877430217154185e-05, "loss": 1.7834, "step": 3506 }, { "epoch": 0.37, "learning_rate": 5.877360568881382e-05, "loss": 0.5515, "step": 3507 }, { "epoch": 0.37, "learning_rate": 5.877290901238862e-05, "loss": 0.3838, "step": 3508 }, { "epoch": 0.37, "learning_rate": 5.8772212142270926e-05, "loss": 0.9553, "step": 3509 }, { "epoch": 0.37, "learning_rate": 5.877151507846543e-05, "loss": 0.6359, "step": 3510 }, { "epoch": 0.37, "learning_rate": 5.877081782097682e-05, "loss": 0.4633, "step": 3511 }, { "epoch": 0.37, "learning_rate": 5.877012036980981e-05, "loss": 0.7726, "step": 3512 }, { "epoch": 0.37, "learning_rate": 5.876942272496908e-05, "loss": 0.6326, "step": 3513 }, { "epoch": 0.37, "learning_rate": 5.876872488645931e-05, "loss": 0.5296, "step": 3514 }, { "epoch": 0.37, "learning_rate": 5.876802685428524e-05, "loss": 0.5312, "step": 3515 }, { "epoch": 0.37, "learning_rate": 5.876732862845153e-05, "loss": 1.5944, "step": 3516 }, { "epoch": 0.37, "learning_rate": 5.87666302089629e-05, "loss": 0.6731, "step": 3517 }, { "epoch": 0.37, "learning_rate": 5.876593159582404e-05, "loss": 0.362, "step": 3518 }, { "epoch": 0.37, "learning_rate": 5.876523278903967e-05, "loss": 0.3227, "step": 3519 }, { "epoch": 0.37, "learning_rate": 5.8764533788614474e-05, "loss": 0.859, "step": 3520 }, { "epoch": 0.37, "learning_rate": 5.876383459455317e-05, "loss": 0.6265, "step": 3521 }, { "epoch": 0.37, "learning_rate": 5.876313520686046e-05, "loss": 0.3654, "step": 3522 }, { "epoch": 0.37, "learning_rate": 5.8762435625541056e-05, "loss": 0.6459, "step": 3523 }, { "epoch": 0.37, "learning_rate": 5.876173585059967e-05, "loss": 0.613, "step": 3524 }, { "epoch": 0.37, "learning_rate": 5.876103588204099e-05, "loss": 0.8492, "step": 3525 }, { "epoch": 0.37, "learning_rate": 5.8760335719869764e-05, "loss": 0.3294, "step": 3526 }, { "epoch": 0.37, "learning_rate": 5.8759635364090676e-05, "loss": 0.8043, "step": 3527 }, { "epoch": 0.37, "learning_rate": 5.875893481470846e-05, "loss": 1.181, "step": 3528 }, { "epoch": 0.37, "learning_rate": 5.8758234071727816e-05, "loss": 0.8389, "step": 3529 }, { "epoch": 0.37, "learning_rate": 5.8757533135153476e-05, "loss": 0.8333, "step": 3530 }, { "epoch": 0.37, "learning_rate": 5.875683200499015e-05, "loss": 0.3152, "step": 3531 }, { "epoch": 0.37, "learning_rate": 5.8756130681242565e-05, "loss": 1.3371, "step": 3532 }, { "epoch": 0.37, "learning_rate": 5.875542916391543e-05, "loss": 1.3999, "step": 3533 }, { "epoch": 0.37, "learning_rate": 5.875472745301347e-05, "loss": 1.2524, "step": 3534 }, { "epoch": 0.37, "learning_rate": 5.875402554854142e-05, "loss": 1.0732, "step": 3535 }, { "epoch": 0.37, "learning_rate": 5.8753323450503995e-05, "loss": 1.8582, "step": 3536 }, { "epoch": 0.37, "learning_rate": 5.875262115890593e-05, "loss": 1.4279, "step": 3537 }, { "epoch": 0.37, "learning_rate": 5.875191867375194e-05, "loss": 1.2708, "step": 3538 }, { "epoch": 0.37, "learning_rate": 5.8751215995046766e-05, "loss": 0.5283, "step": 3539 }, { "epoch": 0.37, "learning_rate": 5.875051312279513e-05, "loss": 0.5426, "step": 3540 }, { "epoch": 0.37, "learning_rate": 5.874981005700177e-05, "loss": 1.43, "step": 3541 }, { "epoch": 0.37, "learning_rate": 5.8749106797671404e-05, "loss": 1.0301, "step": 3542 }, { "epoch": 0.37, "learning_rate": 5.8748403344808786e-05, "loss": 0.7097, "step": 3543 }, { "epoch": 0.37, "learning_rate": 5.874769969841864e-05, "loss": 0.3813, "step": 3544 }, { "epoch": 0.37, "learning_rate": 5.874699585850572e-05, "loss": 1.1058, "step": 3545 }, { "epoch": 0.37, "learning_rate": 5.8746291825074735e-05, "loss": 0.8527, "step": 3546 }, { "epoch": 0.37, "learning_rate": 5.8745587598130445e-05, "loss": 0.936, "step": 3547 }, { "epoch": 0.37, "learning_rate": 5.874488317767758e-05, "loss": 1.1181, "step": 3548 }, { "epoch": 0.37, "learning_rate": 5.8744178563720896e-05, "loss": 0.5568, "step": 3549 }, { "epoch": 0.37, "learning_rate": 5.8743473756265124e-05, "loss": 0.5733, "step": 3550 }, { "epoch": 0.37, "learning_rate": 5.8742768755315015e-05, "loss": 0.5269, "step": 3551 }, { "epoch": 0.37, "learning_rate": 5.87420635608753e-05, "loss": 0.3571, "step": 3552 }, { "epoch": 0.37, "learning_rate": 5.874135817295075e-05, "loss": 0.551, "step": 3553 }, { "epoch": 0.37, "learning_rate": 5.87406525915461e-05, "loss": 0.9166, "step": 3554 }, { "epoch": 0.37, "learning_rate": 5.87399468166661e-05, "loss": 1.1461, "step": 3555 }, { "epoch": 0.37, "learning_rate": 5.87392408483155e-05, "loss": 0.9957, "step": 3556 }, { "epoch": 0.37, "learning_rate": 5.8738534686499055e-05, "loss": 1.7096, "step": 3557 }, { "epoch": 0.37, "learning_rate": 5.873782833122153e-05, "loss": 1.685, "step": 3558 }, { "epoch": 0.37, "learning_rate": 5.8737121782487656e-05, "loss": 1.3715, "step": 3559 }, { "epoch": 0.37, "learning_rate": 5.873641504030221e-05, "loss": 0.6659, "step": 3560 }, { "epoch": 0.37, "learning_rate": 5.873570810466994e-05, "loss": 1.0925, "step": 3561 }, { "epoch": 0.37, "learning_rate": 5.873500097559561e-05, "loss": 0.5769, "step": 3562 }, { "epoch": 0.37, "learning_rate": 5.873429365308396e-05, "loss": 1.3849, "step": 3563 }, { "epoch": 0.37, "learning_rate": 5.873358613713978e-05, "loss": 0.6461, "step": 3564 }, { "epoch": 0.37, "learning_rate": 5.873287842776783e-05, "loss": 0.9773, "step": 3565 }, { "epoch": 0.37, "learning_rate": 5.873217052497285e-05, "loss": 1.1094, "step": 3566 }, { "epoch": 0.37, "learning_rate": 5.8731462428759636e-05, "loss": 0.4646, "step": 3567 }, { "epoch": 0.37, "learning_rate": 5.8730754139132925e-05, "loss": 1.0562, "step": 3568 }, { "epoch": 0.37, "learning_rate": 5.87300456560975e-05, "loss": 0.5369, "step": 3569 }, { "epoch": 0.37, "learning_rate": 5.872933697965814e-05, "loss": 0.696, "step": 3570 }, { "epoch": 0.37, "learning_rate": 5.8728628109819606e-05, "loss": 0.8077, "step": 3571 }, { "epoch": 0.37, "learning_rate": 5.872791904658666e-05, "loss": 1.2501, "step": 3572 }, { "epoch": 0.37, "learning_rate": 5.872720978996409e-05, "loss": 0.8288, "step": 3573 }, { "epoch": 0.37, "learning_rate": 5.872650033995666e-05, "loss": 1.1958, "step": 3574 }, { "epoch": 0.37, "learning_rate": 5.872579069656916e-05, "loss": 0.532, "step": 3575 }, { "epoch": 0.37, "learning_rate": 5.872508085980635e-05, "loss": 0.3122, "step": 3576 }, { "epoch": 0.37, "learning_rate": 5.872437082967302e-05, "loss": 0.6004, "step": 3577 }, { "epoch": 0.37, "learning_rate": 5.872366060617395e-05, "loss": 1.0017, "step": 3578 }, { "epoch": 0.37, "learning_rate": 5.872295018931392e-05, "loss": 0.5023, "step": 3579 }, { "epoch": 0.37, "learning_rate": 5.87222395790977e-05, "loss": 0.7966, "step": 3580 }, { "epoch": 0.37, "learning_rate": 5.872152877553009e-05, "loss": 1.7947, "step": 3581 }, { "epoch": 0.37, "learning_rate": 5.872081777861587e-05, "loss": 1.214, "step": 3582 }, { "epoch": 0.37, "learning_rate": 5.872010658835982e-05, "loss": 1.4445, "step": 3583 }, { "epoch": 0.37, "learning_rate": 5.871939520476674e-05, "loss": 0.3722, "step": 3584 }, { "epoch": 0.37, "learning_rate": 5.87186836278414e-05, "loss": 0.6102, "step": 3585 }, { "epoch": 0.37, "learning_rate": 5.871797185758861e-05, "loss": 0.9428, "step": 3586 }, { "epoch": 0.37, "learning_rate": 5.8717259894013144e-05, "loss": 1.0264, "step": 3587 }, { "epoch": 0.37, "learning_rate": 5.871654773711982e-05, "loss": 0.5022, "step": 3588 }, { "epoch": 0.37, "learning_rate": 5.87158353869134e-05, "loss": 1.1674, "step": 3589 }, { "epoch": 0.37, "learning_rate": 5.87151228433987e-05, "loss": 0.4308, "step": 3590 }, { "epoch": 0.38, "learning_rate": 5.871441010658051e-05, "loss": 1.6545, "step": 3591 }, { "epoch": 0.38, "learning_rate": 5.871369717646364e-05, "loss": 0.7506, "step": 3592 }, { "epoch": 0.38, "learning_rate": 5.8712984053052866e-05, "loss": 1.584, "step": 3593 }, { "epoch": 0.38, "learning_rate": 5.871227073635301e-05, "loss": 1.0212, "step": 3594 }, { "epoch": 0.38, "learning_rate": 5.871155722636886e-05, "loss": 1.2534, "step": 3595 }, { "epoch": 0.38, "learning_rate": 5.871084352310523e-05, "loss": 0.5846, "step": 3596 }, { "epoch": 0.38, "learning_rate": 5.871012962656692e-05, "loss": 0.6972, "step": 3597 }, { "epoch": 0.38, "learning_rate": 5.8709415536758726e-05, "loss": 1.224, "step": 3598 }, { "epoch": 0.38, "learning_rate": 5.870870125368547e-05, "loss": 0.2722, "step": 3599 }, { "epoch": 0.38, "learning_rate": 5.8707986777351954e-05, "loss": 0.4456, "step": 3600 }, { "epoch": 0.38, "learning_rate": 5.870727210776299e-05, "loss": 1.1482, "step": 3601 }, { "epoch": 0.38, "learning_rate": 5.870655724492339e-05, "loss": 0.2133, "step": 3602 }, { "epoch": 0.38, "learning_rate": 5.870584218883795e-05, "loss": 0.857, "step": 3603 }, { "epoch": 0.38, "learning_rate": 5.870512693951151e-05, "loss": 0.6982, "step": 3604 }, { "epoch": 0.38, "learning_rate": 5.870441149694887e-05, "loss": 0.9778, "step": 3605 }, { "epoch": 0.38, "learning_rate": 5.870369586115484e-05, "loss": 0.5296, "step": 3606 }, { "epoch": 0.38, "learning_rate": 5.870298003213426e-05, "loss": 0.9032, "step": 3607 }, { "epoch": 0.38, "learning_rate": 5.870226400989192e-05, "loss": 0.8571, "step": 3608 }, { "epoch": 0.38, "learning_rate": 5.870154779443266e-05, "loss": 0.614, "step": 3609 }, { "epoch": 0.38, "learning_rate": 5.87008313857613e-05, "loss": 0.5058, "step": 3610 }, { "epoch": 0.38, "learning_rate": 5.870011478388265e-05, "loss": 0.5335, "step": 3611 }, { "epoch": 0.38, "learning_rate": 5.8699397988801544e-05, "loss": 1.1196, "step": 3612 }, { "epoch": 0.38, "learning_rate": 5.869868100052281e-05, "loss": 0.5873, "step": 3613 }, { "epoch": 0.38, "learning_rate": 5.869796381905127e-05, "loss": 1.0219, "step": 3614 }, { "epoch": 0.38, "learning_rate": 5.869724644439175e-05, "loss": 0.9093, "step": 3615 }, { "epoch": 0.38, "learning_rate": 5.8696528876549084e-05, "loss": 0.8364, "step": 3616 }, { "epoch": 0.38, "learning_rate": 5.869581111552811e-05, "loss": 0.4292, "step": 3617 }, { "epoch": 0.38, "learning_rate": 5.869509316133364e-05, "loss": 0.5324, "step": 3618 }, { "epoch": 0.38, "learning_rate": 5.869437501397052e-05, "loss": 0.4004, "step": 3619 }, { "epoch": 0.38, "learning_rate": 5.869365667344358e-05, "loss": 0.527, "step": 3620 }, { "epoch": 0.38, "learning_rate": 5.8692938139757644e-05, "loss": 1.3082, "step": 3621 }, { "epoch": 0.38, "learning_rate": 5.8692219412917576e-05, "loss": 0.4157, "step": 3622 }, { "epoch": 0.38, "learning_rate": 5.869150049292819e-05, "loss": 0.879, "step": 3623 }, { "epoch": 0.38, "learning_rate": 5.8690781379794346e-05, "loss": 0.4874, "step": 3624 }, { "epoch": 0.38, "learning_rate": 5.869006207352087e-05, "loss": 0.8056, "step": 3625 }, { "epoch": 0.38, "learning_rate": 5.868934257411262e-05, "loss": 1.2822, "step": 3626 }, { "epoch": 0.38, "learning_rate": 5.8688622881574415e-05, "loss": 0.5804, "step": 3627 }, { "epoch": 0.38, "learning_rate": 5.868790299591112e-05, "loss": 1.162, "step": 3628 }, { "epoch": 0.38, "learning_rate": 5.868718291712757e-05, "loss": 1.7382, "step": 3629 }, { "epoch": 0.38, "learning_rate": 5.868646264522862e-05, "loss": 1.2386, "step": 3630 }, { "epoch": 0.38, "learning_rate": 5.8685742180219115e-05, "loss": 0.3315, "step": 3631 }, { "epoch": 0.38, "learning_rate": 5.868502152210391e-05, "loss": 0.6612, "step": 3632 }, { "epoch": 0.38, "learning_rate": 5.868430067088784e-05, "loss": 0.7141, "step": 3633 }, { "epoch": 0.38, "learning_rate": 5.868357962657578e-05, "loss": 1.8738, "step": 3634 }, { "epoch": 0.38, "learning_rate": 5.8682858389172566e-05, "loss": 2.3722, "step": 3635 }, { "epoch": 0.38, "learning_rate": 5.868213695868307e-05, "loss": 0.521, "step": 3636 }, { "epoch": 0.38, "learning_rate": 5.868141533511214e-05, "loss": 1.3868, "step": 3637 }, { "epoch": 0.38, "learning_rate": 5.8680693518464615e-05, "loss": 1.2156, "step": 3638 }, { "epoch": 0.38, "learning_rate": 5.867997150874539e-05, "loss": 1.1932, "step": 3639 }, { "epoch": 0.38, "learning_rate": 5.86792493059593e-05, "loss": 0.7405, "step": 3640 }, { "epoch": 0.38, "learning_rate": 5.8678526910111215e-05, "loss": 1.1575, "step": 3641 }, { "epoch": 0.38, "learning_rate": 5.8677804321206e-05, "loss": 0.9333, "step": 3642 }, { "epoch": 0.38, "learning_rate": 5.8677081539248505e-05, "loss": 1.0171, "step": 3643 }, { "epoch": 0.38, "learning_rate": 5.8676358564243616e-05, "loss": 1.2239, "step": 3644 }, { "epoch": 0.38, "learning_rate": 5.86756353961962e-05, "loss": 0.3905, "step": 3645 }, { "epoch": 0.38, "learning_rate": 5.8674912035111104e-05, "loss": 1.0443, "step": 3646 }, { "epoch": 0.38, "learning_rate": 5.867418848099321e-05, "loss": 0.5164, "step": 3647 }, { "epoch": 0.38, "learning_rate": 5.867346473384739e-05, "loss": 1.0966, "step": 3648 }, { "epoch": 0.38, "learning_rate": 5.867274079367852e-05, "loss": 0.941, "step": 3649 }, { "epoch": 0.38, "learning_rate": 5.867201666049146e-05, "loss": 0.512, "step": 3650 }, { "epoch": 0.38, "learning_rate": 5.8671292334291094e-05, "loss": 1.572, "step": 3651 }, { "epoch": 0.38, "learning_rate": 5.86705678150823e-05, "loss": 0.7347, "step": 3652 }, { "epoch": 0.38, "learning_rate": 5.866984310286995e-05, "loss": 0.7112, "step": 3653 }, { "epoch": 0.38, "learning_rate": 5.866911819765892e-05, "loss": 1.2443, "step": 3654 }, { "epoch": 0.38, "learning_rate": 5.8668393099454106e-05, "loss": 0.6166, "step": 3655 }, { "epoch": 0.38, "learning_rate": 5.8667667808260364e-05, "loss": 0.8656, "step": 3656 }, { "epoch": 0.38, "learning_rate": 5.86669423240826e-05, "loss": 0.4411, "step": 3657 }, { "epoch": 0.38, "learning_rate": 5.8666216646925684e-05, "loss": 0.8239, "step": 3658 }, { "epoch": 0.38, "learning_rate": 5.8665490776794504e-05, "loss": 1.1871, "step": 3659 }, { "epoch": 0.38, "learning_rate": 5.8664764713693945e-05, "loss": 0.9, "step": 3660 }, { "epoch": 0.38, "learning_rate": 5.8664038457628906e-05, "loss": 1.4029, "step": 3661 }, { "epoch": 0.38, "learning_rate": 5.866331200860425e-05, "loss": 0.3827, "step": 3662 }, { "epoch": 0.38, "learning_rate": 5.8662585366624905e-05, "loss": 1.439, "step": 3663 }, { "epoch": 0.38, "learning_rate": 5.8661858531695734e-05, "loss": 0.4309, "step": 3664 }, { "epoch": 0.38, "learning_rate": 5.8661131503821626e-05, "loss": 1.3963, "step": 3665 }, { "epoch": 0.38, "learning_rate": 5.86604042830075e-05, "loss": 0.2058, "step": 3666 }, { "epoch": 0.38, "learning_rate": 5.8659676869258236e-05, "loss": 0.4785, "step": 3667 }, { "epoch": 0.38, "learning_rate": 5.865894926257873e-05, "loss": 0.4952, "step": 3668 }, { "epoch": 0.38, "learning_rate": 5.865822146297389e-05, "loss": 0.4961, "step": 3669 }, { "epoch": 0.38, "learning_rate": 5.8657493470448604e-05, "loss": 1.8289, "step": 3670 }, { "epoch": 0.38, "learning_rate": 5.865676528500778e-05, "loss": 1.5466, "step": 3671 }, { "epoch": 0.38, "learning_rate": 5.865603690665631e-05, "loss": 0.4894, "step": 3672 }, { "epoch": 0.38, "learning_rate": 5.865530833539912e-05, "loss": 0.4497, "step": 3673 }, { "epoch": 0.38, "learning_rate": 5.865457957124108e-05, "loss": 1.2863, "step": 3674 }, { "epoch": 0.38, "learning_rate": 5.865385061418712e-05, "loss": 0.5444, "step": 3675 }, { "epoch": 0.38, "learning_rate": 5.865312146424215e-05, "loss": 1.1171, "step": 3676 }, { "epoch": 0.38, "learning_rate": 5.8652392121411066e-05, "loss": 0.5688, "step": 3677 }, { "epoch": 0.38, "learning_rate": 5.865166258569878e-05, "loss": 0.6634, "step": 3678 }, { "epoch": 0.38, "learning_rate": 5.865093285711021e-05, "loss": 1.1162, "step": 3679 }, { "epoch": 0.38, "learning_rate": 5.865020293565026e-05, "loss": 0.7774, "step": 3680 }, { "epoch": 0.38, "learning_rate": 5.864947282132385e-05, "loss": 1.1797, "step": 3681 }, { "epoch": 0.38, "learning_rate": 5.864874251413589e-05, "loss": 1.0072, "step": 3682 }, { "epoch": 0.38, "learning_rate": 5.86480120140913e-05, "loss": 0.4141, "step": 3683 }, { "epoch": 0.38, "learning_rate": 5.864728132119499e-05, "loss": 0.704, "step": 3684 }, { "epoch": 0.38, "learning_rate": 5.86465504354519e-05, "loss": 0.4051, "step": 3685 }, { "epoch": 0.38, "learning_rate": 5.8645819356866926e-05, "loss": 0.3452, "step": 3686 }, { "epoch": 0.39, "learning_rate": 5.8645088085444997e-05, "loss": 0.6748, "step": 3687 }, { "epoch": 0.39, "learning_rate": 5.8644356621191036e-05, "loss": 0.3624, "step": 3688 }, { "epoch": 0.39, "learning_rate": 5.864362496410997e-05, "loss": 0.6224, "step": 3689 }, { "epoch": 0.39, "learning_rate": 5.864289311420673e-05, "loss": 0.2874, "step": 3690 }, { "epoch": 0.39, "learning_rate": 5.864216107148623e-05, "loss": 1.2529, "step": 3691 }, { "epoch": 0.39, "learning_rate": 5.8641428835953405e-05, "loss": 1.1429, "step": 3692 }, { "epoch": 0.39, "learning_rate": 5.8640696407613184e-05, "loss": 0.7474, "step": 3693 }, { "epoch": 0.39, "learning_rate": 5.863996378647049e-05, "loss": 1.1742, "step": 3694 }, { "epoch": 0.39, "learning_rate": 5.8639230972530265e-05, "loss": 1.3015, "step": 3695 }, { "epoch": 0.39, "learning_rate": 5.863849796579744e-05, "loss": 0.4498, "step": 3696 }, { "epoch": 0.39, "learning_rate": 5.8637764766276946e-05, "loss": 0.5083, "step": 3697 }, { "epoch": 0.39, "learning_rate": 5.863703137397372e-05, "loss": 1.3449, "step": 3698 }, { "epoch": 0.39, "learning_rate": 5.86362977888927e-05, "loss": 0.4967, "step": 3699 }, { "epoch": 0.39, "learning_rate": 5.863556401103882e-05, "loss": 0.7102, "step": 3700 }, { "epoch": 0.39, "learning_rate": 5.863483004041702e-05, "loss": 1.0397, "step": 3701 }, { "epoch": 0.39, "learning_rate": 5.8634095877032253e-05, "loss": 0.2213, "step": 3702 }, { "epoch": 0.39, "learning_rate": 5.863336152088945e-05, "loss": 0.3639, "step": 3703 }, { "epoch": 0.39, "learning_rate": 5.863262697199355e-05, "loss": 0.2878, "step": 3704 }, { "epoch": 0.39, "learning_rate": 5.863189223034952e-05, "loss": 0.8102, "step": 3705 }, { "epoch": 0.39, "learning_rate": 5.863115729596228e-05, "loss": 1.1826, "step": 3706 }, { "epoch": 0.39, "learning_rate": 5.863042216883679e-05, "loss": 0.8733, "step": 3707 }, { "epoch": 0.39, "learning_rate": 5.8629686848978e-05, "loss": 1.0063, "step": 3708 }, { "epoch": 0.39, "learning_rate": 5.862895133639085e-05, "loss": 1.1045, "step": 3709 }, { "epoch": 0.39, "learning_rate": 5.86282156310803e-05, "loss": 1.5285, "step": 3710 }, { "epoch": 0.39, "learning_rate": 5.862747973305131e-05, "loss": 0.0725, "step": 3711 }, { "epoch": 0.39, "learning_rate": 5.862674364230882e-05, "loss": 1.6087, "step": 3712 }, { "epoch": 0.39, "learning_rate": 5.862600735885778e-05, "loss": 0.4739, "step": 3713 }, { "epoch": 0.39, "learning_rate": 5.862527088270316e-05, "loss": 0.8882, "step": 3714 }, { "epoch": 0.39, "learning_rate": 5.862453421384992e-05, "loss": 0.6897, "step": 3715 }, { "epoch": 0.39, "learning_rate": 5.862379735230302e-05, "loss": 1.277, "step": 3716 }, { "epoch": 0.39, "learning_rate": 5.86230602980674e-05, "loss": 0.4953, "step": 3717 }, { "epoch": 0.39, "learning_rate": 5.8622323051148036e-05, "loss": 0.7237, "step": 3718 }, { "epoch": 0.39, "learning_rate": 5.86215856115499e-05, "loss": 1.4562, "step": 3719 }, { "epoch": 0.39, "learning_rate": 5.862084797927794e-05, "loss": 0.4587, "step": 3720 }, { "epoch": 0.39, "learning_rate": 5.8620110154337133e-05, "loss": 0.2484, "step": 3721 }, { "epoch": 0.39, "learning_rate": 5.8619372136732436e-05, "loss": 0.8551, "step": 3722 }, { "epoch": 0.39, "learning_rate": 5.861863392646883e-05, "loss": 1.6235, "step": 3723 }, { "epoch": 0.39, "learning_rate": 5.861789552355126e-05, "loss": 1.2879, "step": 3724 }, { "epoch": 0.39, "learning_rate": 5.8617156927984724e-05, "loss": 1.3114, "step": 3725 }, { "epoch": 0.39, "learning_rate": 5.8616418139774185e-05, "loss": 0.3563, "step": 3726 }, { "epoch": 0.39, "learning_rate": 5.861567915892462e-05, "loss": 1.1884, "step": 3727 }, { "epoch": 0.39, "learning_rate": 5.8614939985440984e-05, "loss": 1.7242, "step": 3728 }, { "epoch": 0.39, "learning_rate": 5.861420061932828e-05, "loss": 1.3253, "step": 3729 }, { "epoch": 0.39, "learning_rate": 5.861346106059147e-05, "loss": 0.3777, "step": 3730 }, { "epoch": 0.39, "learning_rate": 5.8612721309235527e-05, "loss": 0.594, "step": 3731 }, { "epoch": 0.39, "learning_rate": 5.8611981365265444e-05, "loss": 1.107, "step": 3732 }, { "epoch": 0.39, "learning_rate": 5.86112412286862e-05, "loss": 1.8505, "step": 3733 }, { "epoch": 0.39, "learning_rate": 5.861050089950277e-05, "loss": 1.4804, "step": 3734 }, { "epoch": 0.39, "learning_rate": 5.8609760377720146e-05, "loss": 0.9762, "step": 3735 }, { "epoch": 0.39, "learning_rate": 5.860901966334331e-05, "loss": 1.1181, "step": 3736 }, { "epoch": 0.39, "learning_rate": 5.860827875637724e-05, "loss": 0.4958, "step": 3737 }, { "epoch": 0.39, "learning_rate": 5.860753765682694e-05, "loss": 0.6916, "step": 3738 }, { "epoch": 0.39, "learning_rate": 5.8606796364697386e-05, "loss": 1.5346, "step": 3739 }, { "epoch": 0.39, "learning_rate": 5.860605487999357e-05, "loss": 0.7765, "step": 3740 }, { "epoch": 0.39, "learning_rate": 5.860531320272049e-05, "loss": 0.6365, "step": 3741 }, { "epoch": 0.39, "learning_rate": 5.8604571332883134e-05, "loss": 1.022, "step": 3742 }, { "epoch": 0.39, "learning_rate": 5.8603829270486496e-05, "loss": 0.883, "step": 3743 }, { "epoch": 0.39, "learning_rate": 5.860308701553558e-05, "loss": 0.5021, "step": 3744 }, { "epoch": 0.39, "learning_rate": 5.8602344568035365e-05, "loss": 1.297, "step": 3745 }, { "epoch": 0.39, "learning_rate": 5.860160192799086e-05, "loss": 1.0674, "step": 3746 }, { "epoch": 0.39, "learning_rate": 5.8600859095407065e-05, "loss": 1.3773, "step": 3747 }, { "epoch": 0.39, "learning_rate": 5.8600116070288975e-05, "loss": 1.3831, "step": 3748 }, { "epoch": 0.39, "learning_rate": 5.8599372852641596e-05, "loss": 0.4973, "step": 3749 }, { "epoch": 0.39, "learning_rate": 5.859862944246994e-05, "loss": 0.6359, "step": 3750 }, { "epoch": 0.39, "learning_rate": 5.859788583977899e-05, "loss": 1.0705, "step": 3751 }, { "epoch": 0.39, "learning_rate": 5.859714204457377e-05, "loss": 1.2934, "step": 3752 }, { "epoch": 0.39, "learning_rate": 5.8596398056859284e-05, "loss": 0.8286, "step": 3753 }, { "epoch": 0.39, "learning_rate": 5.859565387664053e-05, "loss": 1.1748, "step": 3754 }, { "epoch": 0.39, "learning_rate": 5.8594909503922535e-05, "loss": 1.0996, "step": 3755 }, { "epoch": 0.39, "learning_rate": 5.859416493871029e-05, "loss": 0.7364, "step": 3756 }, { "epoch": 0.39, "learning_rate": 5.8593420181008815e-05, "loss": 0.8289, "step": 3757 }, { "epoch": 0.39, "learning_rate": 5.859267523082313e-05, "loss": 0.5907, "step": 3758 }, { "epoch": 0.39, "learning_rate": 5.8591930088158247e-05, "loss": 1.5059, "step": 3759 }, { "epoch": 0.39, "learning_rate": 5.859118475301918e-05, "loss": 0.6204, "step": 3760 }, { "epoch": 0.39, "learning_rate": 5.8590439225410945e-05, "loss": 0.7582, "step": 3761 }, { "epoch": 0.39, "learning_rate": 5.858969350533857e-05, "loss": 1.4598, "step": 3762 }, { "epoch": 0.39, "learning_rate": 5.858894759280706e-05, "loss": 0.914, "step": 3763 }, { "epoch": 0.39, "learning_rate": 5.8588201487821445e-05, "loss": 1.1325, "step": 3764 }, { "epoch": 0.39, "learning_rate": 5.8587455190386746e-05, "loss": 0.9127, "step": 3765 }, { "epoch": 0.39, "learning_rate": 5.8586708700508e-05, "loss": 0.6128, "step": 3766 }, { "epoch": 0.39, "learning_rate": 5.858596201819021e-05, "loss": 0.6922, "step": 3767 }, { "epoch": 0.39, "learning_rate": 5.85852151434384e-05, "loss": 0.7019, "step": 3768 }, { "epoch": 0.39, "learning_rate": 5.858446807625763e-05, "loss": 0.4918, "step": 3769 }, { "epoch": 0.39, "learning_rate": 5.85837208166529e-05, "loss": 0.4155, "step": 3770 }, { "epoch": 0.39, "learning_rate": 5.858297336462925e-05, "loss": 0.3136, "step": 3771 }, { "epoch": 0.39, "learning_rate": 5.8582225720191706e-05, "loss": 0.4862, "step": 3772 }, { "epoch": 0.39, "learning_rate": 5.8581477883345316e-05, "loss": 1.2567, "step": 3773 }, { "epoch": 0.39, "learning_rate": 5.858072985409511e-05, "loss": 1.4081, "step": 3774 }, { "epoch": 0.39, "learning_rate": 5.85799816324461e-05, "loss": 0.7386, "step": 3775 }, { "epoch": 0.39, "learning_rate": 5.8579233218403346e-05, "loss": 0.9427, "step": 3776 }, { "epoch": 0.39, "learning_rate": 5.857848461197188e-05, "loss": 0.3387, "step": 3777 }, { "epoch": 0.39, "learning_rate": 5.8577735813156755e-05, "loss": 0.6647, "step": 3778 }, { "epoch": 0.39, "learning_rate": 5.857698682196299e-05, "loss": 1.153, "step": 3779 }, { "epoch": 0.39, "learning_rate": 5.857623763839564e-05, "loss": 0.4873, "step": 3780 }, { "epoch": 0.39, "learning_rate": 5.8575488262459734e-05, "loss": 0.8553, "step": 3781 }, { "epoch": 0.39, "learning_rate": 5.857473869416034e-05, "loss": 0.5176, "step": 3782 }, { "epoch": 0.4, "learning_rate": 5.857398893350248e-05, "loss": 1.1427, "step": 3783 }, { "epoch": 0.4, "learning_rate": 5.857323898049122e-05, "loss": 1.4629, "step": 3784 }, { "epoch": 0.4, "learning_rate": 5.85724888351316e-05, "loss": 0.5488, "step": 3785 }, { "epoch": 0.4, "learning_rate": 5.857173849742867e-05, "loss": 1.1967, "step": 3786 }, { "epoch": 0.4, "learning_rate": 5.857098796738749e-05, "loss": 0.8827, "step": 3787 }, { "epoch": 0.4, "learning_rate": 5.857023724501309e-05, "loss": 0.557, "step": 3788 }, { "epoch": 0.4, "learning_rate": 5.856948633031054e-05, "loss": 1.1167, "step": 3789 }, { "epoch": 0.4, "learning_rate": 5.8568735223284895e-05, "loss": 1.0796, "step": 3790 }, { "epoch": 0.4, "learning_rate": 5.856798392394121e-05, "loss": 0.5468, "step": 3791 }, { "epoch": 0.4, "learning_rate": 5.8567232432284543e-05, "loss": 0.8081, "step": 3792 }, { "epoch": 0.4, "learning_rate": 5.856648074831995e-05, "loss": 0.211, "step": 3793 }, { "epoch": 0.4, "learning_rate": 5.8565728872052495e-05, "loss": 0.3919, "step": 3794 }, { "epoch": 0.4, "learning_rate": 5.856497680348723e-05, "loss": 1.1514, "step": 3795 }, { "epoch": 0.4, "learning_rate": 5.856422454262923e-05, "loss": 0.3898, "step": 3796 }, { "epoch": 0.4, "learning_rate": 5.8563472089483555e-05, "loss": 0.6506, "step": 3797 }, { "epoch": 0.4, "learning_rate": 5.856271944405527e-05, "loss": 1.1461, "step": 3798 }, { "epoch": 0.4, "learning_rate": 5.8561966606349436e-05, "loss": 1.4331, "step": 3799 }, { "epoch": 0.4, "learning_rate": 5.856121357637113e-05, "loss": 0.8731, "step": 3800 }, { "epoch": 0.4, "learning_rate": 5.856046035412541e-05, "loss": 1.3931, "step": 3801 }, { "epoch": 0.4, "learning_rate": 5.855970693961736e-05, "loss": 0.6894, "step": 3802 }, { "epoch": 0.4, "learning_rate": 5.855895333285205e-05, "loss": 1.2032, "step": 3803 }, { "epoch": 0.4, "learning_rate": 5.8558199533834534e-05, "loss": 0.5719, "step": 3804 }, { "epoch": 0.4, "learning_rate": 5.8557445542569915e-05, "loss": 0.3555, "step": 3805 }, { "epoch": 0.4, "learning_rate": 5.8556691359063246e-05, "loss": 1.6837, "step": 3806 }, { "epoch": 0.4, "learning_rate": 5.855593698331961e-05, "loss": 0.3721, "step": 3807 }, { "epoch": 0.4, "learning_rate": 5.85551824153441e-05, "loss": 0.2514, "step": 3808 }, { "epoch": 0.4, "learning_rate": 5.855442765514177e-05, "loss": 0.429, "step": 3809 }, { "epoch": 0.4, "learning_rate": 5.855367270271772e-05, "loss": 0.328, "step": 3810 }, { "epoch": 0.4, "learning_rate": 5.855291755807703e-05, "loss": 0.4299, "step": 3811 }, { "epoch": 0.4, "learning_rate": 5.855216222122477e-05, "loss": 0.3827, "step": 3812 }, { "epoch": 0.4, "learning_rate": 5.8551406692166044e-05, "loss": 0.3966, "step": 3813 }, { "epoch": 0.4, "learning_rate": 5.855065097090593e-05, "loss": 0.5921, "step": 3814 }, { "epoch": 0.4, "learning_rate": 5.8549895057449515e-05, "loss": 0.6933, "step": 3815 }, { "epoch": 0.4, "learning_rate": 5.854913895180187e-05, "loss": 0.5853, "step": 3816 }, { "epoch": 0.4, "learning_rate": 5.854838265396812e-05, "loss": 0.2249, "step": 3817 }, { "epoch": 0.4, "learning_rate": 5.854762616395333e-05, "loss": 0.3314, "step": 3818 }, { "epoch": 0.4, "learning_rate": 5.85468694817626e-05, "loss": 0.5572, "step": 3819 }, { "epoch": 0.4, "learning_rate": 5.854611260740103e-05, "loss": 1.3723, "step": 3820 }, { "epoch": 0.4, "learning_rate": 5.8545355540873695e-05, "loss": 1.2748, "step": 3821 }, { "epoch": 0.4, "learning_rate": 5.854459828218572e-05, "loss": 0.8409, "step": 3822 }, { "epoch": 0.4, "learning_rate": 5.8543840831342187e-05, "loss": 1.0304, "step": 3823 }, { "epoch": 0.4, "learning_rate": 5.854308318834819e-05, "loss": 0.224, "step": 3824 }, { "epoch": 0.4, "learning_rate": 5.854232535320884e-05, "loss": 1.1064, "step": 3825 }, { "epoch": 0.4, "learning_rate": 5.854156732592923e-05, "loss": 0.7745, "step": 3826 }, { "epoch": 0.4, "learning_rate": 5.854080910651447e-05, "loss": 0.2903, "step": 3827 }, { "epoch": 0.4, "learning_rate": 5.8540050694969664e-05, "loss": 1.4688, "step": 3828 }, { "epoch": 0.4, "learning_rate": 5.85392920912999e-05, "loss": 0.6094, "step": 3829 }, { "epoch": 0.4, "learning_rate": 5.853853329551031e-05, "loss": 1.5619, "step": 3830 }, { "epoch": 0.4, "learning_rate": 5.8537774307606e-05, "loss": 1.5297, "step": 3831 }, { "epoch": 0.4, "learning_rate": 5.853701512759206e-05, "loss": 0.3774, "step": 3832 }, { "epoch": 0.4, "learning_rate": 5.853625575547361e-05, "loss": 0.9813, "step": 3833 }, { "epoch": 0.4, "learning_rate": 5.8535496191255775e-05, "loss": 0.4449, "step": 3834 }, { "epoch": 0.4, "learning_rate": 5.8534736434943645e-05, "loss": 1.1658, "step": 3835 }, { "epoch": 0.4, "learning_rate": 5.8533976486542345e-05, "loss": 0.737, "step": 3836 }, { "epoch": 0.4, "learning_rate": 5.8533216346057e-05, "loss": 1.4426, "step": 3837 }, { "epoch": 0.4, "learning_rate": 5.8532456013492715e-05, "loss": 0.5047, "step": 3838 }, { "epoch": 0.4, "learning_rate": 5.853169548885461e-05, "loss": 0.5868, "step": 3839 }, { "epoch": 0.4, "learning_rate": 5.853093477214781e-05, "loss": 0.3955, "step": 3840 }, { "epoch": 0.4, "learning_rate": 5.8530173863377434e-05, "loss": 0.9236, "step": 3841 }, { "epoch": 0.4, "learning_rate": 5.852941276254859e-05, "loss": 0.4942, "step": 3842 }, { "epoch": 0.4, "learning_rate": 5.852865146966643e-05, "loss": 0.7555, "step": 3843 }, { "epoch": 0.4, "learning_rate": 5.852788998473606e-05, "loss": 0.5384, "step": 3844 }, { "epoch": 0.4, "learning_rate": 5.8527128307762605e-05, "loss": 0.8594, "step": 3845 }, { "epoch": 0.4, "learning_rate": 5.85263664387512e-05, "loss": 0.3813, "step": 3846 }, { "epoch": 0.4, "learning_rate": 5.852560437770697e-05, "loss": 1.2834, "step": 3847 }, { "epoch": 0.4, "learning_rate": 5.852484212463505e-05, "loss": 0.8165, "step": 3848 }, { "epoch": 0.4, "learning_rate": 5.8524079679540564e-05, "loss": 1.1141, "step": 3849 }, { "epoch": 0.4, "learning_rate": 5.852331704242864e-05, "loss": 0.6914, "step": 3850 }, { "epoch": 0.4, "learning_rate": 5.852255421330443e-05, "loss": 0.5716, "step": 3851 }, { "epoch": 0.4, "learning_rate": 5.8521791192173054e-05, "loss": 0.3973, "step": 3852 }, { "epoch": 0.4, "learning_rate": 5.8521027979039656e-05, "loss": 1.3258, "step": 3853 }, { "epoch": 0.4, "learning_rate": 5.852026457390937e-05, "loss": 1.1047, "step": 3854 }, { "epoch": 0.4, "learning_rate": 5.8519500976787344e-05, "loss": 0.7976, "step": 3855 }, { "epoch": 0.4, "learning_rate": 5.85187371876787e-05, "loss": 0.3679, "step": 3856 }, { "epoch": 0.4, "learning_rate": 5.8517973206588595e-05, "loss": 1.2458, "step": 3857 }, { "epoch": 0.4, "learning_rate": 5.851720903352216e-05, "loss": 1.5588, "step": 3858 }, { "epoch": 0.4, "learning_rate": 5.851644466848455e-05, "loss": 1.7341, "step": 3859 }, { "epoch": 0.4, "learning_rate": 5.851568011148091e-05, "loss": 1.5487, "step": 3860 }, { "epoch": 0.4, "learning_rate": 5.8514915362516385e-05, "loss": 0.2488, "step": 3861 }, { "epoch": 0.4, "learning_rate": 5.851415042159612e-05, "loss": 1.1587, "step": 3862 }, { "epoch": 0.4, "learning_rate": 5.8513385288725255e-05, "loss": 0.8573, "step": 3863 }, { "epoch": 0.4, "learning_rate": 5.8512619963908964e-05, "loss": 1.2334, "step": 3864 }, { "epoch": 0.4, "learning_rate": 5.851185444715239e-05, "loss": 0.7958, "step": 3865 }, { "epoch": 0.4, "learning_rate": 5.8511088738460674e-05, "loss": 0.7873, "step": 3866 }, { "epoch": 0.4, "learning_rate": 5.851032283783898e-05, "loss": 0.7291, "step": 3867 }, { "epoch": 0.4, "learning_rate": 5.8509556745292465e-05, "loss": 0.5877, "step": 3868 }, { "epoch": 0.4, "learning_rate": 5.8508790460826293e-05, "loss": 1.4291, "step": 3869 }, { "epoch": 0.4, "learning_rate": 5.850802398444561e-05, "loss": 0.6762, "step": 3870 }, { "epoch": 0.4, "learning_rate": 5.8507257316155566e-05, "loss": 1.5797, "step": 3871 }, { "epoch": 0.4, "learning_rate": 5.850649045596136e-05, "loss": 1.0529, "step": 3872 }, { "epoch": 0.4, "learning_rate": 5.8505723403868115e-05, "loss": 1.051, "step": 3873 }, { "epoch": 0.4, "learning_rate": 5.850495615988101e-05, "loss": 0.2723, "step": 3874 }, { "epoch": 0.4, "learning_rate": 5.850418872400521e-05, "loss": 1.0104, "step": 3875 }, { "epoch": 0.4, "learning_rate": 5.8503421096245884e-05, "loss": 0.1714, "step": 3876 }, { "epoch": 0.4, "learning_rate": 5.850265327660819e-05, "loss": 1.2218, "step": 3877 }, { "epoch": 0.41, "learning_rate": 5.8501885265097306e-05, "loss": 0.5463, "step": 3878 }, { "epoch": 0.41, "learning_rate": 5.8501117061718406e-05, "loss": 0.7836, "step": 3879 }, { "epoch": 0.41, "learning_rate": 5.850034866647665e-05, "loss": 0.7627, "step": 3880 }, { "epoch": 0.41, "learning_rate": 5.8499580079377216e-05, "loss": 0.2575, "step": 3881 }, { "epoch": 0.41, "learning_rate": 5.849881130042527e-05, "loss": 0.8558, "step": 3882 }, { "epoch": 0.41, "learning_rate": 5.849804232962601e-05, "loss": 1.3443, "step": 3883 }, { "epoch": 0.41, "learning_rate": 5.849727316698459e-05, "loss": 0.5269, "step": 3884 }, { "epoch": 0.41, "learning_rate": 5.849650381250619e-05, "loss": 0.7561, "step": 3885 }, { "epoch": 0.41, "learning_rate": 5.8495734266196e-05, "loss": 0.66, "step": 3886 }, { "epoch": 0.41, "learning_rate": 5.849496452805919e-05, "loss": 1.2332, "step": 3887 }, { "epoch": 0.41, "learning_rate": 5.8494194598100945e-05, "loss": 1.2675, "step": 3888 }, { "epoch": 0.41, "learning_rate": 5.8493424476326454e-05, "loss": 0.9609, "step": 3889 }, { "epoch": 0.41, "learning_rate": 5.84926541627409e-05, "loss": 0.3813, "step": 3890 }, { "epoch": 0.41, "learning_rate": 5.8491883657349455e-05, "loss": 1.2454, "step": 3891 }, { "epoch": 0.41, "learning_rate": 5.849111296015732e-05, "loss": 0.3375, "step": 3892 }, { "epoch": 0.41, "learning_rate": 5.849034207116968e-05, "loss": 1.3088, "step": 3893 }, { "epoch": 0.41, "learning_rate": 5.848957099039172e-05, "loss": 0.9248, "step": 3894 }, { "epoch": 0.41, "learning_rate": 5.848879971782864e-05, "loss": 1.3981, "step": 3895 }, { "epoch": 0.41, "learning_rate": 5.8488028253485624e-05, "loss": 1.3548, "step": 3896 }, { "epoch": 0.41, "learning_rate": 5.848725659736786e-05, "loss": 0.3816, "step": 3897 }, { "epoch": 0.41, "learning_rate": 5.848648474948056e-05, "loss": 2.2916, "step": 3898 }, { "epoch": 0.41, "learning_rate": 5.8485712709828905e-05, "loss": 0.7075, "step": 3899 }, { "epoch": 0.41, "learning_rate": 5.8484940478418096e-05, "loss": 0.6733, "step": 3900 }, { "epoch": 0.41, "learning_rate": 5.848416805525333e-05, "loss": 1.4738, "step": 3901 }, { "epoch": 0.41, "learning_rate": 5.8483395440339825e-05, "loss": 1.8758, "step": 3902 }, { "epoch": 0.41, "learning_rate": 5.848262263368275e-05, "loss": 0.5016, "step": 3903 }, { "epoch": 0.41, "learning_rate": 5.8481849635287335e-05, "loss": 1.1801, "step": 3904 }, { "epoch": 0.41, "learning_rate": 5.848107644515877e-05, "loss": 0.2972, "step": 3905 }, { "epoch": 0.41, "learning_rate": 5.848030306330225e-05, "loss": 0.4243, "step": 3906 }, { "epoch": 0.41, "learning_rate": 5.847952948972301e-05, "loss": 0.6659, "step": 3907 }, { "epoch": 0.41, "learning_rate": 5.847875572442623e-05, "loss": 0.9034, "step": 3908 }, { "epoch": 0.41, "learning_rate": 5.847798176741714e-05, "loss": 1.0896, "step": 3909 }, { "epoch": 0.41, "learning_rate": 5.8477207618700934e-05, "loss": 0.3571, "step": 3910 }, { "epoch": 0.41, "learning_rate": 5.847643327828283e-05, "loss": 0.7848, "step": 3911 }, { "epoch": 0.41, "learning_rate": 5.8475658746168045e-05, "loss": 0.5231, "step": 3912 }, { "epoch": 0.41, "learning_rate": 5.847488402236178e-05, "loss": 0.5464, "step": 3913 }, { "epoch": 0.41, "learning_rate": 5.8474109106869254e-05, "loss": 0.9604, "step": 3914 }, { "epoch": 0.41, "learning_rate": 5.8473333999695706e-05, "loss": 0.8884, "step": 3915 }, { "epoch": 0.41, "learning_rate": 5.8472558700846314e-05, "loss": 1.249, "step": 3916 }, { "epoch": 0.41, "learning_rate": 5.847178321032633e-05, "loss": 0.5202, "step": 3917 }, { "epoch": 0.41, "learning_rate": 5.847100752814097e-05, "loss": 1.2445, "step": 3918 }, { "epoch": 0.41, "learning_rate": 5.847023165429544e-05, "loss": 1.2847, "step": 3919 }, { "epoch": 0.41, "learning_rate": 5.8469455588794974e-05, "loss": 0.7162, "step": 3920 }, { "epoch": 0.41, "learning_rate": 5.8468679331644794e-05, "loss": 0.968, "step": 3921 }, { "epoch": 0.41, "learning_rate": 5.846790288285013e-05, "loss": 1.3577, "step": 3922 }, { "epoch": 0.41, "learning_rate": 5.84671262424162e-05, "loss": 0.4905, "step": 3923 }, { "epoch": 0.41, "learning_rate": 5.8466349410348235e-05, "loss": 0.8902, "step": 3924 }, { "epoch": 0.41, "learning_rate": 5.846557238665147e-05, "loss": 0.4099, "step": 3925 }, { "epoch": 0.41, "learning_rate": 5.846479517133114e-05, "loss": 0.4463, "step": 3926 }, { "epoch": 0.41, "learning_rate": 5.846401776439246e-05, "loss": 0.6163, "step": 3927 }, { "epoch": 0.41, "learning_rate": 5.8463240165840664e-05, "loss": 0.7774, "step": 3928 }, { "epoch": 0.41, "learning_rate": 5.846246237568101e-05, "loss": 1.2931, "step": 3929 }, { "epoch": 0.41, "learning_rate": 5.8461684393918716e-05, "loss": 1.204, "step": 3930 }, { "epoch": 0.41, "learning_rate": 5.8460906220559015e-05, "loss": 1.0996, "step": 3931 }, { "epoch": 0.41, "learning_rate": 5.846012785560715e-05, "loss": 0.4739, "step": 3932 }, { "epoch": 0.41, "learning_rate": 5.845934929906838e-05, "loss": 0.7959, "step": 3933 }, { "epoch": 0.41, "learning_rate": 5.845857055094791e-05, "loss": 1.0216, "step": 3934 }, { "epoch": 0.41, "learning_rate": 5.845779161125102e-05, "loss": 0.3035, "step": 3935 }, { "epoch": 0.41, "learning_rate": 5.845701247998293e-05, "loss": 0.3426, "step": 3936 }, { "epoch": 0.41, "learning_rate": 5.845623315714888e-05, "loss": 1.138, "step": 3937 }, { "epoch": 0.41, "learning_rate": 5.845545364275414e-05, "loss": 1.4323, "step": 3938 }, { "epoch": 0.41, "learning_rate": 5.845467393680393e-05, "loss": 0.3029, "step": 3939 }, { "epoch": 0.41, "learning_rate": 5.845389403930353e-05, "loss": 1.1489, "step": 3940 }, { "epoch": 0.41, "learning_rate": 5.845311395025816e-05, "loss": 0.3045, "step": 3941 }, { "epoch": 0.41, "learning_rate": 5.845233366967309e-05, "loss": 1.2007, "step": 3942 }, { "epoch": 0.41, "learning_rate": 5.845155319755356e-05, "loss": 1.1549, "step": 3943 }, { "epoch": 0.41, "learning_rate": 5.8450772533904846e-05, "loss": 0.7537, "step": 3944 }, { "epoch": 0.41, "learning_rate": 5.8449991678732176e-05, "loss": 0.7495, "step": 3945 }, { "epoch": 0.41, "learning_rate": 5.8449210632040823e-05, "loss": 0.6356, "step": 3946 }, { "epoch": 0.41, "learning_rate": 5.844842939383603e-05, "loss": 1.2864, "step": 3947 }, { "epoch": 0.41, "learning_rate": 5.844764796412308e-05, "loss": 0.9207, "step": 3948 }, { "epoch": 0.41, "learning_rate": 5.844686634290722e-05, "loss": 1.2499, "step": 3949 }, { "epoch": 0.41, "learning_rate": 5.844608453019371e-05, "loss": 1.2738, "step": 3950 }, { "epoch": 0.41, "learning_rate": 5.8445302525987816e-05, "loss": 0.7078, "step": 3951 }, { "epoch": 0.41, "learning_rate": 5.844452033029479e-05, "loss": 1.0601, "step": 3952 }, { "epoch": 0.41, "learning_rate": 5.8443737943119916e-05, "loss": 0.39, "step": 3953 }, { "epoch": 0.41, "learning_rate": 5.844295536446846e-05, "loss": 1.0911, "step": 3954 }, { "epoch": 0.41, "learning_rate": 5.8442172594345676e-05, "loss": 0.7355, "step": 3955 }, { "epoch": 0.41, "learning_rate": 5.8441389632756836e-05, "loss": 0.875, "step": 3956 }, { "epoch": 0.41, "learning_rate": 5.844060647970723e-05, "loss": 1.1555, "step": 3957 }, { "epoch": 0.41, "learning_rate": 5.8439823135202103e-05, "loss": 1.2862, "step": 3958 }, { "epoch": 0.41, "learning_rate": 5.843903959924675e-05, "loss": 0.7391, "step": 3959 }, { "epoch": 0.41, "learning_rate": 5.843825587184643e-05, "loss": 1.1469, "step": 3960 }, { "epoch": 0.41, "learning_rate": 5.843747195300643e-05, "loss": 0.8624, "step": 3961 }, { "epoch": 0.41, "learning_rate": 5.843668784273202e-05, "loss": 0.5749, "step": 3962 }, { "epoch": 0.41, "learning_rate": 5.843590354102848e-05, "loss": 0.9257, "step": 3963 }, { "epoch": 0.41, "learning_rate": 5.843511904790109e-05, "loss": 0.7529, "step": 3964 }, { "epoch": 0.41, "learning_rate": 5.8434334363355135e-05, "loss": 0.786, "step": 3965 }, { "epoch": 0.41, "learning_rate": 5.84335494873959e-05, "loss": 0.5629, "step": 3966 }, { "epoch": 0.41, "learning_rate": 5.843276442002865e-05, "loss": 1.0193, "step": 3967 }, { "epoch": 0.41, "learning_rate": 5.8431979161258696e-05, "loss": 0.3313, "step": 3968 }, { "epoch": 0.41, "learning_rate": 5.84311937110913e-05, "loss": 0.5396, "step": 3969 }, { "epoch": 0.41, "learning_rate": 5.8430408069531766e-05, "loss": 0.4855, "step": 3970 }, { "epoch": 0.41, "learning_rate": 5.842962223658537e-05, "loss": 0.4472, "step": 3971 }, { "epoch": 0.41, "learning_rate": 5.842883621225742e-05, "loss": 0.498, "step": 3972 }, { "epoch": 0.41, "learning_rate": 5.842804999655319e-05, "loss": 1.6983, "step": 3973 }, { "epoch": 0.42, "learning_rate": 5.8427263589477984e-05, "loss": 0.4809, "step": 3974 }, { "epoch": 0.42, "learning_rate": 5.842647699103708e-05, "loss": 0.3043, "step": 3975 }, { "epoch": 0.42, "learning_rate": 5.84256902012358e-05, "loss": 0.9672, "step": 3976 }, { "epoch": 0.42, "learning_rate": 5.842490322007941e-05, "loss": 0.6068, "step": 3977 }, { "epoch": 0.42, "learning_rate": 5.842411604757323e-05, "loss": 1.4328, "step": 3978 }, { "epoch": 0.42, "learning_rate": 5.842332868372255e-05, "loss": 0.2703, "step": 3979 }, { "epoch": 0.42, "learning_rate": 5.842254112853267e-05, "loss": 0.8584, "step": 3980 }, { "epoch": 0.42, "learning_rate": 5.842175338200889e-05, "loss": 0.8987, "step": 3981 }, { "epoch": 0.42, "learning_rate": 5.842096544415653e-05, "loss": 1.0046, "step": 3982 }, { "epoch": 0.42, "learning_rate": 5.8420177314980874e-05, "loss": 1.5754, "step": 3983 }, { "epoch": 0.42, "learning_rate": 5.8419388994487233e-05, "loss": 1.494, "step": 3984 }, { "epoch": 0.42, "learning_rate": 5.841860048268091e-05, "loss": 0.7739, "step": 3985 }, { "epoch": 0.42, "learning_rate": 5.841781177956722e-05, "loss": 0.4901, "step": 3986 }, { "epoch": 0.42, "learning_rate": 5.8417022885151476e-05, "loss": 0.4624, "step": 3987 }, { "epoch": 0.42, "learning_rate": 5.841623379943898e-05, "loss": 1.3797, "step": 3988 }, { "epoch": 0.42, "learning_rate": 5.8415444522435045e-05, "loss": 0.4505, "step": 3989 }, { "epoch": 0.42, "learning_rate": 5.841465505414498e-05, "loss": 1.0032, "step": 3990 }, { "epoch": 0.42, "learning_rate": 5.8413865394574114e-05, "loss": 0.6162, "step": 3991 }, { "epoch": 0.42, "learning_rate": 5.8413075543727765e-05, "loss": 0.4028, "step": 3992 }, { "epoch": 0.42, "learning_rate": 5.841228550161122e-05, "loss": 0.3345, "step": 3993 }, { "epoch": 0.42, "learning_rate": 5.841149526822982e-05, "loss": 1.653, "step": 3994 }, { "epoch": 0.42, "learning_rate": 5.8410704843588895e-05, "loss": 1.0127, "step": 3995 }, { "epoch": 0.42, "learning_rate": 5.840991422769375e-05, "loss": 1.0784, "step": 3996 }, { "epoch": 0.42, "learning_rate": 5.8409123420549696e-05, "loss": 0.1824, "step": 3997 }, { "epoch": 0.42, "learning_rate": 5.8408332422162084e-05, "loss": 1.4669, "step": 3998 }, { "epoch": 0.42, "learning_rate": 5.840754123253621e-05, "loss": 0.5936, "step": 3999 }, { "epoch": 0.42, "learning_rate": 5.840674985167744e-05, "loss": 0.8601, "step": 4000 }, { "epoch": 0.42, "learning_rate": 5.840595827959105e-05, "loss": 1.8871, "step": 4001 }, { "epoch": 0.42, "learning_rate": 5.8405166516282405e-05, "loss": 0.2496, "step": 4002 }, { "epoch": 0.42, "learning_rate": 5.8404374561756826e-05, "loss": 0.6983, "step": 4003 }, { "epoch": 0.42, "learning_rate": 5.840358241601965e-05, "loss": 1.2423, "step": 4004 }, { "epoch": 0.42, "learning_rate": 5.840279007907619e-05, "loss": 0.2963, "step": 4005 }, { "epoch": 0.42, "learning_rate": 5.8401997550931804e-05, "loss": 1.1234, "step": 4006 }, { "epoch": 0.42, "learning_rate": 5.840120483159182e-05, "loss": 0.328, "step": 4007 }, { "epoch": 0.42, "learning_rate": 5.840041192106156e-05, "loss": 0.4476, "step": 4008 }, { "epoch": 0.42, "learning_rate": 5.839961881934638e-05, "loss": 1.0333, "step": 4009 }, { "epoch": 0.42, "learning_rate": 5.83988255264516e-05, "loss": 1.0609, "step": 4010 }, { "epoch": 0.42, "learning_rate": 5.839803204238258e-05, "loss": 1.2776, "step": 4011 }, { "epoch": 0.42, "learning_rate": 5.839723836714465e-05, "loss": 1.2799, "step": 4012 }, { "epoch": 0.42, "learning_rate": 5.839644450074315e-05, "loss": 0.3314, "step": 4013 }, { "epoch": 0.42, "learning_rate": 5.839565044318344e-05, "loss": 0.7606, "step": 4014 }, { "epoch": 0.42, "learning_rate": 5.839485619447085e-05, "loss": 1.0587, "step": 4015 }, { "epoch": 0.42, "learning_rate": 5.839406175461074e-05, "loss": 1.2002, "step": 4016 }, { "epoch": 0.42, "learning_rate": 5.839326712360844e-05, "loss": 1.3735, "step": 4017 }, { "epoch": 0.42, "learning_rate": 5.8392472301469316e-05, "loss": 0.3514, "step": 4018 }, { "epoch": 0.42, "learning_rate": 5.83916772881987e-05, "loss": 1.5793, "step": 4019 }, { "epoch": 0.42, "learning_rate": 5.8390882083801965e-05, "loss": 1.7042, "step": 4020 }, { "epoch": 0.42, "learning_rate": 5.8390086688284455e-05, "loss": 1.2906, "step": 4021 }, { "epoch": 0.42, "learning_rate": 5.8389291101651525e-05, "loss": 0.8871, "step": 4022 }, { "epoch": 0.42, "learning_rate": 5.838849532390852e-05, "loss": 0.861, "step": 4023 }, { "epoch": 0.42, "learning_rate": 5.838769935506082e-05, "loss": 1.5161, "step": 4024 }, { "epoch": 0.42, "learning_rate": 5.8386903195113765e-05, "loss": 0.469, "step": 4025 }, { "epoch": 0.42, "learning_rate": 5.838610684407271e-05, "loss": 1.0671, "step": 4026 }, { "epoch": 0.42, "learning_rate": 5.8385310301943047e-05, "loss": 1.3443, "step": 4027 }, { "epoch": 0.42, "learning_rate": 5.83845135687301e-05, "loss": 1.1617, "step": 4028 }, { "epoch": 0.42, "learning_rate": 5.838371664443926e-05, "loss": 0.6682, "step": 4029 }, { "epoch": 0.42, "learning_rate": 5.838291952907587e-05, "loss": 1.4204, "step": 4030 }, { "epoch": 0.42, "learning_rate": 5.838212222264531e-05, "loss": 0.6927, "step": 4031 }, { "epoch": 0.42, "learning_rate": 5.8381324725152945e-05, "loss": 0.5749, "step": 4032 }, { "epoch": 0.42, "learning_rate": 5.8380527036604144e-05, "loss": 0.9227, "step": 4033 }, { "epoch": 0.42, "learning_rate": 5.837972915700428e-05, "loss": 0.3506, "step": 4034 }, { "epoch": 0.42, "learning_rate": 5.837893108635871e-05, "loss": 1.3373, "step": 4035 }, { "epoch": 0.42, "learning_rate": 5.837813282467282e-05, "loss": 1.202, "step": 4036 }, { "epoch": 0.42, "learning_rate": 5.837733437195198e-05, "loss": 0.2849, "step": 4037 }, { "epoch": 0.42, "learning_rate": 5.8376535728201565e-05, "loss": 0.7425, "step": 4038 }, { "epoch": 0.42, "learning_rate": 5.837573689342695e-05, "loss": 0.3697, "step": 4039 }, { "epoch": 0.42, "learning_rate": 5.837493786763352e-05, "loss": 0.5055, "step": 4040 }, { "epoch": 0.42, "learning_rate": 5.837413865082664e-05, "loss": 1.4215, "step": 4041 }, { "epoch": 0.42, "learning_rate": 5.83733392430117e-05, "loss": 0.3425, "step": 4042 }, { "epoch": 0.42, "learning_rate": 5.837253964419407e-05, "loss": 1.0711, "step": 4043 }, { "epoch": 0.42, "learning_rate": 5.837173985437916e-05, "loss": 1.3856, "step": 4044 }, { "epoch": 0.42, "learning_rate": 5.8370939873572326e-05, "loss": 0.7811, "step": 4045 }, { "epoch": 0.42, "learning_rate": 5.837013970177896e-05, "loss": 1.3714, "step": 4046 }, { "epoch": 0.42, "learning_rate": 5.8369339339004456e-05, "loss": 1.2869, "step": 4047 }, { "epoch": 0.42, "learning_rate": 5.836853878525419e-05, "loss": 0.8449, "step": 4048 }, { "epoch": 0.42, "learning_rate": 5.8367738040533567e-05, "loss": 1.3115, "step": 4049 }, { "epoch": 0.42, "learning_rate": 5.836693710484797e-05, "loss": 0.925, "step": 4050 }, { "epoch": 0.42, "learning_rate": 5.836613597820279e-05, "loss": 0.7537, "step": 4051 }, { "epoch": 0.42, "learning_rate": 5.8365334660603416e-05, "loss": 0.7318, "step": 4052 }, { "epoch": 0.42, "learning_rate": 5.8364533152055245e-05, "loss": 0.9427, "step": 4053 }, { "epoch": 0.42, "learning_rate": 5.8363731452563676e-05, "loss": 0.3285, "step": 4054 }, { "epoch": 0.42, "learning_rate": 5.83629295621341e-05, "loss": 0.8005, "step": 4055 }, { "epoch": 0.42, "learning_rate": 5.836212748077192e-05, "loss": 0.6148, "step": 4056 }, { "epoch": 0.42, "learning_rate": 5.836132520848253e-05, "loss": 0.3447, "step": 4057 }, { "epoch": 0.42, "learning_rate": 5.836052274527134e-05, "loss": 1.5639, "step": 4058 }, { "epoch": 0.42, "learning_rate": 5.835972009114376e-05, "loss": 0.5249, "step": 4059 }, { "epoch": 0.42, "learning_rate": 5.835891724610516e-05, "loss": 1.523, "step": 4060 }, { "epoch": 0.42, "learning_rate": 5.8358114210160974e-05, "loss": 0.737, "step": 4061 }, { "epoch": 0.42, "learning_rate": 5.8357310983316596e-05, "loss": 1.1243, "step": 4062 }, { "epoch": 0.42, "learning_rate": 5.835650756557743e-05, "loss": 0.7091, "step": 4063 }, { "epoch": 0.42, "learning_rate": 5.83557039569489e-05, "loss": 1.3702, "step": 4064 }, { "epoch": 0.42, "learning_rate": 5.83549001574364e-05, "loss": 1.0361, "step": 4065 }, { "epoch": 0.42, "learning_rate": 5.8354096167045346e-05, "loss": 3.0423, "step": 4066 }, { "epoch": 0.42, "learning_rate": 5.835329198578115e-05, "loss": 1.0531, "step": 4067 }, { "epoch": 0.42, "learning_rate": 5.8352487613649226e-05, "loss": 1.297, "step": 4068 }, { "epoch": 0.42, "learning_rate": 5.8351683050654995e-05, "loss": 0.4306, "step": 4069 }, { "epoch": 0.43, "learning_rate": 5.8350878296803865e-05, "loss": 0.7986, "step": 4070 }, { "epoch": 0.43, "learning_rate": 5.835007335210126e-05, "loss": 0.7146, "step": 4071 }, { "epoch": 0.43, "learning_rate": 5.8349268216552585e-05, "loss": 0.2128, "step": 4072 }, { "epoch": 0.43, "learning_rate": 5.834846289016328e-05, "loss": 0.71, "step": 4073 }, { "epoch": 0.43, "learning_rate": 5.834765737293874e-05, "loss": 0.3926, "step": 4074 }, { "epoch": 0.43, "learning_rate": 5.8346851664884415e-05, "loss": 1.524, "step": 4075 }, { "epoch": 0.43, "learning_rate": 5.834604576600572e-05, "loss": 0.1857, "step": 4076 }, { "epoch": 0.43, "learning_rate": 5.834523967630807e-05, "loss": 0.4056, "step": 4077 }, { "epoch": 0.43, "learning_rate": 5.83444333957969e-05, "loss": 1.288, "step": 4078 }, { "epoch": 0.43, "learning_rate": 5.834362692447764e-05, "loss": 0.6353, "step": 4079 }, { "epoch": 0.43, "learning_rate": 5.834282026235572e-05, "loss": 1.2786, "step": 4080 }, { "epoch": 0.43, "learning_rate": 5.834201340943656e-05, "loss": 1.8558, "step": 4081 }, { "epoch": 0.43, "learning_rate": 5.834120636572559e-05, "loss": 1.2884, "step": 4082 }, { "epoch": 0.43, "learning_rate": 5.834039913122826e-05, "loss": 1.6704, "step": 4083 }, { "epoch": 0.43, "learning_rate": 5.833959170594999e-05, "loss": 0.4931, "step": 4084 }, { "epoch": 0.43, "learning_rate": 5.8338784089896224e-05, "loss": 1.1288, "step": 4085 }, { "epoch": 0.43, "learning_rate": 5.833797628307239e-05, "loss": 0.9106, "step": 4086 }, { "epoch": 0.43, "learning_rate": 5.8337168285483924e-05, "loss": 0.9402, "step": 4087 }, { "epoch": 0.43, "learning_rate": 5.833636009713628e-05, "loss": 0.1619, "step": 4088 }, { "epoch": 0.43, "learning_rate": 5.833555171803489e-05, "loss": 1.0897, "step": 4089 }, { "epoch": 0.43, "learning_rate": 5.83347431481852e-05, "loss": 0.6779, "step": 4090 }, { "epoch": 0.43, "learning_rate": 5.833393438759264e-05, "loss": 0.9414, "step": 4091 }, { "epoch": 0.43, "learning_rate": 5.833312543626266e-05, "loss": 1.0842, "step": 4092 }, { "epoch": 0.43, "learning_rate": 5.833231629420072e-05, "loss": 1.0206, "step": 4093 }, { "epoch": 0.43, "learning_rate": 5.833150696141226e-05, "loss": 0.5304, "step": 4094 }, { "epoch": 0.43, "learning_rate": 5.833069743790271e-05, "loss": 1.43, "step": 4095 }, { "epoch": 0.43, "learning_rate": 5.8329887723677537e-05, "loss": 0.8662, "step": 4096 }, { "epoch": 0.43, "learning_rate": 5.832907781874219e-05, "loss": 1.0441, "step": 4097 }, { "epoch": 0.43, "learning_rate": 5.8328267723102125e-05, "loss": 1.1702, "step": 4098 }, { "epoch": 0.43, "learning_rate": 5.8327457436762784e-05, "loss": 1.1745, "step": 4099 }, { "epoch": 0.43, "learning_rate": 5.832664695972963e-05, "loss": 0.6297, "step": 4100 }, { "epoch": 0.43, "learning_rate": 5.832583629200812e-05, "loss": 1.0434, "step": 4101 }, { "epoch": 0.43, "learning_rate": 5.83250254336037e-05, "loss": 1.197, "step": 4102 }, { "epoch": 0.43, "learning_rate": 5.832421438452184e-05, "loss": 1.0704, "step": 4103 }, { "epoch": 0.43, "learning_rate": 5.832340314476799e-05, "loss": 0.9941, "step": 4104 }, { "epoch": 0.43, "learning_rate": 5.8322591714347626e-05, "loss": 0.7657, "step": 4105 }, { "epoch": 0.43, "learning_rate": 5.83217800932662e-05, "loss": 0.294, "step": 4106 }, { "epoch": 0.43, "learning_rate": 5.832096828152918e-05, "loss": 0.5353, "step": 4107 }, { "epoch": 0.43, "learning_rate": 5.832015627914202e-05, "loss": 0.4422, "step": 4108 }, { "epoch": 0.43, "learning_rate": 5.83193440861102e-05, "loss": 1.2178, "step": 4109 }, { "epoch": 0.43, "learning_rate": 5.831853170243918e-05, "loss": 1.0307, "step": 4110 }, { "epoch": 0.43, "learning_rate": 5.8317719128134426e-05, "loss": 0.5284, "step": 4111 }, { "epoch": 0.43, "learning_rate": 5.831690636320142e-05, "loss": 1.3658, "step": 4112 }, { "epoch": 0.43, "learning_rate": 5.831609340764563e-05, "loss": 1.0023, "step": 4113 }, { "epoch": 0.43, "learning_rate": 5.831528026147252e-05, "loss": 0.6807, "step": 4114 }, { "epoch": 0.43, "learning_rate": 5.831446692468756e-05, "loss": 0.4918, "step": 4115 }, { "epoch": 0.43, "learning_rate": 5.831365339729624e-05, "loss": 1.4562, "step": 4116 }, { "epoch": 0.43, "learning_rate": 5.831283967930403e-05, "loss": 1.4441, "step": 4117 }, { "epoch": 0.43, "learning_rate": 5.8312025770716405e-05, "loss": 0.5561, "step": 4118 }, { "epoch": 0.43, "learning_rate": 5.831121167153885e-05, "loss": 1.4393, "step": 4119 }, { "epoch": 0.43, "learning_rate": 5.8310397381776847e-05, "loss": 0.3386, "step": 4120 }, { "epoch": 0.43, "learning_rate": 5.830958290143587e-05, "loss": 0.9003, "step": 4121 }, { "epoch": 0.43, "learning_rate": 5.83087682305214e-05, "loss": 1.3904, "step": 4122 }, { "epoch": 0.43, "learning_rate": 5.830795336903893e-05, "loss": 1.2678, "step": 4123 }, { "epoch": 0.43, "learning_rate": 5.830713831699394e-05, "loss": 1.3405, "step": 4124 }, { "epoch": 0.43, "learning_rate": 5.830632307439192e-05, "loss": 1.8147, "step": 4125 }, { "epoch": 0.43, "learning_rate": 5.8305507641238356e-05, "loss": 0.8777, "step": 4126 }, { "epoch": 0.43, "learning_rate": 5.8304692017538734e-05, "loss": 1.5712, "step": 4127 }, { "epoch": 0.43, "learning_rate": 5.830387620329855e-05, "loss": 1.3453, "step": 4128 }, { "epoch": 0.43, "learning_rate": 5.83030601985233e-05, "loss": 0.6142, "step": 4129 }, { "epoch": 0.43, "learning_rate": 5.830224400321846e-05, "loss": 1.5888, "step": 4130 }, { "epoch": 0.43, "learning_rate": 5.8301427617389546e-05, "loss": 1.1995, "step": 4131 }, { "epoch": 0.43, "learning_rate": 5.830061104104203e-05, "loss": 0.8274, "step": 4132 }, { "epoch": 0.43, "learning_rate": 5.829979427418143e-05, "loss": 0.6451, "step": 4133 }, { "epoch": 0.43, "learning_rate": 5.829897731681325e-05, "loss": 0.8463, "step": 4134 }, { "epoch": 0.43, "learning_rate": 5.8298160168942965e-05, "loss": 1.4882, "step": 4135 }, { "epoch": 0.43, "learning_rate": 5.829734283057608e-05, "loss": 1.037, "step": 4136 }, { "epoch": 0.43, "learning_rate": 5.829652530171811e-05, "loss": 0.5207, "step": 4137 }, { "epoch": 0.43, "learning_rate": 5.829570758237455e-05, "loss": 0.8993, "step": 4138 }, { "epoch": 0.43, "learning_rate": 5.8294889672550916e-05, "loss": 1.5736, "step": 4139 }, { "epoch": 0.43, "learning_rate": 5.8294071572252697e-05, "loss": 1.4845, "step": 4140 }, { "epoch": 0.43, "learning_rate": 5.829325328148541e-05, "loss": 1.321, "step": 4141 }, { "epoch": 0.43, "learning_rate": 5.829243480025456e-05, "loss": 0.7226, "step": 4142 }, { "epoch": 0.43, "learning_rate": 5.829161612856566e-05, "loss": 0.348, "step": 4143 }, { "epoch": 0.43, "learning_rate": 5.8290797266424225e-05, "loss": 1.7727, "step": 4144 }, { "epoch": 0.43, "learning_rate": 5.828997821383576e-05, "loss": 2.0668, "step": 4145 }, { "epoch": 0.43, "learning_rate": 5.8289158970805775e-05, "loss": 1.0448, "step": 4146 }, { "epoch": 0.43, "learning_rate": 5.8288339537339794e-05, "loss": 0.9133, "step": 4147 }, { "epoch": 0.43, "learning_rate": 5.828751991344333e-05, "loss": 1.3464, "step": 4148 }, { "epoch": 0.43, "learning_rate": 5.828670009912191e-05, "loss": 1.1768, "step": 4149 }, { "epoch": 0.43, "learning_rate": 5.8285880094381035e-05, "loss": 0.4628, "step": 4150 }, { "epoch": 0.43, "learning_rate": 5.8285059899226226e-05, "loss": 0.7618, "step": 4151 }, { "epoch": 0.43, "learning_rate": 5.8284239513663026e-05, "loss": 1.555, "step": 4152 }, { "epoch": 0.43, "learning_rate": 5.8283418937696936e-05, "loss": 0.8842, "step": 4153 }, { "epoch": 0.43, "learning_rate": 5.828259817133349e-05, "loss": 0.4436, "step": 4154 }, { "epoch": 0.43, "learning_rate": 5.828177721457821e-05, "loss": 0.3055, "step": 4155 }, { "epoch": 0.43, "learning_rate": 5.828095606743662e-05, "loss": 1.3796, "step": 4156 }, { "epoch": 0.43, "learning_rate": 5.8280134729914255e-05, "loss": 0.2556, "step": 4157 }, { "epoch": 0.43, "learning_rate": 5.827931320201664e-05, "loss": 1.0409, "step": 4158 }, { "epoch": 0.43, "learning_rate": 5.82784914837493e-05, "loss": 0.7655, "step": 4159 }, { "epoch": 0.43, "learning_rate": 5.827766957511778e-05, "loss": 0.7976, "step": 4160 }, { "epoch": 0.43, "learning_rate": 5.82768474761276e-05, "loss": 0.4907, "step": 4161 }, { "epoch": 0.43, "learning_rate": 5.82760251867843e-05, "loss": 0.8778, "step": 4162 }, { "epoch": 0.43, "learning_rate": 5.8275202707093414e-05, "loss": 1.2224, "step": 4163 }, { "epoch": 0.43, "learning_rate": 5.827438003706049e-05, "loss": 0.5192, "step": 4164 }, { "epoch": 0.43, "learning_rate": 5.827355717669104e-05, "loss": 0.7867, "step": 4165 }, { "epoch": 0.44, "learning_rate": 5.827273412599063e-05, "loss": 0.3005, "step": 4166 }, { "epoch": 0.44, "learning_rate": 5.827191088496477e-05, "loss": 0.4216, "step": 4167 }, { "epoch": 0.44, "learning_rate": 5.827108745361904e-05, "loss": 1.1774, "step": 4168 }, { "epoch": 0.44, "learning_rate": 5.827026383195897e-05, "loss": 0.4658, "step": 4169 }, { "epoch": 0.44, "learning_rate": 5.826944001999008e-05, "loss": 0.1419, "step": 4170 }, { "epoch": 0.44, "learning_rate": 5.8268616017717934e-05, "loss": 0.8917, "step": 4171 }, { "epoch": 0.44, "learning_rate": 5.826779182514809e-05, "loss": 0.5046, "step": 4172 }, { "epoch": 0.44, "learning_rate": 5.8266967442286086e-05, "loss": 1.1423, "step": 4173 }, { "epoch": 0.44, "learning_rate": 5.826614286913747e-05, "loss": 0.573, "step": 4174 }, { "epoch": 0.44, "learning_rate": 5.826531810570779e-05, "loss": 0.6035, "step": 4175 }, { "epoch": 0.44, "learning_rate": 5.82644931520026e-05, "loss": 0.7402, "step": 4176 }, { "epoch": 0.44, "learning_rate": 5.826366800802746e-05, "loss": 0.9664, "step": 4177 }, { "epoch": 0.44, "learning_rate": 5.826284267378791e-05, "loss": 0.6707, "step": 4178 }, { "epoch": 0.44, "learning_rate": 5.826201714928953e-05, "loss": 1.2193, "step": 4179 }, { "epoch": 0.44, "learning_rate": 5.826119143453786e-05, "loss": 0.5238, "step": 4180 }, { "epoch": 0.44, "learning_rate": 5.8260365529538456e-05, "loss": 1.5166, "step": 4181 }, { "epoch": 0.44, "learning_rate": 5.825953943429689e-05, "loss": 0.4025, "step": 4182 }, { "epoch": 0.44, "learning_rate": 5.8258713148818716e-05, "loss": 0.7688, "step": 4183 }, { "epoch": 0.44, "learning_rate": 5.8257886673109495e-05, "loss": 0.7314, "step": 4184 }, { "epoch": 0.44, "learning_rate": 5.8257060007174795e-05, "loss": 0.941, "step": 4185 }, { "epoch": 0.44, "learning_rate": 5.825623315102018e-05, "loss": 0.4438, "step": 4186 }, { "epoch": 0.44, "learning_rate": 5.8255406104651206e-05, "loss": 1.5518, "step": 4187 }, { "epoch": 0.44, "learning_rate": 5.825457886807346e-05, "loss": 1.178, "step": 4188 }, { "epoch": 0.44, "learning_rate": 5.82537514412925e-05, "loss": 1.6841, "step": 4189 }, { "epoch": 0.44, "learning_rate": 5.825292382431388e-05, "loss": 1.1573, "step": 4190 }, { "epoch": 0.44, "learning_rate": 5.82520960171432e-05, "loss": 0.6179, "step": 4191 }, { "epoch": 0.44, "learning_rate": 5.825126801978602e-05, "loss": 1.893, "step": 4192 }, { "epoch": 0.44, "learning_rate": 5.825043983224792e-05, "loss": 1.3253, "step": 4193 }, { "epoch": 0.44, "learning_rate": 5.8249611454534456e-05, "loss": 1.4745, "step": 4194 }, { "epoch": 0.44, "learning_rate": 5.824878288665122e-05, "loss": 0.8024, "step": 4195 }, { "epoch": 0.44, "learning_rate": 5.8247954128603796e-05, "loss": 0.3947, "step": 4196 }, { "epoch": 0.44, "learning_rate": 5.824712518039775e-05, "loss": 0.9889, "step": 4197 }, { "epoch": 0.44, "learning_rate": 5.8246296042038666e-05, "loss": 0.5464, "step": 4198 }, { "epoch": 0.44, "learning_rate": 5.824546671353213e-05, "loss": 0.2527, "step": 4199 }, { "epoch": 0.44, "learning_rate": 5.824463719488371e-05, "loss": 0.9048, "step": 4200 }, { "epoch": 0.44, "learning_rate": 5.8243807486099014e-05, "loss": 1.0999, "step": 4201 }, { "epoch": 0.44, "learning_rate": 5.82429775871836e-05, "loss": 0.5872, "step": 4202 }, { "epoch": 0.44, "learning_rate": 5.824214749814308e-05, "loss": 1.3097, "step": 4203 }, { "epoch": 0.44, "learning_rate": 5.824131721898303e-05, "loss": 1.3712, "step": 4204 }, { "epoch": 0.44, "learning_rate": 5.824048674970904e-05, "loss": 0.3649, "step": 4205 }, { "epoch": 0.44, "learning_rate": 5.82396560903267e-05, "loss": 1.4086, "step": 4206 }, { "epoch": 0.44, "learning_rate": 5.8238825240841605e-05, "loss": 1.0889, "step": 4207 }, { "epoch": 0.44, "learning_rate": 5.823799420125935e-05, "loss": 1.542, "step": 4208 }, { "epoch": 0.44, "learning_rate": 5.823716297158551e-05, "loss": 0.5847, "step": 4209 }, { "epoch": 0.44, "learning_rate": 5.823633155182571e-05, "loss": 0.5759, "step": 4210 }, { "epoch": 0.44, "learning_rate": 5.823549994198553e-05, "loss": 1.1728, "step": 4211 }, { "epoch": 0.44, "learning_rate": 5.8234668142070565e-05, "loss": 0.883, "step": 4212 }, { "epoch": 0.44, "learning_rate": 5.823383615208643e-05, "loss": 0.6218, "step": 4213 }, { "epoch": 0.44, "learning_rate": 5.82330039720387e-05, "loss": 1.3044, "step": 4214 }, { "epoch": 0.44, "learning_rate": 5.823217160193301e-05, "loss": 1.5643, "step": 4215 }, { "epoch": 0.44, "learning_rate": 5.823133904177494e-05, "loss": 0.7725, "step": 4216 }, { "epoch": 0.44, "learning_rate": 5.823050629157011e-05, "loss": 1.5963, "step": 4217 }, { "epoch": 0.44, "learning_rate": 5.82296733513241e-05, "loss": 0.9638, "step": 4218 }, { "epoch": 0.44, "learning_rate": 5.822884022104254e-05, "loss": 0.4708, "step": 4219 }, { "epoch": 0.44, "learning_rate": 5.8228006900731037e-05, "loss": 0.6455, "step": 4220 }, { "epoch": 0.44, "learning_rate": 5.822717339039519e-05, "loss": 0.4589, "step": 4221 }, { "epoch": 0.44, "learning_rate": 5.8226339690040624e-05, "loss": 0.5276, "step": 4222 }, { "epoch": 0.44, "learning_rate": 5.8225505799672944e-05, "loss": 0.7611, "step": 4223 }, { "epoch": 0.44, "learning_rate": 5.8224671719297755e-05, "loss": 0.6444, "step": 4224 }, { "epoch": 0.44, "learning_rate": 5.82238374489207e-05, "loss": 0.1326, "step": 4225 }, { "epoch": 0.44, "learning_rate": 5.8223002988547345e-05, "loss": 1.249, "step": 4226 }, { "epoch": 0.44, "learning_rate": 5.822216833818336e-05, "loss": 1.2575, "step": 4227 }, { "epoch": 0.44, "learning_rate": 5.8221333497834326e-05, "loss": 1.0046, "step": 4228 }, { "epoch": 0.44, "learning_rate": 5.822049846750589e-05, "loss": 1.8154, "step": 4229 }, { "epoch": 0.44, "learning_rate": 5.821966324720366e-05, "loss": 1.7803, "step": 4230 }, { "epoch": 0.44, "learning_rate": 5.821882783693326e-05, "loss": 0.6032, "step": 4231 }, { "epoch": 0.44, "learning_rate": 5.821799223670031e-05, "loss": 0.2002, "step": 4232 }, { "epoch": 0.44, "learning_rate": 5.821715644651043e-05, "loss": 0.3273, "step": 4233 }, { "epoch": 0.44, "learning_rate": 5.8216320466369275e-05, "loss": 1.0707, "step": 4234 }, { "epoch": 0.44, "learning_rate": 5.821548429628244e-05, "loss": 0.4958, "step": 4235 }, { "epoch": 0.44, "learning_rate": 5.821464793625557e-05, "loss": 0.926, "step": 4236 }, { "epoch": 0.44, "learning_rate": 5.8213811386294293e-05, "loss": 0.3536, "step": 4237 }, { "epoch": 0.44, "learning_rate": 5.821297464640424e-05, "loss": 0.3747, "step": 4238 }, { "epoch": 0.44, "learning_rate": 5.821213771659104e-05, "loss": 1.1501, "step": 4239 }, { "epoch": 0.44, "learning_rate": 5.8211300596860326e-05, "loss": 0.6477, "step": 4240 }, { "epoch": 0.44, "learning_rate": 5.821046328721773e-05, "loss": 1.0812, "step": 4241 }, { "epoch": 0.44, "learning_rate": 5.8209625787668915e-05, "loss": 0.9188, "step": 4242 }, { "epoch": 0.44, "learning_rate": 5.82087880982195e-05, "loss": 1.387, "step": 4243 }, { "epoch": 0.44, "learning_rate": 5.820795021887511e-05, "loss": 0.2624, "step": 4244 }, { "epoch": 0.44, "learning_rate": 5.820711214964141e-05, "loss": 1.6482, "step": 4245 }, { "epoch": 0.44, "learning_rate": 5.820627389052402e-05, "loss": 1.2605, "step": 4246 }, { "epoch": 0.44, "learning_rate": 5.82054354415286e-05, "loss": 0.3497, "step": 4247 }, { "epoch": 0.44, "learning_rate": 5.820459680266079e-05, "loss": 0.3118, "step": 4248 }, { "epoch": 0.44, "learning_rate": 5.820375797392623e-05, "loss": 0.9545, "step": 4249 }, { "epoch": 0.44, "learning_rate": 5.8202918955330576e-05, "loss": 0.1544, "step": 4250 }, { "epoch": 0.44, "learning_rate": 5.820207974687946e-05, "loss": 1.2277, "step": 4251 }, { "epoch": 0.44, "learning_rate": 5.820124034857855e-05, "loss": 0.4197, "step": 4252 }, { "epoch": 0.44, "learning_rate": 5.820040076043349e-05, "loss": 1.0235, "step": 4253 }, { "epoch": 0.44, "learning_rate": 5.819956098244993e-05, "loss": 1.0562, "step": 4254 }, { "epoch": 0.44, "learning_rate": 5.819872101463352e-05, "loss": 0.726, "step": 4255 }, { "epoch": 0.44, "learning_rate": 5.8197880856989914e-05, "loss": 1.1167, "step": 4256 }, { "epoch": 0.44, "learning_rate": 5.819704050952478e-05, "loss": 1.2819, "step": 4257 }, { "epoch": 0.44, "learning_rate": 5.8196199972243764e-05, "loss": 0.5327, "step": 4258 }, { "epoch": 0.44, "learning_rate": 5.819535924515253e-05, "loss": 1.1187, "step": 4259 }, { "epoch": 0.44, "learning_rate": 5.819451832825673e-05, "loss": 0.5266, "step": 4260 }, { "epoch": 0.45, "learning_rate": 5.8193677221562035e-05, "loss": 0.4646, "step": 4261 }, { "epoch": 0.45, "learning_rate": 5.8192835925074096e-05, "loss": 1.35, "step": 4262 }, { "epoch": 0.45, "learning_rate": 5.819199443879858e-05, "loss": 1.3493, "step": 4263 }, { "epoch": 0.45, "learning_rate": 5.819115276274116e-05, "loss": 1.2636, "step": 4264 }, { "epoch": 0.45, "learning_rate": 5.81903108969075e-05, "loss": 1.161, "step": 4265 }, { "epoch": 0.45, "learning_rate": 5.818946884130325e-05, "loss": 1.2086, "step": 4266 }, { "epoch": 0.45, "learning_rate": 5.81886265959341e-05, "loss": 0.4003, "step": 4267 }, { "epoch": 0.45, "learning_rate": 5.818778416080571e-05, "loss": 0.3419, "step": 4268 }, { "epoch": 0.45, "learning_rate": 5.818694153592376e-05, "loss": 0.5057, "step": 4269 }, { "epoch": 0.45, "learning_rate": 5.8186098721293904e-05, "loss": 1.285, "step": 4270 }, { "epoch": 0.45, "learning_rate": 5.818525571692183e-05, "loss": 1.1492, "step": 4271 }, { "epoch": 0.45, "learning_rate": 5.818441252281321e-05, "loss": 0.6733, "step": 4272 }, { "epoch": 0.45, "learning_rate": 5.818356913897372e-05, "loss": 1.349, "step": 4273 }, { "epoch": 0.45, "learning_rate": 5.818272556540904e-05, "loss": 1.0454, "step": 4274 }, { "epoch": 0.45, "learning_rate": 5.818188180212483e-05, "loss": 1.5851, "step": 4275 }, { "epoch": 0.45, "learning_rate": 5.8181037849126805e-05, "loss": 1.1847, "step": 4276 }, { "epoch": 0.45, "learning_rate": 5.8180193706420616e-05, "loss": 0.9211, "step": 4277 }, { "epoch": 0.45, "learning_rate": 5.8179349374011966e-05, "loss": 0.2586, "step": 4278 }, { "epoch": 0.45, "learning_rate": 5.817850485190652e-05, "loss": 0.8863, "step": 4279 }, { "epoch": 0.45, "learning_rate": 5.817766014010997e-05, "loss": 0.9452, "step": 4280 }, { "epoch": 0.45, "learning_rate": 5.817681523862802e-05, "loss": 0.7152, "step": 4281 }, { "epoch": 0.45, "learning_rate": 5.817597014746633e-05, "loss": 0.5143, "step": 4282 }, { "epoch": 0.45, "learning_rate": 5.81751248666306e-05, "loss": 0.4688, "step": 4283 }, { "epoch": 0.45, "learning_rate": 5.817427939612653e-05, "loss": 0.7837, "step": 4284 }, { "epoch": 0.45, "learning_rate": 5.81734337359598e-05, "loss": 1.0192, "step": 4285 }, { "epoch": 0.45, "learning_rate": 5.8172587886136105e-05, "loss": 0.3322, "step": 4286 }, { "epoch": 0.45, "learning_rate": 5.817174184666115e-05, "loss": 1.6337, "step": 4287 }, { "epoch": 0.45, "learning_rate": 5.817089561754061e-05, "loss": 1.7734, "step": 4288 }, { "epoch": 0.45, "learning_rate": 5.8170049198780196e-05, "loss": 1.2755, "step": 4289 }, { "epoch": 0.45, "learning_rate": 5.81692025903856e-05, "loss": 1.0876, "step": 4290 }, { "epoch": 0.45, "learning_rate": 5.816835579236252e-05, "loss": 0.6077, "step": 4291 }, { "epoch": 0.45, "learning_rate": 5.816750880471666e-05, "loss": 0.9997, "step": 4292 }, { "epoch": 0.45, "learning_rate": 5.816666162745374e-05, "loss": 0.632, "step": 4293 }, { "epoch": 0.45, "learning_rate": 5.8165814260579416e-05, "loss": 1.221, "step": 4294 }, { "epoch": 0.45, "learning_rate": 5.816496670409944e-05, "loss": 1.7382, "step": 4295 }, { "epoch": 0.45, "learning_rate": 5.8164118958019495e-05, "loss": 1.684, "step": 4296 }, { "epoch": 0.45, "learning_rate": 5.8163271022345285e-05, "loss": 1.4086, "step": 4297 }, { "epoch": 0.45, "learning_rate": 5.816242289708252e-05, "loss": 0.7509, "step": 4298 }, { "epoch": 0.45, "learning_rate": 5.816157458223692e-05, "loss": 0.5388, "step": 4299 }, { "epoch": 0.45, "learning_rate": 5.81607260778142e-05, "loss": 0.7457, "step": 4300 }, { "epoch": 0.45, "learning_rate": 5.8159877383820047e-05, "loss": 1.434, "step": 4301 }, { "epoch": 0.45, "learning_rate": 5.815902850026019e-05, "loss": 1.0669, "step": 4302 }, { "epoch": 0.45, "learning_rate": 5.815817942714035e-05, "loss": 0.5164, "step": 4303 }, { "epoch": 0.45, "learning_rate": 5.815733016446623e-05, "loss": 0.6569, "step": 4304 }, { "epoch": 0.45, "learning_rate": 5.815648071224355e-05, "loss": 1.0118, "step": 4305 }, { "epoch": 0.45, "learning_rate": 5.815563107047803e-05, "loss": 1.3056, "step": 4306 }, { "epoch": 0.45, "learning_rate": 5.815478123917538e-05, "loss": 0.6424, "step": 4307 }, { "epoch": 0.45, "learning_rate": 5.815393121834134e-05, "loss": 0.4321, "step": 4308 }, { "epoch": 0.45, "learning_rate": 5.815308100798163e-05, "loss": 0.5118, "step": 4309 }, { "epoch": 0.45, "learning_rate": 5.815223060810196e-05, "loss": 0.7965, "step": 4310 }, { "epoch": 0.45, "learning_rate": 5.815138001870806e-05, "loss": 0.1967, "step": 4311 }, { "epoch": 0.45, "learning_rate": 5.815052923980565e-05, "loss": 1.4654, "step": 4312 }, { "epoch": 0.45, "learning_rate": 5.814967827140047e-05, "loss": 0.3475, "step": 4313 }, { "epoch": 0.45, "learning_rate": 5.814882711349824e-05, "loss": 1.4296, "step": 4314 }, { "epoch": 0.45, "learning_rate": 5.81479757661047e-05, "loss": 0.7026, "step": 4315 }, { "epoch": 0.45, "learning_rate": 5.814712422922557e-05, "loss": 0.9337, "step": 4316 }, { "epoch": 0.45, "learning_rate": 5.814627250286659e-05, "loss": 0.9942, "step": 4317 }, { "epoch": 0.45, "learning_rate": 5.814542058703348e-05, "loss": 1.0771, "step": 4318 }, { "epoch": 0.45, "learning_rate": 5.814456848173198e-05, "loss": 0.0942, "step": 4319 }, { "epoch": 0.45, "learning_rate": 5.814371618696784e-05, "loss": 1.4361, "step": 4320 }, { "epoch": 0.45, "learning_rate": 5.814286370274678e-05, "loss": 0.5769, "step": 4321 }, { "epoch": 0.45, "learning_rate": 5.814201102907455e-05, "loss": 0.3589, "step": 4322 }, { "epoch": 0.45, "learning_rate": 5.814115816595689e-05, "loss": 1.5314, "step": 4323 }, { "epoch": 0.45, "learning_rate": 5.814030511339954e-05, "loss": 0.9183, "step": 4324 }, { "epoch": 0.45, "learning_rate": 5.813945187140824e-05, "loss": 0.6652, "step": 4325 }, { "epoch": 0.45, "learning_rate": 5.8138598439988724e-05, "loss": 1.3083, "step": 4326 }, { "epoch": 0.45, "learning_rate": 5.813774481914675e-05, "loss": 0.4968, "step": 4327 }, { "epoch": 0.45, "learning_rate": 5.8136891008888063e-05, "loss": 0.5625, "step": 4328 }, { "epoch": 0.45, "learning_rate": 5.813603700921842e-05, "loss": 0.6936, "step": 4329 }, { "epoch": 0.45, "learning_rate": 5.8135182820143544e-05, "loss": 1.0558, "step": 4330 }, { "epoch": 0.45, "learning_rate": 5.813432844166921e-05, "loss": 0.2874, "step": 4331 }, { "epoch": 0.45, "learning_rate": 5.8133473873801155e-05, "loss": 1.0923, "step": 4332 }, { "epoch": 0.45, "learning_rate": 5.813261911654514e-05, "loss": 0.2848, "step": 4333 }, { "epoch": 0.45, "learning_rate": 5.8131764169906914e-05, "loss": 0.7243, "step": 4334 }, { "epoch": 0.45, "learning_rate": 5.813090903389223e-05, "loss": 1.3028, "step": 4335 }, { "epoch": 0.45, "learning_rate": 5.813005370850685e-05, "loss": 0.7555, "step": 4336 }, { "epoch": 0.45, "learning_rate": 5.812919819375654e-05, "loss": 0.5212, "step": 4337 }, { "epoch": 0.45, "learning_rate": 5.812834248964704e-05, "loss": 0.9939, "step": 4338 }, { "epoch": 0.45, "learning_rate": 5.8127486596184123e-05, "loss": 0.6984, "step": 4339 }, { "epoch": 0.45, "learning_rate": 5.812663051337354e-05, "loss": 1.248, "step": 4340 }, { "epoch": 0.45, "learning_rate": 5.812577424122107e-05, "loss": 0.3784, "step": 4341 }, { "epoch": 0.45, "learning_rate": 5.812491777973246e-05, "loss": 0.2776, "step": 4342 }, { "epoch": 0.45, "learning_rate": 5.812406112891349e-05, "loss": 0.5441, "step": 4343 }, { "epoch": 0.45, "learning_rate": 5.812320428876993e-05, "loss": 0.9287, "step": 4344 }, { "epoch": 0.45, "learning_rate": 5.812234725930752e-05, "loss": 0.501, "step": 4345 }, { "epoch": 0.45, "learning_rate": 5.812149004053206e-05, "loss": 0.5411, "step": 4346 }, { "epoch": 0.45, "learning_rate": 5.8120632632449315e-05, "loss": 0.9688, "step": 4347 }, { "epoch": 0.45, "learning_rate": 5.811977503506504e-05, "loss": 1.4701, "step": 4348 }, { "epoch": 0.45, "learning_rate": 5.811891724838502e-05, "loss": 0.7068, "step": 4349 }, { "epoch": 0.45, "learning_rate": 5.811805927241504e-05, "loss": 1.4551, "step": 4350 }, { "epoch": 0.45, "learning_rate": 5.8117201107160844e-05, "loss": 0.424, "step": 4351 }, { "epoch": 0.45, "learning_rate": 5.811634275262825e-05, "loss": 1.1066, "step": 4352 }, { "epoch": 0.45, "learning_rate": 5.8115484208823e-05, "loss": 1.0049, "step": 4353 }, { "epoch": 0.45, "learning_rate": 5.81146254757509e-05, "loss": 0.7051, "step": 4354 }, { "epoch": 0.45, "learning_rate": 5.8113766553417703e-05, "loss": 1.0006, "step": 4355 }, { "epoch": 0.45, "learning_rate": 5.811290744182922e-05, "loss": 1.0071, "step": 4356 }, { "epoch": 0.46, "learning_rate": 5.811204814099122e-05, "loss": 1.0682, "step": 4357 }, { "epoch": 0.46, "learning_rate": 5.8111188650909496e-05, "loss": 1.1665, "step": 4358 }, { "epoch": 0.46, "learning_rate": 5.811032897158982e-05, "loss": 1.2697, "step": 4359 }, { "epoch": 0.46, "learning_rate": 5.810946910303799e-05, "loss": 0.8931, "step": 4360 }, { "epoch": 0.46, "learning_rate": 5.810860904525978e-05, "loss": 0.52, "step": 4361 }, { "epoch": 0.46, "learning_rate": 5.8107748798261004e-05, "loss": 0.8264, "step": 4362 }, { "epoch": 0.46, "learning_rate": 5.810688836204744e-05, "loss": 0.3193, "step": 4363 }, { "epoch": 0.46, "learning_rate": 5.8106027736624875e-05, "loss": 0.4497, "step": 4364 }, { "epoch": 0.46, "learning_rate": 5.8105166921999105e-05, "loss": 0.4814, "step": 4365 }, { "epoch": 0.46, "learning_rate": 5.8104305918175935e-05, "loss": 1.5775, "step": 4366 }, { "epoch": 0.46, "learning_rate": 5.8103444725161146e-05, "loss": 0.6068, "step": 4367 }, { "epoch": 0.46, "learning_rate": 5.8102583342960545e-05, "loss": 1.265, "step": 4368 }, { "epoch": 0.46, "learning_rate": 5.8101721771579934e-05, "loss": 0.3884, "step": 4369 }, { "epoch": 0.46, "learning_rate": 5.81008600110251e-05, "loss": 0.9461, "step": 4370 }, { "epoch": 0.46, "learning_rate": 5.809999806130186e-05, "loss": 0.3412, "step": 4371 }, { "epoch": 0.46, "learning_rate": 5.8099135922415996e-05, "loss": 1.8641, "step": 4372 }, { "epoch": 0.46, "learning_rate": 5.809827359437333e-05, "loss": 1.5064, "step": 4373 }, { "epoch": 0.46, "learning_rate": 5.8097411077179664e-05, "loss": 1.3424, "step": 4374 }, { "epoch": 0.46, "learning_rate": 5.8096548370840795e-05, "loss": 1.395, "step": 4375 }, { "epoch": 0.46, "learning_rate": 5.809568547536254e-05, "loss": 0.6954, "step": 4376 }, { "epoch": 0.46, "learning_rate": 5.8094822390750696e-05, "loss": 0.9249, "step": 4377 }, { "epoch": 0.46, "learning_rate": 5.8093959117011094e-05, "loss": 0.4923, "step": 4378 }, { "epoch": 0.46, "learning_rate": 5.8093095654149526e-05, "loss": 1.3086, "step": 4379 }, { "epoch": 0.46, "learning_rate": 5.809223200217181e-05, "loss": 0.3522, "step": 4380 }, { "epoch": 0.46, "learning_rate": 5.809136816108377e-05, "loss": 1.045, "step": 4381 }, { "epoch": 0.46, "learning_rate": 5.80905041308912e-05, "loss": 1.0592, "step": 4382 }, { "epoch": 0.46, "learning_rate": 5.808963991159994e-05, "loss": 0.7722, "step": 4383 }, { "epoch": 0.46, "learning_rate": 5.808877550321579e-05, "loss": 0.9625, "step": 4384 }, { "epoch": 0.46, "learning_rate": 5.808791090574457e-05, "loss": 1.378, "step": 4385 }, { "epoch": 0.46, "learning_rate": 5.808704611919212e-05, "loss": 1.7769, "step": 4386 }, { "epoch": 0.46, "learning_rate": 5.808618114356424e-05, "loss": 0.986, "step": 4387 }, { "epoch": 0.46, "learning_rate": 5.808531597886676e-05, "loss": 0.4626, "step": 4388 }, { "epoch": 0.46, "learning_rate": 5.80844506251055e-05, "loss": 0.5615, "step": 4389 }, { "epoch": 0.46, "learning_rate": 5.808358508228629e-05, "loss": 0.6063, "step": 4390 }, { "epoch": 0.46, "learning_rate": 5.8082719350414966e-05, "loss": 0.7762, "step": 4391 }, { "epoch": 0.46, "learning_rate": 5.808185342949735e-05, "loss": 1.3546, "step": 4392 }, { "epoch": 0.46, "learning_rate": 5.808098731953924e-05, "loss": 0.8436, "step": 4393 }, { "epoch": 0.46, "learning_rate": 5.8080121020546516e-05, "loss": 1.5089, "step": 4394 }, { "epoch": 0.46, "learning_rate": 5.807925453252499e-05, "loss": 1.2807, "step": 4395 }, { "epoch": 0.46, "learning_rate": 5.807838785548049e-05, "loss": 0.9681, "step": 4396 }, { "epoch": 0.46, "learning_rate": 5.807752098941884e-05, "loss": 0.8641, "step": 4397 }, { "epoch": 0.46, "learning_rate": 5.80766539343459e-05, "loss": 0.7863, "step": 4398 }, { "epoch": 0.46, "learning_rate": 5.80757866902675e-05, "loss": 1.4782, "step": 4399 }, { "epoch": 0.46, "learning_rate": 5.807491925718946e-05, "loss": 1.2851, "step": 4400 }, { "epoch": 0.46, "learning_rate": 5.8074051635117636e-05, "loss": 0.6765, "step": 4401 }, { "epoch": 0.46, "learning_rate": 5.807318382405786e-05, "loss": 0.4244, "step": 4402 }, { "epoch": 0.46, "learning_rate": 5.8072315824015996e-05, "loss": 1.7699, "step": 4403 }, { "epoch": 0.46, "learning_rate": 5.8071447634997856e-05, "loss": 0.3899, "step": 4404 }, { "epoch": 0.46, "learning_rate": 5.807057925700931e-05, "loss": 1.2867, "step": 4405 }, { "epoch": 0.46, "learning_rate": 5.806971069005618e-05, "loss": 0.3478, "step": 4406 }, { "epoch": 0.46, "learning_rate": 5.806884193414433e-05, "loss": 0.582, "step": 4407 }, { "epoch": 0.46, "learning_rate": 5.8067972989279605e-05, "loss": 0.3113, "step": 4408 }, { "epoch": 0.46, "learning_rate": 5.806710385546785e-05, "loss": 1.3609, "step": 4409 }, { "epoch": 0.46, "learning_rate": 5.806623453271492e-05, "loss": 1.4187, "step": 4410 }, { "epoch": 0.46, "learning_rate": 5.806536502102667e-05, "loss": 1.8308, "step": 4411 }, { "epoch": 0.46, "learning_rate": 5.806449532040895e-05, "loss": 0.4865, "step": 4412 }, { "epoch": 0.46, "learning_rate": 5.806362543086761e-05, "loss": 0.7916, "step": 4413 }, { "epoch": 0.46, "learning_rate": 5.806275535240852e-05, "loss": 1.3775, "step": 4414 }, { "epoch": 0.46, "learning_rate": 5.8061885085037506e-05, "loss": 1.1649, "step": 4415 }, { "epoch": 0.46, "learning_rate": 5.806101462876046e-05, "loss": 0.8807, "step": 4416 }, { "epoch": 0.46, "learning_rate": 5.806014398358323e-05, "loss": 0.6229, "step": 4417 }, { "epoch": 0.46, "learning_rate": 5.805927314951168e-05, "loss": 1.3261, "step": 4418 }, { "epoch": 0.46, "learning_rate": 5.805840212655166e-05, "loss": 0.7201, "step": 4419 }, { "epoch": 0.46, "learning_rate": 5.805753091470904e-05, "loss": 1.3817, "step": 4420 }, { "epoch": 0.46, "learning_rate": 5.80566595139897e-05, "loss": 0.6824, "step": 4421 }, { "epoch": 0.46, "learning_rate": 5.8055787924399485e-05, "loss": 0.6031, "step": 4422 }, { "epoch": 0.46, "learning_rate": 5.805491614594427e-05, "loss": 1.0545, "step": 4423 }, { "epoch": 0.46, "learning_rate": 5.805404417862993e-05, "loss": 0.6426, "step": 4424 }, { "epoch": 0.46, "learning_rate": 5.805317202246232e-05, "loss": 0.5767, "step": 4425 }, { "epoch": 0.46, "learning_rate": 5.8052299677447326e-05, "loss": 1.1837, "step": 4426 }, { "epoch": 0.46, "learning_rate": 5.805142714359081e-05, "loss": 0.3299, "step": 4427 }, { "epoch": 0.46, "learning_rate": 5.805055442089865e-05, "loss": 0.807, "step": 4428 }, { "epoch": 0.46, "learning_rate": 5.804968150937673e-05, "loss": 1.2434, "step": 4429 }, { "epoch": 0.46, "learning_rate": 5.804880840903091e-05, "loss": 1.3915, "step": 4430 }, { "epoch": 0.46, "learning_rate": 5.804793511986707e-05, "loss": 0.2409, "step": 4431 }, { "epoch": 0.46, "learning_rate": 5.80470616418911e-05, "loss": 0.9215, "step": 4432 }, { "epoch": 0.46, "learning_rate": 5.804618797510887e-05, "loss": 0.8761, "step": 4433 }, { "epoch": 0.46, "learning_rate": 5.8045314119526266e-05, "loss": 0.4818, "step": 4434 }, { "epoch": 0.46, "learning_rate": 5.8044440075149175e-05, "loss": 1.2581, "step": 4435 }, { "epoch": 0.46, "learning_rate": 5.8043565841983464e-05, "loss": 1.1135, "step": 4436 }, { "epoch": 0.46, "learning_rate": 5.8042691420035034e-05, "loss": 1.1741, "step": 4437 }, { "epoch": 0.46, "learning_rate": 5.8041816809309773e-05, "loss": 0.9017, "step": 4438 }, { "epoch": 0.46, "learning_rate": 5.804094200981356e-05, "loss": 0.6185, "step": 4439 }, { "epoch": 0.46, "learning_rate": 5.804006702155228e-05, "loss": 0.9861, "step": 4440 }, { "epoch": 0.46, "learning_rate": 5.803919184453183e-05, "loss": 0.3223, "step": 4441 }, { "epoch": 0.46, "learning_rate": 5.803831647875811e-05, "loss": 0.8305, "step": 4442 }, { "epoch": 0.46, "learning_rate": 5.8037440924237e-05, "loss": 0.4652, "step": 4443 }, { "epoch": 0.46, "learning_rate": 5.80365651809744e-05, "loss": 1.3004, "step": 4444 }, { "epoch": 0.46, "learning_rate": 5.80356892489762e-05, "loss": 1.0058, "step": 4445 }, { "epoch": 0.46, "learning_rate": 5.80348131282483e-05, "loss": 1.135, "step": 4446 }, { "epoch": 0.46, "learning_rate": 5.803393681879659e-05, "loss": 0.5166, "step": 4447 }, { "epoch": 0.46, "learning_rate": 5.8033060320626984e-05, "loss": 0.3894, "step": 4448 }, { "epoch": 0.46, "learning_rate": 5.803218363374537e-05, "loss": 1.3966, "step": 4449 }, { "epoch": 0.46, "learning_rate": 5.8031306758157674e-05, "loss": 1.5592, "step": 4450 }, { "epoch": 0.46, "learning_rate": 5.803042969386976e-05, "loss": 1.5852, "step": 4451 }, { "epoch": 0.46, "learning_rate": 5.802955244088756e-05, "loss": 0.5702, "step": 4452 }, { "epoch": 0.47, "learning_rate": 5.802867499921697e-05, "loss": 1.3685, "step": 4453 }, { "epoch": 0.47, "learning_rate": 5.8027797368863894e-05, "loss": 0.6956, "step": 4454 }, { "epoch": 0.47, "learning_rate": 5.802691954983425e-05, "loss": 0.7271, "step": 4455 }, { "epoch": 0.47, "learning_rate": 5.802604154213394e-05, "loss": 0.8805, "step": 4456 }, { "epoch": 0.47, "learning_rate": 5.802516334576887e-05, "loss": 1.2937, "step": 4457 }, { "epoch": 0.47, "learning_rate": 5.802428496074497e-05, "loss": 0.797, "step": 4458 }, { "epoch": 0.47, "learning_rate": 5.802340638706813e-05, "loss": 0.5408, "step": 4459 }, { "epoch": 0.47, "learning_rate": 5.802252762474428e-05, "loss": 1.4186, "step": 4460 }, { "epoch": 0.47, "learning_rate": 5.802164867377933e-05, "loss": 0.2283, "step": 4461 }, { "epoch": 0.47, "learning_rate": 5.8020769534179194e-05, "loss": 1.0538, "step": 4462 }, { "epoch": 0.47, "learning_rate": 5.8019890205949796e-05, "loss": 0.813, "step": 4463 }, { "epoch": 0.47, "learning_rate": 5.8019010689097056e-05, "loss": 1.0032, "step": 4464 }, { "epoch": 0.47, "learning_rate": 5.8018130983626884e-05, "loss": 0.5625, "step": 4465 }, { "epoch": 0.47, "learning_rate": 5.801725108954522e-05, "loss": 1.0645, "step": 4466 }, { "epoch": 0.47, "learning_rate": 5.8016371006857975e-05, "loss": 0.6703, "step": 4467 }, { "epoch": 0.47, "learning_rate": 5.8015490735571075e-05, "loss": 0.8422, "step": 4468 }, { "epoch": 0.47, "learning_rate": 5.801461027569044e-05, "loss": 1.1102, "step": 4469 }, { "epoch": 0.47, "learning_rate": 5.801372962722201e-05, "loss": 1.0284, "step": 4470 }, { "epoch": 0.47, "learning_rate": 5.80128487901717e-05, "loss": 1.5566, "step": 4471 }, { "epoch": 0.47, "learning_rate": 5.801196776454546e-05, "loss": 0.4446, "step": 4472 }, { "epoch": 0.47, "learning_rate": 5.8011086550349194e-05, "loss": 0.1384, "step": 4473 }, { "epoch": 0.47, "learning_rate": 5.801020514758886e-05, "loss": 1.6116, "step": 4474 }, { "epoch": 0.47, "learning_rate": 5.8009323556270365e-05, "loss": 0.3689, "step": 4475 }, { "epoch": 0.47, "learning_rate": 5.800844177639967e-05, "loss": 0.7841, "step": 4476 }, { "epoch": 0.47, "learning_rate": 5.800755980798269e-05, "loss": 0.386, "step": 4477 }, { "epoch": 0.47, "learning_rate": 5.800667765102537e-05, "loss": 0.8109, "step": 4478 }, { "epoch": 0.47, "learning_rate": 5.800579530553366e-05, "loss": 0.3749, "step": 4479 }, { "epoch": 0.47, "learning_rate": 5.800491277151347e-05, "loss": 0.6781, "step": 4480 }, { "epoch": 0.47, "learning_rate": 5.8004030048970774e-05, "loss": 1.2556, "step": 4481 }, { "epoch": 0.47, "learning_rate": 5.8003147137911494e-05, "loss": 0.9283, "step": 4482 }, { "epoch": 0.47, "learning_rate": 5.8002264038341583e-05, "loss": 1.1568, "step": 4483 }, { "epoch": 0.47, "learning_rate": 5.8001380750266984e-05, "loss": 0.709, "step": 4484 }, { "epoch": 0.47, "learning_rate": 5.8000497273693645e-05, "loss": 0.4965, "step": 4485 }, { "epoch": 0.47, "learning_rate": 5.7999613608627496e-05, "loss": 1.0438, "step": 4486 }, { "epoch": 0.47, "learning_rate": 5.7998729755074514e-05, "loss": 0.6551, "step": 4487 }, { "epoch": 0.47, "learning_rate": 5.799784571304063e-05, "loss": 0.6148, "step": 4488 }, { "epoch": 0.47, "learning_rate": 5.79969614825318e-05, "loss": 1.3957, "step": 4489 }, { "epoch": 0.47, "learning_rate": 5.7996077063553975e-05, "loss": 0.2584, "step": 4490 }, { "epoch": 0.47, "learning_rate": 5.79951924561131e-05, "loss": 0.7375, "step": 4491 }, { "epoch": 0.47, "learning_rate": 5.7994307660215156e-05, "loss": 0.7312, "step": 4492 }, { "epoch": 0.47, "learning_rate": 5.799342267586607e-05, "loss": 1.143, "step": 4493 }, { "epoch": 0.47, "learning_rate": 5.799253750307182e-05, "loss": 0.3792, "step": 4494 }, { "epoch": 0.47, "learning_rate": 5.799165214183836e-05, "loss": 0.5992, "step": 4495 }, { "epoch": 0.47, "learning_rate": 5.7990766592171644e-05, "loss": 1.1887, "step": 4496 }, { "epoch": 0.47, "learning_rate": 5.798988085407763e-05, "loss": 1.2835, "step": 4497 }, { "epoch": 0.47, "learning_rate": 5.7988994927562294e-05, "loss": 0.33, "step": 4498 }, { "epoch": 0.47, "learning_rate": 5.7988108812631595e-05, "loss": 1.6107, "step": 4499 }, { "epoch": 0.47, "learning_rate": 5.798722250929149e-05, "loss": 0.6962, "step": 4500 }, { "epoch": 0.47, "learning_rate": 5.798633601754796e-05, "loss": 0.4379, "step": 4501 }, { "epoch": 0.47, "learning_rate": 5.798544933740695e-05, "loss": 0.6255, "step": 4502 }, { "epoch": 0.47, "learning_rate": 5.7984562468874454e-05, "loss": 0.3665, "step": 4503 }, { "epoch": 0.47, "learning_rate": 5.7983675411956436e-05, "loss": 0.2098, "step": 4504 }, { "epoch": 0.47, "learning_rate": 5.7982788166658845e-05, "loss": 0.2852, "step": 4505 }, { "epoch": 0.47, "learning_rate": 5.798190073298769e-05, "loss": 0.7591, "step": 4506 }, { "epoch": 0.47, "learning_rate": 5.798101311094892e-05, "loss": 0.3543, "step": 4507 }, { "epoch": 0.47, "learning_rate": 5.7980125300548505e-05, "loss": 1.1282, "step": 4508 }, { "epoch": 0.47, "learning_rate": 5.797923730179245e-05, "loss": 0.1736, "step": 4509 }, { "epoch": 0.47, "learning_rate": 5.797834911468671e-05, "loss": 1.064, "step": 4510 }, { "epoch": 0.47, "learning_rate": 5.7977460739237274e-05, "loss": 0.3795, "step": 4511 }, { "epoch": 0.47, "learning_rate": 5.797657217545011e-05, "loss": 0.7484, "step": 4512 }, { "epoch": 0.47, "learning_rate": 5.797568342333121e-05, "loss": 0.4919, "step": 4513 }, { "epoch": 0.47, "learning_rate": 5.797479448288656e-05, "loss": 0.8438, "step": 4514 }, { "epoch": 0.47, "learning_rate": 5.7973905354122145e-05, "loss": 0.7034, "step": 4515 }, { "epoch": 0.47, "learning_rate": 5.797301603704393e-05, "loss": 1.1765, "step": 4516 }, { "epoch": 0.47, "learning_rate": 5.797212653165792e-05, "loss": 0.3676, "step": 4517 }, { "epoch": 0.47, "learning_rate": 5.7971236837970105e-05, "loss": 0.5702, "step": 4518 }, { "epoch": 0.47, "learning_rate": 5.7970346955986465e-05, "loss": 1.3949, "step": 4519 }, { "epoch": 0.47, "learning_rate": 5.796945688571299e-05, "loss": 1.4123, "step": 4520 }, { "epoch": 0.47, "learning_rate": 5.796856662715569e-05, "loss": 0.763, "step": 4521 }, { "epoch": 0.47, "learning_rate": 5.7967676180320526e-05, "loss": 0.9399, "step": 4522 }, { "epoch": 0.47, "learning_rate": 5.796678554521351e-05, "loss": 0.8078, "step": 4523 }, { "epoch": 0.47, "learning_rate": 5.796589472184065e-05, "loss": 1.2323, "step": 4524 }, { "epoch": 0.47, "learning_rate": 5.796500371020793e-05, "loss": 0.3079, "step": 4525 }, { "epoch": 0.47, "learning_rate": 5.796411251032134e-05, "loss": 1.74, "step": 4526 }, { "epoch": 0.47, "learning_rate": 5.796322112218689e-05, "loss": 0.6257, "step": 4527 }, { "epoch": 0.47, "learning_rate": 5.796232954581058e-05, "loss": 0.4905, "step": 4528 }, { "epoch": 0.47, "learning_rate": 5.796143778119841e-05, "loss": 0.5234, "step": 4529 }, { "epoch": 0.47, "learning_rate": 5.796054582835638e-05, "loss": 1.5139, "step": 4530 }, { "epoch": 0.47, "learning_rate": 5.79596536872905e-05, "loss": 1.0244, "step": 4531 }, { "epoch": 0.47, "learning_rate": 5.7958761358006774e-05, "loss": 1.7478, "step": 4532 }, { "epoch": 0.47, "learning_rate": 5.795786884051121e-05, "loss": 0.9017, "step": 4533 }, { "epoch": 0.47, "learning_rate": 5.795697613480981e-05, "loss": 0.9084, "step": 4534 }, { "epoch": 0.47, "learning_rate": 5.795608324090859e-05, "loss": 1.2472, "step": 4535 }, { "epoch": 0.47, "learning_rate": 5.7955190158813565e-05, "loss": 0.8513, "step": 4536 }, { "epoch": 0.47, "learning_rate": 5.7954296888530735e-05, "loss": 1.5966, "step": 4537 }, { "epoch": 0.47, "learning_rate": 5.795340343006612e-05, "loss": 0.9133, "step": 4538 }, { "epoch": 0.47, "learning_rate": 5.795250978342573e-05, "loss": 0.5911, "step": 4539 }, { "epoch": 0.47, "learning_rate": 5.7951615948615586e-05, "loss": 0.4011, "step": 4540 }, { "epoch": 0.47, "learning_rate": 5.795072192564171e-05, "loss": 1.6287, "step": 4541 }, { "epoch": 0.47, "learning_rate": 5.794982771451011e-05, "loss": 0.7597, "step": 4542 }, { "epoch": 0.47, "learning_rate": 5.79489333152268e-05, "loss": 0.2655, "step": 4543 }, { "epoch": 0.47, "learning_rate": 5.7948038727797826e-05, "loss": 1.0317, "step": 4544 }, { "epoch": 0.47, "learning_rate": 5.7947143952229185e-05, "loss": 0.3457, "step": 4545 }, { "epoch": 0.47, "learning_rate": 5.794624898852692e-05, "loss": 1.5272, "step": 4546 }, { "epoch": 0.47, "learning_rate": 5.794535383669703e-05, "loss": 1.0363, "step": 4547 }, { "epoch": 0.47, "learning_rate": 5.7944458496745574e-05, "loss": 1.1508, "step": 4548 }, { "epoch": 0.48, "learning_rate": 5.794356296867855e-05, "loss": 0.8159, "step": 4549 }, { "epoch": 0.48, "learning_rate": 5.7942667252502e-05, "loss": 1.4944, "step": 4550 }, { "epoch": 0.48, "learning_rate": 5.794177134822195e-05, "loss": 1.5423, "step": 4551 }, { "epoch": 0.48, "learning_rate": 5.7940875255844436e-05, "loss": 0.6291, "step": 4552 }, { "epoch": 0.48, "learning_rate": 5.793997897537549e-05, "loss": 0.4075, "step": 4553 }, { "epoch": 0.48, "learning_rate": 5.793908250682114e-05, "loss": 0.7079, "step": 4554 }, { "epoch": 0.48, "learning_rate": 5.7938185850187426e-05, "loss": 0.7533, "step": 4555 }, { "epoch": 0.48, "learning_rate": 5.793728900548038e-05, "loss": 1.4834, "step": 4556 }, { "epoch": 0.48, "learning_rate": 5.7936391972706045e-05, "loss": 0.5989, "step": 4557 }, { "epoch": 0.48, "learning_rate": 5.7935494751870454e-05, "loss": 1.2323, "step": 4558 }, { "epoch": 0.48, "learning_rate": 5.7934597342979644e-05, "loss": 1.1022, "step": 4559 }, { "epoch": 0.48, "learning_rate": 5.7933699746039665e-05, "loss": 0.3779, "step": 4560 }, { "epoch": 0.48, "learning_rate": 5.793280196105655e-05, "loss": 0.4144, "step": 4561 }, { "epoch": 0.48, "learning_rate": 5.7931903988036355e-05, "loss": 1.0378, "step": 4562 }, { "epoch": 0.48, "learning_rate": 5.793100582698512e-05, "loss": 0.5535, "step": 4563 }, { "epoch": 0.48, "learning_rate": 5.793010747790888e-05, "loss": 1.2929, "step": 4564 }, { "epoch": 0.48, "learning_rate": 5.79292089408137e-05, "loss": 1.895, "step": 4565 }, { "epoch": 0.48, "learning_rate": 5.792831021570561e-05, "loss": 1.3079, "step": 4566 }, { "epoch": 0.48, "learning_rate": 5.7927411302590676e-05, "loss": 0.7275, "step": 4567 }, { "epoch": 0.48, "learning_rate": 5.792651220147495e-05, "loss": 0.8277, "step": 4568 }, { "epoch": 0.48, "learning_rate": 5.792561291236447e-05, "loss": 1.2665, "step": 4569 }, { "epoch": 0.48, "learning_rate": 5.79247134352653e-05, "loss": 1.2696, "step": 4570 }, { "epoch": 0.48, "learning_rate": 5.7923813770183496e-05, "loss": 0.3339, "step": 4571 }, { "epoch": 0.48, "learning_rate": 5.792291391712511e-05, "loss": 1.1122, "step": 4572 }, { "epoch": 0.48, "learning_rate": 5.7922013876096194e-05, "loss": 1.0557, "step": 4573 }, { "epoch": 0.48, "learning_rate": 5.792111364710282e-05, "loss": 1.738, "step": 4574 }, { "epoch": 0.48, "learning_rate": 5.7920213230151046e-05, "loss": 1.2348, "step": 4575 }, { "epoch": 0.48, "learning_rate": 5.791931262524693e-05, "loss": 0.808, "step": 4576 }, { "epoch": 0.48, "learning_rate": 5.7918411832396524e-05, "loss": 0.726, "step": 4577 }, { "epoch": 0.48, "learning_rate": 5.7917510851605905e-05, "loss": 0.367, "step": 4578 }, { "epoch": 0.48, "learning_rate": 5.791660968288114e-05, "loss": 1.3442, "step": 4579 }, { "epoch": 0.48, "learning_rate": 5.791570832622828e-05, "loss": 0.7574, "step": 4580 }, { "epoch": 0.48, "learning_rate": 5.7914806781653415e-05, "loss": 0.9991, "step": 4581 }, { "epoch": 0.48, "learning_rate": 5.79139050491626e-05, "loss": 0.7768, "step": 4582 }, { "epoch": 0.48, "learning_rate": 5.7913003128761904e-05, "loss": 0.8881, "step": 4583 }, { "epoch": 0.48, "learning_rate": 5.7912101020457404e-05, "loss": 0.6447, "step": 4584 }, { "epoch": 0.48, "learning_rate": 5.791119872425517e-05, "loss": 1.3459, "step": 4585 }, { "epoch": 0.48, "learning_rate": 5.791029624016128e-05, "loss": 0.905, "step": 4586 }, { "epoch": 0.48, "learning_rate": 5.7909393568181806e-05, "loss": 0.3571, "step": 4587 }, { "epoch": 0.48, "learning_rate": 5.790849070832282e-05, "loss": 0.9185, "step": 4588 }, { "epoch": 0.48, "learning_rate": 5.79075876605904e-05, "loss": 0.4046, "step": 4589 }, { "epoch": 0.48, "learning_rate": 5.7906684424990645e-05, "loss": 0.4362, "step": 4590 }, { "epoch": 0.48, "learning_rate": 5.7905781001529604e-05, "loss": 0.4193, "step": 4591 }, { "epoch": 0.48, "learning_rate": 5.790487739021339e-05, "loss": 0.7486, "step": 4592 }, { "epoch": 0.48, "learning_rate": 5.7903973591048064e-05, "loss": 1.5102, "step": 4593 }, { "epoch": 0.48, "learning_rate": 5.790306960403972e-05, "loss": 1.8942, "step": 4594 }, { "epoch": 0.48, "learning_rate": 5.790216542919443e-05, "loss": 1.45, "step": 4595 }, { "epoch": 0.48, "learning_rate": 5.79012610665183e-05, "loss": 1.0617, "step": 4596 }, { "epoch": 0.48, "learning_rate": 5.790035651601741e-05, "loss": 0.3784, "step": 4597 }, { "epoch": 0.48, "learning_rate": 5.7899451777697835e-05, "loss": 1.1181, "step": 4598 }, { "epoch": 0.48, "learning_rate": 5.789854685156569e-05, "loss": 1.0226, "step": 4599 }, { "epoch": 0.48, "learning_rate": 5.789764173762705e-05, "loss": 0.4235, "step": 4600 }, { "epoch": 0.48, "learning_rate": 5.7896736435888016e-05, "loss": 0.4289, "step": 4601 }, { "epoch": 0.48, "learning_rate": 5.7895830946354687e-05, "loss": 0.3222, "step": 4602 }, { "epoch": 0.48, "learning_rate": 5.789492526903314e-05, "loss": 0.8399, "step": 4603 }, { "epoch": 0.48, "learning_rate": 5.7894019403929484e-05, "loss": 0.8585, "step": 4604 }, { "epoch": 0.48, "learning_rate": 5.789311335104983e-05, "loss": 0.512, "step": 4605 }, { "epoch": 0.48, "learning_rate": 5.789220711040024e-05, "loss": 0.2499, "step": 4606 }, { "epoch": 0.48, "learning_rate": 5.7891300681986855e-05, "loss": 1.2413, "step": 4607 }, { "epoch": 0.48, "learning_rate": 5.789039406581576e-05, "loss": 1.5356, "step": 4608 }, { "epoch": 0.48, "learning_rate": 5.7889487261893046e-05, "loss": 0.9897, "step": 4609 }, { "epoch": 0.48, "learning_rate": 5.788858027022484e-05, "loss": 0.3867, "step": 4610 }, { "epoch": 0.48, "learning_rate": 5.788767309081723e-05, "loss": 1.0175, "step": 4611 }, { "epoch": 0.48, "learning_rate": 5.788676572367634e-05, "loss": 0.2403, "step": 4612 }, { "epoch": 0.48, "learning_rate": 5.788585816880826e-05, "loss": 0.4951, "step": 4613 }, { "epoch": 0.48, "learning_rate": 5.788495042621911e-05, "loss": 1.2928, "step": 4614 }, { "epoch": 0.48, "learning_rate": 5.7884042495914994e-05, "loss": 0.4101, "step": 4615 }, { "epoch": 0.48, "learning_rate": 5.7883134377902036e-05, "loss": 0.8808, "step": 4616 }, { "epoch": 0.48, "learning_rate": 5.788222607218634e-05, "loss": 0.6761, "step": 4617 }, { "epoch": 0.48, "learning_rate": 5.788131757877401e-05, "loss": 1.4907, "step": 4618 }, { "epoch": 0.48, "learning_rate": 5.788040889767119e-05, "loss": 0.6749, "step": 4619 }, { "epoch": 0.48, "learning_rate": 5.787950002888397e-05, "loss": 0.189, "step": 4620 }, { "epoch": 0.48, "learning_rate": 5.7878590972418494e-05, "loss": 1.0105, "step": 4621 }, { "epoch": 0.48, "learning_rate": 5.787768172828085e-05, "loss": 0.4709, "step": 4622 }, { "epoch": 0.48, "learning_rate": 5.7876772296477176e-05, "loss": 0.7088, "step": 4623 }, { "epoch": 0.48, "learning_rate": 5.78758626770136e-05, "loss": 1.2136, "step": 4624 }, { "epoch": 0.48, "learning_rate": 5.7874952869896236e-05, "loss": 0.7165, "step": 4625 }, { "epoch": 0.48, "learning_rate": 5.7874042875131216e-05, "loss": 0.2422, "step": 4626 }, { "epoch": 0.48, "learning_rate": 5.787313269272467e-05, "loss": 0.4876, "step": 4627 }, { "epoch": 0.48, "learning_rate": 5.78722223226827e-05, "loss": 0.9731, "step": 4628 }, { "epoch": 0.48, "learning_rate": 5.787131176501145e-05, "loss": 0.9151, "step": 4629 }, { "epoch": 0.48, "learning_rate": 5.7870401019717054e-05, "loss": 0.6853, "step": 4630 }, { "epoch": 0.48, "learning_rate": 5.786949008680565e-05, "loss": 0.7865, "step": 4631 }, { "epoch": 0.48, "learning_rate": 5.786857896628335e-05, "loss": 0.6423, "step": 4632 }, { "epoch": 0.48, "learning_rate": 5.78676676581563e-05, "loss": 0.93, "step": 4633 }, { "epoch": 0.48, "learning_rate": 5.786675616243063e-05, "loss": 1.5071, "step": 4634 }, { "epoch": 0.48, "learning_rate": 5.7865844479112475e-05, "loss": 0.9038, "step": 4635 }, { "epoch": 0.48, "learning_rate": 5.786493260820798e-05, "loss": 0.6614, "step": 4636 }, { "epoch": 0.48, "learning_rate": 5.786402054972328e-05, "loss": 0.6407, "step": 4637 }, { "epoch": 0.48, "learning_rate": 5.786310830366451e-05, "loss": 1.3374, "step": 4638 }, { "epoch": 0.48, "learning_rate": 5.7862195870037816e-05, "loss": 2.1605, "step": 4639 }, { "epoch": 0.48, "learning_rate": 5.786128324884933e-05, "loss": 1.5428, "step": 4640 }, { "epoch": 0.48, "learning_rate": 5.786037044010522e-05, "loss": 0.259, "step": 4641 }, { "epoch": 0.48, "learning_rate": 5.78594574438116e-05, "loss": 1.3346, "step": 4642 }, { "epoch": 0.48, "learning_rate": 5.785854425997464e-05, "loss": 1.3807, "step": 4643 }, { "epoch": 0.49, "learning_rate": 5.7857630888600474e-05, "loss": 0.6952, "step": 4644 }, { "epoch": 0.49, "learning_rate": 5.785671732969525e-05, "loss": 1.401, "step": 4645 }, { "epoch": 0.49, "learning_rate": 5.785580358326514e-05, "loss": 0.8841, "step": 4646 }, { "epoch": 0.49, "learning_rate": 5.7854889649316264e-05, "loss": 1.24, "step": 4647 }, { "epoch": 0.49, "learning_rate": 5.7853975527854795e-05, "loss": 0.9461, "step": 4648 }, { "epoch": 0.49, "learning_rate": 5.7853061218886876e-05, "loss": 0.4381, "step": 4649 }, { "epoch": 0.49, "learning_rate": 5.785214672241867e-05, "loss": 2.2579, "step": 4650 }, { "epoch": 0.49, "learning_rate": 5.785123203845632e-05, "loss": 0.423, "step": 4651 }, { "epoch": 0.49, "learning_rate": 5.785031716700601e-05, "loss": 0.5088, "step": 4652 }, { "epoch": 0.49, "learning_rate": 5.784940210807387e-05, "loss": 0.7709, "step": 4653 }, { "epoch": 0.49, "learning_rate": 5.784848686166607e-05, "loss": 0.2606, "step": 4654 }, { "epoch": 0.49, "learning_rate": 5.784757142778877e-05, "loss": 0.1687, "step": 4655 }, { "epoch": 0.49, "learning_rate": 5.784665580644814e-05, "loss": 1.4588, "step": 4656 }, { "epoch": 0.49, "learning_rate": 5.7845739997650334e-05, "loss": 1.0258, "step": 4657 }, { "epoch": 0.49, "learning_rate": 5.784482400140153e-05, "loss": 0.6819, "step": 4658 }, { "epoch": 0.49, "learning_rate": 5.7843907817707877e-05, "loss": 0.4341, "step": 4659 }, { "epoch": 0.49, "learning_rate": 5.784299144657556e-05, "loss": 0.9057, "step": 4660 }, { "epoch": 0.49, "learning_rate": 5.7842074888010736e-05, "loss": 0.8937, "step": 4661 }, { "epoch": 0.49, "learning_rate": 5.784115814201958e-05, "loss": 1.354, "step": 4662 }, { "epoch": 0.49, "learning_rate": 5.784024120860826e-05, "loss": 1.4206, "step": 4663 }, { "epoch": 0.49, "learning_rate": 5.783932408778295e-05, "loss": 0.4364, "step": 4664 }, { "epoch": 0.49, "learning_rate": 5.783840677954982e-05, "loss": 1.2888, "step": 4665 }, { "epoch": 0.49, "learning_rate": 5.7837489283915056e-05, "loss": 1.0519, "step": 4666 }, { "epoch": 0.49, "learning_rate": 5.783657160088482e-05, "loss": 0.992, "step": 4667 }, { "epoch": 0.49, "learning_rate": 5.78356537304653e-05, "loss": 1.0886, "step": 4668 }, { "epoch": 0.49, "learning_rate": 5.783473567266268e-05, "loss": 1.3193, "step": 4669 }, { "epoch": 0.49, "learning_rate": 5.7833817427483124e-05, "loss": 0.9736, "step": 4670 }, { "epoch": 0.49, "learning_rate": 5.783289899493283e-05, "loss": 1.0932, "step": 4671 }, { "epoch": 0.49, "learning_rate": 5.783198037501796e-05, "loss": 0.8078, "step": 4672 }, { "epoch": 0.49, "learning_rate": 5.783106156774472e-05, "loss": 0.6565, "step": 4673 }, { "epoch": 0.49, "learning_rate": 5.7830142573119286e-05, "loss": 1.5828, "step": 4674 }, { "epoch": 0.49, "learning_rate": 5.7829223391147836e-05, "loss": 0.4905, "step": 4675 }, { "epoch": 0.49, "learning_rate": 5.7828304021836564e-05, "loss": 1.6942, "step": 4676 }, { "epoch": 0.49, "learning_rate": 5.782738446519167e-05, "loss": 0.4672, "step": 4677 }, { "epoch": 0.49, "learning_rate": 5.7826464721219336e-05, "loss": 1.3987, "step": 4678 }, { "epoch": 0.49, "learning_rate": 5.782554478992574e-05, "loss": 0.5848, "step": 4679 }, { "epoch": 0.49, "learning_rate": 5.78246246713171e-05, "loss": 0.6378, "step": 4680 }, { "epoch": 0.49, "learning_rate": 5.782370436539959e-05, "loss": 0.1905, "step": 4681 }, { "epoch": 0.49, "learning_rate": 5.7822783872179416e-05, "loss": 0.5309, "step": 4682 }, { "epoch": 0.49, "learning_rate": 5.782186319166277e-05, "loss": 0.7465, "step": 4683 }, { "epoch": 0.49, "learning_rate": 5.7820942323855846e-05, "loss": 0.3143, "step": 4684 }, { "epoch": 0.49, "learning_rate": 5.782002126876486e-05, "loss": 1.4639, "step": 4685 }, { "epoch": 0.49, "learning_rate": 5.781910002639599e-05, "loss": 0.7426, "step": 4686 }, { "epoch": 0.49, "learning_rate": 5.781817859675544e-05, "loss": 0.6054, "step": 4687 }, { "epoch": 0.49, "learning_rate": 5.7817256979849434e-05, "loss": 1.4423, "step": 4688 }, { "epoch": 0.49, "learning_rate": 5.7816335175684154e-05, "loss": 0.6762, "step": 4689 }, { "epoch": 0.49, "learning_rate": 5.781541318426582e-05, "loss": 0.8296, "step": 4690 }, { "epoch": 0.49, "learning_rate": 5.781449100560063e-05, "loss": 0.2655, "step": 4691 }, { "epoch": 0.49, "learning_rate": 5.781356863969481e-05, "loss": 0.9746, "step": 4692 }, { "epoch": 0.49, "learning_rate": 5.781264608655454e-05, "loss": 1.1838, "step": 4693 }, { "epoch": 0.49, "learning_rate": 5.7811723346186035e-05, "loss": 0.9313, "step": 4694 }, { "epoch": 0.49, "learning_rate": 5.781080041859553e-05, "loss": 0.3948, "step": 4695 }, { "epoch": 0.49, "learning_rate": 5.780987730378922e-05, "loss": 1.2818, "step": 4696 }, { "epoch": 0.49, "learning_rate": 5.7808954001773324e-05, "loss": 1.2818, "step": 4697 }, { "epoch": 0.49, "learning_rate": 5.780803051255405e-05, "loss": 1.524, "step": 4698 }, { "epoch": 0.49, "learning_rate": 5.7807106836137624e-05, "loss": 1.394, "step": 4699 }, { "epoch": 0.49, "learning_rate": 5.780618297253026e-05, "loss": 1.6549, "step": 4700 }, { "epoch": 0.49, "learning_rate": 5.780525892173819e-05, "loss": 1.332, "step": 4701 }, { "epoch": 0.49, "learning_rate": 5.780433468376761e-05, "loss": 1.5916, "step": 4702 }, { "epoch": 0.49, "learning_rate": 5.780341025862476e-05, "loss": 0.8119, "step": 4703 }, { "epoch": 0.49, "learning_rate": 5.780248564631586e-05, "loss": 0.7034, "step": 4704 }, { "epoch": 0.49, "learning_rate": 5.780156084684712e-05, "loss": 0.7321, "step": 4705 }, { "epoch": 0.49, "learning_rate": 5.7800635860224793e-05, "loss": 0.622, "step": 4706 }, { "epoch": 0.49, "learning_rate": 5.7799710686455086e-05, "loss": 0.4321, "step": 4707 }, { "epoch": 0.49, "learning_rate": 5.7798785325544223e-05, "loss": 1.0623, "step": 4708 }, { "epoch": 0.49, "learning_rate": 5.7797859777498446e-05, "loss": 0.6573, "step": 4709 }, { "epoch": 0.49, "learning_rate": 5.779693404232398e-05, "loss": 1.3951, "step": 4710 }, { "epoch": 0.49, "learning_rate": 5.7796008120027064e-05, "loss": 2.2571, "step": 4711 }, { "epoch": 0.49, "learning_rate": 5.779508201061392e-05, "loss": 0.7278, "step": 4712 }, { "epoch": 0.49, "learning_rate": 5.779415571409079e-05, "loss": 1.3087, "step": 4713 }, { "epoch": 0.49, "learning_rate": 5.77932292304639e-05, "loss": 1.0418, "step": 4714 }, { "epoch": 0.49, "learning_rate": 5.779230255973951e-05, "loss": 0.5162, "step": 4715 }, { "epoch": 0.49, "learning_rate": 5.7791375701923825e-05, "loss": 1.1174, "step": 4716 }, { "epoch": 0.49, "learning_rate": 5.7790448657023114e-05, "loss": 0.81, "step": 4717 }, { "epoch": 0.49, "learning_rate": 5.7789521425043596e-05, "loss": 1.2725, "step": 4718 }, { "epoch": 0.49, "learning_rate": 5.778859400599152e-05, "loss": 1.1188, "step": 4719 }, { "epoch": 0.49, "learning_rate": 5.778766639987314e-05, "loss": 0.3788, "step": 4720 }, { "epoch": 0.49, "learning_rate": 5.7786738606694685e-05, "loss": 0.5794, "step": 4721 }, { "epoch": 0.49, "learning_rate": 5.778581062646241e-05, "loss": 1.2351, "step": 4722 }, { "epoch": 0.49, "learning_rate": 5.778488245918256e-05, "loss": 0.9828, "step": 4723 }, { "epoch": 0.49, "learning_rate": 5.778395410486138e-05, "loss": 0.3268, "step": 4724 }, { "epoch": 0.49, "learning_rate": 5.778302556350513e-05, "loss": 1.3461, "step": 4725 }, { "epoch": 0.49, "learning_rate": 5.7782096835120045e-05, "loss": 0.3135, "step": 4726 }, { "epoch": 0.49, "learning_rate": 5.7781167919712384e-05, "loss": 1.0706, "step": 4727 }, { "epoch": 0.49, "learning_rate": 5.7780238817288406e-05, "loss": 1.462, "step": 4728 }, { "epoch": 0.49, "learning_rate": 5.777930952785435e-05, "loss": 1.2321, "step": 4729 }, { "epoch": 0.49, "learning_rate": 5.777838005141649e-05, "loss": 1.2515, "step": 4730 }, { "epoch": 0.49, "learning_rate": 5.7777450387981073e-05, "loss": 1.0858, "step": 4731 }, { "epoch": 0.49, "learning_rate": 5.777652053755436e-05, "loss": 1.2584, "step": 4732 }, { "epoch": 0.49, "learning_rate": 5.777559050014261e-05, "loss": 0.2868, "step": 4733 }, { "epoch": 0.49, "learning_rate": 5.777466027575209e-05, "loss": 1.2154, "step": 4734 }, { "epoch": 0.49, "learning_rate": 5.7773729864389043e-05, "loss": 0.7208, "step": 4735 }, { "epoch": 0.49, "learning_rate": 5.777279926605976e-05, "loss": 0.4241, "step": 4736 }, { "epoch": 0.49, "learning_rate": 5.7771868480770474e-05, "loss": 0.5881, "step": 4737 }, { "epoch": 0.49, "learning_rate": 5.7770937508527486e-05, "loss": 1.3053, "step": 4738 }, { "epoch": 0.49, "learning_rate": 5.777000634933703e-05, "loss": 0.6891, "step": 4739 }, { "epoch": 0.5, "learning_rate": 5.77690750032054e-05, "loss": 0.3964, "step": 4740 }, { "epoch": 0.5, "learning_rate": 5.7768143470138845e-05, "loss": 0.8774, "step": 4741 }, { "epoch": 0.5, "learning_rate": 5.776721175014365e-05, "loss": 0.3901, "step": 4742 }, { "epoch": 0.5, "learning_rate": 5.776627984322608e-05, "loss": 1.504, "step": 4743 }, { "epoch": 0.5, "learning_rate": 5.776534774939241e-05, "loss": 1.5701, "step": 4744 }, { "epoch": 0.5, "learning_rate": 5.7764415468648916e-05, "loss": 1.1367, "step": 4745 }, { "epoch": 0.5, "learning_rate": 5.7763483001001874e-05, "loss": 0.7181, "step": 4746 }, { "epoch": 0.5, "learning_rate": 5.776255034645756e-05, "loss": 0.9817, "step": 4747 }, { "epoch": 0.5, "learning_rate": 5.776161750502226e-05, "loss": 0.588, "step": 4748 }, { "epoch": 0.5, "learning_rate": 5.776068447670224e-05, "loss": 1.5191, "step": 4749 }, { "epoch": 0.5, "learning_rate": 5.7759751261503785e-05, "loss": 0.5419, "step": 4750 }, { "epoch": 0.5, "learning_rate": 5.7758817859433184e-05, "loss": 1.3388, "step": 4751 }, { "epoch": 0.5, "learning_rate": 5.775788427049672e-05, "loss": 1.5202, "step": 4752 }, { "epoch": 0.5, "learning_rate": 5.775695049470066e-05, "loss": 0.9516, "step": 4753 }, { "epoch": 0.5, "learning_rate": 5.775601653205132e-05, "loss": 1.2863, "step": 4754 }, { "epoch": 0.5, "learning_rate": 5.7755082382554966e-05, "loss": 1.5576, "step": 4755 }, { "epoch": 0.5, "learning_rate": 5.7754148046217886e-05, "loss": 0.7295, "step": 4756 }, { "epoch": 0.5, "learning_rate": 5.775321352304638e-05, "loss": 0.1976, "step": 4757 }, { "epoch": 0.5, "learning_rate": 5.775227881304674e-05, "loss": 1.5135, "step": 4758 }, { "epoch": 0.5, "learning_rate": 5.775134391622524e-05, "loss": 0.59, "step": 4759 }, { "epoch": 0.5, "learning_rate": 5.775040883258819e-05, "loss": 0.5547, "step": 4760 }, { "epoch": 0.5, "learning_rate": 5.774947356214189e-05, "loss": 1.2129, "step": 4761 }, { "epoch": 0.5, "learning_rate": 5.774853810489261e-05, "loss": 1.6685, "step": 4762 }, { "epoch": 0.5, "learning_rate": 5.774760246084668e-05, "loss": 0.6618, "step": 4763 }, { "epoch": 0.5, "learning_rate": 5.7746666630010374e-05, "loss": 0.3532, "step": 4764 }, { "epoch": 0.5, "learning_rate": 5.774573061239e-05, "loss": 0.48, "step": 4765 }, { "epoch": 0.5, "learning_rate": 5.774479440799187e-05, "loss": 1.4942, "step": 4766 }, { "epoch": 0.5, "learning_rate": 5.774385801682227e-05, "loss": 0.5821, "step": 4767 }, { "epoch": 0.5, "learning_rate": 5.774292143888751e-05, "loss": 1.494, "step": 4768 }, { "epoch": 0.5, "learning_rate": 5.7741984674193896e-05, "loss": 0.7091, "step": 4769 }, { "epoch": 0.5, "learning_rate": 5.774104772274772e-05, "loss": 1.7684, "step": 4770 }, { "epoch": 0.5, "learning_rate": 5.774011058455531e-05, "loss": 0.2047, "step": 4771 }, { "epoch": 0.5, "learning_rate": 5.7739173259622975e-05, "loss": 1.1912, "step": 4772 }, { "epoch": 0.5, "learning_rate": 5.7738235747957e-05, "loss": 1.1314, "step": 4773 }, { "epoch": 0.5, "learning_rate": 5.773729804956372e-05, "loss": 1.2608, "step": 4774 }, { "epoch": 0.5, "learning_rate": 5.7736360164449436e-05, "loss": 1.3243, "step": 4775 }, { "epoch": 0.5, "learning_rate": 5.7735422092620474e-05, "loss": 0.4167, "step": 4776 }, { "epoch": 0.5, "learning_rate": 5.773448383408313e-05, "loss": 1.2344, "step": 4777 }, { "epoch": 0.5, "learning_rate": 5.773354538884373e-05, "loss": 0.7558, "step": 4778 }, { "epoch": 0.5, "learning_rate": 5.77326067569086e-05, "loss": 0.8396, "step": 4779 }, { "epoch": 0.5, "learning_rate": 5.773166793828405e-05, "loss": 1.5397, "step": 4780 }, { "epoch": 0.5, "learning_rate": 5.773072893297639e-05, "loss": 1.0036, "step": 4781 }, { "epoch": 0.5, "learning_rate": 5.772978974099196e-05, "loss": 0.891, "step": 4782 }, { "epoch": 0.5, "learning_rate": 5.772885036233708e-05, "loss": 0.3726, "step": 4783 }, { "epoch": 0.5, "learning_rate": 5.772791079701805e-05, "loss": 1.5508, "step": 4784 }, { "epoch": 0.5, "learning_rate": 5.7726971045041226e-05, "loss": 0.4452, "step": 4785 }, { "epoch": 0.5, "learning_rate": 5.772603110641291e-05, "loss": 0.4294, "step": 4786 }, { "epoch": 0.5, "learning_rate": 5.7725090981139445e-05, "loss": 1.2143, "step": 4787 }, { "epoch": 0.5, "learning_rate": 5.772415066922716e-05, "loss": 0.5826, "step": 4788 }, { "epoch": 0.5, "learning_rate": 5.772321017068237e-05, "loss": 0.5034, "step": 4789 }, { "epoch": 0.5, "learning_rate": 5.772226948551143e-05, "loss": 1.4888, "step": 4790 }, { "epoch": 0.5, "learning_rate": 5.772132861372064e-05, "loss": 0.5882, "step": 4791 }, { "epoch": 0.5, "learning_rate": 5.772038755531636e-05, "loss": 0.8743, "step": 4792 }, { "epoch": 0.5, "learning_rate": 5.771944631030492e-05, "loss": 1.1367, "step": 4793 }, { "epoch": 0.5, "learning_rate": 5.771850487869265e-05, "loss": 1.4727, "step": 4794 }, { "epoch": 0.5, "learning_rate": 5.77175632604859e-05, "loss": 0.3073, "step": 4795 }, { "epoch": 0.5, "learning_rate": 5.7716621455690986e-05, "loss": 1.4805, "step": 4796 }, { "epoch": 0.5, "learning_rate": 5.771567946431427e-05, "loss": 0.6223, "step": 4797 }, { "epoch": 0.5, "learning_rate": 5.771473728636208e-05, "loss": 0.6982, "step": 4798 }, { "epoch": 0.5, "learning_rate": 5.771379492184076e-05, "loss": 0.8324, "step": 4799 }, { "epoch": 0.5, "learning_rate": 5.771285237075667e-05, "loss": 0.3667, "step": 4800 }, { "epoch": 0.5, "learning_rate": 5.771190963311613e-05, "loss": 1.0113, "step": 4801 }, { "epoch": 0.5, "learning_rate": 5.771096670892551e-05, "loss": 0.2053, "step": 4802 }, { "epoch": 0.5, "learning_rate": 5.771002359819114e-05, "loss": 1.0461, "step": 4803 }, { "epoch": 0.5, "learning_rate": 5.770908030091937e-05, "loss": 1.4976, "step": 4804 }, { "epoch": 0.5, "learning_rate": 5.770813681711656e-05, "loss": 0.4106, "step": 4805 }, { "epoch": 0.5, "learning_rate": 5.770719314678905e-05, "loss": 1.0204, "step": 4806 }, { "epoch": 0.5, "learning_rate": 5.77062492899432e-05, "loss": 1.4633, "step": 4807 }, { "epoch": 0.5, "learning_rate": 5.770530524658537e-05, "loss": 1.4264, "step": 4808 }, { "epoch": 0.5, "learning_rate": 5.7704361016721904e-05, "loss": 1.3195, "step": 4809 }, { "epoch": 0.5, "learning_rate": 5.770341660035916e-05, "loss": 0.5102, "step": 4810 }, { "epoch": 0.5, "learning_rate": 5.77024719975035e-05, "loss": 1.1758, "step": 4811 }, { "epoch": 0.5, "learning_rate": 5.7701527208161285e-05, "loss": 0.2424, "step": 4812 }, { "epoch": 0.5, "learning_rate": 5.7700582232338866e-05, "loss": 1.9176, "step": 4813 }, { "epoch": 0.5, "learning_rate": 5.76996370700426e-05, "loss": 1.2509, "step": 4814 }, { "epoch": 0.5, "learning_rate": 5.769869172127887e-05, "loss": 1.0768, "step": 4815 }, { "epoch": 0.5, "learning_rate": 5.769774618605402e-05, "loss": 0.5896, "step": 4816 }, { "epoch": 0.5, "learning_rate": 5.769680046437443e-05, "loss": 0.2457, "step": 4817 }, { "epoch": 0.5, "learning_rate": 5.7695854556246454e-05, "loss": 0.2408, "step": 4818 }, { "epoch": 0.5, "learning_rate": 5.769490846167648e-05, "loss": 1.086, "step": 4819 }, { "epoch": 0.5, "learning_rate": 5.769396218067084e-05, "loss": 0.9424, "step": 4820 }, { "epoch": 0.5, "learning_rate": 5.769301571323594e-05, "loss": 0.5274, "step": 4821 }, { "epoch": 0.5, "learning_rate": 5.769206905937814e-05, "loss": 1.3987, "step": 4822 }, { "epoch": 0.5, "learning_rate": 5.769112221910381e-05, "loss": 1.235, "step": 4823 }, { "epoch": 0.5, "learning_rate": 5.769017519241931e-05, "loss": 0.3863, "step": 4824 }, { "epoch": 0.5, "learning_rate": 5.768922797933104e-05, "loss": 0.9984, "step": 4825 }, { "epoch": 0.5, "learning_rate": 5.7688280579845376e-05, "loss": 0.4111, "step": 4826 }, { "epoch": 0.5, "learning_rate": 5.768733299396867e-05, "loss": 0.8851, "step": 4827 }, { "epoch": 0.5, "learning_rate": 5.7686385221707326e-05, "loss": 1.2538, "step": 4828 }, { "epoch": 0.5, "learning_rate": 5.768543726306771e-05, "loss": 0.4921, "step": 4829 }, { "epoch": 0.5, "learning_rate": 5.768448911805621e-05, "loss": 1.4137, "step": 4830 }, { "epoch": 0.5, "learning_rate": 5.768354078667921e-05, "loss": 0.4567, "step": 4831 }, { "epoch": 0.5, "learning_rate": 5.768259226894308e-05, "loss": 0.6082, "step": 4832 }, { "epoch": 0.5, "learning_rate": 5.7681643564854225e-05, "loss": 0.3696, "step": 4833 }, { "epoch": 0.5, "learning_rate": 5.768069467441902e-05, "loss": 0.6124, "step": 4834 }, { "epoch": 0.5, "learning_rate": 5.7679745597643855e-05, "loss": 0.9079, "step": 4835 }, { "epoch": 0.51, "learning_rate": 5.767879633453512e-05, "loss": 0.3037, "step": 4836 }, { "epoch": 0.51, "learning_rate": 5.767784688509921e-05, "loss": 0.8068, "step": 4837 }, { "epoch": 0.51, "learning_rate": 5.76768972493425e-05, "loss": 0.7407, "step": 4838 }, { "epoch": 0.51, "learning_rate": 5.76759474272714e-05, "loss": 1.6071, "step": 4839 }, { "epoch": 0.51, "learning_rate": 5.7674997418892295e-05, "loss": 0.6006, "step": 4840 }, { "epoch": 0.51, "learning_rate": 5.7674047224211585e-05, "loss": 0.9704, "step": 4841 }, { "epoch": 0.51, "learning_rate": 5.767309684323567e-05, "loss": 0.9317, "step": 4842 }, { "epoch": 0.51, "learning_rate": 5.767214627597093e-05, "loss": 1.1066, "step": 4843 }, { "epoch": 0.51, "learning_rate": 5.767119552242378e-05, "loss": 0.9908, "step": 4844 }, { "epoch": 0.51, "learning_rate": 5.767024458260062e-05, "loss": 1.0212, "step": 4845 }, { "epoch": 0.51, "learning_rate": 5.7669293456507846e-05, "loss": 1.3688, "step": 4846 }, { "epoch": 0.51, "learning_rate": 5.766834214415186e-05, "loss": 0.7049, "step": 4847 }, { "epoch": 0.51, "learning_rate": 5.766739064553908e-05, "loss": 1.0143, "step": 4848 }, { "epoch": 0.51, "learning_rate": 5.7666438960675886e-05, "loss": 0.4377, "step": 4849 }, { "epoch": 0.51, "learning_rate": 5.76654870895687e-05, "loss": 0.3702, "step": 4850 }, { "epoch": 0.51, "learning_rate": 5.766453503222393e-05, "loss": 1.6263, "step": 4851 }, { "epoch": 0.51, "learning_rate": 5.7663582788647995e-05, "loss": 0.2318, "step": 4852 }, { "epoch": 0.51, "learning_rate": 5.766263035884727e-05, "loss": 1.1369, "step": 4853 }, { "epoch": 0.51, "learning_rate": 5.7661677742828204e-05, "loss": 1.5087, "step": 4854 }, { "epoch": 0.51, "learning_rate": 5.76607249405972e-05, "loss": 1.428, "step": 4855 }, { "epoch": 0.51, "learning_rate": 5.765977195216066e-05, "loss": 0.5165, "step": 4856 }, { "epoch": 0.51, "learning_rate": 5.765881877752501e-05, "loss": 0.7007, "step": 4857 }, { "epoch": 0.51, "learning_rate": 5.7657865416696666e-05, "loss": 1.6134, "step": 4858 }, { "epoch": 0.51, "learning_rate": 5.7656911869682036e-05, "loss": 0.7583, "step": 4859 }, { "epoch": 0.51, "learning_rate": 5.765595813648755e-05, "loss": 0.6626, "step": 4860 }, { "epoch": 0.51, "learning_rate": 5.7655004217119624e-05, "loss": 1.3218, "step": 4861 }, { "epoch": 0.51, "learning_rate": 5.765405011158469e-05, "loss": 0.384, "step": 4862 }, { "epoch": 0.51, "learning_rate": 5.765309581988915e-05, "loss": 0.8429, "step": 4863 }, { "epoch": 0.51, "learning_rate": 5.765214134203945e-05, "loss": 0.3192, "step": 4864 }, { "epoch": 0.51, "learning_rate": 5.765118667804199e-05, "loss": 1.4395, "step": 4865 }, { "epoch": 0.51, "learning_rate": 5.765023182790322e-05, "loss": 0.2342, "step": 4866 }, { "epoch": 0.51, "learning_rate": 5.764927679162956e-05, "loss": 1.0999, "step": 4867 }, { "epoch": 0.51, "learning_rate": 5.764832156922743e-05, "loss": 1.0997, "step": 4868 }, { "epoch": 0.51, "learning_rate": 5.764736616070328e-05, "loss": 1.5847, "step": 4869 }, { "epoch": 0.51, "learning_rate": 5.764641056606352e-05, "loss": 0.5156, "step": 4870 }, { "epoch": 0.51, "learning_rate": 5.764545478531459e-05, "loss": 1.3166, "step": 4871 }, { "epoch": 0.51, "learning_rate": 5.764449881846294e-05, "loss": 0.8142, "step": 4872 }, { "epoch": 0.51, "learning_rate": 5.7643542665514986e-05, "loss": 0.5934, "step": 4873 }, { "epoch": 0.51, "learning_rate": 5.764258632647717e-05, "loss": 0.5237, "step": 4874 }, { "epoch": 0.51, "learning_rate": 5.7641629801355936e-05, "loss": 1.0833, "step": 4875 }, { "epoch": 0.51, "learning_rate": 5.764067309015771e-05, "loss": 0.5094, "step": 4876 }, { "epoch": 0.51, "learning_rate": 5.7639716192888945e-05, "loss": 1.1479, "step": 4877 }, { "epoch": 0.51, "learning_rate": 5.763875910955608e-05, "loss": 0.3859, "step": 4878 }, { "epoch": 0.51, "learning_rate": 5.7637801840165545e-05, "loss": 1.1306, "step": 4879 }, { "epoch": 0.51, "learning_rate": 5.7636844384723805e-05, "loss": 1.6296, "step": 4880 }, { "epoch": 0.51, "learning_rate": 5.763588674323729e-05, "loss": 0.7339, "step": 4881 }, { "epoch": 0.51, "learning_rate": 5.7634928915712455e-05, "loss": 0.3302, "step": 4882 }, { "epoch": 0.51, "learning_rate": 5.763397090215574e-05, "loss": 0.3041, "step": 4883 }, { "epoch": 0.51, "learning_rate": 5.76330127025736e-05, "loss": 0.766, "step": 4884 }, { "epoch": 0.51, "learning_rate": 5.763205431697249e-05, "loss": 0.808, "step": 4885 }, { "epoch": 0.51, "learning_rate": 5.7631095745358843e-05, "loss": 1.4639, "step": 4886 }, { "epoch": 0.51, "learning_rate": 5.7630136987739134e-05, "loss": 1.0577, "step": 4887 }, { "epoch": 0.51, "learning_rate": 5.762917804411981e-05, "loss": 1.5374, "step": 4888 }, { "epoch": 0.51, "learning_rate": 5.7628218914507315e-05, "loss": 0.602, "step": 4889 }, { "epoch": 0.51, "learning_rate": 5.762725959890813e-05, "loss": 0.6619, "step": 4890 }, { "epoch": 0.51, "learning_rate": 5.762630009732868e-05, "loss": 1.4152, "step": 4891 }, { "epoch": 0.51, "learning_rate": 5.762534040977545e-05, "loss": 0.4051, "step": 4892 }, { "epoch": 0.51, "learning_rate": 5.7624380536254894e-05, "loss": 0.2639, "step": 4893 }, { "epoch": 0.51, "learning_rate": 5.762342047677347e-05, "loss": 1.2938, "step": 4894 }, { "epoch": 0.51, "learning_rate": 5.7622460231337644e-05, "loss": 1.8521, "step": 4895 }, { "epoch": 0.51, "learning_rate": 5.762149979995388e-05, "loss": 1.1306, "step": 4896 }, { "epoch": 0.51, "learning_rate": 5.762053918262863e-05, "loss": 1.2289, "step": 4897 }, { "epoch": 0.51, "learning_rate": 5.761957837936839e-05, "loss": 1.323, "step": 4898 }, { "epoch": 0.51, "learning_rate": 5.7618617390179606e-05, "loss": 0.4529, "step": 4899 }, { "epoch": 0.51, "learning_rate": 5.761765621506875e-05, "loss": 0.6181, "step": 4900 }, { "epoch": 0.51, "learning_rate": 5.761669485404229e-05, "loss": 0.7664, "step": 4901 }, { "epoch": 0.51, "learning_rate": 5.761573330710671e-05, "loss": 0.515, "step": 4902 }, { "epoch": 0.51, "learning_rate": 5.761477157426847e-05, "loss": 0.4697, "step": 4903 }, { "epoch": 0.51, "learning_rate": 5.761380965553405e-05, "loss": 1.2735, "step": 4904 }, { "epoch": 0.51, "learning_rate": 5.761284755090992e-05, "loss": 0.8095, "step": 4905 }, { "epoch": 0.51, "learning_rate": 5.761188526040257e-05, "loss": 1.2994, "step": 4906 }, { "epoch": 0.51, "learning_rate": 5.761092278401846e-05, "loss": 0.369, "step": 4907 }, { "epoch": 0.51, "learning_rate": 5.7609960121764086e-05, "loss": 0.8719, "step": 4908 }, { "epoch": 0.51, "learning_rate": 5.7608997273645915e-05, "loss": 1.0866, "step": 4909 }, { "epoch": 0.51, "learning_rate": 5.760803423967045e-05, "loss": 0.7867, "step": 4910 }, { "epoch": 0.51, "learning_rate": 5.760707101984414e-05, "loss": 1.4757, "step": 4911 }, { "epoch": 0.51, "learning_rate": 5.760610761417349e-05, "loss": 1.3844, "step": 4912 }, { "epoch": 0.51, "learning_rate": 5.7605144022664984e-05, "loss": 0.7716, "step": 4913 }, { "epoch": 0.51, "learning_rate": 5.760418024532511e-05, "loss": 0.7045, "step": 4914 }, { "epoch": 0.51, "learning_rate": 5.760321628216035e-05, "loss": 0.6658, "step": 4915 }, { "epoch": 0.51, "learning_rate": 5.76022521331772e-05, "loss": 1.1639, "step": 4916 }, { "epoch": 0.51, "learning_rate": 5.7601287798382144e-05, "loss": 0.3685, "step": 4917 }, { "epoch": 0.51, "learning_rate": 5.7600323277781685e-05, "loss": 0.4811, "step": 4918 }, { "epoch": 0.51, "learning_rate": 5.75993585713823e-05, "loss": 1.2719, "step": 4919 }, { "epoch": 0.51, "learning_rate": 5.7598393679190484e-05, "loss": 0.8078, "step": 4920 }, { "epoch": 0.51, "learning_rate": 5.759742860121275e-05, "loss": 0.6245, "step": 4921 }, { "epoch": 0.51, "learning_rate": 5.759646333745559e-05, "loss": 0.5731, "step": 4922 }, { "epoch": 0.51, "learning_rate": 5.759549788792548e-05, "loss": 0.6441, "step": 4923 }, { "epoch": 0.51, "learning_rate": 5.7594532252628945e-05, "loss": 0.3022, "step": 4924 }, { "epoch": 0.51, "learning_rate": 5.759356643157248e-05, "loss": 0.4021, "step": 4925 }, { "epoch": 0.51, "learning_rate": 5.759260042476257e-05, "loss": 0.3871, "step": 4926 }, { "epoch": 0.51, "learning_rate": 5.759163423220573e-05, "loss": 1.5218, "step": 4927 }, { "epoch": 0.51, "learning_rate": 5.759066785390848e-05, "loss": 0.9868, "step": 4928 }, { "epoch": 0.51, "learning_rate": 5.75897012898773e-05, "loss": 0.8339, "step": 4929 }, { "epoch": 0.51, "learning_rate": 5.75887345401187e-05, "loss": 0.5491, "step": 4930 }, { "epoch": 0.51, "learning_rate": 5.7587767604639197e-05, "loss": 1.1096, "step": 4931 }, { "epoch": 0.52, "learning_rate": 5.7586800483445304e-05, "loss": 0.2579, "step": 4932 }, { "epoch": 0.52, "learning_rate": 5.7585833176543523e-05, "loss": 0.9445, "step": 4933 }, { "epoch": 0.52, "learning_rate": 5.758486568394036e-05, "loss": 0.451, "step": 4934 }, { "epoch": 0.52, "learning_rate": 5.758389800564235e-05, "loss": 0.182, "step": 4935 }, { "epoch": 0.52, "learning_rate": 5.758293014165598e-05, "loss": 0.5984, "step": 4936 }, { "epoch": 0.52, "learning_rate": 5.758196209198778e-05, "loss": 1.3615, "step": 4937 }, { "epoch": 0.52, "learning_rate": 5.758099385664427e-05, "loss": 0.6014, "step": 4938 }, { "epoch": 0.52, "learning_rate": 5.758002543563196e-05, "loss": 1.2732, "step": 4939 }, { "epoch": 0.52, "learning_rate": 5.757905682895738e-05, "loss": 1.0615, "step": 4940 }, { "epoch": 0.52, "learning_rate": 5.757808803662703e-05, "loss": 0.65, "step": 4941 }, { "epoch": 0.52, "learning_rate": 5.757711905864744e-05, "loss": 1.2658, "step": 4942 }, { "epoch": 0.52, "learning_rate": 5.7576149895025156e-05, "loss": 1.298, "step": 4943 }, { "epoch": 0.52, "learning_rate": 5.757518054576667e-05, "loss": 0.4269, "step": 4944 }, { "epoch": 0.52, "learning_rate": 5.757421101087852e-05, "loss": 1.0862, "step": 4945 }, { "epoch": 0.52, "learning_rate": 5.757324129036724e-05, "loss": 0.8901, "step": 4946 }, { "epoch": 0.52, "learning_rate": 5.757227138423935e-05, "loss": 0.9251, "step": 4947 }, { "epoch": 0.52, "learning_rate": 5.757130129250137e-05, "loss": 0.7718, "step": 4948 }, { "epoch": 0.52, "learning_rate": 5.7570331015159855e-05, "loss": 1.3213, "step": 4949 }, { "epoch": 0.52, "learning_rate": 5.7569360552221316e-05, "loss": 1.2274, "step": 4950 }, { "epoch": 0.52, "learning_rate": 5.7568389903692295e-05, "loss": 1.139, "step": 4951 }, { "epoch": 0.52, "learning_rate": 5.7567419069579324e-05, "loss": 1.0031, "step": 4952 }, { "epoch": 0.52, "learning_rate": 5.7566448049888936e-05, "loss": 0.3146, "step": 4953 }, { "epoch": 0.52, "learning_rate": 5.756547684462767e-05, "loss": 1.4739, "step": 4954 }, { "epoch": 0.52, "learning_rate": 5.756450545380206e-05, "loss": 1.6139, "step": 4955 }, { "epoch": 0.52, "learning_rate": 5.756353387741866e-05, "loss": 0.94, "step": 4956 }, { "epoch": 0.52, "learning_rate": 5.756256211548399e-05, "loss": 1.3471, "step": 4957 }, { "epoch": 0.52, "learning_rate": 5.7561590168004595e-05, "loss": 1.7169, "step": 4958 }, { "epoch": 0.52, "learning_rate": 5.756061803498703e-05, "loss": 1.0698, "step": 4959 }, { "epoch": 0.52, "learning_rate": 5.7559645716437834e-05, "loss": 1.242, "step": 4960 }, { "epoch": 0.52, "learning_rate": 5.7558673212363555e-05, "loss": 0.9779, "step": 4961 }, { "epoch": 0.52, "learning_rate": 5.755770052277074e-05, "loss": 1.1297, "step": 4962 }, { "epoch": 0.52, "learning_rate": 5.7556727647665926e-05, "loss": 1.1771, "step": 4963 }, { "epoch": 0.52, "learning_rate": 5.755575458705566e-05, "loss": 1.3656, "step": 4964 }, { "epoch": 0.52, "learning_rate": 5.755478134094652e-05, "loss": 0.8479, "step": 4965 }, { "epoch": 0.52, "learning_rate": 5.755380790934503e-05, "loss": 1.5368, "step": 4966 }, { "epoch": 0.52, "learning_rate": 5.755283429225775e-05, "loss": 1.3311, "step": 4967 }, { "epoch": 0.52, "learning_rate": 5.755186048969124e-05, "loss": 1.3312, "step": 4968 }, { "epoch": 0.52, "learning_rate": 5.755088650165205e-05, "loss": 1.2338, "step": 4969 }, { "epoch": 0.52, "learning_rate": 5.754991232814674e-05, "loss": 0.7154, "step": 4970 }, { "epoch": 0.52, "learning_rate": 5.754893796918186e-05, "loss": 1.2546, "step": 4971 }, { "epoch": 0.52, "learning_rate": 5.7547963424763995e-05, "loss": 1.2878, "step": 4972 }, { "epoch": 0.52, "learning_rate": 5.754698869489966e-05, "loss": 1.4736, "step": 4973 }, { "epoch": 0.52, "learning_rate": 5.754601377959545e-05, "loss": 0.7309, "step": 4974 }, { "epoch": 0.52, "learning_rate": 5.754503867885793e-05, "loss": 1.5455, "step": 4975 }, { "epoch": 0.52, "learning_rate": 5.754406339269366e-05, "loss": 0.5171, "step": 4976 }, { "epoch": 0.52, "learning_rate": 5.754308792110919e-05, "loss": 0.4803, "step": 4977 }, { "epoch": 0.52, "learning_rate": 5.754211226411109e-05, "loss": 1.1124, "step": 4978 }, { "epoch": 0.52, "learning_rate": 5.7541136421705946e-05, "loss": 1.0098, "step": 4979 }, { "epoch": 0.52, "learning_rate": 5.754016039390031e-05, "loss": 0.7014, "step": 4980 }, { "epoch": 0.52, "learning_rate": 5.753918418070076e-05, "loss": 0.5441, "step": 4981 }, { "epoch": 0.52, "learning_rate": 5.7538207782113865e-05, "loss": 1.1347, "step": 4982 }, { "epoch": 0.52, "learning_rate": 5.75372311981462e-05, "loss": 0.3553, "step": 4983 }, { "epoch": 0.52, "learning_rate": 5.753625442880434e-05, "loss": 0.8721, "step": 4984 }, { "epoch": 0.52, "learning_rate": 5.7535277474094845e-05, "loss": 0.6571, "step": 4985 }, { "epoch": 0.52, "learning_rate": 5.7534300334024315e-05, "loss": 0.9138, "step": 4986 }, { "epoch": 0.52, "learning_rate": 5.753332300859932e-05, "loss": 1.1118, "step": 4987 }, { "epoch": 0.52, "learning_rate": 5.753234549782644e-05, "loss": 0.7545, "step": 4988 }, { "epoch": 0.52, "learning_rate": 5.753136780171224e-05, "loss": 1.8329, "step": 4989 }, { "epoch": 0.52, "learning_rate": 5.753038992026333e-05, "loss": 0.9505, "step": 4990 }, { "epoch": 0.52, "learning_rate": 5.7529411853486256e-05, "loss": 0.4419, "step": 4991 }, { "epoch": 0.52, "learning_rate": 5.7528433601387626e-05, "loss": 0.5811, "step": 4992 }, { "epoch": 0.52, "learning_rate": 5.752745516397403e-05, "loss": 0.8934, "step": 4993 }, { "epoch": 0.52, "learning_rate": 5.7526476541252054e-05, "loss": 1.4426, "step": 4994 }, { "epoch": 0.52, "learning_rate": 5.7525497733228265e-05, "loss": 0.3792, "step": 4995 }, { "epoch": 0.52, "learning_rate": 5.752451873990928e-05, "loss": 1.3267, "step": 4996 }, { "epoch": 0.52, "learning_rate": 5.752353956130167e-05, "loss": 1.5143, "step": 4997 }, { "epoch": 0.52, "learning_rate": 5.752256019741202e-05, "loss": 0.7883, "step": 4998 }, { "epoch": 0.52, "learning_rate": 5.752158064824694e-05, "loss": 0.6927, "step": 4999 }, { "epoch": 0.52, "learning_rate": 5.7520600913813024e-05, "loss": 1.1995, "step": 5000 }, { "epoch": 0.52, "learning_rate": 5.7519620994116856e-05, "loss": 0.6424, "step": 5001 }, { "epoch": 0.52, "learning_rate": 5.751864088916504e-05, "loss": 1.297, "step": 5002 }, { "epoch": 0.52, "learning_rate": 5.751766059896418e-05, "loss": 0.8915, "step": 5003 }, { "epoch": 0.52, "learning_rate": 5.751668012352086e-05, "loss": 0.7515, "step": 5004 }, { "epoch": 0.52, "learning_rate": 5.751569946284169e-05, "loss": 0.3569, "step": 5005 }, { "epoch": 0.52, "learning_rate": 5.751471861693327e-05, "loss": 0.7895, "step": 5006 }, { "epoch": 0.52, "learning_rate": 5.75137375858022e-05, "loss": 1.2913, "step": 5007 }, { "epoch": 0.52, "learning_rate": 5.7512756369455086e-05, "loss": 1.104, "step": 5008 }, { "epoch": 0.52, "learning_rate": 5.7511774967898534e-05, "loss": 0.6746, "step": 5009 }, { "epoch": 0.52, "learning_rate": 5.751079338113916e-05, "loss": 0.5703, "step": 5010 }, { "epoch": 0.52, "learning_rate": 5.750981160918354e-05, "loss": 0.607, "step": 5011 }, { "epoch": 0.52, "learning_rate": 5.750882965203832e-05, "loss": 0.4806, "step": 5012 }, { "epoch": 0.52, "learning_rate": 5.750784750971009e-05, "loss": 0.7184, "step": 5013 }, { "epoch": 0.52, "learning_rate": 5.7506865182205475e-05, "loss": 0.6134, "step": 5014 }, { "epoch": 0.52, "learning_rate": 5.750588266953107e-05, "loss": 1.1611, "step": 5015 }, { "epoch": 0.52, "learning_rate": 5.750489997169351e-05, "loss": 0.6058, "step": 5016 }, { "epoch": 0.52, "learning_rate": 5.750391708869939e-05, "loss": 0.897, "step": 5017 }, { "epoch": 0.52, "learning_rate": 5.7502934020555336e-05, "loss": 1.4644, "step": 5018 }, { "epoch": 0.52, "learning_rate": 5.750195076726797e-05, "loss": 1.339, "step": 5019 }, { "epoch": 0.52, "learning_rate": 5.7500967328843906e-05, "loss": 1.137, "step": 5020 }, { "epoch": 0.52, "learning_rate": 5.7499983705289765e-05, "loss": 0.3546, "step": 5021 }, { "epoch": 0.52, "learning_rate": 5.749899989661216e-05, "loss": 1.1166, "step": 5022 }, { "epoch": 0.52, "learning_rate": 5.749801590281773e-05, "loss": 0.7982, "step": 5023 }, { "epoch": 0.52, "learning_rate": 5.749703172391309e-05, "loss": 0.8489, "step": 5024 }, { "epoch": 0.52, "learning_rate": 5.7496047359904863e-05, "loss": 1.1865, "step": 5025 }, { "epoch": 0.52, "learning_rate": 5.749506281079968e-05, "loss": 0.3672, "step": 5026 }, { "epoch": 0.53, "learning_rate": 5.7494078076604164e-05, "loss": 1.2389, "step": 5027 }, { "epoch": 0.53, "learning_rate": 5.749309315732495e-05, "loss": 1.2109, "step": 5028 }, { "epoch": 0.53, "learning_rate": 5.749210805296868e-05, "loss": 1.4409, "step": 5029 }, { "epoch": 0.53, "learning_rate": 5.749112276354195e-05, "loss": 0.3386, "step": 5030 }, { "epoch": 0.53, "learning_rate": 5.7490137289051425e-05, "loss": 0.9808, "step": 5031 }, { "epoch": 0.53, "learning_rate": 5.7489151629503724e-05, "loss": 1.0028, "step": 5032 }, { "epoch": 0.53, "learning_rate": 5.748816578490549e-05, "loss": 0.3441, "step": 5033 }, { "epoch": 0.53, "learning_rate": 5.7487179755263355e-05, "loss": 1.1422, "step": 5034 }, { "epoch": 0.53, "learning_rate": 5.748619354058395e-05, "loss": 0.5021, "step": 5035 }, { "epoch": 0.53, "learning_rate": 5.748520714087393e-05, "loss": 0.5895, "step": 5036 }, { "epoch": 0.53, "learning_rate": 5.748422055613992e-05, "loss": 0.6145, "step": 5037 }, { "epoch": 0.53, "learning_rate": 5.7483233786388576e-05, "loss": 0.3172, "step": 5038 }, { "epoch": 0.53, "learning_rate": 5.7482246831626525e-05, "loss": 0.9038, "step": 5039 }, { "epoch": 0.53, "learning_rate": 5.7481259691860424e-05, "loss": 0.7435, "step": 5040 }, { "epoch": 0.53, "learning_rate": 5.748027236709691e-05, "loss": 0.9362, "step": 5041 }, { "epoch": 0.53, "learning_rate": 5.747928485734263e-05, "loss": 0.9954, "step": 5042 }, { "epoch": 0.53, "learning_rate": 5.7478297162604244e-05, "loss": 0.8255, "step": 5043 }, { "epoch": 0.53, "learning_rate": 5.747730928288838e-05, "loss": 0.9047, "step": 5044 }, { "epoch": 0.53, "learning_rate": 5.7476321218201706e-05, "loss": 0.7803, "step": 5045 }, { "epoch": 0.53, "learning_rate": 5.747533296855086e-05, "loss": 1.5333, "step": 5046 }, { "epoch": 0.53, "learning_rate": 5.7474344533942506e-05, "loss": 0.5862, "step": 5047 }, { "epoch": 0.53, "learning_rate": 5.747335591438329e-05, "loss": 0.7617, "step": 5048 }, { "epoch": 0.53, "learning_rate": 5.747236710987987e-05, "loss": 0.8766, "step": 5049 }, { "epoch": 0.53, "learning_rate": 5.747137812043892e-05, "loss": 0.2609, "step": 5050 }, { "epoch": 0.53, "learning_rate": 5.7470388946067055e-05, "loss": 0.9858, "step": 5051 }, { "epoch": 0.53, "learning_rate": 5.746939958677097e-05, "loss": 1.3165, "step": 5052 }, { "epoch": 0.53, "learning_rate": 5.746841004255731e-05, "loss": 1.1814, "step": 5053 }, { "epoch": 0.53, "learning_rate": 5.7467420313432744e-05, "loss": 0.4174, "step": 5054 }, { "epoch": 0.53, "learning_rate": 5.746643039940393e-05, "loss": 1.1996, "step": 5055 }, { "epoch": 0.53, "learning_rate": 5.746544030047754e-05, "loss": 1.3824, "step": 5056 }, { "epoch": 0.53, "learning_rate": 5.7464450016660225e-05, "loss": 0.308, "step": 5057 }, { "epoch": 0.53, "learning_rate": 5.7463459547958655e-05, "loss": 1.0588, "step": 5058 }, { "epoch": 0.53, "learning_rate": 5.7462468894379514e-05, "loss": 1.1838, "step": 5059 }, { "epoch": 0.53, "learning_rate": 5.746147805592945e-05, "loss": 0.353, "step": 5060 }, { "epoch": 0.53, "learning_rate": 5.746048703261514e-05, "loss": 0.8826, "step": 5061 }, { "epoch": 0.53, "learning_rate": 5.745949582444326e-05, "loss": 0.6167, "step": 5062 }, { "epoch": 0.53, "learning_rate": 5.745850443142047e-05, "loss": 0.5875, "step": 5063 }, { "epoch": 0.53, "learning_rate": 5.7457512853553464e-05, "loss": 0.5506, "step": 5064 }, { "epoch": 0.53, "learning_rate": 5.7456521090848904e-05, "loss": 1.4258, "step": 5065 }, { "epoch": 0.53, "learning_rate": 5.745552914331347e-05, "loss": 1.0724, "step": 5066 }, { "epoch": 0.53, "learning_rate": 5.7454537010953835e-05, "loss": 1.3734, "step": 5067 }, { "epoch": 0.53, "learning_rate": 5.7453544693776676e-05, "loss": 0.8369, "step": 5068 }, { "epoch": 0.53, "learning_rate": 5.7452552191788684e-05, "loss": 1.1017, "step": 5069 }, { "epoch": 0.53, "learning_rate": 5.7451559504996535e-05, "loss": 0.8216, "step": 5070 }, { "epoch": 0.53, "learning_rate": 5.74505666334069e-05, "loss": 0.7584, "step": 5071 }, { "epoch": 0.53, "learning_rate": 5.744957357702649e-05, "loss": 1.3298, "step": 5072 }, { "epoch": 0.53, "learning_rate": 5.744858033586197e-05, "loss": 1.1668, "step": 5073 }, { "epoch": 0.53, "learning_rate": 5.744758690992002e-05, "loss": 1.032, "step": 5074 }, { "epoch": 0.53, "learning_rate": 5.744659329920734e-05, "loss": 1.1636, "step": 5075 }, { "epoch": 0.53, "learning_rate": 5.744559950373062e-05, "loss": 1.1378, "step": 5076 }, { "epoch": 0.53, "learning_rate": 5.744460552349654e-05, "loss": 0.4119, "step": 5077 }, { "epoch": 0.53, "learning_rate": 5.74436113585118e-05, "loss": 0.9117, "step": 5078 }, { "epoch": 0.53, "learning_rate": 5.74426170087831e-05, "loss": 0.6298, "step": 5079 }, { "epoch": 0.53, "learning_rate": 5.74416224743171e-05, "loss": 0.7127, "step": 5080 }, { "epoch": 0.53, "learning_rate": 5.7440627755120536e-05, "loss": 1.4629, "step": 5081 }, { "epoch": 0.53, "learning_rate": 5.7439632851200084e-05, "loss": 0.6325, "step": 5082 }, { "epoch": 0.53, "learning_rate": 5.743863776256245e-05, "loss": 1.241, "step": 5083 }, { "epoch": 0.53, "learning_rate": 5.743764248921431e-05, "loss": 0.4021, "step": 5084 }, { "epoch": 0.53, "learning_rate": 5.743664703116238e-05, "loss": 1.0414, "step": 5085 }, { "epoch": 0.53, "learning_rate": 5.743565138841338e-05, "loss": 1.1519, "step": 5086 }, { "epoch": 0.53, "learning_rate": 5.743465556097399e-05, "loss": 0.8434, "step": 5087 }, { "epoch": 0.53, "learning_rate": 5.74336595488509e-05, "loss": 0.7391, "step": 5088 }, { "epoch": 0.53, "learning_rate": 5.7432663352050856e-05, "loss": 0.2504, "step": 5089 }, { "epoch": 0.53, "learning_rate": 5.743166697058054e-05, "loss": 0.2479, "step": 5090 }, { "epoch": 0.53, "learning_rate": 5.743067040444664e-05, "loss": 0.7937, "step": 5091 }, { "epoch": 0.53, "learning_rate": 5.7429673653655904e-05, "loss": 0.6338, "step": 5092 }, { "epoch": 0.53, "learning_rate": 5.7428676718215013e-05, "loss": 0.3347, "step": 5093 }, { "epoch": 0.53, "learning_rate": 5.742767959813069e-05, "loss": 1.6127, "step": 5094 }, { "epoch": 0.53, "learning_rate": 5.742668229340965e-05, "loss": 1.3043, "step": 5095 }, { "epoch": 0.53, "learning_rate": 5.74256848040586e-05, "loss": 0.5689, "step": 5096 }, { "epoch": 0.53, "learning_rate": 5.742468713008425e-05, "loss": 0.7142, "step": 5097 }, { "epoch": 0.53, "learning_rate": 5.742368927149333e-05, "loss": 1.3038, "step": 5098 }, { "epoch": 0.53, "learning_rate": 5.7422691228292554e-05, "loss": 1.2234, "step": 5099 }, { "epoch": 0.53, "learning_rate": 5.742169300048863e-05, "loss": 0.9217, "step": 5100 }, { "epoch": 0.53, "learning_rate": 5.7420694588088295e-05, "loss": 0.4535, "step": 5101 }, { "epoch": 0.53, "learning_rate": 5.7419695991098244e-05, "loss": 1.34, "step": 5102 }, { "epoch": 0.53, "learning_rate": 5.7418697209525226e-05, "loss": 1.0156, "step": 5103 }, { "epoch": 0.53, "learning_rate": 5.741769824337595e-05, "loss": 0.3944, "step": 5104 }, { "epoch": 0.53, "learning_rate": 5.7416699092657143e-05, "loss": 1.5705, "step": 5105 }, { "epoch": 0.53, "learning_rate": 5.741569975737554e-05, "loss": 0.3407, "step": 5106 }, { "epoch": 0.53, "learning_rate": 5.741470023753785e-05, "loss": 0.7793, "step": 5107 }, { "epoch": 0.53, "learning_rate": 5.741370053315083e-05, "loss": 1.5584, "step": 5108 }, { "epoch": 0.53, "learning_rate": 5.741270064422118e-05, "loss": 1.0419, "step": 5109 }, { "epoch": 0.53, "learning_rate": 5.7411700570755646e-05, "loss": 0.7841, "step": 5110 }, { "epoch": 0.53, "learning_rate": 5.7410700312760955e-05, "loss": 0.9672, "step": 5111 }, { "epoch": 0.53, "learning_rate": 5.740969987024384e-05, "loss": 1.3568, "step": 5112 }, { "epoch": 0.53, "learning_rate": 5.7408699243211046e-05, "loss": 1.1944, "step": 5113 }, { "epoch": 0.53, "learning_rate": 5.74076984316693e-05, "loss": 1.132, "step": 5114 }, { "epoch": 0.53, "learning_rate": 5.740669743562534e-05, "loss": 1.4968, "step": 5115 }, { "epoch": 0.53, "learning_rate": 5.7405696255085904e-05, "loss": 1.0386, "step": 5116 }, { "epoch": 0.53, "learning_rate": 5.740469489005774e-05, "loss": 0.9688, "step": 5117 }, { "epoch": 0.53, "learning_rate": 5.740369334054758e-05, "loss": 0.8422, "step": 5118 }, { "epoch": 0.53, "learning_rate": 5.7402691606562164e-05, "loss": 1.0179, "step": 5119 }, { "epoch": 0.53, "learning_rate": 5.740168968810824e-05, "loss": 0.9867, "step": 5120 }, { "epoch": 0.53, "learning_rate": 5.740068758519255e-05, "loss": 1.3259, "step": 5121 }, { "epoch": 0.53, "learning_rate": 5.739968529782185e-05, "loss": 1.2574, "step": 5122 }, { "epoch": 0.54, "learning_rate": 5.739868282600288e-05, "loss": 0.7124, "step": 5123 }, { "epoch": 0.54, "learning_rate": 5.739768016974239e-05, "loss": 0.3479, "step": 5124 }, { "epoch": 0.54, "learning_rate": 5.739667732904711e-05, "loss": 0.8542, "step": 5125 }, { "epoch": 0.54, "learning_rate": 5.739567430392382e-05, "loss": 1.2368, "step": 5126 }, { "epoch": 0.54, "learning_rate": 5.7394671094379266e-05, "loss": 0.5503, "step": 5127 }, { "epoch": 0.54, "learning_rate": 5.7393667700420186e-05, "loss": 0.5219, "step": 5128 }, { "epoch": 0.54, "learning_rate": 5.739266412205335e-05, "loss": 0.4465, "step": 5129 }, { "epoch": 0.54, "learning_rate": 5.739166035928551e-05, "loss": 0.9982, "step": 5130 }, { "epoch": 0.54, "learning_rate": 5.7390656412123425e-05, "loss": 0.7188, "step": 5131 }, { "epoch": 0.54, "learning_rate": 5.7389652280573835e-05, "loss": 1.7714, "step": 5132 }, { "epoch": 0.54, "learning_rate": 5.738864796464353e-05, "loss": 0.9442, "step": 5133 }, { "epoch": 0.54, "learning_rate": 5.738764346433926e-05, "loss": 0.8157, "step": 5134 }, { "epoch": 0.54, "learning_rate": 5.7386638779667754e-05, "loss": 1.3332, "step": 5135 }, { "epoch": 0.54, "learning_rate": 5.738563391063582e-05, "loss": 0.7356, "step": 5136 }, { "epoch": 0.54, "learning_rate": 5.7384628857250214e-05, "loss": 0.4885, "step": 5137 }, { "epoch": 0.54, "learning_rate": 5.738362361951769e-05, "loss": 1.7441, "step": 5138 }, { "epoch": 0.54, "learning_rate": 5.738261819744502e-05, "loss": 0.1265, "step": 5139 }, { "epoch": 0.54, "learning_rate": 5.738161259103896e-05, "loss": 1.3476, "step": 5140 }, { "epoch": 0.54, "learning_rate": 5.738060680030631e-05, "loss": 1.1023, "step": 5141 }, { "epoch": 0.54, "learning_rate": 5.737960082525381e-05, "loss": 0.5581, "step": 5142 }, { "epoch": 0.54, "learning_rate": 5.737859466588823e-05, "loss": 0.5938, "step": 5143 }, { "epoch": 0.54, "learning_rate": 5.737758832221638e-05, "loss": 1.0529, "step": 5144 }, { "epoch": 0.54, "learning_rate": 5.7376581794245004e-05, "loss": 1.2221, "step": 5145 }, { "epoch": 0.54, "learning_rate": 5.737557508198088e-05, "loss": 1.2504, "step": 5146 }, { "epoch": 0.54, "learning_rate": 5.7374568185430796e-05, "loss": 1.1667, "step": 5147 }, { "epoch": 0.54, "learning_rate": 5.737356110460152e-05, "loss": 0.376, "step": 5148 }, { "epoch": 0.54, "learning_rate": 5.7372553839499834e-05, "loss": 0.4164, "step": 5149 }, { "epoch": 0.54, "learning_rate": 5.737154639013253e-05, "loss": 1.376, "step": 5150 }, { "epoch": 0.54, "learning_rate": 5.737053875650637e-05, "loss": 1.3326, "step": 5151 }, { "epoch": 0.54, "learning_rate": 5.7369530938628145e-05, "loss": 1.4313, "step": 5152 }, { "epoch": 0.54, "learning_rate": 5.736852293650465e-05, "loss": 1.6347, "step": 5153 }, { "epoch": 0.54, "learning_rate": 5.7367514750142665e-05, "loss": 0.2294, "step": 5154 }, { "epoch": 0.54, "learning_rate": 5.736650637954896e-05, "loss": 0.8231, "step": 5155 }, { "epoch": 0.54, "learning_rate": 5.736549782473035e-05, "loss": 1.1354, "step": 5156 }, { "epoch": 0.54, "learning_rate": 5.736448908569361e-05, "loss": 0.6562, "step": 5157 }, { "epoch": 0.54, "learning_rate": 5.736348016244553e-05, "loss": 1.1022, "step": 5158 }, { "epoch": 0.54, "learning_rate": 5.73624710549929e-05, "loss": 0.4132, "step": 5159 }, { "epoch": 0.54, "learning_rate": 5.7361461763342524e-05, "loss": 2.0833, "step": 5160 }, { "epoch": 0.54, "learning_rate": 5.7360452287501186e-05, "loss": 2.3342, "step": 5161 }, { "epoch": 0.54, "learning_rate": 5.7359442627475685e-05, "loss": 1.1037, "step": 5162 }, { "epoch": 0.54, "learning_rate": 5.735843278327282e-05, "loss": 0.8055, "step": 5163 }, { "epoch": 0.54, "learning_rate": 5.7357422754899386e-05, "loss": 0.8283, "step": 5164 }, { "epoch": 0.54, "learning_rate": 5.735641254236218e-05, "loss": 1.0191, "step": 5165 }, { "epoch": 0.54, "learning_rate": 5.735540214566801e-05, "loss": 0.568, "step": 5166 }, { "epoch": 0.54, "learning_rate": 5.735439156482367e-05, "loss": 1.1482, "step": 5167 }, { "epoch": 0.54, "learning_rate": 5.7353380799835964e-05, "loss": 1.125, "step": 5168 }, { "epoch": 0.54, "learning_rate": 5.735236985071171e-05, "loss": 1.3853, "step": 5169 }, { "epoch": 0.54, "learning_rate": 5.735135871745769e-05, "loss": 1.259, "step": 5170 }, { "epoch": 0.54, "learning_rate": 5.7350347400080726e-05, "loss": 0.3643, "step": 5171 }, { "epoch": 0.54, "learning_rate": 5.7349335898587634e-05, "loss": 0.8942, "step": 5172 }, { "epoch": 0.54, "learning_rate": 5.7348324212985195e-05, "loss": 1.1811, "step": 5173 }, { "epoch": 0.54, "learning_rate": 5.7347312343280246e-05, "loss": 1.4478, "step": 5174 }, { "epoch": 0.54, "learning_rate": 5.7346300289479585e-05, "loss": 0.6298, "step": 5175 }, { "epoch": 0.54, "learning_rate": 5.734528805159003e-05, "loss": 0.4134, "step": 5176 }, { "epoch": 0.54, "learning_rate": 5.73442756296184e-05, "loss": 0.8509, "step": 5177 }, { "epoch": 0.54, "learning_rate": 5.734326302357149e-05, "loss": 0.7257, "step": 5178 }, { "epoch": 0.54, "learning_rate": 5.734225023345614e-05, "loss": 1.38, "step": 5179 }, { "epoch": 0.54, "learning_rate": 5.7341237259279164e-05, "loss": 1.0129, "step": 5180 }, { "epoch": 0.54, "learning_rate": 5.7340224101047374e-05, "loss": 0.8239, "step": 5181 }, { "epoch": 0.54, "learning_rate": 5.733921075876758e-05, "loss": 0.6485, "step": 5182 }, { "epoch": 0.54, "learning_rate": 5.733819723244663e-05, "loss": 1.3863, "step": 5183 }, { "epoch": 0.54, "learning_rate": 5.733718352209133e-05, "loss": 0.8662, "step": 5184 }, { "epoch": 0.54, "learning_rate": 5.733616962770851e-05, "loss": 0.2261, "step": 5185 }, { "epoch": 0.54, "learning_rate": 5.733515554930498e-05, "loss": 1.0421, "step": 5186 }, { "epoch": 0.54, "learning_rate": 5.733414128688759e-05, "loss": 1.0202, "step": 5187 }, { "epoch": 0.54, "learning_rate": 5.7333126840463156e-05, "loss": 1.4377, "step": 5188 }, { "epoch": 0.54, "learning_rate": 5.73321122100385e-05, "loss": 1.2012, "step": 5189 }, { "epoch": 0.54, "learning_rate": 5.733109739562047e-05, "loss": 1.4877, "step": 5190 }, { "epoch": 0.54, "learning_rate": 5.7330082397215876e-05, "loss": 1.2814, "step": 5191 }, { "epoch": 0.54, "learning_rate": 5.732906721483157e-05, "loss": 1.0104, "step": 5192 }, { "epoch": 0.54, "learning_rate": 5.7328051848474384e-05, "loss": 0.9408, "step": 5193 }, { "epoch": 0.54, "learning_rate": 5.732703629815113e-05, "loss": 1.2018, "step": 5194 }, { "epoch": 0.54, "learning_rate": 5.7326020563868675e-05, "loss": 0.9051, "step": 5195 }, { "epoch": 0.54, "learning_rate": 5.732500464563384e-05, "loss": 0.3852, "step": 5196 }, { "epoch": 0.54, "learning_rate": 5.732398854345346e-05, "loss": 0.1766, "step": 5197 }, { "epoch": 0.54, "learning_rate": 5.732297225733439e-05, "loss": 1.3436, "step": 5198 }, { "epoch": 0.54, "learning_rate": 5.732195578728347e-05, "loss": 1.0214, "step": 5199 }, { "epoch": 0.54, "learning_rate": 5.7320939133307525e-05, "loss": 0.4687, "step": 5200 }, { "epoch": 0.54, "learning_rate": 5.731992229541341e-05, "loss": 0.5818, "step": 5201 }, { "epoch": 0.54, "learning_rate": 5.731890527360798e-05, "loss": 0.9402, "step": 5202 }, { "epoch": 0.54, "learning_rate": 5.731788806789807e-05, "loss": 0.9677, "step": 5203 }, { "epoch": 0.54, "learning_rate": 5.7316870678290524e-05, "loss": 1.3467, "step": 5204 }, { "epoch": 0.54, "learning_rate": 5.73158531047922e-05, "loss": 1.6825, "step": 5205 }, { "epoch": 0.54, "learning_rate": 5.7314835347409945e-05, "loss": 0.8487, "step": 5206 }, { "epoch": 0.54, "learning_rate": 5.7313817406150604e-05, "loss": 1.0878, "step": 5207 }, { "epoch": 0.54, "learning_rate": 5.731279928102104e-05, "loss": 1.2774, "step": 5208 }, { "epoch": 0.54, "learning_rate": 5.7311780972028096e-05, "loss": 0.562, "step": 5209 }, { "epoch": 0.54, "learning_rate": 5.731076247917864e-05, "loss": 1.3754, "step": 5210 }, { "epoch": 0.54, "learning_rate": 5.730974380247952e-05, "loss": 0.696, "step": 5211 }, { "epoch": 0.54, "learning_rate": 5.730872494193759e-05, "loss": 0.3427, "step": 5212 }, { "epoch": 0.54, "learning_rate": 5.730770589755972e-05, "loss": 1.3831, "step": 5213 }, { "epoch": 0.54, "learning_rate": 5.7306686669352757e-05, "loss": 0.9797, "step": 5214 }, { "epoch": 0.54, "learning_rate": 5.7305667257323574e-05, "loss": 0.4993, "step": 5215 }, { "epoch": 0.54, "learning_rate": 5.7304647661479025e-05, "loss": 0.5799, "step": 5216 }, { "epoch": 0.54, "learning_rate": 5.7303627881825976e-05, "loss": 0.3746, "step": 5217 }, { "epoch": 0.54, "learning_rate": 5.73026079183713e-05, "loss": 0.7641, "step": 5218 }, { "epoch": 0.55, "learning_rate": 5.7301587771121846e-05, "loss": 1.5079, "step": 5219 }, { "epoch": 0.55, "learning_rate": 5.7300567440084494e-05, "loss": 1.2814, "step": 5220 }, { "epoch": 0.55, "learning_rate": 5.729954692526611e-05, "loss": 0.7434, "step": 5221 }, { "epoch": 0.55, "learning_rate": 5.7298526226673564e-05, "loss": 0.923, "step": 5222 }, { "epoch": 0.55, "learning_rate": 5.729750534431373e-05, "loss": 1.6735, "step": 5223 }, { "epoch": 0.55, "learning_rate": 5.7296484278193476e-05, "loss": 0.6113, "step": 5224 }, { "epoch": 0.55, "learning_rate": 5.729546302831967e-05, "loss": 1.1999, "step": 5225 }, { "epoch": 0.55, "learning_rate": 5.72944415946992e-05, "loss": 0.6483, "step": 5226 }, { "epoch": 0.55, "learning_rate": 5.729341997733893e-05, "loss": 0.6811, "step": 5227 }, { "epoch": 0.55, "learning_rate": 5.729239817624575e-05, "loss": 1.3284, "step": 5228 }, { "epoch": 0.55, "learning_rate": 5.729137619142653e-05, "loss": 1.1562, "step": 5229 }, { "epoch": 0.55, "learning_rate": 5.729035402288815e-05, "loss": 0.7183, "step": 5230 }, { "epoch": 0.55, "learning_rate": 5.728933167063749e-05, "loss": 0.4606, "step": 5231 }, { "epoch": 0.55, "learning_rate": 5.728830913468143e-05, "loss": 1.2003, "step": 5232 }, { "epoch": 0.55, "learning_rate": 5.728728641502687e-05, "loss": 0.423, "step": 5233 }, { "epoch": 0.55, "learning_rate": 5.7286263511680683e-05, "loss": 1.1137, "step": 5234 }, { "epoch": 0.55, "learning_rate": 5.7285240424649744e-05, "loss": 0.9017, "step": 5235 }, { "epoch": 0.55, "learning_rate": 5.7284217153940956e-05, "loss": 0.7158, "step": 5236 }, { "epoch": 0.55, "learning_rate": 5.72831936995612e-05, "loss": 1.1224, "step": 5237 }, { "epoch": 0.55, "learning_rate": 5.728217006151737e-05, "loss": 0.2327, "step": 5238 }, { "epoch": 0.55, "learning_rate": 5.7281146239816354e-05, "loss": 1.029, "step": 5239 }, { "epoch": 0.55, "learning_rate": 5.7280122234465036e-05, "loss": 1.4157, "step": 5240 }, { "epoch": 0.55, "learning_rate": 5.727909804547033e-05, "loss": 0.8142, "step": 5241 }, { "epoch": 0.55, "learning_rate": 5.727807367283912e-05, "loss": 0.8801, "step": 5242 }, { "epoch": 0.55, "learning_rate": 5.727704911657829e-05, "loss": 1.1424, "step": 5243 }, { "epoch": 0.55, "learning_rate": 5.727602437669476e-05, "loss": 0.4837, "step": 5244 }, { "epoch": 0.55, "learning_rate": 5.7274999453195404e-05, "loss": 0.4408, "step": 5245 }, { "epoch": 0.55, "learning_rate": 5.727397434608714e-05, "loss": 1.2891, "step": 5246 }, { "epoch": 0.55, "learning_rate": 5.7272949055376865e-05, "loss": 0.4092, "step": 5247 }, { "epoch": 0.55, "learning_rate": 5.727192358107147e-05, "loss": 0.6317, "step": 5248 }, { "epoch": 0.55, "learning_rate": 5.7270897923177875e-05, "loss": 0.5861, "step": 5249 }, { "epoch": 0.55, "learning_rate": 5.726987208170297e-05, "loss": 1.412, "step": 5250 }, { "epoch": 0.55, "learning_rate": 5.7268846056653665e-05, "loss": 0.5544, "step": 5251 }, { "epoch": 0.55, "learning_rate": 5.726781984803688e-05, "loss": 0.3583, "step": 5252 }, { "epoch": 0.55, "learning_rate": 5.72667934558595e-05, "loss": 0.7296, "step": 5253 }, { "epoch": 0.55, "learning_rate": 5.726576688012846e-05, "loss": 1.3246, "step": 5254 }, { "epoch": 0.55, "learning_rate": 5.726474012085064e-05, "loss": 0.57, "step": 5255 }, { "epoch": 0.55, "learning_rate": 5.726371317803298e-05, "loss": 0.7935, "step": 5256 }, { "epoch": 0.55, "learning_rate": 5.726268605168238e-05, "loss": 0.8104, "step": 5257 }, { "epoch": 0.55, "learning_rate": 5.7261658741805745e-05, "loss": 0.9844, "step": 5258 }, { "epoch": 0.55, "learning_rate": 5.726063124841002e-05, "loss": 0.0175, "step": 5259 }, { "epoch": 0.55, "learning_rate": 5.725960357150209e-05, "loss": 1.2695, "step": 5260 }, { "epoch": 0.55, "learning_rate": 5.72585757110889e-05, "loss": 0.5953, "step": 5261 }, { "epoch": 0.55, "learning_rate": 5.7257547667177346e-05, "loss": 0.6262, "step": 5262 }, { "epoch": 0.55, "learning_rate": 5.7256519439774365e-05, "loss": 0.5193, "step": 5263 }, { "epoch": 0.55, "learning_rate": 5.725549102888687e-05, "loss": 1.5849, "step": 5264 }, { "epoch": 0.55, "learning_rate": 5.7254462434521784e-05, "loss": 0.9398, "step": 5265 }, { "epoch": 0.55, "learning_rate": 5.725343365668604e-05, "loss": 0.6178, "step": 5266 }, { "epoch": 0.55, "learning_rate": 5.725240469538655e-05, "loss": 0.7164, "step": 5267 }, { "epoch": 0.55, "learning_rate": 5.725137555063025e-05, "loss": 1.4126, "step": 5268 }, { "epoch": 0.55, "learning_rate": 5.725034622242408e-05, "loss": 1.8222, "step": 5269 }, { "epoch": 0.55, "learning_rate": 5.724931671077494e-05, "loss": 0.3312, "step": 5270 }, { "epoch": 0.55, "learning_rate": 5.724828701568978e-05, "loss": 0.5162, "step": 5271 }, { "epoch": 0.55, "learning_rate": 5.724725713717552e-05, "loss": 0.9264, "step": 5272 }, { "epoch": 0.55, "learning_rate": 5.724622707523912e-05, "loss": 0.9874, "step": 5273 }, { "epoch": 0.55, "learning_rate": 5.724519682988748e-05, "loss": 0.1196, "step": 5274 }, { "epoch": 0.55, "learning_rate": 5.724416640112754e-05, "loss": 0.4448, "step": 5275 }, { "epoch": 0.55, "learning_rate": 5.724313578896626e-05, "loss": 0.7071, "step": 5276 }, { "epoch": 0.55, "learning_rate": 5.7242104993410564e-05, "loss": 1.0427, "step": 5277 }, { "epoch": 0.55, "learning_rate": 5.724107401446739e-05, "loss": 1.0082, "step": 5278 }, { "epoch": 0.55, "learning_rate": 5.724004285214369e-05, "loss": 0.2755, "step": 5279 }, { "epoch": 0.55, "learning_rate": 5.7239011506446375e-05, "loss": 1.7207, "step": 5280 }, { "epoch": 0.55, "learning_rate": 5.723797997738241e-05, "loss": 1.1958, "step": 5281 }, { "epoch": 0.55, "learning_rate": 5.7236948264958755e-05, "loss": 0.9162, "step": 5282 }, { "epoch": 0.55, "learning_rate": 5.723591636918232e-05, "loss": 1.4048, "step": 5283 }, { "epoch": 0.55, "learning_rate": 5.7234884290060076e-05, "loss": 1.2315, "step": 5284 }, { "epoch": 0.55, "learning_rate": 5.723385202759896e-05, "loss": 0.2722, "step": 5285 }, { "epoch": 0.55, "learning_rate": 5.723281958180593e-05, "loss": 0.6795, "step": 5286 }, { "epoch": 0.55, "learning_rate": 5.7231786952687925e-05, "loss": 0.359, "step": 5287 }, { "epoch": 0.55, "learning_rate": 5.7230754140251914e-05, "loss": 1.5292, "step": 5288 }, { "epoch": 0.55, "learning_rate": 5.7229721144504826e-05, "loss": 0.4671, "step": 5289 }, { "epoch": 0.55, "learning_rate": 5.722868796545362e-05, "loss": 0.9617, "step": 5290 }, { "epoch": 0.55, "learning_rate": 5.7227654603105275e-05, "loss": 1.2186, "step": 5291 }, { "epoch": 0.55, "learning_rate": 5.722662105746672e-05, "loss": 0.9391, "step": 5292 }, { "epoch": 0.55, "learning_rate": 5.7225587328544925e-05, "loss": 1.5827, "step": 5293 }, { "epoch": 0.55, "learning_rate": 5.722455341634684e-05, "loss": 0.5646, "step": 5294 }, { "epoch": 0.55, "learning_rate": 5.722351932087944e-05, "loss": 0.8314, "step": 5295 }, { "epoch": 0.55, "learning_rate": 5.722248504214967e-05, "loss": 0.5997, "step": 5296 }, { "epoch": 0.55, "learning_rate": 5.722145058016451e-05, "loss": 0.6253, "step": 5297 }, { "epoch": 0.55, "learning_rate": 5.7220415934930904e-05, "loss": 0.5619, "step": 5298 }, { "epoch": 0.55, "learning_rate": 5.721938110645584e-05, "loss": 1.4562, "step": 5299 }, { "epoch": 0.55, "learning_rate": 5.721834609474626e-05, "loss": 0.4799, "step": 5300 }, { "epoch": 0.55, "learning_rate": 5.7217310899809143e-05, "loss": 0.5337, "step": 5301 }, { "epoch": 0.55, "learning_rate": 5.7216275521651456e-05, "loss": 0.8114, "step": 5302 }, { "epoch": 0.55, "learning_rate": 5.7215239960280176e-05, "loss": 0.4778, "step": 5303 }, { "epoch": 0.55, "learning_rate": 5.721420421570227e-05, "loss": 1.5562, "step": 5304 }, { "epoch": 0.55, "learning_rate": 5.7213168287924695e-05, "loss": 1.2742, "step": 5305 }, { "epoch": 0.55, "learning_rate": 5.721213217695445e-05, "loss": 1.3162, "step": 5306 }, { "epoch": 0.55, "learning_rate": 5.7211095882798505e-05, "loss": 0.7709, "step": 5307 }, { "epoch": 0.55, "learning_rate": 5.7210059405463814e-05, "loss": 1.179, "step": 5308 }, { "epoch": 0.55, "learning_rate": 5.720902274495738e-05, "loss": 0.6572, "step": 5309 }, { "epoch": 0.55, "learning_rate": 5.720798590128617e-05, "loss": 1.0328, "step": 5310 }, { "epoch": 0.55, "learning_rate": 5.720694887445717e-05, "loss": 1.0093, "step": 5311 }, { "epoch": 0.55, "learning_rate": 5.720591166447735e-05, "loss": 0.4303, "step": 5312 }, { "epoch": 0.55, "learning_rate": 5.720487427135369e-05, "loss": 0.7174, "step": 5313 }, { "epoch": 0.55, "learning_rate": 5.7203836695093195e-05, "loss": 1.1834, "step": 5314 }, { "epoch": 0.56, "learning_rate": 5.7202798935702825e-05, "loss": 1.3498, "step": 5315 }, { "epoch": 0.56, "learning_rate": 5.720176099318959e-05, "loss": 0.8165, "step": 5316 }, { "epoch": 0.56, "learning_rate": 5.7200722867560455e-05, "loss": 0.9626, "step": 5317 }, { "epoch": 0.56, "learning_rate": 5.719968455882243e-05, "loss": 1.5893, "step": 5318 }, { "epoch": 0.56, "learning_rate": 5.7198646066982475e-05, "loss": 0.1958, "step": 5319 }, { "epoch": 0.56, "learning_rate": 5.7197607392047616e-05, "loss": 1.0287, "step": 5320 }, { "epoch": 0.56, "learning_rate": 5.7196568534024816e-05, "loss": 1.529, "step": 5321 }, { "epoch": 0.56, "learning_rate": 5.719552949292108e-05, "loss": 0.3199, "step": 5322 }, { "epoch": 0.56, "learning_rate": 5.7194490268743404e-05, "loss": 1.3566, "step": 5323 }, { "epoch": 0.56, "learning_rate": 5.719345086149879e-05, "loss": 0.6917, "step": 5324 }, { "epoch": 0.56, "learning_rate": 5.719241127119422e-05, "loss": 1.8874, "step": 5325 }, { "epoch": 0.56, "learning_rate": 5.7191371497836706e-05, "loss": 0.582, "step": 5326 }, { "epoch": 0.56, "learning_rate": 5.7190331541433236e-05, "loss": 1.4607, "step": 5327 }, { "epoch": 0.56, "learning_rate": 5.718929140199082e-05, "loss": 0.6435, "step": 5328 }, { "epoch": 0.56, "learning_rate": 5.718825107951645e-05, "loss": 0.4376, "step": 5329 }, { "epoch": 0.56, "learning_rate": 5.718721057401714e-05, "loss": 0.7655, "step": 5330 }, { "epoch": 0.56, "learning_rate": 5.718616988549989e-05, "loss": 1.0744, "step": 5331 }, { "epoch": 0.56, "learning_rate": 5.718512901397171e-05, "loss": 1.5158, "step": 5332 }, { "epoch": 0.56, "learning_rate": 5.71840879594396e-05, "loss": 0.948, "step": 5333 }, { "epoch": 0.56, "learning_rate": 5.7183046721910565e-05, "loss": 2.0351, "step": 5334 }, { "epoch": 0.56, "learning_rate": 5.7182005301391625e-05, "loss": 1.0999, "step": 5335 }, { "epoch": 0.56, "learning_rate": 5.718096369788978e-05, "loss": 0.5622, "step": 5336 }, { "epoch": 0.56, "learning_rate": 5.717992191141205e-05, "loss": 0.4503, "step": 5337 }, { "epoch": 0.56, "learning_rate": 5.7178879941965444e-05, "loss": 0.8415, "step": 5338 }, { "epoch": 0.56, "learning_rate": 5.717783778955698e-05, "loss": 0.9535, "step": 5339 }, { "epoch": 0.56, "learning_rate": 5.7176795454193667e-05, "loss": 1.4454, "step": 5340 }, { "epoch": 0.56, "learning_rate": 5.717575293588253e-05, "loss": 1.143, "step": 5341 }, { "epoch": 0.56, "learning_rate": 5.717471023463058e-05, "loss": 0.7735, "step": 5342 }, { "epoch": 0.56, "learning_rate": 5.717366735044484e-05, "loss": 0.3984, "step": 5343 }, { "epoch": 0.56, "learning_rate": 5.7172624283332325e-05, "loss": 0.2623, "step": 5344 }, { "epoch": 0.56, "learning_rate": 5.717158103330007e-05, "loss": 0.4652, "step": 5345 }, { "epoch": 0.56, "learning_rate": 5.7170537600355085e-05, "loss": 1.4878, "step": 5346 }, { "epoch": 0.56, "learning_rate": 5.71694939845044e-05, "loss": 1.4453, "step": 5347 }, { "epoch": 0.56, "learning_rate": 5.716845018575504e-05, "loss": 0.77, "step": 5348 }, { "epoch": 0.56, "learning_rate": 5.716740620411402e-05, "loss": 0.5842, "step": 5349 }, { "epoch": 0.56, "learning_rate": 5.716636203958839e-05, "loss": 0.6178, "step": 5350 }, { "epoch": 0.56, "learning_rate": 5.716531769218516e-05, "loss": 0.4623, "step": 5351 }, { "epoch": 0.56, "learning_rate": 5.7164273161911375e-05, "loss": 1.8245, "step": 5352 }, { "epoch": 0.56, "learning_rate": 5.716322844877405e-05, "loss": 0.8763, "step": 5353 }, { "epoch": 0.56, "learning_rate": 5.7162183552780234e-05, "loss": 0.8824, "step": 5354 }, { "epoch": 0.56, "learning_rate": 5.716113847393695e-05, "loss": 0.9595, "step": 5355 }, { "epoch": 0.56, "learning_rate": 5.716009321225124e-05, "loss": 0.6817, "step": 5356 }, { "epoch": 0.56, "learning_rate": 5.715904776773013e-05, "loss": 1.0636, "step": 5357 }, { "epoch": 0.56, "learning_rate": 5.715800214038068e-05, "loss": 0.1573, "step": 5358 }, { "epoch": 0.56, "learning_rate": 5.71569563302099e-05, "loss": 0.9549, "step": 5359 }, { "epoch": 0.56, "learning_rate": 5.715591033722485e-05, "loss": 0.7521, "step": 5360 }, { "epoch": 0.56, "learning_rate": 5.715486416143256e-05, "loss": 1.4305, "step": 5361 }, { "epoch": 0.56, "learning_rate": 5.715381780284008e-05, "loss": 0.7687, "step": 5362 }, { "epoch": 0.56, "learning_rate": 5.715277126145446e-05, "loss": 0.4968, "step": 5363 }, { "epoch": 0.56, "learning_rate": 5.715172453728273e-05, "loss": 1.2629, "step": 5364 }, { "epoch": 0.56, "learning_rate": 5.7150677630331947e-05, "loss": 0.7444, "step": 5365 }, { "epoch": 0.56, "learning_rate": 5.714963054060916e-05, "loss": 1.3222, "step": 5366 }, { "epoch": 0.56, "learning_rate": 5.71485832681214e-05, "loss": 1.3843, "step": 5367 }, { "epoch": 0.56, "learning_rate": 5.714753581287573e-05, "loss": 1.4169, "step": 5368 }, { "epoch": 0.56, "learning_rate": 5.714648817487921e-05, "loss": 0.3972, "step": 5369 }, { "epoch": 0.56, "learning_rate": 5.714544035413888e-05, "loss": 0.367, "step": 5370 }, { "epoch": 0.56, "learning_rate": 5.7144392350661794e-05, "loss": 1.2693, "step": 5371 }, { "epoch": 0.56, "learning_rate": 5.714334416445502e-05, "loss": 0.9827, "step": 5372 }, { "epoch": 0.56, "learning_rate": 5.714229579552559e-05, "loss": 1.5128, "step": 5373 }, { "epoch": 0.56, "learning_rate": 5.714124724388059e-05, "loss": 0.7463, "step": 5374 }, { "epoch": 0.56, "learning_rate": 5.714019850952705e-05, "loss": 0.2786, "step": 5375 }, { "epoch": 0.56, "learning_rate": 5.713914959247205e-05, "loss": 0.3414, "step": 5376 }, { "epoch": 0.56, "learning_rate": 5.7138100492722644e-05, "loss": 0.9141, "step": 5377 }, { "epoch": 0.56, "learning_rate": 5.7137051210285895e-05, "loss": 1.1583, "step": 5378 }, { "epoch": 0.56, "learning_rate": 5.713600174516887e-05, "loss": 1.7596, "step": 5379 }, { "epoch": 0.56, "learning_rate": 5.713495209737863e-05, "loss": 1.1714, "step": 5380 }, { "epoch": 0.56, "learning_rate": 5.713390226692224e-05, "loss": 0.5469, "step": 5381 }, { "epoch": 0.56, "learning_rate": 5.7132852253806774e-05, "loss": 1.5031, "step": 5382 }, { "epoch": 0.56, "learning_rate": 5.713180205803929e-05, "loss": 0.5998, "step": 5383 }, { "epoch": 0.56, "learning_rate": 5.713075167962686e-05, "loss": 0.7423, "step": 5384 }, { "epoch": 0.56, "learning_rate": 5.712970111857656e-05, "loss": 0.8881, "step": 5385 }, { "epoch": 0.56, "learning_rate": 5.7128650374895465e-05, "loss": 0.9769, "step": 5386 }, { "epoch": 0.56, "learning_rate": 5.712759944859064e-05, "loss": 0.72, "step": 5387 }, { "epoch": 0.56, "learning_rate": 5.7126548339669164e-05, "loss": 1.1042, "step": 5388 }, { "epoch": 0.56, "learning_rate": 5.712549704813811e-05, "loss": 1.2591, "step": 5389 }, { "epoch": 0.56, "learning_rate": 5.712444557400455e-05, "loss": 0.3654, "step": 5390 }, { "epoch": 0.56, "learning_rate": 5.712339391727558e-05, "loss": 0.2502, "step": 5391 }, { "epoch": 0.56, "learning_rate": 5.7122342077958265e-05, "loss": 0.7003, "step": 5392 }, { "epoch": 0.56, "learning_rate": 5.712129005605968e-05, "loss": 1.2287, "step": 5393 }, { "epoch": 0.56, "learning_rate": 5.712023785158693e-05, "loss": 0.6263, "step": 5394 }, { "epoch": 0.56, "learning_rate": 5.711918546454708e-05, "loss": 0.8937, "step": 5395 }, { "epoch": 0.56, "learning_rate": 5.711813289494722e-05, "loss": 0.2584, "step": 5396 }, { "epoch": 0.56, "learning_rate": 5.7117080142794435e-05, "loss": 0.9739, "step": 5397 }, { "epoch": 0.56, "learning_rate": 5.71160272080958e-05, "loss": 1.2014, "step": 5398 }, { "epoch": 0.56, "learning_rate": 5.7114974090858426e-05, "loss": 0.9106, "step": 5399 }, { "epoch": 0.56, "learning_rate": 5.7113920791089386e-05, "loss": 0.2775, "step": 5400 }, { "epoch": 0.56, "learning_rate": 5.7112867308795784e-05, "loss": 0.8008, "step": 5401 }, { "epoch": 0.56, "learning_rate": 5.7111813643984696e-05, "loss": 1.5228, "step": 5402 }, { "epoch": 0.56, "learning_rate": 5.711075979666322e-05, "loss": 0.5189, "step": 5403 }, { "epoch": 0.56, "learning_rate": 5.7109705766838456e-05, "loss": 0.9529, "step": 5404 }, { "epoch": 0.56, "learning_rate": 5.7108651554517495e-05, "loss": 0.9023, "step": 5405 }, { "epoch": 0.56, "learning_rate": 5.710759715970744e-05, "loss": 1.4581, "step": 5406 }, { "epoch": 0.56, "learning_rate": 5.710654258241537e-05, "loss": 0.8916, "step": 5407 }, { "epoch": 0.56, "learning_rate": 5.710548782264841e-05, "loss": 0.7107, "step": 5408 }, { "epoch": 0.56, "learning_rate": 5.7104432880413644e-05, "loss": 0.4174, "step": 5409 }, { "epoch": 0.57, "learning_rate": 5.710337775571818e-05, "loss": 1.6162, "step": 5410 }, { "epoch": 0.57, "learning_rate": 5.710232244856911e-05, "loss": 1.1194, "step": 5411 }, { "epoch": 0.57, "learning_rate": 5.7101266958973564e-05, "loss": 0.9976, "step": 5412 }, { "epoch": 0.57, "learning_rate": 5.7100211286938614e-05, "loss": 0.9739, "step": 5413 }, { "epoch": 0.57, "learning_rate": 5.709915543247139e-05, "loss": 0.5259, "step": 5414 }, { "epoch": 0.57, "learning_rate": 5.709809939557899e-05, "loss": 1.2846, "step": 5415 }, { "epoch": 0.57, "learning_rate": 5.709704317626852e-05, "loss": 1.2317, "step": 5416 }, { "epoch": 0.57, "learning_rate": 5.709598677454711e-05, "loss": 1.3512, "step": 5417 }, { "epoch": 0.57, "learning_rate": 5.7094930190421846e-05, "loss": 0.961, "step": 5418 }, { "epoch": 0.57, "learning_rate": 5.709387342389985e-05, "loss": 0.648, "step": 5419 }, { "epoch": 0.57, "learning_rate": 5.709281647498825e-05, "loss": 0.3648, "step": 5420 }, { "epoch": 0.57, "learning_rate": 5.709175934369413e-05, "loss": 0.8668, "step": 5421 }, { "epoch": 0.57, "learning_rate": 5.709070203002463e-05, "loss": 0.5445, "step": 5422 }, { "epoch": 0.57, "learning_rate": 5.708964453398687e-05, "loss": 0.8639, "step": 5423 }, { "epoch": 0.57, "learning_rate": 5.708858685558797e-05, "loss": 0.5959, "step": 5424 }, { "epoch": 0.57, "learning_rate": 5.708752899483503e-05, "loss": 1.7775, "step": 5425 }, { "epoch": 0.57, "learning_rate": 5.708647095173518e-05, "loss": 0.713, "step": 5426 }, { "epoch": 0.57, "learning_rate": 5.708541272629555e-05, "loss": 0.7212, "step": 5427 }, { "epoch": 0.57, "learning_rate": 5.7084354318523256e-05, "loss": 0.4809, "step": 5428 }, { "epoch": 0.57, "learning_rate": 5.7083295728425427e-05, "loss": 0.3858, "step": 5429 }, { "epoch": 0.57, "learning_rate": 5.708223695600919e-05, "loss": 1.022, "step": 5430 }, { "epoch": 0.57, "learning_rate": 5.7081178001281674e-05, "loss": 1.6421, "step": 5431 }, { "epoch": 0.57, "learning_rate": 5.7080118864249995e-05, "loss": 0.6686, "step": 5432 }, { "epoch": 0.57, "learning_rate": 5.7079059544921294e-05, "loss": 1.2722, "step": 5433 }, { "epoch": 0.57, "learning_rate": 5.7078000043302714e-05, "loss": 1.5146, "step": 5434 }, { "epoch": 0.57, "learning_rate": 5.707694035940135e-05, "loss": 1.1458, "step": 5435 }, { "epoch": 0.57, "learning_rate": 5.7075880493224384e-05, "loss": 1.156, "step": 5436 }, { "epoch": 0.57, "learning_rate": 5.7074820444778904e-05, "loss": 1.5158, "step": 5437 }, { "epoch": 0.57, "learning_rate": 5.707376021407208e-05, "loss": 0.7348, "step": 5438 }, { "epoch": 0.57, "learning_rate": 5.707269980111104e-05, "loss": 0.4332, "step": 5439 }, { "epoch": 0.57, "learning_rate": 5.7071639205902906e-05, "loss": 1.4026, "step": 5440 }, { "epoch": 0.57, "learning_rate": 5.707057842845484e-05, "loss": 1.229, "step": 5441 }, { "epoch": 0.57, "learning_rate": 5.706951746877397e-05, "loss": 0.9576, "step": 5442 }, { "epoch": 0.57, "learning_rate": 5.706845632686744e-05, "loss": 0.366, "step": 5443 }, { "epoch": 0.57, "learning_rate": 5.706739500274241e-05, "loss": 1.1607, "step": 5444 }, { "epoch": 0.57, "learning_rate": 5.7066333496406e-05, "loss": 1.4956, "step": 5445 }, { "epoch": 0.57, "learning_rate": 5.706527180786536e-05, "loss": 0.8489, "step": 5446 }, { "epoch": 0.57, "learning_rate": 5.7064209937127654e-05, "loss": 0.6591, "step": 5447 }, { "epoch": 0.57, "learning_rate": 5.7063147884200015e-05, "loss": 0.7351, "step": 5448 }, { "epoch": 0.57, "learning_rate": 5.706208564908959e-05, "loss": 0.7678, "step": 5449 }, { "epoch": 0.57, "learning_rate": 5.706102323180355e-05, "loss": 0.5567, "step": 5450 }, { "epoch": 0.57, "learning_rate": 5.705996063234903e-05, "loss": 0.8857, "step": 5451 }, { "epoch": 0.57, "learning_rate": 5.705889785073318e-05, "loss": 0.5908, "step": 5452 }, { "epoch": 0.57, "learning_rate": 5.7057834886963165e-05, "loss": 0.9587, "step": 5453 }, { "epoch": 0.57, "learning_rate": 5.7056771741046127e-05, "loss": 0.5857, "step": 5454 }, { "epoch": 0.57, "learning_rate": 5.705570841298924e-05, "loss": 1.1606, "step": 5455 }, { "epoch": 0.57, "learning_rate": 5.705464490279966e-05, "loss": 1.4957, "step": 5456 }, { "epoch": 0.57, "learning_rate": 5.7053581210484537e-05, "loss": 0.4076, "step": 5457 }, { "epoch": 0.57, "learning_rate": 5.705251733605103e-05, "loss": 0.4558, "step": 5458 }, { "epoch": 0.57, "learning_rate": 5.705145327950631e-05, "loss": 0.9931, "step": 5459 }, { "epoch": 0.57, "learning_rate": 5.705038904085754e-05, "loss": 2.368, "step": 5460 }, { "epoch": 0.57, "learning_rate": 5.7049324620111874e-05, "loss": 0.9542, "step": 5461 }, { "epoch": 0.57, "learning_rate": 5.704826001727648e-05, "loss": 0.9491, "step": 5462 }, { "epoch": 0.57, "learning_rate": 5.704719523235854e-05, "loss": 1.4371, "step": 5463 }, { "epoch": 0.57, "learning_rate": 5.704613026536521e-05, "loss": 1.3423, "step": 5464 }, { "epoch": 0.57, "learning_rate": 5.704506511630365e-05, "loss": 1.0208, "step": 5465 }, { "epoch": 0.57, "learning_rate": 5.7043999785181045e-05, "loss": 0.6882, "step": 5466 }, { "epoch": 0.57, "learning_rate": 5.7042934272004555e-05, "loss": 0.3355, "step": 5467 }, { "epoch": 0.57, "learning_rate": 5.7041868576781366e-05, "loss": 0.3384, "step": 5468 }, { "epoch": 0.57, "learning_rate": 5.704080269951864e-05, "loss": 0.743, "step": 5469 }, { "epoch": 0.57, "learning_rate": 5.7039736640223565e-05, "loss": 1.3909, "step": 5470 }, { "epoch": 0.57, "learning_rate": 5.703867039890331e-05, "loss": 0.5634, "step": 5471 }, { "epoch": 0.57, "learning_rate": 5.703760397556504e-05, "loss": 1.5126, "step": 5472 }, { "epoch": 0.57, "learning_rate": 5.703653737021596e-05, "loss": 1.2346, "step": 5473 }, { "epoch": 0.57, "learning_rate": 5.703547058286322e-05, "loss": 1.1417, "step": 5474 }, { "epoch": 0.57, "learning_rate": 5.703440361351403e-05, "loss": 0.9691, "step": 5475 }, { "epoch": 0.57, "learning_rate": 5.7033336462175566e-05, "loss": 0.8891, "step": 5476 }, { "epoch": 0.57, "learning_rate": 5.7032269128855e-05, "loss": 0.6344, "step": 5477 }, { "epoch": 0.57, "learning_rate": 5.7031201613559516e-05, "loss": 0.7965, "step": 5478 }, { "epoch": 0.57, "learning_rate": 5.7030133916296314e-05, "loss": 1.3602, "step": 5479 }, { "epoch": 0.57, "learning_rate": 5.702906603707257e-05, "loss": 0.2424, "step": 5480 }, { "epoch": 0.57, "learning_rate": 5.702799797589549e-05, "loss": 0.8162, "step": 5481 }, { "epoch": 0.57, "learning_rate": 5.702692973277225e-05, "loss": 0.5431, "step": 5482 }, { "epoch": 0.57, "learning_rate": 5.702586130771003e-05, "loss": 0.9667, "step": 5483 }, { "epoch": 0.57, "learning_rate": 5.702479270071604e-05, "loss": 1.5382, "step": 5484 }, { "epoch": 0.57, "learning_rate": 5.702372391179747e-05, "loss": 0.9697, "step": 5485 }, { "epoch": 0.57, "learning_rate": 5.7022654940961515e-05, "loss": 1.1263, "step": 5486 }, { "epoch": 0.57, "learning_rate": 5.7021585788215374e-05, "loss": 0.3255, "step": 5487 }, { "epoch": 0.57, "learning_rate": 5.702051645356623e-05, "loss": 0.9044, "step": 5488 }, { "epoch": 0.57, "learning_rate": 5.7019446937021305e-05, "loss": 1.0983, "step": 5489 }, { "epoch": 0.57, "learning_rate": 5.701837723858777e-05, "loss": 1.3073, "step": 5490 }, { "epoch": 0.57, "learning_rate": 5.701730735827285e-05, "loss": 1.3492, "step": 5491 }, { "epoch": 0.57, "learning_rate": 5.701623729608374e-05, "loss": 0.8244, "step": 5492 }, { "epoch": 0.57, "learning_rate": 5.7015167052027636e-05, "loss": 1.1304, "step": 5493 }, { "epoch": 0.57, "learning_rate": 5.701409662611175e-05, "loss": 0.7381, "step": 5494 }, { "epoch": 0.57, "learning_rate": 5.701302601834329e-05, "loss": 1.6987, "step": 5495 }, { "epoch": 0.57, "learning_rate": 5.7011955228729456e-05, "loss": 0.5357, "step": 5496 }, { "epoch": 0.57, "learning_rate": 5.701088425727746e-05, "loss": 1.4596, "step": 5497 }, { "epoch": 0.57, "learning_rate": 5.700981310399451e-05, "loss": 1.4283, "step": 5498 }, { "epoch": 0.57, "learning_rate": 5.7008741768887816e-05, "loss": 0.543, "step": 5499 }, { "epoch": 0.57, "learning_rate": 5.70076702519646e-05, "loss": 1.4556, "step": 5500 }, { "epoch": 0.57, "learning_rate": 5.700659855323206e-05, "loss": 0.8678, "step": 5501 }, { "epoch": 0.57, "learning_rate": 5.700552667269742e-05, "loss": 1.3065, "step": 5502 }, { "epoch": 0.57, "learning_rate": 5.70044546103679e-05, "loss": 1.2555, "step": 5503 }, { "epoch": 0.57, "learning_rate": 5.70033823662507e-05, "loss": 0.9533, "step": 5504 }, { "epoch": 0.57, "learning_rate": 5.7002309940353046e-05, "loss": 0.6876, "step": 5505 }, { "epoch": 0.58, "learning_rate": 5.7001237332682165e-05, "loss": 1.3592, "step": 5506 }, { "epoch": 0.58, "learning_rate": 5.7000164543245284e-05, "loss": 1.3135, "step": 5507 }, { "epoch": 0.58, "learning_rate": 5.699909157204959e-05, "loss": 0.4013, "step": 5508 }, { "epoch": 0.58, "learning_rate": 5.699801841910235e-05, "loss": 1.0917, "step": 5509 }, { "epoch": 0.58, "learning_rate": 5.6996945084410756e-05, "loss": 1.161, "step": 5510 }, { "epoch": 0.58, "learning_rate": 5.6995871567982044e-05, "loss": 1.2838, "step": 5511 }, { "epoch": 0.58, "learning_rate": 5.6994797869823436e-05, "loss": 0.7042, "step": 5512 }, { "epoch": 0.58, "learning_rate": 5.699372398994218e-05, "loss": 0.4176, "step": 5513 }, { "epoch": 0.58, "learning_rate": 5.6992649928345475e-05, "loss": 0.6117, "step": 5514 }, { "epoch": 0.58, "learning_rate": 5.6991575685040576e-05, "loss": 0.7554, "step": 5515 }, { "epoch": 0.58, "learning_rate": 5.699050126003471e-05, "loss": 0.4664, "step": 5516 }, { "epoch": 0.58, "learning_rate": 5.698942665333509e-05, "loss": 1.3067, "step": 5517 }, { "epoch": 0.58, "learning_rate": 5.698835186494897e-05, "loss": 0.8784, "step": 5518 }, { "epoch": 0.58, "learning_rate": 5.698727689488358e-05, "loss": 0.8988, "step": 5519 }, { "epoch": 0.58, "learning_rate": 5.698620174314616e-05, "loss": 0.7173, "step": 5520 }, { "epoch": 0.58, "learning_rate": 5.6985126409743944e-05, "loss": 0.2941, "step": 5521 }, { "epoch": 0.58, "learning_rate": 5.6984050894684173e-05, "loss": 2.2374, "step": 5522 }, { "epoch": 0.58, "learning_rate": 5.698297519797408e-05, "loss": 0.3776, "step": 5523 }, { "epoch": 0.58, "learning_rate": 5.698189931962091e-05, "loss": 1.2403, "step": 5524 }, { "epoch": 0.58, "learning_rate": 5.698082325963191e-05, "loss": 1.4667, "step": 5525 }, { "epoch": 0.58, "learning_rate": 5.697974701801432e-05, "loss": 1.1091, "step": 5526 }, { "epoch": 0.58, "learning_rate": 5.697867059477539e-05, "loss": 1.4699, "step": 5527 }, { "epoch": 0.58, "learning_rate": 5.6977593989922366e-05, "loss": 0.9727, "step": 5528 }, { "epoch": 0.58, "learning_rate": 5.697651720346247e-05, "loss": 0.5317, "step": 5529 }, { "epoch": 0.58, "learning_rate": 5.697544023540299e-05, "loss": 1.172, "step": 5530 }, { "epoch": 0.58, "learning_rate": 5.697436308575116e-05, "loss": 1.0366, "step": 5531 }, { "epoch": 0.58, "learning_rate": 5.697328575451422e-05, "loss": 1.4094, "step": 5532 }, { "epoch": 0.58, "learning_rate": 5.697220824169944e-05, "loss": 0.453, "step": 5533 }, { "epoch": 0.58, "learning_rate": 5.697113054731406e-05, "loss": 0.619, "step": 5534 }, { "epoch": 0.58, "learning_rate": 5.6970052671365335e-05, "loss": 0.5204, "step": 5535 }, { "epoch": 0.58, "learning_rate": 5.696897461386054e-05, "loss": 0.9519, "step": 5536 }, { "epoch": 0.58, "learning_rate": 5.696789637480691e-05, "loss": 0.3845, "step": 5537 }, { "epoch": 0.58, "learning_rate": 5.6966817954211704e-05, "loss": 1.0258, "step": 5538 }, { "epoch": 0.58, "learning_rate": 5.69657393520822e-05, "loss": 0.6104, "step": 5539 }, { "epoch": 0.58, "learning_rate": 5.696466056842564e-05, "loss": 0.7483, "step": 5540 }, { "epoch": 0.58, "learning_rate": 5.69635816032493e-05, "loss": 0.5874, "step": 5541 }, { "epoch": 0.58, "learning_rate": 5.696250245656044e-05, "loss": 1.6218, "step": 5542 }, { "epoch": 0.58, "learning_rate": 5.6961423128366314e-05, "loss": 1.0397, "step": 5543 }, { "epoch": 0.58, "learning_rate": 5.6960343618674195e-05, "loss": 1.6566, "step": 5544 }, { "epoch": 0.58, "learning_rate": 5.695926392749137e-05, "loss": 1.3114, "step": 5545 }, { "epoch": 0.58, "learning_rate": 5.695818405482507e-05, "loss": 1.6704, "step": 5546 }, { "epoch": 0.58, "learning_rate": 5.695710400068258e-05, "loss": 1.7185, "step": 5547 }, { "epoch": 0.58, "learning_rate": 5.695602376507119e-05, "loss": 0.8416, "step": 5548 }, { "epoch": 0.58, "learning_rate": 5.695494334799815e-05, "loss": 1.3132, "step": 5549 }, { "epoch": 0.58, "learning_rate": 5.695386274947073e-05, "loss": 0.9472, "step": 5550 }, { "epoch": 0.58, "learning_rate": 5.6952781969496214e-05, "loss": 1.3452, "step": 5551 }, { "epoch": 0.58, "learning_rate": 5.695170100808188e-05, "loss": 1.3127, "step": 5552 }, { "epoch": 0.58, "learning_rate": 5.6950619865234996e-05, "loss": 0.5665, "step": 5553 }, { "epoch": 0.58, "learning_rate": 5.694953854096285e-05, "loss": 0.4046, "step": 5554 }, { "epoch": 0.58, "learning_rate": 5.6948457035272716e-05, "loss": 0.3505, "step": 5555 }, { "epoch": 0.58, "learning_rate": 5.694737534817188e-05, "loss": 1.4733, "step": 5556 }, { "epoch": 0.58, "learning_rate": 5.694629347966761e-05, "loss": 0.6477, "step": 5557 }, { "epoch": 0.58, "learning_rate": 5.69452114297672e-05, "loss": 1.1129, "step": 5558 }, { "epoch": 0.58, "learning_rate": 5.6944129198477934e-05, "loss": 0.8128, "step": 5559 }, { "epoch": 0.58, "learning_rate": 5.694304678580709e-05, "loss": 0.6488, "step": 5560 }, { "epoch": 0.58, "learning_rate": 5.694196419176196e-05, "loss": 0.8141, "step": 5561 }, { "epoch": 0.58, "learning_rate": 5.694088141634983e-05, "loss": 1.7062, "step": 5562 }, { "epoch": 0.58, "learning_rate": 5.693979845957799e-05, "loss": 1.1478, "step": 5563 }, { "epoch": 0.58, "learning_rate": 5.693871532145373e-05, "loss": 0.6459, "step": 5564 }, { "epoch": 0.58, "learning_rate": 5.6937632001984343e-05, "loss": 0.6215, "step": 5565 }, { "epoch": 0.58, "learning_rate": 5.693654850117712e-05, "loss": 1.2197, "step": 5566 }, { "epoch": 0.58, "learning_rate": 5.693546481903937e-05, "loss": 0.4981, "step": 5567 }, { "epoch": 0.58, "learning_rate": 5.693438095557835e-05, "loss": 0.8281, "step": 5568 }, { "epoch": 0.58, "learning_rate": 5.6933296910801384e-05, "loss": 0.566, "step": 5569 }, { "epoch": 0.58, "learning_rate": 5.6932212684715776e-05, "loss": 0.8539, "step": 5570 }, { "epoch": 0.58, "learning_rate": 5.693112827732881e-05, "loss": 1.0627, "step": 5571 }, { "epoch": 0.58, "learning_rate": 5.6930043688647784e-05, "loss": 1.1327, "step": 5572 }, { "epoch": 0.58, "learning_rate": 5.692895891868001e-05, "loss": 1.0683, "step": 5573 }, { "epoch": 0.58, "learning_rate": 5.6927873967432786e-05, "loss": 0.7132, "step": 5574 }, { "epoch": 0.58, "learning_rate": 5.692678883491341e-05, "loss": 1.4129, "step": 5575 }, { "epoch": 0.58, "learning_rate": 5.692570352112919e-05, "loss": 0.7016, "step": 5576 }, { "epoch": 0.58, "learning_rate": 5.6924618026087446e-05, "loss": 0.3315, "step": 5577 }, { "epoch": 0.58, "learning_rate": 5.692353234979546e-05, "loss": 1.6872, "step": 5578 }, { "epoch": 0.58, "learning_rate": 5.6922446492260565e-05, "loss": 1.0083, "step": 5579 }, { "epoch": 0.58, "learning_rate": 5.692136045349006e-05, "loss": 1.6375, "step": 5580 }, { "epoch": 0.58, "learning_rate": 5.6920274233491246e-05, "loss": 0.7508, "step": 5581 }, { "epoch": 0.58, "learning_rate": 5.6919187832271446e-05, "loss": 0.4996, "step": 5582 }, { "epoch": 0.58, "learning_rate": 5.6918101249837974e-05, "loss": 1.3982, "step": 5583 }, { "epoch": 0.58, "learning_rate": 5.691701448619815e-05, "loss": 0.5155, "step": 5584 }, { "epoch": 0.58, "learning_rate": 5.691592754135927e-05, "loss": 0.248, "step": 5585 }, { "epoch": 0.58, "learning_rate": 5.691484041532867e-05, "loss": 1.2656, "step": 5586 }, { "epoch": 0.58, "learning_rate": 5.691375310811366e-05, "loss": 0.1655, "step": 5587 }, { "epoch": 0.58, "learning_rate": 5.6912665619721565e-05, "loss": 0.498, "step": 5588 }, { "epoch": 0.58, "learning_rate": 5.6911577950159706e-05, "loss": 0.005, "step": 5589 }, { "epoch": 0.58, "learning_rate": 5.691049009943539e-05, "loss": 1.3817, "step": 5590 }, { "epoch": 0.58, "learning_rate": 5.690940206755596e-05, "loss": 0.763, "step": 5591 }, { "epoch": 0.58, "learning_rate": 5.690831385452872e-05, "loss": 0.818, "step": 5592 }, { "epoch": 0.58, "learning_rate": 5.690722546036102e-05, "loss": 0.1097, "step": 5593 }, { "epoch": 0.58, "learning_rate": 5.690613688506017e-05, "loss": 1.5773, "step": 5594 }, { "epoch": 0.58, "learning_rate": 5.69050481286335e-05, "loss": 1.1177, "step": 5595 }, { "epoch": 0.58, "learning_rate": 5.6903959191088346e-05, "loss": 0.4314, "step": 5596 }, { "epoch": 0.58, "learning_rate": 5.690287007243203e-05, "loss": 1.1959, "step": 5597 }, { "epoch": 0.58, "learning_rate": 5.690178077267189e-05, "loss": 0.1285, "step": 5598 }, { "epoch": 0.58, "learning_rate": 5.6900691291815255e-05, "loss": 1.2156, "step": 5599 }, { "epoch": 0.58, "learning_rate": 5.689960162986946e-05, "loss": 1.3348, "step": 5600 }, { "epoch": 0.58, "learning_rate": 5.689851178684184e-05, "loss": 0.9969, "step": 5601 }, { "epoch": 0.59, "learning_rate": 5.689742176273973e-05, "loss": 1.3448, "step": 5602 }, { "epoch": 0.59, "learning_rate": 5.6896331557570474e-05, "loss": 1.1198, "step": 5603 }, { "epoch": 0.59, "learning_rate": 5.6895241171341405e-05, "loss": 0.9911, "step": 5604 }, { "epoch": 0.59, "learning_rate": 5.6894150604059864e-05, "loss": 1.4209, "step": 5605 }, { "epoch": 0.59, "learning_rate": 5.68930598557332e-05, "loss": 0.6084, "step": 5606 }, { "epoch": 0.59, "learning_rate": 5.6891968926368746e-05, "loss": 0.4879, "step": 5607 }, { "epoch": 0.59, "learning_rate": 5.6890877815973846e-05, "loss": 0.7499, "step": 5608 }, { "epoch": 0.59, "learning_rate": 5.688978652455585e-05, "loss": 0.5502, "step": 5609 }, { "epoch": 0.59, "learning_rate": 5.688869505212211e-05, "loss": 0.1655, "step": 5610 }, { "epoch": 0.59, "learning_rate": 5.688760339867996e-05, "loss": 0.167, "step": 5611 }, { "epoch": 0.59, "learning_rate": 5.6886511564236756e-05, "loss": 0.8543, "step": 5612 }, { "epoch": 0.59, "learning_rate": 5.6885419548799844e-05, "loss": 0.3893, "step": 5613 }, { "epoch": 0.59, "learning_rate": 5.688432735237658e-05, "loss": 0.8786, "step": 5614 }, { "epoch": 0.59, "learning_rate": 5.6883234974974324e-05, "loss": 1.4922, "step": 5615 }, { "epoch": 0.59, "learning_rate": 5.6882142416600406e-05, "loss": 1.3676, "step": 5616 }, { "epoch": 0.59, "learning_rate": 5.68810496772622e-05, "loss": 0.6998, "step": 5617 }, { "epoch": 0.59, "learning_rate": 5.6879956756967064e-05, "loss": 1.128, "step": 5618 }, { "epoch": 0.59, "learning_rate": 5.6878863655722344e-05, "loss": 0.3698, "step": 5619 }, { "epoch": 0.59, "learning_rate": 5.68777703735354e-05, "loss": 1.0449, "step": 5620 }, { "epoch": 0.59, "learning_rate": 5.68766769104136e-05, "loss": 1.2853, "step": 5621 }, { "epoch": 0.59, "learning_rate": 5.687558326636429e-05, "loss": 0.6092, "step": 5622 }, { "epoch": 0.59, "learning_rate": 5.6874489441394855e-05, "loss": 1.221, "step": 5623 }, { "epoch": 0.59, "learning_rate": 5.687339543551264e-05, "loss": 0.288, "step": 5624 }, { "epoch": 0.59, "learning_rate": 5.6872301248725015e-05, "loss": 1.69, "step": 5625 }, { "epoch": 0.59, "learning_rate": 5.687120688103934e-05, "loss": 1.2552, "step": 5626 }, { "epoch": 0.59, "learning_rate": 5.6870112332462996e-05, "loss": 0.7449, "step": 5627 }, { "epoch": 0.59, "learning_rate": 5.686901760300334e-05, "loss": 1.1114, "step": 5628 }, { "epoch": 0.59, "learning_rate": 5.6867922692667744e-05, "loss": 1.4438, "step": 5629 }, { "epoch": 0.59, "learning_rate": 5.686682760146357e-05, "loss": 0.4127, "step": 5630 }, { "epoch": 0.59, "learning_rate": 5.686573232939821e-05, "loss": 0.9125, "step": 5631 }, { "epoch": 0.59, "learning_rate": 5.686463687647902e-05, "loss": 0.3823, "step": 5632 }, { "epoch": 0.59, "learning_rate": 5.686354124271339e-05, "loss": 1.6509, "step": 5633 }, { "epoch": 0.59, "learning_rate": 5.6862445428108676e-05, "loss": 1.175, "step": 5634 }, { "epoch": 0.59, "learning_rate": 5.6861349432672274e-05, "loss": 1.1661, "step": 5635 }, { "epoch": 0.59, "learning_rate": 5.6860253256411544e-05, "loss": 0.4221, "step": 5636 }, { "epoch": 0.59, "learning_rate": 5.6859156899333876e-05, "loss": 0.6849, "step": 5637 }, { "epoch": 0.59, "learning_rate": 5.6858060361446646e-05, "loss": 0.2338, "step": 5638 }, { "epoch": 0.59, "learning_rate": 5.6856963642757244e-05, "loss": 0.4989, "step": 5639 }, { "epoch": 0.59, "learning_rate": 5.6855866743273044e-05, "loss": 0.4929, "step": 5640 }, { "epoch": 0.59, "learning_rate": 5.685476966300143e-05, "loss": 1.2876, "step": 5641 }, { "epoch": 0.59, "learning_rate": 5.685367240194979e-05, "loss": 0.8616, "step": 5642 }, { "epoch": 0.59, "learning_rate": 5.685257496012553e-05, "loss": 0.6164, "step": 5643 }, { "epoch": 0.59, "learning_rate": 5.6851477337535995e-05, "loss": 0.7026, "step": 5644 }, { "epoch": 0.59, "learning_rate": 5.68503795341886e-05, "loss": 1.3732, "step": 5645 }, { "epoch": 0.59, "learning_rate": 5.6849281550090734e-05, "loss": 0.923, "step": 5646 }, { "epoch": 0.59, "learning_rate": 5.68481833852498e-05, "loss": 0.8315, "step": 5647 }, { "epoch": 0.59, "learning_rate": 5.684708503967316e-05, "loss": 0.373, "step": 5648 }, { "epoch": 0.59, "learning_rate": 5.684598651336823e-05, "loss": 1.4152, "step": 5649 }, { "epoch": 0.59, "learning_rate": 5.6844887806342406e-05, "loss": 1.0505, "step": 5650 }, { "epoch": 0.59, "learning_rate": 5.684378891860308e-05, "loss": 1.3968, "step": 5651 }, { "epoch": 0.59, "learning_rate": 5.684268985015763e-05, "loss": 1.2367, "step": 5652 }, { "epoch": 0.59, "learning_rate": 5.684159060101349e-05, "loss": 1.497, "step": 5653 }, { "epoch": 0.59, "learning_rate": 5.684049117117804e-05, "loss": 0.1329, "step": 5654 }, { "epoch": 0.59, "learning_rate": 5.683939156065868e-05, "loss": 1.441, "step": 5655 }, { "epoch": 0.59, "learning_rate": 5.683829176946282e-05, "loss": 1.435, "step": 5656 }, { "epoch": 0.59, "learning_rate": 5.6837191797597855e-05, "loss": 1.353, "step": 5657 }, { "epoch": 0.59, "learning_rate": 5.6836091645071194e-05, "loss": 1.1061, "step": 5658 }, { "epoch": 0.59, "learning_rate": 5.6834991311890246e-05, "loss": 1.2467, "step": 5659 }, { "epoch": 0.59, "learning_rate": 5.683389079806242e-05, "loss": 1.3253, "step": 5660 }, { "epoch": 0.59, "learning_rate": 5.683279010359511e-05, "loss": 1.0921, "step": 5661 }, { "epoch": 0.59, "learning_rate": 5.683168922849574e-05, "loss": 0.3137, "step": 5662 }, { "epoch": 0.59, "learning_rate": 5.683058817277172e-05, "loss": 1.1582, "step": 5663 }, { "epoch": 0.59, "learning_rate": 5.6829486936430455e-05, "loss": 0.4285, "step": 5664 }, { "epoch": 0.59, "learning_rate": 5.6828385519479356e-05, "loss": 0.5358, "step": 5665 }, { "epoch": 0.59, "learning_rate": 5.682728392192585e-05, "loss": 1.4715, "step": 5666 }, { "epoch": 0.59, "learning_rate": 5.682618214377734e-05, "loss": 1.6735, "step": 5667 }, { "epoch": 0.59, "learning_rate": 5.682508018504126e-05, "loss": 1.5955, "step": 5668 }, { "epoch": 0.59, "learning_rate": 5.682397804572501e-05, "loss": 1.1853, "step": 5669 }, { "epoch": 0.59, "learning_rate": 5.682287572583601e-05, "loss": 0.9008, "step": 5670 }, { "epoch": 0.59, "learning_rate": 5.6821773225381696e-05, "loss": 0.4486, "step": 5671 }, { "epoch": 0.59, "learning_rate": 5.6820670544369475e-05, "loss": 0.6879, "step": 5672 }, { "epoch": 0.59, "learning_rate": 5.6819567682806774e-05, "loss": 0.6787, "step": 5673 }, { "epoch": 0.59, "learning_rate": 5.681846464070102e-05, "loss": 1.5937, "step": 5674 }, { "epoch": 0.59, "learning_rate": 5.681736141805964e-05, "loss": 0.5336, "step": 5675 }, { "epoch": 0.59, "learning_rate": 5.681625801489006e-05, "loss": 1.1847, "step": 5676 }, { "epoch": 0.59, "learning_rate": 5.681515443119969e-05, "loss": 1.1316, "step": 5677 }, { "epoch": 0.59, "learning_rate": 5.681405066699599e-05, "loss": 0.6516, "step": 5678 }, { "epoch": 0.59, "learning_rate": 5.681294672228637e-05, "loss": 0.6668, "step": 5679 }, { "epoch": 0.59, "learning_rate": 5.6811842597078267e-05, "loss": 0.1924, "step": 5680 }, { "epoch": 0.59, "learning_rate": 5.6810738291379116e-05, "loss": 0.1555, "step": 5681 }, { "epoch": 0.59, "learning_rate": 5.680963380519635e-05, "loss": 1.0914, "step": 5682 }, { "epoch": 0.59, "learning_rate": 5.680852913853739e-05, "loss": 0.9142, "step": 5683 }, { "epoch": 0.59, "learning_rate": 5.68074242914097e-05, "loss": 1.275, "step": 5684 }, { "epoch": 0.59, "learning_rate": 5.6806319263820694e-05, "loss": 0.721, "step": 5685 }, { "epoch": 0.59, "learning_rate": 5.680521405577782e-05, "loss": 1.2662, "step": 5686 }, { "epoch": 0.59, "learning_rate": 5.680410866728851e-05, "loss": 0.9607, "step": 5687 }, { "epoch": 0.59, "learning_rate": 5.6803003098360223e-05, "loss": 1.0277, "step": 5688 }, { "epoch": 0.59, "learning_rate": 5.6801897349000384e-05, "loss": 0.8125, "step": 5689 }, { "epoch": 0.59, "learning_rate": 5.680079141921645e-05, "loss": 0.331, "step": 5690 }, { "epoch": 0.59, "learning_rate": 5.679968530901586e-05, "loss": 1.3133, "step": 5691 }, { "epoch": 0.59, "learning_rate": 5.6798579018406044e-05, "loss": 0.9949, "step": 5692 }, { "epoch": 0.59, "learning_rate": 5.6797472547394476e-05, "loss": 1.0509, "step": 5693 }, { "epoch": 0.59, "learning_rate": 5.679636589598859e-05, "loss": 0.2652, "step": 5694 }, { "epoch": 0.59, "learning_rate": 5.679525906419584e-05, "loss": 1.7258, "step": 5695 }, { "epoch": 0.59, "learning_rate": 5.6794152052023674e-05, "loss": 0.9269, "step": 5696 }, { "epoch": 0.59, "learning_rate": 5.6793044859479544e-05, "loss": 1.3068, "step": 5697 }, { "epoch": 0.6, "learning_rate": 5.679193748657091e-05, "loss": 0.4287, "step": 5698 }, { "epoch": 0.6, "learning_rate": 5.679082993330522e-05, "loss": 0.3097, "step": 5699 }, { "epoch": 0.6, "learning_rate": 5.678972219968993e-05, "loss": 1.0019, "step": 5700 }, { "epoch": 0.6, "learning_rate": 5.67886142857325e-05, "loss": 0.6897, "step": 5701 }, { "epoch": 0.6, "learning_rate": 5.678750619144038e-05, "loss": 0.3898, "step": 5702 }, { "epoch": 0.6, "learning_rate": 5.678639791682105e-05, "loss": 0.3823, "step": 5703 }, { "epoch": 0.6, "learning_rate": 5.6785289461881947e-05, "loss": 0.8216, "step": 5704 }, { "epoch": 0.6, "learning_rate": 5.678418082663054e-05, "loss": 0.4323, "step": 5705 }, { "epoch": 0.6, "learning_rate": 5.6783072011074304e-05, "loss": 0.2532, "step": 5706 }, { "epoch": 0.6, "learning_rate": 5.678196301522069e-05, "loss": 1.1921, "step": 5707 }, { "epoch": 0.6, "learning_rate": 5.6780853839077155e-05, "loss": 0.5034, "step": 5708 }, { "epoch": 0.6, "learning_rate": 5.67797444826512e-05, "loss": 0.5451, "step": 5709 }, { "epoch": 0.6, "learning_rate": 5.677863494595025e-05, "loss": 0.9897, "step": 5710 }, { "epoch": 0.6, "learning_rate": 5.6777525228981816e-05, "loss": 0.7921, "step": 5711 }, { "epoch": 0.6, "learning_rate": 5.677641533175333e-05, "loss": 0.9523, "step": 5712 }, { "epoch": 0.6, "learning_rate": 5.677530525427229e-05, "loss": 0.5782, "step": 5713 }, { "epoch": 0.6, "learning_rate": 5.677419499654616e-05, "loss": 0.9245, "step": 5714 }, { "epoch": 0.6, "learning_rate": 5.677308455858242e-05, "loss": 0.8985, "step": 5715 }, { "epoch": 0.6, "learning_rate": 5.677197394038853e-05, "loss": 0.6906, "step": 5716 }, { "epoch": 0.6, "learning_rate": 5.677086314197197e-05, "loss": 0.5248, "step": 5717 }, { "epoch": 0.6, "learning_rate": 5.676975216334023e-05, "loss": 0.3917, "step": 5718 }, { "epoch": 0.6, "learning_rate": 5.6768641004500783e-05, "loss": 0.2807, "step": 5719 }, { "epoch": 0.6, "learning_rate": 5.676752966546111e-05, "loss": 0.5052, "step": 5720 }, { "epoch": 0.6, "learning_rate": 5.676641814622869e-05, "loss": 0.9211, "step": 5721 }, { "epoch": 0.6, "learning_rate": 5.6765306446811004e-05, "loss": 1.4528, "step": 5722 }, { "epoch": 0.6, "learning_rate": 5.6764194567215537e-05, "loss": 1.0317, "step": 5723 }, { "epoch": 0.6, "learning_rate": 5.676308250744977e-05, "loss": 1.3298, "step": 5724 }, { "epoch": 0.6, "learning_rate": 5.676197026752121e-05, "loss": 0.4072, "step": 5725 }, { "epoch": 0.6, "learning_rate": 5.676085784743731e-05, "loss": 0.5718, "step": 5726 }, { "epoch": 0.6, "learning_rate": 5.675974524720558e-05, "loss": 0.25, "step": 5727 }, { "epoch": 0.6, "learning_rate": 5.675863246683352e-05, "loss": 1.6864, "step": 5728 }, { "epoch": 0.6, "learning_rate": 5.6757519506328586e-05, "loss": 1.5651, "step": 5729 }, { "epoch": 0.6, "learning_rate": 5.6756406365698303e-05, "loss": 0.3606, "step": 5730 }, { "epoch": 0.6, "learning_rate": 5.675529304495015e-05, "loss": 0.9831, "step": 5731 }, { "epoch": 0.6, "learning_rate": 5.6754179544091636e-05, "loss": 1.0858, "step": 5732 }, { "epoch": 0.6, "learning_rate": 5.6753065863130234e-05, "loss": 0.87, "step": 5733 }, { "epoch": 0.6, "learning_rate": 5.675195200207345e-05, "loss": 0.5771, "step": 5734 }, { "epoch": 0.6, "learning_rate": 5.675083796092879e-05, "loss": 0.8544, "step": 5735 }, { "epoch": 0.6, "learning_rate": 5.674972373970375e-05, "loss": 1.1058, "step": 5736 }, { "epoch": 0.6, "learning_rate": 5.674860933840582e-05, "loss": 0.9284, "step": 5737 }, { "epoch": 0.6, "learning_rate": 5.6747494757042516e-05, "loss": 0.4022, "step": 5738 }, { "epoch": 0.6, "learning_rate": 5.6746379995621336e-05, "loss": 0.5079, "step": 5739 }, { "epoch": 0.6, "learning_rate": 5.6745265054149784e-05, "loss": 0.7664, "step": 5740 }, { "epoch": 0.6, "learning_rate": 5.674414993263537e-05, "loss": 1.3628, "step": 5741 }, { "epoch": 0.6, "learning_rate": 5.674303463108559e-05, "loss": 0.6172, "step": 5742 }, { "epoch": 0.6, "learning_rate": 5.674191914950796e-05, "loss": 0.3842, "step": 5743 }, { "epoch": 0.6, "learning_rate": 5.6740803487909984e-05, "loss": 1.6029, "step": 5744 }, { "epoch": 0.6, "learning_rate": 5.6739687646299176e-05, "loss": 1.1609, "step": 5745 }, { "epoch": 0.6, "learning_rate": 5.673857162468306e-05, "loss": 0.4907, "step": 5746 }, { "epoch": 0.6, "learning_rate": 5.673745542306912e-05, "loss": 0.5984, "step": 5747 }, { "epoch": 0.6, "learning_rate": 5.673633904146489e-05, "loss": 0.4153, "step": 5748 }, { "epoch": 0.6, "learning_rate": 5.673522247987789e-05, "loss": 0.4217, "step": 5749 }, { "epoch": 0.6, "learning_rate": 5.673410573831562e-05, "loss": 1.256, "step": 5750 }, { "epoch": 0.6, "learning_rate": 5.6732988816785605e-05, "loss": 0.4665, "step": 5751 }, { "epoch": 0.6, "learning_rate": 5.673187171529537e-05, "loss": 0.9785, "step": 5752 }, { "epoch": 0.6, "learning_rate": 5.673075443385243e-05, "loss": 1.2425, "step": 5753 }, { "epoch": 0.6, "learning_rate": 5.6729636972464295e-05, "loss": 1.1504, "step": 5754 }, { "epoch": 0.6, "learning_rate": 5.672851933113851e-05, "loss": 0.8954, "step": 5755 }, { "epoch": 0.6, "learning_rate": 5.672740150988258e-05, "loss": 0.7484, "step": 5756 }, { "epoch": 0.6, "learning_rate": 5.6726283508704033e-05, "loss": 1.0256, "step": 5757 }, { "epoch": 0.6, "learning_rate": 5.6725165327610404e-05, "loss": 0.5946, "step": 5758 }, { "epoch": 0.6, "learning_rate": 5.672404696660922e-05, "loss": 0.4834, "step": 5759 }, { "epoch": 0.6, "learning_rate": 5.6722928425707994e-05, "loss": 1.311, "step": 5760 }, { "epoch": 0.6, "learning_rate": 5.672180970491427e-05, "loss": 1.4512, "step": 5761 }, { "epoch": 0.6, "learning_rate": 5.6720690804235584e-05, "loss": 0.2297, "step": 5762 }, { "epoch": 0.6, "learning_rate": 5.671957172367946e-05, "loss": 0.4635, "step": 5763 }, { "epoch": 0.6, "learning_rate": 5.6718452463253416e-05, "loss": 1.4434, "step": 5764 }, { "epoch": 0.6, "learning_rate": 5.671733302296501e-05, "loss": 0.4998, "step": 5765 }, { "epoch": 0.6, "learning_rate": 5.671621340282178e-05, "loss": 0.3188, "step": 5766 }, { "epoch": 0.6, "learning_rate": 5.671509360283124e-05, "loss": 0.2675, "step": 5767 }, { "epoch": 0.6, "learning_rate": 5.671397362300094e-05, "loss": 0.4571, "step": 5768 }, { "epoch": 0.6, "learning_rate": 5.671285346333842e-05, "loss": 1.2157, "step": 5769 }, { "epoch": 0.6, "learning_rate": 5.6711733123851224e-05, "loss": 0.6754, "step": 5770 }, { "epoch": 0.6, "learning_rate": 5.6710612604546894e-05, "loss": 1.0873, "step": 5771 }, { "epoch": 0.6, "learning_rate": 5.670949190543297e-05, "loss": 1.0476, "step": 5772 }, { "epoch": 0.6, "learning_rate": 5.6708371026516984e-05, "loss": 0.2959, "step": 5773 }, { "epoch": 0.6, "learning_rate": 5.67072499678065e-05, "loss": 0.9825, "step": 5774 }, { "epoch": 0.6, "learning_rate": 5.670612872930906e-05, "loss": 1.1558, "step": 5775 }, { "epoch": 0.6, "learning_rate": 5.670500731103222e-05, "loss": 0.4486, "step": 5776 }, { "epoch": 0.6, "learning_rate": 5.67038857129835e-05, "loss": 0.6971, "step": 5777 }, { "epoch": 0.6, "learning_rate": 5.670276393517049e-05, "loss": 0.9155, "step": 5778 }, { "epoch": 0.6, "learning_rate": 5.67016419776007e-05, "loss": 0.3677, "step": 5779 }, { "epoch": 0.6, "learning_rate": 5.670051984028172e-05, "loss": 1.4543, "step": 5780 }, { "epoch": 0.6, "learning_rate": 5.669939752322108e-05, "loss": 1.1759, "step": 5781 }, { "epoch": 0.6, "learning_rate": 5.669827502642635e-05, "loss": 0.4756, "step": 5782 }, { "epoch": 0.6, "learning_rate": 5.669715234990506e-05, "loss": 1.0819, "step": 5783 }, { "epoch": 0.6, "learning_rate": 5.669602949366481e-05, "loss": 0.7253, "step": 5784 }, { "epoch": 0.6, "learning_rate": 5.6694906457713126e-05, "loss": 1.0947, "step": 5785 }, { "epoch": 0.6, "learning_rate": 5.6693783242057584e-05, "loss": 0.6666, "step": 5786 }, { "epoch": 0.6, "learning_rate": 5.6692659846705736e-05, "loss": 1.2276, "step": 5787 }, { "epoch": 0.6, "learning_rate": 5.669153627166514e-05, "loss": 0.8634, "step": 5788 }, { "epoch": 0.6, "learning_rate": 5.669041251694337e-05, "loss": 0.9266, "step": 5789 }, { "epoch": 0.6, "learning_rate": 5.6689288582547994e-05, "loss": 0.6564, "step": 5790 }, { "epoch": 0.6, "learning_rate": 5.668816446848657e-05, "loss": 0.7571, "step": 5791 }, { "epoch": 0.6, "learning_rate": 5.668704017476666e-05, "loss": 1.1896, "step": 5792 }, { "epoch": 0.61, "learning_rate": 5.668591570139585e-05, "loss": 1.0042, "step": 5793 }, { "epoch": 0.61, "learning_rate": 5.66847910483817e-05, "loss": 0.6755, "step": 5794 }, { "epoch": 0.61, "learning_rate": 5.668366621573177e-05, "loss": 1.2518, "step": 5795 }, { "epoch": 0.61, "learning_rate": 5.6682541203453645e-05, "loss": 1.0546, "step": 5796 }, { "epoch": 0.61, "learning_rate": 5.6681416011554905e-05, "loss": 1.2179, "step": 5797 }, { "epoch": 0.61, "learning_rate": 5.6680290640043106e-05, "loss": 1.1748, "step": 5798 }, { "epoch": 0.61, "learning_rate": 5.667916508892584e-05, "loss": 0.4583, "step": 5799 }, { "epoch": 0.61, "learning_rate": 5.667803935821067e-05, "loss": 0.3751, "step": 5800 }, { "epoch": 0.61, "learning_rate": 5.667691344790519e-05, "loss": 0.9977, "step": 5801 }, { "epoch": 0.61, "learning_rate": 5.667578735801696e-05, "loss": 0.6454, "step": 5802 }, { "epoch": 0.61, "learning_rate": 5.667466108855358e-05, "loss": 1.3788, "step": 5803 }, { "epoch": 0.61, "learning_rate": 5.667353463952262e-05, "loss": 0.919, "step": 5804 }, { "epoch": 0.61, "learning_rate": 5.667240801093166e-05, "loss": 1.4449, "step": 5805 }, { "epoch": 0.61, "learning_rate": 5.6671281202788295e-05, "loss": 0.7491, "step": 5806 }, { "epoch": 0.61, "learning_rate": 5.667015421510011e-05, "loss": 1.1111, "step": 5807 }, { "epoch": 0.61, "learning_rate": 5.666902704787469e-05, "loss": 0.286, "step": 5808 }, { "epoch": 0.61, "learning_rate": 5.666789970111961e-05, "loss": 0.7986, "step": 5809 }, { "epoch": 0.61, "learning_rate": 5.6666772174842475e-05, "loss": 1.3188, "step": 5810 }, { "epoch": 0.61, "learning_rate": 5.666564446905086e-05, "loss": 0.7722, "step": 5811 }, { "epoch": 0.61, "learning_rate": 5.666451658375237e-05, "loss": 1.4514, "step": 5812 }, { "epoch": 0.61, "learning_rate": 5.66633885189546e-05, "loss": 1.1461, "step": 5813 }, { "epoch": 0.61, "learning_rate": 5.6662260274665136e-05, "loss": 1.039, "step": 5814 }, { "epoch": 0.61, "learning_rate": 5.6661131850891565e-05, "loss": 0.4685, "step": 5815 }, { "epoch": 0.61, "learning_rate": 5.66600032476415e-05, "loss": 1.1313, "step": 5816 }, { "epoch": 0.61, "learning_rate": 5.665887446492253e-05, "loss": 0.4972, "step": 5817 }, { "epoch": 0.61, "learning_rate": 5.665774550274226e-05, "loss": 0.7652, "step": 5818 }, { "epoch": 0.61, "learning_rate": 5.665661636110828e-05, "loss": 1.1229, "step": 5819 }, { "epoch": 0.61, "learning_rate": 5.665548704002819e-05, "loss": 1.0796, "step": 5820 }, { "epoch": 0.61, "learning_rate": 5.6654357539509606e-05, "loss": 1.2291, "step": 5821 }, { "epoch": 0.61, "learning_rate": 5.665322785956012e-05, "loss": 0.5582, "step": 5822 }, { "epoch": 0.61, "learning_rate": 5.665209800018734e-05, "loss": 1.8677, "step": 5823 }, { "epoch": 0.61, "learning_rate": 5.6650967961398875e-05, "loss": 1.9242, "step": 5824 }, { "epoch": 0.61, "learning_rate": 5.664983774320233e-05, "loss": 0.5864, "step": 5825 }, { "epoch": 0.61, "learning_rate": 5.6648707345605303e-05, "loss": 0.6057, "step": 5826 }, { "epoch": 0.61, "learning_rate": 5.664757676861543e-05, "loss": 0.7136, "step": 5827 }, { "epoch": 0.61, "learning_rate": 5.664644601224029e-05, "loss": 1.3329, "step": 5828 }, { "epoch": 0.61, "learning_rate": 5.664531507648752e-05, "loss": 0.3462, "step": 5829 }, { "epoch": 0.61, "learning_rate": 5.664418396136472e-05, "loss": 1.5509, "step": 5830 }, { "epoch": 0.61, "learning_rate": 5.6643052666879504e-05, "loss": 1.6219, "step": 5831 }, { "epoch": 0.61, "learning_rate": 5.6641921193039494e-05, "loss": 0.8931, "step": 5832 }, { "epoch": 0.61, "learning_rate": 5.664078953985231e-05, "loss": 1.1558, "step": 5833 }, { "epoch": 0.61, "learning_rate": 5.6639657707325564e-05, "loss": 0.5075, "step": 5834 }, { "epoch": 0.61, "learning_rate": 5.663852569546686e-05, "loss": 0.9769, "step": 5835 }, { "epoch": 0.61, "learning_rate": 5.663739350428386e-05, "loss": 0.3959, "step": 5836 }, { "epoch": 0.61, "learning_rate": 5.663626113378414e-05, "loss": 0.5201, "step": 5837 }, { "epoch": 0.61, "learning_rate": 5.6635128583975345e-05, "loss": 0.63, "step": 5838 }, { "epoch": 0.61, "learning_rate": 5.6633995854865095e-05, "loss": 0.7042, "step": 5839 }, { "epoch": 0.61, "learning_rate": 5.663286294646102e-05, "loss": 1.5438, "step": 5840 }, { "epoch": 0.61, "learning_rate": 5.6631729858770745e-05, "loss": 0.5032, "step": 5841 }, { "epoch": 0.61, "learning_rate": 5.66305965918019e-05, "loss": 1.2419, "step": 5842 }, { "epoch": 0.61, "learning_rate": 5.66294631455621e-05, "loss": 1.1286, "step": 5843 }, { "epoch": 0.61, "learning_rate": 5.662832952005899e-05, "loss": 0.5421, "step": 5844 }, { "epoch": 0.61, "learning_rate": 5.66271957153002e-05, "loss": 0.6855, "step": 5845 }, { "epoch": 0.61, "learning_rate": 5.662606173129336e-05, "loss": 1.5229, "step": 5846 }, { "epoch": 0.61, "learning_rate": 5.662492756804609e-05, "loss": 1.3702, "step": 5847 }, { "epoch": 0.61, "learning_rate": 5.6623793225566045e-05, "loss": 0.5481, "step": 5848 }, { "epoch": 0.61, "learning_rate": 5.662265870386086e-05, "loss": 0.5853, "step": 5849 }, { "epoch": 0.61, "learning_rate": 5.662152400293815e-05, "loss": 0.4759, "step": 5850 }, { "epoch": 0.61, "learning_rate": 5.6620389122805575e-05, "loss": 1.051, "step": 5851 }, { "epoch": 0.61, "learning_rate": 5.661925406347078e-05, "loss": 0.341, "step": 5852 }, { "epoch": 0.61, "learning_rate": 5.661811882494138e-05, "loss": 1.0059, "step": 5853 }, { "epoch": 0.61, "learning_rate": 5.661698340722504e-05, "loss": 0.5816, "step": 5854 }, { "epoch": 0.61, "learning_rate": 5.661584781032939e-05, "loss": 0.5453, "step": 5855 }, { "epoch": 0.61, "learning_rate": 5.661471203426209e-05, "loss": 0.6822, "step": 5856 }, { "epoch": 0.61, "learning_rate": 5.661357607903077e-05, "loss": 1.0355, "step": 5857 }, { "epoch": 0.61, "learning_rate": 5.661243994464309e-05, "loss": 0.8012, "step": 5858 }, { "epoch": 0.61, "learning_rate": 5.661130363110668e-05, "loss": 0.6958, "step": 5859 }, { "epoch": 0.61, "learning_rate": 5.6610167138429204e-05, "loss": 0.9254, "step": 5860 }, { "epoch": 0.61, "learning_rate": 5.6609030466618316e-05, "loss": 0.147, "step": 5861 }, { "epoch": 0.61, "learning_rate": 5.660789361568165e-05, "loss": 1.4764, "step": 5862 }, { "epoch": 0.61, "learning_rate": 5.660675658562688e-05, "loss": 0.4752, "step": 5863 }, { "epoch": 0.61, "learning_rate": 5.660561937646164e-05, "loss": 0.8798, "step": 5864 }, { "epoch": 0.61, "learning_rate": 5.66044819881936e-05, "loss": 1.2848, "step": 5865 }, { "epoch": 0.61, "learning_rate": 5.660334442083042e-05, "loss": 1.532, "step": 5866 }, { "epoch": 0.61, "learning_rate": 5.660220667437974e-05, "loss": 1.4855, "step": 5867 }, { "epoch": 0.61, "learning_rate": 5.660106874884923e-05, "loss": 1.2762, "step": 5868 }, { "epoch": 0.61, "learning_rate": 5.659993064424655e-05, "loss": 0.376, "step": 5869 }, { "epoch": 0.61, "learning_rate": 5.659879236057937e-05, "loss": 0.7856, "step": 5870 }, { "epoch": 0.61, "learning_rate": 5.659765389785533e-05, "loss": 0.6524, "step": 5871 }, { "epoch": 0.61, "learning_rate": 5.6596515256082116e-05, "loss": 1.4203, "step": 5872 }, { "epoch": 0.61, "learning_rate": 5.659537643526738e-05, "loss": 1.4686, "step": 5873 }, { "epoch": 0.61, "learning_rate": 5.659423743541879e-05, "loss": 1.442, "step": 5874 }, { "epoch": 0.61, "learning_rate": 5.659309825654403e-05, "loss": 0.3293, "step": 5875 }, { "epoch": 0.61, "learning_rate": 5.6591958898650743e-05, "loss": 1.1896, "step": 5876 }, { "epoch": 0.61, "learning_rate": 5.6590819361746615e-05, "loss": 1.0552, "step": 5877 }, { "epoch": 0.61, "learning_rate": 5.658967964583931e-05, "loss": 0.5205, "step": 5878 }, { "epoch": 0.61, "learning_rate": 5.6588539750936506e-05, "loss": 0.6763, "step": 5879 }, { "epoch": 0.61, "learning_rate": 5.658739967704587e-05, "loss": 0.6901, "step": 5880 }, { "epoch": 0.61, "learning_rate": 5.658625942417508e-05, "loss": 1.2034, "step": 5881 }, { "epoch": 0.61, "learning_rate": 5.658511899233181e-05, "loss": 0.8056, "step": 5882 }, { "epoch": 0.61, "learning_rate": 5.658397838152375e-05, "loss": 0.897, "step": 5883 }, { "epoch": 0.61, "learning_rate": 5.658283759175855e-05, "loss": 0.7963, "step": 5884 }, { "epoch": 0.61, "learning_rate": 5.658169662304393e-05, "loss": 0.7952, "step": 5885 }, { "epoch": 0.61, "learning_rate": 5.658055547538753e-05, "loss": 0.9724, "step": 5886 }, { "epoch": 0.61, "learning_rate": 5.657941414879706e-05, "loss": 0.347, "step": 5887 }, { "epoch": 0.61, "learning_rate": 5.657827264328019e-05, "loss": 2.2936, "step": 5888 }, { "epoch": 0.62, "learning_rate": 5.657713095884461e-05, "loss": 0.3918, "step": 5889 }, { "epoch": 0.62, "learning_rate": 5.6575989095497996e-05, "loss": 1.548, "step": 5890 }, { "epoch": 0.62, "learning_rate": 5.657484705324804e-05, "loss": 0.5734, "step": 5891 }, { "epoch": 0.62, "learning_rate": 5.657370483210244e-05, "loss": 0.8063, "step": 5892 }, { "epoch": 0.62, "learning_rate": 5.6572562432068874e-05, "loss": 0.9186, "step": 5893 }, { "epoch": 0.62, "learning_rate": 5.657141985315504e-05, "loss": 0.9062, "step": 5894 }, { "epoch": 0.62, "learning_rate": 5.657027709536862e-05, "loss": 1.0542, "step": 5895 }, { "epoch": 0.62, "learning_rate": 5.656913415871731e-05, "loss": 0.899, "step": 5896 }, { "epoch": 0.62, "learning_rate": 5.6567991043208804e-05, "loss": 0.8698, "step": 5897 }, { "epoch": 0.62, "learning_rate": 5.656684774885081e-05, "loss": 1.5837, "step": 5898 }, { "epoch": 0.62, "learning_rate": 5.6565704275651e-05, "loss": 0.9012, "step": 5899 }, { "epoch": 0.62, "learning_rate": 5.6564560623617085e-05, "loss": 1.8254, "step": 5900 }, { "epoch": 0.62, "learning_rate": 5.6563416792756775e-05, "loss": 0.985, "step": 5901 }, { "epoch": 0.62, "learning_rate": 5.656227278307775e-05, "loss": 0.6646, "step": 5902 }, { "epoch": 0.62, "learning_rate": 5.656112859458771e-05, "loss": 0.6292, "step": 5903 }, { "epoch": 0.62, "learning_rate": 5.655998422729438e-05, "loss": 1.2672, "step": 5904 }, { "epoch": 0.62, "learning_rate": 5.6558839681205456e-05, "loss": 1.7217, "step": 5905 }, { "epoch": 0.62, "learning_rate": 5.655769495632863e-05, "loss": 0.329, "step": 5906 }, { "epoch": 0.62, "learning_rate": 5.6556550052671615e-05, "loss": 0.8821, "step": 5907 }, { "epoch": 0.62, "learning_rate": 5.655540497024212e-05, "loss": 1.0669, "step": 5908 }, { "epoch": 0.62, "learning_rate": 5.655425970904785e-05, "loss": 1.507, "step": 5909 }, { "epoch": 0.62, "learning_rate": 5.655311426909652e-05, "loss": 1.3659, "step": 5910 }, { "epoch": 0.62, "learning_rate": 5.655196865039584e-05, "loss": 1.0516, "step": 5911 }, { "epoch": 0.62, "learning_rate": 5.655082285295352e-05, "loss": 1.0679, "step": 5912 }, { "epoch": 0.62, "learning_rate": 5.654967687677726e-05, "loss": 0.6943, "step": 5913 }, { "epoch": 0.62, "learning_rate": 5.65485307218748e-05, "loss": 1.0049, "step": 5914 }, { "epoch": 0.62, "learning_rate": 5.6547384388253845e-05, "loss": 0.3622, "step": 5915 }, { "epoch": 0.62, "learning_rate": 5.65462378759221e-05, "loss": 1.4538, "step": 5916 }, { "epoch": 0.62, "learning_rate": 5.65450911848873e-05, "loss": 1.6248, "step": 5917 }, { "epoch": 0.62, "learning_rate": 5.6543944315157154e-05, "loss": 1.4472, "step": 5918 }, { "epoch": 0.62, "learning_rate": 5.6542797266739385e-05, "loss": 0.9281, "step": 5919 }, { "epoch": 0.62, "learning_rate": 5.6541650039641716e-05, "loss": 0.9242, "step": 5920 }, { "epoch": 0.62, "learning_rate": 5.6540502633871865e-05, "loss": 0.8468, "step": 5921 }, { "epoch": 0.62, "learning_rate": 5.6539355049437566e-05, "loss": 0.5763, "step": 5922 }, { "epoch": 0.62, "learning_rate": 5.6538207286346534e-05, "loss": 0.6099, "step": 5923 }, { "epoch": 0.62, "learning_rate": 5.653705934460651e-05, "loss": 0.8545, "step": 5924 }, { "epoch": 0.62, "learning_rate": 5.65359112242252e-05, "loss": 1.457, "step": 5925 }, { "epoch": 0.62, "learning_rate": 5.653476292521035e-05, "loss": 0.4657, "step": 5926 }, { "epoch": 0.62, "learning_rate": 5.653361444756969e-05, "loss": 0.1166, "step": 5927 }, { "epoch": 0.62, "learning_rate": 5.653246579131093e-05, "loss": 0.666, "step": 5928 }, { "epoch": 0.62, "learning_rate": 5.653131695644184e-05, "loss": 0.7322, "step": 5929 }, { "epoch": 0.62, "learning_rate": 5.6530167942970114e-05, "loss": 1.4047, "step": 5930 }, { "epoch": 0.62, "learning_rate": 5.652901875090351e-05, "loss": 1.224, "step": 5931 }, { "epoch": 0.62, "learning_rate": 5.6527869380249766e-05, "loss": 0.4187, "step": 5932 }, { "epoch": 0.62, "learning_rate": 5.6526719831016605e-05, "loss": 1.3084, "step": 5933 }, { "epoch": 0.62, "learning_rate": 5.6525570103211775e-05, "loss": 1.1097, "step": 5934 }, { "epoch": 0.62, "learning_rate": 5.6524420196843015e-05, "loss": 1.0587, "step": 5935 }, { "epoch": 0.62, "learning_rate": 5.652327011191806e-05, "loss": 1.3183, "step": 5936 }, { "epoch": 0.62, "learning_rate": 5.652211984844466e-05, "loss": 0.6581, "step": 5937 }, { "epoch": 0.62, "learning_rate": 5.652096940643056e-05, "loss": 0.7299, "step": 5938 }, { "epoch": 0.62, "learning_rate": 5.6519818785883494e-05, "loss": 0.8118, "step": 5939 }, { "epoch": 0.62, "learning_rate": 5.651866798681121e-05, "loss": 0.6463, "step": 5940 }, { "epoch": 0.62, "learning_rate": 5.651751700922146e-05, "loss": 0.9514, "step": 5941 }, { "epoch": 0.62, "learning_rate": 5.6516365853121993e-05, "loss": 1.0359, "step": 5942 }, { "epoch": 0.62, "learning_rate": 5.651521451852055e-05, "loss": 0.6481, "step": 5943 }, { "epoch": 0.62, "learning_rate": 5.651406300542489e-05, "loss": 0.7264, "step": 5944 }, { "epoch": 0.62, "learning_rate": 5.651291131384276e-05, "loss": 1.0077, "step": 5945 }, { "epoch": 0.62, "learning_rate": 5.651175944378192e-05, "loss": 1.2401, "step": 5946 }, { "epoch": 0.62, "learning_rate": 5.651060739525012e-05, "loss": 1.2869, "step": 5947 }, { "epoch": 0.62, "learning_rate": 5.650945516825511e-05, "loss": 0.9247, "step": 5948 }, { "epoch": 0.62, "learning_rate": 5.650830276280464e-05, "loss": 1.5267, "step": 5949 }, { "epoch": 0.62, "learning_rate": 5.650715017890649e-05, "loss": 0.8696, "step": 5950 }, { "epoch": 0.62, "learning_rate": 5.6505997416568414e-05, "loss": 0.667, "step": 5951 }, { "epoch": 0.62, "learning_rate": 5.650484447579815e-05, "loss": 1.0693, "step": 5952 }, { "epoch": 0.62, "learning_rate": 5.650369135660348e-05, "loss": 0.5828, "step": 5953 }, { "epoch": 0.62, "learning_rate": 5.650253805899216e-05, "loss": 0.522, "step": 5954 }, { "epoch": 0.62, "learning_rate": 5.650138458297196e-05, "loss": 0.5372, "step": 5955 }, { "epoch": 0.62, "learning_rate": 5.650023092855063e-05, "loss": 0.5812, "step": 5956 }, { "epoch": 0.62, "learning_rate": 5.649907709573596e-05, "loss": 1.3733, "step": 5957 }, { "epoch": 0.62, "learning_rate": 5.6497923084535694e-05, "loss": 0.7425, "step": 5958 }, { "epoch": 0.62, "learning_rate": 5.6496768894957614e-05, "loss": 0.9162, "step": 5959 }, { "epoch": 0.62, "learning_rate": 5.6495614527009487e-05, "loss": 0.4523, "step": 5960 }, { "epoch": 0.62, "learning_rate": 5.649445998069908e-05, "loss": 1.3241, "step": 5961 }, { "epoch": 0.62, "learning_rate": 5.6493305256034164e-05, "loss": 1.2177, "step": 5962 }, { "epoch": 0.62, "learning_rate": 5.649215035302251e-05, "loss": 1.6047, "step": 5963 }, { "epoch": 0.62, "learning_rate": 5.649099527167191e-05, "loss": 0.8041, "step": 5964 }, { "epoch": 0.62, "learning_rate": 5.648984001199013e-05, "loss": 1.3562, "step": 5965 }, { "epoch": 0.62, "learning_rate": 5.648868457398494e-05, "loss": 0.4785, "step": 5966 }, { "epoch": 0.62, "learning_rate": 5.6487528957664125e-05, "loss": 0.4458, "step": 5967 }, { "epoch": 0.62, "learning_rate": 5.648637316303546e-05, "loss": 0.5663, "step": 5968 }, { "epoch": 0.62, "learning_rate": 5.6485217190106725e-05, "loss": 0.6468, "step": 5969 }, { "epoch": 0.62, "learning_rate": 5.6484061038885714e-05, "loss": 1.6183, "step": 5970 }, { "epoch": 0.62, "learning_rate": 5.648290470938019e-05, "loss": 1.3549, "step": 5971 }, { "epoch": 0.62, "learning_rate": 5.648174820159796e-05, "loss": 1.0567, "step": 5972 }, { "epoch": 0.62, "learning_rate": 5.648059151554679e-05, "loss": 0.4921, "step": 5973 }, { "epoch": 0.62, "learning_rate": 5.647943465123448e-05, "loss": 0.368, "step": 5974 }, { "epoch": 0.62, "learning_rate": 5.6478277608668804e-05, "loss": 0.8454, "step": 5975 }, { "epoch": 0.62, "learning_rate": 5.647712038785756e-05, "loss": 0.6298, "step": 5976 }, { "epoch": 0.62, "learning_rate": 5.6475962988808545e-05, "loss": 0.7627, "step": 5977 }, { "epoch": 0.62, "learning_rate": 5.6474805411529525e-05, "loss": 1.3292, "step": 5978 }, { "epoch": 0.62, "learning_rate": 5.647364765602832e-05, "loss": 0.5772, "step": 5979 }, { "epoch": 0.62, "learning_rate": 5.647248972231272e-05, "loss": 1.379, "step": 5980 }, { "epoch": 0.62, "learning_rate": 5.64713316103905e-05, "loss": 0.2659, "step": 5981 }, { "epoch": 0.62, "learning_rate": 5.6470173320269486e-05, "loss": 1.1923, "step": 5982 }, { "epoch": 0.62, "learning_rate": 5.646901485195745e-05, "loss": 0.9678, "step": 5983 }, { "epoch": 0.62, "learning_rate": 5.6467856205462195e-05, "loss": 0.2897, "step": 5984 }, { "epoch": 0.63, "learning_rate": 5.646669738079153e-05, "loss": 0.7678, "step": 5985 }, { "epoch": 0.63, "learning_rate": 5.6465538377953244e-05, "loss": 1.8583, "step": 5986 }, { "epoch": 0.63, "learning_rate": 5.6464379196955154e-05, "loss": 0.1951, "step": 5987 }, { "epoch": 0.63, "learning_rate": 5.646321983780505e-05, "loss": 0.6117, "step": 5988 }, { "epoch": 0.63, "learning_rate": 5.646206030051075e-05, "loss": 1.3682, "step": 5989 }, { "epoch": 0.63, "learning_rate": 5.646090058508004e-05, "loss": 0.3359, "step": 5990 }, { "epoch": 0.63, "learning_rate": 5.645974069152074e-05, "loss": 1.5042, "step": 5991 }, { "epoch": 0.63, "learning_rate": 5.6458580619840666e-05, "loss": 1.1147, "step": 5992 }, { "epoch": 0.63, "learning_rate": 5.6457420370047614e-05, "loss": 1.3411, "step": 5993 }, { "epoch": 0.63, "learning_rate": 5.64562599421494e-05, "loss": 0.9945, "step": 5994 }, { "epoch": 0.63, "learning_rate": 5.645509933615383e-05, "loss": 0.789, "step": 5995 }, { "epoch": 0.63, "learning_rate": 5.645393855206873e-05, "loss": 0.7934, "step": 5996 }, { "epoch": 0.63, "learning_rate": 5.645277758990189e-05, "loss": 0.3559, "step": 5997 }, { "epoch": 0.63, "learning_rate": 5.645161644966116e-05, "loss": 0.5272, "step": 5998 }, { "epoch": 0.63, "learning_rate": 5.645045513135433e-05, "loss": 0.3806, "step": 5999 }, { "epoch": 0.63, "learning_rate": 5.6449293634989216e-05, "loss": 0.992, "step": 6000 }, { "epoch": 0.63, "learning_rate": 5.6448131960573645e-05, "loss": 1.3841, "step": 6001 }, { "epoch": 0.63, "learning_rate": 5.644697010811544e-05, "loss": 0.3341, "step": 6002 }, { "epoch": 0.63, "learning_rate": 5.6445808077622434e-05, "loss": 1.1623, "step": 6003 }, { "epoch": 0.63, "learning_rate": 5.644464586910243e-05, "loss": 1.1985, "step": 6004 }, { "epoch": 0.63, "learning_rate": 5.644348348256325e-05, "loss": 1.4134, "step": 6005 }, { "epoch": 0.63, "learning_rate": 5.644232091801273e-05, "loss": 0.2354, "step": 6006 }, { "epoch": 0.63, "learning_rate": 5.64411581754587e-05, "loss": 1.3282, "step": 6007 }, { "epoch": 0.63, "learning_rate": 5.6439995254908976e-05, "loss": 0.599, "step": 6008 }, { "epoch": 0.63, "learning_rate": 5.643883215637138e-05, "loss": 1.3686, "step": 6009 }, { "epoch": 0.63, "learning_rate": 5.643766887985376e-05, "loss": 1.4428, "step": 6010 }, { "epoch": 0.63, "learning_rate": 5.643650542536394e-05, "loss": 1.4, "step": 6011 }, { "epoch": 0.63, "learning_rate": 5.6435341792909755e-05, "loss": 0.931, "step": 6012 }, { "epoch": 0.63, "learning_rate": 5.643417798249903e-05, "loss": 1.2767, "step": 6013 }, { "epoch": 0.63, "learning_rate": 5.643301399413961e-05, "loss": 1.1203, "step": 6014 }, { "epoch": 0.63, "learning_rate": 5.6431849827839314e-05, "loss": 1.3933, "step": 6015 }, { "epoch": 0.63, "learning_rate": 5.6430685483605994e-05, "loss": 1.3972, "step": 6016 }, { "epoch": 0.63, "learning_rate": 5.642952096144748e-05, "loss": 0.7066, "step": 6017 }, { "epoch": 0.63, "learning_rate": 5.6428356261371616e-05, "loss": 0.836, "step": 6018 }, { "epoch": 0.63, "learning_rate": 5.6427191383386244e-05, "loss": 1.1278, "step": 6019 }, { "epoch": 0.63, "learning_rate": 5.64260263274992e-05, "loss": 0.3201, "step": 6020 }, { "epoch": 0.63, "learning_rate": 5.642486109371834e-05, "loss": 1.0147, "step": 6021 }, { "epoch": 0.63, "learning_rate": 5.642369568205149e-05, "loss": 0.8216, "step": 6022 }, { "epoch": 0.63, "learning_rate": 5.642253009250651e-05, "loss": 0.1381, "step": 6023 }, { "epoch": 0.63, "learning_rate": 5.6421364325091235e-05, "loss": 0.6234, "step": 6024 }, { "epoch": 0.63, "learning_rate": 5.6420198379813504e-05, "loss": 1.5204, "step": 6025 }, { "epoch": 0.63, "learning_rate": 5.6419032256681196e-05, "loss": 0.7605, "step": 6026 }, { "epoch": 0.63, "learning_rate": 5.641786595570214e-05, "loss": 0.6772, "step": 6027 }, { "epoch": 0.63, "learning_rate": 5.64166994768842e-05, "loss": 1.1739, "step": 6028 }, { "epoch": 0.63, "learning_rate": 5.64155328202352e-05, "loss": 0.4575, "step": 6029 }, { "epoch": 0.63, "learning_rate": 5.641436598576303e-05, "loss": 1.5878, "step": 6030 }, { "epoch": 0.63, "learning_rate": 5.6413198973475524e-05, "loss": 0.5582, "step": 6031 }, { "epoch": 0.63, "learning_rate": 5.641203178338054e-05, "loss": 0.5301, "step": 6032 }, { "epoch": 0.63, "learning_rate": 5.6410864415485944e-05, "loss": 0.6033, "step": 6033 }, { "epoch": 0.63, "learning_rate": 5.640969686979959e-05, "loss": 0.3636, "step": 6034 }, { "epoch": 0.63, "learning_rate": 5.6408529146329326e-05, "loss": 0.9008, "step": 6035 }, { "epoch": 0.63, "learning_rate": 5.640736124508303e-05, "loss": 1.5207, "step": 6036 }, { "epoch": 0.63, "learning_rate": 5.640619316606856e-05, "loss": 1.148, "step": 6037 }, { "epoch": 0.63, "learning_rate": 5.6405024909293764e-05, "loss": 1.048, "step": 6038 }, { "epoch": 0.63, "learning_rate": 5.6403856474766524e-05, "loss": 1.1801, "step": 6039 }, { "epoch": 0.63, "learning_rate": 5.6402687862494706e-05, "loss": 0.8986, "step": 6040 }, { "epoch": 0.63, "learning_rate": 5.640151907248617e-05, "loss": 0.8377, "step": 6041 }, { "epoch": 0.63, "learning_rate": 5.640035010474878e-05, "loss": 0.8185, "step": 6042 }, { "epoch": 0.63, "learning_rate": 5.639918095929042e-05, "loss": 0.5417, "step": 6043 }, { "epoch": 0.63, "learning_rate": 5.639801163611894e-05, "loss": 1.019, "step": 6044 }, { "epoch": 0.63, "learning_rate": 5.6396842135242224e-05, "loss": 0.8884, "step": 6045 }, { "epoch": 0.63, "learning_rate": 5.639567245666814e-05, "loss": 0.5022, "step": 6046 }, { "epoch": 0.63, "learning_rate": 5.6394502600404575e-05, "loss": 1.0388, "step": 6047 }, { "epoch": 0.63, "learning_rate": 5.639333256645939e-05, "loss": 0.9923, "step": 6048 }, { "epoch": 0.63, "learning_rate": 5.6392162354840467e-05, "loss": 1.1564, "step": 6049 }, { "epoch": 0.63, "learning_rate": 5.6390991965555684e-05, "loss": 1.1097, "step": 6050 }, { "epoch": 0.63, "learning_rate": 5.638982139861291e-05, "loss": 0.4188, "step": 6051 }, { "epoch": 0.63, "learning_rate": 5.638865065402004e-05, "loss": 1.175, "step": 6052 }, { "epoch": 0.63, "learning_rate": 5.6387479731784944e-05, "loss": 1.3921, "step": 6053 }, { "epoch": 0.63, "learning_rate": 5.63863086319155e-05, "loss": 0.8771, "step": 6054 }, { "epoch": 0.63, "learning_rate": 5.6385137354419616e-05, "loss": 1.7059, "step": 6055 }, { "epoch": 0.63, "learning_rate": 5.638396589930516e-05, "loss": 0.8506, "step": 6056 }, { "epoch": 0.63, "learning_rate": 5.6382794266580014e-05, "loss": 0.3521, "step": 6057 }, { "epoch": 0.63, "learning_rate": 5.638162245625207e-05, "loss": 1.086, "step": 6058 }, { "epoch": 0.63, "learning_rate": 5.638045046832921e-05, "loss": 0.9973, "step": 6059 }, { "epoch": 0.63, "learning_rate": 5.6379278302819344e-05, "loss": 1.2519, "step": 6060 }, { "epoch": 0.63, "learning_rate": 5.637810595973034e-05, "loss": 0.8602, "step": 6061 }, { "epoch": 0.63, "learning_rate": 5.637693343907011e-05, "loss": 0.4807, "step": 6062 }, { "epoch": 0.63, "learning_rate": 5.637576074084652e-05, "loss": 1.2701, "step": 6063 }, { "epoch": 0.63, "learning_rate": 5.6374587865067476e-05, "loss": 0.7174, "step": 6064 }, { "epoch": 0.63, "learning_rate": 5.6373414811740896e-05, "loss": 0.4296, "step": 6065 }, { "epoch": 0.63, "learning_rate": 5.637224158087465e-05, "loss": 1.2778, "step": 6066 }, { "epoch": 0.63, "learning_rate": 5.637106817247664e-05, "loss": 0.5707, "step": 6067 }, { "epoch": 0.63, "learning_rate": 5.636989458655478e-05, "loss": 0.406, "step": 6068 }, { "epoch": 0.63, "learning_rate": 5.636872082311695e-05, "loss": 0.2312, "step": 6069 }, { "epoch": 0.63, "learning_rate": 5.636754688217106e-05, "loss": 0.3848, "step": 6070 }, { "epoch": 0.63, "learning_rate": 5.636637276372502e-05, "loss": 1.2528, "step": 6071 }, { "epoch": 0.63, "learning_rate": 5.6365198467786733e-05, "loss": 0.8088, "step": 6072 }, { "epoch": 0.63, "learning_rate": 5.636402399436409e-05, "loss": 0.6457, "step": 6073 }, { "epoch": 0.63, "learning_rate": 5.636284934346501e-05, "loss": 0.4511, "step": 6074 }, { "epoch": 0.63, "learning_rate": 5.63616745150974e-05, "loss": 1.4779, "step": 6075 }, { "epoch": 0.63, "learning_rate": 5.6360499509269155e-05, "loss": 1.1238, "step": 6076 }, { "epoch": 0.63, "learning_rate": 5.63593243259882e-05, "loss": 1.0709, "step": 6077 }, { "epoch": 0.63, "learning_rate": 5.635814896526244e-05, "loss": 0.4854, "step": 6078 }, { "epoch": 0.63, "learning_rate": 5.63569734270998e-05, "loss": 1.0094, "step": 6079 }, { "epoch": 0.63, "learning_rate": 5.635579771150817e-05, "loss": 0.8117, "step": 6080 }, { "epoch": 0.64, "learning_rate": 5.6354621818495474e-05, "loss": 0.8485, "step": 6081 }, { "epoch": 0.64, "learning_rate": 5.635344574806964e-05, "loss": 1.6228, "step": 6082 }, { "epoch": 0.64, "learning_rate": 5.635226950023857e-05, "loss": 0.9374, "step": 6083 }, { "epoch": 0.64, "learning_rate": 5.635109307501019e-05, "loss": 0.3116, "step": 6084 }, { "epoch": 0.64, "learning_rate": 5.6349916472392414e-05, "loss": 0.5461, "step": 6085 }, { "epoch": 0.64, "learning_rate": 5.634873969239317e-05, "loss": 1.2054, "step": 6086 }, { "epoch": 0.64, "learning_rate": 5.634756273502037e-05, "loss": 0.8218, "step": 6087 }, { "epoch": 0.64, "learning_rate": 5.634638560028194e-05, "loss": 0.8189, "step": 6088 }, { "epoch": 0.64, "learning_rate": 5.6345208288185815e-05, "loss": 0.4367, "step": 6089 }, { "epoch": 0.64, "learning_rate": 5.634403079873991e-05, "loss": 1.0603, "step": 6090 }, { "epoch": 0.64, "learning_rate": 5.634285313195215e-05, "loss": 0.8205, "step": 6091 }, { "epoch": 0.64, "learning_rate": 5.634167528783047e-05, "loss": 1.6135, "step": 6092 }, { "epoch": 0.64, "learning_rate": 5.6340497266382794e-05, "loss": 0.2217, "step": 6093 }, { "epoch": 0.64, "learning_rate": 5.633931906761706e-05, "loss": 0.6823, "step": 6094 }, { "epoch": 0.64, "learning_rate": 5.633814069154118e-05, "loss": 0.3484, "step": 6095 }, { "epoch": 0.64, "learning_rate": 5.6336962138163104e-05, "loss": 0.6446, "step": 6096 }, { "epoch": 0.64, "learning_rate": 5.633578340749077e-05, "loss": 0.4449, "step": 6097 }, { "epoch": 0.64, "learning_rate": 5.6334604499532096e-05, "loss": 0.7151, "step": 6098 }, { "epoch": 0.64, "learning_rate": 5.633342541429503e-05, "loss": 1.5055, "step": 6099 }, { "epoch": 0.64, "learning_rate": 5.6332246151787506e-05, "loss": 1.248, "step": 6100 }, { "epoch": 0.64, "learning_rate": 5.6331066712017454e-05, "loss": 0.6568, "step": 6101 }, { "epoch": 0.64, "learning_rate": 5.632988709499283e-05, "loss": 0.555, "step": 6102 }, { "epoch": 0.64, "learning_rate": 5.6328707300721566e-05, "loss": 0.4922, "step": 6103 }, { "epoch": 0.64, "learning_rate": 5.6327527329211606e-05, "loss": 0.5069, "step": 6104 }, { "epoch": 0.64, "learning_rate": 5.6326347180470884e-05, "loss": 1.2863, "step": 6105 }, { "epoch": 0.64, "learning_rate": 5.632516685450736e-05, "loss": 0.3974, "step": 6106 }, { "epoch": 0.64, "learning_rate": 5.632398635132897e-05, "loss": 0.9586, "step": 6107 }, { "epoch": 0.64, "learning_rate": 5.6322805670943663e-05, "loss": 1.156, "step": 6108 }, { "epoch": 0.64, "learning_rate": 5.632162481335939e-05, "loss": 1.0483, "step": 6109 }, { "epoch": 0.64, "learning_rate": 5.63204437785841e-05, "loss": 0.686, "step": 6110 }, { "epoch": 0.64, "learning_rate": 5.631926256662574e-05, "loss": 0.5537, "step": 6111 }, { "epoch": 0.64, "learning_rate": 5.631808117749226e-05, "loss": 0.7644, "step": 6112 }, { "epoch": 0.64, "learning_rate": 5.6316899611191605e-05, "loss": 1.0051, "step": 6113 }, { "epoch": 0.64, "learning_rate": 5.6315717867731756e-05, "loss": 0.7781, "step": 6114 }, { "epoch": 0.64, "learning_rate": 5.631453594712064e-05, "loss": 0.38, "step": 6115 }, { "epoch": 0.64, "learning_rate": 5.631335384936624e-05, "loss": 0.7031, "step": 6116 }, { "epoch": 0.64, "learning_rate": 5.631217157447648e-05, "loss": 0.6466, "step": 6117 }, { "epoch": 0.64, "learning_rate": 5.631098912245935e-05, "loss": 0.6344, "step": 6118 }, { "epoch": 0.64, "learning_rate": 5.6309806493322794e-05, "loss": 1.3416, "step": 6119 }, { "epoch": 0.64, "learning_rate": 5.630862368707477e-05, "loss": 0.4257, "step": 6120 }, { "epoch": 0.64, "learning_rate": 5.630744070372326e-05, "loss": 1.147, "step": 6121 }, { "epoch": 0.64, "learning_rate": 5.630625754327621e-05, "loss": 1.3633, "step": 6122 }, { "epoch": 0.64, "learning_rate": 5.630507420574158e-05, "loss": 1.0298, "step": 6123 }, { "epoch": 0.64, "learning_rate": 5.630389069112735e-05, "loss": 0.5685, "step": 6124 }, { "epoch": 0.64, "learning_rate": 5.630270699944149e-05, "loss": 0.6119, "step": 6125 }, { "epoch": 0.64, "learning_rate": 5.630152313069195e-05, "loss": 0.2413, "step": 6126 }, { "epoch": 0.64, "learning_rate": 5.630033908488672e-05, "loss": 0.2961, "step": 6127 }, { "epoch": 0.64, "learning_rate": 5.629915486203376e-05, "loss": 1.1165, "step": 6128 }, { "epoch": 0.64, "learning_rate": 5.629797046214104e-05, "loss": 1.2618, "step": 6129 }, { "epoch": 0.64, "learning_rate": 5.629678588521654e-05, "loss": 0.2758, "step": 6130 }, { "epoch": 0.64, "learning_rate": 5.6295601131268226e-05, "loss": 1.1177, "step": 6131 }, { "epoch": 0.64, "learning_rate": 5.629441620030408e-05, "loss": 1.1656, "step": 6132 }, { "epoch": 0.64, "learning_rate": 5.6293231092332076e-05, "loss": 0.6746, "step": 6133 }, { "epoch": 0.64, "learning_rate": 5.629204580736019e-05, "loss": 0.8028, "step": 6134 }, { "epoch": 0.64, "learning_rate": 5.6290860345396404e-05, "loss": 1.5024, "step": 6135 }, { "epoch": 0.64, "learning_rate": 5.62896747064487e-05, "loss": 1.1153, "step": 6136 }, { "epoch": 0.64, "learning_rate": 5.628848889052506e-05, "loss": 0.8896, "step": 6137 }, { "epoch": 0.64, "learning_rate": 5.628730289763346e-05, "loss": 0.4525, "step": 6138 }, { "epoch": 0.64, "learning_rate": 5.628611672778188e-05, "loss": 0.6761, "step": 6139 }, { "epoch": 0.64, "learning_rate": 5.6284930380978326e-05, "loss": 1.3478, "step": 6140 }, { "epoch": 0.64, "learning_rate": 5.6283743857230765e-05, "loss": 0.4287, "step": 6141 }, { "epoch": 0.64, "learning_rate": 5.628255715654719e-05, "loss": 0.7173, "step": 6142 }, { "epoch": 0.64, "learning_rate": 5.628137027893559e-05, "loss": 0.5805, "step": 6143 }, { "epoch": 0.64, "learning_rate": 5.6280183224403955e-05, "loss": 0.4761, "step": 6144 }, { "epoch": 0.64, "learning_rate": 5.627899599296028e-05, "loss": 0.7129, "step": 6145 }, { "epoch": 0.64, "learning_rate": 5.6277808584612545e-05, "loss": 0.7623, "step": 6146 }, { "epoch": 0.64, "learning_rate": 5.627662099936875e-05, "loss": 0.8775, "step": 6147 }, { "epoch": 0.64, "learning_rate": 5.62754332372369e-05, "loss": 1.159, "step": 6148 }, { "epoch": 0.64, "learning_rate": 5.627424529822498e-05, "loss": 1.2727, "step": 6149 }, { "epoch": 0.64, "learning_rate": 5.627305718234098e-05, "loss": 1.2845, "step": 6150 }, { "epoch": 0.64, "learning_rate": 5.627186888959291e-05, "loss": 0.3932, "step": 6151 }, { "epoch": 0.64, "learning_rate": 5.627068041998877e-05, "loss": 1.3127, "step": 6152 }, { "epoch": 0.64, "learning_rate": 5.626949177353655e-05, "loss": 1.4397, "step": 6153 }, { "epoch": 0.64, "learning_rate": 5.626830295024426e-05, "loss": 1.2382, "step": 6154 }, { "epoch": 0.64, "learning_rate": 5.6267113950119904e-05, "loss": 0.7211, "step": 6155 }, { "epoch": 0.64, "learning_rate": 5.626592477317149e-05, "loss": 0.3717, "step": 6156 }, { "epoch": 0.64, "learning_rate": 5.6264735419407e-05, "loss": 0.5944, "step": 6157 }, { "epoch": 0.64, "learning_rate": 5.626354588883446e-05, "loss": 0.8888, "step": 6158 }, { "epoch": 0.64, "learning_rate": 5.6262356181461886e-05, "loss": 0.516, "step": 6159 }, { "epoch": 0.64, "learning_rate": 5.626116629729726e-05, "loss": 0.6741, "step": 6160 }, { "epoch": 0.64, "learning_rate": 5.625997623634862e-05, "loss": 0.8431, "step": 6161 }, { "epoch": 0.64, "learning_rate": 5.6258785998623964e-05, "loss": 1.0238, "step": 6162 }, { "epoch": 0.64, "learning_rate": 5.625759558413129e-05, "loss": 0.443, "step": 6163 }, { "epoch": 0.64, "learning_rate": 5.625640499287864e-05, "loss": 0.8947, "step": 6164 }, { "epoch": 0.64, "learning_rate": 5.625521422487401e-05, "loss": 1.7384, "step": 6165 }, { "epoch": 0.64, "learning_rate": 5.625402328012542e-05, "loss": 0.2736, "step": 6166 }, { "epoch": 0.64, "learning_rate": 5.625283215864089e-05, "loss": 1.079, "step": 6167 }, { "epoch": 0.64, "learning_rate": 5.625164086042844e-05, "loss": 1.3775, "step": 6168 }, { "epoch": 0.64, "learning_rate": 5.6250449385496085e-05, "loss": 0.9958, "step": 6169 }, { "epoch": 0.64, "learning_rate": 5.624925773385184e-05, "loss": 0.9018, "step": 6170 }, { "epoch": 0.64, "learning_rate": 5.624806590550374e-05, "loss": 0.6945, "step": 6171 }, { "epoch": 0.64, "learning_rate": 5.62468739004598e-05, "loss": 0.3336, "step": 6172 }, { "epoch": 0.64, "learning_rate": 5.624568171872805e-05, "loss": 0.4134, "step": 6173 }, { "epoch": 0.64, "learning_rate": 5.62444893603165e-05, "loss": 1.1592, "step": 6174 }, { "epoch": 0.64, "learning_rate": 5.62432968252332e-05, "loss": 1.3843, "step": 6175 }, { "epoch": 0.65, "learning_rate": 5.6242104113486164e-05, "loss": 1.1634, "step": 6176 }, { "epoch": 0.65, "learning_rate": 5.6240911225083416e-05, "loss": 1.1482, "step": 6177 }, { "epoch": 0.65, "learning_rate": 5.6239718160033e-05, "loss": 0.8942, "step": 6178 }, { "epoch": 0.65, "learning_rate": 5.623852491834294e-05, "loss": 1.7214, "step": 6179 }, { "epoch": 0.65, "learning_rate": 5.623733150002127e-05, "loss": 0.7431, "step": 6180 }, { "epoch": 0.65, "learning_rate": 5.6236137905076025e-05, "loss": 1.3164, "step": 6181 }, { "epoch": 0.65, "learning_rate": 5.6234944133515235e-05, "loss": 0.3551, "step": 6182 }, { "epoch": 0.65, "learning_rate": 5.623375018534694e-05, "loss": 1.4797, "step": 6183 }, { "epoch": 0.65, "learning_rate": 5.623255606057919e-05, "loss": 0.9591, "step": 6184 }, { "epoch": 0.65, "learning_rate": 5.623136175921999e-05, "loss": 0.346, "step": 6185 }, { "epoch": 0.65, "learning_rate": 5.6230167281277414e-05, "loss": 0.505, "step": 6186 }, { "epoch": 0.65, "learning_rate": 5.6228972626759485e-05, "loss": 0.6986, "step": 6187 }, { "epoch": 0.65, "learning_rate": 5.622777779567425e-05, "loss": 0.2947, "step": 6188 }, { "epoch": 0.65, "learning_rate": 5.622658278802975e-05, "loss": 1.2212, "step": 6189 }, { "epoch": 0.65, "learning_rate": 5.622538760383403e-05, "loss": 1.0178, "step": 6190 }, { "epoch": 0.65, "learning_rate": 5.622419224309514e-05, "loss": 0.6564, "step": 6191 }, { "epoch": 0.65, "learning_rate": 5.622299670582112e-05, "loss": 0.5176, "step": 6192 }, { "epoch": 0.65, "learning_rate": 5.622180099202002e-05, "loss": 1.0095, "step": 6193 }, { "epoch": 0.65, "learning_rate": 5.62206051016999e-05, "loss": 1.2945, "step": 6194 }, { "epoch": 0.65, "learning_rate": 5.621940903486879e-05, "loss": 0.7451, "step": 6195 }, { "epoch": 0.65, "learning_rate": 5.621821279153476e-05, "loss": 0.5372, "step": 6196 }, { "epoch": 0.65, "learning_rate": 5.6217016371705865e-05, "loss": 0.1548, "step": 6197 }, { "epoch": 0.65, "learning_rate": 5.621581977539014e-05, "loss": 0.4843, "step": 6198 }, { "epoch": 0.65, "learning_rate": 5.621462300259565e-05, "loss": 0.9855, "step": 6199 }, { "epoch": 0.65, "learning_rate": 5.621342605333045e-05, "loss": 0.4517, "step": 6200 }, { "epoch": 0.65, "learning_rate": 5.62122289276026e-05, "loss": 0.4509, "step": 6201 }, { "epoch": 0.65, "learning_rate": 5.6211031625420166e-05, "loss": 0.5499, "step": 6202 }, { "epoch": 0.65, "learning_rate": 5.62098341467912e-05, "loss": 0.627, "step": 6203 }, { "epoch": 0.65, "learning_rate": 5.6208636491723755e-05, "loss": 0.4333, "step": 6204 }, { "epoch": 0.65, "learning_rate": 5.62074386602259e-05, "loss": 0.541, "step": 6205 }, { "epoch": 0.65, "learning_rate": 5.6206240652305704e-05, "loss": 0.3248, "step": 6206 }, { "epoch": 0.65, "learning_rate": 5.620504246797123e-05, "loss": 1.1438, "step": 6207 }, { "epoch": 0.65, "learning_rate": 5.620384410723054e-05, "loss": 0.7464, "step": 6208 }, { "epoch": 0.65, "learning_rate": 5.6202645570091696e-05, "loss": 0.6251, "step": 6209 }, { "epoch": 0.65, "learning_rate": 5.620144685656278e-05, "loss": 0.2663, "step": 6210 }, { "epoch": 0.65, "learning_rate": 5.620024796665186e-05, "loss": 0.4472, "step": 6211 }, { "epoch": 0.65, "learning_rate": 5.619904890036699e-05, "loss": 0.7908, "step": 6212 }, { "epoch": 0.65, "learning_rate": 5.619784965771626e-05, "loss": 0.3709, "step": 6213 }, { "epoch": 0.65, "learning_rate": 5.619665023870773e-05, "loss": 0.5401, "step": 6214 }, { "epoch": 0.65, "learning_rate": 5.6195450643349487e-05, "loss": 0.4026, "step": 6215 }, { "epoch": 0.65, "learning_rate": 5.6194250871649597e-05, "loss": 0.4625, "step": 6216 }, { "epoch": 0.65, "learning_rate": 5.6193050923616136e-05, "loss": 0.943, "step": 6217 }, { "epoch": 0.65, "learning_rate": 5.619185079925719e-05, "loss": 1.1036, "step": 6218 }, { "epoch": 0.65, "learning_rate": 5.6190650498580825e-05, "loss": 0.8375, "step": 6219 }, { "epoch": 0.65, "learning_rate": 5.618945002159514e-05, "loss": 0.4552, "step": 6220 }, { "epoch": 0.65, "learning_rate": 5.61882493683082e-05, "loss": 1.0835, "step": 6221 }, { "epoch": 0.65, "learning_rate": 5.6187048538728086e-05, "loss": 1.2604, "step": 6222 }, { "epoch": 0.65, "learning_rate": 5.618584753286289e-05, "loss": 0.6594, "step": 6223 }, { "epoch": 0.65, "learning_rate": 5.6184646350720704e-05, "loss": 0.7163, "step": 6224 }, { "epoch": 0.65, "learning_rate": 5.61834449923096e-05, "loss": 1.1604, "step": 6225 }, { "epoch": 0.65, "learning_rate": 5.6182243457637674e-05, "loss": 1.3349, "step": 6226 }, { "epoch": 0.65, "learning_rate": 5.618104174671301e-05, "loss": 0.8889, "step": 6227 }, { "epoch": 0.65, "learning_rate": 5.61798398595437e-05, "loss": 0.3117, "step": 6228 }, { "epoch": 0.65, "learning_rate": 5.6178637796137826e-05, "loss": 1.8374, "step": 6229 }, { "epoch": 0.65, "learning_rate": 5.61774355565035e-05, "loss": 1.2793, "step": 6230 }, { "epoch": 0.65, "learning_rate": 5.6176233140648794e-05, "loss": 0.8264, "step": 6231 }, { "epoch": 0.65, "learning_rate": 5.6175030548581815e-05, "loss": 1.3193, "step": 6232 }, { "epoch": 0.65, "learning_rate": 5.6173827780310645e-05, "loss": 1.2628, "step": 6233 }, { "epoch": 0.65, "learning_rate": 5.61726248358434e-05, "loss": 1.3431, "step": 6234 }, { "epoch": 0.65, "learning_rate": 5.6171421715188176e-05, "loss": 0.7926, "step": 6235 }, { "epoch": 0.65, "learning_rate": 5.6170218418353044e-05, "loss": 0.8338, "step": 6236 }, { "epoch": 0.65, "learning_rate": 5.616901494534614e-05, "loss": 1.0196, "step": 6237 }, { "epoch": 0.65, "learning_rate": 5.616781129617555e-05, "loss": 0.3837, "step": 6238 }, { "epoch": 0.65, "learning_rate": 5.616660747084938e-05, "loss": 0.9832, "step": 6239 }, { "epoch": 0.65, "learning_rate": 5.616540346937572e-05, "loss": 0.522, "step": 6240 }, { "epoch": 0.65, "learning_rate": 5.616419929176269e-05, "loss": 0.3974, "step": 6241 }, { "epoch": 0.65, "learning_rate": 5.6162994938018395e-05, "loss": 0.5779, "step": 6242 }, { "epoch": 0.65, "learning_rate": 5.616179040815094e-05, "loss": 1.0606, "step": 6243 }, { "epoch": 0.65, "learning_rate": 5.616058570216843e-05, "loss": 1.389, "step": 6244 }, { "epoch": 0.65, "learning_rate": 5.6159380820078985e-05, "loss": 0.3447, "step": 6245 }, { "epoch": 0.65, "learning_rate": 5.61581757618907e-05, "loss": 0.2643, "step": 6246 }, { "epoch": 0.65, "learning_rate": 5.61569705276117e-05, "loss": 0.9901, "step": 6247 }, { "epoch": 0.65, "learning_rate": 5.6155765117250095e-05, "loss": 1.6044, "step": 6248 }, { "epoch": 0.65, "learning_rate": 5.6154559530814e-05, "loss": 0.4401, "step": 6249 }, { "epoch": 0.65, "learning_rate": 5.615335376831153e-05, "loss": 1.3755, "step": 6250 }, { "epoch": 0.65, "learning_rate": 5.61521478297508e-05, "loss": 0.4822, "step": 6251 }, { "epoch": 0.65, "learning_rate": 5.615094171513993e-05, "loss": 0.2907, "step": 6252 }, { "epoch": 0.65, "learning_rate": 5.614973542448704e-05, "loss": 1.2254, "step": 6253 }, { "epoch": 0.65, "learning_rate": 5.6148528957800255e-05, "loss": 0.6023, "step": 6254 }, { "epoch": 0.65, "learning_rate": 5.614732231508768e-05, "loss": 0.3254, "step": 6255 }, { "epoch": 0.65, "learning_rate": 5.614611549635746e-05, "loss": 1.2734, "step": 6256 }, { "epoch": 0.65, "learning_rate": 5.61449085016177e-05, "loss": 1.3482, "step": 6257 }, { "epoch": 0.65, "learning_rate": 5.614370133087654e-05, "loss": 1.3307, "step": 6258 }, { "epoch": 0.65, "learning_rate": 5.61424939841421e-05, "loss": 0.5308, "step": 6259 }, { "epoch": 0.65, "learning_rate": 5.61412864614225e-05, "loss": 1.3424, "step": 6260 }, { "epoch": 0.65, "learning_rate": 5.6140078762725875e-05, "loss": 0.5539, "step": 6261 }, { "epoch": 0.65, "learning_rate": 5.613887088806037e-05, "loss": 1.715, "step": 6262 }, { "epoch": 0.65, "learning_rate": 5.6137662837434084e-05, "loss": 0.8547, "step": 6263 }, { "epoch": 0.65, "learning_rate": 5.613645461085518e-05, "loss": 0.9986, "step": 6264 }, { "epoch": 0.65, "learning_rate": 5.613524620833178e-05, "loss": 1.4423, "step": 6265 }, { "epoch": 0.65, "learning_rate": 5.613403762987201e-05, "loss": 0.644, "step": 6266 }, { "epoch": 0.65, "learning_rate": 5.6132828875484014e-05, "loss": 1.103, "step": 6267 }, { "epoch": 0.65, "learning_rate": 5.613161994517594e-05, "loss": 1.2261, "step": 6268 }, { "epoch": 0.65, "learning_rate": 5.6130410838955907e-05, "loss": 0.7492, "step": 6269 }, { "epoch": 0.65, "learning_rate": 5.612920155683206e-05, "loss": 1.051, "step": 6270 }, { "epoch": 0.65, "learning_rate": 5.6127992098812546e-05, "loss": 0.8439, "step": 6271 }, { "epoch": 0.66, "learning_rate": 5.6126782464905495e-05, "loss": 1.054, "step": 6272 }, { "epoch": 0.66, "learning_rate": 5.6125572655119066e-05, "loss": 0.5032, "step": 6273 }, { "epoch": 0.66, "learning_rate": 5.612436266946139e-05, "loss": 0.4175, "step": 6274 }, { "epoch": 0.66, "learning_rate": 5.612315250794062e-05, "loss": 1.6403, "step": 6275 }, { "epoch": 0.66, "learning_rate": 5.6121942170564894e-05, "loss": 1.191, "step": 6276 }, { "epoch": 0.66, "learning_rate": 5.612073165734237e-05, "loss": 0.2948, "step": 6277 }, { "epoch": 0.66, "learning_rate": 5.61195209682812e-05, "loss": 0.8323, "step": 6278 }, { "epoch": 0.66, "learning_rate": 5.6118310103389514e-05, "loss": 0.7082, "step": 6279 }, { "epoch": 0.66, "learning_rate": 5.611709906267548e-05, "loss": 0.4071, "step": 6280 }, { "epoch": 0.66, "learning_rate": 5.611588784614725e-05, "loss": 0.3881, "step": 6281 }, { "epoch": 0.66, "learning_rate": 5.6114676453812966e-05, "loss": 0.5793, "step": 6282 }, { "epoch": 0.66, "learning_rate": 5.6113464885680796e-05, "loss": 0.6522, "step": 6283 }, { "epoch": 0.66, "learning_rate": 5.611225314175889e-05, "loss": 0.4407, "step": 6284 }, { "epoch": 0.66, "learning_rate": 5.6111041222055404e-05, "loss": 1.532, "step": 6285 }, { "epoch": 0.66, "learning_rate": 5.61098291265785e-05, "loss": 1.1865, "step": 6286 }, { "epoch": 0.66, "learning_rate": 5.610861685533633e-05, "loss": 1.2212, "step": 6287 }, { "epoch": 0.66, "learning_rate": 5.610740440833706e-05, "loss": 0.4063, "step": 6288 }, { "epoch": 0.66, "learning_rate": 5.6106191785588856e-05, "loss": 1.0155, "step": 6289 }, { "epoch": 0.66, "learning_rate": 5.6104978987099886e-05, "loss": 0.8917, "step": 6290 }, { "epoch": 0.66, "learning_rate": 5.610376601287829e-05, "loss": 0.6514, "step": 6291 }, { "epoch": 0.66, "learning_rate": 5.610255286293226e-05, "loss": 0.9886, "step": 6292 }, { "epoch": 0.66, "learning_rate": 5.6101339537269954e-05, "loss": 0.2787, "step": 6293 }, { "epoch": 0.66, "learning_rate": 5.610012603589952e-05, "loss": 0.7478, "step": 6294 }, { "epoch": 0.66, "learning_rate": 5.6098912358829156e-05, "loss": 1.1252, "step": 6295 }, { "epoch": 0.66, "learning_rate": 5.609769850606703e-05, "loss": 0.4612, "step": 6296 }, { "epoch": 0.66, "learning_rate": 5.609648447762128e-05, "loss": 0.5721, "step": 6297 }, { "epoch": 0.66, "learning_rate": 5.6095270273500124e-05, "loss": 0.7539, "step": 6298 }, { "epoch": 0.66, "learning_rate": 5.6094055893711706e-05, "loss": 1.9768, "step": 6299 }, { "epoch": 0.66, "learning_rate": 5.609284133826421e-05, "loss": 0.8529, "step": 6300 }, { "epoch": 0.66, "learning_rate": 5.6091626607165815e-05, "loss": 0.2275, "step": 6301 }, { "epoch": 0.66, "learning_rate": 5.609041170042469e-05, "loss": 1.2331, "step": 6302 }, { "epoch": 0.66, "learning_rate": 5.6089196618049015e-05, "loss": 0.8809, "step": 6303 }, { "epoch": 0.66, "learning_rate": 5.6087981360046984e-05, "loss": 1.0345, "step": 6304 }, { "epoch": 0.66, "learning_rate": 5.608676592642675e-05, "loss": 1.2109, "step": 6305 }, { "epoch": 0.66, "learning_rate": 5.608555031719653e-05, "loss": 0.2769, "step": 6306 }, { "epoch": 0.66, "learning_rate": 5.6084334532364484e-05, "loss": 1.4927, "step": 6307 }, { "epoch": 0.66, "learning_rate": 5.6083118571938785e-05, "loss": 1.5636, "step": 6308 }, { "epoch": 0.66, "learning_rate": 5.608190243592765e-05, "loss": 0.8566, "step": 6309 }, { "epoch": 0.66, "learning_rate": 5.608068612433925e-05, "loss": 1.0198, "step": 6310 }, { "epoch": 0.66, "learning_rate": 5.6079469637181774e-05, "loss": 0.9274, "step": 6311 }, { "epoch": 0.66, "learning_rate": 5.607825297446341e-05, "loss": 1.8074, "step": 6312 }, { "epoch": 0.66, "learning_rate": 5.6077036136192344e-05, "loss": 1.3805, "step": 6313 }, { "epoch": 0.66, "learning_rate": 5.6075819122376774e-05, "loss": 1.7449, "step": 6314 }, { "epoch": 0.66, "learning_rate": 5.6074601933024894e-05, "loss": 1.1402, "step": 6315 }, { "epoch": 0.66, "learning_rate": 5.607338456814489e-05, "loss": 0.1796, "step": 6316 }, { "epoch": 0.66, "learning_rate": 5.607216702774496e-05, "loss": 0.4659, "step": 6317 }, { "epoch": 0.66, "learning_rate": 5.607094931183331e-05, "loss": 0.8482, "step": 6318 }, { "epoch": 0.66, "learning_rate": 5.606973142041813e-05, "loss": 0.4422, "step": 6319 }, { "epoch": 0.66, "learning_rate": 5.6068513353507604e-05, "loss": 0.9774, "step": 6320 }, { "epoch": 0.66, "learning_rate": 5.606729511110996e-05, "loss": 0.4083, "step": 6321 }, { "epoch": 0.66, "learning_rate": 5.6066076693233376e-05, "loss": 1.3223, "step": 6322 }, { "epoch": 0.66, "learning_rate": 5.6064858099886067e-05, "loss": 0.6546, "step": 6323 }, { "epoch": 0.66, "learning_rate": 5.606363933107623e-05, "loss": 1.4945, "step": 6324 }, { "epoch": 0.66, "learning_rate": 5.606242038681208e-05, "loss": 0.8084, "step": 6325 }, { "epoch": 0.66, "learning_rate": 5.60612012671018e-05, "loss": 0.4586, "step": 6326 }, { "epoch": 0.66, "learning_rate": 5.605998197195362e-05, "loss": 0.6769, "step": 6327 }, { "epoch": 0.66, "learning_rate": 5.605876250137574e-05, "loss": 1.2327, "step": 6328 }, { "epoch": 0.66, "learning_rate": 5.605754285537636e-05, "loss": 0.5527, "step": 6329 }, { "epoch": 0.66, "learning_rate": 5.605632303396371e-05, "loss": 1.6074, "step": 6330 }, { "epoch": 0.66, "learning_rate": 5.605510303714598e-05, "loss": 0.9831, "step": 6331 }, { "epoch": 0.66, "learning_rate": 5.605388286493139e-05, "loss": 1.2822, "step": 6332 }, { "epoch": 0.66, "learning_rate": 5.605266251732817e-05, "loss": 1.3384, "step": 6333 }, { "epoch": 0.66, "learning_rate": 5.6051441994344515e-05, "loss": 0.5165, "step": 6334 }, { "epoch": 0.66, "learning_rate": 5.605022129598864e-05, "loss": 0.2112, "step": 6335 }, { "epoch": 0.66, "learning_rate": 5.6049000422268786e-05, "loss": 1.2564, "step": 6336 }, { "epoch": 0.66, "learning_rate": 5.604777937319314e-05, "loss": 1.2538, "step": 6337 }, { "epoch": 0.66, "learning_rate": 5.6046558148769944e-05, "loss": 1.5281, "step": 6338 }, { "epoch": 0.66, "learning_rate": 5.604533674900742e-05, "loss": 0.5516, "step": 6339 }, { "epoch": 0.66, "learning_rate": 5.6044115173913776e-05, "loss": 1.0311, "step": 6340 }, { "epoch": 0.66, "learning_rate": 5.604289342349725e-05, "loss": 0.6005, "step": 6341 }, { "epoch": 0.66, "learning_rate": 5.604167149776605e-05, "loss": 0.8519, "step": 6342 }, { "epoch": 0.66, "learning_rate": 5.604044939672841e-05, "loss": 0.3748, "step": 6343 }, { "epoch": 0.66, "learning_rate": 5.6039227120392565e-05, "loss": 0.3544, "step": 6344 }, { "epoch": 0.66, "learning_rate": 5.603800466876673e-05, "loss": 1.4785, "step": 6345 }, { "epoch": 0.66, "learning_rate": 5.6036782041859144e-05, "loss": 0.1681, "step": 6346 }, { "epoch": 0.66, "learning_rate": 5.603555923967804e-05, "loss": 0.2834, "step": 6347 }, { "epoch": 0.66, "learning_rate": 5.6034336262231634e-05, "loss": 1.5426, "step": 6348 }, { "epoch": 0.66, "learning_rate": 5.603311310952816e-05, "loss": 0.7017, "step": 6349 }, { "epoch": 0.66, "learning_rate": 5.603188978157587e-05, "loss": 0.7545, "step": 6350 }, { "epoch": 0.66, "learning_rate": 5.603066627838299e-05, "loss": 1.057, "step": 6351 }, { "epoch": 0.66, "learning_rate": 5.602944259995776e-05, "loss": 0.5785, "step": 6352 }, { "epoch": 0.66, "learning_rate": 5.60282187463084e-05, "loss": 1.603, "step": 6353 }, { "epoch": 0.66, "learning_rate": 5.602699471744316e-05, "loss": 1.4773, "step": 6354 }, { "epoch": 0.66, "learning_rate": 5.602577051337029e-05, "loss": 1.4485, "step": 6355 }, { "epoch": 0.66, "learning_rate": 5.6024546134098026e-05, "loss": 1.3708, "step": 6356 }, { "epoch": 0.66, "learning_rate": 5.60233215796346e-05, "loss": 0.9463, "step": 6357 }, { "epoch": 0.66, "learning_rate": 5.602209684998826e-05, "loss": 1.427, "step": 6358 }, { "epoch": 0.66, "learning_rate": 5.602087194516726e-05, "loss": 0.9516, "step": 6359 }, { "epoch": 0.66, "learning_rate": 5.6019646865179836e-05, "loss": 1.1644, "step": 6360 }, { "epoch": 0.66, "learning_rate": 5.601842161003423e-05, "loss": 0.4191, "step": 6361 }, { "epoch": 0.66, "learning_rate": 5.6017196179738716e-05, "loss": 1.0989, "step": 6362 }, { "epoch": 0.66, "learning_rate": 5.601597057430151e-05, "loss": 1.2784, "step": 6363 }, { "epoch": 0.66, "learning_rate": 5.6014744793730876e-05, "loss": 0.9709, "step": 6364 }, { "epoch": 0.66, "learning_rate": 5.6013518838035076e-05, "loss": 1.15, "step": 6365 }, { "epoch": 0.66, "learning_rate": 5.601229270722235e-05, "loss": 1.1813, "step": 6366 }, { "epoch": 0.66, "learning_rate": 5.601106640130095e-05, "loss": 1.1472, "step": 6367 }, { "epoch": 0.67, "learning_rate": 5.600983992027914e-05, "loss": 1.1422, "step": 6368 }, { "epoch": 0.67, "learning_rate": 5.6008613264165175e-05, "loss": 1.1252, "step": 6369 }, { "epoch": 0.67, "learning_rate": 5.600738643296731e-05, "loss": 0.848, "step": 6370 }, { "epoch": 0.67, "learning_rate": 5.600615942669381e-05, "loss": 0.1584, "step": 6371 }, { "epoch": 0.67, "learning_rate": 5.600493224535293e-05, "loss": 1.5357, "step": 6372 }, { "epoch": 0.67, "learning_rate": 5.6003704888952924e-05, "loss": 0.9185, "step": 6373 }, { "epoch": 0.67, "learning_rate": 5.600247735750206e-05, "loss": 0.489, "step": 6374 }, { "epoch": 0.67, "learning_rate": 5.600124965100861e-05, "loss": 0.253, "step": 6375 }, { "epoch": 0.67, "learning_rate": 5.600002176948083e-05, "loss": 0.4433, "step": 6376 }, { "epoch": 0.67, "learning_rate": 5.599879371292699e-05, "loss": 0.2548, "step": 6377 }, { "epoch": 0.67, "learning_rate": 5.5997565481355344e-05, "loss": 0.4571, "step": 6378 }, { "epoch": 0.67, "learning_rate": 5.599633707477418e-05, "loss": 0.9087, "step": 6379 }, { "epoch": 0.67, "learning_rate": 5.599510849319175e-05, "loss": 0.7203, "step": 6380 }, { "epoch": 0.67, "learning_rate": 5.599387973661632e-05, "loss": 0.8448, "step": 6381 }, { "epoch": 0.67, "learning_rate": 5.59926508050562e-05, "loss": 0.97, "step": 6382 }, { "epoch": 0.67, "learning_rate": 5.599142169851962e-05, "loss": 1.37, "step": 6383 }, { "epoch": 0.67, "learning_rate": 5.599019241701487e-05, "loss": 1.0824, "step": 6384 }, { "epoch": 0.67, "learning_rate": 5.5988962960550234e-05, "loss": 0.8385, "step": 6385 }, { "epoch": 0.67, "learning_rate": 5.5987733329133974e-05, "loss": 0.4264, "step": 6386 }, { "epoch": 0.67, "learning_rate": 5.598650352277439e-05, "loss": 0.4878, "step": 6387 }, { "epoch": 0.67, "learning_rate": 5.5985273541479723e-05, "loss": 0.5675, "step": 6388 }, { "epoch": 0.67, "learning_rate": 5.598404338525828e-05, "loss": 2.4869, "step": 6389 }, { "epoch": 0.67, "learning_rate": 5.598281305411834e-05, "loss": 1.3302, "step": 6390 }, { "epoch": 0.67, "learning_rate": 5.598158254806818e-05, "loss": 1.3418, "step": 6391 }, { "epoch": 0.67, "learning_rate": 5.598035186711609e-05, "loss": 1.2864, "step": 6392 }, { "epoch": 0.67, "learning_rate": 5.597912101127035e-05, "loss": 1.6089, "step": 6393 }, { "epoch": 0.67, "learning_rate": 5.5977889980539246e-05, "loss": 0.3274, "step": 6394 }, { "epoch": 0.67, "learning_rate": 5.5976658774931064e-05, "loss": 1.1161, "step": 6395 }, { "epoch": 0.67, "learning_rate": 5.5975427394454096e-05, "loss": 1.283, "step": 6396 }, { "epoch": 0.67, "learning_rate": 5.597419583911662e-05, "loss": 1.0793, "step": 6397 }, { "epoch": 0.67, "learning_rate": 5.5972964108926935e-05, "loss": 0.8691, "step": 6398 }, { "epoch": 0.67, "learning_rate": 5.5971732203893345e-05, "loss": 0.8276, "step": 6399 }, { "epoch": 0.67, "learning_rate": 5.597050012402412e-05, "loss": 1.176, "step": 6400 }, { "epoch": 0.67, "learning_rate": 5.596926786932756e-05, "loss": 1.1695, "step": 6401 }, { "epoch": 0.67, "learning_rate": 5.596803543981198e-05, "loss": 0.97, "step": 6402 }, { "epoch": 0.67, "learning_rate": 5.596680283548565e-05, "loss": 0.7084, "step": 6403 }, { "epoch": 0.67, "learning_rate": 5.5965570056356886e-05, "loss": 0.1954, "step": 6404 }, { "epoch": 0.67, "learning_rate": 5.5964337102433975e-05, "loss": 0.8837, "step": 6405 }, { "epoch": 0.67, "learning_rate": 5.596310397372523e-05, "loss": 1.3594, "step": 6406 }, { "epoch": 0.67, "learning_rate": 5.596187067023893e-05, "loss": 0.8146, "step": 6407 }, { "epoch": 0.67, "learning_rate": 5.5960637191983394e-05, "loss": 0.5932, "step": 6408 }, { "epoch": 0.67, "learning_rate": 5.5959403538966935e-05, "loss": 0.8716, "step": 6409 }, { "epoch": 0.67, "learning_rate": 5.5958169711197835e-05, "loss": 1.361, "step": 6410 }, { "epoch": 0.67, "learning_rate": 5.5956935708684415e-05, "loss": 0.6268, "step": 6411 }, { "epoch": 0.67, "learning_rate": 5.5955701531434976e-05, "loss": 0.3727, "step": 6412 }, { "epoch": 0.67, "learning_rate": 5.5954467179457825e-05, "loss": 1.4142, "step": 6413 }, { "epoch": 0.67, "learning_rate": 5.5953232652761285e-05, "loss": 0.7099, "step": 6414 }, { "epoch": 0.67, "learning_rate": 5.5951997951353634e-05, "loss": 0.4037, "step": 6415 }, { "epoch": 0.67, "learning_rate": 5.595076307524323e-05, "loss": 0.6311, "step": 6416 }, { "epoch": 0.67, "learning_rate": 5.5949528024438344e-05, "loss": 1.5139, "step": 6417 }, { "epoch": 0.67, "learning_rate": 5.5948292798947306e-05, "loss": 1.0621, "step": 6418 }, { "epoch": 0.67, "learning_rate": 5.594705739877844e-05, "loss": 0.734, "step": 6419 }, { "epoch": 0.67, "learning_rate": 5.594582182394005e-05, "loss": 0.5214, "step": 6420 }, { "epoch": 0.67, "learning_rate": 5.594458607444047e-05, "loss": 1.6753, "step": 6421 }, { "epoch": 0.67, "learning_rate": 5.594335015028799e-05, "loss": 1.0723, "step": 6422 }, { "epoch": 0.67, "learning_rate": 5.5942114051490956e-05, "loss": 0.666, "step": 6423 }, { "epoch": 0.67, "learning_rate": 5.5940877778057674e-05, "loss": 1.091, "step": 6424 }, { "epoch": 0.67, "learning_rate": 5.593964132999648e-05, "loss": 1.3749, "step": 6425 }, { "epoch": 0.67, "learning_rate": 5.5938404707315686e-05, "loss": 0.3747, "step": 6426 }, { "epoch": 0.67, "learning_rate": 5.593716791002362e-05, "loss": 1.8556, "step": 6427 }, { "epoch": 0.67, "learning_rate": 5.593593093812861e-05, "loss": 1.5014, "step": 6428 }, { "epoch": 0.67, "learning_rate": 5.593469379163899e-05, "loss": 0.6545, "step": 6429 }, { "epoch": 0.67, "learning_rate": 5.593345647056307e-05, "loss": 0.8357, "step": 6430 }, { "epoch": 0.67, "learning_rate": 5.593221897490919e-05, "loss": 1.6399, "step": 6431 }, { "epoch": 0.67, "learning_rate": 5.593098130468568e-05, "loss": 0.4264, "step": 6432 }, { "epoch": 0.67, "learning_rate": 5.5929743459900874e-05, "loss": 1.6251, "step": 6433 }, { "epoch": 0.67, "learning_rate": 5.592850544056309e-05, "loss": 1.248, "step": 6434 }, { "epoch": 0.67, "learning_rate": 5.5927267246680686e-05, "loss": 1.2832, "step": 6435 }, { "epoch": 0.67, "learning_rate": 5.592602887826199e-05, "loss": 1.3056, "step": 6436 }, { "epoch": 0.67, "learning_rate": 5.592479033531532e-05, "loss": 1.5641, "step": 6437 }, { "epoch": 0.67, "learning_rate": 5.592355161784904e-05, "loss": 1.2775, "step": 6438 }, { "epoch": 0.67, "learning_rate": 5.592231272587147e-05, "loss": 1.0071, "step": 6439 }, { "epoch": 0.67, "learning_rate": 5.592107365939095e-05, "loss": 0.7726, "step": 6440 }, { "epoch": 0.67, "learning_rate": 5.591983441841584e-05, "loss": 0.9503, "step": 6441 }, { "epoch": 0.67, "learning_rate": 5.591859500295447e-05, "loss": 0.8063, "step": 6442 }, { "epoch": 0.67, "learning_rate": 5.591735541301517e-05, "loss": 0.678, "step": 6443 }, { "epoch": 0.67, "learning_rate": 5.5916115648606303e-05, "loss": 0.2325, "step": 6444 }, { "epoch": 0.67, "learning_rate": 5.5914875709736205e-05, "loss": 1.1324, "step": 6445 }, { "epoch": 0.67, "learning_rate": 5.5913635596413234e-05, "loss": 0.4137, "step": 6446 }, { "epoch": 0.67, "learning_rate": 5.591239530864573e-05, "loss": 1.199, "step": 6447 }, { "epoch": 0.67, "learning_rate": 5.5911154846442045e-05, "loss": 0.5357, "step": 6448 }, { "epoch": 0.67, "learning_rate": 5.5909914209810525e-05, "loss": 1.665, "step": 6449 }, { "epoch": 0.67, "learning_rate": 5.5908673398759525e-05, "loss": 1.4002, "step": 6450 }, { "epoch": 0.67, "learning_rate": 5.59074324132974e-05, "loss": 0.9352, "step": 6451 }, { "epoch": 0.67, "learning_rate": 5.59061912534325e-05, "loss": 0.9599, "step": 6452 }, { "epoch": 0.67, "learning_rate": 5.5904949919173186e-05, "loss": 1.4137, "step": 6453 }, { "epoch": 0.67, "learning_rate": 5.5903708410527806e-05, "loss": 1.2089, "step": 6454 }, { "epoch": 0.67, "learning_rate": 5.590246672750472e-05, "loss": 0.4093, "step": 6455 }, { "epoch": 0.67, "learning_rate": 5.5901224870112296e-05, "loss": 0.5586, "step": 6456 }, { "epoch": 0.67, "learning_rate": 5.5899982838358884e-05, "loss": 1.9534, "step": 6457 }, { "epoch": 0.67, "learning_rate": 5.5898740632252846e-05, "loss": 1.1122, "step": 6458 }, { "epoch": 0.67, "learning_rate": 5.5897498251802545e-05, "loss": 0.3548, "step": 6459 }, { "epoch": 0.67, "learning_rate": 5.589625569701635e-05, "loss": 0.9397, "step": 6460 }, { "epoch": 0.67, "learning_rate": 5.589501296790261e-05, "loss": 0.8877, "step": 6461 }, { "epoch": 0.67, "learning_rate": 5.589377006446972e-05, "loss": 1.16, "step": 6462 }, { "epoch": 0.67, "learning_rate": 5.589252698672602e-05, "loss": 0.9629, "step": 6463 }, { "epoch": 0.68, "learning_rate": 5.589128373467988e-05, "loss": 0.6801, "step": 6464 }, { "epoch": 0.68, "learning_rate": 5.589004030833968e-05, "loss": 0.8534, "step": 6465 }, { "epoch": 0.68, "learning_rate": 5.588879670771379e-05, "loss": 0.3483, "step": 6466 }, { "epoch": 0.68, "learning_rate": 5.588755293281057e-05, "loss": 0.9808, "step": 6467 }, { "epoch": 0.68, "learning_rate": 5.5886308983638405e-05, "loss": 1.0926, "step": 6468 }, { "epoch": 0.68, "learning_rate": 5.588506486020567e-05, "loss": 0.0763, "step": 6469 }, { "epoch": 0.68, "learning_rate": 5.588382056252073e-05, "loss": 0.7443, "step": 6470 }, { "epoch": 0.68, "learning_rate": 5.588257609059196e-05, "loss": 1.1835, "step": 6471 }, { "epoch": 0.68, "learning_rate": 5.588133144442775e-05, "loss": 0.6598, "step": 6472 }, { "epoch": 0.68, "learning_rate": 5.5880086624036476e-05, "loss": 0.4309, "step": 6473 }, { "epoch": 0.68, "learning_rate": 5.5878841629426514e-05, "loss": 1.3757, "step": 6474 }, { "epoch": 0.68, "learning_rate": 5.587759646060625e-05, "loss": 0.6619, "step": 6475 }, { "epoch": 0.68, "learning_rate": 5.587635111758405e-05, "loss": 1.0242, "step": 6476 }, { "epoch": 0.68, "learning_rate": 5.587510560036831e-05, "loss": 0.423, "step": 6477 }, { "epoch": 0.68, "learning_rate": 5.5873859908967415e-05, "loss": 0.6765, "step": 6478 }, { "epoch": 0.68, "learning_rate": 5.587261404338975e-05, "loss": 1.1274, "step": 6479 }, { "epoch": 0.68, "learning_rate": 5.58713680036437e-05, "loss": 0.9347, "step": 6480 }, { "epoch": 0.68, "learning_rate": 5.5870121789737656e-05, "loss": 0.8372, "step": 6481 }, { "epoch": 0.68, "learning_rate": 5.5868875401680006e-05, "loss": 0.7838, "step": 6482 }, { "epoch": 0.68, "learning_rate": 5.586762883947914e-05, "loss": 1.0191, "step": 6483 }, { "epoch": 0.68, "learning_rate": 5.586638210314345e-05, "loss": 0.6734, "step": 6484 }, { "epoch": 0.68, "learning_rate": 5.5865135192681324e-05, "loss": 0.9413, "step": 6485 }, { "epoch": 0.68, "learning_rate": 5.586388810810116e-05, "loss": 1.5458, "step": 6486 }, { "epoch": 0.68, "learning_rate": 5.5862640849411354e-05, "loss": 1.3622, "step": 6487 }, { "epoch": 0.68, "learning_rate": 5.5861393416620305e-05, "loss": 1.2969, "step": 6488 }, { "epoch": 0.68, "learning_rate": 5.58601458097364e-05, "loss": 1.1958, "step": 6489 }, { "epoch": 0.68, "learning_rate": 5.585889802876805e-05, "loss": 1.8371, "step": 6490 }, { "epoch": 0.68, "learning_rate": 5.585765007372365e-05, "loss": 1.3645, "step": 6491 }, { "epoch": 0.68, "learning_rate": 5.5856401944611596e-05, "loss": 0.5814, "step": 6492 }, { "epoch": 0.68, "learning_rate": 5.585515364144029e-05, "loss": 0.4114, "step": 6493 }, { "epoch": 0.68, "learning_rate": 5.585390516421815e-05, "loss": 1.5228, "step": 6494 }, { "epoch": 0.68, "learning_rate": 5.585265651295356e-05, "loss": 0.3628, "step": 6495 }, { "epoch": 0.68, "learning_rate": 5.585140768765494e-05, "loss": 0.5569, "step": 6496 }, { "epoch": 0.68, "learning_rate": 5.585015868833069e-05, "loss": 1.1617, "step": 6497 }, { "epoch": 0.68, "learning_rate": 5.584890951498922e-05, "loss": 0.7368, "step": 6498 }, { "epoch": 0.68, "learning_rate": 5.584766016763894e-05, "loss": 1.4822, "step": 6499 }, { "epoch": 0.68, "learning_rate": 5.584641064628826e-05, "loss": 0.5602, "step": 6500 }, { "epoch": 0.68, "learning_rate": 5.584516095094559e-05, "loss": 0.5458, "step": 6501 }, { "epoch": 0.68, "learning_rate": 5.584391108161935e-05, "loss": 0.467, "step": 6502 }, { "epoch": 0.68, "learning_rate": 5.584266103831794e-05, "loss": 1.3781, "step": 6503 }, { "epoch": 0.68, "learning_rate": 5.584141082104978e-05, "loss": 0.7834, "step": 6504 }, { "epoch": 0.68, "learning_rate": 5.5840160429823296e-05, "loss": 0.3758, "step": 6505 }, { "epoch": 0.68, "learning_rate": 5.58389098646469e-05, "loss": 1.2403, "step": 6506 }, { "epoch": 0.68, "learning_rate": 5.5837659125529006e-05, "loss": 0.4589, "step": 6507 }, { "epoch": 0.68, "learning_rate": 5.5836408212478027e-05, "loss": 1.0634, "step": 6508 }, { "epoch": 0.68, "learning_rate": 5.583515712550241e-05, "loss": 0.9109, "step": 6509 }, { "epoch": 0.68, "learning_rate": 5.583390586461055e-05, "loss": 0.9422, "step": 6510 }, { "epoch": 0.68, "learning_rate": 5.583265442981088e-05, "loss": 0.2841, "step": 6511 }, { "epoch": 0.68, "learning_rate": 5.583140282111183e-05, "loss": 1.3113, "step": 6512 }, { "epoch": 0.68, "learning_rate": 5.583015103852182e-05, "loss": 0.3733, "step": 6513 }, { "epoch": 0.68, "learning_rate": 5.582889908204926e-05, "loss": 1.1464, "step": 6514 }, { "epoch": 0.68, "learning_rate": 5.582764695170261e-05, "loss": 1.0359, "step": 6515 }, { "epoch": 0.68, "learning_rate": 5.582639464749028e-05, "loss": 1.1523, "step": 6516 }, { "epoch": 0.68, "learning_rate": 5.5825142169420715e-05, "loss": 0.8962, "step": 6517 }, { "epoch": 0.68, "learning_rate": 5.5823889517502324e-05, "loss": 0.3544, "step": 6518 }, { "epoch": 0.68, "learning_rate": 5.582263669174355e-05, "loss": 0.7677, "step": 6519 }, { "epoch": 0.68, "learning_rate": 5.582138369215284e-05, "loss": 0.6674, "step": 6520 }, { "epoch": 0.68, "learning_rate": 5.58201305187386e-05, "loss": 1.3408, "step": 6521 }, { "epoch": 0.68, "learning_rate": 5.58188771715093e-05, "loss": 0.6215, "step": 6522 }, { "epoch": 0.68, "learning_rate": 5.581762365047334e-05, "loss": 1.6574, "step": 6523 }, { "epoch": 0.68, "learning_rate": 5.58163699556392e-05, "loss": 1.0791, "step": 6524 }, { "epoch": 0.68, "learning_rate": 5.5815116087015285e-05, "loss": 1.5015, "step": 6525 }, { "epoch": 0.68, "learning_rate": 5.5813862044610054e-05, "loss": 1.1299, "step": 6526 }, { "epoch": 0.68, "learning_rate": 5.581260782843194e-05, "loss": 1.1176, "step": 6527 }, { "epoch": 0.68, "learning_rate": 5.581135343848939e-05, "loss": 0.4022, "step": 6528 }, { "epoch": 0.68, "learning_rate": 5.581009887479085e-05, "loss": 0.8988, "step": 6529 }, { "epoch": 0.68, "learning_rate": 5.580884413734476e-05, "loss": 1.617, "step": 6530 }, { "epoch": 0.68, "learning_rate": 5.580758922615957e-05, "loss": 0.3102, "step": 6531 }, { "epoch": 0.68, "learning_rate": 5.580633414124374e-05, "loss": 1.4044, "step": 6532 }, { "epoch": 0.68, "learning_rate": 5.5805078882605694e-05, "loss": 0.6467, "step": 6533 }, { "epoch": 0.68, "learning_rate": 5.5803823450253894e-05, "loss": 1.3861, "step": 6534 }, { "epoch": 0.68, "learning_rate": 5.580256784419679e-05, "loss": 1.0177, "step": 6535 }, { "epoch": 0.68, "learning_rate": 5.580131206444285e-05, "loss": 0.5906, "step": 6536 }, { "epoch": 0.68, "learning_rate": 5.580005611100051e-05, "loss": 1.1374, "step": 6537 }, { "epoch": 0.68, "learning_rate": 5.5798799983878224e-05, "loss": 1.0225, "step": 6538 }, { "epoch": 0.68, "learning_rate": 5.579754368308445e-05, "loss": 0.7268, "step": 6539 }, { "epoch": 0.68, "learning_rate": 5.579628720862766e-05, "loss": 1.4708, "step": 6540 }, { "epoch": 0.68, "learning_rate": 5.579503056051629e-05, "loss": 1.3361, "step": 6541 }, { "epoch": 0.68, "learning_rate": 5.579377373875881e-05, "loss": 0.639, "step": 6542 }, { "epoch": 0.68, "learning_rate": 5.5792516743363686e-05, "loss": 0.8404, "step": 6543 }, { "epoch": 0.68, "learning_rate": 5.5791259574339374e-05, "loss": 1.0701, "step": 6544 }, { "epoch": 0.68, "learning_rate": 5.5790002231694335e-05, "loss": 0.6978, "step": 6545 }, { "epoch": 0.68, "learning_rate": 5.578874471543704e-05, "loss": 1.3862, "step": 6546 }, { "epoch": 0.68, "learning_rate": 5.578748702557594e-05, "loss": 1.2904, "step": 6547 }, { "epoch": 0.68, "learning_rate": 5.578622916211952e-05, "loss": 0.836, "step": 6548 }, { "epoch": 0.68, "learning_rate": 5.5784971125076235e-05, "loss": 0.9245, "step": 6549 }, { "epoch": 0.68, "learning_rate": 5.578371291445456e-05, "loss": 1.3269, "step": 6550 }, { "epoch": 0.68, "learning_rate": 5.578245453026295e-05, "loss": 0.5034, "step": 6551 }, { "epoch": 0.68, "learning_rate": 5.578119597250991e-05, "loss": 0.5076, "step": 6552 }, { "epoch": 0.68, "learning_rate": 5.577993724120387e-05, "loss": 0.9881, "step": 6553 }, { "epoch": 0.68, "learning_rate": 5.577867833635334e-05, "loss": 0.3427, "step": 6554 }, { "epoch": 0.68, "learning_rate": 5.5777419257966775e-05, "loss": 1.0078, "step": 6555 }, { "epoch": 0.68, "learning_rate": 5.577616000605265e-05, "loss": 0.697, "step": 6556 }, { "epoch": 0.68, "learning_rate": 5.577490058061945e-05, "loss": 1.0336, "step": 6557 }, { "epoch": 0.68, "learning_rate": 5.577364098167565e-05, "loss": 0.4952, "step": 6558 }, { "epoch": 0.69, "learning_rate": 5.577238120922973e-05, "loss": 1.4554, "step": 6559 }, { "epoch": 0.69, "learning_rate": 5.577112126329018e-05, "loss": 0.37, "step": 6560 }, { "epoch": 0.69, "learning_rate": 5.576986114386545e-05, "loss": 0.5365, "step": 6561 }, { "epoch": 0.69, "learning_rate": 5.576860085096406e-05, "loss": 0.6044, "step": 6562 }, { "epoch": 0.69, "learning_rate": 5.5767340384594476e-05, "loss": 1.0521, "step": 6563 }, { "epoch": 0.69, "learning_rate": 5.576607974476518e-05, "loss": 0.6636, "step": 6564 }, { "epoch": 0.69, "learning_rate": 5.576481893148467e-05, "loss": 0.8058, "step": 6565 }, { "epoch": 0.69, "learning_rate": 5.5763557944761425e-05, "loss": 0.7948, "step": 6566 }, { "epoch": 0.69, "learning_rate": 5.576229678460394e-05, "loss": 0.4508, "step": 6567 }, { "epoch": 0.69, "learning_rate": 5.57610354510207e-05, "loss": 0.7446, "step": 6568 }, { "epoch": 0.69, "learning_rate": 5.5759773944020185e-05, "loss": 0.2944, "step": 6569 }, { "epoch": 0.69, "learning_rate": 5.575851226361092e-05, "loss": 0.8299, "step": 6570 }, { "epoch": 0.69, "learning_rate": 5.5757250409801356e-05, "loss": 0.2761, "step": 6571 }, { "epoch": 0.69, "learning_rate": 5.5755988382600025e-05, "loss": 1.6246, "step": 6572 }, { "epoch": 0.69, "learning_rate": 5.57547261820154e-05, "loss": 0.6689, "step": 6573 }, { "epoch": 0.69, "learning_rate": 5.575346380805599e-05, "loss": 0.6156, "step": 6574 }, { "epoch": 0.69, "learning_rate": 5.575220126073028e-05, "loss": 0.9078, "step": 6575 }, { "epoch": 0.69, "learning_rate": 5.575093854004678e-05, "loss": 0.4868, "step": 6576 }, { "epoch": 0.69, "learning_rate": 5.574967564601399e-05, "loss": 0.8835, "step": 6577 }, { "epoch": 0.69, "learning_rate": 5.57484125786404e-05, "loss": 1.0993, "step": 6578 }, { "epoch": 0.69, "learning_rate": 5.574714933793452e-05, "loss": 0.9234, "step": 6579 }, { "epoch": 0.69, "learning_rate": 5.574588592390486e-05, "loss": 0.4301, "step": 6580 }, { "epoch": 0.69, "learning_rate": 5.574462233655993e-05, "loss": 0.3809, "step": 6581 }, { "epoch": 0.69, "learning_rate": 5.5743358575908215e-05, "loss": 0.8743, "step": 6582 }, { "epoch": 0.69, "learning_rate": 5.5742094641958234e-05, "loss": 0.8922, "step": 6583 }, { "epoch": 0.69, "learning_rate": 5.57408305347185e-05, "loss": 1.0285, "step": 6584 }, { "epoch": 0.69, "learning_rate": 5.573956625419752e-05, "loss": 1.446, "step": 6585 }, { "epoch": 0.69, "learning_rate": 5.5738301800403794e-05, "loss": 0.9999, "step": 6586 }, { "epoch": 0.69, "learning_rate": 5.573703717334584e-05, "loss": 1.5071, "step": 6587 }, { "epoch": 0.69, "learning_rate": 5.5735772373032184e-05, "loss": 0.4334, "step": 6588 }, { "epoch": 0.69, "learning_rate": 5.573450739947132e-05, "loss": 0.5014, "step": 6589 }, { "epoch": 0.69, "learning_rate": 5.573324225267178e-05, "loss": 1.9156, "step": 6590 }, { "epoch": 0.69, "learning_rate": 5.5731976932642075e-05, "loss": 1.2268, "step": 6591 }, { "epoch": 0.69, "learning_rate": 5.573071143939072e-05, "loss": 0.5898, "step": 6592 }, { "epoch": 0.69, "learning_rate": 5.5729445772926236e-05, "loss": 1.1068, "step": 6593 }, { "epoch": 0.69, "learning_rate": 5.572817993325714e-05, "loss": 0.5068, "step": 6594 }, { "epoch": 0.69, "learning_rate": 5.572691392039196e-05, "loss": 1.1103, "step": 6595 }, { "epoch": 0.69, "learning_rate": 5.572564773433922e-05, "loss": 1.6844, "step": 6596 }, { "epoch": 0.69, "learning_rate": 5.5724381375107435e-05, "loss": 0.9816, "step": 6597 }, { "epoch": 0.69, "learning_rate": 5.572311484270513e-05, "loss": 2.1115, "step": 6598 }, { "epoch": 0.69, "learning_rate": 5.572184813714084e-05, "loss": 1.4159, "step": 6599 }, { "epoch": 0.69, "learning_rate": 5.572058125842308e-05, "loss": 0.8157, "step": 6600 }, { "epoch": 0.69, "learning_rate": 5.571931420656039e-05, "loss": 0.5154, "step": 6601 }, { "epoch": 0.69, "learning_rate": 5.571804698156129e-05, "loss": 1.755, "step": 6602 }, { "epoch": 0.69, "learning_rate": 5.571677958343432e-05, "loss": 0.7023, "step": 6603 }, { "epoch": 0.69, "learning_rate": 5.571551201218801e-05, "loss": 1.2111, "step": 6604 }, { "epoch": 0.69, "learning_rate": 5.571424426783089e-05, "loss": 0.969, "step": 6605 }, { "epoch": 0.69, "learning_rate": 5.571297635037149e-05, "loss": 0.8171, "step": 6606 }, { "epoch": 0.69, "learning_rate": 5.571170825981835e-05, "loss": 0.862, "step": 6607 }, { "epoch": 0.69, "learning_rate": 5.571043999618001e-05, "loss": 0.8397, "step": 6608 }, { "epoch": 0.69, "learning_rate": 5.5709171559464997e-05, "loss": 0.8138, "step": 6609 }, { "epoch": 0.69, "learning_rate": 5.5707902949681856e-05, "loss": 0.5561, "step": 6610 }, { "epoch": 0.69, "learning_rate": 5.570663416683913e-05, "loss": 0.1569, "step": 6611 }, { "epoch": 0.69, "learning_rate": 5.570536521094536e-05, "loss": 1.1475, "step": 6612 }, { "epoch": 0.69, "learning_rate": 5.570409608200908e-05, "loss": 1.3976, "step": 6613 }, { "epoch": 0.69, "learning_rate": 5.570282678003885e-05, "loss": 0.2596, "step": 6614 }, { "epoch": 0.69, "learning_rate": 5.570155730504319e-05, "loss": 0.9896, "step": 6615 }, { "epoch": 0.69, "learning_rate": 5.570028765703067e-05, "loss": 0.9382, "step": 6616 }, { "epoch": 0.69, "learning_rate": 5.569901783600983e-05, "loss": 0.753, "step": 6617 }, { "epoch": 0.69, "learning_rate": 5.569774784198921e-05, "loss": 0.7153, "step": 6618 }, { "epoch": 0.69, "learning_rate": 5.569647767497736e-05, "loss": 0.5925, "step": 6619 }, { "epoch": 0.69, "learning_rate": 5.5695207334982836e-05, "loss": 0.9748, "step": 6620 }, { "epoch": 0.69, "learning_rate": 5.56939368220142e-05, "loss": 1.2951, "step": 6621 }, { "epoch": 0.69, "learning_rate": 5.5692666136079987e-05, "loss": 0.3588, "step": 6622 }, { "epoch": 0.69, "learning_rate": 5.5691395277188753e-05, "loss": 0.467, "step": 6623 }, { "epoch": 0.69, "learning_rate": 5.569012424534906e-05, "loss": 1.8132, "step": 6624 }, { "epoch": 0.69, "learning_rate": 5.5688853040569456e-05, "loss": 0.8475, "step": 6625 }, { "epoch": 0.69, "learning_rate": 5.568758166285851e-05, "loss": 1.1647, "step": 6626 }, { "epoch": 0.69, "learning_rate": 5.5686310112224774e-05, "loss": 0.9239, "step": 6627 }, { "epoch": 0.69, "learning_rate": 5.5685038388676806e-05, "loss": 0.8868, "step": 6628 }, { "epoch": 0.69, "learning_rate": 5.5683766492223176e-05, "loss": 0.4094, "step": 6629 }, { "epoch": 0.69, "learning_rate": 5.568249442287243e-05, "loss": 1.4328, "step": 6630 }, { "epoch": 0.69, "learning_rate": 5.568122218063315e-05, "loss": 1.5318, "step": 6631 }, { "epoch": 0.69, "learning_rate": 5.567994976551388e-05, "loss": 1.2574, "step": 6632 }, { "epoch": 0.69, "learning_rate": 5.567867717752321e-05, "loss": 0.945, "step": 6633 }, { "epoch": 0.69, "learning_rate": 5.5677404416669684e-05, "loss": 0.2782, "step": 6634 }, { "epoch": 0.69, "learning_rate": 5.567613148296188e-05, "loss": 0.4425, "step": 6635 }, { "epoch": 0.69, "learning_rate": 5.567485837640838e-05, "loss": 1.1893, "step": 6636 }, { "epoch": 0.69, "learning_rate": 5.567358509701772e-05, "loss": 1.4271, "step": 6637 }, { "epoch": 0.69, "learning_rate": 5.56723116447985e-05, "loss": 0.3177, "step": 6638 }, { "epoch": 0.69, "learning_rate": 5.567103801975929e-05, "loss": 0.7385, "step": 6639 }, { "epoch": 0.69, "learning_rate": 5.5669764221908654e-05, "loss": 0.8819, "step": 6640 }, { "epoch": 0.69, "learning_rate": 5.5668490251255176e-05, "loss": 0.9173, "step": 6641 }, { "epoch": 0.69, "learning_rate": 5.566721610780743e-05, "loss": 0.5407, "step": 6642 }, { "epoch": 0.69, "learning_rate": 5.566594179157398e-05, "loss": 1.2265, "step": 6643 }, { "epoch": 0.69, "learning_rate": 5.566466730256342e-05, "loss": 1.0954, "step": 6644 }, { "epoch": 0.69, "learning_rate": 5.566339264078433e-05, "loss": 0.2185, "step": 6645 }, { "epoch": 0.69, "learning_rate": 5.566211780624528e-05, "loss": 1.2876, "step": 6646 }, { "epoch": 0.69, "learning_rate": 5.566084279895486e-05, "loss": 1.2038, "step": 6647 }, { "epoch": 0.69, "learning_rate": 5.565956761892164e-05, "loss": 0.7536, "step": 6648 }, { "epoch": 0.69, "learning_rate": 5.5658292266154225e-05, "loss": 0.3284, "step": 6649 }, { "epoch": 0.69, "learning_rate": 5.56570167406612e-05, "loss": 0.6397, "step": 6650 }, { "epoch": 0.69, "learning_rate": 5.5655741042451125e-05, "loss": 0.8512, "step": 6651 }, { "epoch": 0.69, "learning_rate": 5.565446517153261e-05, "loss": 0.6219, "step": 6652 }, { "epoch": 0.69, "learning_rate": 5.565318912791423e-05, "loss": 0.3717, "step": 6653 }, { "epoch": 0.69, "learning_rate": 5.56519129116046e-05, "loss": 1.3147, "step": 6654 }, { "epoch": 0.7, "learning_rate": 5.5650636522612285e-05, "loss": 1.4807, "step": 6655 }, { "epoch": 0.7, "learning_rate": 5.5649359960945886e-05, "loss": 1.5754, "step": 6656 }, { "epoch": 0.7, "learning_rate": 5.5648083226613996e-05, "loss": 1.4764, "step": 6657 }, { "epoch": 0.7, "learning_rate": 5.564680631962521e-05, "loss": 1.1148, "step": 6658 }, { "epoch": 0.7, "learning_rate": 5.564552923998813e-05, "loss": 1.0298, "step": 6659 }, { "epoch": 0.7, "learning_rate": 5.564425198771135e-05, "loss": 1.2502, "step": 6660 }, { "epoch": 0.7, "learning_rate": 5.564297456280346e-05, "loss": 1.121, "step": 6661 }, { "epoch": 0.7, "learning_rate": 5.5641696965273063e-05, "loss": 0.855, "step": 6662 }, { "epoch": 0.7, "learning_rate": 5.5640419195128766e-05, "loss": 1.524, "step": 6663 }, { "epoch": 0.7, "learning_rate": 5.5639141252379165e-05, "loss": 0.8914, "step": 6664 }, { "epoch": 0.7, "learning_rate": 5.5637863137032865e-05, "loss": 0.6744, "step": 6665 }, { "epoch": 0.7, "learning_rate": 5.563658484909847e-05, "loss": 0.9028, "step": 6666 }, { "epoch": 0.7, "learning_rate": 5.563530638858458e-05, "loss": 2.2177, "step": 6667 }, { "epoch": 0.7, "learning_rate": 5.5634027755499806e-05, "loss": 0.9994, "step": 6668 }, { "epoch": 0.7, "learning_rate": 5.5632748949852764e-05, "loss": 1.3547, "step": 6669 }, { "epoch": 0.7, "learning_rate": 5.5631469971652046e-05, "loss": 1.3861, "step": 6670 }, { "epoch": 0.7, "learning_rate": 5.5630190820906264e-05, "loss": 0.665, "step": 6671 }, { "epoch": 0.7, "learning_rate": 5.5628911497624046e-05, "loss": 0.5237, "step": 6672 }, { "epoch": 0.7, "learning_rate": 5.562763200181399e-05, "loss": 0.7372, "step": 6673 }, { "epoch": 0.7, "learning_rate": 5.5626352333484713e-05, "loss": 0.6515, "step": 6674 }, { "epoch": 0.7, "learning_rate": 5.562507249264482e-05, "loss": 1.0411, "step": 6675 }, { "epoch": 0.7, "learning_rate": 5.562379247930294e-05, "loss": 0.6464, "step": 6676 }, { "epoch": 0.7, "learning_rate": 5.562251229346768e-05, "loss": 0.4784, "step": 6677 }, { "epoch": 0.7, "learning_rate": 5.562123193514767e-05, "loss": 0.3689, "step": 6678 }, { "epoch": 0.7, "learning_rate": 5.561995140435151e-05, "loss": 0.472, "step": 6679 }, { "epoch": 0.7, "learning_rate": 5.5618670701087844e-05, "loss": 1.5843, "step": 6680 }, { "epoch": 0.7, "learning_rate": 5.561738982536528e-05, "loss": 0.5976, "step": 6681 }, { "epoch": 0.7, "learning_rate": 5.561610877719244e-05, "loss": 1.3386, "step": 6682 }, { "epoch": 0.7, "learning_rate": 5.561482755657795e-05, "loss": 1.0458, "step": 6683 }, { "epoch": 0.7, "learning_rate": 5.561354616353043e-05, "loss": 0.4114, "step": 6684 }, { "epoch": 0.7, "learning_rate": 5.5612264598058506e-05, "loss": 0.4216, "step": 6685 }, { "epoch": 0.7, "learning_rate": 5.561098286017082e-05, "loss": 0.9789, "step": 6686 }, { "epoch": 0.7, "learning_rate": 5.560970094987598e-05, "loss": 0.6138, "step": 6687 }, { "epoch": 0.7, "learning_rate": 5.560841886718264e-05, "loss": 1.5581, "step": 6688 }, { "epoch": 0.7, "learning_rate": 5.5607136612099405e-05, "loss": 0.8135, "step": 6689 }, { "epoch": 0.7, "learning_rate": 5.560585418463492e-05, "loss": 0.6398, "step": 6690 }, { "epoch": 0.7, "learning_rate": 5.560457158479782e-05, "loss": 1.0962, "step": 6691 }, { "epoch": 0.7, "learning_rate": 5.560328881259673e-05, "loss": 0.8193, "step": 6692 }, { "epoch": 0.7, "learning_rate": 5.560200586804029e-05, "loss": 1.4208, "step": 6693 }, { "epoch": 0.7, "learning_rate": 5.560072275113713e-05, "loss": 0.7966, "step": 6694 }, { "epoch": 0.7, "learning_rate": 5.559943946189591e-05, "loss": 0.4365, "step": 6695 }, { "epoch": 0.7, "learning_rate": 5.5598156000325246e-05, "loss": 1.1245, "step": 6696 }, { "epoch": 0.7, "learning_rate": 5.559687236643379e-05, "loss": 0.2062, "step": 6697 }, { "epoch": 0.7, "learning_rate": 5.559558856023017e-05, "loss": 1.3996, "step": 6698 }, { "epoch": 0.7, "learning_rate": 5.559430458172304e-05, "loss": 1.1246, "step": 6699 }, { "epoch": 0.7, "learning_rate": 5.559302043092104e-05, "loss": 1.596, "step": 6700 }, { "epoch": 0.7, "learning_rate": 5.5591736107832814e-05, "loss": 1.0894, "step": 6701 }, { "epoch": 0.7, "learning_rate": 5.5590451612467015e-05, "loss": 1.2756, "step": 6702 }, { "epoch": 0.7, "learning_rate": 5.558916694483228e-05, "loss": 1.4973, "step": 6703 }, { "epoch": 0.7, "learning_rate": 5.558788210493725e-05, "loss": 1.0407, "step": 6704 }, { "epoch": 0.7, "learning_rate": 5.55865970927906e-05, "loss": 1.4526, "step": 6705 }, { "epoch": 0.7, "learning_rate": 5.558531190840096e-05, "loss": 0.7222, "step": 6706 }, { "epoch": 0.7, "learning_rate": 5.558402655177699e-05, "loss": 1.6511, "step": 6707 }, { "epoch": 0.7, "learning_rate": 5.558274102292733e-05, "loss": 0.7862, "step": 6708 }, { "epoch": 0.7, "learning_rate": 5.558145532186065e-05, "loss": 1.0471, "step": 6709 }, { "epoch": 0.7, "learning_rate": 5.5580169448585595e-05, "loss": 0.3118, "step": 6710 }, { "epoch": 0.7, "learning_rate": 5.5578883403110824e-05, "loss": 0.9757, "step": 6711 }, { "epoch": 0.7, "learning_rate": 5.5577597185445e-05, "loss": 1.0905, "step": 6712 }, { "epoch": 0.7, "learning_rate": 5.5576310795596776e-05, "loss": 0.5008, "step": 6713 }, { "epoch": 0.7, "learning_rate": 5.557502423357481e-05, "loss": 1.786, "step": 6714 }, { "epoch": 0.7, "learning_rate": 5.557373749938776e-05, "loss": 0.8669, "step": 6715 }, { "epoch": 0.7, "learning_rate": 5.55724505930443e-05, "loss": 1.2039, "step": 6716 }, { "epoch": 0.7, "learning_rate": 5.557116351455309e-05, "loss": 0.4668, "step": 6717 }, { "epoch": 0.7, "learning_rate": 5.556987626392278e-05, "loss": 0.6931, "step": 6718 }, { "epoch": 0.7, "learning_rate": 5.556858884116205e-05, "loss": 0.7924, "step": 6719 }, { "epoch": 0.7, "learning_rate": 5.556730124627957e-05, "loss": 0.4976, "step": 6720 }, { "epoch": 0.7, "learning_rate": 5.5566013479283976e-05, "loss": 1.0348, "step": 6721 }, { "epoch": 0.7, "learning_rate": 5.556472554018398e-05, "loss": 1.3309, "step": 6722 }, { "epoch": 0.7, "learning_rate": 5.556343742898824e-05, "loss": 1.3544, "step": 6723 }, { "epoch": 0.7, "learning_rate": 5.556214914570541e-05, "loss": 1.3185, "step": 6724 }, { "epoch": 0.7, "learning_rate": 5.5560860690344174e-05, "loss": 1.01, "step": 6725 }, { "epoch": 0.7, "learning_rate": 5.55595720629132e-05, "loss": 0.812, "step": 6726 }, { "epoch": 0.7, "learning_rate": 5.5558283263421175e-05, "loss": 0.7414, "step": 6727 }, { "epoch": 0.7, "learning_rate": 5.5556994291876766e-05, "loss": 0.2474, "step": 6728 }, { "epoch": 0.7, "learning_rate": 5.555570514828865e-05, "loss": 0.9426, "step": 6729 }, { "epoch": 0.7, "learning_rate": 5.555441583266551e-05, "loss": 1.5623, "step": 6730 }, { "epoch": 0.7, "learning_rate": 5.5553126345016014e-05, "loss": 1.6343, "step": 6731 }, { "epoch": 0.7, "learning_rate": 5.555183668534886e-05, "loss": 0.8865, "step": 6732 }, { "epoch": 0.7, "learning_rate": 5.555054685367272e-05, "loss": 0.9677, "step": 6733 }, { "epoch": 0.7, "learning_rate": 5.554925684999626e-05, "loss": 0.4473, "step": 6734 }, { "epoch": 0.7, "learning_rate": 5.5547966674328206e-05, "loss": 1.434, "step": 6735 }, { "epoch": 0.7, "learning_rate": 5.5546676326677205e-05, "loss": 0.5546, "step": 6736 }, { "epoch": 0.7, "learning_rate": 5.554538580705195e-05, "loss": 1.2085, "step": 6737 }, { "epoch": 0.7, "learning_rate": 5.5544095115461143e-05, "loss": 1.2563, "step": 6738 }, { "epoch": 0.7, "learning_rate": 5.5542804251913464e-05, "loss": 0.1866, "step": 6739 }, { "epoch": 0.7, "learning_rate": 5.55415132164176e-05, "loss": 1.1769, "step": 6740 }, { "epoch": 0.7, "learning_rate": 5.554022200898225e-05, "loss": 0.6861, "step": 6741 }, { "epoch": 0.7, "learning_rate": 5.55389306296161e-05, "loss": 1.0528, "step": 6742 }, { "epoch": 0.7, "learning_rate": 5.5537639078327835e-05, "loss": 0.3956, "step": 6743 }, { "epoch": 0.7, "learning_rate": 5.553634735512617e-05, "loss": 0.9852, "step": 6744 }, { "epoch": 0.7, "learning_rate": 5.55350554600198e-05, "loss": 0.6425, "step": 6745 }, { "epoch": 0.7, "learning_rate": 5.553376339301739e-05, "loss": 0.5756, "step": 6746 }, { "epoch": 0.7, "learning_rate": 5.553247115412767e-05, "loss": 0.5383, "step": 6747 }, { "epoch": 0.7, "learning_rate": 5.553117874335932e-05, "loss": 0.4995, "step": 6748 }, { "epoch": 0.7, "learning_rate": 5.552988616072106e-05, "loss": 1.4096, "step": 6749 }, { "epoch": 0.7, "learning_rate": 5.5528593406221575e-05, "loss": 0.3564, "step": 6750 }, { "epoch": 0.71, "learning_rate": 5.552730047986957e-05, "loss": 1.0515, "step": 6751 }, { "epoch": 0.71, "learning_rate": 5.552600738167375e-05, "loss": 1.4256, "step": 6752 }, { "epoch": 0.71, "learning_rate": 5.5524714111642824e-05, "loss": 1.1682, "step": 6753 }, { "epoch": 0.71, "learning_rate": 5.55234206697855e-05, "loss": 0.3364, "step": 6754 }, { "epoch": 0.71, "learning_rate": 5.552212705611047e-05, "loss": 1.6782, "step": 6755 }, { "epoch": 0.71, "learning_rate": 5.552083327062646e-05, "loss": 0.8968, "step": 6756 }, { "epoch": 0.71, "learning_rate": 5.551953931334217e-05, "loss": 1.6787, "step": 6757 }, { "epoch": 0.71, "learning_rate": 5.5518245184266315e-05, "loss": 0.2712, "step": 6758 }, { "epoch": 0.71, "learning_rate": 5.5516950883407605e-05, "loss": 1.242, "step": 6759 }, { "epoch": 0.71, "learning_rate": 5.551565641077475e-05, "loss": 1.6704, "step": 6760 }, { "epoch": 0.71, "learning_rate": 5.5514361766376456e-05, "loss": 0.9561, "step": 6761 }, { "epoch": 0.71, "learning_rate": 5.551306695022146e-05, "loss": 1.5906, "step": 6762 }, { "epoch": 0.71, "learning_rate": 5.551177196231846e-05, "loss": 1.54, "step": 6763 }, { "epoch": 0.71, "learning_rate": 5.551047680267619e-05, "loss": 0.757, "step": 6764 }, { "epoch": 0.71, "learning_rate": 5.550918147130335e-05, "loss": 0.971, "step": 6765 }, { "epoch": 0.71, "learning_rate": 5.550788596820867e-05, "loss": 0.3266, "step": 6766 }, { "epoch": 0.71, "learning_rate": 5.550659029340087e-05, "loss": 1.4481, "step": 6767 }, { "epoch": 0.71, "learning_rate": 5.5505294446888676e-05, "loss": 1.0652, "step": 6768 }, { "epoch": 0.71, "learning_rate": 5.550399842868081e-05, "loss": 1.0715, "step": 6769 }, { "epoch": 0.71, "learning_rate": 5.5502702238785986e-05, "loss": 0.5017, "step": 6770 }, { "epoch": 0.71, "learning_rate": 5.5501405877212945e-05, "loss": 1.2107, "step": 6771 }, { "epoch": 0.71, "learning_rate": 5.55001093439704e-05, "loss": 0.5176, "step": 6772 }, { "epoch": 0.71, "learning_rate": 5.549881263906709e-05, "loss": 0.7025, "step": 6773 }, { "epoch": 0.71, "learning_rate": 5.5497515762511734e-05, "loss": 0.1959, "step": 6774 }, { "epoch": 0.71, "learning_rate": 5.549621871431308e-05, "loss": 1.433, "step": 6775 }, { "epoch": 0.71, "learning_rate": 5.549492149447983e-05, "loss": 0.8417, "step": 6776 }, { "epoch": 0.71, "learning_rate": 5.549362410302075e-05, "loss": 2.1322, "step": 6777 }, { "epoch": 0.71, "learning_rate": 5.549232653994455e-05, "loss": 1.1949, "step": 6778 }, { "epoch": 0.71, "learning_rate": 5.549102880525997e-05, "loss": 0.6114, "step": 6779 }, { "epoch": 0.71, "learning_rate": 5.548973089897575e-05, "loss": 0.5651, "step": 6780 }, { "epoch": 0.71, "learning_rate": 5.548843282110063e-05, "loss": 1.2076, "step": 6781 }, { "epoch": 0.71, "learning_rate": 5.548713457164334e-05, "loss": 1.0919, "step": 6782 }, { "epoch": 0.71, "learning_rate": 5.548583615061263e-05, "loss": 0.8736, "step": 6783 }, { "epoch": 0.71, "learning_rate": 5.548453755801722e-05, "loss": 1.5699, "step": 6784 }, { "epoch": 0.71, "learning_rate": 5.5483238793865883e-05, "loss": 0.5318, "step": 6785 }, { "epoch": 0.71, "learning_rate": 5.548193985816733e-05, "loss": 1.0874, "step": 6786 }, { "epoch": 0.71, "learning_rate": 5.548064075093033e-05, "loss": 0.5377, "step": 6787 }, { "epoch": 0.71, "learning_rate": 5.547934147216362e-05, "loss": 1.3198, "step": 6788 }, { "epoch": 0.71, "learning_rate": 5.547804202187594e-05, "loss": 1.0304, "step": 6789 }, { "epoch": 0.71, "learning_rate": 5.547674240007605e-05, "loss": 0.5104, "step": 6790 }, { "epoch": 0.71, "learning_rate": 5.547544260677267e-05, "loss": 1.6218, "step": 6791 }, { "epoch": 0.71, "learning_rate": 5.547414264197459e-05, "loss": 0.5568, "step": 6792 }, { "epoch": 0.71, "learning_rate": 5.547284250569054e-05, "loss": 1.8666, "step": 6793 }, { "epoch": 0.71, "learning_rate": 5.547154219792927e-05, "loss": 1.5956, "step": 6794 }, { "epoch": 0.71, "learning_rate": 5.547024171869954e-05, "loss": 0.7951, "step": 6795 }, { "epoch": 0.71, "learning_rate": 5.5468941068010104e-05, "loss": 0.9288, "step": 6796 }, { "epoch": 0.71, "learning_rate": 5.5467640245869724e-05, "loss": 0.8598, "step": 6797 }, { "epoch": 0.71, "learning_rate": 5.546633925228713e-05, "loss": 1.4025, "step": 6798 }, { "epoch": 0.71, "learning_rate": 5.5465038087271115e-05, "loss": 0.6315, "step": 6799 }, { "epoch": 0.71, "learning_rate": 5.546373675083041e-05, "loss": 1.5168, "step": 6800 }, { "epoch": 0.71, "learning_rate": 5.546243524297379e-05, "loss": 1.0931, "step": 6801 }, { "epoch": 0.71, "learning_rate": 5.546113356371002e-05, "loss": 1.8359, "step": 6802 }, { "epoch": 0.71, "learning_rate": 5.545983171304785e-05, "loss": 1.312, "step": 6803 }, { "epoch": 0.71, "learning_rate": 5.545852969099605e-05, "loss": 1.4507, "step": 6804 }, { "epoch": 0.71, "learning_rate": 5.545722749756339e-05, "loss": 1.4869, "step": 6805 }, { "epoch": 0.71, "learning_rate": 5.545592513275863e-05, "loss": 1.0158, "step": 6806 }, { "epoch": 0.71, "learning_rate": 5.545462259659053e-05, "loss": 0.6117, "step": 6807 }, { "epoch": 0.71, "learning_rate": 5.545331988906788e-05, "loss": 1.1164, "step": 6808 }, { "epoch": 0.71, "learning_rate": 5.545201701019943e-05, "loss": 0.6878, "step": 6809 }, { "epoch": 0.71, "learning_rate": 5.545071395999395e-05, "loss": 0.3282, "step": 6810 }, { "epoch": 0.71, "learning_rate": 5.544941073846023e-05, "loss": 1.4518, "step": 6811 }, { "epoch": 0.71, "learning_rate": 5.544810734560703e-05, "loss": 0.8774, "step": 6812 }, { "epoch": 0.71, "learning_rate": 5.5446803781443114e-05, "loss": 0.5627, "step": 6813 }, { "epoch": 0.71, "learning_rate": 5.5445500045977283e-05, "loss": 0.768, "step": 6814 }, { "epoch": 0.71, "learning_rate": 5.544419613921829e-05, "loss": 0.4812, "step": 6815 }, { "epoch": 0.71, "learning_rate": 5.5442892061174924e-05, "loss": 0.5664, "step": 6816 }, { "epoch": 0.71, "learning_rate": 5.544158781185596e-05, "loss": 1.0342, "step": 6817 }, { "epoch": 0.71, "learning_rate": 5.544028339127018e-05, "loss": 0.1567, "step": 6818 }, { "epoch": 0.71, "learning_rate": 5.543897879942638e-05, "loss": 1.8834, "step": 6819 }, { "epoch": 0.71, "learning_rate": 5.5437674036333304e-05, "loss": 0.609, "step": 6820 }, { "epoch": 0.71, "learning_rate": 5.5436369101999774e-05, "loss": 0.5804, "step": 6821 }, { "epoch": 0.71, "learning_rate": 5.5435063996434556e-05, "loss": 1.0137, "step": 6822 }, { "epoch": 0.71, "learning_rate": 5.543375871964643e-05, "loss": 1.1241, "step": 6823 }, { "epoch": 0.71, "learning_rate": 5.5432453271644204e-05, "loss": 1.0101, "step": 6824 }, { "epoch": 0.71, "learning_rate": 5.543114765243664e-05, "loss": 1.2421, "step": 6825 }, { "epoch": 0.71, "learning_rate": 5.5429841862032554e-05, "loss": 1.2424, "step": 6826 }, { "epoch": 0.71, "learning_rate": 5.5428535900440714e-05, "loss": 0.7504, "step": 6827 }, { "epoch": 0.71, "learning_rate": 5.542722976766992e-05, "loss": 1.2877, "step": 6828 }, { "epoch": 0.71, "learning_rate": 5.5425923463728973e-05, "loss": 1.3287, "step": 6829 }, { "epoch": 0.71, "learning_rate": 5.542461698862664e-05, "loss": 0.4132, "step": 6830 }, { "epoch": 0.71, "learning_rate": 5.5423310342371757e-05, "loss": 1.0702, "step": 6831 }, { "epoch": 0.71, "learning_rate": 5.5422003524973084e-05, "loss": 1.1336, "step": 6832 }, { "epoch": 0.71, "learning_rate": 5.5420696536439435e-05, "loss": 0.6169, "step": 6833 }, { "epoch": 0.71, "learning_rate": 5.54193893767796e-05, "loss": 0.5611, "step": 6834 }, { "epoch": 0.71, "learning_rate": 5.5418082046002395e-05, "loss": 1.2054, "step": 6835 }, { "epoch": 0.71, "learning_rate": 5.5416774544116594e-05, "loss": 0.7848, "step": 6836 }, { "epoch": 0.71, "learning_rate": 5.5415466871131026e-05, "loss": 1.0208, "step": 6837 }, { "epoch": 0.71, "learning_rate": 5.541415902705447e-05, "loss": 0.6389, "step": 6838 }, { "epoch": 0.71, "learning_rate": 5.5412851011895756e-05, "loss": 1.1344, "step": 6839 }, { "epoch": 0.71, "learning_rate": 5.5411542825663666e-05, "loss": 0.7857, "step": 6840 }, { "epoch": 0.71, "learning_rate": 5.541023446836702e-05, "loss": 0.6592, "step": 6841 }, { "epoch": 0.71, "learning_rate": 5.5408925940014626e-05, "loss": 0.3238, "step": 6842 }, { "epoch": 0.71, "learning_rate": 5.5407617240615275e-05, "loss": 1.5531, "step": 6843 }, { "epoch": 0.71, "learning_rate": 5.540630837017779e-05, "loss": 0.568, "step": 6844 }, { "epoch": 0.71, "learning_rate": 5.540499932871099e-05, "loss": 1.5416, "step": 6845 }, { "epoch": 0.71, "learning_rate": 5.540369011622368e-05, "loss": 0.425, "step": 6846 }, { "epoch": 0.72, "learning_rate": 5.540238073272467e-05, "loss": 0.6589, "step": 6847 }, { "epoch": 0.72, "learning_rate": 5.540107117822277e-05, "loss": 2.1108, "step": 6848 }, { "epoch": 0.72, "learning_rate": 5.539976145272681e-05, "loss": 0.3202, "step": 6849 }, { "epoch": 0.72, "learning_rate": 5.5398451556245595e-05, "loss": 0.5212, "step": 6850 }, { "epoch": 0.72, "learning_rate": 5.539714148878795e-05, "loss": 2.0469, "step": 6851 }, { "epoch": 0.72, "learning_rate": 5.5395831250362686e-05, "loss": 0.9439, "step": 6852 }, { "epoch": 0.72, "learning_rate": 5.539452084097863e-05, "loss": 1.4867, "step": 6853 }, { "epoch": 0.72, "learning_rate": 5.53932102606446e-05, "loss": 0.4142, "step": 6854 }, { "epoch": 0.72, "learning_rate": 5.539189950936943e-05, "loss": 0.9268, "step": 6855 }, { "epoch": 0.72, "learning_rate": 5.5390588587161916e-05, "loss": 1.1309, "step": 6856 }, { "epoch": 0.72, "learning_rate": 5.538927749403091e-05, "loss": 0.4331, "step": 6857 }, { "epoch": 0.72, "learning_rate": 5.538796622998523e-05, "loss": 1.0551, "step": 6858 }, { "epoch": 0.72, "learning_rate": 5.538665479503371e-05, "loss": 1.1585, "step": 6859 }, { "epoch": 0.72, "learning_rate": 5.538534318918515e-05, "loss": 0.6708, "step": 6860 }, { "epoch": 0.72, "learning_rate": 5.538403141244841e-05, "loss": 1.1105, "step": 6861 }, { "epoch": 0.72, "learning_rate": 5.5382719464832314e-05, "loss": 0.5136, "step": 6862 }, { "epoch": 0.72, "learning_rate": 5.5381407346345684e-05, "loss": 0.303, "step": 6863 }, { "epoch": 0.72, "learning_rate": 5.538009505699736e-05, "loss": 0.5969, "step": 6864 }, { "epoch": 0.72, "learning_rate": 5.5378782596796175e-05, "loss": 0.3982, "step": 6865 }, { "epoch": 0.72, "learning_rate": 5.537746996575097e-05, "loss": 1.0208, "step": 6866 }, { "epoch": 0.72, "learning_rate": 5.537615716387056e-05, "loss": 0.6825, "step": 6867 }, { "epoch": 0.72, "learning_rate": 5.537484419116381e-05, "loss": 0.5561, "step": 6868 }, { "epoch": 0.72, "learning_rate": 5.537353104763954e-05, "loss": 0.2379, "step": 6869 }, { "epoch": 0.72, "learning_rate": 5.53722177333066e-05, "loss": 0.7799, "step": 6870 }, { "epoch": 0.72, "learning_rate": 5.537090424817382e-05, "loss": 0.5324, "step": 6871 }, { "epoch": 0.72, "learning_rate": 5.536959059225006e-05, "loss": 0.7107, "step": 6872 }, { "epoch": 0.72, "learning_rate": 5.536827676554414e-05, "loss": 1.1977, "step": 6873 }, { "epoch": 0.72, "learning_rate": 5.536696276806493e-05, "loss": 0.829, "step": 6874 }, { "epoch": 0.72, "learning_rate": 5.536564859982125e-05, "loss": 1.1023, "step": 6875 }, { "epoch": 0.72, "learning_rate": 5.5364334260821964e-05, "loss": 0.6105, "step": 6876 }, { "epoch": 0.72, "learning_rate": 5.5363019751075916e-05, "loss": 0.1351, "step": 6877 }, { "epoch": 0.72, "learning_rate": 5.536170507059195e-05, "loss": 1.281, "step": 6878 }, { "epoch": 0.72, "learning_rate": 5.536039021937892e-05, "loss": 0.7677, "step": 6879 }, { "epoch": 0.72, "learning_rate": 5.5359075197445685e-05, "loss": 0.3315, "step": 6880 }, { "epoch": 0.72, "learning_rate": 5.535776000480108e-05, "loss": 0.8385, "step": 6881 }, { "epoch": 0.72, "learning_rate": 5.535644464145397e-05, "loss": 1.0024, "step": 6882 }, { "epoch": 0.72, "learning_rate": 5.53551291074132e-05, "loss": 0.4098, "step": 6883 }, { "epoch": 0.72, "learning_rate": 5.535381340268764e-05, "loss": 0.3282, "step": 6884 }, { "epoch": 0.72, "learning_rate": 5.535249752728615e-05, "loss": 0.8526, "step": 6885 }, { "epoch": 0.72, "learning_rate": 5.535118148121756e-05, "loss": 1.2518, "step": 6886 }, { "epoch": 0.72, "learning_rate": 5.534986526449076e-05, "loss": 1.1711, "step": 6887 }, { "epoch": 0.72, "learning_rate": 5.53485488771146e-05, "loss": 1.1441, "step": 6888 }, { "epoch": 0.72, "learning_rate": 5.5347232319097937e-05, "loss": 1.0252, "step": 6889 }, { "epoch": 0.72, "learning_rate": 5.534591559044964e-05, "loss": 1.5965, "step": 6890 }, { "epoch": 0.72, "learning_rate": 5.534459869117856e-05, "loss": 1.4621, "step": 6891 }, { "epoch": 0.72, "learning_rate": 5.534328162129358e-05, "loss": 0.5466, "step": 6892 }, { "epoch": 0.72, "learning_rate": 5.5341964380803565e-05, "loss": 1.0915, "step": 6893 }, { "epoch": 0.72, "learning_rate": 5.534064696971736e-05, "loss": 0.7359, "step": 6894 }, { "epoch": 0.72, "learning_rate": 5.533932938804386e-05, "loss": 0.5449, "step": 6895 }, { "epoch": 0.72, "learning_rate": 5.533801163579191e-05, "loss": 0.9019, "step": 6896 }, { "epoch": 0.72, "learning_rate": 5.533669371297041e-05, "loss": 1.5809, "step": 6897 }, { "epoch": 0.72, "learning_rate": 5.5335375619588206e-05, "loss": 1.0969, "step": 6898 }, { "epoch": 0.72, "learning_rate": 5.5334057355654184e-05, "loss": 0.5547, "step": 6899 }, { "epoch": 0.72, "learning_rate": 5.5332738921177214e-05, "loss": 0.6027, "step": 6900 }, { "epoch": 0.72, "learning_rate": 5.533142031616617e-05, "loss": 1.0363, "step": 6901 }, { "epoch": 0.72, "learning_rate": 5.533010154062994e-05, "loss": 0.891, "step": 6902 }, { "epoch": 0.72, "learning_rate": 5.5328782594577395e-05, "loss": 0.3641, "step": 6903 }, { "epoch": 0.72, "learning_rate": 5.5327463478017406e-05, "loss": 0.5615, "step": 6904 }, { "epoch": 0.72, "learning_rate": 5.5326144190958854e-05, "loss": 2.0131, "step": 6905 }, { "epoch": 0.72, "learning_rate": 5.5324824733410635e-05, "loss": 0.4697, "step": 6906 }, { "epoch": 0.72, "learning_rate": 5.532350510538162e-05, "loss": 1.4754, "step": 6907 }, { "epoch": 0.72, "learning_rate": 5.532218530688069e-05, "loss": 0.5624, "step": 6908 }, { "epoch": 0.72, "learning_rate": 5.532086533791673e-05, "loss": 1.1733, "step": 6909 }, { "epoch": 0.72, "learning_rate": 5.5319545198498635e-05, "loss": 0.7399, "step": 6910 }, { "epoch": 0.72, "learning_rate": 5.531822488863528e-05, "loss": 0.7927, "step": 6911 }, { "epoch": 0.72, "learning_rate": 5.531690440833557e-05, "loss": 0.5308, "step": 6912 }, { "epoch": 0.72, "learning_rate": 5.5315583757608375e-05, "loss": 1.2799, "step": 6913 }, { "epoch": 0.72, "learning_rate": 5.53142629364626e-05, "loss": 0.2729, "step": 6914 }, { "epoch": 0.72, "learning_rate": 5.5312941944907126e-05, "loss": 0.8055, "step": 6915 }, { "epoch": 0.72, "learning_rate": 5.531162078295085e-05, "loss": 0.646, "step": 6916 }, { "epoch": 0.72, "learning_rate": 5.531029945060266e-05, "loss": 1.4057, "step": 6917 }, { "epoch": 0.72, "learning_rate": 5.530897794787146e-05, "loss": 1.3158, "step": 6918 }, { "epoch": 0.72, "learning_rate": 5.530765627476615e-05, "loss": 1.662, "step": 6919 }, { "epoch": 0.72, "learning_rate": 5.5306334431295606e-05, "loss": 0.7056, "step": 6920 }, { "epoch": 0.72, "learning_rate": 5.530501241746875e-05, "loss": 0.4124, "step": 6921 }, { "epoch": 0.72, "learning_rate": 5.5303690233294466e-05, "loss": 0.4385, "step": 6922 }, { "epoch": 0.72, "learning_rate": 5.5302367878781664e-05, "loss": 0.4715, "step": 6923 }, { "epoch": 0.72, "learning_rate": 5.530104535393924e-05, "loss": 0.7169, "step": 6924 }, { "epoch": 0.72, "learning_rate": 5.5299722658776096e-05, "loss": 0.7045, "step": 6925 }, { "epoch": 0.72, "learning_rate": 5.5298399793301144e-05, "loss": 0.2107, "step": 6926 }, { "epoch": 0.72, "learning_rate": 5.529707675752328e-05, "loss": 0.8755, "step": 6927 }, { "epoch": 0.72, "learning_rate": 5.529575355145141e-05, "loss": 0.5812, "step": 6928 }, { "epoch": 0.72, "learning_rate": 5.5294430175094455e-05, "loss": 1.4526, "step": 6929 }, { "epoch": 0.72, "learning_rate": 5.529310662846131e-05, "loss": 0.5285, "step": 6930 }, { "epoch": 0.72, "learning_rate": 5.529178291156089e-05, "loss": 0.6255, "step": 6931 }, { "epoch": 0.72, "learning_rate": 5.52904590244021e-05, "loss": 0.2842, "step": 6932 }, { "epoch": 0.72, "learning_rate": 5.5289134966993865e-05, "loss": 0.4712, "step": 6933 }, { "epoch": 0.72, "learning_rate": 5.5287810739345085e-05, "loss": 1.1982, "step": 6934 }, { "epoch": 0.72, "learning_rate": 5.5286486341464685e-05, "loss": 0.3602, "step": 6935 }, { "epoch": 0.72, "learning_rate": 5.528516177336157e-05, "loss": 0.4145, "step": 6936 }, { "epoch": 0.72, "learning_rate": 5.528383703504467e-05, "loss": 0.8951, "step": 6937 }, { "epoch": 0.72, "learning_rate": 5.528251212652288e-05, "loss": 0.859, "step": 6938 }, { "epoch": 0.72, "learning_rate": 5.528118704780515e-05, "loss": 1.415, "step": 6939 }, { "epoch": 0.72, "learning_rate": 5.527986179890039e-05, "loss": 1.2857, "step": 6940 }, { "epoch": 0.72, "learning_rate": 5.5278536379817496e-05, "loss": 0.953, "step": 6941 }, { "epoch": 0.73, "learning_rate": 5.5277210790565415e-05, "loss": 0.8244, "step": 6942 }, { "epoch": 0.73, "learning_rate": 5.527588503115307e-05, "loss": 0.8103, "step": 6943 }, { "epoch": 0.73, "learning_rate": 5.527455910158937e-05, "loss": 1.4222, "step": 6944 }, { "epoch": 0.73, "learning_rate": 5.527323300188327e-05, "loss": 0.9421, "step": 6945 }, { "epoch": 0.73, "learning_rate": 5.527190673204367e-05, "loss": 1.3049, "step": 6946 }, { "epoch": 0.73, "learning_rate": 5.527058029207951e-05, "loss": 0.2148, "step": 6947 }, { "epoch": 0.73, "learning_rate": 5.526925368199971e-05, "loss": 1.1214, "step": 6948 }, { "epoch": 0.73, "learning_rate": 5.526792690181321e-05, "loss": 1.553, "step": 6949 }, { "epoch": 0.73, "learning_rate": 5.526659995152894e-05, "loss": 2.0255, "step": 6950 }, { "epoch": 0.73, "learning_rate": 5.526527283115583e-05, "loss": 0.5784, "step": 6951 }, { "epoch": 0.73, "learning_rate": 5.526394554070281e-05, "loss": 0.7912, "step": 6952 }, { "epoch": 0.73, "learning_rate": 5.526261808017883e-05, "loss": 1.2126, "step": 6953 }, { "epoch": 0.73, "learning_rate": 5.526129044959281e-05, "loss": 1.573, "step": 6954 }, { "epoch": 0.73, "learning_rate": 5.5259962648953695e-05, "loss": 0.3501, "step": 6955 }, { "epoch": 0.73, "learning_rate": 5.5258634678270414e-05, "loss": 1.1049, "step": 6956 }, { "epoch": 0.73, "learning_rate": 5.525730653755192e-05, "loss": 1.1153, "step": 6957 }, { "epoch": 0.73, "learning_rate": 5.525597822680715e-05, "loss": 0.9551, "step": 6958 }, { "epoch": 0.73, "learning_rate": 5.525464974604504e-05, "loss": 0.7421, "step": 6959 }, { "epoch": 0.73, "learning_rate": 5.525332109527454e-05, "loss": 1.6352, "step": 6960 }, { "epoch": 0.73, "learning_rate": 5.52519922745046e-05, "loss": 1.0244, "step": 6961 }, { "epoch": 0.73, "learning_rate": 5.5250663283744147e-05, "loss": 0.5109, "step": 6962 }, { "epoch": 0.73, "learning_rate": 5.524933412300213e-05, "loss": 0.9639, "step": 6963 }, { "epoch": 0.73, "learning_rate": 5.524800479228752e-05, "loss": 1.7783, "step": 6964 }, { "epoch": 0.73, "learning_rate": 5.524667529160924e-05, "loss": 0.8124, "step": 6965 }, { "epoch": 0.73, "learning_rate": 5.524534562097625e-05, "loss": 0.7043, "step": 6966 }, { "epoch": 0.73, "learning_rate": 5.524401578039749e-05, "loss": 0.9214, "step": 6967 }, { "epoch": 0.73, "learning_rate": 5.524268576988194e-05, "loss": 0.9015, "step": 6968 }, { "epoch": 0.73, "learning_rate": 5.5241355589438525e-05, "loss": 1.2469, "step": 6969 }, { "epoch": 0.73, "learning_rate": 5.524002523907621e-05, "loss": 0.798, "step": 6970 }, { "epoch": 0.73, "learning_rate": 5.523869471880396e-05, "loss": 1.0248, "step": 6971 }, { "epoch": 0.73, "learning_rate": 5.523736402863072e-05, "loss": 0.3526, "step": 6972 }, { "epoch": 0.73, "learning_rate": 5.523603316856544e-05, "loss": 1.5942, "step": 6973 }, { "epoch": 0.73, "learning_rate": 5.52347021386171e-05, "loss": 1.1483, "step": 6974 }, { "epoch": 0.73, "learning_rate": 5.523337093879464e-05, "loss": 0.4318, "step": 6975 }, { "epoch": 0.73, "learning_rate": 5.523203956910704e-05, "loss": 0.53, "step": 6976 }, { "epoch": 0.73, "learning_rate": 5.5230708029563254e-05, "loss": 0.3498, "step": 6977 }, { "epoch": 0.73, "learning_rate": 5.522937632017224e-05, "loss": 1.6458, "step": 6978 }, { "epoch": 0.73, "learning_rate": 5.522804444094297e-05, "loss": 0.2004, "step": 6979 }, { "epoch": 0.73, "learning_rate": 5.522671239188441e-05, "loss": 0.4703, "step": 6980 }, { "epoch": 0.73, "learning_rate": 5.522538017300552e-05, "loss": 0.7313, "step": 6981 }, { "epoch": 0.73, "learning_rate": 5.5224047784315274e-05, "loss": 0.3797, "step": 6982 }, { "epoch": 0.73, "learning_rate": 5.522271522582264e-05, "loss": 0.7715, "step": 6983 }, { "epoch": 0.73, "learning_rate": 5.522138249753659e-05, "loss": 0.2577, "step": 6984 }, { "epoch": 0.73, "learning_rate": 5.522004959946609e-05, "loss": 1.1992, "step": 6985 }, { "epoch": 0.73, "learning_rate": 5.521871653162013e-05, "loss": 0.4647, "step": 6986 }, { "epoch": 0.73, "learning_rate": 5.5217383294007654e-05, "loss": 0.3042, "step": 6987 }, { "epoch": 0.73, "learning_rate": 5.521604988663767e-05, "loss": 1.4383, "step": 6988 }, { "epoch": 0.73, "learning_rate": 5.521471630951912e-05, "loss": 1.124, "step": 6989 }, { "epoch": 0.73, "learning_rate": 5.521338256266102e-05, "loss": 1.2631, "step": 6990 }, { "epoch": 0.73, "learning_rate": 5.521204864607231e-05, "loss": 1.1891, "step": 6991 }, { "epoch": 0.73, "learning_rate": 5.5210714559761996e-05, "loss": 0.3825, "step": 6992 }, { "epoch": 0.73, "learning_rate": 5.520938030373905e-05, "loss": 1.7078, "step": 6993 }, { "epoch": 0.73, "learning_rate": 5.520804587801246e-05, "loss": 1.5049, "step": 6994 }, { "epoch": 0.73, "learning_rate": 5.5206711282591196e-05, "loss": 2.0512, "step": 6995 }, { "epoch": 0.73, "learning_rate": 5.520537651748425e-05, "loss": 1.043, "step": 6996 }, { "epoch": 0.73, "learning_rate": 5.520404158270061e-05, "loss": 0.4305, "step": 6997 }, { "epoch": 0.73, "learning_rate": 5.520270647824926e-05, "loss": 0.1639, "step": 6998 }, { "epoch": 0.73, "learning_rate": 5.5201371204139186e-05, "loss": 0.9466, "step": 6999 }, { "epoch": 0.73, "learning_rate": 5.520003576037937e-05, "loss": 0.1297, "step": 7000 }, { "epoch": 0.73, "learning_rate": 5.5198700146978814e-05, "loss": 1.0794, "step": 7001 }, { "epoch": 0.73, "learning_rate": 5.519736436394651e-05, "loss": 1.1579, "step": 7002 }, { "epoch": 0.73, "learning_rate": 5.5196028411291445e-05, "loss": 0.6269, "step": 7003 }, { "epoch": 0.73, "learning_rate": 5.519469228902261e-05, "loss": 0.4144, "step": 7004 }, { "epoch": 0.73, "learning_rate": 5.5193355997148994e-05, "loss": 1.0552, "step": 7005 }, { "epoch": 0.73, "learning_rate": 5.51920195356796e-05, "loss": 1.3197, "step": 7006 }, { "epoch": 0.73, "learning_rate": 5.519068290462344e-05, "loss": 0.4868, "step": 7007 }, { "epoch": 0.73, "learning_rate": 5.5189346103989486e-05, "loss": 1.774, "step": 7008 }, { "epoch": 0.73, "learning_rate": 5.518800913378675e-05, "loss": 1.1756, "step": 7009 }, { "epoch": 0.73, "learning_rate": 5.518667199402424e-05, "loss": 0.3091, "step": 7010 }, { "epoch": 0.73, "learning_rate": 5.518533468471093e-05, "loss": 1.035, "step": 7011 }, { "epoch": 0.73, "learning_rate": 5.518399720585585e-05, "loss": 0.9333, "step": 7012 }, { "epoch": 0.73, "learning_rate": 5.518265955746799e-05, "loss": 1.5918, "step": 7013 }, { "epoch": 0.73, "learning_rate": 5.5181321739556356e-05, "loss": 0.7551, "step": 7014 }, { "epoch": 0.73, "learning_rate": 5.5179983752129963e-05, "loss": 0.3233, "step": 7015 }, { "epoch": 0.73, "learning_rate": 5.5178645595197804e-05, "loss": 1.0407, "step": 7016 }, { "epoch": 0.73, "learning_rate": 5.51773072687689e-05, "loss": 1.7283, "step": 7017 }, { "epoch": 0.73, "learning_rate": 5.5175968772852255e-05, "loss": 0.955, "step": 7018 }, { "epoch": 0.73, "learning_rate": 5.5174630107456864e-05, "loss": 1.1952, "step": 7019 }, { "epoch": 0.73, "learning_rate": 5.5173291272591775e-05, "loss": 1.6944, "step": 7020 }, { "epoch": 0.73, "learning_rate": 5.517195226826596e-05, "loss": 0.7677, "step": 7021 }, { "epoch": 0.73, "learning_rate": 5.517061309448846e-05, "loss": 0.7111, "step": 7022 }, { "epoch": 0.73, "learning_rate": 5.516927375126829e-05, "loss": 0.7453, "step": 7023 }, { "epoch": 0.73, "learning_rate": 5.516793423861445e-05, "loss": 1.5088, "step": 7024 }, { "epoch": 0.73, "learning_rate": 5.516659455653596e-05, "loss": 0.6103, "step": 7025 }, { "epoch": 0.73, "learning_rate": 5.516525470504186e-05, "loss": 1.3866, "step": 7026 }, { "epoch": 0.73, "learning_rate": 5.5163914684141146e-05, "loss": 1.1531, "step": 7027 }, { "epoch": 0.73, "learning_rate": 5.516257449384284e-05, "loss": 1.4847, "step": 7028 }, { "epoch": 0.73, "learning_rate": 5.516123413415598e-05, "loss": 1.6209, "step": 7029 }, { "epoch": 0.73, "learning_rate": 5.5159893605089574e-05, "loss": 0.3048, "step": 7030 }, { "epoch": 0.73, "learning_rate": 5.515855290665265e-05, "loss": 0.3852, "step": 7031 }, { "epoch": 0.73, "learning_rate": 5.515721203885423e-05, "loss": 0.8361, "step": 7032 }, { "epoch": 0.73, "learning_rate": 5.5155871001703355e-05, "loss": 0.6217, "step": 7033 }, { "epoch": 0.73, "learning_rate": 5.515452979520904e-05, "loss": 0.475, "step": 7034 }, { "epoch": 0.73, "learning_rate": 5.515318841938031e-05, "loss": 1.5357, "step": 7035 }, { "epoch": 0.73, "learning_rate": 5.515184687422621e-05, "loss": 0.8708, "step": 7036 }, { "epoch": 0.73, "learning_rate": 5.515050515975575e-05, "loss": 2.0134, "step": 7037 }, { "epoch": 0.74, "learning_rate": 5.514916327597799e-05, "loss": 0.5114, "step": 7038 }, { "epoch": 0.74, "learning_rate": 5.514782122290193e-05, "loss": 1.7546, "step": 7039 }, { "epoch": 0.74, "learning_rate": 5.514647900053663e-05, "loss": 0.2997, "step": 7040 }, { "epoch": 0.74, "learning_rate": 5.514513660889112e-05, "loss": 1.3143, "step": 7041 }, { "epoch": 0.74, "learning_rate": 5.514379404797443e-05, "loss": 1.2483, "step": 7042 }, { "epoch": 0.74, "learning_rate": 5.5142451317795597e-05, "loss": 1.0303, "step": 7043 }, { "epoch": 0.74, "learning_rate": 5.514110841836367e-05, "loss": 0.4681, "step": 7044 }, { "epoch": 0.74, "learning_rate": 5.513976534968769e-05, "loss": 0.3394, "step": 7045 }, { "epoch": 0.74, "learning_rate": 5.5138422111776684e-05, "loss": 0.1764, "step": 7046 }, { "epoch": 0.74, "learning_rate": 5.51370787046397e-05, "loss": 0.7782, "step": 7047 }, { "epoch": 0.74, "learning_rate": 5.5135735128285786e-05, "loss": 0.866, "step": 7048 }, { "epoch": 0.74, "learning_rate": 5.513439138272399e-05, "loss": 0.6796, "step": 7049 }, { "epoch": 0.74, "learning_rate": 5.513304746796335e-05, "loss": 0.9749, "step": 7050 }, { "epoch": 0.74, "learning_rate": 5.513170338401291e-05, "loss": 0.4692, "step": 7051 }, { "epoch": 0.74, "learning_rate": 5.513035913088173e-05, "loss": 1.2631, "step": 7052 }, { "epoch": 0.74, "learning_rate": 5.5129014708578844e-05, "loss": 1.3923, "step": 7053 }, { "epoch": 0.74, "learning_rate": 5.512767011711332e-05, "loss": 0.8922, "step": 7054 }, { "epoch": 0.74, "learning_rate": 5.5126325356494195e-05, "loss": 1.1151, "step": 7055 }, { "epoch": 0.74, "learning_rate": 5.512498042673053e-05, "loss": 1.0293, "step": 7056 }, { "epoch": 0.74, "learning_rate": 5.512363532783137e-05, "loss": 0.3136, "step": 7057 }, { "epoch": 0.74, "learning_rate": 5.512229005980578e-05, "loss": 1.3211, "step": 7058 }, { "epoch": 0.74, "learning_rate": 5.512094462266281e-05, "loss": 0.0282, "step": 7059 }, { "epoch": 0.74, "learning_rate": 5.511959901641151e-05, "loss": 1.5658, "step": 7060 }, { "epoch": 0.74, "learning_rate": 5.5118253241060964e-05, "loss": 1.4244, "step": 7061 }, { "epoch": 0.74, "learning_rate": 5.51169072966202e-05, "loss": 0.8821, "step": 7062 }, { "epoch": 0.74, "learning_rate": 5.5115561183098306e-05, "loss": 0.3723, "step": 7063 }, { "epoch": 0.74, "learning_rate": 5.511421490050432e-05, "loss": 0.831, "step": 7064 }, { "epoch": 0.74, "learning_rate": 5.5112868448847324e-05, "loss": 0.3917, "step": 7065 }, { "epoch": 0.74, "learning_rate": 5.511152182813637e-05, "loss": 1.3757, "step": 7066 }, { "epoch": 0.74, "learning_rate": 5.511017503838052e-05, "loss": 1.1304, "step": 7067 }, { "epoch": 0.74, "learning_rate": 5.5108828079588855e-05, "loss": 1.5215, "step": 7068 }, { "epoch": 0.74, "learning_rate": 5.510748095177043e-05, "loss": 0.3374, "step": 7069 }, { "epoch": 0.74, "learning_rate": 5.510613365493431e-05, "loss": 0.3409, "step": 7070 }, { "epoch": 0.74, "learning_rate": 5.510478618908959e-05, "loss": 1.5392, "step": 7071 }, { "epoch": 0.74, "learning_rate": 5.510343855424532e-05, "loss": 1.2527, "step": 7072 }, { "epoch": 0.74, "learning_rate": 5.5102090750410566e-05, "loss": 0.6973, "step": 7073 }, { "epoch": 0.74, "learning_rate": 5.5100742777594413e-05, "loss": 0.179, "step": 7074 }, { "epoch": 0.74, "learning_rate": 5.509939463580594e-05, "loss": 0.635, "step": 7075 }, { "epoch": 0.74, "learning_rate": 5.509804632505421e-05, "loss": 0.3098, "step": 7076 }, { "epoch": 0.74, "learning_rate": 5.50966978453483e-05, "loss": 0.3333, "step": 7077 }, { "epoch": 0.74, "learning_rate": 5.509534919669731e-05, "loss": 2.015, "step": 7078 }, { "epoch": 0.74, "learning_rate": 5.509400037911029e-05, "loss": 1.1737, "step": 7079 }, { "epoch": 0.74, "learning_rate": 5.509265139259632e-05, "loss": 0.2479, "step": 7080 }, { "epoch": 0.74, "learning_rate": 5.50913022371645e-05, "loss": 0.2961, "step": 7081 }, { "epoch": 0.74, "learning_rate": 5.5089952912823905e-05, "loss": 1.1575, "step": 7082 }, { "epoch": 0.74, "learning_rate": 5.508860341958362e-05, "loss": 0.6108, "step": 7083 }, { "epoch": 0.74, "learning_rate": 5.508725375745272e-05, "loss": 0.7635, "step": 7084 }, { "epoch": 0.74, "learning_rate": 5.508590392644031e-05, "loss": 1.4074, "step": 7085 }, { "epoch": 0.74, "learning_rate": 5.508455392655545e-05, "loss": 0.9646, "step": 7086 }, { "epoch": 0.74, "learning_rate": 5.508320375780726e-05, "loss": 0.7947, "step": 7087 }, { "epoch": 0.74, "learning_rate": 5.5081853420204795e-05, "loss": 1.618, "step": 7088 }, { "epoch": 0.74, "learning_rate": 5.508050291375717e-05, "loss": 1.6061, "step": 7089 }, { "epoch": 0.74, "learning_rate": 5.507915223847346e-05, "loss": 1.4975, "step": 7090 }, { "epoch": 0.74, "learning_rate": 5.507780139436277e-05, "loss": 0.8898, "step": 7091 }, { "epoch": 0.74, "learning_rate": 5.507645038143419e-05, "loss": 0.4285, "step": 7092 }, { "epoch": 0.74, "learning_rate": 5.5075099199696814e-05, "loss": 0.5539, "step": 7093 }, { "epoch": 0.74, "learning_rate": 5.5073747849159735e-05, "loss": 0.559, "step": 7094 }, { "epoch": 0.74, "learning_rate": 5.5072396329832055e-05, "loss": 1.0905, "step": 7095 }, { "epoch": 0.74, "learning_rate": 5.507104464172286e-05, "loss": 1.3864, "step": 7096 }, { "epoch": 0.74, "learning_rate": 5.506969278484127e-05, "loss": 0.2438, "step": 7097 }, { "epoch": 0.74, "learning_rate": 5.5068340759196375e-05, "loss": 1.2832, "step": 7098 }, { "epoch": 0.74, "learning_rate": 5.5066988564797267e-05, "loss": 0.7185, "step": 7099 }, { "epoch": 0.74, "learning_rate": 5.506563620165306e-05, "loss": 1.1522, "step": 7100 }, { "epoch": 0.74, "learning_rate": 5.506428366977286e-05, "loss": 0.6821, "step": 7101 }, { "epoch": 0.74, "learning_rate": 5.5062930969165753e-05, "loss": 1.6129, "step": 7102 }, { "epoch": 0.74, "learning_rate": 5.506157809984087e-05, "loss": 1.4818, "step": 7103 }, { "epoch": 0.74, "learning_rate": 5.506022506180731e-05, "loss": 1.4354, "step": 7104 }, { "epoch": 0.74, "learning_rate": 5.505887185507417e-05, "loss": 1.1818, "step": 7105 }, { "epoch": 0.74, "learning_rate": 5.505751847965058e-05, "loss": 1.0373, "step": 7106 }, { "epoch": 0.74, "learning_rate": 5.5056164935545625e-05, "loss": 1.4578, "step": 7107 }, { "epoch": 0.74, "learning_rate": 5.505481122276844e-05, "loss": 0.6891, "step": 7108 }, { "epoch": 0.74, "learning_rate": 5.505345734132812e-05, "loss": 1.3112, "step": 7109 }, { "epoch": 0.74, "learning_rate": 5.5052103291233796e-05, "loss": 0.3333, "step": 7110 }, { "epoch": 0.74, "learning_rate": 5.505074907249457e-05, "loss": 0.9466, "step": 7111 }, { "epoch": 0.74, "learning_rate": 5.504939468511956e-05, "loss": 1.256, "step": 7112 }, { "epoch": 0.74, "learning_rate": 5.504804012911789e-05, "loss": 1.3895, "step": 7113 }, { "epoch": 0.74, "learning_rate": 5.504668540449867e-05, "loss": 1.3098, "step": 7114 }, { "epoch": 0.74, "learning_rate": 5.504533051127103e-05, "loss": 0.6726, "step": 7115 }, { "epoch": 0.74, "learning_rate": 5.504397544944408e-05, "loss": 1.2503, "step": 7116 }, { "epoch": 0.74, "learning_rate": 5.5042620219026954e-05, "loss": 0.994, "step": 7117 }, { "epoch": 0.74, "learning_rate": 5.504126482002877e-05, "loss": 0.2583, "step": 7118 }, { "epoch": 0.74, "learning_rate": 5.503990925245864e-05, "loss": 0.7819, "step": 7119 }, { "epoch": 0.74, "learning_rate": 5.5038553516325694e-05, "loss": 1.158, "step": 7120 }, { "epoch": 0.74, "learning_rate": 5.5037197611639083e-05, "loss": 1.6284, "step": 7121 }, { "epoch": 0.74, "learning_rate": 5.5035841538407905e-05, "loss": 0.6647, "step": 7122 }, { "epoch": 0.74, "learning_rate": 5.50344852966413e-05, "loss": 1.4899, "step": 7123 }, { "epoch": 0.74, "learning_rate": 5.5033128886348405e-05, "loss": 0.5099, "step": 7124 }, { "epoch": 0.74, "learning_rate": 5.503177230753834e-05, "loss": 1.0005, "step": 7125 }, { "epoch": 0.74, "learning_rate": 5.5030415560220234e-05, "loss": 0.7586, "step": 7126 }, { "epoch": 0.74, "learning_rate": 5.5029058644403235e-05, "loss": 0.3254, "step": 7127 }, { "epoch": 0.74, "learning_rate": 5.502770156009647e-05, "loss": 1.6528, "step": 7128 }, { "epoch": 0.74, "learning_rate": 5.502634430730907e-05, "loss": 1.3648, "step": 7129 }, { "epoch": 0.74, "learning_rate": 5.5024986886050176e-05, "loss": 0.9494, "step": 7130 }, { "epoch": 0.74, "learning_rate": 5.5023629296328926e-05, "loss": 0.9258, "step": 7131 }, { "epoch": 0.74, "learning_rate": 5.502227153815446e-05, "loss": 1.1563, "step": 7132 }, { "epoch": 0.74, "learning_rate": 5.502091361153591e-05, "loss": 0.6293, "step": 7133 }, { "epoch": 0.75, "learning_rate": 5.501955551648243e-05, "loss": 0.8518, "step": 7134 }, { "epoch": 0.75, "learning_rate": 5.5018197253003164e-05, "loss": 0.5558, "step": 7135 }, { "epoch": 0.75, "learning_rate": 5.501683882110723e-05, "loss": 1.6403, "step": 7136 }, { "epoch": 0.75, "learning_rate": 5.50154802208038e-05, "loss": 0.3144, "step": 7137 }, { "epoch": 0.75, "learning_rate": 5.5014121452102016e-05, "loss": 0.3089, "step": 7138 }, { "epoch": 0.75, "learning_rate": 5.501276251501101e-05, "loss": 0.4591, "step": 7139 }, { "epoch": 0.75, "learning_rate": 5.5011403409539944e-05, "loss": 1.2086, "step": 7140 }, { "epoch": 0.75, "learning_rate": 5.501004413569797e-05, "loss": 0.7284, "step": 7141 }, { "epoch": 0.75, "learning_rate": 5.5008684693494216e-05, "loss": 1.3956, "step": 7142 }, { "epoch": 0.75, "learning_rate": 5.5007325082937854e-05, "loss": 0.3474, "step": 7143 }, { "epoch": 0.75, "learning_rate": 5.500596530403804e-05, "loss": 0.9592, "step": 7144 }, { "epoch": 0.75, "learning_rate": 5.50046053568039e-05, "loss": 1.5955, "step": 7145 }, { "epoch": 0.75, "learning_rate": 5.500324524124462e-05, "loss": 1.406, "step": 7146 }, { "epoch": 0.75, "learning_rate": 5.5001884957369345e-05, "loss": 0.9779, "step": 7147 }, { "epoch": 0.75, "learning_rate": 5.500052450518722e-05, "loss": 1.0864, "step": 7148 }, { "epoch": 0.75, "learning_rate": 5.499916388470742e-05, "loss": 0.5895, "step": 7149 }, { "epoch": 0.75, "learning_rate": 5.49978030959391e-05, "loss": 1.524, "step": 7150 }, { "epoch": 0.75, "learning_rate": 5.499644213889141e-05, "loss": 1.6361, "step": 7151 }, { "epoch": 0.75, "learning_rate": 5.4995081013573526e-05, "loss": 0.5865, "step": 7152 }, { "epoch": 0.75, "learning_rate": 5.499371971999461e-05, "loss": 0.5071, "step": 7153 }, { "epoch": 0.75, "learning_rate": 5.499235825816382e-05, "loss": 0.6541, "step": 7154 }, { "epoch": 0.75, "learning_rate": 5.499099662809032e-05, "loss": 0.5044, "step": 7155 }, { "epoch": 0.75, "learning_rate": 5.4989634829783276e-05, "loss": 0.7931, "step": 7156 }, { "epoch": 0.75, "learning_rate": 5.498827286325185e-05, "loss": 1.3593, "step": 7157 }, { "epoch": 0.75, "learning_rate": 5.498691072850523e-05, "loss": 0.6029, "step": 7158 }, { "epoch": 0.75, "learning_rate": 5.4985548425552573e-05, "loss": 1.2673, "step": 7159 }, { "epoch": 0.75, "learning_rate": 5.498418595440304e-05, "loss": 0.5969, "step": 7160 }, { "epoch": 0.75, "learning_rate": 5.4982823315065825e-05, "loss": 1.2174, "step": 7161 }, { "epoch": 0.75, "learning_rate": 5.498146050755008e-05, "loss": 0.8981, "step": 7162 }, { "epoch": 0.75, "learning_rate": 5.4980097531864995e-05, "loss": 0.7862, "step": 7163 }, { "epoch": 0.75, "learning_rate": 5.497873438801973e-05, "loss": 0.9232, "step": 7164 }, { "epoch": 0.75, "learning_rate": 5.497737107602348e-05, "loss": 1.3609, "step": 7165 }, { "epoch": 0.75, "learning_rate": 5.497600759588541e-05, "loss": 0.9492, "step": 7166 }, { "epoch": 0.75, "learning_rate": 5.497464394761469e-05, "loss": 0.6184, "step": 7167 }, { "epoch": 0.75, "learning_rate": 5.497328013122052e-05, "loss": 0.3394, "step": 7168 }, { "epoch": 0.75, "learning_rate": 5.497191614671207e-05, "loss": 0.6994, "step": 7169 }, { "epoch": 0.75, "learning_rate": 5.497055199409852e-05, "loss": 1.7355, "step": 7170 }, { "epoch": 0.75, "learning_rate": 5.496918767338906e-05, "loss": 0.4654, "step": 7171 }, { "epoch": 0.75, "learning_rate": 5.496782318459287e-05, "loss": 1.2558, "step": 7172 }, { "epoch": 0.75, "learning_rate": 5.496645852771914e-05, "loss": 0.8964, "step": 7173 }, { "epoch": 0.75, "learning_rate": 5.496509370277705e-05, "loss": 0.7523, "step": 7174 }, { "epoch": 0.75, "learning_rate": 5.496372870977579e-05, "loss": 0.3107, "step": 7175 }, { "epoch": 0.75, "learning_rate": 5.4962363548724545e-05, "loss": 0.5136, "step": 7176 }, { "epoch": 0.75, "learning_rate": 5.496099821963251e-05, "loss": 0.341, "step": 7177 }, { "epoch": 0.75, "learning_rate": 5.495963272250888e-05, "loss": 1.5695, "step": 7178 }, { "epoch": 0.75, "learning_rate": 5.495826705736285e-05, "loss": 0.5547, "step": 7179 }, { "epoch": 0.75, "learning_rate": 5.4956901224203595e-05, "loss": 1.16, "step": 7180 }, { "epoch": 0.75, "learning_rate": 5.495553522304032e-05, "loss": 1.2899, "step": 7181 }, { "epoch": 0.75, "learning_rate": 5.4954169053882224e-05, "loss": 1.3225, "step": 7182 }, { "epoch": 0.75, "learning_rate": 5.49528027167385e-05, "loss": 1.8116, "step": 7183 }, { "epoch": 0.75, "learning_rate": 5.495143621161835e-05, "loss": 1.8043, "step": 7184 }, { "epoch": 0.75, "learning_rate": 5.4950069538530974e-05, "loss": 0.3809, "step": 7185 }, { "epoch": 0.75, "learning_rate": 5.494870269748556e-05, "loss": 0.6184, "step": 7186 }, { "epoch": 0.75, "learning_rate": 5.4947335688491314e-05, "loss": 2.2766, "step": 7187 }, { "epoch": 0.75, "learning_rate": 5.494596851155744e-05, "loss": 1.2174, "step": 7188 }, { "epoch": 0.75, "learning_rate": 5.494460116669315e-05, "loss": 1.1723, "step": 7189 }, { "epoch": 0.75, "learning_rate": 5.494323365390764e-05, "loss": 1.3724, "step": 7190 }, { "epoch": 0.75, "learning_rate": 5.494186597321012e-05, "loss": 0.6632, "step": 7191 }, { "epoch": 0.75, "learning_rate": 5.494049812460978e-05, "loss": 1.219, "step": 7192 }, { "epoch": 0.75, "learning_rate": 5.493913010811586e-05, "loss": 1.1004, "step": 7193 }, { "epoch": 0.75, "learning_rate": 5.493776192373755e-05, "loss": 1.1847, "step": 7194 }, { "epoch": 0.75, "learning_rate": 5.4936393571484055e-05, "loss": 1.1143, "step": 7195 }, { "epoch": 0.75, "learning_rate": 5.493502505136459e-05, "loss": 0.7117, "step": 7196 }, { "epoch": 0.75, "learning_rate": 5.493365636338838e-05, "loss": 0.3791, "step": 7197 }, { "epoch": 0.75, "learning_rate": 5.493228750756462e-05, "loss": 0.7245, "step": 7198 }, { "epoch": 0.75, "learning_rate": 5.493091848390254e-05, "loss": 0.3929, "step": 7199 }, { "epoch": 0.75, "learning_rate": 5.492954929241135e-05, "loss": 0.5069, "step": 7200 }, { "epoch": 0.75, "learning_rate": 5.4928179933100266e-05, "loss": 1.0948, "step": 7201 }, { "epoch": 0.75, "learning_rate": 5.492681040597851e-05, "loss": 1.0545, "step": 7202 }, { "epoch": 0.75, "learning_rate": 5.4925440711055296e-05, "loss": 1.2774, "step": 7203 }, { "epoch": 0.75, "learning_rate": 5.492407084833985e-05, "loss": 0.7657, "step": 7204 }, { "epoch": 0.75, "learning_rate": 5.492270081784139e-05, "loss": 0.6182, "step": 7205 }, { "epoch": 0.75, "learning_rate": 5.492133061956913e-05, "loss": 0.7131, "step": 7206 }, { "epoch": 0.75, "learning_rate": 5.491996025353231e-05, "loss": 0.5905, "step": 7207 }, { "epoch": 0.75, "learning_rate": 5.491858971974015e-05, "loss": 1.4884, "step": 7208 }, { "epoch": 0.75, "learning_rate": 5.491721901820188e-05, "loss": 1.452, "step": 7209 }, { "epoch": 0.75, "learning_rate": 5.491584814892671e-05, "loss": 1.1628, "step": 7210 }, { "epoch": 0.75, "learning_rate": 5.491447711192389e-05, "loss": 0.6576, "step": 7211 }, { "epoch": 0.75, "learning_rate": 5.491310590720263e-05, "loss": 1.2377, "step": 7212 }, { "epoch": 0.75, "learning_rate": 5.4911734534772175e-05, "loss": 0.4939, "step": 7213 }, { "epoch": 0.75, "learning_rate": 5.491036299464176e-05, "loss": 0.6298, "step": 7214 }, { "epoch": 0.75, "learning_rate": 5.49089912868206e-05, "loss": 1.2372, "step": 7215 }, { "epoch": 0.75, "learning_rate": 5.490761941131794e-05, "loss": 0.4436, "step": 7216 }, { "epoch": 0.75, "learning_rate": 5.490624736814302e-05, "loss": 0.6929, "step": 7217 }, { "epoch": 0.75, "learning_rate": 5.4904875157305065e-05, "loss": 1.3503, "step": 7218 }, { "epoch": 0.75, "learning_rate": 5.490350277881331e-05, "loss": 0.679, "step": 7219 }, { "epoch": 0.75, "learning_rate": 5.4902130232677016e-05, "loss": 0.4309, "step": 7220 }, { "epoch": 0.75, "learning_rate": 5.4900757518905404e-05, "loss": 0.9804, "step": 7221 }, { "epoch": 0.75, "learning_rate": 5.4899384637507714e-05, "loss": 1.5908, "step": 7222 }, { "epoch": 0.75, "learning_rate": 5.4898011588493195e-05, "loss": 1.284, "step": 7223 }, { "epoch": 0.75, "learning_rate": 5.489663837187109e-05, "loss": 0.4211, "step": 7224 }, { "epoch": 0.75, "learning_rate": 5.489526498765063e-05, "loss": 0.8064, "step": 7225 }, { "epoch": 0.75, "learning_rate": 5.489389143584109e-05, "loss": 0.6105, "step": 7226 }, { "epoch": 0.75, "learning_rate": 5.489251771645169e-05, "loss": 1.2162, "step": 7227 }, { "epoch": 0.75, "learning_rate": 5.489114382949167e-05, "loss": 0.3118, "step": 7228 }, { "epoch": 0.75, "learning_rate": 5.488976977497031e-05, "loss": 0.7394, "step": 7229 }, { "epoch": 0.76, "learning_rate": 5.488839555289684e-05, "loss": 0.9337, "step": 7230 }, { "epoch": 0.76, "learning_rate": 5.488702116328051e-05, "loss": 0.8574, "step": 7231 }, { "epoch": 0.76, "learning_rate": 5.4885646606130585e-05, "loss": 1.4781, "step": 7232 }, { "epoch": 0.76, "learning_rate": 5.488427188145631e-05, "loss": 1.0186, "step": 7233 }, { "epoch": 0.76, "learning_rate": 5.4882896989266935e-05, "loss": 0.611, "step": 7234 }, { "epoch": 0.76, "learning_rate": 5.488152192957172e-05, "loss": 0.2208, "step": 7235 }, { "epoch": 0.76, "learning_rate": 5.4880146702379916e-05, "loss": 1.3994, "step": 7236 }, { "epoch": 0.76, "learning_rate": 5.487877130770079e-05, "loss": 0.6013, "step": 7237 }, { "epoch": 0.76, "learning_rate": 5.4877395745543605e-05, "loss": 1.5346, "step": 7238 }, { "epoch": 0.76, "learning_rate": 5.4876020015917595e-05, "loss": 0.6621, "step": 7239 }, { "epoch": 0.76, "learning_rate": 5.487464411883205e-05, "loss": 2.0236, "step": 7240 }, { "epoch": 0.76, "learning_rate": 5.487326805429623e-05, "loss": 1.1604, "step": 7241 }, { "epoch": 0.76, "learning_rate": 5.487189182231937e-05, "loss": 0.7497, "step": 7242 }, { "epoch": 0.76, "learning_rate": 5.487051542291077e-05, "loss": 1.1723, "step": 7243 }, { "epoch": 0.76, "learning_rate": 5.486913885607967e-05, "loss": 0.4192, "step": 7244 }, { "epoch": 0.76, "learning_rate": 5.486776212183535e-05, "loss": 1.5541, "step": 7245 }, { "epoch": 0.76, "learning_rate": 5.486638522018707e-05, "loss": 0.8469, "step": 7246 }, { "epoch": 0.76, "learning_rate": 5.4865008151144104e-05, "loss": 0.5268, "step": 7247 }, { "epoch": 0.76, "learning_rate": 5.486363091471573e-05, "loss": 0.3654, "step": 7248 }, { "epoch": 0.76, "learning_rate": 5.4862253510911195e-05, "loss": 1.496, "step": 7249 }, { "epoch": 0.76, "learning_rate": 5.4860875939739795e-05, "loss": 0.4949, "step": 7250 }, { "epoch": 0.76, "learning_rate": 5.485949820121079e-05, "loss": 0.9103, "step": 7251 }, { "epoch": 0.76, "learning_rate": 5.485812029533347e-05, "loss": 0.8064, "step": 7252 }, { "epoch": 0.76, "learning_rate": 5.48567422221171e-05, "loss": 1.2013, "step": 7253 }, { "epoch": 0.76, "learning_rate": 5.485536398157095e-05, "loss": 0.7528, "step": 7254 }, { "epoch": 0.76, "learning_rate": 5.4853985573704305e-05, "loss": 0.4423, "step": 7255 }, { "epoch": 0.76, "learning_rate": 5.485260699852645e-05, "loss": 0.7114, "step": 7256 }, { "epoch": 0.76, "learning_rate": 5.485122825604665e-05, "loss": 1.6986, "step": 7257 }, { "epoch": 0.76, "learning_rate": 5.484984934627421e-05, "loss": 0.6773, "step": 7258 }, { "epoch": 0.76, "learning_rate": 5.484847026921839e-05, "loss": 0.8854, "step": 7259 }, { "epoch": 0.76, "learning_rate": 5.4847091024888485e-05, "loss": 0.605, "step": 7260 }, { "epoch": 0.76, "learning_rate": 5.484571161329377e-05, "loss": 0.9929, "step": 7261 }, { "epoch": 0.76, "learning_rate": 5.484433203444355e-05, "loss": 0.9841, "step": 7262 }, { "epoch": 0.76, "learning_rate": 5.48429522883471e-05, "loss": 0.3015, "step": 7263 }, { "epoch": 0.76, "learning_rate": 5.48415723750137e-05, "loss": 1.8962, "step": 7264 }, { "epoch": 0.76, "learning_rate": 5.4840192294452636e-05, "loss": 0.7383, "step": 7265 }, { "epoch": 0.76, "learning_rate": 5.483881204667323e-05, "loss": 0.4364, "step": 7266 }, { "epoch": 0.76, "learning_rate": 5.4837431631684744e-05, "loss": 0.2697, "step": 7267 }, { "epoch": 0.76, "learning_rate": 5.483605104949648e-05, "loss": 1.3341, "step": 7268 }, { "epoch": 0.76, "learning_rate": 5.483467030011773e-05, "loss": 0.986, "step": 7269 }, { "epoch": 0.76, "learning_rate": 5.4833289383557795e-05, "loss": 0.6735, "step": 7270 }, { "epoch": 0.76, "learning_rate": 5.483190829982596e-05, "loss": 1.3953, "step": 7271 }, { "epoch": 0.76, "learning_rate": 5.483052704893154e-05, "loss": 1.5728, "step": 7272 }, { "epoch": 0.76, "learning_rate": 5.482914563088381e-05, "loss": 0.7053, "step": 7273 }, { "epoch": 0.76, "learning_rate": 5.482776404569209e-05, "loss": 1.0185, "step": 7274 }, { "epoch": 0.76, "learning_rate": 5.482638229336567e-05, "loss": 0.8101, "step": 7275 }, { "epoch": 0.76, "learning_rate": 5.482500037391385e-05, "loss": 0.4847, "step": 7276 }, { "epoch": 0.76, "learning_rate": 5.4823618287345936e-05, "loss": 0.2372, "step": 7277 }, { "epoch": 0.76, "learning_rate": 5.482223603367124e-05, "loss": 0.9764, "step": 7278 }, { "epoch": 0.76, "learning_rate": 5.482085361289904e-05, "loss": 0.2141, "step": 7279 }, { "epoch": 0.76, "learning_rate": 5.481947102503867e-05, "loss": 1.0054, "step": 7280 }, { "epoch": 0.76, "learning_rate": 5.481808827009944e-05, "loss": 1.3235, "step": 7281 }, { "epoch": 0.76, "learning_rate": 5.4816705348090636e-05, "loss": 1.103, "step": 7282 }, { "epoch": 0.76, "learning_rate": 5.481532225902158e-05, "loss": 0.3696, "step": 7283 }, { "epoch": 0.76, "learning_rate": 5.481393900290158e-05, "loss": 1.0689, "step": 7284 }, { "epoch": 0.76, "learning_rate": 5.4812555579739945e-05, "loss": 0.9996, "step": 7285 }, { "epoch": 0.76, "learning_rate": 5.4811171989546e-05, "loss": 0.5532, "step": 7286 }, { "epoch": 0.76, "learning_rate": 5.4809788232329054e-05, "loss": 1.5369, "step": 7287 }, { "epoch": 0.76, "learning_rate": 5.48084043080984e-05, "loss": 1.8996, "step": 7288 }, { "epoch": 0.76, "learning_rate": 5.480702021686339e-05, "loss": 0.7268, "step": 7289 }, { "epoch": 0.76, "learning_rate": 5.4805635958633324e-05, "loss": 0.3869, "step": 7290 }, { "epoch": 0.76, "learning_rate": 5.480425153341751e-05, "loss": 0.7127, "step": 7291 }, { "epoch": 0.76, "learning_rate": 5.480286694122529e-05, "loss": 1.521, "step": 7292 }, { "epoch": 0.76, "learning_rate": 5.480148218206596e-05, "loss": 1.4782, "step": 7293 }, { "epoch": 0.76, "learning_rate": 5.480009725594887e-05, "loss": 1.0246, "step": 7294 }, { "epoch": 0.76, "learning_rate": 5.479871216288332e-05, "loss": 0.9917, "step": 7295 }, { "epoch": 0.76, "learning_rate": 5.479732690287864e-05, "loss": 1.2927, "step": 7296 }, { "epoch": 0.76, "learning_rate": 5.479594147594416e-05, "loss": 0.74, "step": 7297 }, { "epoch": 0.76, "learning_rate": 5.4794555882089204e-05, "loss": 1.4761, "step": 7298 }, { "epoch": 0.76, "learning_rate": 5.479317012132309e-05, "loss": 1.5183, "step": 7299 }, { "epoch": 0.76, "learning_rate": 5.4791784193655166e-05, "loss": 0.6624, "step": 7300 }, { "epoch": 0.76, "learning_rate": 5.479039809909476e-05, "loss": 0.906, "step": 7301 }, { "epoch": 0.76, "learning_rate": 5.478901183765117e-05, "loss": 1.335, "step": 7302 }, { "epoch": 0.76, "learning_rate": 5.478762540933377e-05, "loss": 0.7783, "step": 7303 }, { "epoch": 0.76, "learning_rate": 5.478623881415188e-05, "loss": 0.6538, "step": 7304 }, { "epoch": 0.76, "learning_rate": 5.478485205211482e-05, "loss": 0.553, "step": 7305 }, { "epoch": 0.76, "learning_rate": 5.478346512323194e-05, "loss": 0.5666, "step": 7306 }, { "epoch": 0.76, "learning_rate": 5.478207802751257e-05, "loss": 0.3987, "step": 7307 }, { "epoch": 0.76, "learning_rate": 5.478069076496604e-05, "loss": 0.3916, "step": 7308 }, { "epoch": 0.76, "learning_rate": 5.477930333560171e-05, "loss": 0.2431, "step": 7309 }, { "epoch": 0.76, "learning_rate": 5.477791573942891e-05, "loss": 1.0305, "step": 7310 }, { "epoch": 0.76, "learning_rate": 5.477652797645697e-05, "loss": 0.7551, "step": 7311 }, { "epoch": 0.76, "learning_rate": 5.4775140046695245e-05, "loss": 1.0158, "step": 7312 }, { "epoch": 0.76, "learning_rate": 5.4773751950153076e-05, "loss": 1.2526, "step": 7313 }, { "epoch": 0.76, "learning_rate": 5.47723636868398e-05, "loss": 1.1751, "step": 7314 }, { "epoch": 0.76, "learning_rate": 5.477097525676477e-05, "loss": 1.6693, "step": 7315 }, { "epoch": 0.76, "learning_rate": 5.4769586659937335e-05, "loss": 0.3819, "step": 7316 }, { "epoch": 0.76, "learning_rate": 5.476819789636683e-05, "loss": 0.6677, "step": 7317 }, { "epoch": 0.76, "learning_rate": 5.4766808966062626e-05, "loss": 0.3296, "step": 7318 }, { "epoch": 0.76, "learning_rate": 5.4765419869034046e-05, "loss": 1.1203, "step": 7319 }, { "epoch": 0.76, "learning_rate": 5.476403060529046e-05, "loss": 0.5036, "step": 7320 }, { "epoch": 0.76, "learning_rate": 5.476264117484121e-05, "loss": 0.7564, "step": 7321 }, { "epoch": 0.76, "learning_rate": 5.4761251577695646e-05, "loss": 1.2654, "step": 7322 }, { "epoch": 0.76, "learning_rate": 5.475986181386315e-05, "loss": 1.3051, "step": 7323 }, { "epoch": 0.76, "learning_rate": 5.475847188335304e-05, "loss": 1.5017, "step": 7324 }, { "epoch": 0.77, "learning_rate": 5.47570817861747e-05, "loss": 0.7739, "step": 7325 }, { "epoch": 0.77, "learning_rate": 5.4755691522337473e-05, "loss": 2.6829, "step": 7326 }, { "epoch": 0.77, "learning_rate": 5.475430109185072e-05, "loss": 1.1969, "step": 7327 }, { "epoch": 0.77, "learning_rate": 5.475291049472382e-05, "loss": 0.3259, "step": 7328 }, { "epoch": 0.77, "learning_rate": 5.475151973096611e-05, "loss": 0.5318, "step": 7329 }, { "epoch": 0.77, "learning_rate": 5.475012880058696e-05, "loss": 0.4311, "step": 7330 }, { "epoch": 0.77, "learning_rate": 5.4748737703595725e-05, "loss": 1.8947, "step": 7331 }, { "epoch": 0.77, "learning_rate": 5.474734644000179e-05, "loss": 0.529, "step": 7332 }, { "epoch": 0.77, "learning_rate": 5.47459550098145e-05, "loss": 1.2714, "step": 7333 }, { "epoch": 0.77, "learning_rate": 5.474456341304324e-05, "loss": 1.164, "step": 7334 }, { "epoch": 0.77, "learning_rate": 5.4743171649697366e-05, "loss": 0.2707, "step": 7335 }, { "epoch": 0.77, "learning_rate": 5.4741779719786245e-05, "loss": 0.9352, "step": 7336 }, { "epoch": 0.77, "learning_rate": 5.474038762331926e-05, "loss": 1.0157, "step": 7337 }, { "epoch": 0.77, "learning_rate": 5.473899536030577e-05, "loss": 1.4534, "step": 7338 }, { "epoch": 0.77, "learning_rate": 5.473760293075515e-05, "loss": 0.7282, "step": 7339 }, { "epoch": 0.77, "learning_rate": 5.473621033467678e-05, "loss": 1.1009, "step": 7340 }, { "epoch": 0.77, "learning_rate": 5.4734817572080025e-05, "loss": 1.5443, "step": 7341 }, { "epoch": 0.77, "learning_rate": 5.4733424642974266e-05, "loss": 1.3321, "step": 7342 }, { "epoch": 0.77, "learning_rate": 5.473203154736889e-05, "loss": 0.606, "step": 7343 }, { "epoch": 0.77, "learning_rate": 5.473063828527325e-05, "loss": 0.9276, "step": 7344 }, { "epoch": 0.77, "learning_rate": 5.4729244856696744e-05, "loss": 1.8007, "step": 7345 }, { "epoch": 0.77, "learning_rate": 5.472785126164875e-05, "loss": 1.23, "step": 7346 }, { "epoch": 0.77, "learning_rate": 5.472645750013865e-05, "loss": 0.7513, "step": 7347 }, { "epoch": 0.77, "learning_rate": 5.472506357217582e-05, "loss": 1.118, "step": 7348 }, { "epoch": 0.77, "learning_rate": 5.4723669477769647e-05, "loss": 0.9943, "step": 7349 }, { "epoch": 0.77, "learning_rate": 5.472227521692951e-05, "loss": 1.0213, "step": 7350 }, { "epoch": 0.77, "learning_rate": 5.472088078966481e-05, "loss": 1.5194, "step": 7351 }, { "epoch": 0.77, "learning_rate": 5.4719486195984917e-05, "loss": 1.4756, "step": 7352 }, { "epoch": 0.77, "learning_rate": 5.471809143589924e-05, "loss": 1.0781, "step": 7353 }, { "epoch": 0.77, "learning_rate": 5.471669650941714e-05, "loss": 1.227, "step": 7354 }, { "epoch": 0.77, "learning_rate": 5.471530141654803e-05, "loss": 1.7489, "step": 7355 }, { "epoch": 0.77, "learning_rate": 5.471390615730129e-05, "loss": 1.6887, "step": 7356 }, { "epoch": 0.77, "learning_rate": 5.471251073168632e-05, "loss": 1.2444, "step": 7357 }, { "epoch": 0.77, "learning_rate": 5.471111513971251e-05, "loss": 1.2008, "step": 7358 }, { "epoch": 0.77, "learning_rate": 5.470971938138925e-05, "loss": 1.128, "step": 7359 }, { "epoch": 0.77, "learning_rate": 5.470832345672595e-05, "loss": 1.4472, "step": 7360 }, { "epoch": 0.77, "learning_rate": 5.470692736573199e-05, "loss": 0.8561, "step": 7361 }, { "epoch": 0.77, "learning_rate": 5.470553110841678e-05, "loss": 0.4904, "step": 7362 }, { "epoch": 0.77, "learning_rate": 5.4704134684789705e-05, "loss": 1.8388, "step": 7363 }, { "epoch": 0.77, "learning_rate": 5.4702738094860187e-05, "loss": 1.3963, "step": 7364 }, { "epoch": 0.77, "learning_rate": 5.4701341338637616e-05, "loss": 0.5287, "step": 7365 }, { "epoch": 0.77, "learning_rate": 5.469994441613139e-05, "loss": 0.2937, "step": 7366 }, { "epoch": 0.77, "learning_rate": 5.469854732735092e-05, "loss": 0.4323, "step": 7367 }, { "epoch": 0.77, "learning_rate": 5.469715007230561e-05, "loss": 0.5024, "step": 7368 }, { "epoch": 0.77, "learning_rate": 5.469575265100486e-05, "loss": 1.1504, "step": 7369 }, { "epoch": 0.77, "learning_rate": 5.469435506345809e-05, "loss": 1.2713, "step": 7370 }, { "epoch": 0.77, "learning_rate": 5.469295730967469e-05, "loss": 1.338, "step": 7371 }, { "epoch": 0.77, "learning_rate": 5.4691559389664075e-05, "loss": 0.9626, "step": 7372 }, { "epoch": 0.77, "learning_rate": 5.4690161303435676e-05, "loss": 1.3513, "step": 7373 }, { "epoch": 0.77, "learning_rate": 5.4688763050998874e-05, "loss": 0.4766, "step": 7374 }, { "epoch": 0.77, "learning_rate": 5.46873646323631e-05, "loss": 1.5659, "step": 7375 }, { "epoch": 0.77, "learning_rate": 5.4685966047537764e-05, "loss": 0.9395, "step": 7376 }, { "epoch": 0.77, "learning_rate": 5.468456729653229e-05, "loss": 1.0276, "step": 7377 }, { "epoch": 0.77, "learning_rate": 5.468316837935606e-05, "loss": 0.753, "step": 7378 }, { "epoch": 0.77, "learning_rate": 5.468176929601854e-05, "loss": 0.7641, "step": 7379 }, { "epoch": 0.77, "learning_rate": 5.468037004652912e-05, "loss": 0.9077, "step": 7380 }, { "epoch": 0.77, "learning_rate": 5.467897063089722e-05, "loss": 1.0061, "step": 7381 }, { "epoch": 0.77, "learning_rate": 5.467757104913226e-05, "loss": 0.3753, "step": 7382 }, { "epoch": 0.77, "learning_rate": 5.4676171301243674e-05, "loss": 1.3206, "step": 7383 }, { "epoch": 0.77, "learning_rate": 5.467477138724087e-05, "loss": 0.3528, "step": 7384 }, { "epoch": 0.77, "learning_rate": 5.467337130713328e-05, "loss": 1.6005, "step": 7385 }, { "epoch": 0.77, "learning_rate": 5.467197106093033e-05, "loss": 0.2999, "step": 7386 }, { "epoch": 0.77, "learning_rate": 5.4670570648641435e-05, "loss": 0.5423, "step": 7387 }, { "epoch": 0.77, "learning_rate": 5.4669170070276046e-05, "loss": 0.3964, "step": 7388 }, { "epoch": 0.77, "learning_rate": 5.466776932584356e-05, "loss": 0.4471, "step": 7389 }, { "epoch": 0.77, "learning_rate": 5.4666368415353436e-05, "loss": 0.5415, "step": 7390 }, { "epoch": 0.77, "learning_rate": 5.466496733881508e-05, "loss": 0.915, "step": 7391 }, { "epoch": 0.77, "learning_rate": 5.466356609623795e-05, "loss": 0.616, "step": 7392 }, { "epoch": 0.77, "learning_rate": 5.466216468763145e-05, "loss": 1.9236, "step": 7393 }, { "epoch": 0.77, "learning_rate": 5.466076311300503e-05, "loss": 0.6846, "step": 7394 }, { "epoch": 0.77, "learning_rate": 5.465936137236812e-05, "loss": 0.4216, "step": 7395 }, { "epoch": 0.77, "learning_rate": 5.465795946573017e-05, "loss": 1.2679, "step": 7396 }, { "epoch": 0.77, "learning_rate": 5.46565573931006e-05, "loss": 1.5297, "step": 7397 }, { "epoch": 0.77, "learning_rate": 5.465515515448886e-05, "loss": 0.9452, "step": 7398 }, { "epoch": 0.77, "learning_rate": 5.465375274990438e-05, "loss": 1.211, "step": 7399 }, { "epoch": 0.77, "learning_rate": 5.46523501793566e-05, "loss": 0.9236, "step": 7400 }, { "epoch": 0.77, "learning_rate": 5.465094744285497e-05, "loss": 1.644, "step": 7401 }, { "epoch": 0.77, "learning_rate": 5.464954454040894e-05, "loss": 0.1665, "step": 7402 }, { "epoch": 0.77, "learning_rate": 5.464814147202792e-05, "loss": 0.721, "step": 7403 }, { "epoch": 0.77, "learning_rate": 5.46467382377214e-05, "loss": 0.318, "step": 7404 }, { "epoch": 0.77, "learning_rate": 5.46453348374988e-05, "loss": 1.305, "step": 7405 }, { "epoch": 0.77, "learning_rate": 5.4643931271369575e-05, "loss": 0.1873, "step": 7406 }, { "epoch": 0.77, "learning_rate": 5.464252753934316e-05, "loss": 1.5179, "step": 7407 }, { "epoch": 0.77, "learning_rate": 5.4641123641429025e-05, "loss": 1.035, "step": 7408 }, { "epoch": 0.77, "learning_rate": 5.463971957763661e-05, "loss": 0.6992, "step": 7409 }, { "epoch": 0.77, "learning_rate": 5.463831534797537e-05, "loss": 0.9324, "step": 7410 }, { "epoch": 0.77, "learning_rate": 5.4636910952454755e-05, "loss": 1.0461, "step": 7411 }, { "epoch": 0.77, "learning_rate": 5.463550639108421e-05, "loss": 0.5497, "step": 7412 }, { "epoch": 0.77, "learning_rate": 5.4634101663873215e-05, "loss": 0.3842, "step": 7413 }, { "epoch": 0.77, "learning_rate": 5.46326967708312e-05, "loss": 1.0629, "step": 7414 }, { "epoch": 0.77, "learning_rate": 5.463129171196764e-05, "loss": 0.8813, "step": 7415 }, { "epoch": 0.77, "learning_rate": 5.462988648729199e-05, "loss": 1.0093, "step": 7416 }, { "epoch": 0.77, "learning_rate": 5.46284810968137e-05, "loss": 0.8658, "step": 7417 }, { "epoch": 0.77, "learning_rate": 5.462707554054224e-05, "loss": 0.4923, "step": 7418 }, { "epoch": 0.77, "learning_rate": 5.462566981848707e-05, "loss": 0.2756, "step": 7419 }, { "epoch": 0.77, "learning_rate": 5.4624263930657654e-05, "loss": 1.5059, "step": 7420 }, { "epoch": 0.78, "learning_rate": 5.4622857877063454e-05, "loss": 0.6471, "step": 7421 }, { "epoch": 0.78, "learning_rate": 5.462145165771394e-05, "loss": 1.3311, "step": 7422 }, { "epoch": 0.78, "learning_rate": 5.462004527261857e-05, "loss": 0.3398, "step": 7423 }, { "epoch": 0.78, "learning_rate": 5.461863872178682e-05, "loss": 1.1072, "step": 7424 }, { "epoch": 0.78, "learning_rate": 5.461723200522815e-05, "loss": 1.602, "step": 7425 }, { "epoch": 0.78, "learning_rate": 5.461582512295203e-05, "loss": 1.0842, "step": 7426 }, { "epoch": 0.78, "learning_rate": 5.461441807496794e-05, "loss": 1.4618, "step": 7427 }, { "epoch": 0.78, "learning_rate": 5.461301086128535e-05, "loss": 1.266, "step": 7428 }, { "epoch": 0.78, "learning_rate": 5.461160348191372e-05, "loss": 1.6906, "step": 7429 }, { "epoch": 0.78, "learning_rate": 5.461019593686254e-05, "loss": 0.8836, "step": 7430 }, { "epoch": 0.78, "learning_rate": 5.460878822614129e-05, "loss": 1.2393, "step": 7431 }, { "epoch": 0.78, "learning_rate": 5.460738034975943e-05, "loss": 0.6109, "step": 7432 }, { "epoch": 0.78, "learning_rate": 5.460597230772643e-05, "loss": 0.476, "step": 7433 }, { "epoch": 0.78, "learning_rate": 5.4604564100051786e-05, "loss": 1.4037, "step": 7434 }, { "epoch": 0.78, "learning_rate": 5.460315572674498e-05, "loss": 1.0031, "step": 7435 }, { "epoch": 0.78, "learning_rate": 5.4601747187815486e-05, "loss": 1.7483, "step": 7436 }, { "epoch": 0.78, "learning_rate": 5.460033848327279e-05, "loss": 0.8807, "step": 7437 }, { "epoch": 0.78, "learning_rate": 5.459892961312636e-05, "loss": 0.2652, "step": 7438 }, { "epoch": 0.78, "learning_rate": 5.4597520577385695e-05, "loss": 1.0913, "step": 7439 }, { "epoch": 0.78, "learning_rate": 5.459611137606029e-05, "loss": 0.5554, "step": 7440 }, { "epoch": 0.78, "learning_rate": 5.45947020091596e-05, "loss": 1.6254, "step": 7441 }, { "epoch": 0.78, "learning_rate": 5.459329247669314e-05, "loss": 0.9314, "step": 7442 }, { "epoch": 0.78, "learning_rate": 5.459188277867039e-05, "loss": 0.2582, "step": 7443 }, { "epoch": 0.78, "learning_rate": 5.4590472915100835e-05, "loss": 0.877, "step": 7444 }, { "epoch": 0.78, "learning_rate": 5.458906288599397e-05, "loss": 0.8429, "step": 7445 }, { "epoch": 0.78, "learning_rate": 5.458765269135929e-05, "loss": 0.9128, "step": 7446 }, { "epoch": 0.78, "learning_rate": 5.4586242331206285e-05, "loss": 1.3008, "step": 7447 }, { "epoch": 0.78, "learning_rate": 5.458483180554445e-05, "loss": 0.691, "step": 7448 }, { "epoch": 0.78, "learning_rate": 5.4583421114383276e-05, "loss": 0.6323, "step": 7449 }, { "epoch": 0.78, "learning_rate": 5.4582010257732266e-05, "loss": 1.0536, "step": 7450 }, { "epoch": 0.78, "learning_rate": 5.4580599235600914e-05, "loss": 0.8825, "step": 7451 }, { "epoch": 0.78, "learning_rate": 5.457918804799872e-05, "loss": 0.6058, "step": 7452 }, { "epoch": 0.78, "learning_rate": 5.4577776694935184e-05, "loss": 0.3302, "step": 7453 }, { "epoch": 0.78, "learning_rate": 5.4576365176419804e-05, "loss": 0.2188, "step": 7454 }, { "epoch": 0.78, "learning_rate": 5.4574953492462086e-05, "loss": 0.3238, "step": 7455 }, { "epoch": 0.78, "learning_rate": 5.457354164307153e-05, "loss": 0.6233, "step": 7456 }, { "epoch": 0.78, "learning_rate": 5.457212962825764e-05, "loss": 0.7704, "step": 7457 }, { "epoch": 0.78, "learning_rate": 5.457071744802992e-05, "loss": 0.382, "step": 7458 }, { "epoch": 0.78, "learning_rate": 5.456930510239789e-05, "loss": 0.8186, "step": 7459 }, { "epoch": 0.78, "learning_rate": 5.4567892591371036e-05, "loss": 1.4105, "step": 7460 }, { "epoch": 0.78, "learning_rate": 5.456647991495888e-05, "loss": 0.4886, "step": 7461 }, { "epoch": 0.78, "learning_rate": 5.4565067073170935e-05, "loss": 0.9541, "step": 7462 }, { "epoch": 0.78, "learning_rate": 5.45636540660167e-05, "loss": 0.8384, "step": 7463 }, { "epoch": 0.78, "learning_rate": 5.456224089350569e-05, "loss": 0.933, "step": 7464 }, { "epoch": 0.78, "learning_rate": 5.456082755564743e-05, "loss": 0.8906, "step": 7465 }, { "epoch": 0.78, "learning_rate": 5.455941405245143e-05, "loss": 0.5082, "step": 7466 }, { "epoch": 0.78, "learning_rate": 5.4558000383927185e-05, "loss": 1.1562, "step": 7467 }, { "epoch": 0.78, "learning_rate": 5.455658655008424e-05, "loss": 0.9696, "step": 7468 }, { "epoch": 0.78, "learning_rate": 5.4555172550932095e-05, "loss": 1.5189, "step": 7469 }, { "epoch": 0.78, "learning_rate": 5.4553758386480274e-05, "loss": 0.5034, "step": 7470 }, { "epoch": 0.78, "learning_rate": 5.45523440567383e-05, "loss": 1.5348, "step": 7471 }, { "epoch": 0.78, "learning_rate": 5.4550929561715685e-05, "loss": 1.2486, "step": 7472 }, { "epoch": 0.78, "learning_rate": 5.4549514901421965e-05, "loss": 0.2292, "step": 7473 }, { "epoch": 0.78, "learning_rate": 5.454810007586665e-05, "loss": 0.7005, "step": 7474 }, { "epoch": 0.78, "learning_rate": 5.454668508505926e-05, "loss": 0.1223, "step": 7475 }, { "epoch": 0.78, "learning_rate": 5.454526992900934e-05, "loss": 1.4652, "step": 7476 }, { "epoch": 0.78, "learning_rate": 5.4543854607726406e-05, "loss": 0.9054, "step": 7477 }, { "epoch": 0.78, "learning_rate": 5.454243912121999e-05, "loss": 1.3168, "step": 7478 }, { "epoch": 0.78, "learning_rate": 5.454102346949961e-05, "loss": 1.2481, "step": 7479 }, { "epoch": 0.78, "learning_rate": 5.4539607652574804e-05, "loss": 0.901, "step": 7480 }, { "epoch": 0.78, "learning_rate": 5.45381916704551e-05, "loss": 1.6812, "step": 7481 }, { "epoch": 0.78, "learning_rate": 5.453677552315002e-05, "loss": 0.5492, "step": 7482 }, { "epoch": 0.78, "learning_rate": 5.453535921066912e-05, "loss": 0.5427, "step": 7483 }, { "epoch": 0.78, "learning_rate": 5.4533942733021924e-05, "loss": 0.8225, "step": 7484 }, { "epoch": 0.78, "learning_rate": 5.4532526090217965e-05, "loss": 1.3907, "step": 7485 }, { "epoch": 0.78, "learning_rate": 5.453110928226678e-05, "loss": 0.9122, "step": 7486 }, { "epoch": 0.78, "learning_rate": 5.45296923091779e-05, "loss": 0.249, "step": 7487 }, { "epoch": 0.78, "learning_rate": 5.452827517096088e-05, "loss": 0.4863, "step": 7488 }, { "epoch": 0.78, "learning_rate": 5.452685786762525e-05, "loss": 0.9955, "step": 7489 }, { "epoch": 0.78, "learning_rate": 5.452544039918055e-05, "loss": 1.4758, "step": 7490 }, { "epoch": 0.78, "learning_rate": 5.452402276563632e-05, "loss": 0.9722, "step": 7491 }, { "epoch": 0.78, "learning_rate": 5.452260496700211e-05, "loss": 0.9235, "step": 7492 }, { "epoch": 0.78, "learning_rate": 5.4521187003287464e-05, "loss": 0.3476, "step": 7493 }, { "epoch": 0.78, "learning_rate": 5.451976887450192e-05, "loss": 0.586, "step": 7494 }, { "epoch": 0.78, "learning_rate": 5.451835058065503e-05, "loss": 0.9228, "step": 7495 }, { "epoch": 0.78, "learning_rate": 5.451693212175634e-05, "loss": 1.0105, "step": 7496 }, { "epoch": 0.78, "learning_rate": 5.4515513497815404e-05, "loss": 1.331, "step": 7497 }, { "epoch": 0.78, "learning_rate": 5.451409470884176e-05, "loss": 1.1384, "step": 7498 }, { "epoch": 0.78, "learning_rate": 5.4512675754844974e-05, "loss": 1.1058, "step": 7499 }, { "epoch": 0.78, "learning_rate": 5.4511256635834586e-05, "loss": 1.2858, "step": 7500 }, { "epoch": 0.78, "learning_rate": 5.450983735182015e-05, "loss": 0.3759, "step": 7501 }, { "epoch": 0.78, "learning_rate": 5.450841790281123e-05, "loss": 1.156, "step": 7502 }, { "epoch": 0.78, "learning_rate": 5.450699828881738e-05, "loss": 1.2901, "step": 7503 }, { "epoch": 0.78, "learning_rate": 5.450557850984815e-05, "loss": 1.0522, "step": 7504 }, { "epoch": 0.78, "learning_rate": 5.4504158565913084e-05, "loss": 0.3546, "step": 7505 }, { "epoch": 0.78, "learning_rate": 5.450273845702176e-05, "loss": 1.1103, "step": 7506 }, { "epoch": 0.78, "learning_rate": 5.4501318183183745e-05, "loss": 1.5007, "step": 7507 }, { "epoch": 0.78, "learning_rate": 5.4499897744408585e-05, "loss": 0.2858, "step": 7508 }, { "epoch": 0.78, "learning_rate": 5.4498477140705844e-05, "loss": 1.69, "step": 7509 }, { "epoch": 0.78, "learning_rate": 5.44970563720851e-05, "loss": 1.4142, "step": 7510 }, { "epoch": 0.78, "learning_rate": 5.449563543855588e-05, "loss": 1.108, "step": 7511 }, { "epoch": 0.78, "learning_rate": 5.449421434012779e-05, "loss": 1.2764, "step": 7512 }, { "epoch": 0.78, "learning_rate": 5.449279307681038e-05, "loss": 1.2524, "step": 7513 }, { "epoch": 0.78, "learning_rate": 5.449137164861321e-05, "loss": 0.8202, "step": 7514 }, { "epoch": 0.78, "learning_rate": 5.4489950055545864e-05, "loss": 0.2752, "step": 7515 }, { "epoch": 0.78, "learning_rate": 5.44885282976179e-05, "loss": 0.3764, "step": 7516 }, { "epoch": 0.79, "learning_rate": 5.448710637483889e-05, "loss": 0.9951, "step": 7517 }, { "epoch": 0.79, "learning_rate": 5.448568428721842e-05, "loss": 1.2021, "step": 7518 }, { "epoch": 0.79, "learning_rate": 5.448426203476604e-05, "loss": 0.6133, "step": 7519 }, { "epoch": 0.79, "learning_rate": 5.448283961749135e-05, "loss": 1.4749, "step": 7520 }, { "epoch": 0.79, "learning_rate": 5.448141703540391e-05, "loss": 0.8681, "step": 7521 }, { "epoch": 0.79, "learning_rate": 5.447999428851328e-05, "loss": 0.8744, "step": 7522 }, { "epoch": 0.79, "learning_rate": 5.447857137682908e-05, "loss": 1.0931, "step": 7523 }, { "epoch": 0.79, "learning_rate": 5.447714830036086e-05, "loss": 0.2686, "step": 7524 }, { "epoch": 0.79, "learning_rate": 5.4475725059118196e-05, "loss": 1.348, "step": 7525 }, { "epoch": 0.79, "learning_rate": 5.447430165311069e-05, "loss": 1.1803, "step": 7526 }, { "epoch": 0.79, "learning_rate": 5.44728780823479e-05, "loss": 0.3315, "step": 7527 }, { "epoch": 0.79, "learning_rate": 5.447145434683942e-05, "loss": 0.3888, "step": 7528 }, { "epoch": 0.79, "learning_rate": 5.447003044659484e-05, "loss": 1.4105, "step": 7529 }, { "epoch": 0.79, "learning_rate": 5.446860638162375e-05, "loss": 1.1922, "step": 7530 }, { "epoch": 0.79, "learning_rate": 5.446718215193571e-05, "loss": 0.7062, "step": 7531 }, { "epoch": 0.79, "learning_rate": 5.446575775754033e-05, "loss": 0.3641, "step": 7532 }, { "epoch": 0.79, "learning_rate": 5.4464333198447205e-05, "loss": 1.0845, "step": 7533 }, { "epoch": 0.79, "learning_rate": 5.44629084746659e-05, "loss": 1.0116, "step": 7534 }, { "epoch": 0.79, "learning_rate": 5.4461483586206024e-05, "loss": 0.9171, "step": 7535 }, { "epoch": 0.79, "learning_rate": 5.446005853307716e-05, "loss": 0.752, "step": 7536 }, { "epoch": 0.79, "learning_rate": 5.445863331528891e-05, "loss": 0.1474, "step": 7537 }, { "epoch": 0.79, "learning_rate": 5.4457207932850866e-05, "loss": 0.3848, "step": 7538 }, { "epoch": 0.79, "learning_rate": 5.4455782385772616e-05, "loss": 1.0479, "step": 7539 }, { "epoch": 0.79, "learning_rate": 5.445435667406376e-05, "loss": 0.3313, "step": 7540 }, { "epoch": 0.79, "learning_rate": 5.445293079773389e-05, "loss": 0.8726, "step": 7541 }, { "epoch": 0.79, "learning_rate": 5.445150475679262e-05, "loss": 1.0094, "step": 7542 }, { "epoch": 0.79, "learning_rate": 5.4450078551249546e-05, "loss": 1.0467, "step": 7543 }, { "epoch": 0.79, "learning_rate": 5.444865218111426e-05, "loss": 0.7839, "step": 7544 }, { "epoch": 0.79, "learning_rate": 5.444722564639636e-05, "loss": 1.33, "step": 7545 }, { "epoch": 0.79, "learning_rate": 5.4445798947105466e-05, "loss": 0.7035, "step": 7546 }, { "epoch": 0.79, "learning_rate": 5.4444372083251175e-05, "loss": 0.2972, "step": 7547 }, { "epoch": 0.79, "learning_rate": 5.444294505484309e-05, "loss": 1.0741, "step": 7548 }, { "epoch": 0.79, "learning_rate": 5.444151786189082e-05, "loss": 0.3717, "step": 7549 }, { "epoch": 0.79, "learning_rate": 5.444009050440396e-05, "loss": 0.862, "step": 7550 }, { "epoch": 0.79, "learning_rate": 5.443866298239213e-05, "loss": 0.8143, "step": 7551 }, { "epoch": 0.79, "learning_rate": 5.4437235295864946e-05, "loss": 1.2387, "step": 7552 }, { "epoch": 0.79, "learning_rate": 5.443580744483201e-05, "loss": 0.1281, "step": 7553 }, { "epoch": 0.79, "learning_rate": 5.443437942930294e-05, "loss": 1.3297, "step": 7554 }, { "epoch": 0.79, "learning_rate": 5.443295124928734e-05, "loss": 0.902, "step": 7555 }, { "epoch": 0.79, "learning_rate": 5.443152290479482e-05, "loss": 1.1509, "step": 7556 }, { "epoch": 0.79, "learning_rate": 5.443009439583502e-05, "loss": 0.323, "step": 7557 }, { "epoch": 0.79, "learning_rate": 5.4428665722417535e-05, "loss": 1.1135, "step": 7558 }, { "epoch": 0.79, "learning_rate": 5.442723688455199e-05, "loss": 0.5087, "step": 7559 }, { "epoch": 0.79, "learning_rate": 5.4425807882248e-05, "loss": 0.6969, "step": 7560 }, { "epoch": 0.79, "learning_rate": 5.4424378715515185e-05, "loss": 0.626, "step": 7561 }, { "epoch": 0.79, "learning_rate": 5.4422949384363166e-05, "loss": 1.3386, "step": 7562 }, { "epoch": 0.79, "learning_rate": 5.4421519888801564e-05, "loss": 1.0268, "step": 7563 }, { "epoch": 0.79, "learning_rate": 5.4420090228840015e-05, "loss": 1.2321, "step": 7564 }, { "epoch": 0.79, "learning_rate": 5.441866040448813e-05, "loss": 0.8556, "step": 7565 }, { "epoch": 0.79, "learning_rate": 5.4417230415755524e-05, "loss": 0.479, "step": 7566 }, { "epoch": 0.79, "learning_rate": 5.441580026265185e-05, "loss": 1.2226, "step": 7567 }, { "epoch": 0.79, "learning_rate": 5.441436994518671e-05, "loss": 0.7285, "step": 7568 }, { "epoch": 0.79, "learning_rate": 5.4412939463369755e-05, "loss": 0.2798, "step": 7569 }, { "epoch": 0.79, "learning_rate": 5.4411508817210605e-05, "loss": 0.9425, "step": 7570 }, { "epoch": 0.79, "learning_rate": 5.4410078006718885e-05, "loss": 0.5196, "step": 7571 }, { "epoch": 0.79, "learning_rate": 5.440864703190423e-05, "loss": 0.5288, "step": 7572 }, { "epoch": 0.79, "learning_rate": 5.440721589277627e-05, "loss": 0.2477, "step": 7573 }, { "epoch": 0.79, "learning_rate": 5.440578458934465e-05, "loss": 1.5109, "step": 7574 }, { "epoch": 0.79, "learning_rate": 5.440435312161899e-05, "loss": 0.2864, "step": 7575 }, { "epoch": 0.79, "learning_rate": 5.440292148960894e-05, "loss": 0.7587, "step": 7576 }, { "epoch": 0.79, "learning_rate": 5.440148969332414e-05, "loss": 0.6429, "step": 7577 }, { "epoch": 0.79, "learning_rate": 5.440005773277422e-05, "loss": 0.7221, "step": 7578 }, { "epoch": 0.79, "learning_rate": 5.439862560796881e-05, "loss": 0.7607, "step": 7579 }, { "epoch": 0.79, "learning_rate": 5.439719331891756e-05, "loss": 0.4457, "step": 7580 }, { "epoch": 0.79, "learning_rate": 5.439576086563013e-05, "loss": 0.38, "step": 7581 }, { "epoch": 0.79, "learning_rate": 5.4394328248116135e-05, "loss": 0.2514, "step": 7582 }, { "epoch": 0.79, "learning_rate": 5.439289546638523e-05, "loss": 1.7328, "step": 7583 }, { "epoch": 0.79, "learning_rate": 5.439146252044706e-05, "loss": 1.7079, "step": 7584 }, { "epoch": 0.79, "learning_rate": 5.4390029410311276e-05, "loss": 1.6227, "step": 7585 }, { "epoch": 0.79, "learning_rate": 5.438859613598751e-05, "loss": 0.9831, "step": 7586 }, { "epoch": 0.79, "learning_rate": 5.438716269748543e-05, "loss": 1.2388, "step": 7587 }, { "epoch": 0.79, "learning_rate": 5.4385729094814675e-05, "loss": 0.6633, "step": 7588 }, { "epoch": 0.79, "learning_rate": 5.43842953279849e-05, "loss": 1.658, "step": 7589 }, { "epoch": 0.79, "learning_rate": 5.438286139700575e-05, "loss": 0.3705, "step": 7590 }, { "epoch": 0.79, "learning_rate": 5.438142730188689e-05, "loss": 0.4919, "step": 7591 }, { "epoch": 0.79, "learning_rate": 5.4379993042637963e-05, "loss": 1.1413, "step": 7592 }, { "epoch": 0.79, "learning_rate": 5.437855861926863e-05, "loss": 1.2336, "step": 7593 }, { "epoch": 0.79, "learning_rate": 5.437712403178853e-05, "loss": 1.2905, "step": 7594 }, { "epoch": 0.79, "learning_rate": 5.437568928020734e-05, "loss": 0.9319, "step": 7595 }, { "epoch": 0.79, "learning_rate": 5.4374254364534716e-05, "loss": 1.0897, "step": 7596 }, { "epoch": 0.79, "learning_rate": 5.437281928478032e-05, "loss": 0.3774, "step": 7597 }, { "epoch": 0.79, "learning_rate": 5.43713840409538e-05, "loss": 1.1438, "step": 7598 }, { "epoch": 0.79, "learning_rate": 5.436994863306483e-05, "loss": 1.267, "step": 7599 }, { "epoch": 0.79, "learning_rate": 5.436851306112307e-05, "loss": 0.965, "step": 7600 }, { "epoch": 0.79, "learning_rate": 5.436707732513818e-05, "loss": 0.5769, "step": 7601 }, { "epoch": 0.79, "learning_rate": 5.436564142511982e-05, "loss": 1.0487, "step": 7602 }, { "epoch": 0.79, "learning_rate": 5.4364205361077674e-05, "loss": 0.8278, "step": 7603 }, { "epoch": 0.79, "learning_rate": 5.436276913302139e-05, "loss": 0.2878, "step": 7604 }, { "epoch": 0.79, "learning_rate": 5.436133274096064e-05, "loss": 1.3252, "step": 7605 }, { "epoch": 0.79, "learning_rate": 5.435989618490511e-05, "loss": 1.4229, "step": 7606 }, { "epoch": 0.79, "learning_rate": 5.435845946486446e-05, "loss": 0.7717, "step": 7607 }, { "epoch": 0.79, "learning_rate": 5.4357022580848346e-05, "loss": 0.93, "step": 7608 }, { "epoch": 0.79, "learning_rate": 5.4355585532866466e-05, "loss": 0.201, "step": 7609 }, { "epoch": 0.79, "learning_rate": 5.4354148320928485e-05, "loss": 0.9358, "step": 7610 }, { "epoch": 0.79, "learning_rate": 5.4352710945044064e-05, "loss": 0.5867, "step": 7611 }, { "epoch": 0.79, "learning_rate": 5.435127340522291e-05, "loss": 0.337, "step": 7612 }, { "epoch": 0.8, "learning_rate": 5.434983570147466e-05, "loss": 1.0625, "step": 7613 }, { "epoch": 0.8, "learning_rate": 5.434839783380902e-05, "loss": 1.3172, "step": 7614 }, { "epoch": 0.8, "learning_rate": 5.4346959802235666e-05, "loss": 1.4795, "step": 7615 }, { "epoch": 0.8, "learning_rate": 5.434552160676428e-05, "loss": 0.7333, "step": 7616 }, { "epoch": 0.8, "learning_rate": 5.434408324740453e-05, "loss": 0.4697, "step": 7617 }, { "epoch": 0.8, "learning_rate": 5.434264472416611e-05, "loss": 1.2988, "step": 7618 }, { "epoch": 0.8, "learning_rate": 5.4341206037058705e-05, "loss": 0.8069, "step": 7619 }, { "epoch": 0.8, "learning_rate": 5.4339767186091994e-05, "loss": 1.1967, "step": 7620 }, { "epoch": 0.8, "learning_rate": 5.4338328171275664e-05, "loss": 0.7472, "step": 7621 }, { "epoch": 0.8, "learning_rate": 5.433688899261941e-05, "loss": 0.7972, "step": 7622 }, { "epoch": 0.8, "learning_rate": 5.4335449650132904e-05, "loss": 0.494, "step": 7623 }, { "epoch": 0.8, "learning_rate": 5.4334010143825855e-05, "loss": 2.0695, "step": 7624 }, { "epoch": 0.8, "learning_rate": 5.433257047370794e-05, "loss": 0.3635, "step": 7625 }, { "epoch": 0.8, "learning_rate": 5.433113063978885e-05, "loss": 1.1188, "step": 7626 }, { "epoch": 0.8, "learning_rate": 5.432969064207828e-05, "loss": 1.3214, "step": 7627 }, { "epoch": 0.8, "learning_rate": 5.432825048058593e-05, "loss": 0.74, "step": 7628 }, { "epoch": 0.8, "learning_rate": 5.432681015532149e-05, "loss": 0.9433, "step": 7629 }, { "epoch": 0.8, "learning_rate": 5.432536966629466e-05, "loss": 1.4469, "step": 7630 }, { "epoch": 0.8, "learning_rate": 5.4323929013515125e-05, "loss": 1.2416, "step": 7631 }, { "epoch": 0.8, "learning_rate": 5.432248819699259e-05, "loss": 0.9877, "step": 7632 }, { "epoch": 0.8, "learning_rate": 5.4321047216736763e-05, "loss": 0.8708, "step": 7633 }, { "epoch": 0.8, "learning_rate": 5.4319606072757336e-05, "loss": 0.4876, "step": 7634 }, { "epoch": 0.8, "learning_rate": 5.431816476506401e-05, "loss": 1.5531, "step": 7635 }, { "epoch": 0.8, "learning_rate": 5.431672329366648e-05, "loss": 0.907, "step": 7636 }, { "epoch": 0.8, "learning_rate": 5.431528165857447e-05, "loss": 0.5104, "step": 7637 }, { "epoch": 0.8, "learning_rate": 5.431383985979767e-05, "loss": 0.5688, "step": 7638 }, { "epoch": 0.8, "learning_rate": 5.4312397897345787e-05, "loss": 0.764, "step": 7639 }, { "epoch": 0.8, "learning_rate": 5.431095577122854e-05, "loss": 1.1629, "step": 7640 }, { "epoch": 0.8, "learning_rate": 5.430951348145561e-05, "loss": 1.5069, "step": 7641 }, { "epoch": 0.8, "learning_rate": 5.430807102803673e-05, "loss": 0.3366, "step": 7642 }, { "epoch": 0.8, "learning_rate": 5.430662841098161e-05, "loss": 1.0064, "step": 7643 }, { "epoch": 0.8, "learning_rate": 5.4305185630299946e-05, "loss": 0.2546, "step": 7644 }, { "epoch": 0.8, "learning_rate": 5.430374268600147e-05, "loss": 0.7062, "step": 7645 }, { "epoch": 0.8, "learning_rate": 5.4302299578095874e-05, "loss": 1.1577, "step": 7646 }, { "epoch": 0.8, "learning_rate": 5.430085630659289e-05, "loss": 0.7007, "step": 7647 }, { "epoch": 0.8, "learning_rate": 5.429941287150222e-05, "loss": 0.5861, "step": 7648 }, { "epoch": 0.8, "learning_rate": 5.429796927283359e-05, "loss": 1.1199, "step": 7649 }, { "epoch": 0.8, "learning_rate": 5.4296525510596725e-05, "loss": 1.0125, "step": 7650 }, { "epoch": 0.8, "learning_rate": 5.429508158480133e-05, "loss": 1.2669, "step": 7651 }, { "epoch": 0.8, "learning_rate": 5.429363749545713e-05, "loss": 0.4293, "step": 7652 }, { "epoch": 0.8, "learning_rate": 5.429219324257385e-05, "loss": 0.862, "step": 7653 }, { "epoch": 0.8, "learning_rate": 5.4290748826161205e-05, "loss": 0.576, "step": 7654 }, { "epoch": 0.8, "learning_rate": 5.428930424622891e-05, "loss": 1.499, "step": 7655 }, { "epoch": 0.8, "learning_rate": 5.428785950278673e-05, "loss": 1.3335, "step": 7656 }, { "epoch": 0.8, "learning_rate": 5.428641459584434e-05, "loss": 0.4734, "step": 7657 }, { "epoch": 0.8, "learning_rate": 5.4284969525411496e-05, "loss": 0.6098, "step": 7658 }, { "epoch": 0.8, "learning_rate": 5.428352429149793e-05, "loss": 0.338, "step": 7659 }, { "epoch": 0.8, "learning_rate": 5.428207889411335e-05, "loss": 0.7442, "step": 7660 }, { "epoch": 0.8, "learning_rate": 5.42806333332675e-05, "loss": 1.0603, "step": 7661 }, { "epoch": 0.8, "learning_rate": 5.42791876089701e-05, "loss": 0.6395, "step": 7662 }, { "epoch": 0.8, "learning_rate": 5.427774172123089e-05, "loss": 0.3841, "step": 7663 }, { "epoch": 0.8, "learning_rate": 5.4276295670059616e-05, "loss": 1.4295, "step": 7664 }, { "epoch": 0.8, "learning_rate": 5.4274849455465985e-05, "loss": 0.3502, "step": 7665 }, { "epoch": 0.8, "learning_rate": 5.4273403077459756e-05, "loss": 0.6099, "step": 7666 }, { "epoch": 0.8, "learning_rate": 5.427195653605066e-05, "loss": 1.0094, "step": 7667 }, { "epoch": 0.8, "learning_rate": 5.427050983124843e-05, "loss": 0.3739, "step": 7668 }, { "epoch": 0.8, "learning_rate": 5.42690629630628e-05, "loss": 0.2595, "step": 7669 }, { "epoch": 0.8, "learning_rate": 5.426761593150352e-05, "loss": 0.8436, "step": 7670 }, { "epoch": 0.8, "learning_rate": 5.426616873658033e-05, "loss": 0.3979, "step": 7671 }, { "epoch": 0.8, "learning_rate": 5.4264721378302966e-05, "loss": 0.8299, "step": 7672 }, { "epoch": 0.8, "learning_rate": 5.426327385668117e-05, "loss": 0.9509, "step": 7673 }, { "epoch": 0.8, "learning_rate": 5.42618261717247e-05, "loss": 0.8966, "step": 7674 }, { "epoch": 0.8, "learning_rate": 5.42603783234433e-05, "loss": 0.2862, "step": 7675 }, { "epoch": 0.8, "learning_rate": 5.42589303118467e-05, "loss": 0.6141, "step": 7676 }, { "epoch": 0.8, "learning_rate": 5.425748213694466e-05, "loss": 0.3829, "step": 7677 }, { "epoch": 0.8, "learning_rate": 5.425603379874692e-05, "loss": 1.0633, "step": 7678 }, { "epoch": 0.8, "learning_rate": 5.4254585297263244e-05, "loss": 0.5539, "step": 7679 }, { "epoch": 0.8, "learning_rate": 5.4253136632503363e-05, "loss": 0.5406, "step": 7680 }, { "epoch": 0.8, "learning_rate": 5.4251687804477055e-05, "loss": 1.2386, "step": 7681 }, { "epoch": 0.8, "learning_rate": 5.425023881319406e-05, "loss": 1.4304, "step": 7682 }, { "epoch": 0.8, "learning_rate": 5.424878965866412e-05, "loss": 0.6982, "step": 7683 }, { "epoch": 0.8, "learning_rate": 5.424734034089701e-05, "loss": 0.5819, "step": 7684 }, { "epoch": 0.8, "learning_rate": 5.4245890859902476e-05, "loss": 0.7341, "step": 7685 }, { "epoch": 0.8, "learning_rate": 5.424444121569027e-05, "loss": 0.4674, "step": 7686 }, { "epoch": 0.8, "learning_rate": 5.424299140827017e-05, "loss": 1.0278, "step": 7687 }, { "epoch": 0.8, "learning_rate": 5.424154143765191e-05, "loss": 0.3136, "step": 7688 }, { "epoch": 0.8, "learning_rate": 5.424009130384528e-05, "loss": 0.9577, "step": 7689 }, { "epoch": 0.8, "learning_rate": 5.423864100686003e-05, "loss": 0.7888, "step": 7690 }, { "epoch": 0.8, "learning_rate": 5.423719054670591e-05, "loss": 0.7378, "step": 7691 }, { "epoch": 0.8, "learning_rate": 5.4235739923392696e-05, "loss": 0.2516, "step": 7692 }, { "epoch": 0.8, "learning_rate": 5.423428913693015e-05, "loss": 1.2058, "step": 7693 }, { "epoch": 0.8, "learning_rate": 5.423283818732804e-05, "loss": 1.0564, "step": 7694 }, { "epoch": 0.8, "learning_rate": 5.423138707459614e-05, "loss": 0.9224, "step": 7695 }, { "epoch": 0.8, "learning_rate": 5.4229935798744205e-05, "loss": 0.9497, "step": 7696 }, { "epoch": 0.8, "learning_rate": 5.4228484359782015e-05, "loss": 1.0918, "step": 7697 }, { "epoch": 0.8, "learning_rate": 5.422703275771933e-05, "loss": 0.6302, "step": 7698 }, { "epoch": 0.8, "learning_rate": 5.4225580992565944e-05, "loss": 0.5731, "step": 7699 }, { "epoch": 0.8, "learning_rate": 5.42241290643316e-05, "loss": 1.5489, "step": 7700 }, { "epoch": 0.8, "learning_rate": 5.422267697302609e-05, "loss": 0.6447, "step": 7701 }, { "epoch": 0.8, "learning_rate": 5.422122471865919e-05, "loss": 0.2643, "step": 7702 }, { "epoch": 0.8, "learning_rate": 5.421977230124067e-05, "loss": 0.3666, "step": 7703 }, { "epoch": 0.8, "learning_rate": 5.4218319720780316e-05, "loss": 1.2788, "step": 7704 }, { "epoch": 0.8, "learning_rate": 5.421686697728789e-05, "loss": 1.0958, "step": 7705 }, { "epoch": 0.8, "learning_rate": 5.4215414070773195e-05, "loss": 0.3449, "step": 7706 }, { "epoch": 0.8, "learning_rate": 5.421396100124599e-05, "loss": 0.4373, "step": 7707 }, { "epoch": 0.81, "learning_rate": 5.4212507768716066e-05, "loss": 1.5955, "step": 7708 }, { "epoch": 0.81, "learning_rate": 5.421105437319321e-05, "loss": 1.2033, "step": 7709 }, { "epoch": 0.81, "learning_rate": 5.4209600814687195e-05, "loss": 0.4873, "step": 7710 }, { "epoch": 0.81, "learning_rate": 5.420814709320781e-05, "loss": 1.2365, "step": 7711 }, { "epoch": 0.81, "learning_rate": 5.420669320876484e-05, "loss": 1.3781, "step": 7712 }, { "epoch": 0.81, "learning_rate": 5.420523916136809e-05, "loss": 0.6086, "step": 7713 }, { "epoch": 0.81, "learning_rate": 5.420378495102732e-05, "loss": 0.4217, "step": 7714 }, { "epoch": 0.81, "learning_rate": 5.420233057775235e-05, "loss": 1.1313, "step": 7715 }, { "epoch": 0.81, "learning_rate": 5.420087604155294e-05, "loss": 1.28, "step": 7716 }, { "epoch": 0.81, "learning_rate": 5.419942134243889e-05, "loss": 1.2141, "step": 7717 }, { "epoch": 0.81, "learning_rate": 5.419796648042001e-05, "loss": 1.2955, "step": 7718 }, { "epoch": 0.81, "learning_rate": 5.419651145550607e-05, "loss": 0.6563, "step": 7719 }, { "epoch": 0.81, "learning_rate": 5.419505626770688e-05, "loss": 0.31, "step": 7720 }, { "epoch": 0.81, "learning_rate": 5.4193600917032233e-05, "loss": 1.0256, "step": 7721 }, { "epoch": 0.81, "learning_rate": 5.4192145403491926e-05, "loss": 1.1314, "step": 7722 }, { "epoch": 0.81, "learning_rate": 5.4190689727095765e-05, "loss": 1.1916, "step": 7723 }, { "epoch": 0.81, "learning_rate": 5.4189233887853535e-05, "loss": 1.5827, "step": 7724 }, { "epoch": 0.81, "learning_rate": 5.418777788577504e-05, "loss": 0.6444, "step": 7725 }, { "epoch": 0.81, "learning_rate": 5.418632172087008e-05, "loss": 1.0374, "step": 7726 }, { "epoch": 0.81, "learning_rate": 5.4184865393148467e-05, "loss": 0.5119, "step": 7727 }, { "epoch": 0.81, "learning_rate": 5.418340890262e-05, "loss": 1.7456, "step": 7728 }, { "epoch": 0.81, "learning_rate": 5.418195224929448e-05, "loss": 0.6257, "step": 7729 }, { "epoch": 0.81, "learning_rate": 5.418049543318172e-05, "loss": 0.6979, "step": 7730 }, { "epoch": 0.81, "learning_rate": 5.417903845429151e-05, "loss": 0.7741, "step": 7731 }, { "epoch": 0.81, "learning_rate": 5.4177581312633683e-05, "loss": 0.5854, "step": 7732 }, { "epoch": 0.81, "learning_rate": 5.417612400821803e-05, "loss": 1.2448, "step": 7733 }, { "epoch": 0.81, "learning_rate": 5.417466654105437e-05, "loss": 0.4006, "step": 7734 }, { "epoch": 0.81, "learning_rate": 5.4173208911152515e-05, "loss": 0.8435, "step": 7735 }, { "epoch": 0.81, "learning_rate": 5.417175111852227e-05, "loss": 0.8386, "step": 7736 }, { "epoch": 0.81, "learning_rate": 5.417029316317345e-05, "loss": 1.2057, "step": 7737 }, { "epoch": 0.81, "learning_rate": 5.416883504511587e-05, "loss": 1.1321, "step": 7738 }, { "epoch": 0.81, "learning_rate": 5.416737676435935e-05, "loss": 1.1123, "step": 7739 }, { "epoch": 0.81, "learning_rate": 5.416591832091371e-05, "loss": 0.6007, "step": 7740 }, { "epoch": 0.81, "learning_rate": 5.416445971478876e-05, "loss": 0.5435, "step": 7741 }, { "epoch": 0.81, "learning_rate": 5.416300094599432e-05, "loss": 1.1396, "step": 7742 }, { "epoch": 0.81, "learning_rate": 5.41615420145402e-05, "loss": 0.3986, "step": 7743 }, { "epoch": 0.81, "learning_rate": 5.416008292043624e-05, "loss": 0.8521, "step": 7744 }, { "epoch": 0.81, "learning_rate": 5.415862366369226e-05, "loss": 0.7623, "step": 7745 }, { "epoch": 0.81, "learning_rate": 5.415716424431808e-05, "loss": 0.4028, "step": 7746 }, { "epoch": 0.81, "learning_rate": 5.415570466232352e-05, "loss": 1.5001, "step": 7747 }, { "epoch": 0.81, "learning_rate": 5.415424491771841e-05, "loss": 1.4963, "step": 7748 }, { "epoch": 0.81, "learning_rate": 5.415278501051256e-05, "loss": 1.1598, "step": 7749 }, { "epoch": 0.81, "learning_rate": 5.4151324940715825e-05, "loss": 1.2041, "step": 7750 }, { "epoch": 0.81, "learning_rate": 5.414986470833802e-05, "loss": 0.7817, "step": 7751 }, { "epoch": 0.81, "learning_rate": 5.414840431338897e-05, "loss": 0.3462, "step": 7752 }, { "epoch": 0.81, "learning_rate": 5.414694375587852e-05, "loss": 0.39, "step": 7753 }, { "epoch": 0.81, "learning_rate": 5.4145483035816495e-05, "loss": 1.1655, "step": 7754 }, { "epoch": 0.81, "learning_rate": 5.414402215321273e-05, "loss": 0.5509, "step": 7755 }, { "epoch": 0.81, "learning_rate": 5.4142561108077046e-05, "loss": 0.8861, "step": 7756 }, { "epoch": 0.81, "learning_rate": 5.4141099900419306e-05, "loss": 1.0786, "step": 7757 }, { "epoch": 0.81, "learning_rate": 5.4139638530249316e-05, "loss": 0.2744, "step": 7758 }, { "epoch": 0.81, "learning_rate": 5.413817699757694e-05, "loss": 0.9624, "step": 7759 }, { "epoch": 0.81, "learning_rate": 5.413671530241199e-05, "loss": 0.7314, "step": 7760 }, { "epoch": 0.81, "learning_rate": 5.413525344476433e-05, "loss": 0.1968, "step": 7761 }, { "epoch": 0.81, "learning_rate": 5.413379142464378e-05, "loss": 1.049, "step": 7762 }, { "epoch": 0.81, "learning_rate": 5.413232924206021e-05, "loss": 0.9328, "step": 7763 }, { "epoch": 0.81, "learning_rate": 5.413086689702344e-05, "loss": 1.4058, "step": 7764 }, { "epoch": 0.81, "learning_rate": 5.4129404389543305e-05, "loss": 0.2055, "step": 7765 }, { "epoch": 0.81, "learning_rate": 5.4127941719629684e-05, "loss": 1.275, "step": 7766 }, { "epoch": 0.81, "learning_rate": 5.41264788872924e-05, "loss": 0.309, "step": 7767 }, { "epoch": 0.81, "learning_rate": 5.4125015892541306e-05, "loss": 1.0485, "step": 7768 }, { "epoch": 0.81, "learning_rate": 5.412355273538625e-05, "loss": 1.1037, "step": 7769 }, { "epoch": 0.81, "learning_rate": 5.4122089415837074e-05, "loss": 1.3135, "step": 7770 }, { "epoch": 0.81, "learning_rate": 5.4120625933903637e-05, "loss": 1.1699, "step": 7771 }, { "epoch": 0.81, "learning_rate": 5.41191622895958e-05, "loss": 0.676, "step": 7772 }, { "epoch": 0.81, "learning_rate": 5.4117698482923393e-05, "loss": 1.1769, "step": 7773 }, { "epoch": 0.81, "learning_rate": 5.41162345138963e-05, "loss": 1.1994, "step": 7774 }, { "epoch": 0.81, "learning_rate": 5.4114770382524345e-05, "loss": 1.455, "step": 7775 }, { "epoch": 0.81, "learning_rate": 5.41133060888174e-05, "loss": 1.4627, "step": 7776 }, { "epoch": 0.81, "learning_rate": 5.411184163278532e-05, "loss": 0.6352, "step": 7777 }, { "epoch": 0.81, "learning_rate": 5.4110377014437975e-05, "loss": 0.7921, "step": 7778 }, { "epoch": 0.81, "learning_rate": 5.41089122337852e-05, "loss": 0.7238, "step": 7779 }, { "epoch": 0.81, "learning_rate": 5.410744729083687e-05, "loss": 1.7176, "step": 7780 }, { "epoch": 0.81, "learning_rate": 5.410598218560285e-05, "loss": 1.4017, "step": 7781 }, { "epoch": 0.81, "learning_rate": 5.4104516918093e-05, "loss": 0.7407, "step": 7782 }, { "epoch": 0.81, "learning_rate": 5.4103051488317184e-05, "loss": 0.9477, "step": 7783 }, { "epoch": 0.81, "learning_rate": 5.410158589628525e-05, "loss": 0.4827, "step": 7784 }, { "epoch": 0.81, "learning_rate": 5.41001201420071e-05, "loss": 1.2409, "step": 7785 }, { "epoch": 0.81, "learning_rate": 5.409865422549256e-05, "loss": 1.771, "step": 7786 }, { "epoch": 0.81, "learning_rate": 5.409718814675153e-05, "loss": 0.9622, "step": 7787 }, { "epoch": 0.81, "learning_rate": 5.409572190579387e-05, "loss": 0.237, "step": 7788 }, { "epoch": 0.81, "learning_rate": 5.4094255502629444e-05, "loss": 1.1486, "step": 7789 }, { "epoch": 0.81, "learning_rate": 5.4092788937268125e-05, "loss": 1.0718, "step": 7790 }, { "epoch": 0.81, "learning_rate": 5.409132220971979e-05, "loss": 1.2474, "step": 7791 }, { "epoch": 0.81, "learning_rate": 5.408985531999432e-05, "loss": 0.2615, "step": 7792 }, { "epoch": 0.81, "learning_rate": 5.408838826810156e-05, "loss": 1.9193, "step": 7793 }, { "epoch": 0.81, "learning_rate": 5.4086921054051425e-05, "loss": 0.7882, "step": 7794 }, { "epoch": 0.81, "learning_rate": 5.408545367785377e-05, "loss": 0.3122, "step": 7795 }, { "epoch": 0.81, "learning_rate": 5.408398613951847e-05, "loss": 1.0284, "step": 7796 }, { "epoch": 0.81, "learning_rate": 5.408251843905542e-05, "loss": 1.3507, "step": 7797 }, { "epoch": 0.81, "learning_rate": 5.408105057647448e-05, "loss": 0.9823, "step": 7798 }, { "epoch": 0.81, "learning_rate": 5.407958255178555e-05, "loss": 1.0075, "step": 7799 }, { "epoch": 0.81, "learning_rate": 5.4078114364998496e-05, "loss": 1.0147, "step": 7800 }, { "epoch": 0.81, "learning_rate": 5.4076646016123214e-05, "loss": 0.3727, "step": 7801 }, { "epoch": 0.81, "learning_rate": 5.4075177505169586e-05, "loss": 1.3979, "step": 7802 }, { "epoch": 0.81, "learning_rate": 5.4073708832147494e-05, "loss": 1.1704, "step": 7803 }, { "epoch": 0.82, "learning_rate": 5.4072239997066837e-05, "loss": 0.7363, "step": 7804 }, { "epoch": 0.82, "learning_rate": 5.407077099993748e-05, "loss": 0.6528, "step": 7805 }, { "epoch": 0.82, "learning_rate": 5.4069301840769325e-05, "loss": 0.2997, "step": 7806 }, { "epoch": 0.82, "learning_rate": 5.406783251957227e-05, "loss": 1.0244, "step": 7807 }, { "epoch": 0.82, "learning_rate": 5.406636303635619e-05, "loss": 0.2994, "step": 7808 }, { "epoch": 0.82, "learning_rate": 5.4064893391130994e-05, "loss": 1.4675, "step": 7809 }, { "epoch": 0.82, "learning_rate": 5.406342358390656e-05, "loss": 1.0735, "step": 7810 }, { "epoch": 0.82, "learning_rate": 5.4061953614692796e-05, "loss": 0.5387, "step": 7811 }, { "epoch": 0.82, "learning_rate": 5.406048348349958e-05, "loss": 1.347, "step": 7812 }, { "epoch": 0.82, "learning_rate": 5.405901319033682e-05, "loss": 1.1356, "step": 7813 }, { "epoch": 0.82, "learning_rate": 5.405754273521442e-05, "loss": 0.95, "step": 7814 }, { "epoch": 0.82, "learning_rate": 5.405607211814226e-05, "loss": 1.4794, "step": 7815 }, { "epoch": 0.82, "learning_rate": 5.4054601339130266e-05, "loss": 0.6564, "step": 7816 }, { "epoch": 0.82, "learning_rate": 5.405313039818831e-05, "loss": 0.493, "step": 7817 }, { "epoch": 0.82, "learning_rate": 5.4051659295326315e-05, "loss": 1.27, "step": 7818 }, { "epoch": 0.82, "learning_rate": 5.405018803055418e-05, "loss": 0.4889, "step": 7819 }, { "epoch": 0.82, "learning_rate": 5.40487166038818e-05, "loss": 1.0558, "step": 7820 }, { "epoch": 0.82, "learning_rate": 5.4047245015319095e-05, "loss": 0.6856, "step": 7821 }, { "epoch": 0.82, "learning_rate": 5.4045773264875954e-05, "loss": 0.3376, "step": 7822 }, { "epoch": 0.82, "learning_rate": 5.40443013525623e-05, "loss": 0.6462, "step": 7823 }, { "epoch": 0.82, "learning_rate": 5.4042829278388033e-05, "loss": 1.3947, "step": 7824 }, { "epoch": 0.82, "learning_rate": 5.404135704236307e-05, "loss": 1.063, "step": 7825 }, { "epoch": 0.82, "learning_rate": 5.403988464449731e-05, "loss": 0.4187, "step": 7826 }, { "epoch": 0.82, "learning_rate": 5.403841208480068e-05, "loss": 0.9128, "step": 7827 }, { "epoch": 0.82, "learning_rate": 5.403693936328307e-05, "loss": 1.5865, "step": 7828 }, { "epoch": 0.82, "learning_rate": 5.403546647995441e-05, "loss": 0.744, "step": 7829 }, { "epoch": 0.82, "learning_rate": 5.4033993434824627e-05, "loss": 0.2561, "step": 7830 }, { "epoch": 0.82, "learning_rate": 5.4032520227903616e-05, "loss": 0.3036, "step": 7831 }, { "epoch": 0.82, "learning_rate": 5.40310468592013e-05, "loss": 0.1409, "step": 7832 }, { "epoch": 0.82, "learning_rate": 5.40295733287276e-05, "loss": 0.8545, "step": 7833 }, { "epoch": 0.82, "learning_rate": 5.4028099636492435e-05, "loss": 0.4405, "step": 7834 }, { "epoch": 0.82, "learning_rate": 5.402662578250573e-05, "loss": 0.6431, "step": 7835 }, { "epoch": 0.82, "learning_rate": 5.40251517667774e-05, "loss": 0.2635, "step": 7836 }, { "epoch": 0.82, "learning_rate": 5.402367758931736e-05, "loss": 0.7689, "step": 7837 }, { "epoch": 0.82, "learning_rate": 5.402220325013556e-05, "loss": 1.5211, "step": 7838 }, { "epoch": 0.82, "learning_rate": 5.4020728749241895e-05, "loss": 0.9093, "step": 7839 }, { "epoch": 0.82, "learning_rate": 5.401925408664631e-05, "loss": 1.324, "step": 7840 }, { "epoch": 0.82, "learning_rate": 5.401777926235873e-05, "loss": 1.4065, "step": 7841 }, { "epoch": 0.82, "learning_rate": 5.401630427638908e-05, "loss": 1.1574, "step": 7842 }, { "epoch": 0.82, "learning_rate": 5.401482912874728e-05, "loss": 1.3477, "step": 7843 }, { "epoch": 0.82, "learning_rate": 5.401335381944328e-05, "loss": 0.8955, "step": 7844 }, { "epoch": 0.82, "learning_rate": 5.401187834848699e-05, "loss": 0.3934, "step": 7845 }, { "epoch": 0.82, "learning_rate": 5.401040271588836e-05, "loss": 0.5261, "step": 7846 }, { "epoch": 0.82, "learning_rate": 5.400892692165732e-05, "loss": 0.5541, "step": 7847 }, { "epoch": 0.82, "learning_rate": 5.4007450965803796e-05, "loss": 0.2593, "step": 7848 }, { "epoch": 0.82, "learning_rate": 5.4005974848337723e-05, "loss": 0.3978, "step": 7849 }, { "epoch": 0.82, "learning_rate": 5.4004498569269054e-05, "loss": 1.3067, "step": 7850 }, { "epoch": 0.82, "learning_rate": 5.4003022128607714e-05, "loss": 1.4205, "step": 7851 }, { "epoch": 0.82, "learning_rate": 5.4001545526363645e-05, "loss": 0.867, "step": 7852 }, { "epoch": 0.82, "learning_rate": 5.400006876254679e-05, "loss": 1.3773, "step": 7853 }, { "epoch": 0.82, "learning_rate": 5.3998591837167084e-05, "loss": 0.7145, "step": 7854 }, { "epoch": 0.82, "learning_rate": 5.399711475023448e-05, "loss": 0.6745, "step": 7855 }, { "epoch": 0.82, "learning_rate": 5.39956375017589e-05, "loss": 0.3441, "step": 7856 }, { "epoch": 0.82, "learning_rate": 5.399416009175032e-05, "loss": 1.0667, "step": 7857 }, { "epoch": 0.82, "learning_rate": 5.399268252021866e-05, "loss": 1.4527, "step": 7858 }, { "epoch": 0.82, "learning_rate": 5.399120478717388e-05, "loss": 0.3152, "step": 7859 }, { "epoch": 0.82, "learning_rate": 5.398972689262592e-05, "loss": 1.3601, "step": 7860 }, { "epoch": 0.82, "learning_rate": 5.3988248836584725e-05, "loss": 0.8515, "step": 7861 }, { "epoch": 0.82, "learning_rate": 5.398677061906026e-05, "loss": 0.5487, "step": 7862 }, { "epoch": 0.82, "learning_rate": 5.398529224006246e-05, "loss": 1.2751, "step": 7863 }, { "epoch": 0.82, "learning_rate": 5.398381369960129e-05, "loss": 1.1788, "step": 7864 }, { "epoch": 0.82, "learning_rate": 5.39823349976867e-05, "loss": 0.5192, "step": 7865 }, { "epoch": 0.82, "learning_rate": 5.398085613432864e-05, "loss": 1.0258, "step": 7866 }, { "epoch": 0.82, "learning_rate": 5.397937710953707e-05, "loss": 1.4931, "step": 7867 }, { "epoch": 0.82, "learning_rate": 5.397789792332194e-05, "loss": 1.5916, "step": 7868 }, { "epoch": 0.82, "learning_rate": 5.397641857569321e-05, "loss": 0.4611, "step": 7869 }, { "epoch": 0.82, "learning_rate": 5.3974939066660846e-05, "loss": 0.6539, "step": 7870 }, { "epoch": 0.82, "learning_rate": 5.39734593962348e-05, "loss": 0.4564, "step": 7871 }, { "epoch": 0.82, "learning_rate": 5.397197956442503e-05, "loss": 1.5543, "step": 7872 }, { "epoch": 0.82, "learning_rate": 5.3970499571241505e-05, "loss": 0.7332, "step": 7873 }, { "epoch": 0.82, "learning_rate": 5.396901941669419e-05, "loss": 0.4056, "step": 7874 }, { "epoch": 0.82, "learning_rate": 5.396753910079304e-05, "loss": 0.4902, "step": 7875 }, { "epoch": 0.82, "learning_rate": 5.396605862354803e-05, "loss": 1.1562, "step": 7876 }, { "epoch": 0.82, "learning_rate": 5.396457798496911e-05, "loss": 1.3872, "step": 7877 }, { "epoch": 0.82, "learning_rate": 5.396309718506627e-05, "loss": 0.8611, "step": 7878 }, { "epoch": 0.82, "learning_rate": 5.396161622384946e-05, "loss": 1.2499, "step": 7879 }, { "epoch": 0.82, "learning_rate": 5.3960135101328655e-05, "loss": 1.4464, "step": 7880 }, { "epoch": 0.82, "learning_rate": 5.3958653817513836e-05, "loss": 1.0473, "step": 7881 }, { "epoch": 0.82, "learning_rate": 5.3957172372414965e-05, "loss": 0.6236, "step": 7882 }, { "epoch": 0.82, "learning_rate": 5.3955690766042e-05, "loss": 0.5081, "step": 7883 }, { "epoch": 0.82, "learning_rate": 5.3954208998404946e-05, "loss": 0.1944, "step": 7884 }, { "epoch": 0.82, "learning_rate": 5.3952727069513754e-05, "loss": 0.8449, "step": 7885 }, { "epoch": 0.82, "learning_rate": 5.39512449793784e-05, "loss": 0.7471, "step": 7886 }, { "epoch": 0.82, "learning_rate": 5.3949762728008886e-05, "loss": 1.43, "step": 7887 }, { "epoch": 0.82, "learning_rate": 5.394828031541516e-05, "loss": 0.5979, "step": 7888 }, { "epoch": 0.82, "learning_rate": 5.394679774160723e-05, "loss": 1.3388, "step": 7889 }, { "epoch": 0.82, "learning_rate": 5.394531500659505e-05, "loss": 1.7595, "step": 7890 }, { "epoch": 0.82, "learning_rate": 5.3943832110388604e-05, "loss": 0.7422, "step": 7891 }, { "epoch": 0.82, "learning_rate": 5.39423490529979e-05, "loss": 0.4456, "step": 7892 }, { "epoch": 0.82, "learning_rate": 5.3940865834432886e-05, "loss": 0.2409, "step": 7893 }, { "epoch": 0.82, "learning_rate": 5.393938245470358e-05, "loss": 0.334, "step": 7894 }, { "epoch": 0.82, "learning_rate": 5.393789891381995e-05, "loss": 1.7203, "step": 7895 }, { "epoch": 0.82, "learning_rate": 5.393641521179198e-05, "loss": 1.5213, "step": 7896 }, { "epoch": 0.82, "learning_rate": 5.393493134862966e-05, "loss": 1.5761, "step": 7897 }, { "epoch": 0.82, "learning_rate": 5.393344732434299e-05, "loss": 1.2026, "step": 7898 }, { "epoch": 0.82, "learning_rate": 5.3931963138941955e-05, "loss": 1.0267, "step": 7899 }, { "epoch": 0.83, "learning_rate": 5.393047879243653e-05, "loss": 0.2419, "step": 7900 }, { "epoch": 0.83, "learning_rate": 5.392899428483673e-05, "loss": 0.8365, "step": 7901 }, { "epoch": 0.83, "learning_rate": 5.392750961615255e-05, "loss": 0.9197, "step": 7902 }, { "epoch": 0.83, "learning_rate": 5.3926024786393964e-05, "loss": 1.2018, "step": 7903 }, { "epoch": 0.83, "learning_rate": 5.392453979557098e-05, "loss": 0.3989, "step": 7904 }, { "epoch": 0.83, "learning_rate": 5.392305464369358e-05, "loss": 0.5022, "step": 7905 }, { "epoch": 0.83, "learning_rate": 5.392156933077179e-05, "loss": 1.0273, "step": 7906 }, { "epoch": 0.83, "learning_rate": 5.392008385681558e-05, "loss": 1.1702, "step": 7907 }, { "epoch": 0.83, "learning_rate": 5.3918598221834965e-05, "loss": 0.6359, "step": 7908 }, { "epoch": 0.83, "learning_rate": 5.391711242583996e-05, "loss": 1.1611, "step": 7909 }, { "epoch": 0.83, "learning_rate": 5.391562646884053e-05, "loss": 0.9889, "step": 7910 }, { "epoch": 0.83, "learning_rate": 5.39141403508467e-05, "loss": 0.8071, "step": 7911 }, { "epoch": 0.83, "learning_rate": 5.3912654071868486e-05, "loss": 0.5748, "step": 7912 }, { "epoch": 0.83, "learning_rate": 5.391116763191587e-05, "loss": 0.7281, "step": 7913 }, { "epoch": 0.83, "learning_rate": 5.390968103099887e-05, "loss": 0.3723, "step": 7914 }, { "epoch": 0.83, "learning_rate": 5.3908194269127484e-05, "loss": 1.3077, "step": 7915 }, { "epoch": 0.83, "learning_rate": 5.390670734631174e-05, "loss": 0.4688, "step": 7916 }, { "epoch": 0.83, "learning_rate": 5.390522026256163e-05, "loss": 0.9523, "step": 7917 }, { "epoch": 0.83, "learning_rate": 5.3903733017887175e-05, "loss": 0.5171, "step": 7918 }, { "epoch": 0.83, "learning_rate": 5.3902245612298376e-05, "loss": 0.9555, "step": 7919 }, { "epoch": 0.83, "learning_rate": 5.3900758045805254e-05, "loss": 0.8163, "step": 7920 }, { "epoch": 0.83, "learning_rate": 5.3899270318417824e-05, "loss": 1.4359, "step": 7921 }, { "epoch": 0.83, "learning_rate": 5.38977824301461e-05, "loss": 1.0815, "step": 7922 }, { "epoch": 0.83, "learning_rate": 5.389629438100009e-05, "loss": 0.8094, "step": 7923 }, { "epoch": 0.83, "learning_rate": 5.3894806170989824e-05, "loss": 0.8701, "step": 7924 }, { "epoch": 0.83, "learning_rate": 5.389331780012531e-05, "loss": 0.4104, "step": 7925 }, { "epoch": 0.83, "learning_rate": 5.389182926841658e-05, "loss": 1.5109, "step": 7926 }, { "epoch": 0.83, "learning_rate": 5.389034057587363e-05, "loss": 0.3893, "step": 7927 }, { "epoch": 0.83, "learning_rate": 5.3888851722506505e-05, "loss": 1.2388, "step": 7928 }, { "epoch": 0.83, "learning_rate": 5.3887362708325215e-05, "loss": 1.4585, "step": 7929 }, { "epoch": 0.83, "learning_rate": 5.3885873533339796e-05, "loss": 1.3618, "step": 7930 }, { "epoch": 0.83, "learning_rate": 5.388438419756026e-05, "loss": 1.1146, "step": 7931 }, { "epoch": 0.83, "learning_rate": 5.388289470099664e-05, "loss": 0.447, "step": 7932 }, { "epoch": 0.83, "learning_rate": 5.388140504365896e-05, "loss": 1.0799, "step": 7933 }, { "epoch": 0.83, "learning_rate": 5.387991522555725e-05, "loss": 1.158, "step": 7934 }, { "epoch": 0.83, "learning_rate": 5.387842524670154e-05, "loss": 0.6476, "step": 7935 }, { "epoch": 0.83, "learning_rate": 5.3876935107101856e-05, "loss": 0.7239, "step": 7936 }, { "epoch": 0.83, "learning_rate": 5.3875444806768236e-05, "loss": 0.7778, "step": 7937 }, { "epoch": 0.83, "learning_rate": 5.3873954345710704e-05, "loss": 0.4742, "step": 7938 }, { "epoch": 0.83, "learning_rate": 5.3872463723939295e-05, "loss": 0.2055, "step": 7939 }, { "epoch": 0.83, "learning_rate": 5.3870972941464046e-05, "loss": 0.7665, "step": 7940 }, { "epoch": 0.83, "learning_rate": 5.3869481998295e-05, "loss": 0.5379, "step": 7941 }, { "epoch": 0.83, "learning_rate": 5.386799089444218e-05, "loss": 0.6225, "step": 7942 }, { "epoch": 0.83, "learning_rate": 5.386649962991563e-05, "loss": 0.2475, "step": 7943 }, { "epoch": 0.83, "learning_rate": 5.386500820472539e-05, "loss": 0.3692, "step": 7944 }, { "epoch": 0.83, "learning_rate": 5.38635166188815e-05, "loss": 0.7373, "step": 7945 }, { "epoch": 0.83, "learning_rate": 5.3862024872393994e-05, "loss": 0.8984, "step": 7946 }, { "epoch": 0.83, "learning_rate": 5.3860532965272924e-05, "loss": 0.601, "step": 7947 }, { "epoch": 0.83, "learning_rate": 5.385904089752833e-05, "loss": 0.7073, "step": 7948 }, { "epoch": 0.83, "learning_rate": 5.385754866917026e-05, "loss": 0.9364, "step": 7949 }, { "epoch": 0.83, "learning_rate": 5.3856056280208745e-05, "loss": 0.8281, "step": 7950 }, { "epoch": 0.83, "learning_rate": 5.385456373065385e-05, "loss": 1.0941, "step": 7951 }, { "epoch": 0.83, "learning_rate": 5.385307102051561e-05, "loss": 0.4689, "step": 7952 }, { "epoch": 0.83, "learning_rate": 5.385157814980407e-05, "loss": 1.0372, "step": 7953 }, { "epoch": 0.83, "learning_rate": 5.385008511852929e-05, "loss": 1.0468, "step": 7954 }, { "epoch": 0.83, "learning_rate": 5.384859192670132e-05, "loss": 1.3555, "step": 7955 }, { "epoch": 0.83, "learning_rate": 5.384709857433021e-05, "loss": 0.5333, "step": 7956 }, { "epoch": 0.83, "learning_rate": 5.384560506142601e-05, "loss": 0.897, "step": 7957 }, { "epoch": 0.83, "learning_rate": 5.384411138799878e-05, "loss": 0.8622, "step": 7958 }, { "epoch": 0.83, "learning_rate": 5.3842617554058574e-05, "loss": 1.2779, "step": 7959 }, { "epoch": 0.83, "learning_rate": 5.384112355961544e-05, "loss": 0.7752, "step": 7960 }, { "epoch": 0.83, "learning_rate": 5.3839629404679435e-05, "loss": 0.4945, "step": 7961 }, { "epoch": 0.83, "learning_rate": 5.3838135089260635e-05, "loss": 0.1968, "step": 7962 }, { "epoch": 0.83, "learning_rate": 5.383664061336907e-05, "loss": 1.0209, "step": 7963 }, { "epoch": 0.83, "learning_rate": 5.3835145977014834e-05, "loss": 0.3866, "step": 7964 }, { "epoch": 0.83, "learning_rate": 5.3833651180207966e-05, "loss": 0.6202, "step": 7965 }, { "epoch": 0.83, "learning_rate": 5.3832156222958536e-05, "loss": 0.3088, "step": 7966 }, { "epoch": 0.83, "learning_rate": 5.3830661105276606e-05, "loss": 0.6769, "step": 7967 }, { "epoch": 0.83, "learning_rate": 5.382916582717224e-05, "loss": 0.7144, "step": 7968 }, { "epoch": 0.83, "learning_rate": 5.382767038865551e-05, "loss": 0.9849, "step": 7969 }, { "epoch": 0.83, "learning_rate": 5.382617478973647e-05, "loss": 1.2809, "step": 7970 }, { "epoch": 0.83, "learning_rate": 5.382467903042521e-05, "loss": 0.8484, "step": 7971 }, { "epoch": 0.83, "learning_rate": 5.3823183110731774e-05, "loss": 1.2088, "step": 7972 }, { "epoch": 0.83, "learning_rate": 5.382168703066625e-05, "loss": 1.1308, "step": 7973 }, { "epoch": 0.83, "learning_rate": 5.382019079023869e-05, "loss": 0.6371, "step": 7974 }, { "epoch": 0.83, "learning_rate": 5.381869438945919e-05, "loss": 0.6998, "step": 7975 }, { "epoch": 0.83, "learning_rate": 5.3817197828337814e-05, "loss": 0.4268, "step": 7976 }, { "epoch": 0.83, "learning_rate": 5.3815701106884624e-05, "loss": 0.7107, "step": 7977 }, { "epoch": 0.83, "learning_rate": 5.381420422510972e-05, "loss": 1.5081, "step": 7978 }, { "epoch": 0.83, "learning_rate": 5.381270718302316e-05, "loss": 1.6873, "step": 7979 }, { "epoch": 0.83, "learning_rate": 5.381120998063502e-05, "loss": 0.4932, "step": 7980 }, { "epoch": 0.83, "learning_rate": 5.3809712617955394e-05, "loss": 0.6383, "step": 7981 }, { "epoch": 0.83, "learning_rate": 5.380821509499435e-05, "loss": 1.2454, "step": 7982 }, { "epoch": 0.83, "learning_rate": 5.3806717411761975e-05, "loss": 0.6125, "step": 7983 }, { "epoch": 0.83, "learning_rate": 5.380521956826835e-05, "loss": 1.8114, "step": 7984 }, { "epoch": 0.83, "learning_rate": 5.380372156452356e-05, "loss": 1.1557, "step": 7985 }, { "epoch": 0.83, "learning_rate": 5.380222340053767e-05, "loss": 1.216, "step": 7986 }, { "epoch": 0.83, "learning_rate": 5.3800725076320796e-05, "loss": 1.0739, "step": 7987 }, { "epoch": 0.83, "learning_rate": 5.3799226591883004e-05, "loss": 0.4144, "step": 7988 }, { "epoch": 0.83, "learning_rate": 5.379772794723439e-05, "loss": 1.3911, "step": 7989 }, { "epoch": 0.83, "learning_rate": 5.3796229142385043e-05, "loss": 0.4058, "step": 7990 }, { "epoch": 0.83, "learning_rate": 5.379473017734504e-05, "loss": 0.7615, "step": 7991 }, { "epoch": 0.83, "learning_rate": 5.3793231052124487e-05, "loss": 0.535, "step": 7992 }, { "epoch": 0.83, "learning_rate": 5.379173176673347e-05, "loss": 0.9449, "step": 7993 }, { "epoch": 0.83, "learning_rate": 5.3790232321182086e-05, "loss": 0.6978, "step": 7994 }, { "epoch": 0.83, "learning_rate": 5.378873271548041e-05, "loss": 0.3589, "step": 7995 }, { "epoch": 0.84, "learning_rate": 5.378723294963857e-05, "loss": 0.3474, "step": 7996 }, { "epoch": 0.84, "learning_rate": 5.378573302366663e-05, "loss": 0.5044, "step": 7997 }, { "epoch": 0.84, "learning_rate": 5.3784232937574705e-05, "loss": 0.7507, "step": 7998 }, { "epoch": 0.84, "learning_rate": 5.3782732691372886e-05, "loss": 0.429, "step": 7999 }, { "epoch": 0.84, "learning_rate": 5.378123228507127e-05, "loss": 1.5265, "step": 8000 }, { "epoch": 0.84, "learning_rate": 5.377973171867997e-05, "loss": 1.0642, "step": 8001 }, { "epoch": 0.84, "learning_rate": 5.377823099220909e-05, "loss": 0.477, "step": 8002 }, { "epoch": 0.84, "learning_rate": 5.37767301056687e-05, "loss": 1.6565, "step": 8003 }, { "epoch": 0.84, "learning_rate": 5.3775229059068946e-05, "loss": 0.6038, "step": 8004 }, { "epoch": 0.84, "learning_rate": 5.377372785241991e-05, "loss": 1.1274, "step": 8005 }, { "epoch": 0.84, "learning_rate": 5.37722264857317e-05, "loss": 1.0448, "step": 8006 }, { "epoch": 0.84, "learning_rate": 5.377072495901441e-05, "loss": 0.9156, "step": 8007 }, { "epoch": 0.84, "learning_rate": 5.3769223272278175e-05, "loss": 1.08, "step": 8008 }, { "epoch": 0.84, "learning_rate": 5.376772142553309e-05, "loss": 0.5853, "step": 8009 }, { "epoch": 0.84, "learning_rate": 5.376621941878926e-05, "loss": 0.4642, "step": 8010 }, { "epoch": 0.84, "learning_rate": 5.376471725205681e-05, "loss": 0.7689, "step": 8011 }, { "epoch": 0.84, "learning_rate": 5.376321492534584e-05, "loss": 1.7292, "step": 8012 }, { "epoch": 0.84, "learning_rate": 5.376171243866646e-05, "loss": 1.4712, "step": 8013 }, { "epoch": 0.84, "learning_rate": 5.37602097920288e-05, "loss": 0.8967, "step": 8014 }, { "epoch": 0.84, "learning_rate": 5.3758706985442956e-05, "loss": 1.115, "step": 8015 }, { "epoch": 0.84, "learning_rate": 5.375720401891907e-05, "loss": 0.6373, "step": 8016 }, { "epoch": 0.84, "learning_rate": 5.375570089246724e-05, "loss": 1.0996, "step": 8017 }, { "epoch": 0.84, "learning_rate": 5.375419760609759e-05, "loss": 0.731, "step": 8018 }, { "epoch": 0.84, "learning_rate": 5.375269415982024e-05, "loss": 0.563, "step": 8019 }, { "epoch": 0.84, "learning_rate": 5.3751190553645306e-05, "loss": 0.9584, "step": 8020 }, { "epoch": 0.84, "learning_rate": 5.374968678758293e-05, "loss": 1.3644, "step": 8021 }, { "epoch": 0.84, "learning_rate": 5.37481828616432e-05, "loss": 1.3546, "step": 8022 }, { "epoch": 0.84, "learning_rate": 5.374667877583627e-05, "loss": 0.4696, "step": 8023 }, { "epoch": 0.84, "learning_rate": 5.374517453017225e-05, "loss": 0.6023, "step": 8024 }, { "epoch": 0.84, "learning_rate": 5.374367012466127e-05, "loss": 1.469, "step": 8025 }, { "epoch": 0.84, "learning_rate": 5.3742165559313466e-05, "loss": 1.3024, "step": 8026 }, { "epoch": 0.84, "learning_rate": 5.3740660834138954e-05, "loss": 1.0419, "step": 8027 }, { "epoch": 0.84, "learning_rate": 5.3739155949147864e-05, "loss": 0.3383, "step": 8028 }, { "epoch": 0.84, "learning_rate": 5.3737650904350334e-05, "loss": 1.6262, "step": 8029 }, { "epoch": 0.84, "learning_rate": 5.37361456997565e-05, "loss": 0.3452, "step": 8030 }, { "epoch": 0.84, "learning_rate": 5.373464033537648e-05, "loss": 0.3823, "step": 8031 }, { "epoch": 0.84, "learning_rate": 5.3733134811220417e-05, "loss": 0.968, "step": 8032 }, { "epoch": 0.84, "learning_rate": 5.373162912729844e-05, "loss": 1.3504, "step": 8033 }, { "epoch": 0.84, "learning_rate": 5.373012328362069e-05, "loss": 1.054, "step": 8034 }, { "epoch": 0.84, "learning_rate": 5.3728617280197305e-05, "loss": 1.1126, "step": 8035 }, { "epoch": 0.84, "learning_rate": 5.372711111703842e-05, "loss": 0.1948, "step": 8036 }, { "epoch": 0.84, "learning_rate": 5.372560479415416e-05, "loss": 1.0118, "step": 8037 }, { "epoch": 0.84, "learning_rate": 5.3724098311554694e-05, "loss": 0.9817, "step": 8038 }, { "epoch": 0.84, "learning_rate": 5.372259166925015e-05, "loss": 1.2374, "step": 8039 }, { "epoch": 0.84, "learning_rate": 5.3721084867250674e-05, "loss": 1.5492, "step": 8040 }, { "epoch": 0.84, "learning_rate": 5.37195779055664e-05, "loss": 0.7201, "step": 8041 }, { "epoch": 0.84, "learning_rate": 5.371807078420747e-05, "loss": 0.5367, "step": 8042 }, { "epoch": 0.84, "learning_rate": 5.371656350318405e-05, "loss": 1.3274, "step": 8043 }, { "epoch": 0.84, "learning_rate": 5.371505606250626e-05, "loss": 1.4338, "step": 8044 }, { "epoch": 0.84, "learning_rate": 5.3713548462184264e-05, "loss": 1.2752, "step": 8045 }, { "epoch": 0.84, "learning_rate": 5.371204070222821e-05, "loss": 0.8564, "step": 8046 }, { "epoch": 0.84, "learning_rate": 5.371053278264825e-05, "loss": 0.8286, "step": 8047 }, { "epoch": 0.84, "learning_rate": 5.370902470345453e-05, "loss": 1.3125, "step": 8048 }, { "epoch": 0.84, "learning_rate": 5.370751646465721e-05, "loss": 0.5326, "step": 8049 }, { "epoch": 0.84, "learning_rate": 5.370600806626643e-05, "loss": 0.7722, "step": 8050 }, { "epoch": 0.84, "learning_rate": 5.370449950829235e-05, "loss": 1.8924, "step": 8051 }, { "epoch": 0.84, "learning_rate": 5.370299079074512e-05, "loss": 0.8409, "step": 8052 }, { "epoch": 0.84, "learning_rate": 5.3701481913634906e-05, "loss": 0.7727, "step": 8053 }, { "epoch": 0.84, "learning_rate": 5.369997287697186e-05, "loss": 1.473, "step": 8054 }, { "epoch": 0.84, "learning_rate": 5.3698463680766146e-05, "loss": 1.2799, "step": 8055 }, { "epoch": 0.84, "learning_rate": 5.3696954325027926e-05, "loss": 0.8946, "step": 8056 }, { "epoch": 0.84, "learning_rate": 5.369544480976734e-05, "loss": 0.5059, "step": 8057 }, { "epoch": 0.84, "learning_rate": 5.369393513499457e-05, "loss": 1.2399, "step": 8058 }, { "epoch": 0.84, "learning_rate": 5.369242530071978e-05, "loss": 1.5421, "step": 8059 }, { "epoch": 0.84, "learning_rate": 5.369091530695312e-05, "loss": 1.0011, "step": 8060 }, { "epoch": 0.84, "learning_rate": 5.3689405153704755e-05, "loss": 0.4362, "step": 8061 }, { "epoch": 0.84, "learning_rate": 5.368789484098486e-05, "loss": 0.4631, "step": 8062 }, { "epoch": 0.84, "learning_rate": 5.3686384368803605e-05, "loss": 2.2054, "step": 8063 }, { "epoch": 0.84, "learning_rate": 5.368487373717115e-05, "loss": 1.1427, "step": 8064 }, { "epoch": 0.84, "learning_rate": 5.368336294609767e-05, "loss": 1.5325, "step": 8065 }, { "epoch": 0.84, "learning_rate": 5.368185199559333e-05, "loss": 0.3583, "step": 8066 }, { "epoch": 0.84, "learning_rate": 5.36803408856683e-05, "loss": 0.7253, "step": 8067 }, { "epoch": 0.84, "learning_rate": 5.3678829616332754e-05, "loss": 0.6037, "step": 8068 }, { "epoch": 0.84, "learning_rate": 5.367731818759688e-05, "loss": 1.6386, "step": 8069 }, { "epoch": 0.84, "learning_rate": 5.3675806599470826e-05, "loss": 0.3761, "step": 8070 }, { "epoch": 0.84, "learning_rate": 5.367429485196479e-05, "loss": 1.353, "step": 8071 }, { "epoch": 0.84, "learning_rate": 5.367278294508893e-05, "loss": 0.4805, "step": 8072 }, { "epoch": 0.84, "learning_rate": 5.367127087885345e-05, "loss": 0.7096, "step": 8073 }, { "epoch": 0.84, "learning_rate": 5.36697586532685e-05, "loss": 1.4419, "step": 8074 }, { "epoch": 0.84, "learning_rate": 5.366824626834428e-05, "loss": 0.2662, "step": 8075 }, { "epoch": 0.84, "learning_rate": 5.366673372409096e-05, "loss": 1.423, "step": 8076 }, { "epoch": 0.84, "learning_rate": 5.366522102051873e-05, "loss": 0.7406, "step": 8077 }, { "epoch": 0.84, "learning_rate": 5.366370815763777e-05, "loss": 0.3567, "step": 8078 }, { "epoch": 0.84, "learning_rate": 5.366219513545826e-05, "loss": 1.1739, "step": 8079 }, { "epoch": 0.84, "learning_rate": 5.3660681953990394e-05, "loss": 1.5225, "step": 8080 }, { "epoch": 0.84, "learning_rate": 5.365916861324435e-05, "loss": 1.3895, "step": 8081 }, { "epoch": 0.84, "learning_rate": 5.365765511323032e-05, "loss": 1.191, "step": 8082 }, { "epoch": 0.84, "learning_rate": 5.365614145395849e-05, "loss": 0.9705, "step": 8083 }, { "epoch": 0.84, "learning_rate": 5.365462763543904e-05, "loss": 1.3923, "step": 8084 }, { "epoch": 0.84, "learning_rate": 5.3653113657682186e-05, "loss": 1.5302, "step": 8085 }, { "epoch": 0.84, "learning_rate": 5.36515995206981e-05, "loss": 1.5329, "step": 8086 }, { "epoch": 0.84, "learning_rate": 5.365008522449698e-05, "loss": 1.2446, "step": 8087 }, { "epoch": 0.84, "learning_rate": 5.364857076908902e-05, "loss": 1.213, "step": 8088 }, { "epoch": 0.84, "learning_rate": 5.364705615448442e-05, "loss": 1.3523, "step": 8089 }, { "epoch": 0.84, "learning_rate": 5.3645541380693366e-05, "loss": 0.5799, "step": 8090 }, { "epoch": 0.85, "learning_rate": 5.364402644772606e-05, "loss": 1.3355, "step": 8091 }, { "epoch": 0.85, "learning_rate": 5.3642511355592695e-05, "loss": 0.2461, "step": 8092 }, { "epoch": 0.85, "learning_rate": 5.364099610430349e-05, "loss": 1.3848, "step": 8093 }, { "epoch": 0.85, "learning_rate": 5.363948069386862e-05, "loss": 1.2949, "step": 8094 }, { "epoch": 0.85, "learning_rate": 5.36379651242983e-05, "loss": 0.4219, "step": 8095 }, { "epoch": 0.85, "learning_rate": 5.363644939560273e-05, "loss": 0.346, "step": 8096 }, { "epoch": 0.85, "learning_rate": 5.3634933507792105e-05, "loss": 1.3088, "step": 8097 }, { "epoch": 0.85, "learning_rate": 5.3633417460876644e-05, "loss": 0.7918, "step": 8098 }, { "epoch": 0.85, "learning_rate": 5.3631901254866556e-05, "loss": 1.2121, "step": 8099 }, { "epoch": 0.85, "learning_rate": 5.3630384889772024e-05, "loss": 1.0124, "step": 8100 }, { "epoch": 0.85, "learning_rate": 5.3628868365603276e-05, "loss": 0.7904, "step": 8101 }, { "epoch": 0.85, "learning_rate": 5.3627351682370515e-05, "loss": 1.487, "step": 8102 }, { "epoch": 0.85, "learning_rate": 5.362583484008395e-05, "loss": 1.4128, "step": 8103 }, { "epoch": 0.85, "learning_rate": 5.362431783875379e-05, "loss": 0.4204, "step": 8104 }, { "epoch": 0.85, "learning_rate": 5.362280067839025e-05, "loss": 1.5553, "step": 8105 }, { "epoch": 0.85, "learning_rate": 5.362128335900355e-05, "loss": 0.3887, "step": 8106 }, { "epoch": 0.85, "learning_rate": 5.3619765880603894e-05, "loss": 0.3917, "step": 8107 }, { "epoch": 0.85, "learning_rate": 5.36182482432015e-05, "loss": 1.0228, "step": 8108 }, { "epoch": 0.85, "learning_rate": 5.3616730446806586e-05, "loss": 0.1177, "step": 8109 }, { "epoch": 0.85, "learning_rate": 5.361521249142937e-05, "loss": 1.2924, "step": 8110 }, { "epoch": 0.85, "learning_rate": 5.361369437708007e-05, "loss": 0.3744, "step": 8111 }, { "epoch": 0.85, "learning_rate": 5.36121761037689e-05, "loss": 0.9425, "step": 8112 }, { "epoch": 0.85, "learning_rate": 5.361065767150608e-05, "loss": 0.4062, "step": 8113 }, { "epoch": 0.85, "learning_rate": 5.360913908030184e-05, "loss": 1.1248, "step": 8114 }, { "epoch": 0.85, "learning_rate": 5.3607620330166415e-05, "loss": 1.6245, "step": 8115 }, { "epoch": 0.85, "learning_rate": 5.3606101421109996e-05, "loss": 1.5833, "step": 8116 }, { "epoch": 0.85, "learning_rate": 5.360458235314283e-05, "loss": 1.4449, "step": 8117 }, { "epoch": 0.85, "learning_rate": 5.3603063126275144e-05, "loss": 0.1678, "step": 8118 }, { "epoch": 0.85, "learning_rate": 5.3601543740517154e-05, "loss": 0.652, "step": 8119 }, { "epoch": 0.85, "learning_rate": 5.36000241958791e-05, "loss": 0.5068, "step": 8120 }, { "epoch": 0.85, "learning_rate": 5.35985044923712e-05, "loss": 0.7498, "step": 8121 }, { "epoch": 0.85, "learning_rate": 5.359698463000369e-05, "loss": 0.2203, "step": 8122 }, { "epoch": 0.85, "learning_rate": 5.35954646087868e-05, "loss": 1.5869, "step": 8123 }, { "epoch": 0.85, "learning_rate": 5.359394442873076e-05, "loss": 0.7473, "step": 8124 }, { "epoch": 0.85, "learning_rate": 5.359242408984581e-05, "loss": 0.9638, "step": 8125 }, { "epoch": 0.85, "learning_rate": 5.359090359214219e-05, "loss": 0.5222, "step": 8126 }, { "epoch": 0.85, "learning_rate": 5.358938293563012e-05, "loss": 1.2948, "step": 8127 }, { "epoch": 0.85, "learning_rate": 5.3587862120319836e-05, "loss": 0.6408, "step": 8128 }, { "epoch": 0.85, "learning_rate": 5.3586341146221595e-05, "loss": 1.2001, "step": 8129 }, { "epoch": 0.85, "learning_rate": 5.358482001334562e-05, "loss": 0.9625, "step": 8130 }, { "epoch": 0.85, "learning_rate": 5.358329872170215e-05, "loss": 0.4981, "step": 8131 }, { "epoch": 0.85, "learning_rate": 5.358177727130144e-05, "loss": 1.2933, "step": 8132 }, { "epoch": 0.85, "learning_rate": 5.358025566215372e-05, "loss": 0.5903, "step": 8133 }, { "epoch": 0.85, "learning_rate": 5.357873389426922e-05, "loss": 0.7866, "step": 8134 }, { "epoch": 0.85, "learning_rate": 5.357721196765822e-05, "loss": 1.0567, "step": 8135 }, { "epoch": 0.85, "learning_rate": 5.357568988233094e-05, "loss": 1.1228, "step": 8136 }, { "epoch": 0.85, "learning_rate": 5.357416763829763e-05, "loss": 0.4672, "step": 8137 }, { "epoch": 0.85, "learning_rate": 5.357264523556855e-05, "loss": 0.4851, "step": 8138 }, { "epoch": 0.85, "learning_rate": 5.357112267415392e-05, "loss": 1.4538, "step": 8139 }, { "epoch": 0.85, "learning_rate": 5.3569599954064024e-05, "loss": 1.3002, "step": 8140 }, { "epoch": 0.85, "learning_rate": 5.3568077075309086e-05, "loss": 1.0986, "step": 8141 }, { "epoch": 0.85, "learning_rate": 5.3566554037899375e-05, "loss": 0.5984, "step": 8142 }, { "epoch": 0.85, "learning_rate": 5.356503084184513e-05, "loss": 1.1226, "step": 8143 }, { "epoch": 0.85, "learning_rate": 5.3563507487156616e-05, "loss": 0.9925, "step": 8144 }, { "epoch": 0.85, "learning_rate": 5.356198397384408e-05, "loss": 1.5061, "step": 8145 }, { "epoch": 0.85, "learning_rate": 5.356046030191778e-05, "loss": 1.6418, "step": 8146 }, { "epoch": 0.85, "learning_rate": 5.3558936471387975e-05, "loss": 0.5659, "step": 8147 }, { "epoch": 0.85, "learning_rate": 5.355741248226492e-05, "loss": 1.3255, "step": 8148 }, { "epoch": 0.85, "learning_rate": 5.3555888334558885e-05, "loss": 1.4322, "step": 8149 }, { "epoch": 0.85, "learning_rate": 5.355436402828011e-05, "loss": 0.9016, "step": 8150 }, { "epoch": 0.85, "learning_rate": 5.355283956343887e-05, "loss": 1.4309, "step": 8151 }, { "epoch": 0.85, "learning_rate": 5.355131494004543e-05, "loss": 1.2453, "step": 8152 }, { "epoch": 0.85, "learning_rate": 5.354979015811005e-05, "loss": 1.4061, "step": 8153 }, { "epoch": 0.85, "learning_rate": 5.354826521764299e-05, "loss": 1.1341, "step": 8154 }, { "epoch": 0.85, "learning_rate": 5.3546740118654506e-05, "loss": 2.1704, "step": 8155 }, { "epoch": 0.85, "learning_rate": 5.354521486115489e-05, "loss": 0.7039, "step": 8156 }, { "epoch": 0.85, "learning_rate": 5.3543689445154395e-05, "loss": 1.0473, "step": 8157 }, { "epoch": 0.85, "learning_rate": 5.35421638706633e-05, "loss": 0.3883, "step": 8158 }, { "epoch": 0.85, "learning_rate": 5.354063813769186e-05, "loss": 0.7012, "step": 8159 }, { "epoch": 0.85, "learning_rate": 5.353911224625034e-05, "loss": 0.5274, "step": 8160 }, { "epoch": 0.85, "learning_rate": 5.3537586196349036e-05, "loss": 1.1222, "step": 8161 }, { "epoch": 0.85, "learning_rate": 5.35360599879982e-05, "loss": 0.8838, "step": 8162 }, { "epoch": 0.85, "learning_rate": 5.3534533621208127e-05, "loss": 1.4916, "step": 8163 }, { "epoch": 0.85, "learning_rate": 5.353300709598908e-05, "loss": 0.9479, "step": 8164 }, { "epoch": 0.85, "learning_rate": 5.3531480412351325e-05, "loss": 0.6911, "step": 8165 }, { "epoch": 0.85, "learning_rate": 5.352995357030516e-05, "loss": 0.8348, "step": 8166 }, { "epoch": 0.85, "learning_rate": 5.352842656986084e-05, "loss": 0.3905, "step": 8167 }, { "epoch": 0.85, "learning_rate": 5.3526899411028675e-05, "loss": 1.7157, "step": 8168 }, { "epoch": 0.85, "learning_rate": 5.352537209381892e-05, "loss": 0.9433, "step": 8169 }, { "epoch": 0.85, "learning_rate": 5.352384461824186e-05, "loss": 0.6703, "step": 8170 }, { "epoch": 0.85, "learning_rate": 5.352231698430779e-05, "loss": 0.3149, "step": 8171 }, { "epoch": 0.85, "learning_rate": 5.352078919202698e-05, "loss": 0.9537, "step": 8172 }, { "epoch": 0.85, "learning_rate": 5.351926124140972e-05, "loss": 1.2344, "step": 8173 }, { "epoch": 0.85, "learning_rate": 5.35177331324663e-05, "loss": 1.1216, "step": 8174 }, { "epoch": 0.85, "learning_rate": 5.3516204865207e-05, "loss": 2.3465, "step": 8175 }, { "epoch": 0.85, "learning_rate": 5.351467643964211e-05, "loss": 0.2189, "step": 8176 }, { "epoch": 0.85, "learning_rate": 5.3513147855781926e-05, "loss": 0.3592, "step": 8177 }, { "epoch": 0.85, "learning_rate": 5.351161911363674e-05, "loss": 1.21, "step": 8178 }, { "epoch": 0.85, "learning_rate": 5.3510090213216814e-05, "loss": 1.1633, "step": 8179 }, { "epoch": 0.85, "learning_rate": 5.350856115453247e-05, "loss": 1.075, "step": 8180 }, { "epoch": 0.85, "learning_rate": 5.3507031937594e-05, "loss": 0.2589, "step": 8181 }, { "epoch": 0.85, "learning_rate": 5.350550256241169e-05, "loss": 0.7132, "step": 8182 }, { "epoch": 0.85, "learning_rate": 5.350397302899583e-05, "loss": 0.2915, "step": 8183 }, { "epoch": 0.85, "learning_rate": 5.3502443337356725e-05, "loss": 0.6924, "step": 8184 }, { "epoch": 0.85, "learning_rate": 5.350091348750467e-05, "loss": 1.3441, "step": 8185 }, { "epoch": 0.85, "learning_rate": 5.349938347944997e-05, "loss": 1.3511, "step": 8186 }, { "epoch": 0.86, "learning_rate": 5.3497853313202914e-05, "loss": 0.641, "step": 8187 }, { "epoch": 0.86, "learning_rate": 5.3496322988773806e-05, "loss": 1.339, "step": 8188 }, { "epoch": 0.86, "learning_rate": 5.349479250617295e-05, "loss": 0.3865, "step": 8189 }, { "epoch": 0.86, "learning_rate": 5.3493261865410653e-05, "loss": 1.4224, "step": 8190 }, { "epoch": 0.86, "learning_rate": 5.349173106649721e-05, "loss": 1.7355, "step": 8191 }, { "epoch": 0.86, "learning_rate": 5.3490200109442937e-05, "loss": 0.4273, "step": 8192 }, { "epoch": 0.86, "learning_rate": 5.348866899425812e-05, "loss": 1.578, "step": 8193 }, { "epoch": 0.86, "learning_rate": 5.348713772095309e-05, "loss": 0.928, "step": 8194 }, { "epoch": 0.86, "learning_rate": 5.348560628953814e-05, "loss": 1.0491, "step": 8195 }, { "epoch": 0.86, "learning_rate": 5.3484074700023577e-05, "loss": 0.4586, "step": 8196 }, { "epoch": 0.86, "learning_rate": 5.348254295241973e-05, "loss": 1.032, "step": 8197 }, { "epoch": 0.86, "learning_rate": 5.3481011046736895e-05, "loss": 0.6925, "step": 8198 }, { "epoch": 0.86, "learning_rate": 5.347947898298539e-05, "loss": 1.2846, "step": 8199 }, { "epoch": 0.86, "learning_rate": 5.347794676117552e-05, "loss": 1.1561, "step": 8200 }, { "epoch": 0.86, "learning_rate": 5.3476414381317604e-05, "loss": 0.9152, "step": 8201 }, { "epoch": 0.86, "learning_rate": 5.3474881843421964e-05, "loss": 1.0787, "step": 8202 }, { "epoch": 0.86, "learning_rate": 5.347334914749892e-05, "loss": 1.3644, "step": 8203 }, { "epoch": 0.86, "learning_rate": 5.3471816293558766e-05, "loss": 0.7693, "step": 8204 }, { "epoch": 0.86, "learning_rate": 5.347028328161185e-05, "loss": 1.6161, "step": 8205 }, { "epoch": 0.86, "learning_rate": 5.3468750111668475e-05, "loss": 0.6744, "step": 8206 }, { "epoch": 0.86, "learning_rate": 5.3467216783738965e-05, "loss": 0.4304, "step": 8207 }, { "epoch": 0.86, "learning_rate": 5.346568329783364e-05, "loss": 0.8881, "step": 8208 }, { "epoch": 0.86, "learning_rate": 5.346414965396282e-05, "loss": 1.0693, "step": 8209 }, { "epoch": 0.86, "learning_rate": 5.346261585213685e-05, "loss": 1.4795, "step": 8210 }, { "epoch": 0.86, "learning_rate": 5.346108189236603e-05, "loss": 1.4556, "step": 8211 }, { "epoch": 0.86, "learning_rate": 5.345954777466069e-05, "loss": 0.5544, "step": 8212 }, { "epoch": 0.86, "learning_rate": 5.345801349903117e-05, "loss": 1.0429, "step": 8213 }, { "epoch": 0.86, "learning_rate": 5.3456479065487805e-05, "loss": 1.1419, "step": 8214 }, { "epoch": 0.86, "learning_rate": 5.34549444740409e-05, "loss": 0.8194, "step": 8215 }, { "epoch": 0.86, "learning_rate": 5.345340972470079e-05, "loss": 1.4272, "step": 8216 }, { "epoch": 0.86, "learning_rate": 5.3451874817477824e-05, "loss": 1.3194, "step": 8217 }, { "epoch": 0.86, "learning_rate": 5.345033975238232e-05, "loss": 1.5038, "step": 8218 }, { "epoch": 0.86, "learning_rate": 5.3448804529424615e-05, "loss": 1.8374, "step": 8219 }, { "epoch": 0.86, "learning_rate": 5.344726914861505e-05, "loss": 0.6203, "step": 8220 }, { "epoch": 0.86, "learning_rate": 5.344573360996394e-05, "loss": 1.0691, "step": 8221 }, { "epoch": 0.86, "learning_rate": 5.344419791348165e-05, "loss": 1.1587, "step": 8222 }, { "epoch": 0.86, "learning_rate": 5.34426620591785e-05, "loss": 0.8121, "step": 8223 }, { "epoch": 0.86, "learning_rate": 5.3441126047064834e-05, "loss": 0.9923, "step": 8224 }, { "epoch": 0.86, "learning_rate": 5.3439589877151005e-05, "loss": 0.4733, "step": 8225 }, { "epoch": 0.86, "learning_rate": 5.343805354944732e-05, "loss": 1.1511, "step": 8226 }, { "epoch": 0.86, "learning_rate": 5.3436517063964156e-05, "loss": 0.7222, "step": 8227 }, { "epoch": 0.86, "learning_rate": 5.343498042071185e-05, "loss": 1.4537, "step": 8228 }, { "epoch": 0.86, "learning_rate": 5.343344361970072e-05, "loss": 1.0429, "step": 8229 }, { "epoch": 0.86, "learning_rate": 5.3431906660941146e-05, "loss": 0.3729, "step": 8230 }, { "epoch": 0.86, "learning_rate": 5.343036954444345e-05, "loss": 0.661, "step": 8231 }, { "epoch": 0.86, "learning_rate": 5.3428832270218e-05, "loss": 0.5709, "step": 8232 }, { "epoch": 0.86, "learning_rate": 5.3427294838275114e-05, "loss": 0.4597, "step": 8233 }, { "epoch": 0.86, "learning_rate": 5.342575724862518e-05, "loss": 0.7916, "step": 8234 }, { "epoch": 0.86, "learning_rate": 5.342421950127852e-05, "loss": 0.6581, "step": 8235 }, { "epoch": 0.86, "learning_rate": 5.342268159624549e-05, "loss": 1.0832, "step": 8236 }, { "epoch": 0.86, "learning_rate": 5.342114353353646e-05, "loss": 1.3732, "step": 8237 }, { "epoch": 0.86, "learning_rate": 5.341960531316177e-05, "loss": 0.572, "step": 8238 }, { "epoch": 0.86, "learning_rate": 5.341806693513176e-05, "loss": 0.6556, "step": 8239 }, { "epoch": 0.86, "learning_rate": 5.3416528399456824e-05, "loss": 1.2543, "step": 8240 }, { "epoch": 0.86, "learning_rate": 5.3414989706147285e-05, "loss": 0.62, "step": 8241 }, { "epoch": 0.86, "learning_rate": 5.3413450855213526e-05, "loss": 0.7775, "step": 8242 }, { "epoch": 0.86, "learning_rate": 5.341191184666588e-05, "loss": 0.6455, "step": 8243 }, { "epoch": 0.86, "learning_rate": 5.341037268051473e-05, "loss": 1.3574, "step": 8244 }, { "epoch": 0.86, "learning_rate": 5.3408833356770426e-05, "loss": 1.1154, "step": 8245 }, { "epoch": 0.86, "learning_rate": 5.340729387544333e-05, "loss": 0.3726, "step": 8246 }, { "epoch": 0.86, "learning_rate": 5.3405754236543814e-05, "loss": 0.6631, "step": 8247 }, { "epoch": 0.86, "learning_rate": 5.3404214440082233e-05, "loss": 1.4374, "step": 8248 }, { "epoch": 0.86, "learning_rate": 5.340267448606897e-05, "loss": 0.7176, "step": 8249 }, { "epoch": 0.86, "learning_rate": 5.3401134374514356e-05, "loss": 1.3554, "step": 8250 }, { "epoch": 0.86, "learning_rate": 5.3399594105428795e-05, "loss": 1.0273, "step": 8251 }, { "epoch": 0.86, "learning_rate": 5.339805367882264e-05, "loss": 0.6884, "step": 8252 }, { "epoch": 0.86, "learning_rate": 5.339651309470626e-05, "loss": 0.8046, "step": 8253 }, { "epoch": 0.86, "learning_rate": 5.339497235309003e-05, "loss": 0.7999, "step": 8254 }, { "epoch": 0.86, "learning_rate": 5.339343145398432e-05, "loss": 0.7223, "step": 8255 }, { "epoch": 0.86, "learning_rate": 5.339189039739949e-05, "loss": 1.4204, "step": 8256 }, { "epoch": 0.86, "learning_rate": 5.339034918334594e-05, "loss": 0.8214, "step": 8257 }, { "epoch": 0.86, "learning_rate": 5.338880781183404e-05, "loss": 0.9272, "step": 8258 }, { "epoch": 0.86, "learning_rate": 5.338726628287415e-05, "loss": 1.2063, "step": 8259 }, { "epoch": 0.86, "learning_rate": 5.338572459647665e-05, "loss": 0.8955, "step": 8260 }, { "epoch": 0.86, "learning_rate": 5.338418275265193e-05, "loss": 0.4817, "step": 8261 }, { "epoch": 0.86, "learning_rate": 5.338264075141037e-05, "loss": 0.5359, "step": 8262 }, { "epoch": 0.86, "learning_rate": 5.3381098592762334e-05, "loss": 1.3216, "step": 8263 }, { "epoch": 0.86, "learning_rate": 5.3379556276718216e-05, "loss": 0.6101, "step": 8264 }, { "epoch": 0.86, "learning_rate": 5.337801380328839e-05, "loss": 0.9263, "step": 8265 }, { "epoch": 0.86, "learning_rate": 5.337647117248325e-05, "loss": 0.3414, "step": 8266 }, { "epoch": 0.86, "learning_rate": 5.337492838431317e-05, "loss": 0.5465, "step": 8267 }, { "epoch": 0.86, "learning_rate": 5.3373385438788544e-05, "loss": 0.6552, "step": 8268 }, { "epoch": 0.86, "learning_rate": 5.3371842335919766e-05, "loss": 0.5641, "step": 8269 }, { "epoch": 0.86, "learning_rate": 5.337029907571721e-05, "loss": 0.4602, "step": 8270 }, { "epoch": 0.86, "learning_rate": 5.336875565819126e-05, "loss": 0.5717, "step": 8271 }, { "epoch": 0.86, "learning_rate": 5.336721208335232e-05, "loss": 1.6559, "step": 8272 }, { "epoch": 0.86, "learning_rate": 5.3365668351210775e-05, "loss": 1.5237, "step": 8273 }, { "epoch": 0.86, "learning_rate": 5.336412446177702e-05, "loss": 0.7684, "step": 8274 }, { "epoch": 0.86, "learning_rate": 5.3362580415061447e-05, "loss": 1.1684, "step": 8275 }, { "epoch": 0.86, "learning_rate": 5.336103621107444e-05, "loss": 0.4972, "step": 8276 }, { "epoch": 0.86, "learning_rate": 5.335949184982641e-05, "loss": 1.2138, "step": 8277 }, { "epoch": 0.86, "learning_rate": 5.335794733132775e-05, "loss": 0.7439, "step": 8278 }, { "epoch": 0.86, "learning_rate": 5.3356402655588846e-05, "loss": 0.6812, "step": 8279 }, { "epoch": 0.86, "learning_rate": 5.335485782262011e-05, "loss": 0.5832, "step": 8280 }, { "epoch": 0.86, "learning_rate": 5.335331283243194e-05, "loss": 0.8937, "step": 8281 }, { "epoch": 0.86, "learning_rate": 5.335176768503472e-05, "loss": 0.9153, "step": 8282 }, { "epoch": 0.87, "learning_rate": 5.3350222380438875e-05, "loss": 1.2659, "step": 8283 }, { "epoch": 0.87, "learning_rate": 5.334867691865479e-05, "loss": 0.8702, "step": 8284 }, { "epoch": 0.87, "learning_rate": 5.334713129969288e-05, "loss": 0.4074, "step": 8285 }, { "epoch": 0.87, "learning_rate": 5.334558552356354e-05, "loss": 0.8625, "step": 8286 }, { "epoch": 0.87, "learning_rate": 5.334403959027718e-05, "loss": 0.3053, "step": 8287 }, { "epoch": 0.87, "learning_rate": 5.334249349984421e-05, "loss": 1.4848, "step": 8288 }, { "epoch": 0.87, "learning_rate": 5.334094725227504e-05, "loss": 0.7855, "step": 8289 }, { "epoch": 0.87, "learning_rate": 5.333940084758007e-05, "loss": 0.6335, "step": 8290 }, { "epoch": 0.87, "learning_rate": 5.3337854285769716e-05, "loss": 0.3528, "step": 8291 }, { "epoch": 0.87, "learning_rate": 5.333630756685439e-05, "loss": 1.1551, "step": 8292 }, { "epoch": 0.87, "learning_rate": 5.333476069084449e-05, "loss": 0.9824, "step": 8293 }, { "epoch": 0.87, "learning_rate": 5.333321365775045e-05, "loss": 0.8183, "step": 8294 }, { "epoch": 0.87, "learning_rate": 5.3331666467582675e-05, "loss": 0.9586, "step": 8295 }, { "epoch": 0.87, "learning_rate": 5.333011912035158e-05, "loss": 0.9027, "step": 8296 }, { "epoch": 0.87, "learning_rate": 5.332857161606758e-05, "loss": 0.6558, "step": 8297 }, { "epoch": 0.87, "learning_rate": 5.33270239547411e-05, "loss": 1.1462, "step": 8298 }, { "epoch": 0.87, "learning_rate": 5.332547613638255e-05, "loss": 0.5478, "step": 8299 }, { "epoch": 0.87, "learning_rate": 5.3323928161002345e-05, "loss": 1.5972, "step": 8300 }, { "epoch": 0.87, "learning_rate": 5.332238002861092e-05, "loss": 0.9714, "step": 8301 }, { "epoch": 0.87, "learning_rate": 5.332083173921869e-05, "loss": 0.6635, "step": 8302 }, { "epoch": 0.87, "learning_rate": 5.331928329283607e-05, "loss": 1.1901, "step": 8303 }, { "epoch": 0.87, "learning_rate": 5.3317734689473495e-05, "loss": 0.6584, "step": 8304 }, { "epoch": 0.87, "learning_rate": 5.3316185929141396e-05, "loss": 0.3732, "step": 8305 }, { "epoch": 0.87, "learning_rate": 5.331463701185018e-05, "loss": 1.1286, "step": 8306 }, { "epoch": 0.87, "learning_rate": 5.331308793761029e-05, "loss": 0.3594, "step": 8307 }, { "epoch": 0.87, "learning_rate": 5.3311538706432136e-05, "loss": 1.4853, "step": 8308 }, { "epoch": 0.87, "learning_rate": 5.330998931832616e-05, "loss": 0.8601, "step": 8309 }, { "epoch": 0.87, "learning_rate": 5.3308439773302794e-05, "loss": 0.9801, "step": 8310 }, { "epoch": 0.87, "learning_rate": 5.330689007137246e-05, "loss": 0.8002, "step": 8311 }, { "epoch": 0.87, "learning_rate": 5.33053402125456e-05, "loss": 0.5415, "step": 8312 }, { "epoch": 0.87, "learning_rate": 5.3303790196832644e-05, "loss": 0.2541, "step": 8313 }, { "epoch": 0.87, "learning_rate": 5.330224002424402e-05, "loss": 0.4118, "step": 8314 }, { "epoch": 0.87, "learning_rate": 5.330068969479017e-05, "loss": 1.3481, "step": 8315 }, { "epoch": 0.87, "learning_rate": 5.3299139208481526e-05, "loss": 0.7512, "step": 8316 }, { "epoch": 0.87, "learning_rate": 5.329758856532854e-05, "loss": 0.9087, "step": 8317 }, { "epoch": 0.87, "learning_rate": 5.3296037765341634e-05, "loss": 1.1354, "step": 8318 }, { "epoch": 0.87, "learning_rate": 5.3294486808531245e-05, "loss": 1.5386, "step": 8319 }, { "epoch": 0.87, "learning_rate": 5.329293569490783e-05, "loss": 0.718, "step": 8320 }, { "epoch": 0.87, "learning_rate": 5.329138442448181e-05, "loss": 0.3902, "step": 8321 }, { "epoch": 0.87, "learning_rate": 5.328983299726365e-05, "loss": 1.1054, "step": 8322 }, { "epoch": 0.87, "learning_rate": 5.328828141326379e-05, "loss": 1.5218, "step": 8323 }, { "epoch": 0.87, "learning_rate": 5.328672967249265e-05, "loss": 0.8615, "step": 8324 }, { "epoch": 0.87, "learning_rate": 5.3285177774960716e-05, "loss": 0.9248, "step": 8325 }, { "epoch": 0.87, "learning_rate": 5.32836257206784e-05, "loss": 1.2952, "step": 8326 }, { "epoch": 0.87, "learning_rate": 5.328207350965617e-05, "loss": 0.8884, "step": 8327 }, { "epoch": 0.87, "learning_rate": 5.328052114190446e-05, "loss": 1.0639, "step": 8328 }, { "epoch": 0.87, "learning_rate": 5.327896861743374e-05, "loss": 0.3085, "step": 8329 }, { "epoch": 0.87, "learning_rate": 5.327741593625444e-05, "loss": 1.0324, "step": 8330 }, { "epoch": 0.87, "learning_rate": 5.3275863098377024e-05, "loss": 0.4598, "step": 8331 }, { "epoch": 0.87, "learning_rate": 5.3274310103811945e-05, "loss": 1.4498, "step": 8332 }, { "epoch": 0.87, "learning_rate": 5.327275695256966e-05, "loss": 1.6669, "step": 8333 }, { "epoch": 0.87, "learning_rate": 5.3271203644660615e-05, "loss": 0.5286, "step": 8334 }, { "epoch": 0.87, "learning_rate": 5.326965018009527e-05, "loss": 0.2713, "step": 8335 }, { "epoch": 0.87, "learning_rate": 5.32680965588841e-05, "loss": 1.2976, "step": 8336 }, { "epoch": 0.87, "learning_rate": 5.326654278103753e-05, "loss": 0.4223, "step": 8337 }, { "epoch": 0.87, "learning_rate": 5.326498884656605e-05, "loss": 0.4063, "step": 8338 }, { "epoch": 0.87, "learning_rate": 5.32634347554801e-05, "loss": 0.8377, "step": 8339 }, { "epoch": 0.87, "learning_rate": 5.326188050779016e-05, "loss": 0.1055, "step": 8340 }, { "epoch": 0.87, "learning_rate": 5.3260326103506674e-05, "loss": 1.4203, "step": 8341 }, { "epoch": 0.87, "learning_rate": 5.3258771542640124e-05, "loss": 0.5705, "step": 8342 }, { "epoch": 0.87, "learning_rate": 5.325721682520096e-05, "loss": 0.9363, "step": 8343 }, { "epoch": 0.87, "learning_rate": 5.3255661951199655e-05, "loss": 0.5503, "step": 8344 }, { "epoch": 0.87, "learning_rate": 5.325410692064667e-05, "loss": 0.7415, "step": 8345 }, { "epoch": 0.87, "learning_rate": 5.3252551733552486e-05, "loss": 0.2518, "step": 8346 }, { "epoch": 0.87, "learning_rate": 5.325099638992756e-05, "loss": 0.6251, "step": 8347 }, { "epoch": 0.87, "learning_rate": 5.324944088978237e-05, "loss": 0.4259, "step": 8348 }, { "epoch": 0.87, "learning_rate": 5.324788523312738e-05, "loss": 0.4591, "step": 8349 }, { "epoch": 0.87, "learning_rate": 5.3246329419973075e-05, "loss": 1.3964, "step": 8350 }, { "epoch": 0.87, "learning_rate": 5.324477345032991e-05, "loss": 0.813, "step": 8351 }, { "epoch": 0.87, "learning_rate": 5.324321732420837e-05, "loss": 0.7136, "step": 8352 }, { "epoch": 0.87, "learning_rate": 5.324166104161893e-05, "loss": 1.0383, "step": 8353 }, { "epoch": 0.87, "learning_rate": 5.324010460257207e-05, "loss": 1.0352, "step": 8354 }, { "epoch": 0.87, "learning_rate": 5.323854800707826e-05, "loss": 0.5958, "step": 8355 }, { "epoch": 0.87, "learning_rate": 5.323699125514799e-05, "loss": 1.2957, "step": 8356 }, { "epoch": 0.87, "learning_rate": 5.323543434679172e-05, "loss": 0.4764, "step": 8357 }, { "epoch": 0.87, "learning_rate": 5.323387728201995e-05, "loss": 1.0759, "step": 8358 }, { "epoch": 0.87, "learning_rate": 5.323232006084315e-05, "loss": 1.3316, "step": 8359 }, { "epoch": 0.87, "learning_rate": 5.3230762683271813e-05, "loss": 0.7191, "step": 8360 }, { "epoch": 0.87, "learning_rate": 5.322920514931641e-05, "loss": 1.1678, "step": 8361 }, { "epoch": 0.87, "learning_rate": 5.3227647458987436e-05, "loss": 0.9009, "step": 8362 }, { "epoch": 0.87, "learning_rate": 5.3226089612295376e-05, "loss": 0.7873, "step": 8363 }, { "epoch": 0.87, "learning_rate": 5.322453160925072e-05, "loss": 0.7475, "step": 8364 }, { "epoch": 0.87, "learning_rate": 5.322297344986394e-05, "loss": 1.0673, "step": 8365 }, { "epoch": 0.87, "learning_rate": 5.322141513414555e-05, "loss": 1.4841, "step": 8366 }, { "epoch": 0.87, "learning_rate": 5.3219856662106005e-05, "loss": 0.8773, "step": 8367 }, { "epoch": 0.87, "learning_rate": 5.3218298033755835e-05, "loss": 0.8705, "step": 8368 }, { "epoch": 0.87, "learning_rate": 5.321673924910552e-05, "loss": 1.2916, "step": 8369 }, { "epoch": 0.87, "learning_rate": 5.321518030816554e-05, "loss": 0.6009, "step": 8370 }, { "epoch": 0.87, "learning_rate": 5.32136212109464e-05, "loss": 0.3811, "step": 8371 }, { "epoch": 0.87, "learning_rate": 5.321206195745859e-05, "loss": 0.6647, "step": 8372 }, { "epoch": 0.87, "learning_rate": 5.3210502547712605e-05, "loss": 0.7022, "step": 8373 }, { "epoch": 0.87, "learning_rate": 5.320894298171896e-05, "loss": 0.3357, "step": 8374 }, { "epoch": 0.87, "learning_rate": 5.3207383259488145e-05, "loss": 1.3725, "step": 8375 }, { "epoch": 0.87, "learning_rate": 5.3205823381030644e-05, "loss": 0.2222, "step": 8376 }, { "epoch": 0.87, "learning_rate": 5.320426334635697e-05, "loss": 1.374, "step": 8377 }, { "epoch": 0.87, "learning_rate": 5.320270315547763e-05, "loss": 0.3946, "step": 8378 }, { "epoch": 0.88, "learning_rate": 5.320114280840311e-05, "loss": 0.5908, "step": 8379 }, { "epoch": 0.88, "learning_rate": 5.319958230514394e-05, "loss": 0.7375, "step": 8380 }, { "epoch": 0.88, "learning_rate": 5.31980216457106e-05, "loss": 2.0989, "step": 8381 }, { "epoch": 0.88, "learning_rate": 5.319646083011362e-05, "loss": 0.7429, "step": 8382 }, { "epoch": 0.88, "learning_rate": 5.319489985836348e-05, "loss": 0.4426, "step": 8383 }, { "epoch": 0.88, "learning_rate": 5.3193338730470705e-05, "loss": 1.1001, "step": 8384 }, { "epoch": 0.88, "learning_rate": 5.31917774464458e-05, "loss": 0.4892, "step": 8385 }, { "epoch": 0.88, "learning_rate": 5.319021600629928e-05, "loss": 0.9541, "step": 8386 }, { "epoch": 0.88, "learning_rate": 5.318865441004164e-05, "loss": 1.0896, "step": 8387 }, { "epoch": 0.88, "learning_rate": 5.318709265768342e-05, "loss": 1.4224, "step": 8388 }, { "epoch": 0.88, "learning_rate": 5.318553074923511e-05, "loss": 0.27, "step": 8389 }, { "epoch": 0.88, "learning_rate": 5.3183968684707225e-05, "loss": 0.2365, "step": 8390 }, { "epoch": 0.88, "learning_rate": 5.3182406464110295e-05, "loss": 0.8658, "step": 8391 }, { "epoch": 0.88, "learning_rate": 5.3180844087454835e-05, "loss": 0.4083, "step": 8392 }, { "epoch": 0.88, "learning_rate": 5.317928155475135e-05, "loss": 1.1369, "step": 8393 }, { "epoch": 0.88, "learning_rate": 5.317771886601036e-05, "loss": 1.097, "step": 8394 }, { "epoch": 0.88, "learning_rate": 5.31761560212424e-05, "loss": 1.1225, "step": 8395 }, { "epoch": 0.88, "learning_rate": 5.317459302045797e-05, "loss": 0.7997, "step": 8396 }, { "epoch": 0.88, "learning_rate": 5.3173029863667604e-05, "loss": 0.4381, "step": 8397 }, { "epoch": 0.88, "learning_rate": 5.3171466550881834e-05, "loss": 0.9732, "step": 8398 }, { "epoch": 0.88, "learning_rate": 5.316990308211116e-05, "loss": 1.1775, "step": 8399 }, { "epoch": 0.88, "learning_rate": 5.316833945736613e-05, "loss": 1.2726, "step": 8400 }, { "epoch": 0.88, "learning_rate": 5.3166775676657256e-05, "loss": 1.0241, "step": 8401 }, { "epoch": 0.88, "learning_rate": 5.316521173999506e-05, "loss": 0.5646, "step": 8402 }, { "epoch": 0.88, "learning_rate": 5.316364764739009e-05, "loss": 0.8811, "step": 8403 }, { "epoch": 0.88, "learning_rate": 5.316208339885286e-05, "loss": 1.1989, "step": 8404 }, { "epoch": 0.88, "learning_rate": 5.31605189943939e-05, "loss": 1.1167, "step": 8405 }, { "epoch": 0.88, "learning_rate": 5.315895443402376e-05, "loss": 1.0238, "step": 8406 }, { "epoch": 0.88, "learning_rate": 5.3157389717752935e-05, "loss": 1.321, "step": 8407 }, { "epoch": 0.88, "learning_rate": 5.3155824845592004e-05, "loss": 0.2663, "step": 8408 }, { "epoch": 0.88, "learning_rate": 5.3154259817551466e-05, "loss": 1.4388, "step": 8409 }, { "epoch": 0.88, "learning_rate": 5.315269463364187e-05, "loss": 1.3725, "step": 8410 }, { "epoch": 0.88, "learning_rate": 5.3151129293873744e-05, "loss": 0.7334, "step": 8411 }, { "epoch": 0.88, "learning_rate": 5.314956379825764e-05, "loss": 1.6829, "step": 8412 }, { "epoch": 0.88, "learning_rate": 5.314799814680409e-05, "loss": 1.2847, "step": 8413 }, { "epoch": 0.88, "learning_rate": 5.3146432339523634e-05, "loss": 0.3581, "step": 8414 }, { "epoch": 0.88, "learning_rate": 5.31448663764268e-05, "loss": 0.8929, "step": 8415 }, { "epoch": 0.88, "learning_rate": 5.3143300257524155e-05, "loss": 0.7587, "step": 8416 }, { "epoch": 0.88, "learning_rate": 5.3141733982826223e-05, "loss": 0.5128, "step": 8417 }, { "epoch": 0.88, "learning_rate": 5.314016755234355e-05, "loss": 0.8915, "step": 8418 }, { "epoch": 0.88, "learning_rate": 5.313860096608669e-05, "loss": 0.9544, "step": 8419 }, { "epoch": 0.88, "learning_rate": 5.313703422406617e-05, "loss": 1.4859, "step": 8420 }, { "epoch": 0.88, "learning_rate": 5.3135467326292564e-05, "loss": 0.9201, "step": 8421 }, { "epoch": 0.88, "learning_rate": 5.313390027277639e-05, "loss": 0.7769, "step": 8422 }, { "epoch": 0.88, "learning_rate": 5.313233306352822e-05, "loss": 0.6422, "step": 8423 }, { "epoch": 0.88, "learning_rate": 5.3130765698558596e-05, "loss": 1.105, "step": 8424 }, { "epoch": 0.88, "learning_rate": 5.312919817787807e-05, "loss": 1.1712, "step": 8425 }, { "epoch": 0.88, "learning_rate": 5.31276305014972e-05, "loss": 0.4782, "step": 8426 }, { "epoch": 0.88, "learning_rate": 5.3126062669426525e-05, "loss": 0.8964, "step": 8427 }, { "epoch": 0.88, "learning_rate": 5.3124494681676605e-05, "loss": 0.7605, "step": 8428 }, { "epoch": 0.88, "learning_rate": 5.3122926538258e-05, "loss": 1.168, "step": 8429 }, { "epoch": 0.88, "learning_rate": 5.312135823918127e-05, "loss": 0.1315, "step": 8430 }, { "epoch": 0.88, "learning_rate": 5.311978978445696e-05, "loss": 0.2476, "step": 8431 }, { "epoch": 0.88, "learning_rate": 5.3118221174095635e-05, "loss": 0.6361, "step": 8432 }, { "epoch": 0.88, "learning_rate": 5.3116652408107855e-05, "loss": 0.9651, "step": 8433 }, { "epoch": 0.88, "learning_rate": 5.3115083486504184e-05, "loss": 1.2978, "step": 8434 }, { "epoch": 0.88, "learning_rate": 5.311351440929517e-05, "loss": 0.2763, "step": 8435 }, { "epoch": 0.88, "learning_rate": 5.311194517649139e-05, "loss": 0.6841, "step": 8436 }, { "epoch": 0.88, "learning_rate": 5.311037578810341e-05, "loss": 1.7573, "step": 8437 }, { "epoch": 0.88, "learning_rate": 5.3108806244141775e-05, "loss": 1.0915, "step": 8438 }, { "epoch": 0.88, "learning_rate": 5.3107236544617076e-05, "loss": 0.3561, "step": 8439 }, { "epoch": 0.88, "learning_rate": 5.310566668953986e-05, "loss": 1.0402, "step": 8440 }, { "epoch": 0.88, "learning_rate": 5.31040966789207e-05, "loss": 1.149, "step": 8441 }, { "epoch": 0.88, "learning_rate": 5.310252651277017e-05, "loss": 0.451, "step": 8442 }, { "epoch": 0.88, "learning_rate": 5.310095619109884e-05, "loss": 0.8674, "step": 8443 }, { "epoch": 0.88, "learning_rate": 5.309938571391727e-05, "loss": 1.3727, "step": 8444 }, { "epoch": 0.88, "learning_rate": 5.309781508123605e-05, "loss": 0.9443, "step": 8445 }, { "epoch": 0.88, "learning_rate": 5.309624429306574e-05, "loss": 1.1865, "step": 8446 }, { "epoch": 0.88, "learning_rate": 5.3094673349416904e-05, "loss": 0.8549, "step": 8447 }, { "epoch": 0.88, "learning_rate": 5.309310225030015e-05, "loss": 0.6023, "step": 8448 }, { "epoch": 0.88, "learning_rate": 5.3091530995726024e-05, "loss": 1.5883, "step": 8449 }, { "epoch": 0.88, "learning_rate": 5.308995958570512e-05, "loss": 1.5442, "step": 8450 }, { "epoch": 0.88, "learning_rate": 5.3088388020248016e-05, "loss": 0.9315, "step": 8451 }, { "epoch": 0.88, "learning_rate": 5.308681629936528e-05, "loss": 2.8133, "step": 8452 }, { "epoch": 0.88, "learning_rate": 5.308524442306749e-05, "loss": 1.2161, "step": 8453 }, { "epoch": 0.88, "learning_rate": 5.308367239136525e-05, "loss": 0.6885, "step": 8454 }, { "epoch": 0.88, "learning_rate": 5.308210020426912e-05, "loss": 1.1337, "step": 8455 }, { "epoch": 0.88, "learning_rate": 5.30805278617897e-05, "loss": 0.4591, "step": 8456 }, { "epoch": 0.88, "learning_rate": 5.307895536393755e-05, "loss": 0.335, "step": 8457 }, { "epoch": 0.88, "learning_rate": 5.307738271072329e-05, "loss": 0.4348, "step": 8458 }, { "epoch": 0.88, "learning_rate": 5.307580990215748e-05, "loss": 1.2929, "step": 8459 }, { "epoch": 0.88, "learning_rate": 5.307423693825072e-05, "loss": 1.2055, "step": 8460 }, { "epoch": 0.88, "learning_rate": 5.30726638190136e-05, "loss": 0.7773, "step": 8461 }, { "epoch": 0.88, "learning_rate": 5.30710905444567e-05, "loss": 1.63, "step": 8462 }, { "epoch": 0.88, "learning_rate": 5.306951711459062e-05, "loss": 0.3596, "step": 8463 }, { "epoch": 0.88, "learning_rate": 5.306794352942595e-05, "loss": 1.2991, "step": 8464 }, { "epoch": 0.88, "learning_rate": 5.3066369788973276e-05, "loss": 0.3604, "step": 8465 }, { "epoch": 0.88, "learning_rate": 5.3064795893243196e-05, "loss": 1.1121, "step": 8466 }, { "epoch": 0.88, "learning_rate": 5.306322184224631e-05, "loss": 1.8257, "step": 8467 }, { "epoch": 0.88, "learning_rate": 5.306164763599322e-05, "loss": 0.4097, "step": 8468 }, { "epoch": 0.88, "learning_rate": 5.306007327449451e-05, "loss": 0.8243, "step": 8469 }, { "epoch": 0.88, "learning_rate": 5.305849875776077e-05, "loss": 1.0377, "step": 8470 }, { "epoch": 0.88, "learning_rate": 5.305692408580262e-05, "loss": 1.4328, "step": 8471 }, { "epoch": 0.88, "learning_rate": 5.305534925863065e-05, "loss": 0.3805, "step": 8472 }, { "epoch": 0.88, "learning_rate": 5.305377427625547e-05, "loss": 1.3078, "step": 8473 }, { "epoch": 0.89, "learning_rate": 5.305219913868767e-05, "loss": 1.7143, "step": 8474 }, { "epoch": 0.89, "learning_rate": 5.305062384593786e-05, "loss": 0.7299, "step": 8475 }, { "epoch": 0.89, "learning_rate": 5.3049048398016646e-05, "loss": 0.4809, "step": 8476 }, { "epoch": 0.89, "learning_rate": 5.304747279493463e-05, "loss": 0.7378, "step": 8477 }, { "epoch": 0.89, "learning_rate": 5.304589703670241e-05, "loss": 0.2925, "step": 8478 }, { "epoch": 0.89, "learning_rate": 5.304432112333061e-05, "loss": 0.6775, "step": 8479 }, { "epoch": 0.89, "learning_rate": 5.304274505482984e-05, "loss": 0.2728, "step": 8480 }, { "epoch": 0.89, "learning_rate": 5.30411688312107e-05, "loss": 0.2191, "step": 8481 }, { "epoch": 0.89, "learning_rate": 5.30395924524838e-05, "loss": 1.1016, "step": 8482 }, { "epoch": 0.89, "learning_rate": 5.303801591865975e-05, "loss": 1.0102, "step": 8483 }, { "epoch": 0.89, "learning_rate": 5.3036439229749175e-05, "loss": 0.3732, "step": 8484 }, { "epoch": 0.89, "learning_rate": 5.3034862385762676e-05, "loss": 1.276, "step": 8485 }, { "epoch": 0.89, "learning_rate": 5.303328538671088e-05, "loss": 1.4873, "step": 8486 }, { "epoch": 0.89, "learning_rate": 5.3031708232604386e-05, "loss": 0.4721, "step": 8487 }, { "epoch": 0.89, "learning_rate": 5.303013092345382e-05, "loss": 1.3053, "step": 8488 }, { "epoch": 0.89, "learning_rate": 5.302855345926981e-05, "loss": 0.5933, "step": 8489 }, { "epoch": 0.89, "learning_rate": 5.302697584006297e-05, "loss": 0.5727, "step": 8490 }, { "epoch": 0.89, "learning_rate": 5.3025398065843906e-05, "loss": 1.4532, "step": 8491 }, { "epoch": 0.89, "learning_rate": 5.302382013662325e-05, "loss": 1.0742, "step": 8492 }, { "epoch": 0.89, "learning_rate": 5.302224205241163e-05, "loss": 0.9412, "step": 8493 }, { "epoch": 0.89, "learning_rate": 5.3020663813219654e-05, "loss": 0.9744, "step": 8494 }, { "epoch": 0.89, "learning_rate": 5.3019085419057964e-05, "loss": 1.0734, "step": 8495 }, { "epoch": 0.89, "learning_rate": 5.301750686993718e-05, "loss": 1.3403, "step": 8496 }, { "epoch": 0.89, "learning_rate": 5.3015928165867914e-05, "loss": 1.2801, "step": 8497 }, { "epoch": 0.89, "learning_rate": 5.301434930686082e-05, "loss": 0.7837, "step": 8498 }, { "epoch": 0.89, "learning_rate": 5.3012770292926504e-05, "loss": 0.7007, "step": 8499 }, { "epoch": 0.89, "learning_rate": 5.3011191124075595e-05, "loss": 1.4481, "step": 8500 }, { "epoch": 0.89, "learning_rate": 5.3009611800318746e-05, "loss": 1.0982, "step": 8501 }, { "epoch": 0.89, "learning_rate": 5.300803232166656e-05, "loss": 1.3873, "step": 8502 }, { "epoch": 0.89, "learning_rate": 5.3006452688129685e-05, "loss": 0.3775, "step": 8503 }, { "epoch": 0.89, "learning_rate": 5.300487289971877e-05, "loss": 0.8809, "step": 8504 }, { "epoch": 0.89, "learning_rate": 5.3003292956444416e-05, "loss": 0.4397, "step": 8505 }, { "epoch": 0.89, "learning_rate": 5.300171285831729e-05, "loss": 1.4465, "step": 8506 }, { "epoch": 0.89, "learning_rate": 5.300013260534801e-05, "loss": 0.3988, "step": 8507 }, { "epoch": 0.89, "learning_rate": 5.2998552197547206e-05, "loss": 0.7051, "step": 8508 }, { "epoch": 0.89, "learning_rate": 5.2996971634925545e-05, "loss": 0.4958, "step": 8509 }, { "epoch": 0.89, "learning_rate": 5.299539091749365e-05, "loss": 0.4337, "step": 8510 }, { "epoch": 0.89, "learning_rate": 5.299381004526216e-05, "loss": 1.0402, "step": 8511 }, { "epoch": 0.89, "learning_rate": 5.299222901824172e-05, "loss": 0.665, "step": 8512 }, { "epoch": 0.89, "learning_rate": 5.299064783644298e-05, "loss": 1.6077, "step": 8513 }, { "epoch": 0.89, "learning_rate": 5.298906649987657e-05, "loss": 0.4209, "step": 8514 }, { "epoch": 0.89, "learning_rate": 5.298748500855315e-05, "loss": 0.519, "step": 8515 }, { "epoch": 0.89, "learning_rate": 5.298590336248336e-05, "loss": 1.0415, "step": 8516 }, { "epoch": 0.89, "learning_rate": 5.298432156167784e-05, "loss": 0.8523, "step": 8517 }, { "epoch": 0.89, "learning_rate": 5.298273960614725e-05, "loss": 1.3316, "step": 8518 }, { "epoch": 0.89, "learning_rate": 5.298115749590224e-05, "loss": 0.4179, "step": 8519 }, { "epoch": 0.89, "learning_rate": 5.297957523095344e-05, "loss": 0.6221, "step": 8520 }, { "epoch": 0.89, "learning_rate": 5.2977992811311526e-05, "loss": 0.7999, "step": 8521 }, { "epoch": 0.89, "learning_rate": 5.297641023698714e-05, "loss": 1.5548, "step": 8522 }, { "epoch": 0.89, "learning_rate": 5.2974827507990934e-05, "loss": 0.8556, "step": 8523 }, { "epoch": 0.89, "learning_rate": 5.297324462433357e-05, "loss": 0.6108, "step": 8524 }, { "epoch": 0.89, "learning_rate": 5.2971661586025694e-05, "loss": 0.3048, "step": 8525 }, { "epoch": 0.89, "learning_rate": 5.297007839307797e-05, "loss": 1.2885, "step": 8526 }, { "epoch": 0.89, "learning_rate": 5.296849504550105e-05, "loss": 0.699, "step": 8527 }, { "epoch": 0.89, "learning_rate": 5.2966911543305594e-05, "loss": 1.3244, "step": 8528 }, { "epoch": 0.89, "learning_rate": 5.296532788650226e-05, "loss": 1.1551, "step": 8529 }, { "epoch": 0.89, "learning_rate": 5.296374407510173e-05, "loss": 1.6329, "step": 8530 }, { "epoch": 0.89, "learning_rate": 5.296216010911463e-05, "loss": 0.787, "step": 8531 }, { "epoch": 0.89, "learning_rate": 5.296057598855164e-05, "loss": 0.7365, "step": 8532 }, { "epoch": 0.89, "learning_rate": 5.295899171342344e-05, "loss": 1.0673, "step": 8533 }, { "epoch": 0.89, "learning_rate": 5.295740728374067e-05, "loss": 0.3423, "step": 8534 }, { "epoch": 0.89, "learning_rate": 5.2955822699514007e-05, "loss": 1.531, "step": 8535 }, { "epoch": 0.89, "learning_rate": 5.295423796075412e-05, "loss": 0.4989, "step": 8536 }, { "epoch": 0.89, "learning_rate": 5.2952653067471667e-05, "loss": 0.3009, "step": 8537 }, { "epoch": 0.89, "learning_rate": 5.295106801967733e-05, "loss": 1.0761, "step": 8538 }, { "epoch": 0.89, "learning_rate": 5.2949482817381776e-05, "loss": 1.767, "step": 8539 }, { "epoch": 0.89, "learning_rate": 5.2947897460595666e-05, "loss": 1.0742, "step": 8540 }, { "epoch": 0.89, "learning_rate": 5.294631194932968e-05, "loss": 0.5943, "step": 8541 }, { "epoch": 0.89, "learning_rate": 5.29447262835945e-05, "loss": 1.1742, "step": 8542 }, { "epoch": 0.89, "learning_rate": 5.2943140463400776e-05, "loss": 1.0688, "step": 8543 }, { "epoch": 0.89, "learning_rate": 5.2941554488759216e-05, "loss": 0.3075, "step": 8544 }, { "epoch": 0.89, "learning_rate": 5.293996835968047e-05, "loss": 0.5918, "step": 8545 }, { "epoch": 0.89, "learning_rate": 5.293838207617522e-05, "loss": 0.524, "step": 8546 }, { "epoch": 0.89, "learning_rate": 5.2936795638254156e-05, "loss": 1.5568, "step": 8547 }, { "epoch": 0.89, "learning_rate": 5.293520904592795e-05, "loss": 1.1639, "step": 8548 }, { "epoch": 0.89, "learning_rate": 5.293362229920728e-05, "loss": 0.6615, "step": 8549 }, { "epoch": 0.89, "learning_rate": 5.293203539810283e-05, "loss": 1.5347, "step": 8550 }, { "epoch": 0.89, "learning_rate": 5.293044834262528e-05, "loss": 0.3882, "step": 8551 }, { "epoch": 0.89, "learning_rate": 5.2928861132785324e-05, "loss": 1.2867, "step": 8552 }, { "epoch": 0.89, "learning_rate": 5.292727376859364e-05, "loss": 0.7394, "step": 8553 }, { "epoch": 0.89, "learning_rate": 5.292568625006092e-05, "loss": 1.0698, "step": 8554 }, { "epoch": 0.89, "learning_rate": 5.2924098577197834e-05, "loss": 0.4459, "step": 8555 }, { "epoch": 0.89, "learning_rate": 5.292251075001508e-05, "loss": 1.1625, "step": 8556 }, { "epoch": 0.89, "learning_rate": 5.292092276852335e-05, "loss": 1.1373, "step": 8557 }, { "epoch": 0.89, "learning_rate": 5.2919334632733333e-05, "loss": 0.5838, "step": 8558 }, { "epoch": 0.89, "learning_rate": 5.291774634265572e-05, "loss": 0.7605, "step": 8559 }, { "epoch": 0.89, "learning_rate": 5.291615789830119e-05, "loss": 1.0786, "step": 8560 }, { "epoch": 0.89, "learning_rate": 5.291456929968046e-05, "loss": 1.3403, "step": 8561 }, { "epoch": 0.89, "learning_rate": 5.2912980546804205e-05, "loss": 1.2611, "step": 8562 }, { "epoch": 0.89, "learning_rate": 5.291139163968312e-05, "loss": 0.5923, "step": 8563 }, { "epoch": 0.89, "learning_rate": 5.290980257832792e-05, "loss": 1.1337, "step": 8564 }, { "epoch": 0.89, "learning_rate": 5.290821336274928e-05, "loss": 0.1142, "step": 8565 }, { "epoch": 0.89, "learning_rate": 5.290662399295791e-05, "loss": 0.697, "step": 8566 }, { "epoch": 0.89, "learning_rate": 5.29050344689645e-05, "loss": 1.3541, "step": 8567 }, { "epoch": 0.89, "learning_rate": 5.2903444790779774e-05, "loss": 0.3715, "step": 8568 }, { "epoch": 0.89, "learning_rate": 5.29018549584144e-05, "loss": 0.8392, "step": 8569 }, { "epoch": 0.9, "learning_rate": 5.2900264971879106e-05, "loss": 0.5439, "step": 8570 }, { "epoch": 0.9, "learning_rate": 5.2898674831184586e-05, "loss": 1.5851, "step": 8571 }, { "epoch": 0.9, "learning_rate": 5.289708453634154e-05, "loss": 1.1193, "step": 8572 }, { "epoch": 0.9, "learning_rate": 5.289549408736068e-05, "loss": 0.7116, "step": 8573 }, { "epoch": 0.9, "learning_rate": 5.2893903484252714e-05, "loss": 0.8353, "step": 8574 }, { "epoch": 0.9, "learning_rate": 5.289231272702835e-05, "loss": 0.8872, "step": 8575 }, { "epoch": 0.9, "learning_rate": 5.2890721815698294e-05, "loss": 0.5707, "step": 8576 }, { "epoch": 0.9, "learning_rate": 5.288913075027324e-05, "loss": 0.3512, "step": 8577 }, { "epoch": 0.9, "learning_rate": 5.288753953076392e-05, "loss": 0.2695, "step": 8578 }, { "epoch": 0.9, "learning_rate": 5.288594815718105e-05, "loss": 1.5093, "step": 8579 }, { "epoch": 0.9, "learning_rate": 5.288435662953532e-05, "loss": 1.1793, "step": 8580 }, { "epoch": 0.9, "learning_rate": 5.288276494783746e-05, "loss": 1.0574, "step": 8581 }, { "epoch": 0.9, "learning_rate": 5.288117311209818e-05, "loss": 0.8434, "step": 8582 }, { "epoch": 0.9, "learning_rate": 5.2879581122328196e-05, "loss": 1.315, "step": 8583 }, { "epoch": 0.9, "learning_rate": 5.287798897853823e-05, "loss": 0.6691, "step": 8584 }, { "epoch": 0.9, "learning_rate": 5.287639668073899e-05, "loss": 1.0752, "step": 8585 }, { "epoch": 0.9, "learning_rate": 5.28748042289412e-05, "loss": 0.9684, "step": 8586 }, { "epoch": 0.9, "learning_rate": 5.287321162315558e-05, "loss": 0.8255, "step": 8587 }, { "epoch": 0.9, "learning_rate": 5.287161886339285e-05, "loss": 0.7543, "step": 8588 }, { "epoch": 0.9, "learning_rate": 5.287002594966374e-05, "loss": 1.7326, "step": 8589 }, { "epoch": 0.9, "learning_rate": 5.286843288197896e-05, "loss": 1.5378, "step": 8590 }, { "epoch": 0.9, "learning_rate": 5.2866839660349244e-05, "loss": 1.2812, "step": 8591 }, { "epoch": 0.9, "learning_rate": 5.2865246284785314e-05, "loss": 0.4518, "step": 8592 }, { "epoch": 0.9, "learning_rate": 5.286365275529789e-05, "loss": 1.0414, "step": 8593 }, { "epoch": 0.9, "learning_rate": 5.286205907189771e-05, "loss": 0.8899, "step": 8594 }, { "epoch": 0.9, "learning_rate": 5.28604652345955e-05, "loss": 1.4984, "step": 8595 }, { "epoch": 0.9, "learning_rate": 5.285887124340198e-05, "loss": 0.6919, "step": 8596 }, { "epoch": 0.9, "learning_rate": 5.28572770983279e-05, "loss": 1.4158, "step": 8597 }, { "epoch": 0.9, "learning_rate": 5.2855682799383976e-05, "loss": 0.9168, "step": 8598 }, { "epoch": 0.9, "learning_rate": 5.285408834658093e-05, "loss": 0.4865, "step": 8599 }, { "epoch": 0.9, "learning_rate": 5.285249373992952e-05, "loss": 0.8924, "step": 8600 }, { "epoch": 0.9, "learning_rate": 5.285089897944047e-05, "loss": 0.6778, "step": 8601 }, { "epoch": 0.9, "learning_rate": 5.28493040651245e-05, "loss": 0.9682, "step": 8602 }, { "epoch": 0.9, "learning_rate": 5.284770899699238e-05, "loss": 1.2172, "step": 8603 }, { "epoch": 0.9, "learning_rate": 5.284611377505482e-05, "loss": 0.5435, "step": 8604 }, { "epoch": 0.9, "learning_rate": 5.2844518399322564e-05, "loss": 1.1149, "step": 8605 }, { "epoch": 0.9, "learning_rate": 5.2842922869806365e-05, "loss": 0.9058, "step": 8606 }, { "epoch": 0.9, "learning_rate": 5.2841327186516945e-05, "loss": 1.413, "step": 8607 }, { "epoch": 0.9, "learning_rate": 5.283973134946506e-05, "loss": 0.6837, "step": 8608 }, { "epoch": 0.9, "learning_rate": 5.2838135358661445e-05, "loss": 0.9835, "step": 8609 }, { "epoch": 0.9, "learning_rate": 5.2836539214116846e-05, "loss": 0.594, "step": 8610 }, { "epoch": 0.9, "learning_rate": 5.2834942915842014e-05, "loss": 0.6013, "step": 8611 }, { "epoch": 0.9, "learning_rate": 5.283334646384769e-05, "loss": 0.4018, "step": 8612 }, { "epoch": 0.9, "learning_rate": 5.2831749858144615e-05, "loss": 1.7323, "step": 8613 }, { "epoch": 0.9, "learning_rate": 5.283015309874354e-05, "loss": 0.5766, "step": 8614 }, { "epoch": 0.9, "learning_rate": 5.282855618565522e-05, "loss": 1.0538, "step": 8615 }, { "epoch": 0.9, "learning_rate": 5.2826959118890394e-05, "loss": 0.4981, "step": 8616 }, { "epoch": 0.9, "learning_rate": 5.282536189845983e-05, "loss": 0.316, "step": 8617 }, { "epoch": 0.9, "learning_rate": 5.282376452437427e-05, "loss": 1.0868, "step": 8618 }, { "epoch": 0.9, "learning_rate": 5.282216699664446e-05, "loss": 0.4153, "step": 8619 }, { "epoch": 0.9, "learning_rate": 5.282056931528116e-05, "loss": 0.6169, "step": 8620 }, { "epoch": 0.9, "learning_rate": 5.281897148029514e-05, "loss": 0.2874, "step": 8621 }, { "epoch": 0.9, "learning_rate": 5.281737349169714e-05, "loss": 1.4812, "step": 8622 }, { "epoch": 0.9, "learning_rate": 5.281577534949791e-05, "loss": 0.5476, "step": 8623 }, { "epoch": 0.9, "learning_rate": 5.2814177053708235e-05, "loss": 0.7358, "step": 8624 }, { "epoch": 0.9, "learning_rate": 5.281257860433885e-05, "loss": 1.2345, "step": 8625 }, { "epoch": 0.9, "learning_rate": 5.2810980001400516e-05, "loss": 1.2107, "step": 8626 }, { "epoch": 0.9, "learning_rate": 5.280938124490401e-05, "loss": 0.4879, "step": 8627 }, { "epoch": 0.9, "learning_rate": 5.280778233486008e-05, "loss": 1.1322, "step": 8628 }, { "epoch": 0.9, "learning_rate": 5.280618327127951e-05, "loss": 1.0026, "step": 8629 }, { "epoch": 0.9, "learning_rate": 5.280458405417303e-05, "loss": 1.1475, "step": 8630 }, { "epoch": 0.9, "learning_rate": 5.280298468355144e-05, "loss": 0.8947, "step": 8631 }, { "epoch": 0.9, "learning_rate": 5.280138515942549e-05, "loss": 0.6762, "step": 8632 }, { "epoch": 0.9, "learning_rate": 5.2799785481805955e-05, "loss": 1.0434, "step": 8633 }, { "epoch": 0.9, "learning_rate": 5.279818565070359e-05, "loss": 0.7034, "step": 8634 }, { "epoch": 0.9, "learning_rate": 5.2796585666129174e-05, "loss": 0.7602, "step": 8635 }, { "epoch": 0.9, "learning_rate": 5.279498552809349e-05, "loss": 0.681, "step": 8636 }, { "epoch": 0.9, "learning_rate": 5.2793385236607284e-05, "loss": 0.8185, "step": 8637 }, { "epoch": 0.9, "learning_rate": 5.2791784791681344e-05, "loss": 0.4045, "step": 8638 }, { "epoch": 0.9, "learning_rate": 5.279018419332644e-05, "loss": 0.798, "step": 8639 }, { "epoch": 0.9, "learning_rate": 5.2788583441553354e-05, "loss": 0.5703, "step": 8640 }, { "epoch": 0.9, "learning_rate": 5.278698253637285e-05, "loss": 1.2358, "step": 8641 }, { "epoch": 0.9, "learning_rate": 5.278538147779572e-05, "loss": 1.1889, "step": 8642 }, { "epoch": 0.9, "learning_rate": 5.278378026583273e-05, "loss": 0.7142, "step": 8643 }, { "epoch": 0.9, "learning_rate": 5.2782178900494656e-05, "loss": 1.4032, "step": 8644 }, { "epoch": 0.9, "learning_rate": 5.278057738179229e-05, "loss": 0.2834, "step": 8645 }, { "epoch": 0.9, "learning_rate": 5.27789757097364e-05, "loss": 0.7662, "step": 8646 }, { "epoch": 0.9, "learning_rate": 5.277737388433779e-05, "loss": 0.2891, "step": 8647 }, { "epoch": 0.9, "learning_rate": 5.277577190560722e-05, "loss": 1.495, "step": 8648 }, { "epoch": 0.9, "learning_rate": 5.277416977355548e-05, "loss": 0.8935, "step": 8649 }, { "epoch": 0.9, "learning_rate": 5.277256748819336e-05, "loss": 1.6071, "step": 8650 }, { "epoch": 0.9, "learning_rate": 5.2770965049531645e-05, "loss": 1.9685, "step": 8651 }, { "epoch": 0.9, "learning_rate": 5.276936245758113e-05, "loss": 1.2734, "step": 8652 }, { "epoch": 0.9, "learning_rate": 5.276775971235258e-05, "loss": 1.239, "step": 8653 }, { "epoch": 0.9, "learning_rate": 5.2766156813856804e-05, "loss": 0.5053, "step": 8654 }, { "epoch": 0.9, "learning_rate": 5.276455376210458e-05, "loss": 1.6605, "step": 8655 }, { "epoch": 0.9, "learning_rate": 5.276295055710672e-05, "loss": 0.783, "step": 8656 }, { "epoch": 0.9, "learning_rate": 5.2761347198874e-05, "loss": 0.5395, "step": 8657 }, { "epoch": 0.9, "learning_rate": 5.275974368741721e-05, "loss": 1.2098, "step": 8658 }, { "epoch": 0.9, "learning_rate": 5.275814002274716e-05, "loss": 1.2642, "step": 8659 }, { "epoch": 0.9, "learning_rate": 5.2756536204874624e-05, "loss": 1.4165, "step": 8660 }, { "epoch": 0.9, "learning_rate": 5.275493223381042e-05, "loss": 0.7864, "step": 8661 }, { "epoch": 0.9, "learning_rate": 5.275332810956533e-05, "loss": 1.4068, "step": 8662 }, { "epoch": 0.9, "learning_rate": 5.2751723832150166e-05, "loss": 0.2729, "step": 8663 }, { "epoch": 0.9, "learning_rate": 5.275011940157572e-05, "loss": 1.2581, "step": 8664 }, { "epoch": 0.9, "learning_rate": 5.274851481785279e-05, "loss": 1.5083, "step": 8665 }, { "epoch": 0.91, "learning_rate": 5.274691008099218e-05, "loss": 0.8032, "step": 8666 }, { "epoch": 0.91, "learning_rate": 5.274530519100469e-05, "loss": 0.5306, "step": 8667 }, { "epoch": 0.91, "learning_rate": 5.2743700147901136e-05, "loss": 1.0354, "step": 8668 }, { "epoch": 0.91, "learning_rate": 5.274209495169231e-05, "loss": 0.9507, "step": 8669 }, { "epoch": 0.91, "learning_rate": 5.2740489602389026e-05, "loss": 1.0766, "step": 8670 }, { "epoch": 0.91, "learning_rate": 5.2738884100002086e-05, "loss": 1.0954, "step": 8671 }, { "epoch": 0.91, "learning_rate": 5.273727844454229e-05, "loss": 0.8303, "step": 8672 }, { "epoch": 0.91, "learning_rate": 5.2735672636020466e-05, "loss": 0.4374, "step": 8673 }, { "epoch": 0.91, "learning_rate": 5.27340666744474e-05, "loss": 0.3843, "step": 8674 }, { "epoch": 0.91, "learning_rate": 5.273246055983394e-05, "loss": 0.789, "step": 8675 }, { "epoch": 0.91, "learning_rate": 5.273085429219086e-05, "loss": 1.5373, "step": 8676 }, { "epoch": 0.91, "learning_rate": 5.272924787152898e-05, "loss": 0.2541, "step": 8677 }, { "epoch": 0.91, "learning_rate": 5.2727641297859134e-05, "loss": 0.8891, "step": 8678 }, { "epoch": 0.91, "learning_rate": 5.272603457119212e-05, "loss": 1.2887, "step": 8679 }, { "epoch": 0.91, "learning_rate": 5.272442769153876e-05, "loss": 0.519, "step": 8680 }, { "epoch": 0.91, "learning_rate": 5.2722820658909875e-05, "loss": 0.6493, "step": 8681 }, { "epoch": 0.91, "learning_rate": 5.2721213473316266e-05, "loss": 1.3503, "step": 8682 }, { "epoch": 0.91, "learning_rate": 5.2719606134768775e-05, "loss": 0.9855, "step": 8683 }, { "epoch": 0.91, "learning_rate": 5.27179986432782e-05, "loss": 0.7097, "step": 8684 }, { "epoch": 0.91, "learning_rate": 5.2716390998855385e-05, "loss": 0.3609, "step": 8685 }, { "epoch": 0.91, "learning_rate": 5.2714783201511136e-05, "loss": 1.4246, "step": 8686 }, { "epoch": 0.91, "learning_rate": 5.271317525125628e-05, "loss": 0.9743, "step": 8687 }, { "epoch": 0.91, "learning_rate": 5.271156714810164e-05, "loss": 0.6548, "step": 8688 }, { "epoch": 0.91, "learning_rate": 5.2709958892058054e-05, "loss": 1.5972, "step": 8689 }, { "epoch": 0.91, "learning_rate": 5.270835048313633e-05, "loss": 0.6183, "step": 8690 }, { "epoch": 0.91, "learning_rate": 5.270674192134731e-05, "loss": 0.5573, "step": 8691 }, { "epoch": 0.91, "learning_rate": 5.270513320670182e-05, "loss": 0.723, "step": 8692 }, { "epoch": 0.91, "learning_rate": 5.2703524339210675e-05, "loss": 0.6887, "step": 8693 }, { "epoch": 0.91, "learning_rate": 5.270191531888472e-05, "loss": 1.0063, "step": 8694 }, { "epoch": 0.91, "learning_rate": 5.270030614573478e-05, "loss": 1.3852, "step": 8695 }, { "epoch": 0.91, "learning_rate": 5.269869681977169e-05, "loss": 1.0648, "step": 8696 }, { "epoch": 0.91, "learning_rate": 5.2697087341006295e-05, "loss": 0.5422, "step": 8697 }, { "epoch": 0.91, "learning_rate": 5.2695477709449414e-05, "loss": 1.1241, "step": 8698 }, { "epoch": 0.91, "learning_rate": 5.269386792511188e-05, "loss": 0.4193, "step": 8699 }, { "epoch": 0.91, "learning_rate": 5.269225798800454e-05, "loss": 0.237, "step": 8700 }, { "epoch": 0.91, "learning_rate": 5.2690647898138235e-05, "loss": 1.0057, "step": 8701 }, { "epoch": 0.91, "learning_rate": 5.268903765552379e-05, "loss": 0.4913, "step": 8702 }, { "epoch": 0.91, "learning_rate": 5.268742726017206e-05, "loss": 0.6017, "step": 8703 }, { "epoch": 0.91, "learning_rate": 5.268581671209387e-05, "loss": 0.9103, "step": 8704 }, { "epoch": 0.91, "learning_rate": 5.268420601130007e-05, "loss": 0.2234, "step": 8705 }, { "epoch": 0.91, "learning_rate": 5.268259515780151e-05, "loss": 0.3355, "step": 8706 }, { "epoch": 0.91, "learning_rate": 5.268098415160902e-05, "loss": 1.4794, "step": 8707 }, { "epoch": 0.91, "learning_rate": 5.267937299273346e-05, "loss": 1.3767, "step": 8708 }, { "epoch": 0.91, "learning_rate": 5.267776168118566e-05, "loss": 0.2064, "step": 8709 }, { "epoch": 0.91, "learning_rate": 5.267615021697648e-05, "loss": 1.1351, "step": 8710 }, { "epoch": 0.91, "learning_rate": 5.2674538600116754e-05, "loss": 0.0965, "step": 8711 }, { "epoch": 0.91, "learning_rate": 5.267292683061735e-05, "loss": 0.5773, "step": 8712 }, { "epoch": 0.91, "learning_rate": 5.26713149084891e-05, "loss": 1.5106, "step": 8713 }, { "epoch": 0.91, "learning_rate": 5.266970283374286e-05, "loss": 1.0858, "step": 8714 }, { "epoch": 0.91, "learning_rate": 5.2668090606389495e-05, "loss": 1.5803, "step": 8715 }, { "epoch": 0.91, "learning_rate": 5.2666478226439846e-05, "loss": 1.3371, "step": 8716 }, { "epoch": 0.91, "learning_rate": 5.266486569390477e-05, "loss": 1.0434, "step": 8717 }, { "epoch": 0.91, "learning_rate": 5.2663253008795115e-05, "loss": 0.3258, "step": 8718 }, { "epoch": 0.91, "learning_rate": 5.266164017112174e-05, "loss": 1.4812, "step": 8719 }, { "epoch": 0.91, "learning_rate": 5.2660027180895515e-05, "loss": 1.4319, "step": 8720 }, { "epoch": 0.91, "learning_rate": 5.2658414038127286e-05, "loss": 1.0642, "step": 8721 }, { "epoch": 0.91, "learning_rate": 5.265680074282791e-05, "loss": 0.8391, "step": 8722 }, { "epoch": 0.91, "learning_rate": 5.265518729500827e-05, "loss": 0.7396, "step": 8723 }, { "epoch": 0.91, "learning_rate": 5.265357369467919e-05, "loss": 1.6592, "step": 8724 }, { "epoch": 0.91, "learning_rate": 5.265195994185156e-05, "loss": 1.0472, "step": 8725 }, { "epoch": 0.91, "learning_rate": 5.2650346036536236e-05, "loss": 0.8131, "step": 8726 }, { "epoch": 0.91, "learning_rate": 5.2648731978744085e-05, "loss": 0.6059, "step": 8727 }, { "epoch": 0.91, "learning_rate": 5.264711776848597e-05, "loss": 1.2173, "step": 8728 }, { "epoch": 0.91, "learning_rate": 5.264550340577275e-05, "loss": 1.3493, "step": 8729 }, { "epoch": 0.91, "learning_rate": 5.2643888890615306e-05, "loss": 1.1879, "step": 8730 }, { "epoch": 0.91, "learning_rate": 5.26422742230245e-05, "loss": 0.3207, "step": 8731 }, { "epoch": 0.91, "learning_rate": 5.2640659403011195e-05, "loss": 0.5159, "step": 8732 }, { "epoch": 0.91, "learning_rate": 5.2639044430586274e-05, "loss": 1.4306, "step": 8733 }, { "epoch": 0.91, "learning_rate": 5.26374293057606e-05, "loss": 0.3278, "step": 8734 }, { "epoch": 0.91, "learning_rate": 5.263581402854506e-05, "loss": 0.8334, "step": 8735 }, { "epoch": 0.91, "learning_rate": 5.2634198598950495e-05, "loss": 0.523, "step": 8736 }, { "epoch": 0.91, "learning_rate": 5.2632583016987823e-05, "loss": 0.7161, "step": 8737 }, { "epoch": 0.91, "learning_rate": 5.2630967282667886e-05, "loss": 1.2848, "step": 8738 }, { "epoch": 0.91, "learning_rate": 5.262935139600157e-05, "loss": 0.7371, "step": 8739 }, { "epoch": 0.91, "learning_rate": 5.262773535699977e-05, "loss": 0.72, "step": 8740 }, { "epoch": 0.91, "learning_rate": 5.262611916567333e-05, "loss": 1.1469, "step": 8741 }, { "epoch": 0.91, "learning_rate": 5.262450282203316e-05, "loss": 0.6136, "step": 8742 }, { "epoch": 0.91, "learning_rate": 5.2622886326090137e-05, "loss": 0.5017, "step": 8743 }, { "epoch": 0.91, "learning_rate": 5.2621269677855123e-05, "loss": 0.2257, "step": 8744 }, { "epoch": 0.91, "learning_rate": 5.261965287733902e-05, "loss": 0.8334, "step": 8745 }, { "epoch": 0.91, "learning_rate": 5.261803592455271e-05, "loss": 0.7682, "step": 8746 }, { "epoch": 0.91, "learning_rate": 5.2616418819507076e-05, "loss": 1.4301, "step": 8747 }, { "epoch": 0.91, "learning_rate": 5.2614801562212995e-05, "loss": 1.3884, "step": 8748 }, { "epoch": 0.91, "learning_rate": 5.2613184152681364e-05, "loss": 0.9689, "step": 8749 }, { "epoch": 0.91, "learning_rate": 5.2611566590923066e-05, "loss": 0.5464, "step": 8750 }, { "epoch": 0.91, "learning_rate": 5.2609948876948996e-05, "loss": 0.853, "step": 8751 }, { "epoch": 0.91, "learning_rate": 5.260833101077005e-05, "loss": 0.9644, "step": 8752 }, { "epoch": 0.91, "learning_rate": 5.2606712992397094e-05, "loss": 1.5287, "step": 8753 }, { "epoch": 0.91, "learning_rate": 5.260509482184104e-05, "loss": 0.6067, "step": 8754 }, { "epoch": 0.91, "learning_rate": 5.2603476499112775e-05, "loss": 1.2757, "step": 8755 }, { "epoch": 0.91, "learning_rate": 5.2601858024223205e-05, "loss": 1.5215, "step": 8756 }, { "epoch": 0.91, "learning_rate": 5.260023939718321e-05, "loss": 1.0818, "step": 8757 }, { "epoch": 0.91, "learning_rate": 5.2598620618003684e-05, "loss": 1.0915, "step": 8758 }, { "epoch": 0.91, "learning_rate": 5.259700168669554e-05, "loss": 1.2929, "step": 8759 }, { "epoch": 0.91, "learning_rate": 5.2595382603269664e-05, "loss": 0.7925, "step": 8760 }, { "epoch": 0.91, "learning_rate": 5.259376336773696e-05, "loss": 0.7821, "step": 8761 }, { "epoch": 0.92, "learning_rate": 5.2592143980108334e-05, "loss": 1.1107, "step": 8762 }, { "epoch": 0.92, "learning_rate": 5.259052444039468e-05, "loss": 1.4025, "step": 8763 }, { "epoch": 0.92, "learning_rate": 5.2588904748606883e-05, "loss": 1.1005, "step": 8764 }, { "epoch": 0.92, "learning_rate": 5.258728490475588e-05, "loss": 0.5664, "step": 8765 }, { "epoch": 0.92, "learning_rate": 5.2585664908852565e-05, "loss": 1.1442, "step": 8766 }, { "epoch": 0.92, "learning_rate": 5.258404476090783e-05, "loss": 1.2641, "step": 8767 }, { "epoch": 0.92, "learning_rate": 5.258242446093259e-05, "loss": 0.8135, "step": 8768 }, { "epoch": 0.92, "learning_rate": 5.258080400893775e-05, "loss": 1.4097, "step": 8769 }, { "epoch": 0.92, "learning_rate": 5.2579183404934224e-05, "loss": 0.7932, "step": 8770 }, { "epoch": 0.92, "learning_rate": 5.2577562648932926e-05, "loss": 0.7791, "step": 8771 }, { "epoch": 0.92, "learning_rate": 5.2575941740944745e-05, "loss": 1.1086, "step": 8772 }, { "epoch": 0.92, "learning_rate": 5.257432068098062e-05, "loss": 1.1634, "step": 8773 }, { "epoch": 0.92, "learning_rate": 5.2572699469051436e-05, "loss": 0.8, "step": 8774 }, { "epoch": 0.92, "learning_rate": 5.2571078105168135e-05, "loss": 1.0982, "step": 8775 }, { "epoch": 0.92, "learning_rate": 5.25694565893416e-05, "loss": 1.5366, "step": 8776 }, { "epoch": 0.92, "learning_rate": 5.256783492158278e-05, "loss": 1.0828, "step": 8777 }, { "epoch": 0.92, "learning_rate": 5.256621310190257e-05, "loss": 0.6684, "step": 8778 }, { "epoch": 0.92, "learning_rate": 5.2564591130311896e-05, "loss": 0.3569, "step": 8779 }, { "epoch": 0.92, "learning_rate": 5.256296900682167e-05, "loss": 1.6056, "step": 8780 }, { "epoch": 0.92, "learning_rate": 5.256134673144282e-05, "loss": 1.1365, "step": 8781 }, { "epoch": 0.92, "learning_rate": 5.255972430418626e-05, "loss": 0.9582, "step": 8782 }, { "epoch": 0.92, "learning_rate": 5.255810172506292e-05, "loss": 1.3956, "step": 8783 }, { "epoch": 0.92, "learning_rate": 5.255647899408371e-05, "loss": 1.4245, "step": 8784 }, { "epoch": 0.92, "learning_rate": 5.255485611125956e-05, "loss": 0.9636, "step": 8785 }, { "epoch": 0.92, "learning_rate": 5.25532330766014e-05, "loss": 0.7567, "step": 8786 }, { "epoch": 0.92, "learning_rate": 5.2551609890120156e-05, "loss": 0.8418, "step": 8787 }, { "epoch": 0.92, "learning_rate": 5.2549986551826744e-05, "loss": 0.9692, "step": 8788 }, { "epoch": 0.92, "learning_rate": 5.25483630617321e-05, "loss": 0.6352, "step": 8789 }, { "epoch": 0.92, "learning_rate": 5.2546739419847164e-05, "loss": 1.5177, "step": 8790 }, { "epoch": 0.92, "learning_rate": 5.254511562618283e-05, "loss": 0.7134, "step": 8791 }, { "epoch": 0.92, "learning_rate": 5.254349168075007e-05, "loss": 1.1483, "step": 8792 }, { "epoch": 0.92, "learning_rate": 5.25418675835598e-05, "loss": 1.6361, "step": 8793 }, { "epoch": 0.92, "learning_rate": 5.254024333462294e-05, "loss": 0.1865, "step": 8794 }, { "epoch": 0.92, "learning_rate": 5.253861893395045e-05, "loss": 0.3679, "step": 8795 }, { "epoch": 0.92, "learning_rate": 5.253699438155324e-05, "loss": 0.6425, "step": 8796 }, { "epoch": 0.92, "learning_rate": 5.253536967744226e-05, "loss": 1.5338, "step": 8797 }, { "epoch": 0.92, "learning_rate": 5.2533744821628446e-05, "loss": 1.3522, "step": 8798 }, { "epoch": 0.92, "learning_rate": 5.253211981412273e-05, "loss": 2.099, "step": 8799 }, { "epoch": 0.92, "learning_rate": 5.2530494654936064e-05, "loss": 1.1057, "step": 8800 }, { "epoch": 0.92, "learning_rate": 5.252886934407937e-05, "loss": 1.3925, "step": 8801 }, { "epoch": 0.92, "learning_rate": 5.2527243881563606e-05, "loss": 0.7895, "step": 8802 }, { "epoch": 0.92, "learning_rate": 5.25256182673997e-05, "loss": 0.8145, "step": 8803 }, { "epoch": 0.92, "learning_rate": 5.2523992501598605e-05, "loss": 0.2267, "step": 8804 }, { "epoch": 0.92, "learning_rate": 5.252236658417127e-05, "loss": 0.8438, "step": 8805 }, { "epoch": 0.92, "learning_rate": 5.2520740515128626e-05, "loss": 1.0968, "step": 8806 }, { "epoch": 0.92, "learning_rate": 5.2519114294481624e-05, "loss": 0.6871, "step": 8807 }, { "epoch": 0.92, "learning_rate": 5.251748792224122e-05, "loss": 1.0496, "step": 8808 }, { "epoch": 0.92, "learning_rate": 5.2515861398418355e-05, "loss": 0.7671, "step": 8809 }, { "epoch": 0.92, "learning_rate": 5.2514234723023975e-05, "loss": 1.2808, "step": 8810 }, { "epoch": 0.92, "learning_rate": 5.2512607896069035e-05, "loss": 0.5946, "step": 8811 }, { "epoch": 0.92, "learning_rate": 5.25109809175645e-05, "loss": 0.8599, "step": 8812 }, { "epoch": 0.92, "learning_rate": 5.250935378752129e-05, "loss": 1.3778, "step": 8813 }, { "epoch": 0.92, "learning_rate": 5.250772650595038e-05, "loss": 0.3063, "step": 8814 }, { "epoch": 0.92, "learning_rate": 5.250609907286273e-05, "loss": 0.9123, "step": 8815 }, { "epoch": 0.92, "learning_rate": 5.250447148826928e-05, "loss": 0.979, "step": 8816 }, { "epoch": 0.92, "learning_rate": 5.250284375218099e-05, "loss": 1.2417, "step": 8817 }, { "epoch": 0.92, "learning_rate": 5.250121586460883e-05, "loss": 0.827, "step": 8818 }, { "epoch": 0.92, "learning_rate": 5.249958782556375e-05, "loss": 0.284, "step": 8819 }, { "epoch": 0.92, "learning_rate": 5.249795963505671e-05, "loss": 0.2536, "step": 8820 }, { "epoch": 0.92, "learning_rate": 5.249633129309866e-05, "loss": 1.457, "step": 8821 }, { "epoch": 0.92, "learning_rate": 5.2494702799700586e-05, "loss": 1.4446, "step": 8822 }, { "epoch": 0.92, "learning_rate": 5.2493074154873425e-05, "loss": 1.185, "step": 8823 }, { "epoch": 0.92, "learning_rate": 5.249144535862815e-05, "loss": 0.5425, "step": 8824 }, { "epoch": 0.92, "learning_rate": 5.248981641097574e-05, "loss": 0.8854, "step": 8825 }, { "epoch": 0.92, "learning_rate": 5.2488187311927135e-05, "loss": 1.0427, "step": 8826 }, { "epoch": 0.92, "learning_rate": 5.2486558061493316e-05, "loss": 0.1496, "step": 8827 }, { "epoch": 0.92, "learning_rate": 5.248492865968526e-05, "loss": 0.7506, "step": 8828 }, { "epoch": 0.92, "learning_rate": 5.2483299106513915e-05, "loss": 1.1067, "step": 8829 }, { "epoch": 0.92, "learning_rate": 5.248166940199026e-05, "loss": 1.185, "step": 8830 }, { "epoch": 0.92, "learning_rate": 5.248003954612528e-05, "loss": 0.4018, "step": 8831 }, { "epoch": 0.92, "learning_rate": 5.247840953892992e-05, "loss": 0.6409, "step": 8832 }, { "epoch": 0.92, "learning_rate": 5.247677938041517e-05, "loss": 0.9274, "step": 8833 }, { "epoch": 0.92, "learning_rate": 5.2475149070592e-05, "loss": 0.7477, "step": 8834 }, { "epoch": 0.92, "learning_rate": 5.247351860947139e-05, "loss": 1.0413, "step": 8835 }, { "epoch": 0.92, "learning_rate": 5.2471887997064315e-05, "loss": 0.503, "step": 8836 }, { "epoch": 0.92, "learning_rate": 5.247025723338174e-05, "loss": 1.0074, "step": 8837 }, { "epoch": 0.92, "learning_rate": 5.2468626318434656e-05, "loss": 1.2894, "step": 8838 }, { "epoch": 0.92, "learning_rate": 5.246699525223403e-05, "loss": 0.227, "step": 8839 }, { "epoch": 0.92, "learning_rate": 5.246536403479085e-05, "loss": 1.3095, "step": 8840 }, { "epoch": 0.92, "learning_rate": 5.246373266611609e-05, "loss": 1.146, "step": 8841 }, { "epoch": 0.92, "learning_rate": 5.246210114622075e-05, "loss": 0.9507, "step": 8842 }, { "epoch": 0.92, "learning_rate": 5.24604694751158e-05, "loss": 0.6276, "step": 8843 }, { "epoch": 0.92, "learning_rate": 5.2458837652812224e-05, "loss": 0.924, "step": 8844 }, { "epoch": 0.92, "learning_rate": 5.2457205679321e-05, "loss": 1.2455, "step": 8845 }, { "epoch": 0.92, "learning_rate": 5.245557355465313e-05, "loss": 1.1416, "step": 8846 }, { "epoch": 0.92, "learning_rate": 5.245394127881959e-05, "loss": 1.9665, "step": 8847 }, { "epoch": 0.92, "learning_rate": 5.245230885183137e-05, "loss": 1.1289, "step": 8848 }, { "epoch": 0.92, "learning_rate": 5.245067627369946e-05, "loss": 1.011, "step": 8849 }, { "epoch": 0.92, "learning_rate": 5.2449043544434855e-05, "loss": 0.7186, "step": 8850 }, { "epoch": 0.92, "learning_rate": 5.244741066404854e-05, "loss": 1.3765, "step": 8851 }, { "epoch": 0.92, "learning_rate": 5.244577763255151e-05, "loss": 0.3939, "step": 8852 }, { "epoch": 0.92, "learning_rate": 5.244414444995475e-05, "loss": 0.694, "step": 8853 }, { "epoch": 0.92, "learning_rate": 5.244251111626926e-05, "loss": 1.3486, "step": 8854 }, { "epoch": 0.92, "learning_rate": 5.244087763150604e-05, "loss": 1.0488, "step": 8855 }, { "epoch": 0.92, "learning_rate": 5.243924399567609e-05, "loss": 1.028, "step": 8856 }, { "epoch": 0.93, "learning_rate": 5.243761020879039e-05, "loss": 1.8481, "step": 8857 }, { "epoch": 0.93, "learning_rate": 5.243597627085995e-05, "loss": 0.3689, "step": 8858 }, { "epoch": 0.93, "learning_rate": 5.2434342181895774e-05, "loss": 1.2318, "step": 8859 }, { "epoch": 0.93, "learning_rate": 5.243270794190885e-05, "loss": 1.1665, "step": 8860 }, { "epoch": 0.93, "learning_rate": 5.2431073550910195e-05, "loss": 0.4111, "step": 8861 }, { "epoch": 0.93, "learning_rate": 5.242943900891079e-05, "loss": 1.1659, "step": 8862 }, { "epoch": 0.93, "learning_rate": 5.242780431592164e-05, "loss": 0.338, "step": 8863 }, { "epoch": 0.93, "learning_rate": 5.242616947195377e-05, "loss": 0.5836, "step": 8864 }, { "epoch": 0.93, "learning_rate": 5.242453447701818e-05, "loss": 1.5965, "step": 8865 }, { "epoch": 0.93, "learning_rate": 5.242289933112587e-05, "loss": 1.488, "step": 8866 }, { "epoch": 0.93, "learning_rate": 5.242126403428784e-05, "loss": 1.5077, "step": 8867 }, { "epoch": 0.93, "learning_rate": 5.241962858651511e-05, "loss": 0.7664, "step": 8868 }, { "epoch": 0.93, "learning_rate": 5.241799298781869e-05, "loss": 1.309, "step": 8869 }, { "epoch": 0.93, "learning_rate": 5.241635723820958e-05, "loss": 1.3278, "step": 8870 }, { "epoch": 0.93, "learning_rate": 5.2414721337698804e-05, "loss": 1.3016, "step": 8871 }, { "epoch": 0.93, "learning_rate": 5.241308528629736e-05, "loss": 1.0645, "step": 8872 }, { "epoch": 0.93, "learning_rate": 5.241144908401629e-05, "loss": 1.3088, "step": 8873 }, { "epoch": 0.93, "learning_rate": 5.240981273086657e-05, "loss": 1.3551, "step": 8874 }, { "epoch": 0.93, "learning_rate": 5.2408176226859233e-05, "loss": 0.3018, "step": 8875 }, { "epoch": 0.93, "learning_rate": 5.24065395720053e-05, "loss": 0.9175, "step": 8876 }, { "epoch": 0.93, "learning_rate": 5.240490276631579e-05, "loss": 0.3907, "step": 8877 }, { "epoch": 0.93, "learning_rate": 5.240326580980172e-05, "loss": 0.4112, "step": 8878 }, { "epoch": 0.93, "learning_rate": 5.24016287024741e-05, "loss": 0.3691, "step": 8879 }, { "epoch": 0.93, "learning_rate": 5.2399991444343965e-05, "loss": 0.7403, "step": 8880 }, { "epoch": 0.93, "learning_rate": 5.239835403542232e-05, "loss": 0.7161, "step": 8881 }, { "epoch": 0.93, "learning_rate": 5.2396716475720195e-05, "loss": 1.1094, "step": 8882 }, { "epoch": 0.93, "learning_rate": 5.2395078765248625e-05, "loss": 1.563, "step": 8883 }, { "epoch": 0.93, "learning_rate": 5.2393440904018615e-05, "loss": 0.5769, "step": 8884 }, { "epoch": 0.93, "learning_rate": 5.23918028920412e-05, "loss": 1.977, "step": 8885 }, { "epoch": 0.93, "learning_rate": 5.239016472932741e-05, "loss": 0.5307, "step": 8886 }, { "epoch": 0.93, "learning_rate": 5.238852641588828e-05, "loss": 1.477, "step": 8887 }, { "epoch": 0.93, "learning_rate": 5.2386887951734814e-05, "loss": 1.1202, "step": 8888 }, { "epoch": 0.93, "learning_rate": 5.238524933687806e-05, "loss": 1.4867, "step": 8889 }, { "epoch": 0.93, "learning_rate": 5.238361057132905e-05, "loss": 0.6157, "step": 8890 }, { "epoch": 0.93, "learning_rate": 5.2381971655098813e-05, "loss": 0.4562, "step": 8891 }, { "epoch": 0.93, "learning_rate": 5.238033258819837e-05, "loss": 1.4973, "step": 8892 }, { "epoch": 0.93, "learning_rate": 5.237869337063877e-05, "loss": 0.74, "step": 8893 }, { "epoch": 0.93, "learning_rate": 5.237705400243104e-05, "loss": 1.5475, "step": 8894 }, { "epoch": 0.93, "learning_rate": 5.237541448358622e-05, "loss": 1.4178, "step": 8895 }, { "epoch": 0.93, "learning_rate": 5.2373774814115355e-05, "loss": 1.3767, "step": 8896 }, { "epoch": 0.93, "learning_rate": 5.237213499402946e-05, "loss": 1.0812, "step": 8897 }, { "epoch": 0.93, "learning_rate": 5.237049502333959e-05, "loss": 0.5209, "step": 8898 }, { "epoch": 0.93, "learning_rate": 5.236885490205679e-05, "loss": 0.4083, "step": 8899 }, { "epoch": 0.93, "learning_rate": 5.236721463019208e-05, "loss": 0.7423, "step": 8900 }, { "epoch": 0.93, "learning_rate": 5.236557420775652e-05, "loss": 0.4014, "step": 8901 }, { "epoch": 0.93, "learning_rate": 5.236393363476114e-05, "loss": 1.1063, "step": 8902 }, { "epoch": 0.93, "learning_rate": 5.236229291121701e-05, "loss": 0.5521, "step": 8903 }, { "epoch": 0.93, "learning_rate": 5.236065203713514e-05, "loss": 1.4523, "step": 8904 }, { "epoch": 0.93, "learning_rate": 5.2359011012526595e-05, "loss": 0.7412, "step": 8905 }, { "epoch": 0.93, "learning_rate": 5.235736983740243e-05, "loss": 0.8584, "step": 8906 }, { "epoch": 0.93, "learning_rate": 5.235572851177367e-05, "loss": 0.3333, "step": 8907 }, { "epoch": 0.93, "learning_rate": 5.2354087035651385e-05, "loss": 1.3686, "step": 8908 }, { "epoch": 0.93, "learning_rate": 5.235244540904661e-05, "loss": 0.5902, "step": 8909 }, { "epoch": 0.93, "learning_rate": 5.23508036319704e-05, "loss": 1.4817, "step": 8910 }, { "epoch": 0.93, "learning_rate": 5.234916170443382e-05, "loss": 0.5857, "step": 8911 }, { "epoch": 0.93, "learning_rate": 5.234751962644791e-05, "loss": 1.0149, "step": 8912 }, { "epoch": 0.93, "learning_rate": 5.234587739802373e-05, "loss": 0.3746, "step": 8913 }, { "epoch": 0.93, "learning_rate": 5.2344235019172315e-05, "loss": 1.5837, "step": 8914 }, { "epoch": 0.93, "learning_rate": 5.2342592489904755e-05, "loss": 0.4814, "step": 8915 }, { "epoch": 0.93, "learning_rate": 5.234094981023208e-05, "loss": 0.7751, "step": 8916 }, { "epoch": 0.93, "learning_rate": 5.233930698016537e-05, "loss": 0.7296, "step": 8917 }, { "epoch": 0.93, "learning_rate": 5.233766399971566e-05, "loss": 1.0977, "step": 8918 }, { "epoch": 0.93, "learning_rate": 5.2336020868894024e-05, "loss": 0.5911, "step": 8919 }, { "epoch": 0.93, "learning_rate": 5.233437758771153e-05, "loss": 0.51, "step": 8920 }, { "epoch": 0.93, "learning_rate": 5.233273415617922e-05, "loss": 0.5311, "step": 8921 }, { "epoch": 0.93, "learning_rate": 5.233109057430818e-05, "loss": 0.9157, "step": 8922 }, { "epoch": 0.93, "learning_rate": 5.232944684210946e-05, "loss": 0.4391, "step": 8923 }, { "epoch": 0.93, "learning_rate": 5.232780295959413e-05, "loss": 0.3119, "step": 8924 }, { "epoch": 0.93, "learning_rate": 5.232615892677326e-05, "loss": 0.6101, "step": 8925 }, { "epoch": 0.93, "learning_rate": 5.232451474365789e-05, "loss": 0.5943, "step": 8926 }, { "epoch": 0.93, "learning_rate": 5.232287041025913e-05, "loss": 0.8092, "step": 8927 }, { "epoch": 0.93, "learning_rate": 5.232122592658803e-05, "loss": 1.4692, "step": 8928 }, { "epoch": 0.93, "learning_rate": 5.231958129265565e-05, "loss": 0.7479, "step": 8929 }, { "epoch": 0.93, "learning_rate": 5.231793650847308e-05, "loss": 0.8416, "step": 8930 }, { "epoch": 0.93, "learning_rate": 5.231629157405138e-05, "loss": 0.9086, "step": 8931 }, { "epoch": 0.93, "learning_rate": 5.231464648940162e-05, "loss": 0.2372, "step": 8932 }, { "epoch": 0.93, "learning_rate": 5.231300125453489e-05, "loss": 0.7727, "step": 8933 }, { "epoch": 0.93, "learning_rate": 5.231135586946225e-05, "loss": 0.8847, "step": 8934 }, { "epoch": 0.93, "learning_rate": 5.230971033419479e-05, "loss": 0.5531, "step": 8935 }, { "epoch": 0.93, "learning_rate": 5.230806464874357e-05, "loss": 0.6884, "step": 8936 }, { "epoch": 0.93, "learning_rate": 5.2306418813119686e-05, "loss": 1.0464, "step": 8937 }, { "epoch": 0.93, "learning_rate": 5.230477282733421e-05, "loss": 0.641, "step": 8938 }, { "epoch": 0.93, "learning_rate": 5.230312669139822e-05, "loss": 0.8907, "step": 8939 }, { "epoch": 0.93, "learning_rate": 5.23014804053228e-05, "loss": 0.5814, "step": 8940 }, { "epoch": 0.93, "learning_rate": 5.229983396911903e-05, "loss": 0.7078, "step": 8941 }, { "epoch": 0.93, "learning_rate": 5.2298187382798e-05, "loss": 0.2653, "step": 8942 }, { "epoch": 0.93, "learning_rate": 5.229654064637079e-05, "loss": 1.439, "step": 8943 }, { "epoch": 0.93, "learning_rate": 5.2294893759848484e-05, "loss": 1.4648, "step": 8944 }, { "epoch": 0.93, "learning_rate": 5.229324672324216e-05, "loss": 0.6681, "step": 8945 }, { "epoch": 0.93, "learning_rate": 5.229159953656293e-05, "loss": 1.3885, "step": 8946 }, { "epoch": 0.93, "learning_rate": 5.228995219982186e-05, "loss": 0.6206, "step": 8947 }, { "epoch": 0.93, "learning_rate": 5.228830471303005e-05, "loss": 0.9483, "step": 8948 }, { "epoch": 0.93, "learning_rate": 5.228665707619858e-05, "loss": 1.0233, "step": 8949 }, { "epoch": 0.93, "learning_rate": 5.2285009289338563e-05, "loss": 0.6193, "step": 8950 }, { "epoch": 0.93, "learning_rate": 5.228336135246107e-05, "loss": 0.7772, "step": 8951 }, { "epoch": 0.93, "learning_rate": 5.2281713265577206e-05, "loss": 2.1789, "step": 8952 }, { "epoch": 0.94, "learning_rate": 5.228006502869805e-05, "loss": 1.2281, "step": 8953 }, { "epoch": 0.94, "learning_rate": 5.2278416641834724e-05, "loss": 0.9686, "step": 8954 }, { "epoch": 0.94, "learning_rate": 5.22767681049983e-05, "loss": 1.0252, "step": 8955 }, { "epoch": 0.94, "learning_rate": 5.2275119418199894e-05, "loss": 0.5873, "step": 8956 }, { "epoch": 0.94, "learning_rate": 5.227347058145059e-05, "loss": 1.1421, "step": 8957 }, { "epoch": 0.94, "learning_rate": 5.22718215947615e-05, "loss": 0.4404, "step": 8958 }, { "epoch": 0.94, "learning_rate": 5.2270172458143705e-05, "loss": 1.1172, "step": 8959 }, { "epoch": 0.94, "learning_rate": 5.226852317160833e-05, "loss": 1.9414, "step": 8960 }, { "epoch": 0.94, "learning_rate": 5.2266873735166463e-05, "loss": 1.3167, "step": 8961 }, { "epoch": 0.94, "learning_rate": 5.226522414882921e-05, "loss": 0.9046, "step": 8962 }, { "epoch": 0.94, "learning_rate": 5.2263574412607686e-05, "loss": 0.7831, "step": 8963 }, { "epoch": 0.94, "learning_rate": 5.226192452651299e-05, "loss": 0.3529, "step": 8964 }, { "epoch": 0.94, "learning_rate": 5.226027449055622e-05, "loss": 0.5386, "step": 8965 }, { "epoch": 0.94, "learning_rate": 5.2258624304748496e-05, "loss": 0.5718, "step": 8966 }, { "epoch": 0.94, "learning_rate": 5.225697396910091e-05, "loss": 0.7299, "step": 8967 }, { "epoch": 0.94, "learning_rate": 5.225532348362459e-05, "loss": 1.3948, "step": 8968 }, { "epoch": 0.94, "learning_rate": 5.225367284833065e-05, "loss": 1.2895, "step": 8969 }, { "epoch": 0.94, "learning_rate": 5.225202206323018e-05, "loss": 0.8796, "step": 8970 }, { "epoch": 0.94, "learning_rate": 5.22503711283343e-05, "loss": 0.4258, "step": 8971 }, { "epoch": 0.94, "learning_rate": 5.224872004365413e-05, "loss": 0.278, "step": 8972 }, { "epoch": 0.94, "learning_rate": 5.2247068809200786e-05, "loss": 0.4139, "step": 8973 }, { "epoch": 0.94, "learning_rate": 5.224541742498538e-05, "loss": 1.3449, "step": 8974 }, { "epoch": 0.94, "learning_rate": 5.224376589101903e-05, "loss": 0.5675, "step": 8975 }, { "epoch": 0.94, "learning_rate": 5.224211420731285e-05, "loss": 1.15, "step": 8976 }, { "epoch": 0.94, "learning_rate": 5.224046237387796e-05, "loss": 1.3047, "step": 8977 }, { "epoch": 0.94, "learning_rate": 5.223881039072549e-05, "loss": 1.1723, "step": 8978 }, { "epoch": 0.94, "learning_rate": 5.2237158257866543e-05, "loss": 1.0248, "step": 8979 }, { "epoch": 0.94, "learning_rate": 5.223550597531225e-05, "loss": 0.954, "step": 8980 }, { "epoch": 0.94, "learning_rate": 5.223385354307374e-05, "loss": 1.009, "step": 8981 }, { "epoch": 0.94, "learning_rate": 5.223220096116213e-05, "loss": 0.674, "step": 8982 }, { "epoch": 0.94, "learning_rate": 5.2230548229588537e-05, "loss": 1.0253, "step": 8983 }, { "epoch": 0.94, "learning_rate": 5.2228895348364096e-05, "loss": 0.7072, "step": 8984 }, { "epoch": 0.94, "learning_rate": 5.222724231749994e-05, "loss": 1.0052, "step": 8985 }, { "epoch": 0.94, "learning_rate": 5.2225589137007186e-05, "loss": 0.9483, "step": 8986 }, { "epoch": 0.94, "learning_rate": 5.222393580689696e-05, "loss": 0.7216, "step": 8987 }, { "epoch": 0.94, "learning_rate": 5.222228232718041e-05, "loss": 0.9725, "step": 8988 }, { "epoch": 0.94, "learning_rate": 5.2220628697868645e-05, "loss": 0.6578, "step": 8989 }, { "epoch": 0.94, "learning_rate": 5.221897491897281e-05, "loss": 0.786, "step": 8990 }, { "epoch": 0.94, "learning_rate": 5.221732099050404e-05, "loss": 1.3696, "step": 8991 }, { "epoch": 0.94, "learning_rate": 5.221566691247346e-05, "loss": 1.2313, "step": 8992 }, { "epoch": 0.94, "learning_rate": 5.22140126848922e-05, "loss": 0.9692, "step": 8993 }, { "epoch": 0.94, "learning_rate": 5.2212358307771414e-05, "loss": 0.3376, "step": 8994 }, { "epoch": 0.94, "learning_rate": 5.221070378112222e-05, "loss": 0.4526, "step": 8995 }, { "epoch": 0.94, "learning_rate": 5.220904910495577e-05, "loss": 1.1636, "step": 8996 }, { "epoch": 0.94, "learning_rate": 5.220739427928321e-05, "loss": 1.1323, "step": 8997 }, { "epoch": 0.94, "learning_rate": 5.220573930411565e-05, "loss": 1.4465, "step": 8998 }, { "epoch": 0.94, "learning_rate": 5.220408417946426e-05, "loss": 0.9702, "step": 8999 }, { "epoch": 0.94, "learning_rate": 5.2202428905340164e-05, "loss": 0.3031, "step": 9000 }, { "epoch": 0.94, "learning_rate": 5.220077348175452e-05, "loss": 1.2069, "step": 9001 }, { "epoch": 0.94, "learning_rate": 5.219911790871846e-05, "loss": 0.4323, "step": 9002 }, { "epoch": 0.94, "learning_rate": 5.2197462186243116e-05, "loss": 1.4566, "step": 9003 }, { "epoch": 0.94, "learning_rate": 5.219580631433967e-05, "loss": 0.6323, "step": 9004 }, { "epoch": 0.94, "learning_rate": 5.219415029301924e-05, "loss": 0.4088, "step": 9005 }, { "epoch": 0.94, "learning_rate": 5.219249412229299e-05, "loss": 1.1888, "step": 9006 }, { "epoch": 0.94, "learning_rate": 5.219083780217206e-05, "loss": 0.5296, "step": 9007 }, { "epoch": 0.94, "learning_rate": 5.21891813326676e-05, "loss": 1.6835, "step": 9008 }, { "epoch": 0.94, "learning_rate": 5.218752471379077e-05, "loss": 1.5268, "step": 9009 }, { "epoch": 0.94, "learning_rate": 5.2185867945552706e-05, "loss": 0.7683, "step": 9010 }, { "epoch": 0.94, "learning_rate": 5.2184211027964574e-05, "loss": 1.0071, "step": 9011 }, { "epoch": 0.94, "learning_rate": 5.218255396103752e-05, "loss": 1.3959, "step": 9012 }, { "epoch": 0.94, "learning_rate": 5.2180896744782704e-05, "loss": 1.1738, "step": 9013 }, { "epoch": 0.94, "learning_rate": 5.217923937921129e-05, "loss": 0.1744, "step": 9014 }, { "epoch": 0.94, "learning_rate": 5.2177581864334416e-05, "loss": 0.3819, "step": 9015 }, { "epoch": 0.94, "learning_rate": 5.217592420016326e-05, "loss": 1.3976, "step": 9016 }, { "epoch": 0.94, "learning_rate": 5.2174266386708955e-05, "loss": 0.3683, "step": 9017 }, { "epoch": 0.94, "learning_rate": 5.217260842398269e-05, "loss": 0.3765, "step": 9018 }, { "epoch": 0.94, "learning_rate": 5.217095031199561e-05, "loss": 0.6714, "step": 9019 }, { "epoch": 0.94, "learning_rate": 5.216929205075889e-05, "loss": 1.1252, "step": 9020 }, { "epoch": 0.94, "learning_rate": 5.216763364028366e-05, "loss": 0.4733, "step": 9021 }, { "epoch": 0.94, "learning_rate": 5.2165975080581124e-05, "loss": 2.4628, "step": 9022 }, { "epoch": 0.94, "learning_rate": 5.2164316371662426e-05, "loss": 1.5832, "step": 9023 }, { "epoch": 0.94, "learning_rate": 5.216265751353873e-05, "loss": 1.551, "step": 9024 }, { "epoch": 0.94, "learning_rate": 5.216099850622122e-05, "loss": 1.0788, "step": 9025 }, { "epoch": 0.94, "learning_rate": 5.2159339349721046e-05, "loss": 0.4675, "step": 9026 }, { "epoch": 0.94, "learning_rate": 5.2157680044049395e-05, "loss": 0.4846, "step": 9027 }, { "epoch": 0.94, "learning_rate": 5.215602058921742e-05, "loss": 0.5557, "step": 9028 }, { "epoch": 0.94, "learning_rate": 5.2154360985236294e-05, "loss": 1.1584, "step": 9029 }, { "epoch": 0.94, "learning_rate": 5.21527012321172e-05, "loss": 1.5304, "step": 9030 }, { "epoch": 0.94, "learning_rate": 5.2151041329871304e-05, "loss": 0.4554, "step": 9031 }, { "epoch": 0.94, "learning_rate": 5.214938127850978e-05, "loss": 0.4204, "step": 9032 }, { "epoch": 0.94, "learning_rate": 5.21477210780438e-05, "loss": 0.5169, "step": 9033 }, { "epoch": 0.94, "learning_rate": 5.214606072848455e-05, "loss": 0.7128, "step": 9034 }, { "epoch": 0.94, "learning_rate": 5.2144400229843185e-05, "loss": 1.0187, "step": 9035 }, { "epoch": 0.94, "learning_rate": 5.214273958213091e-05, "loss": 0.363, "step": 9036 }, { "epoch": 0.94, "learning_rate": 5.21410787853589e-05, "loss": 0.2793, "step": 9037 }, { "epoch": 0.94, "learning_rate": 5.213941783953832e-05, "loss": 0.3419, "step": 9038 }, { "epoch": 0.94, "learning_rate": 5.2137756744680366e-05, "loss": 0.9184, "step": 9039 }, { "epoch": 0.94, "learning_rate": 5.2136095500796206e-05, "loss": 0.7243, "step": 9040 }, { "epoch": 0.94, "learning_rate": 5.213443410789703e-05, "loss": 1.019, "step": 9041 }, { "epoch": 0.94, "learning_rate": 5.2132772565994036e-05, "loss": 1.6479, "step": 9042 }, { "epoch": 0.94, "learning_rate": 5.2131110875098375e-05, "loss": 0.5016, "step": 9043 }, { "epoch": 0.94, "learning_rate": 5.2129449035221266e-05, "loss": 1.1035, "step": 9044 }, { "epoch": 0.94, "learning_rate": 5.212778704637389e-05, "loss": 1.8806, "step": 9045 }, { "epoch": 0.94, "learning_rate": 5.2126124908567426e-05, "loss": 1.2718, "step": 9046 }, { "epoch": 0.94, "learning_rate": 5.2124462621813054e-05, "loss": 0.7396, "step": 9047 }, { "epoch": 0.94, "learning_rate": 5.212280018612199e-05, "loss": 0.8277, "step": 9048 }, { "epoch": 0.95, "learning_rate": 5.21211376015054e-05, "loss": 1.0733, "step": 9049 }, { "epoch": 0.95, "learning_rate": 5.211947486797449e-05, "loss": 0.6471, "step": 9050 }, { "epoch": 0.95, "learning_rate": 5.211781198554046e-05, "loss": 0.2233, "step": 9051 }, { "epoch": 0.95, "learning_rate": 5.211614895421449e-05, "loss": 0.6923, "step": 9052 }, { "epoch": 0.95, "learning_rate": 5.211448577400778e-05, "loss": 1.05, "step": 9053 }, { "epoch": 0.95, "learning_rate": 5.211282244493152e-05, "loss": 0.7122, "step": 9054 }, { "epoch": 0.95, "learning_rate": 5.2111158966996924e-05, "loss": 1.4545, "step": 9055 }, { "epoch": 0.95, "learning_rate": 5.2109495340215174e-05, "loss": 1.1213, "step": 9056 }, { "epoch": 0.95, "learning_rate": 5.210783156459747e-05, "loss": 0.7949, "step": 9057 }, { "epoch": 0.95, "learning_rate": 5.210616764015503e-05, "loss": 0.763, "step": 9058 }, { "epoch": 0.95, "learning_rate": 5.210450356689902e-05, "loss": 1.3302, "step": 9059 }, { "epoch": 0.95, "learning_rate": 5.2102839344840684e-05, "loss": 0.8358, "step": 9060 }, { "epoch": 0.95, "learning_rate": 5.21011749739912e-05, "loss": 0.3042, "step": 9061 }, { "epoch": 0.95, "learning_rate": 5.2099510454361774e-05, "loss": 2.0908, "step": 9062 }, { "epoch": 0.95, "learning_rate": 5.209784578596362e-05, "loss": 0.8662, "step": 9063 }, { "epoch": 0.95, "learning_rate": 5.209618096880793e-05, "loss": 1.1068, "step": 9064 }, { "epoch": 0.95, "learning_rate": 5.209451600290593e-05, "loss": 1.189, "step": 9065 }, { "epoch": 0.95, "learning_rate": 5.209285088826881e-05, "loss": 0.7468, "step": 9066 }, { "epoch": 0.95, "learning_rate": 5.209118562490779e-05, "loss": 0.4679, "step": 9067 }, { "epoch": 0.95, "learning_rate": 5.2089520212834075e-05, "loss": 0.5532, "step": 9068 }, { "epoch": 0.95, "learning_rate": 5.208785465205888e-05, "loss": 0.3282, "step": 9069 }, { "epoch": 0.95, "learning_rate": 5.208618894259342e-05, "loss": 0.6634, "step": 9070 }, { "epoch": 0.95, "learning_rate": 5.2084523084448896e-05, "loss": 0.1399, "step": 9071 }, { "epoch": 0.95, "learning_rate": 5.208285707763654e-05, "loss": 1.4682, "step": 9072 }, { "epoch": 0.95, "learning_rate": 5.2081190922167546e-05, "loss": 0.8819, "step": 9073 }, { "epoch": 0.95, "learning_rate": 5.207952461805315e-05, "loss": 0.5943, "step": 9074 }, { "epoch": 0.95, "learning_rate": 5.2077858165304555e-05, "loss": 1.6024, "step": 9075 }, { "epoch": 0.95, "learning_rate": 5.2076191563932984e-05, "loss": 0.6535, "step": 9076 }, { "epoch": 0.95, "learning_rate": 5.207452481394966e-05, "loss": 1.6491, "step": 9077 }, { "epoch": 0.95, "learning_rate": 5.2072857915365795e-05, "loss": 1.0638, "step": 9078 }, { "epoch": 0.95, "learning_rate": 5.2071190868192626e-05, "loss": 0.4612, "step": 9079 }, { "epoch": 0.95, "learning_rate": 5.2069523672441354e-05, "loss": 1.3028, "step": 9080 }, { "epoch": 0.95, "learning_rate": 5.206785632812321e-05, "loss": 0.2793, "step": 9081 }, { "epoch": 0.95, "learning_rate": 5.206618883524943e-05, "loss": 1.0318, "step": 9082 }, { "epoch": 0.95, "learning_rate": 5.206452119383123e-05, "loss": 0.5487, "step": 9083 }, { "epoch": 0.95, "learning_rate": 5.2062853403879835e-05, "loss": 0.87, "step": 9084 }, { "epoch": 0.95, "learning_rate": 5.206118546540647e-05, "loss": 0.9277, "step": 9085 }, { "epoch": 0.95, "learning_rate": 5.205951737842236e-05, "loss": 1.313, "step": 9086 }, { "epoch": 0.95, "learning_rate": 5.205784914293876e-05, "loss": 0.3539, "step": 9087 }, { "epoch": 0.95, "learning_rate": 5.205618075896686e-05, "loss": 1.9604, "step": 9088 }, { "epoch": 0.95, "learning_rate": 5.2054512226517915e-05, "loss": 0.862, "step": 9089 }, { "epoch": 0.95, "learning_rate": 5.205284354560316e-05, "loss": 0.5257, "step": 9090 }, { "epoch": 0.95, "learning_rate": 5.205117471623383e-05, "loss": 0.7922, "step": 9091 }, { "epoch": 0.95, "learning_rate": 5.204950573842114e-05, "loss": 0.6917, "step": 9092 }, { "epoch": 0.95, "learning_rate": 5.204783661217634e-05, "loss": 1.289, "step": 9093 }, { "epoch": 0.95, "learning_rate": 5.204616733751066e-05, "loss": 0.2694, "step": 9094 }, { "epoch": 0.95, "learning_rate": 5.204449791443535e-05, "loss": 1.5146, "step": 9095 }, { "epoch": 0.95, "learning_rate": 5.204282834296162e-05, "loss": 0.6731, "step": 9096 }, { "epoch": 0.95, "learning_rate": 5.204115862310074e-05, "loss": 0.2811, "step": 9097 }, { "epoch": 0.95, "learning_rate": 5.2039488754863934e-05, "loss": 0.3596, "step": 9098 }, { "epoch": 0.95, "learning_rate": 5.2037818738262446e-05, "loss": 0.3429, "step": 9099 }, { "epoch": 0.95, "learning_rate": 5.203614857330752e-05, "loss": 0.3675, "step": 9100 }, { "epoch": 0.95, "learning_rate": 5.20344782600104e-05, "loss": 1.3502, "step": 9101 }, { "epoch": 0.95, "learning_rate": 5.203280779838232e-05, "loss": 1.3517, "step": 9102 }, { "epoch": 0.95, "learning_rate": 5.2031137188434546e-05, "loss": 0.9361, "step": 9103 }, { "epoch": 0.95, "learning_rate": 5.20294664301783e-05, "loss": 2.0619, "step": 9104 }, { "epoch": 0.95, "learning_rate": 5.202779552362485e-05, "loss": 1.0982, "step": 9105 }, { "epoch": 0.95, "learning_rate": 5.202612446878542e-05, "loss": 1.2828, "step": 9106 }, { "epoch": 0.95, "learning_rate": 5.202445326567129e-05, "loss": 1.6982, "step": 9107 }, { "epoch": 0.95, "learning_rate": 5.2022781914293686e-05, "loss": 1.079, "step": 9108 }, { "epoch": 0.95, "learning_rate": 5.202111041466387e-05, "loss": 1.3418, "step": 9109 }, { "epoch": 0.95, "learning_rate": 5.201943876679309e-05, "loss": 0.6453, "step": 9110 }, { "epoch": 0.95, "learning_rate": 5.20177669706926e-05, "loss": 1.387, "step": 9111 }, { "epoch": 0.95, "learning_rate": 5.201609502637365e-05, "loss": 0.6004, "step": 9112 }, { "epoch": 0.95, "learning_rate": 5.2014422933847504e-05, "loss": 1.2726, "step": 9113 }, { "epoch": 0.95, "learning_rate": 5.2012750693125415e-05, "loss": 1.4383, "step": 9114 }, { "epoch": 0.95, "learning_rate": 5.201107830421864e-05, "loss": 1.0086, "step": 9115 }, { "epoch": 0.95, "learning_rate": 5.200940576713843e-05, "loss": 0.89, "step": 9116 }, { "epoch": 0.95, "learning_rate": 5.2007733081896055e-05, "loss": 1.3027, "step": 9117 }, { "epoch": 0.95, "learning_rate": 5.200606024850277e-05, "loss": 0.8772, "step": 9118 }, { "epoch": 0.95, "learning_rate": 5.200438726696983e-05, "loss": 1.4754, "step": 9119 }, { "epoch": 0.95, "learning_rate": 5.200271413730852e-05, "loss": 1.3167, "step": 9120 }, { "epoch": 0.95, "learning_rate": 5.2001040859530066e-05, "loss": 0.712, "step": 9121 }, { "epoch": 0.95, "learning_rate": 5.1999367433645767e-05, "loss": 0.7987, "step": 9122 }, { "epoch": 0.95, "learning_rate": 5.199769385966686e-05, "loss": 1.2243, "step": 9123 }, { "epoch": 0.95, "learning_rate": 5.199602013760464e-05, "loss": 0.5366, "step": 9124 }, { "epoch": 0.95, "learning_rate": 5.1994346267470355e-05, "loss": 0.1901, "step": 9125 }, { "epoch": 0.95, "learning_rate": 5.1992672249275275e-05, "loss": 1.3447, "step": 9126 }, { "epoch": 0.95, "learning_rate": 5.1990998083030674e-05, "loss": 1.1201, "step": 9127 }, { "epoch": 0.95, "learning_rate": 5.1989323768747814e-05, "loss": 0.4144, "step": 9128 }, { "epoch": 0.95, "learning_rate": 5.198764930643797e-05, "loss": 1.0841, "step": 9129 }, { "epoch": 0.95, "learning_rate": 5.198597469611243e-05, "loss": 1.4114, "step": 9130 }, { "epoch": 0.95, "learning_rate": 5.1984299937782445e-05, "loss": 0.8345, "step": 9131 }, { "epoch": 0.95, "learning_rate": 5.1982625031459286e-05, "loss": 1.0662, "step": 9132 }, { "epoch": 0.95, "learning_rate": 5.198094997715425e-05, "loss": 1.0687, "step": 9133 }, { "epoch": 0.95, "learning_rate": 5.1979274774878596e-05, "loss": 0.9788, "step": 9134 }, { "epoch": 0.95, "learning_rate": 5.1977599424643616e-05, "loss": 1.0382, "step": 9135 }, { "epoch": 0.95, "learning_rate": 5.197592392646057e-05, "loss": 2.3874, "step": 9136 }, { "epoch": 0.95, "learning_rate": 5.1974248280340745e-05, "loss": 0.8723, "step": 9137 }, { "epoch": 0.95, "learning_rate": 5.1972572486295426e-05, "loss": 0.6173, "step": 9138 }, { "epoch": 0.95, "learning_rate": 5.1970896544335894e-05, "loss": 0.9847, "step": 9139 }, { "epoch": 0.95, "learning_rate": 5.1969220454473426e-05, "loss": 1.3301, "step": 9140 }, { "epoch": 0.95, "learning_rate": 5.19675442167193e-05, "loss": 1.183, "step": 9141 }, { "epoch": 0.95, "learning_rate": 5.196586783108482e-05, "loss": 0.7967, "step": 9142 }, { "epoch": 0.95, "learning_rate": 5.196419129758124e-05, "loss": 1.4572, "step": 9143 }, { "epoch": 0.95, "learning_rate": 5.196251461621987e-05, "loss": 0.443, "step": 9144 }, { "epoch": 0.96, "learning_rate": 5.1960837787011996e-05, "loss": 0.6584, "step": 9145 }, { "epoch": 0.96, "learning_rate": 5.19591608099689e-05, "loss": 1.1265, "step": 9146 }, { "epoch": 0.96, "learning_rate": 5.195748368510186e-05, "loss": 0.7731, "step": 9147 }, { "epoch": 0.96, "learning_rate": 5.195580641242219e-05, "loss": 2.3693, "step": 9148 }, { "epoch": 0.96, "learning_rate": 5.195412899194116e-05, "loss": 0.842, "step": 9149 }, { "epoch": 0.96, "learning_rate": 5.195245142367008e-05, "loss": 0.7407, "step": 9150 }, { "epoch": 0.96, "learning_rate": 5.1950773707620225e-05, "loss": 0.43, "step": 9151 }, { "epoch": 0.96, "learning_rate": 5.19490958438029e-05, "loss": 0.7286, "step": 9152 }, { "epoch": 0.96, "learning_rate": 5.1947417832229396e-05, "loss": 1.6735, "step": 9153 }, { "epoch": 0.96, "learning_rate": 5.194573967291102e-05, "loss": 1.2462, "step": 9154 }, { "epoch": 0.96, "learning_rate": 5.1944061365859046e-05, "loss": 0.9622, "step": 9155 }, { "epoch": 0.96, "learning_rate": 5.194238291108479e-05, "loss": 1.1012, "step": 9156 }, { "epoch": 0.96, "learning_rate": 5.194070430859955e-05, "loss": 0.7191, "step": 9157 }, { "epoch": 0.96, "learning_rate": 5.1939025558414616e-05, "loss": 0.7067, "step": 9158 }, { "epoch": 0.96, "learning_rate": 5.19373466605413e-05, "loss": 0.9398, "step": 9159 }, { "epoch": 0.96, "learning_rate": 5.1935667614990895e-05, "loss": 0.783, "step": 9160 }, { "epoch": 0.96, "learning_rate": 5.193398842177471e-05, "loss": 0.4123, "step": 9161 }, { "epoch": 0.96, "learning_rate": 5.193230908090405e-05, "loss": 1.2476, "step": 9162 }, { "epoch": 0.96, "learning_rate": 5.193062959239022e-05, "loss": 0.4669, "step": 9163 }, { "epoch": 0.96, "learning_rate": 5.192894995624452e-05, "loss": 1.0465, "step": 9164 }, { "epoch": 0.96, "learning_rate": 5.192727017247825e-05, "loss": 0.667, "step": 9165 }, { "epoch": 0.96, "learning_rate": 5.1925590241102733e-05, "loss": 0.4248, "step": 9166 }, { "epoch": 0.96, "learning_rate": 5.192391016212928e-05, "loss": 0.3945, "step": 9167 }, { "epoch": 0.96, "learning_rate": 5.192222993556919e-05, "loss": 0.614, "step": 9168 }, { "epoch": 0.96, "learning_rate": 5.192054956143377e-05, "loss": 0.2518, "step": 9169 }, { "epoch": 0.96, "learning_rate": 5.1918869039734346e-05, "loss": 0.6815, "step": 9170 }, { "epoch": 0.96, "learning_rate": 5.191718837048223e-05, "loss": 0.4212, "step": 9171 }, { "epoch": 0.96, "learning_rate": 5.191550755368872e-05, "loss": 0.1638, "step": 9172 }, { "epoch": 0.96, "learning_rate": 5.191382658936515e-05, "loss": 0.4189, "step": 9173 }, { "epoch": 0.96, "learning_rate": 5.191214547752282e-05, "loss": 1.1041, "step": 9174 }, { "epoch": 0.96, "learning_rate": 5.191046421817305e-05, "loss": 0.967, "step": 9175 }, { "epoch": 0.96, "learning_rate": 5.190878281132717e-05, "loss": 0.5998, "step": 9176 }, { "epoch": 0.96, "learning_rate": 5.190710125699649e-05, "loss": 0.6918, "step": 9177 }, { "epoch": 0.96, "learning_rate": 5.190541955519233e-05, "loss": 1.5313, "step": 9178 }, { "epoch": 0.96, "learning_rate": 5.1903737705926e-05, "loss": 1.7092, "step": 9179 }, { "epoch": 0.96, "learning_rate": 5.1902055709208844e-05, "loss": 1.9303, "step": 9180 }, { "epoch": 0.96, "learning_rate": 5.190037356505217e-05, "loss": 0.9554, "step": 9181 }, { "epoch": 0.96, "learning_rate": 5.189869127346731e-05, "loss": 0.6619, "step": 9182 }, { "epoch": 0.96, "learning_rate": 5.189700883446558e-05, "loss": 1.2369, "step": 9183 }, { "epoch": 0.96, "learning_rate": 5.189532624805831e-05, "loss": 0.5808, "step": 9184 }, { "epoch": 0.96, "learning_rate": 5.189364351425683e-05, "loss": 0.2132, "step": 9185 }, { "epoch": 0.96, "learning_rate": 5.189196063307246e-05, "loss": 1.5696, "step": 9186 }, { "epoch": 0.96, "learning_rate": 5.1890277604516536e-05, "loss": 0.1437, "step": 9187 }, { "epoch": 0.96, "learning_rate": 5.1888594428600386e-05, "loss": 1.6883, "step": 9188 }, { "epoch": 0.96, "learning_rate": 5.188691110533534e-05, "loss": 0.225, "step": 9189 }, { "epoch": 0.96, "learning_rate": 5.188522763473273e-05, "loss": 0.4675, "step": 9190 }, { "epoch": 0.96, "learning_rate": 5.188354401680388e-05, "loss": 1.5949, "step": 9191 }, { "epoch": 0.96, "learning_rate": 5.1881860251560145e-05, "loss": 0.5123, "step": 9192 }, { "epoch": 0.96, "learning_rate": 5.1880176339012834e-05, "loss": 1.6432, "step": 9193 }, { "epoch": 0.96, "learning_rate": 5.1878492279173304e-05, "loss": 1.4043, "step": 9194 }, { "epoch": 0.96, "learning_rate": 5.1876808072052884e-05, "loss": 0.535, "step": 9195 }, { "epoch": 0.96, "learning_rate": 5.1875123717662905e-05, "loss": 0.3276, "step": 9196 }, { "epoch": 0.96, "learning_rate": 5.187343921601472e-05, "loss": 1.1659, "step": 9197 }, { "epoch": 0.96, "learning_rate": 5.187175456711964e-05, "loss": 0.2871, "step": 9198 }, { "epoch": 0.96, "learning_rate": 5.1870069770989045e-05, "loss": 1.5101, "step": 9199 }, { "epoch": 0.96, "learning_rate": 5.1868384827634255e-05, "loss": 0.9151, "step": 9200 }, { "epoch": 0.96, "learning_rate": 5.186669973706662e-05, "loss": 0.8272, "step": 9201 }, { "epoch": 0.96, "learning_rate": 5.186501449929746e-05, "loss": 0.3142, "step": 9202 }, { "epoch": 0.96, "learning_rate": 5.186332911433815e-05, "loss": 0.7335, "step": 9203 }, { "epoch": 0.96, "learning_rate": 5.186164358220002e-05, "loss": 0.5236, "step": 9204 }, { "epoch": 0.96, "learning_rate": 5.185995790289443e-05, "loss": 1.2825, "step": 9205 }, { "epoch": 0.96, "learning_rate": 5.185827207643271e-05, "loss": 0.4661, "step": 9206 }, { "epoch": 0.96, "learning_rate": 5.1856586102826225e-05, "loss": 1.1994, "step": 9207 }, { "epoch": 0.96, "learning_rate": 5.185489998208631e-05, "loss": 0.4151, "step": 9208 }, { "epoch": 0.96, "learning_rate": 5.1853213714224316e-05, "loss": 0.4333, "step": 9209 }, { "epoch": 0.96, "learning_rate": 5.185152729925161e-05, "loss": 0.9381, "step": 9210 }, { "epoch": 0.96, "learning_rate": 5.184984073717953e-05, "loss": 0.4921, "step": 9211 }, { "epoch": 0.96, "learning_rate": 5.184815402801944e-05, "loss": 0.2677, "step": 9212 }, { "epoch": 0.96, "learning_rate": 5.184646717178269e-05, "loss": 1.2095, "step": 9213 }, { "epoch": 0.96, "learning_rate": 5.184478016848063e-05, "loss": 0.6048, "step": 9214 }, { "epoch": 0.96, "learning_rate": 5.1843093018124614e-05, "loss": 1.057, "step": 9215 }, { "epoch": 0.96, "learning_rate": 5.184140572072602e-05, "loss": 0.3418, "step": 9216 }, { "epoch": 0.96, "learning_rate": 5.183971827629619e-05, "loss": 1.7853, "step": 9217 }, { "epoch": 0.96, "learning_rate": 5.183803068484648e-05, "loss": 1.1648, "step": 9218 }, { "epoch": 0.96, "learning_rate": 5.1836342946388264e-05, "loss": 0.9933, "step": 9219 }, { "epoch": 0.96, "learning_rate": 5.183465506093289e-05, "loss": 0.9673, "step": 9220 }, { "epoch": 0.96, "learning_rate": 5.183296702849173e-05, "loss": 0.4182, "step": 9221 }, { "epoch": 0.96, "learning_rate": 5.183127884907615e-05, "loss": 0.4207, "step": 9222 }, { "epoch": 0.96, "learning_rate": 5.18295905226975e-05, "loss": 0.4086, "step": 9223 }, { "epoch": 0.96, "learning_rate": 5.1827902049367165e-05, "loss": 0.9087, "step": 9224 }, { "epoch": 0.96, "learning_rate": 5.18262134290965e-05, "loss": 1.0421, "step": 9225 }, { "epoch": 0.96, "learning_rate": 5.182452466189686e-05, "loss": 0.4867, "step": 9226 }, { "epoch": 0.96, "learning_rate": 5.1822835747779636e-05, "loss": 0.3946, "step": 9227 }, { "epoch": 0.96, "learning_rate": 5.1821146686756184e-05, "loss": 0.3559, "step": 9228 }, { "epoch": 0.96, "learning_rate": 5.181945747883788e-05, "loss": 0.6355, "step": 9229 }, { "epoch": 0.96, "learning_rate": 5.18177681240361e-05, "loss": 0.8113, "step": 9230 }, { "epoch": 0.96, "learning_rate": 5.1816078622362195e-05, "loss": 1.5947, "step": 9231 }, { "epoch": 0.96, "learning_rate": 5.181438897382756e-05, "loss": 0.5044, "step": 9232 }, { "epoch": 0.96, "learning_rate": 5.1812699178443574e-05, "loss": 0.6318, "step": 9233 }, { "epoch": 0.96, "learning_rate": 5.181100923622159e-05, "loss": 1.3664, "step": 9234 }, { "epoch": 0.96, "learning_rate": 5.1809319147172995e-05, "loss": 1.4666, "step": 9235 }, { "epoch": 0.96, "learning_rate": 5.180762891130917e-05, "loss": 0.765, "step": 9236 }, { "epoch": 0.96, "learning_rate": 5.180593852864149e-05, "loss": 0.8233, "step": 9237 }, { "epoch": 0.96, "learning_rate": 5.180424799918133e-05, "loss": 0.7393, "step": 9238 }, { "epoch": 0.96, "learning_rate": 5.180255732294007e-05, "loss": 0.7325, "step": 9239 }, { "epoch": 0.97, "learning_rate": 5.1800866499929106e-05, "loss": 1.3866, "step": 9240 }, { "epoch": 0.97, "learning_rate": 5.179917553015981e-05, "loss": 0.8774, "step": 9241 }, { "epoch": 0.97, "learning_rate": 5.1797484413643555e-05, "loss": 0.9149, "step": 9242 }, { "epoch": 0.97, "learning_rate": 5.179579315039174e-05, "loss": 1.3861, "step": 9243 }, { "epoch": 0.97, "learning_rate": 5.179410174041575e-05, "loss": 1.0656, "step": 9244 }, { "epoch": 0.97, "learning_rate": 5.179241018372696e-05, "loss": 0.4661, "step": 9245 }, { "epoch": 0.97, "learning_rate": 5.179071848033677e-05, "loss": 0.5776, "step": 9246 }, { "epoch": 0.97, "learning_rate": 5.1789026630256555e-05, "loss": 0.2783, "step": 9247 }, { "epoch": 0.97, "learning_rate": 5.1787334633497714e-05, "loss": 1.5358, "step": 9248 }, { "epoch": 0.97, "learning_rate": 5.178564249007163e-05, "loss": 1.1025, "step": 9249 }, { "epoch": 0.97, "learning_rate": 5.1783950199989704e-05, "loss": 0.4371, "step": 9250 }, { "epoch": 0.97, "learning_rate": 5.178225776326332e-05, "loss": 0.1335, "step": 9251 }, { "epoch": 0.97, "learning_rate": 5.178056517990388e-05, "loss": 0.6737, "step": 9252 }, { "epoch": 0.97, "learning_rate": 5.177887244992276e-05, "loss": 1.3698, "step": 9253 }, { "epoch": 0.97, "learning_rate": 5.177717957333137e-05, "loss": 0.3904, "step": 9254 }, { "epoch": 0.97, "learning_rate": 5.1775486550141104e-05, "loss": 1.0802, "step": 9255 }, { "epoch": 0.97, "learning_rate": 5.177379338036336e-05, "loss": 0.4352, "step": 9256 }, { "epoch": 0.97, "learning_rate": 5.177210006400953e-05, "loss": 0.3131, "step": 9257 }, { "epoch": 0.97, "learning_rate": 5.177040660109103e-05, "loss": 1.1362, "step": 9258 }, { "epoch": 0.97, "learning_rate": 5.1768712991619226e-05, "loss": 0.6696, "step": 9259 }, { "epoch": 0.97, "learning_rate": 5.176701923560554e-05, "loss": 0.2509, "step": 9260 }, { "epoch": 0.97, "learning_rate": 5.176532533306139e-05, "loss": 1.0162, "step": 9261 }, { "epoch": 0.97, "learning_rate": 5.176363128399815e-05, "loss": 0.4207, "step": 9262 }, { "epoch": 0.97, "learning_rate": 5.176193708842724e-05, "loss": 0.5493, "step": 9263 }, { "epoch": 0.97, "learning_rate": 5.176024274636007e-05, "loss": 0.2736, "step": 9264 }, { "epoch": 0.97, "learning_rate": 5.1758548257808024e-05, "loss": 0.3338, "step": 9265 }, { "epoch": 0.97, "learning_rate": 5.175685362278253e-05, "loss": 1.1807, "step": 9266 }, { "epoch": 0.97, "learning_rate": 5.1755158841294984e-05, "loss": 1.0457, "step": 9267 }, { "epoch": 0.97, "learning_rate": 5.17534639133568e-05, "loss": 0.3816, "step": 9268 }, { "epoch": 0.97, "learning_rate": 5.1751768838979386e-05, "loss": 0.9894, "step": 9269 }, { "epoch": 0.97, "learning_rate": 5.175007361817416e-05, "loss": 1.2866, "step": 9270 }, { "epoch": 0.97, "learning_rate": 5.1748378250952525e-05, "loss": 0.9905, "step": 9271 }, { "epoch": 0.97, "learning_rate": 5.174668273732589e-05, "loss": 0.3321, "step": 9272 }, { "epoch": 0.97, "learning_rate": 5.1744987077305686e-05, "loss": 0.6055, "step": 9273 }, { "epoch": 0.97, "learning_rate": 5.17432912709033e-05, "loss": 0.8284, "step": 9274 }, { "epoch": 0.97, "learning_rate": 5.174159531813018e-05, "loss": 1.2312, "step": 9275 }, { "epoch": 0.97, "learning_rate": 5.173989921899772e-05, "loss": 1.0928, "step": 9276 }, { "epoch": 0.97, "learning_rate": 5.173820297351735e-05, "loss": 1.3864, "step": 9277 }, { "epoch": 0.97, "learning_rate": 5.1736506581700485e-05, "loss": 1.4992, "step": 9278 }, { "epoch": 0.97, "learning_rate": 5.173481004355854e-05, "loss": 0.4155, "step": 9279 }, { "epoch": 0.97, "learning_rate": 5.173311335910294e-05, "loss": 1.5683, "step": 9280 }, { "epoch": 0.97, "learning_rate": 5.17314165283451e-05, "loss": 1.0005, "step": 9281 }, { "epoch": 0.97, "learning_rate": 5.1729719551296455e-05, "loss": 0.5899, "step": 9282 }, { "epoch": 0.97, "learning_rate": 5.172802242796842e-05, "loss": 0.5353, "step": 9283 }, { "epoch": 0.97, "learning_rate": 5.1726325158372435e-05, "loss": 0.7818, "step": 9284 }, { "epoch": 0.97, "learning_rate": 5.172462774251989e-05, "loss": 1.5645, "step": 9285 }, { "epoch": 0.97, "learning_rate": 5.1722930180422254e-05, "loss": 1.0975, "step": 9286 }, { "epoch": 0.97, "learning_rate": 5.172123247209092e-05, "loss": 1.6023, "step": 9287 }, { "epoch": 0.97, "learning_rate": 5.171953461753734e-05, "loss": 1.4246, "step": 9288 }, { "epoch": 0.97, "learning_rate": 5.1717836616772936e-05, "loss": 1.51, "step": 9289 }, { "epoch": 0.97, "learning_rate": 5.171613846980914e-05, "loss": 0.4446, "step": 9290 }, { "epoch": 0.97, "learning_rate": 5.1714440176657364e-05, "loss": 0.477, "step": 9291 }, { "epoch": 0.97, "learning_rate": 5.171274173732907e-05, "loss": 1.2544, "step": 9292 }, { "epoch": 0.97, "learning_rate": 5.171104315183568e-05, "loss": 0.4402, "step": 9293 }, { "epoch": 0.97, "learning_rate": 5.170934442018863e-05, "loss": 0.2809, "step": 9294 }, { "epoch": 0.97, "learning_rate": 5.170764554239934e-05, "loss": 0.6241, "step": 9295 }, { "epoch": 0.97, "learning_rate": 5.1705946518479263e-05, "loss": 0.463, "step": 9296 }, { "epoch": 0.97, "learning_rate": 5.170424734843984e-05, "loss": 0.7856, "step": 9297 }, { "epoch": 0.97, "learning_rate": 5.17025480322925e-05, "loss": 1.4477, "step": 9298 }, { "epoch": 0.97, "learning_rate": 5.1700848570048684e-05, "loss": 0.4855, "step": 9299 }, { "epoch": 0.97, "learning_rate": 5.169914896171983e-05, "loss": 0.6466, "step": 9300 }, { "epoch": 0.97, "learning_rate": 5.1697449207317383e-05, "loss": 0.3222, "step": 9301 }, { "epoch": 0.97, "learning_rate": 5.16957493068528e-05, "loss": 0.8621, "step": 9302 }, { "epoch": 0.97, "learning_rate": 5.1694049260337485e-05, "loss": 0.3497, "step": 9303 }, { "epoch": 0.97, "learning_rate": 5.169234906778292e-05, "loss": 0.848, "step": 9304 }, { "epoch": 0.97, "learning_rate": 5.169064872920053e-05, "loss": 1.0783, "step": 9305 }, { "epoch": 0.97, "learning_rate": 5.1688948244601774e-05, "loss": 1.5188, "step": 9306 }, { "epoch": 0.97, "learning_rate": 5.1687247613998094e-05, "loss": 0.9163, "step": 9307 }, { "epoch": 0.97, "learning_rate": 5.168554683740094e-05, "loss": 0.5116, "step": 9308 }, { "epoch": 0.97, "learning_rate": 5.168384591482175e-05, "loss": 1.1358, "step": 9309 }, { "epoch": 0.97, "learning_rate": 5.168214484627199e-05, "loss": 1.5549, "step": 9310 }, { "epoch": 0.97, "learning_rate": 5.16804436317631e-05, "loss": 0.8174, "step": 9311 }, { "epoch": 0.97, "learning_rate": 5.167874227130653e-05, "loss": 0.8278, "step": 9312 }, { "epoch": 0.97, "learning_rate": 5.167704076491374e-05, "loss": 1.1376, "step": 9313 }, { "epoch": 0.97, "learning_rate": 5.1675339112596194e-05, "loss": 0.9143, "step": 9314 }, { "epoch": 0.97, "learning_rate": 5.167363731436534e-05, "loss": 0.4493, "step": 9315 }, { "epoch": 0.97, "learning_rate": 5.1671935370232624e-05, "loss": 0.3476, "step": 9316 }, { "epoch": 0.97, "learning_rate": 5.1670233280209506e-05, "loss": 0.7151, "step": 9317 }, { "epoch": 0.97, "learning_rate": 5.1668531044307446e-05, "loss": 0.1402, "step": 9318 }, { "epoch": 0.97, "learning_rate": 5.166682866253792e-05, "loss": 0.0974, "step": 9319 }, { "epoch": 0.97, "learning_rate": 5.1665126134912364e-05, "loss": 0.5634, "step": 9320 }, { "epoch": 0.97, "learning_rate": 5.166342346144224e-05, "loss": 0.3992, "step": 9321 }, { "epoch": 0.97, "learning_rate": 5.1661720642139027e-05, "loss": 1.0781, "step": 9322 }, { "epoch": 0.97, "learning_rate": 5.166001767701417e-05, "loss": 1.4342, "step": 9323 }, { "epoch": 0.97, "learning_rate": 5.165831456607916e-05, "loss": 1.3993, "step": 9324 }, { "epoch": 0.97, "learning_rate": 5.165661130934543e-05, "loss": 0.3582, "step": 9325 }, { "epoch": 0.97, "learning_rate": 5.1654907906824466e-05, "loss": 1.1004, "step": 9326 }, { "epoch": 0.97, "learning_rate": 5.165320435852773e-05, "loss": 0.407, "step": 9327 }, { "epoch": 0.97, "learning_rate": 5.165150066446668e-05, "loss": 0.4033, "step": 9328 }, { "epoch": 0.97, "learning_rate": 5.16497968246528e-05, "loss": 0.4699, "step": 9329 }, { "epoch": 0.97, "learning_rate": 5.164809283909755e-05, "loss": 0.2532, "step": 9330 }, { "epoch": 0.97, "learning_rate": 5.1646388707812404e-05, "loss": 1.255, "step": 9331 }, { "epoch": 0.97, "learning_rate": 5.1644684430808844e-05, "loss": 0.9476, "step": 9332 }, { "epoch": 0.97, "learning_rate": 5.1642980008098325e-05, "loss": 0.7658, "step": 9333 }, { "epoch": 0.97, "learning_rate": 5.164127543969233e-05, "loss": 1.5072, "step": 9334 }, { "epoch": 0.97, "learning_rate": 5.163957072560233e-05, "loss": 1.4047, "step": 9335 }, { "epoch": 0.98, "learning_rate": 5.1637865865839796e-05, "loss": 1.3513, "step": 9336 }, { "epoch": 0.98, "learning_rate": 5.163616086041623e-05, "loss": 1.1728, "step": 9337 }, { "epoch": 0.98, "learning_rate": 5.163445570934308e-05, "loss": 1.1652, "step": 9338 }, { "epoch": 0.98, "learning_rate": 5.163275041263184e-05, "loss": 0.4694, "step": 9339 }, { "epoch": 0.98, "learning_rate": 5.1631044970293976e-05, "loss": 1.4626, "step": 9340 }, { "epoch": 0.98, "learning_rate": 5.1629339382340986e-05, "loss": 0.8667, "step": 9341 }, { "epoch": 0.98, "learning_rate": 5.162763364878435e-05, "loss": 1.4664, "step": 9342 }, { "epoch": 0.98, "learning_rate": 5.162592776963554e-05, "loss": 1.1749, "step": 9343 }, { "epoch": 0.98, "learning_rate": 5.162422174490604e-05, "loss": 0.9436, "step": 9344 }, { "epoch": 0.98, "learning_rate": 5.162251557460735e-05, "loss": 0.9546, "step": 9345 }, { "epoch": 0.98, "learning_rate": 5.162080925875094e-05, "loss": 1.2784, "step": 9346 }, { "epoch": 0.98, "learning_rate": 5.1619102797348286e-05, "loss": 0.9411, "step": 9347 }, { "epoch": 0.98, "learning_rate": 5.1617396190410906e-05, "loss": 0.5079, "step": 9348 }, { "epoch": 0.98, "learning_rate": 5.161568943795028e-05, "loss": 0.7317, "step": 9349 }, { "epoch": 0.98, "learning_rate": 5.161398253997787e-05, "loss": 1.1902, "step": 9350 }, { "epoch": 0.98, "learning_rate": 5.161227549650519e-05, "loss": 1.1279, "step": 9351 }, { "epoch": 0.98, "learning_rate": 5.161056830754374e-05, "loss": 0.2492, "step": 9352 }, { "epoch": 0.98, "learning_rate": 5.1608860973104996e-05, "loss": 0.4719, "step": 9353 }, { "epoch": 0.98, "learning_rate": 5.160715349320046e-05, "loss": 0.4485, "step": 9354 }, { "epoch": 0.98, "learning_rate": 5.1605445867841615e-05, "loss": 1.4412, "step": 9355 }, { "epoch": 0.98, "learning_rate": 5.160373809703997e-05, "loss": 0.5626, "step": 9356 }, { "epoch": 0.98, "learning_rate": 5.160203018080701e-05, "loss": 0.5869, "step": 9357 }, { "epoch": 0.98, "learning_rate": 5.160032211915424e-05, "loss": 0.4251, "step": 9358 }, { "epoch": 0.98, "learning_rate": 5.159861391209315e-05, "loss": 1.0137, "step": 9359 }, { "epoch": 0.98, "learning_rate": 5.1596905559635245e-05, "loss": 0.9441, "step": 9360 }, { "epoch": 0.98, "learning_rate": 5.159519706179203e-05, "loss": 0.3762, "step": 9361 }, { "epoch": 0.98, "learning_rate": 5.1593488418575e-05, "loss": 1.1283, "step": 9362 }, { "epoch": 0.98, "learning_rate": 5.1591779629995654e-05, "loss": 0.9212, "step": 9363 }, { "epoch": 0.98, "learning_rate": 5.159007069606551e-05, "loss": 0.384, "step": 9364 }, { "epoch": 0.98, "learning_rate": 5.1588361616796066e-05, "loss": 0.0859, "step": 9365 }, { "epoch": 0.98, "learning_rate": 5.158665239219881e-05, "loss": 1.7234, "step": 9366 }, { "epoch": 0.98, "learning_rate": 5.158494302228525e-05, "loss": 0.6738, "step": 9367 }, { "epoch": 0.98, "learning_rate": 5.1583233507066924e-05, "loss": 0.9733, "step": 9368 }, { "epoch": 0.98, "learning_rate": 5.158152384655532e-05, "loss": 0.5288, "step": 9369 }, { "epoch": 0.98, "learning_rate": 5.157981404076193e-05, "loss": 1.319, "step": 9370 }, { "epoch": 0.98, "learning_rate": 5.15781040896983e-05, "loss": 0.6882, "step": 9371 }, { "epoch": 0.98, "learning_rate": 5.157639399337592e-05, "loss": 0.4915, "step": 9372 }, { "epoch": 0.98, "learning_rate": 5.157468375180629e-05, "loss": 1.7982, "step": 9373 }, { "epoch": 0.98, "learning_rate": 5.157297336500095e-05, "loss": 0.343, "step": 9374 }, { "epoch": 0.98, "learning_rate": 5.1571262832971395e-05, "loss": 1.2723, "step": 9375 }, { "epoch": 0.98, "learning_rate": 5.1569552155729146e-05, "loss": 0.4895, "step": 9376 }, { "epoch": 0.98, "learning_rate": 5.156784133328572e-05, "loss": 1.51, "step": 9377 }, { "epoch": 0.98, "learning_rate": 5.156613036565264e-05, "loss": 0.5174, "step": 9378 }, { "epoch": 0.98, "learning_rate": 5.1564419252841404e-05, "loss": 0.1348, "step": 9379 }, { "epoch": 0.98, "learning_rate": 5.156270799486355e-05, "loss": 0.1778, "step": 9380 }, { "epoch": 0.98, "learning_rate": 5.156099659173059e-05, "loss": 0.4081, "step": 9381 }, { "epoch": 0.98, "learning_rate": 5.155928504345404e-05, "loss": 0.6211, "step": 9382 }, { "epoch": 0.98, "learning_rate": 5.1557573350045434e-05, "loss": 1.013, "step": 9383 }, { "epoch": 0.98, "learning_rate": 5.155586151151629e-05, "loss": 0.4348, "step": 9384 }, { "epoch": 0.98, "learning_rate": 5.155414952787813e-05, "loss": 1.9017, "step": 9385 }, { "epoch": 0.98, "learning_rate": 5.1552437399142476e-05, "loss": 0.7385, "step": 9386 }, { "epoch": 0.98, "learning_rate": 5.155072512532085e-05, "loss": 0.8081, "step": 9387 }, { "epoch": 0.98, "learning_rate": 5.15490127064248e-05, "loss": 0.4349, "step": 9388 }, { "epoch": 0.98, "learning_rate": 5.1547300142465826e-05, "loss": 0.2914, "step": 9389 }, { "epoch": 0.98, "learning_rate": 5.154558743345547e-05, "loss": 1.4256, "step": 9390 }, { "epoch": 0.98, "learning_rate": 5.1543874579405266e-05, "loss": 1.3798, "step": 9391 }, { "epoch": 0.98, "learning_rate": 5.154216158032673e-05, "loss": 0.1833, "step": 9392 }, { "epoch": 0.98, "learning_rate": 5.154044843623142e-05, "loss": 0.1945, "step": 9393 }, { "epoch": 0.98, "learning_rate": 5.153873514713084e-05, "loss": 0.2981, "step": 9394 }, { "epoch": 0.98, "learning_rate": 5.153702171303654e-05, "loss": 0.991, "step": 9395 }, { "epoch": 0.98, "learning_rate": 5.153530813396003e-05, "loss": 0.3453, "step": 9396 }, { "epoch": 0.98, "learning_rate": 5.153359440991289e-05, "loss": 1.415, "step": 9397 }, { "epoch": 0.98, "learning_rate": 5.153188054090662e-05, "loss": 0.2786, "step": 9398 }, { "epoch": 0.98, "learning_rate": 5.153016652695276e-05, "loss": 0.8424, "step": 9399 }, { "epoch": 0.98, "learning_rate": 5.152845236806286e-05, "loss": 0.884, "step": 9400 }, { "epoch": 0.98, "learning_rate": 5.1526738064248465e-05, "loss": 0.3911, "step": 9401 }, { "epoch": 0.98, "learning_rate": 5.1525023615521104e-05, "loss": 1.2156, "step": 9402 }, { "epoch": 0.98, "learning_rate": 5.1523309021892325e-05, "loss": 0.8004, "step": 9403 }, { "epoch": 0.98, "learning_rate": 5.1521594283373655e-05, "loss": 0.6103, "step": 9404 }, { "epoch": 0.98, "learning_rate": 5.1519879399976646e-05, "loss": 1.1681, "step": 9405 }, { "epoch": 0.98, "learning_rate": 5.151816437171286e-05, "loss": 0.3959, "step": 9406 }, { "epoch": 0.98, "learning_rate": 5.151644919859381e-05, "loss": 0.2283, "step": 9407 }, { "epoch": 0.98, "learning_rate": 5.151473388063107e-05, "loss": 0.9531, "step": 9408 }, { "epoch": 0.98, "learning_rate": 5.1513018417836164e-05, "loss": 1.5941, "step": 9409 }, { "epoch": 0.98, "learning_rate": 5.151130281022065e-05, "loss": 1.0823, "step": 9410 }, { "epoch": 0.98, "learning_rate": 5.150958705779609e-05, "loss": 1.0539, "step": 9411 }, { "epoch": 0.98, "learning_rate": 5.1507871160574024e-05, "loss": 0.8529, "step": 9412 }, { "epoch": 0.98, "learning_rate": 5.150615511856599e-05, "loss": 0.5795, "step": 9413 }, { "epoch": 0.98, "learning_rate": 5.1504438931783566e-05, "loss": 1.0194, "step": 9414 }, { "epoch": 0.98, "learning_rate": 5.150272260023829e-05, "loss": 1.0402, "step": 9415 }, { "epoch": 0.98, "learning_rate": 5.1501006123941707e-05, "loss": 0.7297, "step": 9416 }, { "epoch": 0.98, "learning_rate": 5.149928950290538e-05, "loss": 0.2904, "step": 9417 }, { "epoch": 0.98, "learning_rate": 5.1497572737140875e-05, "loss": 1.0807, "step": 9418 }, { "epoch": 0.98, "learning_rate": 5.149585582665974e-05, "loss": 0.8638, "step": 9419 }, { "epoch": 0.98, "learning_rate": 5.149413877147353e-05, "loss": 1.87, "step": 9420 }, { "epoch": 0.98, "learning_rate": 5.1492421571593804e-05, "loss": 0.9258, "step": 9421 }, { "epoch": 0.98, "learning_rate": 5.149070422703213e-05, "loss": 0.3554, "step": 9422 }, { "epoch": 0.98, "learning_rate": 5.148898673780006e-05, "loss": 1.1527, "step": 9423 }, { "epoch": 0.98, "learning_rate": 5.1487269103909156e-05, "loss": 0.3467, "step": 9424 }, { "epoch": 0.98, "learning_rate": 5.1485551325370996e-05, "loss": 0.4628, "step": 9425 }, { "epoch": 0.98, "learning_rate": 5.148383340219712e-05, "loss": 0.7677, "step": 9426 }, { "epoch": 0.98, "learning_rate": 5.1482115334399115e-05, "loss": 0.3976, "step": 9427 }, { "epoch": 0.98, "learning_rate": 5.148039712198853e-05, "loss": 0.4618, "step": 9428 }, { "epoch": 0.98, "learning_rate": 5.1478678764976934e-05, "loss": 0.6109, "step": 9429 }, { "epoch": 0.98, "learning_rate": 5.14769602633759e-05, "loss": 1.5641, "step": 9430 }, { "epoch": 0.98, "learning_rate": 5.1475241617197e-05, "loss": 1.3774, "step": 9431 }, { "epoch": 0.99, "learning_rate": 5.147352282645179e-05, "loss": 1.041, "step": 9432 }, { "epoch": 0.99, "learning_rate": 5.147180389115186e-05, "loss": 1.3668, "step": 9433 }, { "epoch": 0.99, "learning_rate": 5.147008481130876e-05, "loss": 1.3188, "step": 9434 }, { "epoch": 0.99, "learning_rate": 5.1468365586934074e-05, "loss": 0.6578, "step": 9435 }, { "epoch": 0.99, "learning_rate": 5.146664621803938e-05, "loss": 0.889, "step": 9436 }, { "epoch": 0.99, "learning_rate": 5.146492670463624e-05, "loss": 0.6336, "step": 9437 }, { "epoch": 0.99, "learning_rate": 5.146320704673624e-05, "loss": 0.5685, "step": 9438 }, { "epoch": 0.99, "learning_rate": 5.1461487244350946e-05, "loss": 1.4287, "step": 9439 }, { "epoch": 0.99, "learning_rate": 5.1459767297491945e-05, "loss": 1.2095, "step": 9440 }, { "epoch": 0.99, "learning_rate": 5.145804720617082e-05, "loss": 1.0918, "step": 9441 }, { "epoch": 0.99, "learning_rate": 5.1456326970399124e-05, "loss": 1.0067, "step": 9442 }, { "epoch": 0.99, "learning_rate": 5.145460659018848e-05, "loss": 1.0975, "step": 9443 }, { "epoch": 0.99, "learning_rate": 5.145288606555042e-05, "loss": 0.4304, "step": 9444 }, { "epoch": 0.99, "learning_rate": 5.145116539649656e-05, "loss": 0.2467, "step": 9445 }, { "epoch": 0.99, "learning_rate": 5.144944458303848e-05, "loss": 1.453, "step": 9446 }, { "epoch": 0.99, "learning_rate": 5.144772362518775e-05, "loss": 1.3743, "step": 9447 }, { "epoch": 0.99, "learning_rate": 5.144600252295597e-05, "loss": 0.308, "step": 9448 }, { "epoch": 0.99, "learning_rate": 5.144428127635471e-05, "loss": 0.2833, "step": 9449 }, { "epoch": 0.99, "learning_rate": 5.144255988539557e-05, "loss": 1.4551, "step": 9450 }, { "epoch": 0.99, "learning_rate": 5.1440838350090136e-05, "loss": 1.3224, "step": 9451 }, { "epoch": 0.99, "learning_rate": 5.1439116670449994e-05, "loss": 0.3288, "step": 9452 }, { "epoch": 0.99, "learning_rate": 5.143739484648674e-05, "loss": 0.9937, "step": 9453 }, { "epoch": 0.99, "learning_rate": 5.143567287821195e-05, "loss": 1.0446, "step": 9454 }, { "epoch": 0.99, "learning_rate": 5.143395076563723e-05, "loss": 0.382, "step": 9455 }, { "epoch": 0.99, "learning_rate": 5.143222850877417e-05, "loss": 1.2878, "step": 9456 }, { "epoch": 0.99, "learning_rate": 5.143050610763436e-05, "loss": 0.9292, "step": 9457 }, { "epoch": 0.99, "learning_rate": 5.142878356222939e-05, "loss": 0.8825, "step": 9458 }, { "epoch": 0.99, "learning_rate": 5.1427060872570874e-05, "loss": 1.0213, "step": 9459 }, { "epoch": 0.99, "learning_rate": 5.1425338038670396e-05, "loss": 1.4437, "step": 9460 }, { "epoch": 0.99, "learning_rate": 5.142361506053956e-05, "loss": 1.0212, "step": 9461 }, { "epoch": 0.99, "learning_rate": 5.1421891938189954e-05, "loss": 0.8916, "step": 9462 }, { "epoch": 0.99, "learning_rate": 5.142016867163318e-05, "loss": 0.3664, "step": 9463 }, { "epoch": 0.99, "learning_rate": 5.141844526088085e-05, "loss": 1.4484, "step": 9464 }, { "epoch": 0.99, "learning_rate": 5.1416721705944556e-05, "loss": 0.4954, "step": 9465 }, { "epoch": 0.99, "learning_rate": 5.141499800683589e-05, "loss": 1.0438, "step": 9466 }, { "epoch": 0.99, "learning_rate": 5.141327416356649e-05, "loss": 0.4269, "step": 9467 }, { "epoch": 0.99, "learning_rate": 5.1411550176147926e-05, "loss": 1.1077, "step": 9468 }, { "epoch": 0.99, "learning_rate": 5.1409826044591814e-05, "loss": 0.2044, "step": 9469 }, { "epoch": 0.99, "learning_rate": 5.1408101768909766e-05, "loss": 0.7019, "step": 9470 }, { "epoch": 0.99, "learning_rate": 5.1406377349113384e-05, "loss": 0.4612, "step": 9471 }, { "epoch": 0.99, "learning_rate": 5.140465278521429e-05, "loss": 0.61, "step": 9472 }, { "epoch": 0.99, "learning_rate": 5.1402928077224074e-05, "loss": 1.3617, "step": 9473 }, { "epoch": 0.99, "learning_rate": 5.1401203225154354e-05, "loss": 0.3503, "step": 9474 }, { "epoch": 0.99, "learning_rate": 5.139947822901674e-05, "loss": 0.3412, "step": 9475 }, { "epoch": 0.99, "learning_rate": 5.1397753088822845e-05, "loss": 2.2677, "step": 9476 }, { "epoch": 0.99, "learning_rate": 5.139602780458428e-05, "loss": 1.7971, "step": 9477 }, { "epoch": 0.99, "learning_rate": 5.139430237631267e-05, "loss": 1.1714, "step": 9478 }, { "epoch": 0.99, "learning_rate": 5.139257680401962e-05, "loss": 0.4604, "step": 9479 }, { "epoch": 0.99, "learning_rate": 5.139085108771674e-05, "loss": 1.7489, "step": 9480 }, { "epoch": 0.99, "learning_rate": 5.138912522741566e-05, "loss": 1.2209, "step": 9481 }, { "epoch": 0.99, "learning_rate": 5.1387399223128e-05, "loss": 0.7548, "step": 9482 }, { "epoch": 0.99, "learning_rate": 5.1385673074865364e-05, "loss": 0.592, "step": 9483 }, { "epoch": 0.99, "learning_rate": 5.1383946782639395e-05, "loss": 0.7263, "step": 9484 }, { "epoch": 0.99, "learning_rate": 5.138222034646168e-05, "loss": 0.6995, "step": 9485 }, { "epoch": 0.99, "learning_rate": 5.138049376634388e-05, "loss": 0.739, "step": 9486 }, { "epoch": 0.99, "learning_rate": 5.137876704229759e-05, "loss": 0.4684, "step": 9487 }, { "epoch": 0.99, "learning_rate": 5.137704017433443e-05, "loss": 1.1523, "step": 9488 }, { "epoch": 0.99, "learning_rate": 5.1375313162466054e-05, "loss": 0.6601, "step": 9489 }, { "epoch": 0.99, "learning_rate": 5.137358600670407e-05, "loss": 0.8779, "step": 9490 }, { "epoch": 0.99, "learning_rate": 5.1371858707060096e-05, "loss": 1.206, "step": 9491 }, { "epoch": 0.99, "learning_rate": 5.137013126354578e-05, "loss": 0.3896, "step": 9492 }, { "epoch": 0.99, "learning_rate": 5.136840367617274e-05, "loss": 1.471, "step": 9493 }, { "epoch": 0.99, "learning_rate": 5.13666759449526e-05, "loss": 1.1453, "step": 9494 }, { "epoch": 0.99, "learning_rate": 5.1364948069896994e-05, "loss": 1.0322, "step": 9495 }, { "epoch": 0.99, "learning_rate": 5.1363220051017565e-05, "loss": 1.0106, "step": 9496 }, { "epoch": 0.99, "learning_rate": 5.136149188832594e-05, "loss": 1.4804, "step": 9497 }, { "epoch": 0.99, "learning_rate": 5.135976358183374e-05, "loss": 0.8264, "step": 9498 }, { "epoch": 0.99, "learning_rate": 5.1358035131552613e-05, "loss": 0.5217, "step": 9499 }, { "epoch": 0.99, "learning_rate": 5.1356306537494194e-05, "loss": 1.1459, "step": 9500 }, { "epoch": 0.99, "learning_rate": 5.135457779967012e-05, "loss": 0.998, "step": 9501 }, { "epoch": 0.99, "learning_rate": 5.135284891809201e-05, "loss": 1.0409, "step": 9502 }, { "epoch": 0.99, "learning_rate": 5.135111989277152e-05, "loss": 0.5417, "step": 9503 }, { "epoch": 0.99, "learning_rate": 5.13493907237203e-05, "loss": 0.9369, "step": 9504 }, { "epoch": 0.99, "learning_rate": 5.134766141094997e-05, "loss": 0.5949, "step": 9505 }, { "epoch": 0.99, "learning_rate": 5.134593195447218e-05, "loss": 0.496, "step": 9506 }, { "epoch": 0.99, "learning_rate": 5.134420235429857e-05, "loss": 0.2992, "step": 9507 }, { "epoch": 0.99, "learning_rate": 5.134247261044078e-05, "loss": 1.634, "step": 9508 }, { "epoch": 0.99, "learning_rate": 5.134074272291046e-05, "loss": 0.2602, "step": 9509 }, { "epoch": 0.99, "learning_rate": 5.133901269171925e-05, "loss": 0.5283, "step": 9510 }, { "epoch": 0.99, "learning_rate": 5.133728251687881e-05, "loss": 0.8557, "step": 9511 }, { "epoch": 0.99, "learning_rate": 5.133555219840077e-05, "loss": 0.3399, "step": 9512 }, { "epoch": 0.99, "learning_rate": 5.133382173629678e-05, "loss": 0.7833, "step": 9513 }, { "epoch": 0.99, "learning_rate": 5.13320911305785e-05, "loss": 1.337, "step": 9514 }, { "epoch": 0.99, "learning_rate": 5.133036038125758e-05, "loss": 0.8972, "step": 9515 }, { "epoch": 0.99, "learning_rate": 5.132862948834565e-05, "loss": 2.3711, "step": 9516 }, { "epoch": 0.99, "learning_rate": 5.132689845185439e-05, "loss": 0.5413, "step": 9517 }, { "epoch": 0.99, "learning_rate": 5.1325167271795435e-05, "loss": 1.315, "step": 9518 }, { "epoch": 0.99, "learning_rate": 5.132343594818044e-05, "loss": 0.93, "step": 9519 }, { "epoch": 0.99, "learning_rate": 5.1321704481021064e-05, "loss": 0.6025, "step": 9520 }, { "epoch": 0.99, "learning_rate": 5.131997287032896e-05, "loss": 0.983, "step": 9521 }, { "epoch": 0.99, "learning_rate": 5.1318241116115794e-05, "loss": 0.7774, "step": 9522 }, { "epoch": 0.99, "learning_rate": 5.131650921839321e-05, "loss": 1.2514, "step": 9523 }, { "epoch": 0.99, "learning_rate": 5.131477717717288e-05, "loss": 1.0395, "step": 9524 }, { "epoch": 0.99, "learning_rate": 5.131304499246646e-05, "loss": 0.1832, "step": 9525 }, { "epoch": 0.99, "learning_rate": 5.13113126642856e-05, "loss": 1.4602, "step": 9526 }, { "epoch": 0.99, "learning_rate": 5.1309580192641977e-05, "loss": 0.7632, "step": 9527 }, { "epoch": 1.0, "learning_rate": 5.130784757754724e-05, "loss": 0.986, "step": 9528 }, { "epoch": 1.0, "learning_rate": 5.1306114819013055e-05, "loss": 0.5805, "step": 9529 }, { "epoch": 1.0, "learning_rate": 5.13043819170511e-05, "loss": 1.0715, "step": 9530 }, { "epoch": 1.0, "learning_rate": 5.1302648871673024e-05, "loss": 0.432, "step": 9531 }, { "epoch": 1.0, "learning_rate": 5.130091568289051e-05, "loss": 0.7595, "step": 9532 }, { "epoch": 1.0, "learning_rate": 5.12991823507152e-05, "loss": 1.3401, "step": 9533 }, { "epoch": 1.0, "learning_rate": 5.12974488751588e-05, "loss": 1.0664, "step": 9534 }, { "epoch": 1.0, "learning_rate": 5.1295715256232933e-05, "loss": 0.37, "step": 9535 }, { "epoch": 1.0, "learning_rate": 5.129398149394931e-05, "loss": 1.0782, "step": 9536 }, { "epoch": 1.0, "learning_rate": 5.129224758831958e-05, "loss": 0.4073, "step": 9537 }, { "epoch": 1.0, "learning_rate": 5.1290513539355414e-05, "loss": 1.0897, "step": 9538 }, { "epoch": 1.0, "learning_rate": 5.128877934706851e-05, "loss": 0.289, "step": 9539 }, { "epoch": 1.0, "learning_rate": 5.128704501147051e-05, "loss": 0.5758, "step": 9540 }, { "epoch": 1.0, "learning_rate": 5.1285310532573106e-05, "loss": 1.6004, "step": 9541 }, { "epoch": 1.0, "learning_rate": 5.128357591038797e-05, "loss": 0.7598, "step": 9542 }, { "epoch": 1.0, "learning_rate": 5.128184114492679e-05, "loss": 1.0593, "step": 9543 }, { "epoch": 1.0, "learning_rate": 5.128010623620123e-05, "loss": 1.1365, "step": 9544 }, { "epoch": 1.0, "learning_rate": 5.127837118422297e-05, "loss": 1.0489, "step": 9545 }, { "epoch": 1.0, "learning_rate": 5.1276635989003705e-05, "loss": 0.5503, "step": 9546 }, { "epoch": 1.0, "learning_rate": 5.1274900650555096e-05, "loss": 0.8789, "step": 9547 }, { "epoch": 1.0, "learning_rate": 5.127316516888884e-05, "loss": 1.357, "step": 9548 }, { "epoch": 1.0, "learning_rate": 5.1271429544016605e-05, "loss": 1.3899, "step": 9549 }, { "epoch": 1.0, "learning_rate": 5.126969377595009e-05, "loss": 0.3202, "step": 9550 }, { "epoch": 1.0, "learning_rate": 5.126795786470097e-05, "loss": 0.933, "step": 9551 }, { "epoch": 1.0, "learning_rate": 5.126622181028093e-05, "loss": 0.4586, "step": 9552 }, { "epoch": 1.0, "learning_rate": 5.126448561270168e-05, "loss": 0.1695, "step": 9553 }, { "epoch": 1.0, "learning_rate": 5.1262749271974866e-05, "loss": 1.3111, "step": 9554 }, { "epoch": 1.0, "learning_rate": 5.1261012788112206e-05, "loss": 0.725, "step": 9555 }, { "epoch": 1.0, "learning_rate": 5.1259276161125385e-05, "loss": 0.4074, "step": 9556 }, { "epoch": 1.0, "learning_rate": 5.125753939102609e-05, "loss": 0.7848, "step": 9557 }, { "epoch": 1.0, "learning_rate": 5.1255802477826014e-05, "loss": 0.583, "step": 9558 }, { "epoch": 1.0, "learning_rate": 5.125406542153685e-05, "loss": 0.8072, "step": 9559 }, { "epoch": 1.0, "learning_rate": 5.125232822217029e-05, "loss": 1.1767, "step": 9560 }, { "epoch": 1.0, "learning_rate": 5.125059087973803e-05, "loss": 0.9618, "step": 9561 }, { "epoch": 1.0, "learning_rate": 5.124885339425175e-05, "loss": 1.1766, "step": 9562 }, { "epoch": 1.0, "learning_rate": 5.124711576572318e-05, "loss": 1.3533, "step": 9563 }, { "epoch": 1.0, "learning_rate": 5.124537799416399e-05, "loss": 0.2651, "step": 9564 }, { "epoch": 1.0, "learning_rate": 5.1243640079585896e-05, "loss": 0.3644, "step": 9565 }, { "epoch": 1.0, "learning_rate": 5.124190202200057e-05, "loss": 0.9501, "step": 9566 }, { "epoch": 1.0, "learning_rate": 5.1240163821419745e-05, "loss": 0.7825, "step": 9567 }, { "epoch": 1.0, "learning_rate": 5.12384254778551e-05, "loss": 0.687, "step": 9568 }, { "epoch": 1.0, "learning_rate": 5.123668699131835e-05, "loss": 0.5993, "step": 9569 }, { "epoch": 1.0, "learning_rate": 5.123494836182119e-05, "loss": 0.6949, "step": 9570 }, { "epoch": 1.0, "learning_rate": 5.123320958937532e-05, "loss": 1.0941, "step": 9571 }, { "epoch": 1.0, "learning_rate": 5.1231470673992455e-05, "loss": 1.3244, "step": 9572 }, { "epoch": 1.0, "learning_rate": 5.1229731615684295e-05, "loss": 1.3357, "step": 9573 }, { "epoch": 1.0, "learning_rate": 5.122799241446255e-05, "loss": 1.428, "step": 9574 }, { "epoch": 1.0, "learning_rate": 5.122625307033893e-05, "loss": 1.7404, "step": 9575 }, { "epoch": 1.0, "learning_rate": 5.122451358332514e-05, "loss": 1.4001, "step": 9576 }, { "epoch": 1.0, "learning_rate": 5.122277395343288e-05, "loss": 0.8833, "step": 9577 }, { "epoch": 1.0, "learning_rate": 5.122103418067389e-05, "loss": 0.3943, "step": 9578 }, { "epoch": 1.0, "learning_rate": 5.1219294265059855e-05, "loss": 2.0173, "step": 9579 }, { "epoch": 1.0, "learning_rate": 5.121755420660249e-05, "loss": 1.8144, "step": 9580 }, { "epoch": 1.0, "learning_rate": 5.121581400531353e-05, "loss": 1.2383, "step": 9581 }, { "epoch": 1.0, "learning_rate": 5.121407366120467e-05, "loss": 0.9148, "step": 9582 }, { "epoch": 1.0, "learning_rate": 5.121233317428761e-05, "loss": 1.3213, "step": 9583 }, { "epoch": 1.0, "learning_rate": 5.12105925445741e-05, "loss": 1.0048, "step": 9584 }, { "epoch": 1.0, "learning_rate": 5.120885177207585e-05, "loss": 0.9154, "step": 9585 }, { "epoch": 1.0, "learning_rate": 5.1207110856804564e-05, "loss": 1.3446, "step": 9586 }, { "epoch": 1.0, "learning_rate": 5.1205369798771976e-05, "loss": 0.9722, "step": 9587 }, { "epoch": 1.0, "learning_rate": 5.12036285979898e-05, "loss": 0.323, "step": 9588 }, { "epoch": 1.0, "learning_rate": 5.120188725446975e-05, "loss": 1.2183, "step": 9589 }, { "epoch": 1.0, "learning_rate": 5.120014576822356e-05, "loss": 1.2507, "step": 9590 }, { "epoch": 1.0, "learning_rate": 5.119840413926295e-05, "loss": 0.2978, "step": 9591 }, { "epoch": 1.0, "learning_rate": 5.119666236759965e-05, "loss": 0.7127, "step": 9592 }, { "epoch": 1.0, "learning_rate": 5.1194920453245366e-05, "loss": 0.6296, "step": 9593 }, { "epoch": 1.0, "learning_rate": 5.119317839621184e-05, "loss": 1.4017, "step": 9594 }, { "epoch": 1.0, "learning_rate": 5.1191436196510794e-05, "loss": 1.1003, "step": 9595 }, { "epoch": 1.0, "learning_rate": 5.118969385415396e-05, "loss": 0.5396, "step": 9596 }, { "epoch": 1.0, "learning_rate": 5.118795136915307e-05, "loss": 1.4138, "step": 9597 }, { "epoch": 1.0, "learning_rate": 5.1186208741519845e-05, "loss": 0.5037, "step": 9598 }, { "epoch": 1.0, "learning_rate": 5.1184465971266025e-05, "loss": 0.1877, "step": 9599 }, { "epoch": 1.0, "learning_rate": 5.118272305840333e-05, "loss": 0.7416, "step": 9600 }, { "epoch": 1.0, "learning_rate": 5.11809800029435e-05, "loss": 0.6412, "step": 9601 }, { "epoch": 1.0, "learning_rate": 5.1179236804898274e-05, "loss": 1.4192, "step": 9602 }, { "epoch": 1.0, "learning_rate": 5.117749346427938e-05, "loss": 1.4495, "step": 9603 }, { "epoch": 1.0, "learning_rate": 5.117574998109855e-05, "loss": 0.7899, "step": 9604 }, { "epoch": 1.0, "learning_rate": 5.117400635536753e-05, "loss": 0.7013, "step": 9605 }, { "epoch": 1.0, "learning_rate": 5.117226258709805e-05, "loss": 0.9718, "step": 9606 }, { "epoch": 1.0, "learning_rate": 5.117051867630186e-05, "loss": 0.4856, "step": 9607 }, { "epoch": 1.0, "learning_rate": 5.1168774622990686e-05, "loss": 1.3812, "step": 9608 }, { "epoch": 1.0, "learning_rate": 5.116703042717627e-05, "loss": 0.5088, "step": 9609 }, { "epoch": 1.0, "learning_rate": 5.116528608887036e-05, "loss": 1.0062, "step": 9610 }, { "epoch": 1.0, "learning_rate": 5.116354160808471e-05, "loss": 0.9495, "step": 9611 }, { "epoch": 1.0, "learning_rate": 5.1161796984831036e-05, "loss": 1.8196, "step": 9612 }, { "epoch": 1.0, "learning_rate": 5.1160052219121095e-05, "loss": 0.6125, "step": 9613 }, { "epoch": 1.0, "learning_rate": 5.115830731096664e-05, "loss": 0.36, "step": 9614 }, { "epoch": 1.0, "learning_rate": 5.1156562260379405e-05, "loss": 1.1244, "step": 9615 }, { "epoch": 1.0, "learning_rate": 5.1154817067371144e-05, "loss": 1.5127, "step": 9616 }, { "epoch": 1.0, "learning_rate": 5.115307173195361e-05, "loss": 0.7847, "step": 9617 }, { "epoch": 1.0, "learning_rate": 5.1151326254138545e-05, "loss": 0.8676, "step": 9618 }, { "epoch": 1.0, "learning_rate": 5.11495806339377e-05, "loss": 0.5505, "step": 9619 }, { "epoch": 1.0, "learning_rate": 5.114783487136283e-05, "loss": 1.0315, "step": 9620 }, { "epoch": 1.0, "learning_rate": 5.114608896642567e-05, "loss": 0.5758, "step": 9621 }, { "epoch": 1.0, "learning_rate": 5.1144342919137995e-05, "loss": 1.5187, "step": 9622 }, { "epoch": 1.01, "learning_rate": 5.1142596729511554e-05, "loss": 1.347, "step": 9623 }, { "epoch": 1.01, "learning_rate": 5.11408503975581e-05, "loss": 0.4632, "step": 9624 }, { "epoch": 1.01, "learning_rate": 5.113910392328939e-05, "loss": 0.9605, "step": 9625 }, { "epoch": 1.01, "learning_rate": 5.1137357306717174e-05, "loss": 0.9684, "step": 9626 }, { "epoch": 1.01, "learning_rate": 5.1135610547853205e-05, "loss": 1.304, "step": 9627 }, { "epoch": 1.01, "learning_rate": 5.113386364670927e-05, "loss": 1.0727, "step": 9628 }, { "epoch": 1.01, "learning_rate": 5.113211660329709e-05, "loss": 1.0008, "step": 9629 }, { "epoch": 1.01, "learning_rate": 5.113036941762846e-05, "loss": 0.843, "step": 9630 }, { "epoch": 1.01, "learning_rate": 5.112862208971512e-05, "loss": 0.9073, "step": 9631 }, { "epoch": 1.01, "learning_rate": 5.1126874619568845e-05, "loss": 0.514, "step": 9632 }, { "epoch": 1.01, "learning_rate": 5.112512700720139e-05, "loss": 1.1082, "step": 9633 }, { "epoch": 1.01, "learning_rate": 5.1123379252624515e-05, "loss": 1.1709, "step": 9634 }, { "epoch": 1.01, "learning_rate": 5.112163135585001e-05, "loss": 0.8343, "step": 9635 }, { "epoch": 1.01, "learning_rate": 5.111988331688962e-05, "loss": 1.363, "step": 9636 }, { "epoch": 1.01, "learning_rate": 5.111813513575511e-05, "loss": 0.6347, "step": 9637 }, { "epoch": 1.01, "learning_rate": 5.111638681245826e-05, "loss": 1.4699, "step": 9638 }, { "epoch": 1.01, "learning_rate": 5.111463834701083e-05, "loss": 0.8915, "step": 9639 }, { "epoch": 1.01, "learning_rate": 5.11128897394246e-05, "loss": 1.0085, "step": 9640 }, { "epoch": 1.01, "learning_rate": 5.111114098971133e-05, "loss": 1.5193, "step": 9641 }, { "epoch": 1.01, "learning_rate": 5.110939209788281e-05, "loss": 0.1514, "step": 9642 }, { "epoch": 1.01, "learning_rate": 5.11076430639508e-05, "loss": 0.9337, "step": 9643 }, { "epoch": 1.01, "learning_rate": 5.110589388792706e-05, "loss": 0.5302, "step": 9644 }, { "epoch": 1.01, "learning_rate": 5.11041445698234e-05, "loss": 0.8044, "step": 9645 }, { "epoch": 1.01, "learning_rate": 5.1102395109651566e-05, "loss": 0.2891, "step": 9646 }, { "epoch": 1.01, "learning_rate": 5.110064550742335e-05, "loss": 1.5047, "step": 9647 }, { "epoch": 1.01, "learning_rate": 5.109889576315052e-05, "loss": 0.718, "step": 9648 }, { "epoch": 1.01, "learning_rate": 5.1097145876844874e-05, "loss": 0.6565, "step": 9649 }, { "epoch": 1.01, "learning_rate": 5.109539584851817e-05, "loss": 0.955, "step": 9650 }, { "epoch": 1.01, "learning_rate": 5.109364567818219e-05, "loss": 0.8506, "step": 9651 }, { "epoch": 1.01, "learning_rate": 5.1091895365848727e-05, "loss": 0.6031, "step": 9652 }, { "epoch": 1.01, "learning_rate": 5.109014491152956e-05, "loss": 1.0391, "step": 9653 }, { "epoch": 1.01, "learning_rate": 5.108839431523647e-05, "loss": 0.4973, "step": 9654 }, { "epoch": 1.01, "learning_rate": 5.108664357698125e-05, "loss": 1.3111, "step": 9655 }, { "epoch": 1.01, "learning_rate": 5.1084892696775687e-05, "loss": 0.4176, "step": 9656 }, { "epoch": 1.01, "learning_rate": 5.108314167463154e-05, "loss": 1.9081, "step": 9657 }, { "epoch": 1.01, "learning_rate": 5.108139051056062e-05, "loss": 0.9307, "step": 9658 }, { "epoch": 1.01, "learning_rate": 5.1079639204574725e-05, "loss": 0.6598, "step": 9659 }, { "epoch": 1.01, "learning_rate": 5.107788775668562e-05, "loss": 0.8926, "step": 9660 }, { "epoch": 1.01, "learning_rate": 5.107613616690511e-05, "loss": 0.6584, "step": 9661 }, { "epoch": 1.01, "learning_rate": 5.107438443524498e-05, "loss": 1.3235, "step": 9662 }, { "epoch": 1.01, "learning_rate": 5.107263256171703e-05, "loss": 1.0744, "step": 9663 }, { "epoch": 1.01, "learning_rate": 5.107088054633305e-05, "loss": 0.4897, "step": 9664 }, { "epoch": 1.01, "learning_rate": 5.1069128389104824e-05, "loss": 0.4405, "step": 9665 }, { "epoch": 1.01, "learning_rate": 5.106737609004416e-05, "loss": 1.2411, "step": 9666 }, { "epoch": 1.01, "learning_rate": 5.106562364916285e-05, "loss": 1.3032, "step": 9667 }, { "epoch": 1.01, "learning_rate": 5.106387106647269e-05, "loss": 0.6918, "step": 9668 }, { "epoch": 1.01, "learning_rate": 5.1062118341985476e-05, "loss": 0.505, "step": 9669 }, { "epoch": 1.01, "learning_rate": 5.1060365475713016e-05, "loss": 0.8974, "step": 9670 }, { "epoch": 1.01, "learning_rate": 5.1058612467667096e-05, "loss": 1.3412, "step": 9671 }, { "epoch": 1.01, "learning_rate": 5.105685931785953e-05, "loss": 0.5684, "step": 9672 }, { "epoch": 1.01, "learning_rate": 5.10551060263021e-05, "loss": 0.7035, "step": 9673 }, { "epoch": 1.01, "learning_rate": 5.1053352593006645e-05, "loss": 1.5096, "step": 9674 }, { "epoch": 1.01, "learning_rate": 5.105159901798494e-05, "loss": 1.0671, "step": 9675 }, { "epoch": 1.01, "learning_rate": 5.104984530124878e-05, "loss": 0.631, "step": 9676 }, { "epoch": 1.01, "learning_rate": 5.1048091442810006e-05, "loss": 1.1853, "step": 9677 }, { "epoch": 1.01, "learning_rate": 5.10463374426804e-05, "loss": 1.3835, "step": 9678 }, { "epoch": 1.01, "learning_rate": 5.1044583300871764e-05, "loss": 0.9704, "step": 9679 }, { "epoch": 1.01, "learning_rate": 5.1042829017395934e-05, "loss": 0.4405, "step": 9680 }, { "epoch": 1.01, "learning_rate": 5.104107459226469e-05, "loss": 0.8412, "step": 9681 }, { "epoch": 1.01, "learning_rate": 5.103932002548986e-05, "loss": 1.2438, "step": 9682 }, { "epoch": 1.01, "learning_rate": 5.1037565317083245e-05, "loss": 1.6138, "step": 9683 }, { "epoch": 1.01, "learning_rate": 5.103581046705667e-05, "loss": 1.2196, "step": 9684 }, { "epoch": 1.01, "learning_rate": 5.103405547542194e-05, "loss": 1.2892, "step": 9685 }, { "epoch": 1.01, "learning_rate": 5.1032300342190864e-05, "loss": 0.5408, "step": 9686 }, { "epoch": 1.01, "learning_rate": 5.103054506737527e-05, "loss": 0.2403, "step": 9687 }, { "epoch": 1.01, "learning_rate": 5.102878965098696e-05, "loss": 1.3387, "step": 9688 }, { "epoch": 1.01, "learning_rate": 5.102703409303777e-05, "loss": 1.5611, "step": 9689 }, { "epoch": 1.01, "learning_rate": 5.10252783935395e-05, "loss": 0.1427, "step": 9690 }, { "epoch": 1.01, "learning_rate": 5.102352255250397e-05, "loss": 0.811, "step": 9691 }, { "epoch": 1.01, "learning_rate": 5.102176656994301e-05, "loss": 1.7013, "step": 9692 }, { "epoch": 1.01, "learning_rate": 5.102001044586844e-05, "loss": 0.6433, "step": 9693 }, { "epoch": 1.01, "learning_rate": 5.101825418029208e-05, "loss": 1.306, "step": 9694 }, { "epoch": 1.01, "learning_rate": 5.1016497773225736e-05, "loss": 1.255, "step": 9695 }, { "epoch": 1.01, "learning_rate": 5.1014741224681266e-05, "loss": 0.5993, "step": 9696 }, { "epoch": 1.01, "learning_rate": 5.101298453467046e-05, "loss": 0.8919, "step": 9697 }, { "epoch": 1.01, "learning_rate": 5.1011227703205165e-05, "loss": 1.5929, "step": 9698 }, { "epoch": 1.01, "learning_rate": 5.10094707302972e-05, "loss": 1.3306, "step": 9699 }, { "epoch": 1.01, "learning_rate": 5.1007713615958394e-05, "loss": 1.0447, "step": 9700 }, { "epoch": 1.01, "learning_rate": 5.100595636020058e-05, "loss": 1.0535, "step": 9701 }, { "epoch": 1.01, "learning_rate": 5.1004198963035576e-05, "loss": 0.782, "step": 9702 }, { "epoch": 1.01, "learning_rate": 5.100244142447522e-05, "loss": 1.1141, "step": 9703 }, { "epoch": 1.01, "learning_rate": 5.100068374453134e-05, "loss": 0.6613, "step": 9704 }, { "epoch": 1.01, "learning_rate": 5.0998925923215774e-05, "loss": 0.512, "step": 9705 }, { "epoch": 1.01, "learning_rate": 5.099716796054035e-05, "loss": 1.7289, "step": 9706 }, { "epoch": 1.01, "learning_rate": 5.099540985651691e-05, "loss": 1.2956, "step": 9707 }, { "epoch": 1.01, "learning_rate": 5.099365161115729e-05, "loss": 1.2023, "step": 9708 }, { "epoch": 1.01, "learning_rate": 5.09918932244733e-05, "loss": 1.2269, "step": 9709 }, { "epoch": 1.01, "learning_rate": 5.099013469647681e-05, "loss": 2.2327, "step": 9710 }, { "epoch": 1.01, "learning_rate": 5.0988376027179636e-05, "loss": 1.4262, "step": 9711 }, { "epoch": 1.01, "learning_rate": 5.0986617216593627e-05, "loss": 0.6045, "step": 9712 }, { "epoch": 1.01, "learning_rate": 5.0984858264730626e-05, "loss": 0.2032, "step": 9713 }, { "epoch": 1.01, "learning_rate": 5.0983099171602465e-05, "loss": 0.5039, "step": 9714 }, { "epoch": 1.01, "learning_rate": 5.0981339937221e-05, "loss": 1.4681, "step": 9715 }, { "epoch": 1.01, "learning_rate": 5.097958056159805e-05, "loss": 0.6244, "step": 9716 }, { "epoch": 1.01, "learning_rate": 5.097782104474549e-05, "loss": 0.8746, "step": 9717 }, { "epoch": 1.01, "learning_rate": 5.097606138667513e-05, "loss": 0.2756, "step": 9718 }, { "epoch": 1.02, "learning_rate": 5.0974301587398836e-05, "loss": 0.5474, "step": 9719 }, { "epoch": 1.02, "learning_rate": 5.097254164692846e-05, "loss": 1.3356, "step": 9720 }, { "epoch": 1.02, "learning_rate": 5.097078156527583e-05, "loss": 1.2357, "step": 9721 }, { "epoch": 1.02, "learning_rate": 5.096902134245281e-05, "loss": 0.6861, "step": 9722 }, { "epoch": 1.02, "learning_rate": 5.0967260978471244e-05, "loss": 0.3161, "step": 9723 }, { "epoch": 1.0, "learning_rate": 5.096550047334298e-05, "loss": 0.7663, "step": 9724 }, { "epoch": 1.0, "learning_rate": 5.096373982707988e-05, "loss": 0.2108, "step": 9725 }, { "epoch": 1.0, "learning_rate": 5.096197903969379e-05, "loss": 0.5431, "step": 9726 }, { "epoch": 1.0, "learning_rate": 5.096021811119656e-05, "loss": 0.5657, "step": 9727 }, { "epoch": 1.0, "learning_rate": 5.095845704160004e-05, "loss": 0.5055, "step": 9728 }, { "epoch": 1.0, "learning_rate": 5.095669583091609e-05, "loss": 0.6659, "step": 9729 }, { "epoch": 1.0, "learning_rate": 5.095493447915657e-05, "loss": 0.3531, "step": 9730 }, { "epoch": 1.0, "learning_rate": 5.0953172986333345e-05, "loss": 0.6446, "step": 9731 }, { "epoch": 1.0, "learning_rate": 5.095141135245825e-05, "loss": 0.5036, "step": 9732 }, { "epoch": 1.0, "learning_rate": 5.094964957754315e-05, "loss": 0.6281, "step": 9733 }, { "epoch": 1.0, "learning_rate": 5.0947887661599924e-05, "loss": 0.373, "step": 9734 }, { "epoch": 1.0, "learning_rate": 5.094612560464041e-05, "loss": 0.7681, "step": 9735 }, { "epoch": 1.0, "learning_rate": 5.094436340667648e-05, "loss": 0.6953, "step": 9736 }, { "epoch": 1.0, "learning_rate": 5.094260106772001e-05, "loss": 0.4597, "step": 9737 }, { "epoch": 1.0, "learning_rate": 5.094083858778284e-05, "loss": 0.4209, "step": 9738 }, { "epoch": 1.0, "learning_rate": 5.093907596687685e-05, "loss": 0.4639, "step": 9739 }, { "epoch": 1.0, "learning_rate": 5.093731320501389e-05, "loss": 0.5519, "step": 9740 }, { "epoch": 1.0, "learning_rate": 5.0935550302205835e-05, "loss": 1.1647, "step": 9741 }, { "epoch": 1.0, "learning_rate": 5.0933787258464556e-05, "loss": 0.537, "step": 9742 }, { "epoch": 1.0, "learning_rate": 5.093202407380193e-05, "loss": 0.3901, "step": 9743 }, { "epoch": 1.0, "learning_rate": 5.0930260748229806e-05, "loss": 0.4865, "step": 9744 }, { "epoch": 1.0, "learning_rate": 5.092849728176007e-05, "loss": 0.3061, "step": 9745 }, { "epoch": 1.0, "learning_rate": 5.092673367440458e-05, "loss": 0.557, "step": 9746 }, { "epoch": 1.0, "learning_rate": 5.0924969926175226e-05, "loss": 0.2219, "step": 9747 }, { "epoch": 1.0, "learning_rate": 5.0923206037083865e-05, "loss": 0.9334, "step": 9748 }, { "epoch": 1.0, "learning_rate": 5.0921442007142377e-05, "loss": 0.9527, "step": 9749 }, { "epoch": 1.0, "learning_rate": 5.091967783636263e-05, "loss": 0.3247, "step": 9750 }, { "epoch": 1.0, "learning_rate": 5.091791352475652e-05, "loss": 0.4754, "step": 9751 }, { "epoch": 1.0, "learning_rate": 5.0916149072335895e-05, "loss": 1.1432, "step": 9752 }, { "epoch": 1.0, "learning_rate": 5.091438447911266e-05, "loss": 0.2002, "step": 9753 }, { "epoch": 1.0, "learning_rate": 5.091261974509867e-05, "loss": 0.6603, "step": 9754 }, { "epoch": 1.0, "learning_rate": 5.091085487030583e-05, "loss": 0.4935, "step": 9755 }, { "epoch": 1.0, "learning_rate": 5.0909089854746e-05, "loss": 0.5977, "step": 9756 }, { "epoch": 1.0, "learning_rate": 5.090732469843107e-05, "loss": 0.2959, "step": 9757 }, { "epoch": 1.0, "learning_rate": 5.090555940137293e-05, "loss": 1.0001, "step": 9758 }, { "epoch": 1.0, "learning_rate": 5.090379396358345e-05, "loss": 0.5844, "step": 9759 }, { "epoch": 1.0, "learning_rate": 5.090202838507451e-05, "loss": 0.3845, "step": 9760 }, { "epoch": 1.0, "learning_rate": 5.0900262665858015e-05, "loss": 0.9408, "step": 9761 }, { "epoch": 1.0, "learning_rate": 5.0898496805945855e-05, "loss": 0.3389, "step": 9762 }, { "epoch": 1.0, "learning_rate": 5.0896730805349886e-05, "loss": 0.2815, "step": 9763 }, { "epoch": 1.0, "learning_rate": 5.089496466408202e-05, "loss": 0.7027, "step": 9764 }, { "epoch": 1.0, "learning_rate": 5.0893198382154144e-05, "loss": 0.4439, "step": 9765 }, { "epoch": 1.0, "learning_rate": 5.089143195957814e-05, "loss": 0.306, "step": 9766 }, { "epoch": 1.0, "learning_rate": 5.08896653963659e-05, "loss": 0.3408, "step": 9767 }, { "epoch": 1.0, "learning_rate": 5.088789869252934e-05, "loss": 0.281, "step": 9768 }, { "epoch": 1.0, "learning_rate": 5.088613184808031e-05, "loss": 0.275, "step": 9769 }, { "epoch": 1.0, "learning_rate": 5.088436486303074e-05, "loss": 0.2322, "step": 9770 }, { "epoch": 1.01, "learning_rate": 5.0882597737392506e-05, "loss": 0.1811, "step": 9771 }, { "epoch": 1.01, "learning_rate": 5.0880830471177525e-05, "loss": 0.1406, "step": 9772 }, { "epoch": 1.01, "learning_rate": 5.087906306439766e-05, "loss": 0.5829, "step": 9773 }, { "epoch": 1.01, "learning_rate": 5.087729551706484e-05, "loss": 0.4091, "step": 9774 }, { "epoch": 1.01, "learning_rate": 5.0875527829190955e-05, "loss": 0.4816, "step": 9775 }, { "epoch": 1.01, "learning_rate": 5.087376000078789e-05, "loss": 0.3146, "step": 9776 }, { "epoch": 1.01, "learning_rate": 5.087199203186756e-05, "loss": 0.191, "step": 9777 }, { "epoch": 1.01, "learning_rate": 5.087022392244186e-05, "loss": 1.2493, "step": 9778 }, { "epoch": 1.01, "learning_rate": 5.08684556725227e-05, "loss": 0.5396, "step": 9779 }, { "epoch": 1.01, "learning_rate": 5.0866687282121985e-05, "loss": 1.0155, "step": 9780 }, { "epoch": 1.01, "learning_rate": 5.0864918751251604e-05, "loss": 0.52, "step": 9781 }, { "epoch": 1.01, "learning_rate": 5.0863150079923475e-05, "loss": 0.2388, "step": 9782 }, { "epoch": 1.01, "learning_rate": 5.08613812681495e-05, "loss": 0.8405, "step": 9783 }, { "epoch": 1.01, "learning_rate": 5.085961231594159e-05, "loss": 0.5752, "step": 9784 }, { "epoch": 1.01, "learning_rate": 5.085784322331164e-05, "loss": 0.8428, "step": 9785 }, { "epoch": 1.01, "learning_rate": 5.085607399027159e-05, "loss": 0.9399, "step": 9786 }, { "epoch": 1.01, "learning_rate": 5.085430461683332e-05, "loss": 1.1409, "step": 9787 }, { "epoch": 1.01, "learning_rate": 5.085253510300875e-05, "loss": 0.6929, "step": 9788 }, { "epoch": 1.01, "learning_rate": 5.0850765448809794e-05, "loss": 1.1604, "step": 9789 }, { "epoch": 1.01, "learning_rate": 5.0848995654248364e-05, "loss": 0.7991, "step": 9790 }, { "epoch": 1.01, "learning_rate": 5.084722571933637e-05, "loss": 0.4843, "step": 9791 }, { "epoch": 1.01, "learning_rate": 5.084545564408573e-05, "loss": 0.8401, "step": 9792 }, { "epoch": 1.01, "learning_rate": 5.0843685428508365e-05, "loss": 0.449, "step": 9793 }, { "epoch": 1.01, "learning_rate": 5.084191507261619e-05, "loss": 0.7046, "step": 9794 }, { "epoch": 1.01, "learning_rate": 5.084014457642111e-05, "loss": 0.747, "step": 9795 }, { "epoch": 1.01, "learning_rate": 5.083837393993506e-05, "loss": 0.8046, "step": 9796 }, { "epoch": 1.01, "learning_rate": 5.083660316316994e-05, "loss": 1.3578, "step": 9797 }, { "epoch": 1.01, "learning_rate": 5.0834832246137696e-05, "loss": 0.6921, "step": 9798 }, { "epoch": 1.01, "learning_rate": 5.083306118885023e-05, "loss": 0.1883, "step": 9799 }, { "epoch": 1.01, "learning_rate": 5.083128999131948e-05, "loss": 0.3021, "step": 9800 }, { "epoch": 1.01, "learning_rate": 5.082951865355735e-05, "loss": 0.6613, "step": 9801 }, { "epoch": 1.01, "learning_rate": 5.082774717557578e-05, "loss": 0.6478, "step": 9802 }, { "epoch": 1.01, "learning_rate": 5.082597555738668e-05, "loss": 0.6161, "step": 9803 }, { "epoch": 1.01, "learning_rate": 5.082420379900199e-05, "loss": 0.4458, "step": 9804 }, { "epoch": 1.01, "learning_rate": 5.082243190043363e-05, "loss": 0.5727, "step": 9805 }, { "epoch": 1.01, "learning_rate": 5.0820659861693536e-05, "loss": 1.058, "step": 9806 }, { "epoch": 1.01, "learning_rate": 5.081888768279363e-05, "loss": 0.3167, "step": 9807 }, { "epoch": 1.01, "learning_rate": 5.081711536374585e-05, "loss": 0.7888, "step": 9808 }, { "epoch": 1.01, "learning_rate": 5.081534290456211e-05, "loss": 0.3917, "step": 9809 }, { "epoch": 1.01, "learning_rate": 5.081357030525435e-05, "loss": 0.5624, "step": 9810 }, { "epoch": 1.01, "learning_rate": 5.0811797565834514e-05, "loss": 0.2556, "step": 9811 }, { "epoch": 1.01, "learning_rate": 5.0810024686314526e-05, "loss": 0.4831, "step": 9812 }, { "epoch": 1.01, "learning_rate": 5.080825166670631e-05, "loss": 0.2267, "step": 9813 }, { "epoch": 1.01, "learning_rate": 5.080647850702183e-05, "loss": 0.3107, "step": 9814 }, { "epoch": 1.01, "learning_rate": 5.080470520727299e-05, "loss": 1.6469, "step": 9815 }, { "epoch": 1.01, "learning_rate": 5.0802931767471745e-05, "loss": 0.957, "step": 9816 }, { "epoch": 1.01, "learning_rate": 5.080115818763004e-05, "loss": 0.9495, "step": 9817 }, { "epoch": 1.01, "learning_rate": 5.07993844677598e-05, "loss": 0.4689, "step": 9818 }, { "epoch": 1.01, "learning_rate": 5.079761060787297e-05, "loss": 0.5929, "step": 9819 }, { "epoch": 1.01, "learning_rate": 5.079583660798149e-05, "loss": 0.6695, "step": 9820 }, { "epoch": 1.01, "learning_rate": 5.07940624680973e-05, "loss": 0.1995, "step": 9821 }, { "epoch": 1.01, "learning_rate": 5.0792288188232356e-05, "loss": 0.7712, "step": 9822 }, { "epoch": 1.01, "learning_rate": 5.0790513768398586e-05, "loss": 0.73, "step": 9823 }, { "epoch": 1.01, "learning_rate": 5.078873920860794e-05, "loss": 0.8904, "step": 9824 }, { "epoch": 1.01, "learning_rate": 5.078696450887237e-05, "loss": 1.1168, "step": 9825 }, { "epoch": 1.01, "learning_rate": 5.078518966920382e-05, "loss": 0.594, "step": 9826 }, { "epoch": 1.01, "learning_rate": 5.0783414689614236e-05, "loss": 0.609, "step": 9827 }, { "epoch": 1.01, "learning_rate": 5.0781639570115564e-05, "loss": 0.312, "step": 9828 }, { "epoch": 1.01, "learning_rate": 5.0779864310719766e-05, "loss": 0.78, "step": 9829 }, { "epoch": 1.01, "learning_rate": 5.077808891143877e-05, "loss": 0.4351, "step": 9830 }, { "epoch": 1.01, "learning_rate": 5.077631337228455e-05, "loss": 0.8724, "step": 9831 }, { "epoch": 1.01, "learning_rate": 5.0774537693269045e-05, "loss": 0.5777, "step": 9832 }, { "epoch": 1.01, "learning_rate": 5.0772761874404223e-05, "loss": 0.379, "step": 9833 }, { "epoch": 1.01, "learning_rate": 5.0770985915702015e-05, "loss": 0.3928, "step": 9834 }, { "epoch": 1.01, "learning_rate": 5.07692098171744e-05, "loss": 0.6148, "step": 9835 }, { "epoch": 1.01, "learning_rate": 5.0767433578833315e-05, "loss": 0.7568, "step": 9836 }, { "epoch": 1.01, "learning_rate": 5.076565720069074e-05, "loss": 0.5438, "step": 9837 }, { "epoch": 1.01, "learning_rate": 5.0763880682758614e-05, "loss": 0.8437, "step": 9838 }, { "epoch": 1.01, "learning_rate": 5.0762104025048896e-05, "loss": 0.2293, "step": 9839 }, { "epoch": 1.01, "learning_rate": 5.076032722757355e-05, "loss": 1.0339, "step": 9840 }, { "epoch": 1.01, "learning_rate": 5.0758550290344546e-05, "loss": 0.4716, "step": 9841 }, { "epoch": 1.01, "learning_rate": 5.075677321337383e-05, "loss": 0.2978, "step": 9842 }, { "epoch": 1.01, "learning_rate": 5.075499599667338e-05, "loss": 0.3697, "step": 9843 }, { "epoch": 1.01, "learning_rate": 5.075321864025515e-05, "loss": 0.7638, "step": 9844 }, { "epoch": 1.01, "learning_rate": 5.075144114413112e-05, "loss": 0.2533, "step": 9845 }, { "epoch": 1.01, "learning_rate": 5.0749663508313225e-05, "loss": 0.8906, "step": 9846 }, { "epoch": 1.01, "learning_rate": 5.074788573281346e-05, "loss": 0.5389, "step": 9847 }, { "epoch": 1.01, "learning_rate": 5.074610781764378e-05, "loss": 0.4265, "step": 9848 }, { "epoch": 1.01, "learning_rate": 5.074432976281616e-05, "loss": 0.9473, "step": 9849 }, { "epoch": 1.01, "learning_rate": 5.074255156834257e-05, "loss": 0.4963, "step": 9850 }, { "epoch": 1.01, "learning_rate": 5.074077323423497e-05, "loss": 0.3673, "step": 9851 }, { "epoch": 1.01, "learning_rate": 5.073899476050534e-05, "loss": 0.8807, "step": 9852 }, { "epoch": 1.01, "learning_rate": 5.073721614716565e-05, "loss": 0.6205, "step": 9853 }, { "epoch": 1.01, "learning_rate": 5.073543739422787e-05, "loss": 0.3221, "step": 9854 }, { "epoch": 1.01, "learning_rate": 5.073365850170397e-05, "loss": 0.4245, "step": 9855 }, { "epoch": 1.01, "learning_rate": 5.0731879469605946e-05, "loss": 0.9043, "step": 9856 }, { "epoch": 1.01, "learning_rate": 5.0730100297945754e-05, "loss": 0.1462, "step": 9857 }, { "epoch": 1.01, "learning_rate": 5.0728320986735374e-05, "loss": 0.9946, "step": 9858 }, { "epoch": 1.01, "learning_rate": 5.072654153598679e-05, "loss": 0.4555, "step": 9859 }, { "epoch": 1.01, "learning_rate": 5.072476194571198e-05, "loss": 0.2302, "step": 9860 }, { "epoch": 1.01, "learning_rate": 5.072298221592292e-05, "loss": 0.7245, "step": 9861 }, { "epoch": 1.01, "learning_rate": 5.0721202346631595e-05, "loss": 0.9194, "step": 9862 }, { "epoch": 1.01, "learning_rate": 5.071942233784998e-05, "loss": 0.7427, "step": 9863 }, { "epoch": 1.01, "learning_rate": 5.071764218959007e-05, "loss": 0.2688, "step": 9864 }, { "epoch": 1.01, "learning_rate": 5.071586190186383e-05, "loss": 0.9234, "step": 9865 }, { "epoch": 1.01, "learning_rate": 5.071408147468326e-05, "loss": 0.3241, "step": 9866 }, { "epoch": 1.02, "learning_rate": 5.0712300908060335e-05, "loss": 0.3108, "step": 9867 }, { "epoch": 1.02, "learning_rate": 5.0710520202007054e-05, "loss": 0.8721, "step": 9868 }, { "epoch": 1.02, "learning_rate": 5.0708739356535396e-05, "loss": 0.5262, "step": 9869 }, { "epoch": 1.02, "learning_rate": 5.0706958371657355e-05, "loss": 0.6592, "step": 9870 }, { "epoch": 1.02, "learning_rate": 5.0705177247384904e-05, "loss": 0.5024, "step": 9871 }, { "epoch": 1.02, "learning_rate": 5.0703395983730044e-05, "loss": 0.3694, "step": 9872 }, { "epoch": 1.02, "learning_rate": 5.0701614580704776e-05, "loss": 1.1477, "step": 9873 }, { "epoch": 1.02, "learning_rate": 5.069983303832108e-05, "loss": 0.3215, "step": 9874 }, { "epoch": 1.02, "learning_rate": 5.0698051356590945e-05, "loss": 0.4461, "step": 9875 }, { "epoch": 1.02, "learning_rate": 5.0696269535526375e-05, "loss": 0.4644, "step": 9876 }, { "epoch": 1.02, "learning_rate": 5.0694487575139354e-05, "loss": 0.8434, "step": 9877 }, { "epoch": 1.02, "learning_rate": 5.0692705475441894e-05, "loss": 0.4178, "step": 9878 }, { "epoch": 1.02, "learning_rate": 5.069092323644599e-05, "loss": 0.4258, "step": 9879 }, { "epoch": 1.02, "learning_rate": 5.0689140858163616e-05, "loss": 0.1975, "step": 9880 }, { "epoch": 1.02, "learning_rate": 5.06873583406068e-05, "loss": 0.4521, "step": 9881 }, { "epoch": 1.02, "learning_rate": 5.068557568378751e-05, "loss": 0.489, "step": 9882 }, { "epoch": 1.02, "learning_rate": 5.0683792887717785e-05, "loss": 0.2218, "step": 9883 }, { "epoch": 1.02, "learning_rate": 5.068200995240959e-05, "loss": 1.0921, "step": 9884 }, { "epoch": 1.02, "learning_rate": 5.0680226877874956e-05, "loss": 0.3611, "step": 9885 }, { "epoch": 1.02, "learning_rate": 5.067844366412587e-05, "loss": 0.3071, "step": 9886 }, { "epoch": 1.02, "learning_rate": 5.067666031117433e-05, "loss": 1.3178, "step": 9887 }, { "epoch": 1.02, "learning_rate": 5.0674876819032355e-05, "loss": 0.9497, "step": 9888 }, { "epoch": 1.02, "learning_rate": 5.0673093187711956e-05, "loss": 0.4715, "step": 9889 }, { "epoch": 1.02, "learning_rate": 5.0671309417225115e-05, "loss": 0.6479, "step": 9890 }, { "epoch": 1.02, "learning_rate": 5.066952550758387e-05, "loss": 0.7825, "step": 9891 }, { "epoch": 1.02, "learning_rate": 5.066774145880021e-05, "loss": 0.1598, "step": 9892 }, { "epoch": 1.02, "learning_rate": 5.066595727088615e-05, "loss": 0.0919, "step": 9893 }, { "epoch": 1.02, "learning_rate": 5.066417294385371e-05, "loss": 1.0645, "step": 9894 }, { "epoch": 1.02, "learning_rate": 5.066238847771488e-05, "loss": 0.3342, "step": 9895 }, { "epoch": 1.02, "learning_rate": 5.066060387248169e-05, "loss": 0.447, "step": 9896 }, { "epoch": 1.02, "learning_rate": 5.065881912816615e-05, "loss": 0.1846, "step": 9897 }, { "epoch": 1.02, "learning_rate": 5.065703424478027e-05, "loss": 0.3099, "step": 9898 }, { "epoch": 1.02, "learning_rate": 5.0655249222336076e-05, "loss": 0.1696, "step": 9899 }, { "epoch": 1.02, "learning_rate": 5.065346406084557e-05, "loss": 0.8132, "step": 9900 }, { "epoch": 1.02, "learning_rate": 5.0651678760320776e-05, "loss": 0.9429, "step": 9901 }, { "epoch": 1.02, "learning_rate": 5.0649893320773713e-05, "loss": 0.827, "step": 9902 }, { "epoch": 1.02, "learning_rate": 5.064810774221641e-05, "loss": 0.2548, "step": 9903 }, { "epoch": 1.02, "learning_rate": 5.064632202466086e-05, "loss": 0.4303, "step": 9904 }, { "epoch": 1.02, "learning_rate": 5.064453616811912e-05, "loss": 0.1244, "step": 9905 }, { "epoch": 1.02, "learning_rate": 5.064275017260318e-05, "loss": 0.3511, "step": 9906 }, { "epoch": 1.02, "learning_rate": 5.064096403812508e-05, "loss": 0.485, "step": 9907 }, { "epoch": 1.02, "learning_rate": 5.0639177764696844e-05, "loss": 0.5775, "step": 9908 }, { "epoch": 1.02, "learning_rate": 5.0637391352330476e-05, "loss": 0.6122, "step": 9909 }, { "epoch": 1.02, "learning_rate": 5.063560480103804e-05, "loss": 0.244, "step": 9910 }, { "epoch": 1.02, "learning_rate": 5.063381811083152e-05, "loss": 0.7412, "step": 9911 }, { "epoch": 1.02, "learning_rate": 5.063203128172298e-05, "loss": 0.6345, "step": 9912 }, { "epoch": 1.02, "learning_rate": 5.063024431372443e-05, "loss": 0.3757, "step": 9913 }, { "epoch": 1.02, "learning_rate": 5.062845720684789e-05, "loss": 0.9049, "step": 9914 }, { "epoch": 1.02, "learning_rate": 5.062666996110542e-05, "loss": 0.8794, "step": 9915 }, { "epoch": 1.02, "learning_rate": 5.062488257650903e-05, "loss": 0.2151, "step": 9916 }, { "epoch": 1.02, "learning_rate": 5.062309505307075e-05, "loss": 0.3714, "step": 9917 }, { "epoch": 1.02, "learning_rate": 5.062130739080262e-05, "loss": 0.3326, "step": 9918 }, { "epoch": 1.02, "learning_rate": 5.061951958971668e-05, "loss": 1.0478, "step": 9919 }, { "epoch": 1.02, "learning_rate": 5.0617731649824954e-05, "loss": 0.853, "step": 9920 }, { "epoch": 1.02, "learning_rate": 5.0615943571139485e-05, "loss": 0.9197, "step": 9921 }, { "epoch": 1.02, "learning_rate": 5.061415535367231e-05, "loss": 0.8816, "step": 9922 }, { "epoch": 1.02, "learning_rate": 5.061236699743545e-05, "loss": 0.8996, "step": 9923 }, { "epoch": 1.02, "learning_rate": 5.061057850244098e-05, "loss": 0.4632, "step": 9924 }, { "epoch": 1.02, "learning_rate": 5.060878986870091e-05, "loss": 0.6522, "step": 9925 }, { "epoch": 1.02, "learning_rate": 5.060700109622728e-05, "loss": 0.444, "step": 9926 }, { "epoch": 1.02, "learning_rate": 5.0605212185032146e-05, "loss": 1.089, "step": 9927 }, { "epoch": 1.02, "learning_rate": 5.060342313512754e-05, "loss": 0.927, "step": 9928 }, { "epoch": 1.02, "learning_rate": 5.0601633946525516e-05, "loss": 1.2402, "step": 9929 }, { "epoch": 1.02, "learning_rate": 5.059984461923812e-05, "loss": 1.0235, "step": 9930 }, { "epoch": 1.02, "learning_rate": 5.059805515327738e-05, "loss": 0.2625, "step": 9931 }, { "epoch": 1.02, "learning_rate": 5.059626554865536e-05, "loss": 0.2802, "step": 9932 }, { "epoch": 1.02, "learning_rate": 5.059447580538408e-05, "loss": 0.4309, "step": 9933 }, { "epoch": 1.02, "learning_rate": 5.059268592347563e-05, "loss": 0.1226, "step": 9934 }, { "epoch": 1.02, "learning_rate": 5.0590895902942024e-05, "loss": 0.3301, "step": 9935 }, { "epoch": 1.02, "learning_rate": 5.0589105743795325e-05, "loss": 0.8502, "step": 9936 }, { "epoch": 1.02, "learning_rate": 5.058731544604759e-05, "loss": 0.8253, "step": 9937 }, { "epoch": 1.02, "learning_rate": 5.058552500971087e-05, "loss": 0.3674, "step": 9938 }, { "epoch": 1.02, "learning_rate": 5.05837344347972e-05, "loss": 0.7725, "step": 9939 }, { "epoch": 1.02, "learning_rate": 5.0581943721318643e-05, "loss": 0.5868, "step": 9940 }, { "epoch": 1.02, "learning_rate": 5.058015286928727e-05, "loss": 0.8508, "step": 9941 }, { "epoch": 1.02, "learning_rate": 5.0578361878715114e-05, "loss": 0.6719, "step": 9942 }, { "epoch": 1.02, "learning_rate": 5.057657074961424e-05, "loss": 0.8408, "step": 9943 }, { "epoch": 1.02, "learning_rate": 5.057477948199672e-05, "loss": 0.2419, "step": 9944 }, { "epoch": 1.02, "learning_rate": 5.057298807587459e-05, "loss": 0.8666, "step": 9945 }, { "epoch": 1.02, "learning_rate": 5.0571196531259915e-05, "loss": 1.9222, "step": 9946 }, { "epoch": 1.02, "learning_rate": 5.0569404848164754e-05, "loss": 0.4015, "step": 9947 }, { "epoch": 1.02, "learning_rate": 5.0567613026601174e-05, "loss": 0.5652, "step": 9948 }, { "epoch": 1.02, "learning_rate": 5.056582106658124e-05, "loss": 0.5125, "step": 9949 }, { "epoch": 1.02, "learning_rate": 5.0564028968117014e-05, "loss": 0.5089, "step": 9950 }, { "epoch": 1.02, "learning_rate": 5.056223673122055e-05, "loss": 0.2683, "step": 9951 }, { "epoch": 1.02, "learning_rate": 5.056044435590393e-05, "loss": 0.9348, "step": 9952 }, { "epoch": 1.02, "learning_rate": 5.05586518421792e-05, "loss": 0.322, "step": 9953 }, { "epoch": 1.02, "learning_rate": 5.055685919005844e-05, "loss": 1.1796, "step": 9954 }, { "epoch": 1.02, "learning_rate": 5.05550663995537e-05, "loss": 1.0203, "step": 9955 }, { "epoch": 1.02, "learning_rate": 5.0553273470677075e-05, "loss": 0.503, "step": 9956 }, { "epoch": 1.02, "learning_rate": 5.0551480403440625e-05, "loss": 0.997, "step": 9957 }, { "epoch": 1.02, "learning_rate": 5.054968719785641e-05, "loss": 0.3145, "step": 9958 }, { "epoch": 1.02, "learning_rate": 5.0547893853936515e-05, "loss": 0.4691, "step": 9959 }, { "epoch": 1.02, "learning_rate": 5.0546100371693006e-05, "loss": 0.1323, "step": 9960 }, { "epoch": 1.02, "learning_rate": 5.0544306751137955e-05, "loss": 0.5668, "step": 9961 }, { "epoch": 1.02, "learning_rate": 5.054251299228344e-05, "loss": 0.299, "step": 9962 }, { "epoch": 1.03, "learning_rate": 5.054071909514153e-05, "loss": 0.4366, "step": 9963 }, { "epoch": 1.03, "learning_rate": 5.053892505972431e-05, "loss": 0.7203, "step": 9964 }, { "epoch": 1.03, "learning_rate": 5.0537130886043846e-05, "loss": 0.2279, "step": 9965 }, { "epoch": 1.03, "learning_rate": 5.0535336574112225e-05, "loss": 0.9065, "step": 9966 }, { "epoch": 1.03, "learning_rate": 5.053354212394152e-05, "loss": 1.1457, "step": 9967 }, { "epoch": 1.03, "learning_rate": 5.053174753554382e-05, "loss": 0.2685, "step": 9968 }, { "epoch": 1.03, "learning_rate": 5.0529952808931204e-05, "loss": 0.478, "step": 9969 }, { "epoch": 1.03, "learning_rate": 5.052815794411574e-05, "loss": 0.4791, "step": 9970 }, { "epoch": 1.03, "learning_rate": 5.0526362941109516e-05, "loss": 0.7451, "step": 9971 }, { "epoch": 1.03, "learning_rate": 5.052456779992463e-05, "loss": 0.334, "step": 9972 }, { "epoch": 1.03, "learning_rate": 5.0522772520573155e-05, "loss": 0.4126, "step": 9973 }, { "epoch": 1.03, "learning_rate": 5.052097710306717e-05, "loss": 0.521, "step": 9974 }, { "epoch": 1.03, "learning_rate": 5.0519181547418776e-05, "loss": 0.6263, "step": 9975 }, { "epoch": 1.03, "learning_rate": 5.051738585364005e-05, "loss": 0.1798, "step": 9976 }, { "epoch": 1.03, "learning_rate": 5.0515590021743086e-05, "loss": 0.2278, "step": 9977 }, { "epoch": 1.03, "learning_rate": 5.051379405173996e-05, "loss": 0.4639, "step": 9978 }, { "epoch": 1.03, "learning_rate": 5.051199794364278e-05, "loss": 0.4233, "step": 9979 }, { "epoch": 1.03, "learning_rate": 5.051020169746363e-05, "loss": 0.3831, "step": 9980 }, { "epoch": 1.03, "learning_rate": 5.0508405313214596e-05, "loss": 0.5071, "step": 9981 }, { "epoch": 1.03, "learning_rate": 5.050660879090779e-05, "loss": 0.3694, "step": 9982 }, { "epoch": 1.03, "learning_rate": 5.050481213055528e-05, "loss": 1.0409, "step": 9983 }, { "epoch": 1.03, "learning_rate": 5.0503015332169176e-05, "loss": 0.466, "step": 9984 }, { "epoch": 1.03, "learning_rate": 5.050121839576156e-05, "loss": 0.3876, "step": 9985 }, { "epoch": 1.03, "learning_rate": 5.0499421321344544e-05, "loss": 0.4508, "step": 9986 }, { "epoch": 1.03, "learning_rate": 5.0497624108930234e-05, "loss": 0.5515, "step": 9987 }, { "epoch": 1.03, "learning_rate": 5.04958267585307e-05, "loss": 1.1274, "step": 9988 }, { "epoch": 1.03, "learning_rate": 5.049402927015806e-05, "loss": 1.3047, "step": 9989 }, { "epoch": 1.03, "learning_rate": 5.049223164382441e-05, "loss": 0.4095, "step": 9990 }, { "epoch": 1.03, "learning_rate": 5.049043387954185e-05, "loss": 0.5972, "step": 9991 }, { "epoch": 1.03, "learning_rate": 5.048863597732248e-05, "loss": 0.5261, "step": 9992 }, { "epoch": 1.03, "learning_rate": 5.0486837937178416e-05, "loss": 0.431, "step": 9993 }, { "epoch": 1.03, "learning_rate": 5.0485039759121744e-05, "loss": 0.9189, "step": 9994 }, { "epoch": 1.03, "learning_rate": 5.0483241443164586e-05, "loss": 1.0757, "step": 9995 }, { "epoch": 1.03, "learning_rate": 5.048144298931903e-05, "loss": 0.2819, "step": 9996 }, { "epoch": 1.03, "learning_rate": 5.0479644397597195e-05, "loss": 0.8849, "step": 9997 }, { "epoch": 1.03, "learning_rate": 5.0477845668011195e-05, "loss": 0.7774, "step": 9998 }, { "epoch": 1.03, "learning_rate": 5.047604680057312e-05, "loss": 0.92, "step": 9999 }, { "epoch": 1.03, "learning_rate": 5.0474247795295084e-05, "loss": 0.7151, "step": 10000 }, { "epoch": 1.03, "learning_rate": 5.047244865218921e-05, "loss": 0.6591, "step": 10001 }, { "epoch": 1.03, "learning_rate": 5.0470649371267604e-05, "loss": 0.4237, "step": 10002 }, { "epoch": 1.03, "learning_rate": 5.0468849952542374e-05, "loss": 0.3045, "step": 10003 }, { "epoch": 1.03, "learning_rate": 5.046705039602563e-05, "loss": 0.4323, "step": 10004 }, { "epoch": 1.03, "learning_rate": 5.04652507017295e-05, "loss": 0.3803, "step": 10005 }, { "epoch": 1.03, "learning_rate": 5.0463450869666087e-05, "loss": 0.8336, "step": 10006 }, { "epoch": 1.03, "learning_rate": 5.0461650899847506e-05, "loss": 0.6114, "step": 10007 }, { "epoch": 1.03, "learning_rate": 5.045985079228589e-05, "loss": 0.4841, "step": 10008 }, { "epoch": 1.03, "learning_rate": 5.045805054699334e-05, "loss": 0.4828, "step": 10009 }, { "epoch": 1.03, "learning_rate": 5.045625016398198e-05, "loss": 0.4492, "step": 10010 }, { "epoch": 1.03, "learning_rate": 5.045444964326393e-05, "loss": 0.3188, "step": 10011 }, { "epoch": 1.03, "learning_rate": 5.04526489848513e-05, "loss": 0.9093, "step": 10012 }, { "epoch": 1.03, "learning_rate": 5.0450848188756243e-05, "loss": 0.4812, "step": 10013 }, { "epoch": 1.03, "learning_rate": 5.0449047254990845e-05, "loss": 0.7405, "step": 10014 }, { "epoch": 1.03, "learning_rate": 5.044724618356726e-05, "loss": 0.2908, "step": 10015 }, { "epoch": 1.03, "learning_rate": 5.044544497449759e-05, "loss": 0.5086, "step": 10016 }, { "epoch": 1.03, "learning_rate": 5.044364362779396e-05, "loss": 1.2119, "step": 10017 }, { "epoch": 1.03, "learning_rate": 5.044184214346851e-05, "loss": 1.8045, "step": 10018 }, { "epoch": 1.03, "learning_rate": 5.044004052153337e-05, "loss": 0.3926, "step": 10019 }, { "epoch": 1.03, "learning_rate": 5.0438238762000656e-05, "loss": 0.879, "step": 10020 }, { "epoch": 1.03, "learning_rate": 5.0436436864882495e-05, "loss": 1.0983, "step": 10021 }, { "epoch": 1.03, "learning_rate": 5.043463483019103e-05, "loss": 0.1922, "step": 10022 }, { "epoch": 1.03, "learning_rate": 5.043283265793839e-05, "loss": 0.8359, "step": 10023 }, { "epoch": 1.03, "learning_rate": 5.0431030348136695e-05, "loss": 0.4823, "step": 10024 }, { "epoch": 1.03, "learning_rate": 5.042922790079807e-05, "loss": 0.48, "step": 10025 }, { "epoch": 1.03, "learning_rate": 5.042742531593468e-05, "loss": 0.5021, "step": 10026 }, { "epoch": 1.03, "learning_rate": 5.042562259355864e-05, "loss": 0.2364, "step": 10027 }, { "epoch": 1.03, "learning_rate": 5.0423819733682085e-05, "loss": 0.3578, "step": 10028 }, { "epoch": 1.03, "learning_rate": 5.042201673631716e-05, "loss": 0.1633, "step": 10029 }, { "epoch": 1.03, "learning_rate": 5.0420213601475986e-05, "loss": 0.2114, "step": 10030 }, { "epoch": 1.03, "learning_rate": 5.041841032917071e-05, "loss": 1.5635, "step": 10031 }, { "epoch": 1.03, "learning_rate": 5.041660691941349e-05, "loss": 0.743, "step": 10032 }, { "epoch": 1.03, "learning_rate": 5.041480337221644e-05, "loss": 0.9365, "step": 10033 }, { "epoch": 1.03, "learning_rate": 5.041299968759171e-05, "loss": 0.2424, "step": 10034 }, { "epoch": 1.03, "learning_rate": 5.041119586555144e-05, "loss": 0.5946, "step": 10035 }, { "epoch": 1.03, "learning_rate": 5.040939190610778e-05, "loss": 0.6009, "step": 10036 }, { "epoch": 1.03, "learning_rate": 5.040758780927287e-05, "loss": 0.5792, "step": 10037 }, { "epoch": 1.03, "learning_rate": 5.040578357505885e-05, "loss": 0.5742, "step": 10038 }, { "epoch": 1.03, "learning_rate": 5.0403979203477876e-05, "loss": 0.5904, "step": 10039 }, { "epoch": 1.03, "learning_rate": 5.040217469454208e-05, "loss": 0.7729, "step": 10040 }, { "epoch": 1.03, "learning_rate": 5.0400370048263616e-05, "loss": 1.1375, "step": 10041 }, { "epoch": 1.03, "learning_rate": 5.039856526465464e-05, "loss": 0.5972, "step": 10042 }, { "epoch": 1.03, "learning_rate": 5.0396760343727294e-05, "loss": 0.785, "step": 10043 }, { "epoch": 1.03, "learning_rate": 5.039495528549373e-05, "loss": 0.6106, "step": 10044 }, { "epoch": 1.03, "learning_rate": 5.0393150089966094e-05, "loss": 0.241, "step": 10045 }, { "epoch": 1.03, "learning_rate": 5.039134475715656e-05, "loss": 0.6262, "step": 10046 }, { "epoch": 1.03, "learning_rate": 5.038953928707724e-05, "loss": 0.7838, "step": 10047 }, { "epoch": 1.03, "learning_rate": 5.038773367974033e-05, "loss": 1.3245, "step": 10048 }, { "epoch": 1.03, "learning_rate": 5.038592793515796e-05, "loss": 0.1028, "step": 10049 }, { "epoch": 1.03, "learning_rate": 5.03841220533423e-05, "loss": 0.4912, "step": 10050 }, { "epoch": 1.03, "learning_rate": 5.0382316034305496e-05, "loss": 0.1262, "step": 10051 }, { "epoch": 1.03, "learning_rate": 5.0380509878059704e-05, "loss": 0.3146, "step": 10052 }, { "epoch": 1.03, "learning_rate": 5.03787035846171e-05, "loss": 1.4902, "step": 10053 }, { "epoch": 1.03, "learning_rate": 5.037689715398981e-05, "loss": 0.3877, "step": 10054 }, { "epoch": 1.03, "learning_rate": 5.037509058619003e-05, "loss": 0.9561, "step": 10055 }, { "epoch": 1.03, "learning_rate": 5.037328388122991e-05, "loss": 0.1865, "step": 10056 }, { "epoch": 1.03, "learning_rate": 5.0371477039121613e-05, "loss": 0.766, "step": 10057 }, { "epoch": 1.03, "learning_rate": 5.036967005987729e-05, "loss": 0.425, "step": 10058 }, { "epoch": 1.04, "learning_rate": 5.036786294350912e-05, "loss": 0.3067, "step": 10059 }, { "epoch": 1.04, "learning_rate": 5.036605569002927e-05, "loss": 1.1027, "step": 10060 }, { "epoch": 1.04, "learning_rate": 5.036424829944988e-05, "loss": 0.3852, "step": 10061 }, { "epoch": 1.04, "learning_rate": 5.036244077178315e-05, "loss": 0.8482, "step": 10062 }, { "epoch": 1.04, "learning_rate": 5.036063310704122e-05, "loss": 0.9241, "step": 10063 }, { "epoch": 1.04, "learning_rate": 5.035882530523628e-05, "loss": 0.7014, "step": 10064 }, { "epoch": 1.04, "learning_rate": 5.0357017366380495e-05, "loss": 0.481, "step": 10065 }, { "epoch": 1.04, "learning_rate": 5.0355209290486026e-05, "loss": 1.0572, "step": 10066 }, { "epoch": 1.04, "learning_rate": 5.035340107756506e-05, "loss": 0.4083, "step": 10067 }, { "epoch": 1.04, "learning_rate": 5.035159272762974e-05, "loss": 0.4138, "step": 10068 }, { "epoch": 1.04, "learning_rate": 5.034978424069228e-05, "loss": 0.5927, "step": 10069 }, { "epoch": 1.04, "learning_rate": 5.0347975616764836e-05, "loss": 0.3804, "step": 10070 }, { "epoch": 1.04, "learning_rate": 5.034616685585957e-05, "loss": 0.5796, "step": 10071 }, { "epoch": 1.04, "learning_rate": 5.034435795798866e-05, "loss": 0.9944, "step": 10072 }, { "epoch": 1.04, "learning_rate": 5.034254892316431e-05, "loss": 0.4208, "step": 10073 }, { "epoch": 1.04, "learning_rate": 5.034073975139867e-05, "loss": 0.3996, "step": 10074 }, { "epoch": 1.04, "learning_rate": 5.0338930442703934e-05, "loss": 1.0239, "step": 10075 }, { "epoch": 1.04, "learning_rate": 5.0337120997092285e-05, "loss": 0.4734, "step": 10076 }, { "epoch": 1.04, "learning_rate": 5.033531141457588e-05, "loss": 0.1702, "step": 10077 }, { "epoch": 1.04, "learning_rate": 5.0333501695166924e-05, "loss": 1.0602, "step": 10078 }, { "epoch": 1.04, "learning_rate": 5.033169183887759e-05, "loss": 1.1617, "step": 10079 }, { "epoch": 1.04, "learning_rate": 5.032988184572007e-05, "loss": 0.3312, "step": 10080 }, { "epoch": 1.04, "learning_rate": 5.032807171570654e-05, "loss": 0.2213, "step": 10081 }, { "epoch": 1.04, "learning_rate": 5.032626144884919e-05, "loss": 0.9137, "step": 10082 }, { "epoch": 1.04, "learning_rate": 5.0324451045160193e-05, "loss": 0.2901, "step": 10083 }, { "epoch": 1.04, "learning_rate": 5.0322640504651755e-05, "loss": 0.0829, "step": 10084 }, { "epoch": 1.04, "learning_rate": 5.0320829827336064e-05, "loss": 0.9233, "step": 10085 }, { "epoch": 1.04, "learning_rate": 5.031901901322529e-05, "loss": 0.2928, "step": 10086 }, { "epoch": 1.04, "learning_rate": 5.031720806233164e-05, "loss": 0.296, "step": 10087 }, { "epoch": 1.04, "learning_rate": 5.031539697466729e-05, "loss": 0.8385, "step": 10088 }, { "epoch": 1.04, "learning_rate": 5.031358575024445e-05, "loss": 0.3716, "step": 10089 }, { "epoch": 1.04, "learning_rate": 5.03117743890753e-05, "loss": 0.7123, "step": 10090 }, { "epoch": 1.04, "learning_rate": 5.030996289117204e-05, "loss": 0.6794, "step": 10091 }, { "epoch": 1.04, "learning_rate": 5.030815125654686e-05, "loss": 0.2635, "step": 10092 }, { "epoch": 1.04, "learning_rate": 5.030633948521196e-05, "loss": 0.2109, "step": 10093 }, { "epoch": 1.04, "learning_rate": 5.0304527577179526e-05, "loss": 0.2959, "step": 10094 }, { "epoch": 1.04, "learning_rate": 5.030271553246177e-05, "loss": 0.9968, "step": 10095 }, { "epoch": 1.04, "learning_rate": 5.0300903351070884e-05, "loss": 0.4388, "step": 10096 }, { "epoch": 1.04, "learning_rate": 5.029909103301907e-05, "loss": 1.8454, "step": 10097 }, { "epoch": 1.04, "learning_rate": 5.029727857831852e-05, "loss": 0.2205, "step": 10098 }, { "epoch": 1.04, "learning_rate": 5.0295465986981436e-05, "loss": 1.2185, "step": 10099 }, { "epoch": 1.04, "learning_rate": 5.029365325902002e-05, "loss": 0.4585, "step": 10100 }, { "epoch": 1.04, "learning_rate": 5.02918403944465e-05, "loss": 0.5955, "step": 10101 }, { "epoch": 1.04, "learning_rate": 5.029002739327304e-05, "loss": 0.4439, "step": 10102 }, { "epoch": 1.04, "learning_rate": 5.028821425551187e-05, "loss": 0.4166, "step": 10103 }, { "epoch": 1.04, "learning_rate": 5.028640098117518e-05, "loss": 0.5325, "step": 10104 }, { "epoch": 1.04, "learning_rate": 5.028458757027519e-05, "loss": 0.5594, "step": 10105 }, { "epoch": 1.04, "learning_rate": 5.028277402282411e-05, "loss": 0.526, "step": 10106 }, { "epoch": 1.04, "learning_rate": 5.028096033883414e-05, "loss": 0.2706, "step": 10107 }, { "epoch": 1.04, "learning_rate": 5.0279146518317475e-05, "loss": 0.2755, "step": 10108 }, { "epoch": 1.04, "learning_rate": 5.027733256128635e-05, "loss": 0.9418, "step": 10109 }, { "epoch": 1.04, "learning_rate": 5.027551846775297e-05, "loss": 0.5784, "step": 10110 }, { "epoch": 1.04, "learning_rate": 5.0273704237729545e-05, "loss": 0.291, "step": 10111 }, { "epoch": 1.04, "learning_rate": 5.027188987122828e-05, "loss": 0.2091, "step": 10112 }, { "epoch": 1.04, "learning_rate": 5.02700753682614e-05, "loss": 0.4104, "step": 10113 }, { "epoch": 1.04, "learning_rate": 5.026826072884111e-05, "loss": 0.7488, "step": 10114 }, { "epoch": 1.04, "learning_rate": 5.0266445952979645e-05, "loss": 0.4372, "step": 10115 }, { "epoch": 1.04, "learning_rate": 5.026463104068919e-05, "loss": 1.0588, "step": 10116 }, { "epoch": 1.04, "learning_rate": 5.026281599198199e-05, "loss": 0.281, "step": 10117 }, { "epoch": 1.04, "learning_rate": 5.0261000806870255e-05, "loss": 0.9004, "step": 10118 }, { "epoch": 1.04, "learning_rate": 5.02591854853662e-05, "loss": 0.7946, "step": 10119 }, { "epoch": 1.04, "learning_rate": 5.0257370027482054e-05, "loss": 0.7647, "step": 10120 }, { "epoch": 1.04, "learning_rate": 5.0255554433230025e-05, "loss": 0.3448, "step": 10121 }, { "epoch": 1.04, "learning_rate": 5.025373870262235e-05, "loss": 0.8468, "step": 10122 }, { "epoch": 1.04, "learning_rate": 5.025192283567123e-05, "loss": 0.4089, "step": 10123 }, { "epoch": 1.04, "learning_rate": 5.025010683238892e-05, "loss": 0.5395, "step": 10124 }, { "epoch": 1.04, "learning_rate": 5.024829069278762e-05, "loss": 0.8262, "step": 10125 }, { "epoch": 1.04, "learning_rate": 5.024647441687958e-05, "loss": 0.789, "step": 10126 }, { "epoch": 1.04, "learning_rate": 5.024465800467699e-05, "loss": 0.8109, "step": 10127 }, { "epoch": 1.04, "learning_rate": 5.024284145619212e-05, "loss": 0.6182, "step": 10128 }, { "epoch": 1.04, "learning_rate": 5.024102477143716e-05, "loss": 0.8665, "step": 10129 }, { "epoch": 1.04, "learning_rate": 5.023920795042436e-05, "loss": 0.4547, "step": 10130 }, { "epoch": 1.04, "learning_rate": 5.023739099316596e-05, "loss": 0.3096, "step": 10131 }, { "epoch": 1.04, "learning_rate": 5.023557389967417e-05, "loss": 0.2791, "step": 10132 }, { "epoch": 1.04, "learning_rate": 5.023375666996124e-05, "loss": 0.6337, "step": 10133 }, { "epoch": 1.04, "learning_rate": 5.023193930403938e-05, "loss": 0.567, "step": 10134 }, { "epoch": 1.04, "learning_rate": 5.023012180192085e-05, "loss": 0.3635, "step": 10135 }, { "epoch": 1.04, "learning_rate": 5.0228304163617864e-05, "loss": 0.1994, "step": 10136 }, { "epoch": 1.04, "learning_rate": 5.0226486389142677e-05, "loss": 0.7539, "step": 10137 }, { "epoch": 1.04, "learning_rate": 5.0224668478507506e-05, "loss": 0.7302, "step": 10138 }, { "epoch": 1.04, "learning_rate": 5.0222850431724616e-05, "loss": 0.8045, "step": 10139 }, { "epoch": 1.04, "learning_rate": 5.022103224880621e-05, "loss": 0.6611, "step": 10140 }, { "epoch": 1.04, "learning_rate": 5.021921392976456e-05, "loss": 0.6789, "step": 10141 }, { "epoch": 1.04, "learning_rate": 5.021739547461188e-05, "loss": 0.097, "step": 10142 }, { "epoch": 1.04, "learning_rate": 5.021557688336044e-05, "loss": 0.857, "step": 10143 }, { "epoch": 1.04, "learning_rate": 5.021375815602245e-05, "loss": 0.937, "step": 10144 }, { "epoch": 1.04, "learning_rate": 5.021193929261018e-05, "loss": 0.4163, "step": 10145 }, { "epoch": 1.04, "learning_rate": 5.021012029313586e-05, "loss": 0.3503, "step": 10146 }, { "epoch": 1.04, "learning_rate": 5.0208301157611745e-05, "loss": 0.9365, "step": 10147 }, { "epoch": 1.04, "learning_rate": 5.0206481886050066e-05, "loss": 0.4569, "step": 10148 }, { "epoch": 1.04, "learning_rate": 5.0204662478463074e-05, "loss": 0.3215, "step": 10149 }, { "epoch": 1.04, "learning_rate": 5.020284293486303e-05, "loss": 0.2907, "step": 10150 }, { "epoch": 1.04, "learning_rate": 5.020102325526219e-05, "loss": 0.9766, "step": 10151 }, { "epoch": 1.04, "learning_rate": 5.019920343967276e-05, "loss": 0.9925, "step": 10152 }, { "epoch": 1.04, "learning_rate": 5.019738348810704e-05, "loss": 1.0872, "step": 10153 }, { "epoch": 1.05, "learning_rate": 5.0195563400577246e-05, "loss": 0.1763, "step": 10154 }, { "epoch": 1.05, "learning_rate": 5.0193743177095654e-05, "loss": 0.8355, "step": 10155 }, { "epoch": 1.05, "learning_rate": 5.01919228176745e-05, "loss": 1.0536, "step": 10156 }, { "epoch": 1.05, "learning_rate": 5.019010232232605e-05, "loss": 0.3515, "step": 10157 }, { "epoch": 1.05, "learning_rate": 5.018828169106256e-05, "loss": 0.5913, "step": 10158 }, { "epoch": 1.05, "learning_rate": 5.0186460923896276e-05, "loss": 1.2775, "step": 10159 }, { "epoch": 1.05, "learning_rate": 5.018464002083946e-05, "loss": 1.1715, "step": 10160 }, { "epoch": 1.05, "learning_rate": 5.018281898190437e-05, "loss": 0.3549, "step": 10161 }, { "epoch": 1.05, "learning_rate": 5.018099780710326e-05, "loss": 0.7507, "step": 10162 }, { "epoch": 1.05, "learning_rate": 5.0179176496448406e-05, "loss": 0.6338, "step": 10163 }, { "epoch": 1.05, "learning_rate": 5.0177355049952056e-05, "loss": 0.4835, "step": 10164 }, { "epoch": 1.05, "learning_rate": 5.017553346762647e-05, "loss": 0.736, "step": 10165 }, { "epoch": 1.05, "learning_rate": 5.017371174948391e-05, "loss": 0.4883, "step": 10166 }, { "epoch": 1.05, "learning_rate": 5.0171889895536645e-05, "loss": 0.5613, "step": 10167 }, { "epoch": 1.05, "learning_rate": 5.0170067905796934e-05, "loss": 0.2891, "step": 10168 }, { "epoch": 1.05, "learning_rate": 5.0168245780277056e-05, "loss": 0.5258, "step": 10169 }, { "epoch": 1.05, "learning_rate": 5.0166423518989254e-05, "loss": 0.5988, "step": 10170 }, { "epoch": 1.05, "learning_rate": 5.0164601121945816e-05, "loss": 0.8301, "step": 10171 }, { "epoch": 1.05, "learning_rate": 5.0162778589159005e-05, "loss": 1.9103, "step": 10172 }, { "epoch": 1.05, "learning_rate": 5.016095592064108e-05, "loss": 1.2391, "step": 10173 }, { "epoch": 1.05, "learning_rate": 5.015913311640431e-05, "loss": 0.7491, "step": 10174 }, { "epoch": 1.05, "learning_rate": 5.015731017646098e-05, "loss": 1.0683, "step": 10175 }, { "epoch": 1.05, "learning_rate": 5.015548710082336e-05, "loss": 0.2545, "step": 10176 }, { "epoch": 1.05, "learning_rate": 5.015366388950371e-05, "loss": 0.8338, "step": 10177 }, { "epoch": 1.05, "learning_rate": 5.01518405425143e-05, "loss": 0.4913, "step": 10178 }, { "epoch": 1.05, "learning_rate": 5.0150017059867434e-05, "loss": 0.2979, "step": 10179 }, { "epoch": 1.05, "learning_rate": 5.0148193441575355e-05, "loss": 0.5743, "step": 10180 }, { "epoch": 1.05, "learning_rate": 5.0146369687650356e-05, "loss": 0.9877, "step": 10181 }, { "epoch": 1.05, "learning_rate": 5.01445457981047e-05, "loss": 0.4019, "step": 10182 }, { "epoch": 1.05, "learning_rate": 5.0142721772950695e-05, "loss": 0.3655, "step": 10183 }, { "epoch": 1.05, "learning_rate": 5.014089761220059e-05, "loss": 0.5513, "step": 10184 }, { "epoch": 1.05, "learning_rate": 5.013907331586667e-05, "loss": 0.3311, "step": 10185 }, { "epoch": 1.05, "learning_rate": 5.0137248883961235e-05, "loss": 0.5379, "step": 10186 }, { "epoch": 1.05, "learning_rate": 5.0135424316496544e-05, "loss": 0.5976, "step": 10187 }, { "epoch": 1.05, "learning_rate": 5.013359961348488e-05, "loss": 0.3993, "step": 10188 }, { "epoch": 1.05, "learning_rate": 5.0131774774938547e-05, "loss": 0.2412, "step": 10189 }, { "epoch": 1.05, "learning_rate": 5.012994980086981e-05, "loss": 0.5006, "step": 10190 }, { "epoch": 1.05, "learning_rate": 5.012812469129096e-05, "loss": 0.4612, "step": 10191 }, { "epoch": 1.05, "learning_rate": 5.0126299446214285e-05, "loss": 0.5853, "step": 10192 }, { "epoch": 1.05, "learning_rate": 5.012447406565208e-05, "loss": 0.526, "step": 10193 }, { "epoch": 1.05, "learning_rate": 5.0122648549616604e-05, "loss": 0.6569, "step": 10194 }, { "epoch": 1.05, "learning_rate": 5.012082289812019e-05, "loss": 0.2774, "step": 10195 }, { "epoch": 1.05, "learning_rate": 5.011899711117509e-05, "loss": 0.5488, "step": 10196 }, { "epoch": 1.05, "learning_rate": 5.011717118879361e-05, "loss": 0.3853, "step": 10197 }, { "epoch": 1.05, "learning_rate": 5.011534513098805e-05, "loss": 0.4051, "step": 10198 }, { "epoch": 1.05, "learning_rate": 5.011351893777068e-05, "loss": 0.2892, "step": 10199 }, { "epoch": 1.05, "learning_rate": 5.0111692609153815e-05, "loss": 0.5636, "step": 10200 }, { "epoch": 1.05, "learning_rate": 5.010986614514974e-05, "loss": 1.0908, "step": 10201 }, { "epoch": 1.05, "learning_rate": 5.010803954577075e-05, "loss": 0.4361, "step": 10202 }, { "epoch": 1.05, "learning_rate": 5.0106212811029145e-05, "loss": 0.8368, "step": 10203 }, { "epoch": 1.05, "learning_rate": 5.0104385940937215e-05, "loss": 0.3933, "step": 10204 }, { "epoch": 1.05, "learning_rate": 5.010255893550726e-05, "loss": 0.6594, "step": 10205 }, { "epoch": 1.05, "learning_rate": 5.01007317947516e-05, "loss": 0.4183, "step": 10206 }, { "epoch": 1.05, "learning_rate": 5.00989045186825e-05, "loss": 0.6826, "step": 10207 }, { "epoch": 1.05, "learning_rate": 5.009707710731228e-05, "loss": 0.789, "step": 10208 }, { "epoch": 1.05, "learning_rate": 5.0095249560653244e-05, "loss": 0.7893, "step": 10209 }, { "epoch": 1.05, "learning_rate": 5.0093421878717686e-05, "loss": 0.5913, "step": 10210 }, { "epoch": 1.05, "learning_rate": 5.009159406151791e-05, "loss": 0.531, "step": 10211 }, { "epoch": 1.05, "learning_rate": 5.0089766109066235e-05, "loss": 0.8545, "step": 10212 }, { "epoch": 1.05, "learning_rate": 5.008793802137494e-05, "loss": 0.5845, "step": 10213 }, { "epoch": 1.05, "learning_rate": 5.008610979845636e-05, "loss": 0.947, "step": 10214 }, { "epoch": 1.05, "learning_rate": 5.008428144032278e-05, "loss": 0.584, "step": 10215 }, { "epoch": 1.05, "learning_rate": 5.008245294698652e-05, "loss": 0.2289, "step": 10216 }, { "epoch": 1.05, "learning_rate": 5.0080624318459894e-05, "loss": 0.8011, "step": 10217 }, { "epoch": 1.05, "learning_rate": 5.007879555475519e-05, "loss": 1.2589, "step": 10218 }, { "epoch": 1.05, "learning_rate": 5.007696665588474e-05, "loss": 0.2982, "step": 10219 }, { "epoch": 1.05, "learning_rate": 5.0075137621860845e-05, "loss": 0.6796, "step": 10220 }, { "epoch": 1.05, "learning_rate": 5.007330845269582e-05, "loss": 1.0594, "step": 10221 }, { "epoch": 1.05, "learning_rate": 5.007147914840198e-05, "loss": 1.1066, "step": 10222 }, { "epoch": 1.05, "learning_rate": 5.006964970899164e-05, "loss": 0.1922, "step": 10223 }, { "epoch": 1.05, "learning_rate": 5.0067820134477116e-05, "loss": 0.5528, "step": 10224 }, { "epoch": 1.05, "learning_rate": 5.0065990424870714e-05, "loss": 0.3565, "step": 10225 }, { "epoch": 1.05, "learning_rate": 5.006416058018476e-05, "loss": 0.4942, "step": 10226 }, { "epoch": 1.05, "learning_rate": 5.0062330600431596e-05, "loss": 0.6149, "step": 10227 }, { "epoch": 1.05, "learning_rate": 5.006050048562349e-05, "loss": 0.6795, "step": 10228 }, { "epoch": 1.05, "learning_rate": 5.0058670235772793e-05, "loss": 0.1742, "step": 10229 }, { "epoch": 1.05, "learning_rate": 5.0056839850891824e-05, "loss": 0.96, "step": 10230 }, { "epoch": 1.05, "learning_rate": 5.005500933099291e-05, "loss": 0.9652, "step": 10231 }, { "epoch": 1.05, "learning_rate": 5.0053178676088346e-05, "loss": 0.6203, "step": 10232 }, { "epoch": 1.05, "learning_rate": 5.0051347886190496e-05, "loss": 0.865, "step": 10233 }, { "epoch": 1.05, "learning_rate": 5.004951696131166e-05, "loss": 0.604, "step": 10234 }, { "epoch": 1.05, "learning_rate": 5.0047685901464156e-05, "loss": 0.3103, "step": 10235 }, { "epoch": 1.05, "learning_rate": 5.0045854706660325e-05, "loss": 1.2228, "step": 10236 }, { "epoch": 1.05, "learning_rate": 5.004402337691249e-05, "loss": 0.267, "step": 10237 }, { "epoch": 1.05, "learning_rate": 5.004219191223299e-05, "loss": 0.4669, "step": 10238 }, { "epoch": 1.05, "learning_rate": 5.0040360312634135e-05, "loss": 0.4821, "step": 10239 }, { "epoch": 1.05, "learning_rate": 5.0038528578128266e-05, "loss": 0.558, "step": 10240 }, { "epoch": 1.05, "learning_rate": 5.003669670872771e-05, "loss": 1.204, "step": 10241 }, { "epoch": 1.05, "learning_rate": 5.00348647044448e-05, "loss": 0.976, "step": 10242 }, { "epoch": 1.05, "learning_rate": 5.003303256529186e-05, "loss": 0.2045, "step": 10243 }, { "epoch": 1.05, "learning_rate": 5.003120029128125e-05, "loss": 0.7649, "step": 10244 }, { "epoch": 1.05, "learning_rate": 5.002936788242527e-05, "loss": 0.5449, "step": 10245 }, { "epoch": 1.05, "learning_rate": 5.0027535338736286e-05, "loss": 0.2749, "step": 10246 }, { "epoch": 1.05, "learning_rate": 5.002570266022661e-05, "loss": 0.303, "step": 10247 }, { "epoch": 1.05, "learning_rate": 5.00238698469086e-05, "loss": 0.4667, "step": 10248 }, { "epoch": 1.05, "learning_rate": 5.002203689879458e-05, "loss": 0.6073, "step": 10249 }, { "epoch": 1.06, "learning_rate": 5.002020381589688e-05, "loss": 1.0791, "step": 10250 }, { "epoch": 1.06, "learning_rate": 5.001837059822786e-05, "loss": 0.4261, "step": 10251 }, { "epoch": 1.06, "learning_rate": 5.001653724579985e-05, "loss": 0.1399, "step": 10252 }, { "epoch": 1.06, "learning_rate": 5.001470375862521e-05, "loss": 0.4342, "step": 10253 }, { "epoch": 1.06, "learning_rate": 5.001287013671625e-05, "loss": 0.3562, "step": 10254 }, { "epoch": 1.06, "learning_rate": 5.001103638008533e-05, "loss": 0.263, "step": 10255 }, { "epoch": 1.06, "learning_rate": 5.00092024887448e-05, "loss": 1.1345, "step": 10256 }, { "epoch": 1.06, "learning_rate": 5.0007368462707005e-05, "loss": 0.4043, "step": 10257 }, { "epoch": 1.06, "learning_rate": 5.000553430198429e-05, "loss": 0.6243, "step": 10258 }, { "epoch": 1.06, "learning_rate": 5.000370000658899e-05, "loss": 0.5823, "step": 10259 }, { "epoch": 1.06, "learning_rate": 5.000186557653347e-05, "loss": 0.4894, "step": 10260 }, { "epoch": 1.06, "learning_rate": 5.000003101183006e-05, "loss": 0.8979, "step": 10261 }, { "epoch": 1.06, "learning_rate": 4.999819631249113e-05, "loss": 0.4425, "step": 10262 }, { "epoch": 1.06, "learning_rate": 4.9996361478529015e-05, "loss": 0.4078, "step": 10263 }, { "epoch": 1.06, "learning_rate": 4.9994526509956086e-05, "loss": 0.4386, "step": 10264 }, { "epoch": 1.06, "learning_rate": 4.9992691406784675e-05, "loss": 0.6414, "step": 10265 }, { "epoch": 1.06, "learning_rate": 4.999085616902714e-05, "loss": 0.8492, "step": 10266 }, { "epoch": 1.06, "learning_rate": 4.998902079669584e-05, "loss": 0.205, "step": 10267 }, { "epoch": 1.06, "learning_rate": 4.9987185289803135e-05, "loss": 1.0025, "step": 10268 }, { "epoch": 1.06, "learning_rate": 4.998534964836136e-05, "loss": 0.5725, "step": 10269 }, { "epoch": 1.06, "learning_rate": 4.9983513872382904e-05, "loss": 0.864, "step": 10270 }, { "epoch": 1.06, "learning_rate": 4.99816779618801e-05, "loss": 0.5954, "step": 10271 }, { "epoch": 1.06, "learning_rate": 4.997984191686532e-05, "loss": 0.3438, "step": 10272 }, { "epoch": 1.06, "learning_rate": 4.997800573735091e-05, "loss": 1.0288, "step": 10273 }, { "epoch": 1.06, "learning_rate": 4.9976169423349255e-05, "loss": 0.857, "step": 10274 }, { "epoch": 1.06, "learning_rate": 4.997433297487269e-05, "loss": 0.9646, "step": 10275 }, { "epoch": 1.06, "learning_rate": 4.997249639193359e-05, "loss": 1.4259, "step": 10276 }, { "epoch": 1.06, "learning_rate": 4.9970659674544324e-05, "loss": 0.7542, "step": 10277 }, { "epoch": 1.06, "learning_rate": 4.996882282271724e-05, "loss": 0.2447, "step": 10278 }, { "epoch": 1.06, "learning_rate": 4.9966985836464724e-05, "loss": 0.3885, "step": 10279 }, { "epoch": 1.06, "learning_rate": 4.996514871579912e-05, "loss": 0.6307, "step": 10280 }, { "epoch": 1.06, "learning_rate": 4.996331146073281e-05, "loss": 0.7303, "step": 10281 }, { "epoch": 1.06, "learning_rate": 4.996147407127816e-05, "loss": 0.4496, "step": 10282 }, { "epoch": 1.06, "learning_rate": 4.9959636547447545e-05, "loss": 0.3057, "step": 10283 }, { "epoch": 1.06, "learning_rate": 4.995779888925332e-05, "loss": 0.654, "step": 10284 }, { "epoch": 1.06, "learning_rate": 4.995596109670786e-05, "loss": 1.1827, "step": 10285 }, { "epoch": 1.06, "learning_rate": 4.995412316982354e-05, "loss": 0.8247, "step": 10286 }, { "epoch": 1.06, "learning_rate": 4.9952285108612736e-05, "loss": 1.1566, "step": 10287 }, { "epoch": 1.06, "learning_rate": 4.995044691308781e-05, "loss": 0.1323, "step": 10288 }, { "epoch": 1.06, "learning_rate": 4.994860858326115e-05, "loss": 0.6878, "step": 10289 }, { "epoch": 1.06, "learning_rate": 4.9946770119145124e-05, "loss": 0.5538, "step": 10290 }, { "epoch": 1.06, "learning_rate": 4.994493152075211e-05, "loss": 0.9729, "step": 10291 }, { "epoch": 1.06, "learning_rate": 4.9943092788094475e-05, "loss": 0.1588, "step": 10292 }, { "epoch": 1.06, "learning_rate": 4.994125392118461e-05, "loss": 0.7628, "step": 10293 }, { "epoch": 1.06, "learning_rate": 4.993941492003489e-05, "loss": 0.8314, "step": 10294 }, { "epoch": 1.06, "learning_rate": 4.99375757846577e-05, "loss": 0.5522, "step": 10295 }, { "epoch": 1.06, "learning_rate": 4.9935736515065416e-05, "loss": 0.6891, "step": 10296 }, { "epoch": 1.06, "learning_rate": 4.99338971112704e-05, "loss": 0.4959, "step": 10297 }, { "epoch": 1.06, "learning_rate": 4.993205757328507e-05, "loss": 0.8453, "step": 10298 }, { "epoch": 1.06, "learning_rate": 4.993021790112179e-05, "loss": 0.6356, "step": 10299 }, { "epoch": 1.06, "learning_rate": 4.992837809479294e-05, "loss": 0.944, "step": 10300 }, { "epoch": 1.06, "learning_rate": 4.992653815431091e-05, "loss": 0.6114, "step": 10301 }, { "epoch": 1.06, "learning_rate": 4.99246980796881e-05, "loss": 0.8279, "step": 10302 }, { "epoch": 1.06, "learning_rate": 4.992285787093688e-05, "loss": 0.8423, "step": 10303 }, { "epoch": 1.06, "learning_rate": 4.992101752806963e-05, "loss": 1.0836, "step": 10304 }, { "epoch": 1.06, "learning_rate": 4.9919177051098765e-05, "loss": 0.7561, "step": 10305 }, { "epoch": 1.06, "learning_rate": 4.9917336440036654e-05, "loss": 0.2994, "step": 10306 }, { "epoch": 1.06, "learning_rate": 4.9915495694895706e-05, "loss": 0.3362, "step": 10307 }, { "epoch": 1.06, "learning_rate": 4.991365481568829e-05, "loss": 0.6158, "step": 10308 }, { "epoch": 1.06, "learning_rate": 4.9911813802426804e-05, "loss": 0.1351, "step": 10309 }, { "epoch": 1.06, "learning_rate": 4.990997265512366e-05, "loss": 0.5174, "step": 10310 }, { "epoch": 1.06, "learning_rate": 4.9908131373791224e-05, "loss": 0.5609, "step": 10311 }, { "epoch": 1.06, "learning_rate": 4.990628995844192e-05, "loss": 0.5696, "step": 10312 }, { "epoch": 1.06, "learning_rate": 4.990444840908812e-05, "loss": 0.2112, "step": 10313 }, { "epoch": 1.06, "learning_rate": 4.9902606725742236e-05, "loss": 1.0769, "step": 10314 }, { "epoch": 1.06, "learning_rate": 4.990076490841666e-05, "loss": 0.7966, "step": 10315 }, { "epoch": 1.06, "learning_rate": 4.989892295712379e-05, "loss": 0.5871, "step": 10316 }, { "epoch": 1.06, "learning_rate": 4.989708087187602e-05, "loss": 0.9447, "step": 10317 }, { "epoch": 1.06, "learning_rate": 4.989523865268577e-05, "loss": 0.5087, "step": 10318 }, { "epoch": 1.06, "learning_rate": 4.989339629956542e-05, "loss": 0.1807, "step": 10319 }, { "epoch": 1.06, "learning_rate": 4.989155381252738e-05, "loss": 0.1245, "step": 10320 }, { "epoch": 1.06, "learning_rate": 4.988971119158406e-05, "loss": 1.2537, "step": 10321 }, { "epoch": 1.06, "learning_rate": 4.988786843674786e-05, "loss": 0.4581, "step": 10322 }, { "epoch": 1.06, "learning_rate": 4.988602554803117e-05, "loss": 0.395, "step": 10323 }, { "epoch": 1.06, "learning_rate": 4.988418252544642e-05, "loss": 0.8412, "step": 10324 }, { "epoch": 1.06, "learning_rate": 4.9882339369006e-05, "loss": 0.8857, "step": 10325 }, { "epoch": 1.06, "learning_rate": 4.988049607872232e-05, "loss": 0.9173, "step": 10326 }, { "epoch": 1.06, "learning_rate": 4.9878652654607807e-05, "loss": 0.506, "step": 10327 }, { "epoch": 1.06, "learning_rate": 4.987680909667484e-05, "loss": 0.5846, "step": 10328 }, { "epoch": 1.06, "learning_rate": 4.9874965404935854e-05, "loss": 1.1765, "step": 10329 }, { "epoch": 1.06, "learning_rate": 4.987312157940325e-05, "loss": 0.6837, "step": 10330 }, { "epoch": 1.06, "learning_rate": 4.9871277620089444e-05, "loss": 0.6748, "step": 10331 }, { "epoch": 1.06, "learning_rate": 4.9869433527006835e-05, "loss": 0.6174, "step": 10332 }, { "epoch": 1.06, "learning_rate": 4.9867589300167856e-05, "loss": 0.3447, "step": 10333 }, { "epoch": 1.06, "learning_rate": 4.9865744939584923e-05, "loss": 0.5917, "step": 10334 }, { "epoch": 1.06, "learning_rate": 4.986390044527043e-05, "loss": 0.2888, "step": 10335 }, { "epoch": 1.06, "learning_rate": 4.986205581723682e-05, "loss": 0.1991, "step": 10336 }, { "epoch": 1.06, "learning_rate": 4.986021105549648e-05, "loss": 0.438, "step": 10337 }, { "epoch": 1.06, "learning_rate": 4.9858366160061855e-05, "loss": 0.295, "step": 10338 }, { "epoch": 1.06, "learning_rate": 4.9856521130945354e-05, "loss": 0.9896, "step": 10339 }, { "epoch": 1.06, "learning_rate": 4.9854675968159405e-05, "loss": 0.9856, "step": 10340 }, { "epoch": 1.06, "learning_rate": 4.9852830671716426e-05, "loss": 0.4324, "step": 10341 }, { "epoch": 1.06, "learning_rate": 4.9850985241628826e-05, "loss": 0.3436, "step": 10342 }, { "epoch": 1.06, "learning_rate": 4.984913967790904e-05, "loss": 0.2199, "step": 10343 }, { "epoch": 1.06, "learning_rate": 4.9847293980569494e-05, "loss": 0.5597, "step": 10344 }, { "epoch": 1.06, "learning_rate": 4.9845448149622604e-05, "loss": 0.2749, "step": 10345 }, { "epoch": 1.07, "learning_rate": 4.984360218508081e-05, "loss": 0.6976, "step": 10346 }, { "epoch": 1.07, "learning_rate": 4.984175608695652e-05, "loss": 0.5749, "step": 10347 }, { "epoch": 1.07, "learning_rate": 4.9839909855262165e-05, "loss": 0.966, "step": 10348 }, { "epoch": 1.07, "learning_rate": 4.9838063490010185e-05, "loss": 0.3521, "step": 10349 }, { "epoch": 1.07, "learning_rate": 4.9836216991213004e-05, "loss": 1.0296, "step": 10350 }, { "epoch": 1.07, "learning_rate": 4.983437035888306e-05, "loss": 0.3794, "step": 10351 }, { "epoch": 1.07, "learning_rate": 4.983252359303277e-05, "loss": 0.382, "step": 10352 }, { "epoch": 1.07, "learning_rate": 4.983067669367456e-05, "loss": 0.9092, "step": 10353 }, { "epoch": 1.07, "learning_rate": 4.9828829660820884e-05, "loss": 0.6652, "step": 10354 }, { "epoch": 1.07, "learning_rate": 4.9826982494484174e-05, "loss": 1.1036, "step": 10355 }, { "epoch": 1.07, "learning_rate": 4.982513519467684e-05, "loss": 0.674, "step": 10356 }, { "epoch": 1.07, "learning_rate": 4.9823287761411346e-05, "loss": 0.2667, "step": 10357 }, { "epoch": 1.07, "learning_rate": 4.982144019470011e-05, "loss": 0.6028, "step": 10358 }, { "epoch": 1.07, "learning_rate": 4.981959249455559e-05, "loss": 0.7908, "step": 10359 }, { "epoch": 1.07, "learning_rate": 4.98177446609902e-05, "loss": 0.6414, "step": 10360 }, { "epoch": 1.07, "learning_rate": 4.98158966940164e-05, "loss": 0.1848, "step": 10361 }, { "epoch": 1.07, "learning_rate": 4.9814048593646604e-05, "loss": 1.2569, "step": 10362 }, { "epoch": 1.07, "learning_rate": 4.981220035989328e-05, "loss": 0.2112, "step": 10363 }, { "epoch": 1.07, "learning_rate": 4.9810351992768855e-05, "loss": 1.2109, "step": 10364 }, { "epoch": 1.07, "learning_rate": 4.9808503492285775e-05, "loss": 0.8755, "step": 10365 }, { "epoch": 1.07, "learning_rate": 4.980665485845649e-05, "loss": 0.7165, "step": 10366 }, { "epoch": 1.07, "learning_rate": 4.980480609129343e-05, "loss": 0.7797, "step": 10367 }, { "epoch": 1.07, "learning_rate": 4.980295719080906e-05, "loss": 1.0033, "step": 10368 }, { "epoch": 1.07, "learning_rate": 4.980110815701581e-05, "loss": 0.381, "step": 10369 }, { "epoch": 1.07, "learning_rate": 4.979925898992613e-05, "loss": 1.002, "step": 10370 }, { "epoch": 1.07, "learning_rate": 4.979740968955247e-05, "loss": 0.4116, "step": 10371 }, { "epoch": 1.07, "learning_rate": 4.9795560255907294e-05, "loss": 0.6831, "step": 10372 }, { "epoch": 1.07, "learning_rate": 4.979371068900302e-05, "loss": 0.7565, "step": 10373 }, { "epoch": 1.07, "learning_rate": 4.9791860988852125e-05, "loss": 0.9487, "step": 10374 }, { "epoch": 1.07, "learning_rate": 4.979001115546705e-05, "loss": 0.2394, "step": 10375 }, { "epoch": 1.07, "learning_rate": 4.978816118886025e-05, "loss": 0.1617, "step": 10376 }, { "epoch": 1.07, "learning_rate": 4.978631108904419e-05, "loss": 0.2727, "step": 10377 }, { "epoch": 1.07, "learning_rate": 4.9784460856031305e-05, "loss": 0.9327, "step": 10378 }, { "epoch": 1.07, "learning_rate": 4.9782610489834065e-05, "loss": 0.4241, "step": 10379 }, { "epoch": 1.07, "learning_rate": 4.978075999046491e-05, "loss": 0.3377, "step": 10380 }, { "epoch": 1.07, "learning_rate": 4.9778909357936306e-05, "loss": 0.8321, "step": 10381 }, { "epoch": 1.07, "learning_rate": 4.9777058592260724e-05, "loss": 1.162, "step": 10382 }, { "epoch": 1.07, "learning_rate": 4.97752076934506e-05, "loss": 0.8613, "step": 10383 }, { "epoch": 1.07, "learning_rate": 4.977335666151841e-05, "loss": 0.4782, "step": 10384 }, { "epoch": 1.07, "learning_rate": 4.9771505496476615e-05, "loss": 0.1158, "step": 10385 }, { "epoch": 1.07, "learning_rate": 4.9769654198337664e-05, "loss": 0.5092, "step": 10386 }, { "epoch": 1.07, "learning_rate": 4.976780276711403e-05, "loss": 0.3773, "step": 10387 }, { "epoch": 1.07, "learning_rate": 4.976595120281817e-05, "loss": 0.2525, "step": 10388 }, { "epoch": 1.07, "learning_rate": 4.9764099505462545e-05, "loss": 0.2278, "step": 10389 }, { "epoch": 1.07, "learning_rate": 4.9762247675059636e-05, "loss": 0.2927, "step": 10390 }, { "epoch": 1.07, "learning_rate": 4.97603957116219e-05, "loss": 0.3789, "step": 10391 }, { "epoch": 1.07, "learning_rate": 4.9758543615161804e-05, "loss": 0.6171, "step": 10392 }, { "epoch": 1.07, "learning_rate": 4.975669138569181e-05, "loss": 0.3085, "step": 10393 }, { "epoch": 1.07, "learning_rate": 4.975483902322439e-05, "loss": 0.3066, "step": 10394 }, { "epoch": 1.07, "learning_rate": 4.9752986527772024e-05, "loss": 0.5206, "step": 10395 }, { "epoch": 1.07, "learning_rate": 4.9751133899347175e-05, "loss": 0.3719, "step": 10396 }, { "epoch": 1.07, "learning_rate": 4.974928113796231e-05, "loss": 0.5578, "step": 10397 }, { "epoch": 1.07, "learning_rate": 4.9747428243629905e-05, "loss": 0.2665, "step": 10398 }, { "epoch": 1.07, "learning_rate": 4.974557521636243e-05, "loss": 0.2374, "step": 10399 }, { "epoch": 1.07, "learning_rate": 4.974372205617236e-05, "loss": 0.7806, "step": 10400 }, { "epoch": 1.07, "learning_rate": 4.974186876307219e-05, "loss": 0.722, "step": 10401 }, { "epoch": 1.07, "learning_rate": 4.974001533707437e-05, "loss": 1.0455, "step": 10402 }, { "epoch": 1.07, "learning_rate": 4.9738161778191386e-05, "loss": 1.544, "step": 10403 }, { "epoch": 1.07, "learning_rate": 4.973630808643571e-05, "loss": 0.3633, "step": 10404 }, { "epoch": 1.07, "learning_rate": 4.9734454261819826e-05, "loss": 1.1859, "step": 10405 }, { "epoch": 1.07, "learning_rate": 4.9732600304356216e-05, "loss": 0.5627, "step": 10406 }, { "epoch": 1.07, "learning_rate": 4.9730746214057356e-05, "loss": 0.6787, "step": 10407 }, { "epoch": 1.07, "learning_rate": 4.972889199093573e-05, "loss": 0.9089, "step": 10408 }, { "epoch": 1.07, "learning_rate": 4.9727037635003824e-05, "loss": 0.5738, "step": 10409 }, { "epoch": 1.07, "learning_rate": 4.9725183146274116e-05, "loss": 0.5494, "step": 10410 }, { "epoch": 1.07, "learning_rate": 4.9723328524759085e-05, "loss": 1.4243, "step": 10411 }, { "epoch": 1.07, "learning_rate": 4.9721473770471226e-05, "loss": 0.3205, "step": 10412 }, { "epoch": 1.07, "learning_rate": 4.9719618883423014e-05, "loss": 0.2978, "step": 10413 }, { "epoch": 1.07, "learning_rate": 4.9717763863626945e-05, "loss": 0.6496, "step": 10414 }, { "epoch": 1.07, "learning_rate": 4.9715908711095514e-05, "loss": 0.9629, "step": 10415 }, { "epoch": 1.07, "learning_rate": 4.9714053425841176e-05, "loss": 0.4627, "step": 10416 }, { "epoch": 1.07, "learning_rate": 4.971219800787646e-05, "loss": 0.4071, "step": 10417 }, { "epoch": 1.07, "learning_rate": 4.971034245721384e-05, "loss": 0.6436, "step": 10418 }, { "epoch": 1.07, "learning_rate": 4.97084867738658e-05, "loss": 1.1187, "step": 10419 }, { "epoch": 1.07, "learning_rate": 4.970663095784484e-05, "loss": 0.7459, "step": 10420 }, { "epoch": 1.07, "learning_rate": 4.970477500916345e-05, "loss": 1.015, "step": 10421 }, { "epoch": 1.07, "learning_rate": 4.970291892783413e-05, "loss": 0.2305, "step": 10422 }, { "epoch": 1.07, "learning_rate": 4.970106271386936e-05, "loss": 0.4985, "step": 10423 }, { "epoch": 1.07, "learning_rate": 4.969920636728166e-05, "loss": 0.386, "step": 10424 }, { "epoch": 1.07, "learning_rate": 4.9697349888083494e-05, "loss": 0.9322, "step": 10425 }, { "epoch": 1.07, "learning_rate": 4.969549327628739e-05, "loss": 0.4331, "step": 10426 }, { "epoch": 1.07, "learning_rate": 4.969363653190583e-05, "loss": 0.9338, "step": 10427 }, { "epoch": 1.07, "learning_rate": 4.969177965495132e-05, "loss": 0.5024, "step": 10428 }, { "epoch": 1.07, "learning_rate": 4.9689922645436355e-05, "loss": 0.7465, "step": 10429 }, { "epoch": 1.07, "learning_rate": 4.968806550337344e-05, "loss": 1.2379, "step": 10430 }, { "epoch": 1.07, "learning_rate": 4.968620822877507e-05, "loss": 0.9577, "step": 10431 }, { "epoch": 1.07, "learning_rate": 4.968435082165375e-05, "loss": 0.2022, "step": 10432 }, { "epoch": 1.07, "learning_rate": 4.9682493282021997e-05, "loss": 0.6962, "step": 10433 }, { "epoch": 1.07, "learning_rate": 4.96806356098923e-05, "loss": 0.2908, "step": 10434 }, { "epoch": 1.07, "learning_rate": 4.9678777805277164e-05, "loss": 0.1597, "step": 10435 }, { "epoch": 1.07, "learning_rate": 4.9676919868189103e-05, "loss": 0.5172, "step": 10436 }, { "epoch": 1.07, "learning_rate": 4.9675061798640615e-05, "loss": 0.8264, "step": 10437 }, { "epoch": 1.07, "learning_rate": 4.967320359664422e-05, "loss": 0.2773, "step": 10438 }, { "epoch": 1.07, "learning_rate": 4.967134526221243e-05, "loss": 0.5448, "step": 10439 }, { "epoch": 1.07, "learning_rate": 4.966948679535774e-05, "loss": 0.428, "step": 10440 }, { "epoch": 1.07, "learning_rate": 4.966762819609266e-05, "loss": 0.2461, "step": 10441 }, { "epoch": 1.08, "learning_rate": 4.966576946442972e-05, "loss": 1.5348, "step": 10442 }, { "epoch": 1.08, "learning_rate": 4.9663910600381417e-05, "loss": 0.8542, "step": 10443 }, { "epoch": 1.08, "learning_rate": 4.966205160396027e-05, "loss": 0.8167, "step": 10444 }, { "epoch": 1.08, "learning_rate": 4.966019247517879e-05, "loss": 0.4318, "step": 10445 }, { "epoch": 1.08, "learning_rate": 4.96583332140495e-05, "loss": 0.4785, "step": 10446 }, { "epoch": 1.08, "learning_rate": 4.965647382058489e-05, "loss": 1.0711, "step": 10447 }, { "epoch": 1.08, "learning_rate": 4.965461429479753e-05, "loss": 0.1886, "step": 10448 }, { "epoch": 1.08, "learning_rate": 4.965275463669988e-05, "loss": 0.6962, "step": 10449 }, { "epoch": 1.08, "learning_rate": 4.965089484630449e-05, "loss": 0.6259, "step": 10450 }, { "epoch": 1.08, "learning_rate": 4.9649034923623874e-05, "loss": 0.6501, "step": 10451 }, { "epoch": 1.08, "learning_rate": 4.9647174868670557e-05, "loss": 1.1049, "step": 10452 }, { "epoch": 1.08, "learning_rate": 4.9645314681457056e-05, "loss": 0.7632, "step": 10453 }, { "epoch": 1.08, "learning_rate": 4.964345436199589e-05, "loss": 0.1895, "step": 10454 }, { "epoch": 1.08, "learning_rate": 4.964159391029959e-05, "loss": 0.6379, "step": 10455 }, { "epoch": 1.08, "learning_rate": 4.963973332638067e-05, "loss": 0.2943, "step": 10456 }, { "epoch": 1.08, "learning_rate": 4.963787261025166e-05, "loss": 1.1325, "step": 10457 }, { "epoch": 1.08, "learning_rate": 4.963601176192508e-05, "loss": 0.9106, "step": 10458 }, { "epoch": 1.08, "learning_rate": 4.9634150781413484e-05, "loss": 0.9521, "step": 10459 }, { "epoch": 1.08, "learning_rate": 4.963228966872937e-05, "loss": 0.7031, "step": 10460 }, { "epoch": 1.08, "learning_rate": 4.9630428423885273e-05, "loss": 0.3136, "step": 10461 }, { "epoch": 1.08, "learning_rate": 4.962856704689372e-05, "loss": 0.6883, "step": 10462 }, { "epoch": 1.08, "learning_rate": 4.962670553776725e-05, "loss": 0.5492, "step": 10463 }, { "epoch": 1.08, "learning_rate": 4.962484389651839e-05, "loss": 0.7371, "step": 10464 }, { "epoch": 1.08, "learning_rate": 4.9622982123159675e-05, "loss": 0.5869, "step": 10465 }, { "epoch": 1.08, "learning_rate": 4.9621120217703635e-05, "loss": 0.6483, "step": 10466 }, { "epoch": 1.08, "learning_rate": 4.961925818016281e-05, "loss": 0.6784, "step": 10467 }, { "epoch": 1.08, "learning_rate": 4.961739601054972e-05, "loss": 0.7215, "step": 10468 }, { "epoch": 1.08, "learning_rate": 4.961553370887691e-05, "loss": 0.7009, "step": 10469 }, { "epoch": 1.08, "learning_rate": 4.961367127515692e-05, "loss": 0.5517, "step": 10470 }, { "epoch": 1.08, "learning_rate": 4.961180870940229e-05, "loss": 0.8846, "step": 10471 }, { "epoch": 1.08, "learning_rate": 4.960994601162555e-05, "loss": 0.5954, "step": 10472 }, { "epoch": 1.08, "learning_rate": 4.960808318183924e-05, "loss": 0.2438, "step": 10473 }, { "epoch": 1.08, "learning_rate": 4.96062202200559e-05, "loss": 0.8805, "step": 10474 }, { "epoch": 1.08, "learning_rate": 4.960435712628808e-05, "loss": 0.6494, "step": 10475 }, { "epoch": 1.08, "learning_rate": 4.96024939005483e-05, "loss": 0.7096, "step": 10476 }, { "epoch": 1.08, "learning_rate": 4.9600630542849134e-05, "loss": 1.1029, "step": 10477 }, { "epoch": 1.08, "learning_rate": 4.95987670532031e-05, "loss": 1.1296, "step": 10478 }, { "epoch": 1.08, "learning_rate": 4.9596903431622756e-05, "loss": 0.7528, "step": 10479 }, { "epoch": 1.08, "learning_rate": 4.9595039678120634e-05, "loss": 0.2811, "step": 10480 }, { "epoch": 1.08, "learning_rate": 4.9593175792709305e-05, "loss": 0.5094, "step": 10481 }, { "epoch": 1.08, "learning_rate": 4.959131177540129e-05, "loss": 0.3821, "step": 10482 }, { "epoch": 1.08, "learning_rate": 4.958944762620915e-05, "loss": 0.2408, "step": 10483 }, { "epoch": 1.08, "learning_rate": 4.9587583345145434e-05, "loss": 0.5001, "step": 10484 }, { "epoch": 1.08, "learning_rate": 4.9585718932222694e-05, "loss": 0.4618, "step": 10485 }, { "epoch": 1.08, "learning_rate": 4.9583854387453476e-05, "loss": 0.1615, "step": 10486 }, { "epoch": 1.08, "learning_rate": 4.958198971085032e-05, "loss": 1.0348, "step": 10487 }, { "epoch": 1.08, "learning_rate": 4.958012490242581e-05, "loss": 0.6378, "step": 10488 }, { "epoch": 1.08, "learning_rate": 4.957825996219247e-05, "loss": 0.6508, "step": 10489 }, { "epoch": 1.08, "learning_rate": 4.957639489016287e-05, "loss": 0.2979, "step": 10490 }, { "epoch": 1.08, "learning_rate": 4.9574529686349554e-05, "loss": 0.5343, "step": 10491 }, { "epoch": 1.08, "learning_rate": 4.957266435076509e-05, "loss": 0.4031, "step": 10492 }, { "epoch": 1.08, "learning_rate": 4.957079888342202e-05, "loss": 0.8669, "step": 10493 }, { "epoch": 1.08, "learning_rate": 4.956893328433291e-05, "loss": 0.705, "step": 10494 }, { "epoch": 1.08, "learning_rate": 4.956706755351033e-05, "loss": 0.5582, "step": 10495 }, { "epoch": 1.08, "learning_rate": 4.956520169096683e-05, "loss": 0.4317, "step": 10496 }, { "epoch": 1.08, "learning_rate": 4.9563335696714966e-05, "loss": 0.5366, "step": 10497 }, { "epoch": 1.08, "learning_rate": 4.95614695707673e-05, "loss": 0.6448, "step": 10498 }, { "epoch": 1.08, "learning_rate": 4.95596033131364e-05, "loss": 0.3596, "step": 10499 }, { "epoch": 1.08, "learning_rate": 4.9557736923834834e-05, "loss": 0.5753, "step": 10500 }, { "epoch": 1.08, "learning_rate": 4.955587040287515e-05, "loss": 0.7209, "step": 10501 }, { "epoch": 1.08, "learning_rate": 4.9554003750269924e-05, "loss": 0.2928, "step": 10502 }, { "epoch": 1.08, "learning_rate": 4.9552136966031726e-05, "loss": 0.152, "step": 10503 }, { "epoch": 1.08, "learning_rate": 4.955027005017311e-05, "loss": 1.2195, "step": 10504 }, { "epoch": 1.08, "learning_rate": 4.954840300270665e-05, "loss": 0.0712, "step": 10505 }, { "epoch": 1.08, "learning_rate": 4.9546535823644924e-05, "loss": 0.5169, "step": 10506 }, { "epoch": 1.08, "learning_rate": 4.954466851300048e-05, "loss": 0.6435, "step": 10507 }, { "epoch": 1.08, "learning_rate": 4.954280107078591e-05, "loss": 0.6521, "step": 10508 }, { "epoch": 1.08, "learning_rate": 4.9540933497013775e-05, "loss": 1.0634, "step": 10509 }, { "epoch": 1.08, "learning_rate": 4.953906579169665e-05, "loss": 0.9213, "step": 10510 }, { "epoch": 1.08, "learning_rate": 4.9537197954847105e-05, "loss": 0.4202, "step": 10511 }, { "epoch": 1.08, "learning_rate": 4.9535329986477704e-05, "loss": 1.4203, "step": 10512 }, { "epoch": 1.08, "learning_rate": 4.9533461886601044e-05, "loss": 0.0719, "step": 10513 }, { "epoch": 1.08, "learning_rate": 4.953159365522968e-05, "loss": 0.6993, "step": 10514 }, { "epoch": 1.08, "learning_rate": 4.95297252923762e-05, "loss": 1.2715, "step": 10515 }, { "epoch": 1.08, "learning_rate": 4.952785679805319e-05, "loss": 0.1988, "step": 10516 }, { "epoch": 1.08, "learning_rate": 4.952598817227321e-05, "loss": 0.4027, "step": 10517 }, { "epoch": 1.08, "learning_rate": 4.9524119415048835e-05, "loss": 0.3615, "step": 10518 }, { "epoch": 1.08, "learning_rate": 4.9522250526392665e-05, "loss": 0.5739, "step": 10519 }, { "epoch": 1.08, "learning_rate": 4.952038150631728e-05, "loss": 0.354, "step": 10520 }, { "epoch": 1.08, "learning_rate": 4.951851235483525e-05, "loss": 0.7943, "step": 10521 }, { "epoch": 1.08, "learning_rate": 4.951664307195916e-05, "loss": 0.4973, "step": 10522 }, { "epoch": 1.08, "learning_rate": 4.951477365770159e-05, "loss": 0.8671, "step": 10523 }, { "epoch": 1.08, "learning_rate": 4.951290411207514e-05, "loss": 1.0357, "step": 10524 }, { "epoch": 1.08, "learning_rate": 4.951103443509237e-05, "loss": 0.456, "step": 10525 }, { "epoch": 1.08, "learning_rate": 4.95091646267659e-05, "loss": 0.2482, "step": 10526 }, { "epoch": 1.08, "learning_rate": 4.950729468710828e-05, "loss": 1.1981, "step": 10527 }, { "epoch": 1.08, "learning_rate": 4.950542461613213e-05, "loss": 0.1143, "step": 10528 }, { "epoch": 1.08, "learning_rate": 4.9503554413850024e-05, "loss": 0.8629, "step": 10529 }, { "epoch": 1.08, "learning_rate": 4.950168408027455e-05, "loss": 0.3784, "step": 10530 }, { "epoch": 1.08, "learning_rate": 4.9499813615418305e-05, "loss": 0.5844, "step": 10531 }, { "epoch": 1.08, "learning_rate": 4.949794301929387e-05, "loss": 0.4343, "step": 10532 }, { "epoch": 1.08, "learning_rate": 4.949607229191385e-05, "loss": 0.9227, "step": 10533 }, { "epoch": 1.08, "learning_rate": 4.9494201433290835e-05, "loss": 0.5064, "step": 10534 }, { "epoch": 1.08, "learning_rate": 4.949233044343741e-05, "loss": 0.9346, "step": 10535 }, { "epoch": 1.08, "learning_rate": 4.9490459322366196e-05, "loss": 0.5661, "step": 10536 }, { "epoch": 1.09, "learning_rate": 4.948858807008975e-05, "loss": 1.0999, "step": 10537 }, { "epoch": 1.09, "learning_rate": 4.9486716686620695e-05, "loss": 0.4835, "step": 10538 }, { "epoch": 1.09, "learning_rate": 4.948484517197162e-05, "loss": 0.4157, "step": 10539 }, { "epoch": 1.09, "learning_rate": 4.948297352615513e-05, "loss": 0.6433, "step": 10540 }, { "epoch": 1.09, "learning_rate": 4.948110174918383e-05, "loss": 0.4191, "step": 10541 }, { "epoch": 1.09, "learning_rate": 4.9479229841070294e-05, "loss": 0.8345, "step": 10542 }, { "epoch": 1.09, "learning_rate": 4.947735780182715e-05, "loss": 0.4317, "step": 10543 }, { "epoch": 1.09, "learning_rate": 4.947548563146699e-05, "loss": 1.2701, "step": 10544 }, { "epoch": 1.09, "learning_rate": 4.947361333000241e-05, "loss": 0.185, "step": 10545 }, { "epoch": 1.09, "learning_rate": 4.947174089744603e-05, "loss": 0.4755, "step": 10546 }, { "epoch": 1.09, "learning_rate": 4.946986833381045e-05, "loss": 0.4554, "step": 10547 }, { "epoch": 1.09, "learning_rate": 4.946799563910827e-05, "loss": 0.9442, "step": 10548 }, { "epoch": 1.09, "learning_rate": 4.946612281335209e-05, "loss": 0.4042, "step": 10549 }, { "epoch": 1.09, "learning_rate": 4.9464249856554526e-05, "loss": 0.6052, "step": 10550 }, { "epoch": 1.09, "learning_rate": 4.946237676872819e-05, "loss": 0.7434, "step": 10551 }, { "epoch": 1.09, "learning_rate": 4.946050354988569e-05, "loss": 1.3929, "step": 10552 }, { "epoch": 1.09, "learning_rate": 4.9458630200039634e-05, "loss": 0.749, "step": 10553 }, { "epoch": 1.09, "learning_rate": 4.9456756719202635e-05, "loss": 0.8861, "step": 10554 }, { "epoch": 1.09, "learning_rate": 4.94548831073873e-05, "loss": 0.403, "step": 10555 }, { "epoch": 1.09, "learning_rate": 4.9453009364606234e-05, "loss": 0.5603, "step": 10556 }, { "epoch": 1.09, "learning_rate": 4.9451135490872073e-05, "loss": 0.263, "step": 10557 }, { "epoch": 1.09, "learning_rate": 4.9449261486197414e-05, "loss": 0.6285, "step": 10558 }, { "epoch": 1.09, "learning_rate": 4.944738735059488e-05, "loss": 1.0637, "step": 10559 }, { "epoch": 1.09, "learning_rate": 4.944551308407708e-05, "loss": 0.5767, "step": 10560 }, { "epoch": 1.09, "learning_rate": 4.944363868665664e-05, "loss": 0.6165, "step": 10561 }, { "epoch": 1.09, "learning_rate": 4.9441764158346175e-05, "loss": 0.9007, "step": 10562 }, { "epoch": 1.09, "learning_rate": 4.9439889499158295e-05, "loss": 0.3562, "step": 10563 }, { "epoch": 1.09, "learning_rate": 4.9438014709105634e-05, "loss": 0.7416, "step": 10564 }, { "epoch": 1.09, "learning_rate": 4.94361397882008e-05, "loss": 1.3626, "step": 10565 }, { "epoch": 1.09, "learning_rate": 4.943426473645643e-05, "loss": 0.1416, "step": 10566 }, { "epoch": 1.09, "learning_rate": 4.943238955388513e-05, "loss": 0.7166, "step": 10567 }, { "epoch": 1.09, "learning_rate": 4.9430514240499534e-05, "loss": 0.1908, "step": 10568 }, { "epoch": 1.09, "learning_rate": 4.942863879631226e-05, "loss": 0.6235, "step": 10569 }, { "epoch": 1.09, "learning_rate": 4.9426763221335934e-05, "loss": 0.29, "step": 10570 }, { "epoch": 1.09, "learning_rate": 4.942488751558319e-05, "loss": 0.8441, "step": 10571 }, { "epoch": 1.09, "learning_rate": 4.942301167906665e-05, "loss": 0.979, "step": 10572 }, { "epoch": 1.09, "learning_rate": 4.942113571179893e-05, "loss": 0.2257, "step": 10573 }, { "epoch": 1.09, "learning_rate": 4.941925961379267e-05, "loss": 0.9191, "step": 10574 }, { "epoch": 1.09, "learning_rate": 4.94173833850605e-05, "loss": 0.4234, "step": 10575 }, { "epoch": 1.09, "learning_rate": 4.9415507025615056e-05, "loss": 0.6065, "step": 10576 }, { "epoch": 1.09, "learning_rate": 4.9413630535468946e-05, "loss": 0.6944, "step": 10577 }, { "epoch": 1.09, "learning_rate": 4.941175391463483e-05, "loss": 1.1715, "step": 10578 }, { "epoch": 1.09, "learning_rate": 4.9409877163125325e-05, "loss": 0.5756, "step": 10579 }, { "epoch": 1.09, "learning_rate": 4.940800028095307e-05, "loss": 0.4304, "step": 10580 }, { "epoch": 1.09, "learning_rate": 4.9406123268130696e-05, "loss": 0.792, "step": 10581 }, { "epoch": 1.09, "learning_rate": 4.940424612467084e-05, "loss": 0.8019, "step": 10582 }, { "epoch": 1.09, "learning_rate": 4.9402368850586134e-05, "loss": 0.6765, "step": 10583 }, { "epoch": 1.09, "learning_rate": 4.9400491445889234e-05, "loss": 0.1657, "step": 10584 }, { "epoch": 1.09, "learning_rate": 4.9398613910592755e-05, "loss": 0.5834, "step": 10585 }, { "epoch": 1.09, "learning_rate": 4.9396736244709356e-05, "loss": 0.3354, "step": 10586 }, { "epoch": 1.09, "learning_rate": 4.9394858448251654e-05, "loss": 0.5024, "step": 10587 }, { "epoch": 1.09, "learning_rate": 4.9392980521232314e-05, "loss": 0.3848, "step": 10588 }, { "epoch": 1.09, "learning_rate": 4.939110246366396e-05, "loss": 0.4167, "step": 10589 }, { "epoch": 1.09, "learning_rate": 4.938922427555925e-05, "loss": 0.4132, "step": 10590 }, { "epoch": 1.09, "learning_rate": 4.9387345956930814e-05, "loss": 0.2883, "step": 10591 }, { "epoch": 1.09, "learning_rate": 4.93854675077913e-05, "loss": 0.6858, "step": 10592 }, { "epoch": 1.09, "learning_rate": 4.9383588928153353e-05, "loss": 0.7619, "step": 10593 }, { "epoch": 1.09, "learning_rate": 4.938171021802962e-05, "loss": 1.1054, "step": 10594 }, { "epoch": 1.09, "learning_rate": 4.937983137743275e-05, "loss": 0.5713, "step": 10595 }, { "epoch": 1.09, "learning_rate": 4.937795240637539e-05, "loss": 0.2448, "step": 10596 }, { "epoch": 1.09, "learning_rate": 4.937607330487019e-05, "loss": 0.5733, "step": 10597 }, { "epoch": 1.09, "learning_rate": 4.93741940729298e-05, "loss": 0.7391, "step": 10598 }, { "epoch": 1.09, "learning_rate": 4.9372314710566866e-05, "loss": 0.397, "step": 10599 }, { "epoch": 1.09, "learning_rate": 4.9370435217794044e-05, "loss": 1.2155, "step": 10600 }, { "epoch": 1.09, "learning_rate": 4.936855559462398e-05, "loss": 0.9082, "step": 10601 }, { "epoch": 1.09, "learning_rate": 4.9366675841069335e-05, "loss": 0.7899, "step": 10602 }, { "epoch": 1.09, "learning_rate": 4.9364795957142756e-05, "loss": 1.239, "step": 10603 }, { "epoch": 1.09, "learning_rate": 4.9362915942856906e-05, "loss": 0.5318, "step": 10604 }, { "epoch": 1.09, "learning_rate": 4.9361035798224435e-05, "loss": 1.3653, "step": 10605 }, { "epoch": 1.09, "learning_rate": 4.9359155523258e-05, "loss": 1.1696, "step": 10606 }, { "epoch": 1.09, "learning_rate": 4.9357275117970264e-05, "loss": 0.6313, "step": 10607 }, { "epoch": 1.09, "learning_rate": 4.935539458237387e-05, "loss": 0.8178, "step": 10608 }, { "epoch": 1.09, "learning_rate": 4.93535139164815e-05, "loss": 0.6485, "step": 10609 }, { "epoch": 1.09, "learning_rate": 4.9351633120305804e-05, "loss": 0.4929, "step": 10610 }, { "epoch": 1.09, "learning_rate": 4.9349752193859436e-05, "loss": 0.933, "step": 10611 }, { "epoch": 1.09, "learning_rate": 4.934787113715506e-05, "loss": 0.3018, "step": 10612 }, { "epoch": 1.09, "learning_rate": 4.934598995020535e-05, "loss": 0.5129, "step": 10613 }, { "epoch": 1.09, "learning_rate": 4.9344108633022954e-05, "loss": 0.4474, "step": 10614 }, { "epoch": 1.09, "learning_rate": 4.934222718562056e-05, "loss": 0.2466, "step": 10615 }, { "epoch": 1.09, "learning_rate": 4.934034560801081e-05, "loss": 0.5435, "step": 10616 }, { "epoch": 1.09, "learning_rate": 4.933846390020637e-05, "loss": 1.4749, "step": 10617 }, { "epoch": 1.09, "learning_rate": 4.9336582062219925e-05, "loss": 0.2495, "step": 10618 }, { "epoch": 1.09, "learning_rate": 4.9334700094064134e-05, "loss": 0.283, "step": 10619 }, { "epoch": 1.09, "learning_rate": 4.933281799575166e-05, "loss": 1.0687, "step": 10620 }, { "epoch": 1.09, "learning_rate": 4.9330935767295185e-05, "loss": 0.4512, "step": 10621 }, { "epoch": 1.09, "learning_rate": 4.932905340870737e-05, "loss": 0.307, "step": 10622 }, { "epoch": 1.09, "learning_rate": 4.9327170920000894e-05, "loss": 0.4451, "step": 10623 }, { "epoch": 1.09, "learning_rate": 4.932528830118843e-05, "loss": 0.6088, "step": 10624 }, { "epoch": 1.09, "learning_rate": 4.9323405552282635e-05, "loss": 1.0182, "step": 10625 }, { "epoch": 1.09, "learning_rate": 4.932152267329619e-05, "loss": 0.3176, "step": 10626 }, { "epoch": 1.09, "learning_rate": 4.931963966424179e-05, "loss": 0.4896, "step": 10627 }, { "epoch": 1.09, "learning_rate": 4.9317756525132086e-05, "loss": 0.5046, "step": 10628 }, { "epoch": 1.09, "learning_rate": 4.931587325597978e-05, "loss": 0.6369, "step": 10629 }, { "epoch": 1.09, "learning_rate": 4.931398985679752e-05, "loss": 0.2316, "step": 10630 }, { "epoch": 1.09, "learning_rate": 4.931210632759801e-05, "loss": 0.6489, "step": 10631 }, { "epoch": 1.09, "learning_rate": 4.931022266839391e-05, "loss": 0.2696, "step": 10632 }, { "epoch": 1.1, "learning_rate": 4.9308338879197906e-05, "loss": 0.2159, "step": 10633 }, { "epoch": 1.1, "learning_rate": 4.930645496002269e-05, "loss": 1.1086, "step": 10634 }, { "epoch": 1.1, "learning_rate": 4.930457091088094e-05, "loss": 0.3626, "step": 10635 }, { "epoch": 1.1, "learning_rate": 4.930268673178534e-05, "loss": 0.8948, "step": 10636 }, { "epoch": 1.1, "learning_rate": 4.930080242274855e-05, "loss": 1.7135, "step": 10637 }, { "epoch": 1.1, "learning_rate": 4.929891798378329e-05, "loss": 0.9645, "step": 10638 }, { "epoch": 1.1, "learning_rate": 4.929703341490222e-05, "loss": 0.5311, "step": 10639 }, { "epoch": 1.1, "learning_rate": 4.929514871611804e-05, "loss": 0.4911, "step": 10640 }, { "epoch": 1.1, "learning_rate": 4.929326388744344e-05, "loss": 1.1228, "step": 10641 }, { "epoch": 1.1, "learning_rate": 4.929137892889109e-05, "loss": 0.3249, "step": 10642 }, { "epoch": 1.1, "learning_rate": 4.9289493840473704e-05, "loss": 0.4093, "step": 10643 }, { "epoch": 1.1, "learning_rate": 4.928760862220394e-05, "loss": 0.7981, "step": 10644 }, { "epoch": 1.1, "learning_rate": 4.9285723274094526e-05, "loss": 0.6314, "step": 10645 }, { "epoch": 1.1, "learning_rate": 4.928383779615813e-05, "loss": 0.7817, "step": 10646 }, { "epoch": 1.1, "learning_rate": 4.9281952188407445e-05, "loss": 0.971, "step": 10647 }, { "epoch": 1.1, "learning_rate": 4.928006645085518e-05, "loss": 0.8255, "step": 10648 }, { "epoch": 1.1, "learning_rate": 4.927818058351401e-05, "loss": 0.0956, "step": 10649 }, { "epoch": 1.1, "learning_rate": 4.9276294586396645e-05, "loss": 0.6911, "step": 10650 }, { "epoch": 1.1, "learning_rate": 4.9274408459515766e-05, "loss": 0.8729, "step": 10651 }, { "epoch": 1.1, "learning_rate": 4.9272522202884076e-05, "loss": 0.3427, "step": 10652 }, { "epoch": 1.1, "learning_rate": 4.927063581651429e-05, "loss": 0.4396, "step": 10653 }, { "epoch": 1.1, "learning_rate": 4.92687493004191e-05, "loss": 0.5923, "step": 10654 }, { "epoch": 1.1, "learning_rate": 4.926686265461118e-05, "loss": 0.339, "step": 10655 }, { "epoch": 1.1, "learning_rate": 4.926497587910325e-05, "loss": 0.7956, "step": 10656 }, { "epoch": 1.1, "learning_rate": 4.926308897390801e-05, "loss": 0.2062, "step": 10657 }, { "epoch": 1.1, "learning_rate": 4.926120193903817e-05, "loss": 0.6452, "step": 10658 }, { "epoch": 1.1, "learning_rate": 4.925931477450642e-05, "loss": 0.2266, "step": 10659 }, { "epoch": 1.1, "learning_rate": 4.925742748032548e-05, "loss": 0.4617, "step": 10660 }, { "epoch": 1.1, "learning_rate": 4.925554005650803e-05, "loss": 0.7429, "step": 10661 }, { "epoch": 1.1, "learning_rate": 4.92536525030668e-05, "loss": 0.1575, "step": 10662 }, { "epoch": 1.1, "learning_rate": 4.9251764820014476e-05, "loss": 0.6685, "step": 10663 }, { "epoch": 1.1, "learning_rate": 4.924987700736378e-05, "loss": 0.4193, "step": 10664 }, { "epoch": 1.1, "learning_rate": 4.924798906512741e-05, "loss": 0.5169, "step": 10665 }, { "epoch": 1.1, "learning_rate": 4.924610099331809e-05, "loss": 0.8994, "step": 10666 }, { "epoch": 1.1, "learning_rate": 4.924421279194853e-05, "loss": 0.9311, "step": 10667 }, { "epoch": 1.1, "learning_rate": 4.9242324461031414e-05, "loss": 0.773, "step": 10668 }, { "epoch": 1.1, "learning_rate": 4.924043600057948e-05, "loss": 0.5241, "step": 10669 }, { "epoch": 1.1, "learning_rate": 4.923854741060543e-05, "loss": 1.0181, "step": 10670 }, { "epoch": 1.1, "learning_rate": 4.923665869112198e-05, "loss": 0.4106, "step": 10671 }, { "epoch": 1.1, "learning_rate": 4.923476984214185e-05, "loss": 0.8038, "step": 10672 }, { "epoch": 1.1, "learning_rate": 4.9232880863677754e-05, "loss": 0.3706, "step": 10673 }, { "epoch": 1.1, "learning_rate": 4.92309917557424e-05, "loss": 0.9764, "step": 10674 }, { "epoch": 1.1, "learning_rate": 4.92291025183485e-05, "loss": 0.099, "step": 10675 }, { "epoch": 1.1, "learning_rate": 4.922721315150879e-05, "loss": 1.21, "step": 10676 }, { "epoch": 1.1, "learning_rate": 4.922532365523598e-05, "loss": 0.4029, "step": 10677 }, { "epoch": 1.1, "learning_rate": 4.922343402954278e-05, "loss": 0.8675, "step": 10678 }, { "epoch": 1.1, "learning_rate": 4.922154427444194e-05, "loss": 0.3243, "step": 10679 }, { "epoch": 1.1, "learning_rate": 4.9219654389946145e-05, "loss": 0.3829, "step": 10680 }, { "epoch": 1.1, "learning_rate": 4.921776437606813e-05, "loss": 1.0498, "step": 10681 }, { "epoch": 1.1, "learning_rate": 4.921587423282063e-05, "loss": 0.3876, "step": 10682 }, { "epoch": 1.1, "learning_rate": 4.921398396021636e-05, "loss": 0.5074, "step": 10683 }, { "epoch": 1.1, "learning_rate": 4.921209355826805e-05, "loss": 0.8367, "step": 10684 }, { "epoch": 1.1, "learning_rate": 4.921020302698842e-05, "loss": 0.4773, "step": 10685 }, { "epoch": 1.1, "learning_rate": 4.92083123663902e-05, "loss": 0.2893, "step": 10686 }, { "epoch": 1.1, "learning_rate": 4.9206421576486105e-05, "loss": 0.735, "step": 10687 }, { "epoch": 1.1, "learning_rate": 4.9204530657288875e-05, "loss": 0.3041, "step": 10688 }, { "epoch": 1.1, "learning_rate": 4.920263960881125e-05, "loss": 0.2579, "step": 10689 }, { "epoch": 1.1, "learning_rate": 4.920074843106593e-05, "loss": 0.127, "step": 10690 }, { "epoch": 1.1, "learning_rate": 4.9198857124065685e-05, "loss": 0.2009, "step": 10691 }, { "epoch": 1.1, "learning_rate": 4.919696568782322e-05, "loss": 0.8381, "step": 10692 }, { "epoch": 1.1, "learning_rate": 4.919507412235127e-05, "loss": 0.4605, "step": 10693 }, { "epoch": 1.1, "learning_rate": 4.919318242766257e-05, "loss": 0.4852, "step": 10694 }, { "epoch": 1.1, "learning_rate": 4.919129060376985e-05, "loss": 0.3582, "step": 10695 }, { "epoch": 1.1, "learning_rate": 4.918939865068586e-05, "loss": 0.4942, "step": 10696 }, { "epoch": 1.1, "learning_rate": 4.918750656842333e-05, "loss": 0.7398, "step": 10697 }, { "epoch": 1.1, "learning_rate": 4.9185614356995e-05, "loss": 0.7179, "step": 10698 }, { "epoch": 1.1, "learning_rate": 4.91837220164136e-05, "loss": 0.1476, "step": 10699 }, { "epoch": 1.1, "learning_rate": 4.918182954669187e-05, "loss": 1.1242, "step": 10700 }, { "epoch": 1.1, "learning_rate": 4.917993694784255e-05, "loss": 0.4531, "step": 10701 }, { "epoch": 1.1, "learning_rate": 4.917804421987838e-05, "loss": 0.8312, "step": 10702 }, { "epoch": 1.1, "learning_rate": 4.9176151362812114e-05, "loss": 0.6452, "step": 10703 }, { "epoch": 1.1, "learning_rate": 4.9174258376656476e-05, "loss": 0.9637, "step": 10704 }, { "epoch": 1.1, "learning_rate": 4.917236526142422e-05, "loss": 0.425, "step": 10705 }, { "epoch": 1.1, "learning_rate": 4.917047201712809e-05, "loss": 0.7229, "step": 10706 }, { "epoch": 1.1, "learning_rate": 4.916857864378082e-05, "loss": 0.4189, "step": 10707 }, { "epoch": 1.1, "learning_rate": 4.916668514139516e-05, "loss": 0.7534, "step": 10708 }, { "epoch": 1.1, "learning_rate": 4.916479150998387e-05, "loss": 0.3574, "step": 10709 }, { "epoch": 1.1, "learning_rate": 4.916289774955969e-05, "loss": 0.4626, "step": 10710 }, { "epoch": 1.1, "learning_rate": 4.916100386013536e-05, "loss": 0.4389, "step": 10711 }, { "epoch": 1.1, "learning_rate": 4.9159109841723636e-05, "loss": 0.4805, "step": 10712 }, { "epoch": 1.1, "learning_rate": 4.9157215694337266e-05, "loss": 0.8033, "step": 10713 }, { "epoch": 1.1, "learning_rate": 4.9155321417989e-05, "loss": 0.3364, "step": 10714 }, { "epoch": 1.1, "learning_rate": 4.91534270126916e-05, "loss": 0.837, "step": 10715 }, { "epoch": 1.1, "learning_rate": 4.91515324784578e-05, "loss": 0.3538, "step": 10716 }, { "epoch": 1.1, "learning_rate": 4.914963781530038e-05, "loss": 1.2527, "step": 10717 }, { "epoch": 1.1, "learning_rate": 4.9147743023232075e-05, "loss": 1.0377, "step": 10718 }, { "epoch": 1.1, "learning_rate": 4.914584810226563e-05, "loss": 1.0829, "step": 10719 }, { "epoch": 1.1, "learning_rate": 4.914395305241383e-05, "loss": 0.9946, "step": 10720 }, { "epoch": 1.1, "learning_rate": 4.9142057873689414e-05, "loss": 1.4267, "step": 10721 }, { "epoch": 1.1, "learning_rate": 4.9140162566105136e-05, "loss": 0.7855, "step": 10722 }, { "epoch": 1.1, "learning_rate": 4.913826712967378e-05, "loss": 0.9727, "step": 10723 }, { "epoch": 1.1, "learning_rate": 4.9136371564408065e-05, "loss": 0.9454, "step": 10724 }, { "epoch": 1.1, "learning_rate": 4.913447587032079e-05, "loss": 0.6431, "step": 10725 }, { "epoch": 1.1, "learning_rate": 4.9132580047424693e-05, "loss": 0.9976, "step": 10726 }, { "epoch": 1.1, "learning_rate": 4.913068409573255e-05, "loss": 0.5527, "step": 10727 }, { "epoch": 1.1, "learning_rate": 4.9128788015257107e-05, "loss": 0.7437, "step": 10728 }, { "epoch": 1.11, "learning_rate": 4.912689180601115e-05, "loss": 0.6641, "step": 10729 }, { "epoch": 1.11, "learning_rate": 4.912499546800742e-05, "loss": 0.8535, "step": 10730 }, { "epoch": 1.11, "learning_rate": 4.912309900125871e-05, "loss": 0.8207, "step": 10731 }, { "epoch": 1.11, "learning_rate": 4.912120240577777e-05, "loss": 0.6679, "step": 10732 }, { "epoch": 1.11, "learning_rate": 4.911930568157736e-05, "loss": 0.7927, "step": 10733 }, { "epoch": 1.11, "learning_rate": 4.911740882867026e-05, "loss": 0.8919, "step": 10734 }, { "epoch": 1.11, "learning_rate": 4.9115511847069245e-05, "loss": 0.9575, "step": 10735 }, { "epoch": 1.11, "learning_rate": 4.911361473678708e-05, "loss": 0.384, "step": 10736 }, { "epoch": 1.11, "learning_rate": 4.911171749783652e-05, "loss": 0.2776, "step": 10737 }, { "epoch": 1.11, "learning_rate": 4.910982013023035e-05, "loss": 0.3841, "step": 10738 }, { "epoch": 1.11, "learning_rate": 4.910792263398135e-05, "loss": 0.5185, "step": 10739 }, { "epoch": 1.11, "learning_rate": 4.910602500910228e-05, "loss": 0.5759, "step": 10740 }, { "epoch": 1.11, "learning_rate": 4.9104127255605924e-05, "loss": 0.9647, "step": 10741 }, { "epoch": 1.11, "learning_rate": 4.910222937350505e-05, "loss": 0.4016, "step": 10742 }, { "epoch": 1.11, "learning_rate": 4.910033136281245e-05, "loss": 0.2733, "step": 10743 }, { "epoch": 1.11, "learning_rate": 4.909843322354087e-05, "loss": 0.6054, "step": 10744 }, { "epoch": 1.11, "learning_rate": 4.909653495570311e-05, "loss": 0.6338, "step": 10745 }, { "epoch": 1.11, "learning_rate": 4.909463655931194e-05, "loss": 0.4997, "step": 10746 }, { "epoch": 1.11, "learning_rate": 4.909273803438016e-05, "loss": 0.8256, "step": 10747 }, { "epoch": 1.11, "learning_rate": 4.9090839380920536e-05, "loss": 0.6774, "step": 10748 }, { "epoch": 1.11, "learning_rate": 4.908894059894583e-05, "loss": 0.1733, "step": 10749 }, { "epoch": 1.11, "learning_rate": 4.908704168846886e-05, "loss": 0.6789, "step": 10750 }, { "epoch": 1.11, "learning_rate": 4.908514264950238e-05, "loss": 1.008, "step": 10751 }, { "epoch": 1.11, "learning_rate": 4.908324348205919e-05, "loss": 0.1862, "step": 10752 }, { "epoch": 1.11, "learning_rate": 4.908134418615207e-05, "loss": 0.3574, "step": 10753 }, { "epoch": 1.11, "learning_rate": 4.90794447617938e-05, "loss": 0.2892, "step": 10754 }, { "epoch": 1.11, "learning_rate": 4.907754520899718e-05, "loss": 0.7938, "step": 10755 }, { "epoch": 1.11, "learning_rate": 4.907564552777499e-05, "loss": 0.9414, "step": 10756 }, { "epoch": 1.11, "learning_rate": 4.907374571814001e-05, "loss": 0.5588, "step": 10757 }, { "epoch": 1.11, "learning_rate": 4.907184578010504e-05, "loss": 0.2283, "step": 10758 }, { "epoch": 1.11, "learning_rate": 4.906994571368287e-05, "loss": 0.5398, "step": 10759 }, { "epoch": 1.11, "learning_rate": 4.906804551888628e-05, "loss": 0.7974, "step": 10760 }, { "epoch": 1.11, "learning_rate": 4.906614519572808e-05, "loss": 1.1798, "step": 10761 }, { "epoch": 1.11, "learning_rate": 4.906424474422104e-05, "loss": 0.7731, "step": 10762 }, { "epoch": 1.11, "learning_rate": 4.906234416437798e-05, "loss": 0.5589, "step": 10763 }, { "epoch": 1.11, "learning_rate": 4.906044345621167e-05, "loss": 0.2017, "step": 10764 }, { "epoch": 1.11, "learning_rate": 4.9058542619734915e-05, "loss": 1.3822, "step": 10765 }, { "epoch": 1.11, "learning_rate": 4.905664165496051e-05, "loss": 0.336, "step": 10766 }, { "epoch": 1.11, "learning_rate": 4.905474056190125e-05, "loss": 0.5141, "step": 10767 }, { "epoch": 1.11, "learning_rate": 4.905283934056994e-05, "loss": 0.3174, "step": 10768 }, { "epoch": 1.11, "learning_rate": 4.905093799097936e-05, "loss": 0.8626, "step": 10769 }, { "epoch": 1.11, "learning_rate": 4.904903651314234e-05, "loss": 0.4548, "step": 10770 }, { "epoch": 1.11, "learning_rate": 4.904713490707165e-05, "loss": 0.3337, "step": 10771 }, { "epoch": 1.11, "learning_rate": 4.904523317278011e-05, "loss": 0.7951, "step": 10772 }, { "epoch": 1.11, "learning_rate": 4.9043331310280525e-05, "loss": 1.363, "step": 10773 }, { "epoch": 1.11, "learning_rate": 4.904142931958569e-05, "loss": 0.9732, "step": 10774 }, { "epoch": 1.11, "learning_rate": 4.903952720070839e-05, "loss": 0.5724, "step": 10775 }, { "epoch": 1.11, "learning_rate": 4.903762495366146e-05, "loss": 0.7249, "step": 10776 }, { "epoch": 1.11, "learning_rate": 4.9035722578457694e-05, "loss": 0.3432, "step": 10777 }, { "epoch": 1.11, "learning_rate": 4.903382007510989e-05, "loss": 0.3522, "step": 10778 }, { "epoch": 1.11, "learning_rate": 4.903191744363087e-05, "loss": 0.0997, "step": 10779 }, { "epoch": 1.11, "learning_rate": 4.903001468403344e-05, "loss": 0.3561, "step": 10780 }, { "epoch": 1.11, "learning_rate": 4.902811179633039e-05, "loss": 0.535, "step": 10781 }, { "epoch": 1.11, "learning_rate": 4.902620878053455e-05, "loss": 0.3909, "step": 10782 }, { "epoch": 1.11, "learning_rate": 4.9024305636658726e-05, "loss": 0.7554, "step": 10783 }, { "epoch": 1.11, "learning_rate": 4.902240236471572e-05, "loss": 0.2764, "step": 10784 }, { "epoch": 1.11, "learning_rate": 4.902049896471836e-05, "loss": 0.2205, "step": 10785 }, { "epoch": 1.11, "learning_rate": 4.901859543667945e-05, "loss": 0.7586, "step": 10786 }, { "epoch": 1.11, "learning_rate": 4.9016691780611805e-05, "loss": 0.1925, "step": 10787 }, { "epoch": 1.11, "learning_rate": 4.901478799652824e-05, "loss": 0.5917, "step": 10788 }, { "epoch": 1.11, "learning_rate": 4.901288408444157e-05, "loss": 1.1602, "step": 10789 }, { "epoch": 1.11, "learning_rate": 4.901098004436462e-05, "loss": 0.8559, "step": 10790 }, { "epoch": 1.11, "learning_rate": 4.9009075876310204e-05, "loss": 0.7361, "step": 10791 }, { "epoch": 1.11, "learning_rate": 4.9007171580291127e-05, "loss": 0.9393, "step": 10792 }, { "epoch": 1.11, "learning_rate": 4.900526715632022e-05, "loss": 1.025, "step": 10793 }, { "epoch": 1.11, "learning_rate": 4.9003362604410295e-05, "loss": 0.1444, "step": 10794 }, { "epoch": 1.11, "learning_rate": 4.900145792457419e-05, "loss": 1.0073, "step": 10795 }, { "epoch": 1.11, "learning_rate": 4.899955311682471e-05, "loss": 0.5844, "step": 10796 }, { "epoch": 1.11, "learning_rate": 4.899764818117469e-05, "loss": 0.0582, "step": 10797 }, { "epoch": 1.11, "learning_rate": 4.899574311763694e-05, "loss": 0.4959, "step": 10798 }, { "epoch": 1.11, "learning_rate": 4.899383792622429e-05, "loss": 0.5669, "step": 10799 }, { "epoch": 1.11, "learning_rate": 4.899193260694958e-05, "loss": 0.4949, "step": 10800 }, { "epoch": 1.11, "learning_rate": 4.8990027159825614e-05, "loss": 0.3859, "step": 10801 }, { "epoch": 1.11, "learning_rate": 4.898812158486523e-05, "loss": 0.805, "step": 10802 }, { "epoch": 1.11, "learning_rate": 4.8986215882081256e-05, "loss": 0.4746, "step": 10803 }, { "epoch": 1.11, "learning_rate": 4.898431005148652e-05, "loss": 0.5238, "step": 10804 }, { "epoch": 1.11, "learning_rate": 4.8982404093093845e-05, "loss": 0.2226, "step": 10805 }, { "epoch": 1.11, "learning_rate": 4.8980498006916076e-05, "loss": 0.78, "step": 10806 }, { "epoch": 1.11, "learning_rate": 4.897859179296603e-05, "loss": 0.4872, "step": 10807 }, { "epoch": 1.11, "learning_rate": 4.8976685451256546e-05, "loss": 0.4688, "step": 10808 }, { "epoch": 1.11, "learning_rate": 4.897477898180045e-05, "loss": 0.2943, "step": 10809 }, { "epoch": 1.11, "learning_rate": 4.897287238461059e-05, "loss": 1.2704, "step": 10810 }, { "epoch": 1.11, "learning_rate": 4.897096565969979e-05, "loss": 0.9783, "step": 10811 }, { "epoch": 1.11, "learning_rate": 4.896905880708088e-05, "loss": 0.8187, "step": 10812 }, { "epoch": 1.11, "learning_rate": 4.8967151826766715e-05, "loss": 0.3279, "step": 10813 }, { "epoch": 1.11, "learning_rate": 4.896524471877012e-05, "loss": 1.5176, "step": 10814 }, { "epoch": 1.11, "learning_rate": 4.896333748310393e-05, "loss": 0.4559, "step": 10815 }, { "epoch": 1.11, "learning_rate": 4.896143011978099e-05, "loss": 0.4577, "step": 10816 }, { "epoch": 1.11, "learning_rate": 4.8959522628814144e-05, "loss": 1.1013, "step": 10817 }, { "epoch": 1.11, "learning_rate": 4.8957615010216235e-05, "loss": 0.3583, "step": 10818 }, { "epoch": 1.11, "learning_rate": 4.895570726400009e-05, "loss": 0.6722, "step": 10819 }, { "epoch": 1.11, "learning_rate": 4.895379939017855e-05, "loss": 0.759, "step": 10820 }, { "epoch": 1.11, "learning_rate": 4.895189138876447e-05, "loss": 0.7769, "step": 10821 }, { "epoch": 1.11, "learning_rate": 4.89499832597707e-05, "loss": 0.4258, "step": 10822 }, { "epoch": 1.11, "learning_rate": 4.894807500321007e-05, "loss": 1.167, "step": 10823 }, { "epoch": 1.11, "learning_rate": 4.8946166619095435e-05, "loss": 0.607, "step": 10824 }, { "epoch": 1.12, "learning_rate": 4.894425810743964e-05, "loss": 0.6981, "step": 10825 }, { "epoch": 1.12, "learning_rate": 4.894234946825553e-05, "loss": 0.3944, "step": 10826 }, { "epoch": 1.12, "learning_rate": 4.894044070155596e-05, "loss": 0.545, "step": 10827 }, { "epoch": 1.12, "learning_rate": 4.893853180735377e-05, "loss": 0.4455, "step": 10828 }, { "epoch": 1.12, "learning_rate": 4.8936622785661814e-05, "loss": 0.2432, "step": 10829 }, { "epoch": 1.12, "learning_rate": 4.8934713636492954e-05, "loss": 0.9609, "step": 10830 }, { "epoch": 1.12, "learning_rate": 4.893280435986002e-05, "loss": 0.5092, "step": 10831 }, { "epoch": 1.12, "learning_rate": 4.8930894955775884e-05, "loss": 0.105, "step": 10832 }, { "epoch": 1.12, "learning_rate": 4.892898542425339e-05, "loss": 0.2362, "step": 10833 }, { "epoch": 1.12, "learning_rate": 4.89270757653054e-05, "loss": 1.0765, "step": 10834 }, { "epoch": 1.12, "learning_rate": 4.892516597894476e-05, "loss": 0.563, "step": 10835 }, { "epoch": 1.12, "learning_rate": 4.892325606518433e-05, "loss": 0.4695, "step": 10836 }, { "epoch": 1.12, "learning_rate": 4.8921346024036973e-05, "loss": 0.2279, "step": 10837 }, { "epoch": 1.12, "learning_rate": 4.8919435855515535e-05, "loss": 0.5968, "step": 10838 }, { "epoch": 1.12, "learning_rate": 4.891752555963288e-05, "loss": 0.865, "step": 10839 }, { "epoch": 1.12, "learning_rate": 4.8915615136401873e-05, "loss": 0.5718, "step": 10840 }, { "epoch": 1.12, "learning_rate": 4.891370458583537e-05, "loss": 0.4026, "step": 10841 }, { "epoch": 1.12, "learning_rate": 4.891179390794624e-05, "loss": 1.0131, "step": 10842 }, { "epoch": 1.12, "learning_rate": 4.890988310274733e-05, "loss": 0.2935, "step": 10843 }, { "epoch": 1.12, "learning_rate": 4.890797217025152e-05, "loss": 0.6364, "step": 10844 }, { "epoch": 1.12, "learning_rate": 4.890606111047166e-05, "loss": 1.3774, "step": 10845 }, { "epoch": 1.12, "learning_rate": 4.890414992342062e-05, "loss": 0.3262, "step": 10846 }, { "epoch": 1.12, "learning_rate": 4.890223860911127e-05, "loss": 0.458, "step": 10847 }, { "epoch": 1.12, "learning_rate": 4.890032716755646e-05, "loss": 0.7138, "step": 10848 }, { "epoch": 1.12, "learning_rate": 4.889841559876909e-05, "loss": 0.9265, "step": 10849 }, { "epoch": 1.12, "learning_rate": 4.889650390276199e-05, "loss": 0.7207, "step": 10850 }, { "epoch": 1.12, "learning_rate": 4.889459207954806e-05, "loss": 0.7921, "step": 10851 }, { "epoch": 1.12, "learning_rate": 4.8892680129140146e-05, "loss": 0.6488, "step": 10852 }, { "epoch": 1.12, "learning_rate": 4.889076805155113e-05, "loss": 0.1094, "step": 10853 }, { "epoch": 1.12, "learning_rate": 4.8888855846793897e-05, "loss": 0.973, "step": 10854 }, { "epoch": 1.12, "learning_rate": 4.8886943514881295e-05, "loss": 0.831, "step": 10855 }, { "epoch": 1.12, "learning_rate": 4.888503105582621e-05, "loss": 0.4836, "step": 10856 }, { "epoch": 1.12, "learning_rate": 4.888311846964152e-05, "loss": 0.2533, "step": 10857 }, { "epoch": 1.12, "learning_rate": 4.8881205756340095e-05, "loss": 0.4793, "step": 10858 }, { "epoch": 1.12, "learning_rate": 4.88792929159348e-05, "loss": 0.9442, "step": 10859 }, { "epoch": 1.12, "learning_rate": 4.887737994843853e-05, "loss": 0.6574, "step": 10860 }, { "epoch": 1.12, "learning_rate": 4.8875466853864156e-05, "loss": 0.8076, "step": 10861 }, { "epoch": 1.12, "learning_rate": 4.8873553632224556e-05, "loss": 0.4318, "step": 10862 }, { "epoch": 1.12, "learning_rate": 4.88716402835326e-05, "loss": 0.4594, "step": 10863 }, { "epoch": 1.12, "learning_rate": 4.886972680780118e-05, "loss": 0.8167, "step": 10864 }, { "epoch": 1.12, "learning_rate": 4.886781320504318e-05, "loss": 0.3515, "step": 10865 }, { "epoch": 1.12, "learning_rate": 4.8865899475271474e-05, "loss": 1.0982, "step": 10866 }, { "epoch": 1.12, "learning_rate": 4.886398561849895e-05, "loss": 0.4178, "step": 10867 }, { "epoch": 1.12, "learning_rate": 4.886207163473849e-05, "loss": 0.2831, "step": 10868 }, { "epoch": 1.12, "learning_rate": 4.886015752400297e-05, "loss": 1.0585, "step": 10869 }, { "epoch": 1.12, "learning_rate": 4.885824328630529e-05, "loss": 0.8327, "step": 10870 }, { "epoch": 1.12, "learning_rate": 4.885632892165832e-05, "loss": 0.2905, "step": 10871 }, { "epoch": 1.12, "learning_rate": 4.885441443007496e-05, "loss": 0.5315, "step": 10872 }, { "epoch": 1.12, "learning_rate": 4.885249981156809e-05, "loss": 0.4299, "step": 10873 }, { "epoch": 1.12, "learning_rate": 4.885058506615061e-05, "loss": 1.2056, "step": 10874 }, { "epoch": 1.12, "learning_rate": 4.8848670193835396e-05, "loss": 0.9022, "step": 10875 }, { "epoch": 1.12, "learning_rate": 4.884675519463535e-05, "loss": 0.2982, "step": 10876 }, { "epoch": 1.12, "learning_rate": 4.884484006856336e-05, "loss": 0.6831, "step": 10877 }, { "epoch": 1.12, "learning_rate": 4.88429248156323e-05, "loss": 0.6564, "step": 10878 }, { "epoch": 1.12, "learning_rate": 4.8841009435855094e-05, "loss": 0.5955, "step": 10879 }, { "epoch": 1.12, "learning_rate": 4.883909392924462e-05, "loss": 0.2628, "step": 10880 }, { "epoch": 1.12, "learning_rate": 4.883717829581376e-05, "loss": 0.9454, "step": 10881 }, { "epoch": 1.12, "learning_rate": 4.8835262535575435e-05, "loss": 0.8271, "step": 10882 }, { "epoch": 1.12, "learning_rate": 4.883334664854253e-05, "loss": 0.3038, "step": 10883 }, { "epoch": 1.12, "learning_rate": 4.8831430634727944e-05, "loss": 0.2456, "step": 10884 }, { "epoch": 1.12, "learning_rate": 4.882951449414457e-05, "loss": 0.6716, "step": 10885 }, { "epoch": 1.12, "learning_rate": 4.8827598226805304e-05, "loss": 0.7929, "step": 10886 }, { "epoch": 1.12, "learning_rate": 4.882568183272306e-05, "loss": 0.3922, "step": 10887 }, { "epoch": 1.12, "learning_rate": 4.8823765311910724e-05, "loss": 0.2891, "step": 10888 }, { "epoch": 1.12, "learning_rate": 4.882184866438121e-05, "loss": 0.9161, "step": 10889 }, { "epoch": 1.12, "learning_rate": 4.881993189014741e-05, "loss": 0.2659, "step": 10890 }, { "epoch": 1.12, "learning_rate": 4.881801498922224e-05, "loss": 0.3131, "step": 10891 }, { "epoch": 1.12, "learning_rate": 4.881609796161858e-05, "loss": 0.6148, "step": 10892 }, { "epoch": 1.12, "learning_rate": 4.8814180807349375e-05, "loss": 0.2803, "step": 10893 }, { "epoch": 1.12, "learning_rate": 4.881226352642749e-05, "loss": 0.6898, "step": 10894 }, { "epoch": 1.12, "learning_rate": 4.8810346118865856e-05, "loss": 0.7544, "step": 10895 }, { "epoch": 1.12, "learning_rate": 4.880842858467737e-05, "loss": 1.0215, "step": 10896 }, { "epoch": 1.12, "learning_rate": 4.8806510923874944e-05, "loss": 0.3422, "step": 10897 }, { "epoch": 1.12, "learning_rate": 4.8804593136471486e-05, "loss": 0.4309, "step": 10898 }, { "epoch": 1.12, "learning_rate": 4.880267522247991e-05, "loss": 0.9743, "step": 10899 }, { "epoch": 1.12, "learning_rate": 4.8800757181913124e-05, "loss": 0.9589, "step": 10900 }, { "epoch": 1.12, "learning_rate": 4.879883901478403e-05, "loss": 1.1365, "step": 10901 }, { "epoch": 1.12, "learning_rate": 4.879692072110556e-05, "loss": 0.7902, "step": 10902 }, { "epoch": 1.12, "learning_rate": 4.8795002300890614e-05, "loss": 0.2571, "step": 10903 }, { "epoch": 1.12, "learning_rate": 4.879308375415211e-05, "loss": 1.6171, "step": 10904 }, { "epoch": 1.12, "learning_rate": 4.879116508090297e-05, "loss": 0.8569, "step": 10905 }, { "epoch": 1.12, "learning_rate": 4.8789246281156096e-05, "loss": 1.0695, "step": 10906 }, { "epoch": 1.12, "learning_rate": 4.878732735492441e-05, "loss": 0.6902, "step": 10907 }, { "epoch": 1.12, "learning_rate": 4.8785408302220845e-05, "loss": 1.0336, "step": 10908 }, { "epoch": 1.12, "learning_rate": 4.87834891230583e-05, "loss": 0.7513, "step": 10909 }, { "epoch": 1.12, "learning_rate": 4.8781569817449694e-05, "loss": 0.9421, "step": 10910 }, { "epoch": 1.12, "learning_rate": 4.877965038540796e-05, "loss": 0.2301, "step": 10911 }, { "epoch": 1.12, "learning_rate": 4.8777730826946016e-05, "loss": 0.181, "step": 10912 }, { "epoch": 1.12, "learning_rate": 4.877581114207678e-05, "loss": 1.2683, "step": 10913 }, { "epoch": 1.12, "learning_rate": 4.8773891330813176e-05, "loss": 0.7452, "step": 10914 }, { "epoch": 1.12, "learning_rate": 4.877197139316814e-05, "loss": 1.2895, "step": 10915 }, { "epoch": 1.12, "learning_rate": 4.8770051329154566e-05, "loss": 1.1806, "step": 10916 }, { "epoch": 1.12, "learning_rate": 4.87681311387854e-05, "loss": 0.769, "step": 10917 }, { "epoch": 1.12, "learning_rate": 4.876621082207358e-05, "loss": 0.663, "step": 10918 }, { "epoch": 1.12, "learning_rate": 4.876429037903201e-05, "loss": 1.1041, "step": 10919 }, { "epoch": 1.13, "learning_rate": 4.876236980967363e-05, "loss": 0.4152, "step": 10920 }, { "epoch": 1.13, "learning_rate": 4.876044911401137e-05, "loss": 1.0852, "step": 10921 }, { "epoch": 1.13, "learning_rate": 4.875852829205815e-05, "loss": 0.7837, "step": 10922 }, { "epoch": 1.13, "learning_rate": 4.87566073438269e-05, "loss": 0.7402, "step": 10923 }, { "epoch": 1.13, "learning_rate": 4.8754686269330576e-05, "loss": 0.7254, "step": 10924 }, { "epoch": 1.13, "learning_rate": 4.875276506858208e-05, "loss": 1.0954, "step": 10925 }, { "epoch": 1.13, "learning_rate": 4.8750843741594366e-05, "loss": 0.8889, "step": 10926 }, { "epoch": 1.13, "learning_rate": 4.8748922288380346e-05, "loss": 0.2229, "step": 10927 }, { "epoch": 1.13, "learning_rate": 4.8747000708952976e-05, "loss": 0.6694, "step": 10928 }, { "epoch": 1.13, "learning_rate": 4.8745079003325185e-05, "loss": 1.2039, "step": 10929 }, { "epoch": 1.13, "learning_rate": 4.874315717150991e-05, "loss": 0.5828, "step": 10930 }, { "epoch": 1.13, "learning_rate": 4.874123521352008e-05, "loss": 0.5207, "step": 10931 }, { "epoch": 1.13, "learning_rate": 4.873931312936864e-05, "loss": 0.8301, "step": 10932 }, { "epoch": 1.13, "learning_rate": 4.873739091906854e-05, "loss": 0.3035, "step": 10933 }, { "epoch": 1.13, "learning_rate": 4.87354685826327e-05, "loss": 0.6099, "step": 10934 }, { "epoch": 1.13, "learning_rate": 4.8733546120074065e-05, "loss": 0.2, "step": 10935 }, { "epoch": 1.13, "learning_rate": 4.873162353140558e-05, "loss": 0.644, "step": 10936 }, { "epoch": 1.13, "learning_rate": 4.87297008166402e-05, "loss": 1.1602, "step": 10937 }, { "epoch": 1.13, "learning_rate": 4.8727777975790854e-05, "loss": 0.5207, "step": 10938 }, { "epoch": 1.13, "learning_rate": 4.8725855008870486e-05, "loss": 0.7759, "step": 10939 }, { "epoch": 1.13, "learning_rate": 4.872393191589204e-05, "loss": 0.6962, "step": 10940 }, { "epoch": 1.13, "learning_rate": 4.872200869686847e-05, "loss": 0.2367, "step": 10941 }, { "epoch": 1.13, "learning_rate": 4.872008535181271e-05, "loss": 1.0707, "step": 10942 }, { "epoch": 1.13, "learning_rate": 4.871816188073773e-05, "loss": 0.6284, "step": 10943 }, { "epoch": 1.13, "learning_rate": 4.871623828365645e-05, "loss": 1.0837, "step": 10944 }, { "epoch": 1.13, "learning_rate": 4.871431456058184e-05, "loss": 0.9842, "step": 10945 }, { "epoch": 1.13, "learning_rate": 4.871239071152684e-05, "loss": 0.8861, "step": 10946 }, { "epoch": 1.13, "learning_rate": 4.8710466736504405e-05, "loss": 0.9368, "step": 10947 }, { "epoch": 1.13, "learning_rate": 4.870854263552748e-05, "loss": 0.3342, "step": 10948 }, { "epoch": 1.13, "learning_rate": 4.8706618408609034e-05, "loss": 0.2235, "step": 10949 }, { "epoch": 1.13, "learning_rate": 4.870469405576201e-05, "loss": 0.6445, "step": 10950 }, { "epoch": 1.13, "learning_rate": 4.870276957699936e-05, "loss": 0.9075, "step": 10951 }, { "epoch": 1.13, "learning_rate": 4.8700844972334036e-05, "loss": 0.5643, "step": 10952 }, { "epoch": 1.13, "learning_rate": 4.8698920241779e-05, "loss": 0.7705, "step": 10953 }, { "epoch": 1.13, "learning_rate": 4.869699538534721e-05, "loss": 0.3181, "step": 10954 }, { "epoch": 1.13, "learning_rate": 4.869507040305163e-05, "loss": 1.0996, "step": 10955 }, { "epoch": 1.13, "learning_rate": 4.86931452949052e-05, "loss": 0.2491, "step": 10956 }, { "epoch": 1.13, "learning_rate": 4.869122006092089e-05, "loss": 0.3828, "step": 10957 }, { "epoch": 1.13, "learning_rate": 4.868929470111166e-05, "loss": 0.13, "step": 10958 }, { "epoch": 1.13, "learning_rate": 4.868736921549047e-05, "loss": 0.8595, "step": 10959 }, { "epoch": 1.13, "learning_rate": 4.868544360407029e-05, "loss": 0.7269, "step": 10960 }, { "epoch": 1.13, "learning_rate": 4.868351786686407e-05, "loss": 0.5565, "step": 10961 }, { "epoch": 1.13, "learning_rate": 4.868159200388478e-05, "loss": 0.2, "step": 10962 }, { "epoch": 1.13, "learning_rate": 4.86796660151454e-05, "loss": 0.6922, "step": 10963 }, { "epoch": 1.13, "learning_rate": 4.867773990065885e-05, "loss": 0.6184, "step": 10964 }, { "epoch": 1.13, "learning_rate": 4.867581366043815e-05, "loss": 1.2795, "step": 10965 }, { "epoch": 1.13, "learning_rate": 4.8673887294496225e-05, "loss": 0.7196, "step": 10966 }, { "epoch": 1.13, "learning_rate": 4.8671960802846066e-05, "loss": 0.821, "step": 10967 }, { "epoch": 1.13, "learning_rate": 4.867003418550064e-05, "loss": 0.2687, "step": 10968 }, { "epoch": 1.13, "learning_rate": 4.866810744247291e-05, "loss": 1.001, "step": 10969 }, { "epoch": 1.13, "learning_rate": 4.8666180573775846e-05, "loss": 0.3218, "step": 10970 }, { "epoch": 1.13, "learning_rate": 4.866425357942242e-05, "loss": 0.5629, "step": 10971 }, { "epoch": 1.13, "learning_rate": 4.866232645942561e-05, "loss": 0.5063, "step": 10972 }, { "epoch": 1.13, "learning_rate": 4.866039921379838e-05, "loss": 0.4406, "step": 10973 }, { "epoch": 1.13, "learning_rate": 4.865847184255372e-05, "loss": 0.5144, "step": 10974 }, { "epoch": 1.13, "learning_rate": 4.865654434570459e-05, "loss": 0.2677, "step": 10975 }, { "epoch": 1.13, "learning_rate": 4.865461672326396e-05, "loss": 0.2545, "step": 10976 }, { "epoch": 1.13, "learning_rate": 4.8652688975244824e-05, "loss": 0.8041, "step": 10977 }, { "epoch": 1.13, "learning_rate": 4.865076110166014e-05, "loss": 0.4651, "step": 10978 }, { "epoch": 1.13, "learning_rate": 4.86488331025229e-05, "loss": 0.8658, "step": 10979 }, { "epoch": 1.13, "learning_rate": 4.864690497784609e-05, "loss": 0.523, "step": 10980 }, { "epoch": 1.13, "learning_rate": 4.864497672764267e-05, "loss": 0.865, "step": 10981 }, { "epoch": 1.13, "learning_rate": 4.864304835192563e-05, "loss": 0.4059, "step": 10982 }, { "epoch": 1.13, "learning_rate": 4.864111985070795e-05, "loss": 0.3944, "step": 10983 }, { "epoch": 1.13, "learning_rate": 4.8639191224002615e-05, "loss": 1.3979, "step": 10984 }, { "epoch": 1.13, "learning_rate": 4.86372624718226e-05, "loss": 0.7006, "step": 10985 }, { "epoch": 1.13, "learning_rate": 4.86353335941809e-05, "loss": 0.517, "step": 10986 }, { "epoch": 1.13, "learning_rate": 4.863340459109049e-05, "loss": 0.5088, "step": 10987 }, { "epoch": 1.13, "learning_rate": 4.863147546256437e-05, "loss": 0.7742, "step": 10988 }, { "epoch": 1.13, "learning_rate": 4.8629546208615506e-05, "loss": 0.5851, "step": 10989 }, { "epoch": 1.13, "learning_rate": 4.8627616829256896e-05, "loss": 0.433, "step": 10990 }, { "epoch": 1.13, "learning_rate": 4.8625687324501536e-05, "loss": 0.5772, "step": 10991 }, { "epoch": 1.13, "learning_rate": 4.86237576943624e-05, "loss": 0.6244, "step": 10992 }, { "epoch": 1.13, "learning_rate": 4.862182793885249e-05, "loss": 0.2542, "step": 10993 }, { "epoch": 1.13, "learning_rate": 4.8619898057984785e-05, "loss": 0.3246, "step": 10994 }, { "epoch": 1.13, "learning_rate": 4.861796805177229e-05, "loss": 1.1299, "step": 10995 }, { "epoch": 1.13, "learning_rate": 4.861603792022799e-05, "loss": 0.9167, "step": 10996 }, { "epoch": 1.13, "learning_rate": 4.8614107663364876e-05, "loss": 0.1337, "step": 10997 }, { "epoch": 1.13, "learning_rate": 4.861217728119594e-05, "loss": 1.4668, "step": 10998 }, { "epoch": 1.13, "learning_rate": 4.861024677373419e-05, "loss": 0.2826, "step": 10999 }, { "epoch": 1.13, "learning_rate": 4.860831614099261e-05, "loss": 0.7982, "step": 11000 }, { "epoch": 1.13, "learning_rate": 4.8606385382984194e-05, "loss": 0.9723, "step": 11001 }, { "epoch": 1.13, "learning_rate": 4.8604454499721954e-05, "loss": 1.2023, "step": 11002 }, { "epoch": 1.13, "learning_rate": 4.860252349121887e-05, "loss": 0.6122, "step": 11003 }, { "epoch": 1.13, "learning_rate": 4.860059235748796e-05, "loss": 0.925, "step": 11004 }, { "epoch": 1.13, "learning_rate": 4.8598661098542206e-05, "loss": 0.806, "step": 11005 }, { "epoch": 1.13, "learning_rate": 4.859672971439463e-05, "loss": 0.4328, "step": 11006 }, { "epoch": 1.13, "learning_rate": 4.859479820505821e-05, "loss": 1.0505, "step": 11007 }, { "epoch": 1.13, "learning_rate": 4.8592866570545964e-05, "loss": 0.8575, "step": 11008 }, { "epoch": 1.13, "learning_rate": 4.8590934810870884e-05, "loss": 0.6797, "step": 11009 }, { "epoch": 1.13, "learning_rate": 4.8589002926045994e-05, "loss": 0.7685, "step": 11010 }, { "epoch": 1.13, "learning_rate": 4.858707091608427e-05, "loss": 0.35, "step": 11011 }, { "epoch": 1.13, "learning_rate": 4.858513878099875e-05, "loss": 0.8755, "step": 11012 }, { "epoch": 1.13, "learning_rate": 4.8583206520802405e-05, "loss": 0.5416, "step": 11013 }, { "epoch": 1.13, "learning_rate": 4.8581274135508274e-05, "loss": 0.7976, "step": 11014 }, { "epoch": 1.13, "learning_rate": 4.8579341625129346e-05, "loss": 0.8451, "step": 11015 }, { "epoch": 1.14, "learning_rate": 4.8577408989678644e-05, "loss": 0.6201, "step": 11016 }, { "epoch": 1.14, "learning_rate": 4.8575476229169165e-05, "loss": 0.3243, "step": 11017 }, { "epoch": 1.14, "learning_rate": 4.857354334361394e-05, "loss": 0.4307, "step": 11018 }, { "epoch": 1.14, "learning_rate": 4.857161033302596e-05, "loss": 0.9813, "step": 11019 }, { "epoch": 1.14, "learning_rate": 4.8569677197418234e-05, "loss": 0.3364, "step": 11020 }, { "epoch": 1.14, "learning_rate": 4.85677439368038e-05, "loss": 0.4162, "step": 11021 }, { "epoch": 1.14, "learning_rate": 4.8565810551195654e-05, "loss": 0.2064, "step": 11022 }, { "epoch": 1.14, "learning_rate": 4.856387704060681e-05, "loss": 0.5023, "step": 11023 }, { "epoch": 1.14, "learning_rate": 4.856194340505029e-05, "loss": 1.3298, "step": 11024 }, { "epoch": 1.14, "learning_rate": 4.8560009644539124e-05, "loss": 0.8444, "step": 11025 }, { "epoch": 1.14, "learning_rate": 4.8558075759086306e-05, "loss": 0.496, "step": 11026 }, { "epoch": 1.14, "learning_rate": 4.855614174870486e-05, "loss": 0.5029, "step": 11027 }, { "epoch": 1.14, "learning_rate": 4.855420761340781e-05, "loss": 0.3777, "step": 11028 }, { "epoch": 1.14, "learning_rate": 4.8552273353208184e-05, "loss": 0.8231, "step": 11029 }, { "epoch": 1.14, "learning_rate": 4.855033896811899e-05, "loss": 0.3525, "step": 11030 }, { "epoch": 1.14, "learning_rate": 4.854840445815326e-05, "loss": 0.5426, "step": 11031 }, { "epoch": 1.14, "learning_rate": 4.8546469823324e-05, "loss": 0.8467, "step": 11032 }, { "epoch": 1.14, "learning_rate": 4.854453506364426e-05, "loss": 0.442, "step": 11033 }, { "epoch": 1.14, "learning_rate": 4.8542600179127036e-05, "loss": 0.6186, "step": 11034 }, { "epoch": 1.14, "learning_rate": 4.854066516978536e-05, "loss": 0.5099, "step": 11035 }, { "epoch": 1.14, "learning_rate": 4.8538730035632276e-05, "loss": 0.4564, "step": 11036 }, { "epoch": 1.14, "learning_rate": 4.85367947766808e-05, "loss": 0.2987, "step": 11037 }, { "epoch": 1.14, "learning_rate": 4.8534859392943954e-05, "loss": 0.7278, "step": 11038 }, { "epoch": 1.14, "learning_rate": 4.853292388443478e-05, "loss": 0.9474, "step": 11039 }, { "epoch": 1.14, "learning_rate": 4.853098825116629e-05, "loss": 0.1596, "step": 11040 }, { "epoch": 1.14, "learning_rate": 4.8529052493151524e-05, "loss": 0.7178, "step": 11041 }, { "epoch": 1.14, "learning_rate": 4.8527116610403514e-05, "loss": 0.9345, "step": 11042 }, { "epoch": 1.14, "learning_rate": 4.852518060293529e-05, "loss": 0.6165, "step": 11043 }, { "epoch": 1.14, "learning_rate": 4.852324447075988e-05, "loss": 0.3311, "step": 11044 }, { "epoch": 1.14, "learning_rate": 4.852130821389032e-05, "loss": 0.9296, "step": 11045 }, { "epoch": 1.14, "learning_rate": 4.851937183233966e-05, "loss": 0.424, "step": 11046 }, { "epoch": 1.14, "learning_rate": 4.851743532612092e-05, "loss": 0.2909, "step": 11047 }, { "epoch": 1.14, "learning_rate": 4.851549869524713e-05, "loss": 0.748, "step": 11048 }, { "epoch": 1.14, "learning_rate": 4.8513561939731334e-05, "loss": 1.0238, "step": 11049 }, { "epoch": 1.14, "learning_rate": 4.851162505958658e-05, "loss": 0.804, "step": 11050 }, { "epoch": 1.14, "learning_rate": 4.8509688054825896e-05, "loss": 0.2381, "step": 11051 }, { "epoch": 1.14, "learning_rate": 4.8507750925462316e-05, "loss": 0.9722, "step": 11052 }, { "epoch": 1.14, "learning_rate": 4.8505813671508884e-05, "loss": 0.7324, "step": 11053 }, { "epoch": 1.14, "learning_rate": 4.850387629297866e-05, "loss": 0.9671, "step": 11054 }, { "epoch": 1.14, "learning_rate": 4.850193878988465e-05, "loss": 0.5822, "step": 11055 }, { "epoch": 1.14, "learning_rate": 4.8500001162239934e-05, "loss": 0.9889, "step": 11056 }, { "epoch": 1.14, "learning_rate": 4.8498063410057535e-05, "loss": 0.788, "step": 11057 }, { "epoch": 1.14, "learning_rate": 4.8496125533350504e-05, "loss": 0.5178, "step": 11058 }, { "epoch": 1.14, "learning_rate": 4.849418753213187e-05, "loss": 0.5111, "step": 11059 }, { "epoch": 1.14, "learning_rate": 4.849224940641471e-05, "loss": 0.5635, "step": 11060 }, { "epoch": 1.14, "learning_rate": 4.849031115621204e-05, "loss": 0.3027, "step": 11061 }, { "epoch": 1.14, "learning_rate": 4.848837278153693e-05, "loss": 0.6226, "step": 11062 }, { "epoch": 1.14, "learning_rate": 4.8486434282402414e-05, "loss": 0.4628, "step": 11063 }, { "epoch": 1.14, "learning_rate": 4.848449565882155e-05, "loss": 0.2222, "step": 11064 }, { "epoch": 1.14, "learning_rate": 4.8482556910807395e-05, "loss": 1.0546, "step": 11065 }, { "epoch": 1.14, "learning_rate": 4.848061803837298e-05, "loss": 0.6551, "step": 11066 }, { "epoch": 1.14, "learning_rate": 4.847867904153138e-05, "loss": 0.5327, "step": 11067 }, { "epoch": 1.14, "learning_rate": 4.847673992029563e-05, "loss": 0.5901, "step": 11068 }, { "epoch": 1.14, "learning_rate": 4.847480067467879e-05, "loss": 0.269, "step": 11069 }, { "epoch": 1.14, "learning_rate": 4.8472861304693914e-05, "loss": 0.7248, "step": 11070 }, { "epoch": 1.14, "learning_rate": 4.847092181035406e-05, "loss": 0.3284, "step": 11071 }, { "epoch": 1.14, "learning_rate": 4.846898219167227e-05, "loss": 0.7107, "step": 11072 }, { "epoch": 1.14, "learning_rate": 4.8467042448661626e-05, "loss": 0.9073, "step": 11073 }, { "epoch": 1.14, "learning_rate": 4.8465102581335175e-05, "loss": 0.8933, "step": 11074 }, { "epoch": 1.14, "learning_rate": 4.846316258970597e-05, "loss": 0.4301, "step": 11075 }, { "epoch": 1.14, "learning_rate": 4.8461222473787076e-05, "loss": 0.8118, "step": 11076 }, { "epoch": 1.14, "learning_rate": 4.8459282233591544e-05, "loss": 0.2921, "step": 11077 }, { "epoch": 1.14, "learning_rate": 4.845734186913245e-05, "loss": 0.1895, "step": 11078 }, { "epoch": 1.14, "learning_rate": 4.8455401380422845e-05, "loss": 0.9427, "step": 11079 }, { "epoch": 1.14, "learning_rate": 4.84534607674758e-05, "loss": 0.6071, "step": 11080 }, { "epoch": 1.14, "learning_rate": 4.845152003030438e-05, "loss": 0.7235, "step": 11081 }, { "epoch": 1.14, "learning_rate": 4.844957916892163e-05, "loss": 0.3146, "step": 11082 }, { "epoch": 1.14, "learning_rate": 4.844763818334064e-05, "loss": 0.9627, "step": 11083 }, { "epoch": 1.14, "learning_rate": 4.844569707357447e-05, "loss": 0.6858, "step": 11084 }, { "epoch": 1.14, "learning_rate": 4.844375583963618e-05, "loss": 0.5987, "step": 11085 }, { "epoch": 1.14, "learning_rate": 4.844181448153883e-05, "loss": 0.7002, "step": 11086 }, { "epoch": 1.14, "learning_rate": 4.843987299929552e-05, "loss": 0.2465, "step": 11087 }, { "epoch": 1.14, "learning_rate": 4.843793139291928e-05, "loss": 0.7923, "step": 11088 }, { "epoch": 1.14, "learning_rate": 4.843598966242321e-05, "loss": 0.4938, "step": 11089 }, { "epoch": 1.14, "learning_rate": 4.8434047807820376e-05, "loss": 0.6834, "step": 11090 }, { "epoch": 1.14, "learning_rate": 4.843210582912383e-05, "loss": 0.653, "step": 11091 }, { "epoch": 1.14, "learning_rate": 4.843016372634668e-05, "loss": 0.2863, "step": 11092 }, { "epoch": 1.14, "learning_rate": 4.8428221499501965e-05, "loss": 0.8069, "step": 11093 }, { "epoch": 1.14, "learning_rate": 4.842627914860279e-05, "loss": 0.7915, "step": 11094 }, { "epoch": 1.14, "learning_rate": 4.84243366736622e-05, "loss": 0.3909, "step": 11095 }, { "epoch": 1.14, "learning_rate": 4.8422394074693286e-05, "loss": 0.6639, "step": 11096 }, { "epoch": 1.14, "learning_rate": 4.8420451351709135e-05, "loss": 0.53, "step": 11097 }, { "epoch": 1.14, "learning_rate": 4.841850850472281e-05, "loss": 0.6725, "step": 11098 }, { "epoch": 1.14, "learning_rate": 4.841656553374739e-05, "loss": 0.1854, "step": 11099 }, { "epoch": 1.14, "learning_rate": 4.841462243879597e-05, "loss": 1.0865, "step": 11100 }, { "epoch": 1.14, "learning_rate": 4.841267921988162e-05, "loss": 0.7024, "step": 11101 }, { "epoch": 1.14, "learning_rate": 4.841073587701742e-05, "loss": 0.2042, "step": 11102 }, { "epoch": 1.14, "learning_rate": 4.840879241021646e-05, "loss": 0.1922, "step": 11103 }, { "epoch": 1.14, "learning_rate": 4.84068488194918e-05, "loss": 1.0148, "step": 11104 }, { "epoch": 1.14, "learning_rate": 4.8404905104856555e-05, "loss": 0.2654, "step": 11105 }, { "epoch": 1.14, "learning_rate": 4.840296126632379e-05, "loss": 0.456, "step": 11106 }, { "epoch": 1.14, "learning_rate": 4.840101730390659e-05, "loss": 0.5711, "step": 11107 }, { "epoch": 1.14, "learning_rate": 4.839907321761805e-05, "loss": 0.3237, "step": 11108 }, { "epoch": 1.14, "learning_rate": 4.839712900747126e-05, "loss": 0.7188, "step": 11109 }, { "epoch": 1.14, "learning_rate": 4.83951846734793e-05, "loss": 0.7638, "step": 11110 }, { "epoch": 1.14, "learning_rate": 4.839324021565526e-05, "loss": 0.8114, "step": 11111 }, { "epoch": 1.15, "learning_rate": 4.839129563401224e-05, "loss": 0.8542, "step": 11112 }, { "epoch": 1.15, "learning_rate": 4.838935092856331e-05, "loss": 0.2789, "step": 11113 }, { "epoch": 1.15, "learning_rate": 4.838740609932157e-05, "loss": 0.4446, "step": 11114 }, { "epoch": 1.15, "learning_rate": 4.838546114630012e-05, "loss": 0.7284, "step": 11115 }, { "epoch": 1.15, "learning_rate": 4.838351606951204e-05, "loss": 0.6752, "step": 11116 }, { "epoch": 1.15, "learning_rate": 4.838157086897044e-05, "loss": 0.6516, "step": 11117 }, { "epoch": 1.15, "learning_rate": 4.83796255446884e-05, "loss": 0.6919, "step": 11118 }, { "epoch": 1.15, "learning_rate": 4.837768009667903e-05, "loss": 0.6339, "step": 11119 }, { "epoch": 1.15, "learning_rate": 4.8375734524955416e-05, "loss": 0.8967, "step": 11120 }, { "epoch": 1.15, "learning_rate": 4.837378882953065e-05, "loss": 0.8729, "step": 11121 }, { "epoch": 1.15, "learning_rate": 4.837184301041784e-05, "loss": 1.0914, "step": 11122 }, { "epoch": 1.15, "learning_rate": 4.8369897067630076e-05, "loss": 0.8647, "step": 11123 }, { "epoch": 1.15, "learning_rate": 4.836795100118047e-05, "loss": 0.9881, "step": 11124 }, { "epoch": 1.15, "learning_rate": 4.836600481108212e-05, "loss": 1.353, "step": 11125 }, { "epoch": 1.15, "learning_rate": 4.836405849734812e-05, "loss": 0.167, "step": 11126 }, { "epoch": 1.15, "learning_rate": 4.8362112059991564e-05, "loss": 0.856, "step": 11127 }, { "epoch": 1.15, "learning_rate": 4.836016549902557e-05, "loss": 0.9597, "step": 11128 }, { "epoch": 1.15, "learning_rate": 4.8358218814463246e-05, "loss": 1.3954, "step": 11129 }, { "epoch": 1.15, "learning_rate": 4.835627200631768e-05, "loss": 0.8747, "step": 11130 }, { "epoch": 1.15, "learning_rate": 4.835432507460199e-05, "loss": 0.2859, "step": 11131 }, { "epoch": 1.15, "learning_rate": 4.8352378019329286e-05, "loss": 0.3569, "step": 11132 }, { "epoch": 1.15, "learning_rate": 4.835043084051265e-05, "loss": 0.6384, "step": 11133 }, { "epoch": 1.15, "learning_rate": 4.834848353816522e-05, "loss": 0.2942, "step": 11134 }, { "epoch": 1.15, "learning_rate": 4.834653611230008e-05, "loss": 0.3401, "step": 11135 }, { "epoch": 1.15, "learning_rate": 4.834458856293037e-05, "loss": 0.3326, "step": 11136 }, { "epoch": 1.15, "learning_rate": 4.834264089006917e-05, "loss": 0.1452, "step": 11137 }, { "epoch": 1.15, "learning_rate": 4.834069309372961e-05, "loss": 0.716, "step": 11138 }, { "epoch": 1.15, "learning_rate": 4.833874517392479e-05, "loss": 1.1712, "step": 11139 }, { "epoch": 1.15, "learning_rate": 4.833679713066783e-05, "loss": 0.3265, "step": 11140 }, { "epoch": 1.15, "learning_rate": 4.833484896397184e-05, "loss": 0.8051, "step": 11141 }, { "epoch": 1.15, "learning_rate": 4.833290067384994e-05, "loss": 1.2527, "step": 11142 }, { "epoch": 1.15, "learning_rate": 4.833095226031524e-05, "loss": 0.8329, "step": 11143 }, { "epoch": 1.15, "learning_rate": 4.832900372338086e-05, "loss": 0.8149, "step": 11144 }, { "epoch": 1.15, "learning_rate": 4.832705506305991e-05, "loss": 1.2008, "step": 11145 }, { "epoch": 1.15, "learning_rate": 4.8325106279365526e-05, "loss": 0.3941, "step": 11146 }, { "epoch": 1.15, "learning_rate": 4.83231573723108e-05, "loss": 1.0342, "step": 11147 }, { "epoch": 1.15, "learning_rate": 4.832120834190887e-05, "loss": 0.5732, "step": 11148 }, { "epoch": 1.15, "learning_rate": 4.831925918817286e-05, "loss": 0.6226, "step": 11149 }, { "epoch": 1.15, "learning_rate": 4.831730991111588e-05, "loss": 0.9023, "step": 11150 }, { "epoch": 1.15, "learning_rate": 4.8315360510751054e-05, "loss": 0.6946, "step": 11151 }, { "epoch": 1.15, "learning_rate": 4.83134109870915e-05, "loss": 0.3246, "step": 11152 }, { "epoch": 1.15, "learning_rate": 4.831146134015036e-05, "loss": 0.7, "step": 11153 }, { "epoch": 1.15, "learning_rate": 4.8309511569940734e-05, "loss": 0.1115, "step": 11154 }, { "epoch": 1.15, "learning_rate": 4.8307561676475773e-05, "loss": 0.1898, "step": 11155 }, { "epoch": 1.15, "learning_rate": 4.830561165976858e-05, "loss": 0.331, "step": 11156 }, { "epoch": 1.15, "learning_rate": 4.83036615198323e-05, "loss": 0.8896, "step": 11157 }, { "epoch": 1.15, "learning_rate": 4.830171125668005e-05, "loss": 0.2772, "step": 11158 }, { "epoch": 1.15, "learning_rate": 4.829976087032496e-05, "loss": 0.6579, "step": 11159 }, { "epoch": 1.15, "learning_rate": 4.829781036078016e-05, "loss": 0.2227, "step": 11160 }, { "epoch": 1.15, "learning_rate": 4.829585972805879e-05, "loss": 0.8906, "step": 11161 }, { "epoch": 1.15, "learning_rate": 4.829390897217397e-05, "loss": 0.3096, "step": 11162 }, { "epoch": 1.15, "learning_rate": 4.8291958093138845e-05, "loss": 0.4999, "step": 11163 }, { "epoch": 1.15, "learning_rate": 4.829000709096652e-05, "loss": 1.0423, "step": 11164 }, { "epoch": 1.15, "learning_rate": 4.828805596567016e-05, "loss": 1.0734, "step": 11165 }, { "epoch": 1.15, "learning_rate": 4.8286104717262874e-05, "loss": 1.6001, "step": 11166 }, { "epoch": 1.15, "learning_rate": 4.8284153345757825e-05, "loss": 0.3457, "step": 11167 }, { "epoch": 1.15, "learning_rate": 4.828220185116812e-05, "loss": 0.2597, "step": 11168 }, { "epoch": 1.15, "learning_rate": 4.8280250233506915e-05, "loss": 1.0412, "step": 11169 }, { "epoch": 1.15, "learning_rate": 4.8278298492787346e-05, "loss": 1.2235, "step": 11170 }, { "epoch": 1.15, "learning_rate": 4.827634662902254e-05, "loss": 0.6915, "step": 11171 }, { "epoch": 1.15, "learning_rate": 4.827439464222565e-05, "loss": 0.4299, "step": 11172 }, { "epoch": 1.15, "learning_rate": 4.827244253240981e-05, "loss": 0.6372, "step": 11173 }, { "epoch": 1.15, "learning_rate": 4.827049029958817e-05, "loss": 0.4335, "step": 11174 }, { "epoch": 1.15, "learning_rate": 4.826853794377386e-05, "loss": 0.7746, "step": 11175 }, { "epoch": 1.15, "learning_rate": 4.826658546498002e-05, "loss": 0.6965, "step": 11176 }, { "epoch": 1.15, "learning_rate": 4.826463286321981e-05, "loss": 0.5629, "step": 11177 }, { "epoch": 1.15, "learning_rate": 4.826268013850636e-05, "loss": 0.4692, "step": 11178 }, { "epoch": 1.15, "learning_rate": 4.8260727290852816e-05, "loss": 0.4924, "step": 11179 }, { "epoch": 1.15, "learning_rate": 4.825877432027233e-05, "loss": 0.6562, "step": 11180 }, { "epoch": 1.15, "learning_rate": 4.8256821226778055e-05, "loss": 1.2052, "step": 11181 }, { "epoch": 1.15, "learning_rate": 4.825486801038313e-05, "loss": 0.4278, "step": 11182 }, { "epoch": 1.15, "learning_rate": 4.8252914671100706e-05, "loss": 0.4948, "step": 11183 }, { "epoch": 1.15, "learning_rate": 4.825096120894392e-05, "loss": 0.6715, "step": 11184 }, { "epoch": 1.15, "learning_rate": 4.8249007623925944e-05, "loss": 0.8075, "step": 11185 }, { "epoch": 1.15, "learning_rate": 4.824705391605992e-05, "loss": 0.5838, "step": 11186 }, { "epoch": 1.15, "learning_rate": 4.824510008535898e-05, "loss": 0.5472, "step": 11187 }, { "epoch": 1.15, "learning_rate": 4.824314613183632e-05, "loss": 1.0382, "step": 11188 }, { "epoch": 1.15, "learning_rate": 4.824119205550506e-05, "loss": 0.9532, "step": 11189 }, { "epoch": 1.15, "learning_rate": 4.8239237856378355e-05, "loss": 0.8438, "step": 11190 }, { "epoch": 1.15, "learning_rate": 4.823728353446938e-05, "loss": 0.8351, "step": 11191 }, { "epoch": 1.15, "learning_rate": 4.823532908979127e-05, "loss": 0.2124, "step": 11192 }, { "epoch": 1.15, "learning_rate": 4.8233374522357196e-05, "loss": 0.7932, "step": 11193 }, { "epoch": 1.15, "learning_rate": 4.823141983218031e-05, "loss": 0.1639, "step": 11194 }, { "epoch": 1.15, "learning_rate": 4.822946501927378e-05, "loss": 0.7444, "step": 11195 }, { "epoch": 1.15, "learning_rate": 4.8227510083650744e-05, "loss": 0.9388, "step": 11196 }, { "epoch": 1.15, "learning_rate": 4.822555502532438e-05, "loss": 0.6626, "step": 11197 }, { "epoch": 1.15, "learning_rate": 4.8223599844307847e-05, "loss": 1.0873, "step": 11198 }, { "epoch": 1.15, "learning_rate": 4.822164454061429e-05, "loss": 0.4381, "step": 11199 }, { "epoch": 1.15, "learning_rate": 4.82196891142569e-05, "loss": 0.8576, "step": 11200 }, { "epoch": 1.15, "learning_rate": 4.8217733565248824e-05, "loss": 0.353, "step": 11201 }, { "epoch": 1.15, "learning_rate": 4.821577789360322e-05, "loss": 0.2854, "step": 11202 }, { "epoch": 1.15, "learning_rate": 4.821382209933328e-05, "loss": 0.4962, "step": 11203 }, { "epoch": 1.15, "learning_rate": 4.8211866182452136e-05, "loss": 0.5173, "step": 11204 }, { "epoch": 1.15, "learning_rate": 4.8209910142972976e-05, "loss": 0.7506, "step": 11205 }, { "epoch": 1.15, "learning_rate": 4.820795398090895e-05, "loss": 0.6285, "step": 11206 }, { "epoch": 1.15, "learning_rate": 4.820599769627325e-05, "loss": 0.941, "step": 11207 }, { "epoch": 1.16, "learning_rate": 4.820404128907903e-05, "loss": 0.2293, "step": 11208 }, { "epoch": 1.16, "learning_rate": 4.8202084759339464e-05, "loss": 0.2952, "step": 11209 }, { "epoch": 1.16, "learning_rate": 4.8200128107067725e-05, "loss": 0.1781, "step": 11210 }, { "epoch": 1.16, "learning_rate": 4.8198171332276974e-05, "loss": 0.9432, "step": 11211 }, { "epoch": 1.16, "learning_rate": 4.81962144349804e-05, "loss": 0.8794, "step": 11212 }, { "epoch": 1.16, "learning_rate": 4.819425741519117e-05, "loss": 1.0678, "step": 11213 }, { "epoch": 1.16, "learning_rate": 4.819230027292245e-05, "loss": 0.1935, "step": 11214 }, { "epoch": 1.16, "learning_rate": 4.819034300818742e-05, "loss": 0.9792, "step": 11215 }, { "epoch": 1.16, "learning_rate": 4.8188385620999264e-05, "loss": 0.3387, "step": 11216 }, { "epoch": 1.16, "learning_rate": 4.8186428111371144e-05, "loss": 0.7048, "step": 11217 }, { "epoch": 1.16, "learning_rate": 4.818447047931624e-05, "loss": 0.6911, "step": 11218 }, { "epoch": 1.16, "learning_rate": 4.818251272484775e-05, "loss": 0.6885, "step": 11219 }, { "epoch": 1.16, "learning_rate": 4.818055484797884e-05, "loss": 0.2661, "step": 11220 }, { "epoch": 1.16, "learning_rate": 4.817859684872268e-05, "loss": 0.3025, "step": 11221 }, { "epoch": 1.16, "learning_rate": 4.817663872709246e-05, "loss": 0.6066, "step": 11222 }, { "epoch": 1.16, "learning_rate": 4.817468048310136e-05, "loss": 0.6167, "step": 11223 }, { "epoch": 1.16, "learning_rate": 4.817272211676256e-05, "loss": 0.5171, "step": 11224 }, { "epoch": 1.16, "learning_rate": 4.817076362808926e-05, "loss": 1.5919, "step": 11225 }, { "epoch": 1.16, "learning_rate": 4.8168805017094615e-05, "loss": 0.9948, "step": 11226 }, { "epoch": 1.16, "learning_rate": 4.8166846283791837e-05, "loss": 0.6783, "step": 11227 }, { "epoch": 1.16, "learning_rate": 4.8164887428194096e-05, "loss": 1.0405, "step": 11228 }, { "epoch": 1.16, "learning_rate": 4.8162928450314586e-05, "loss": 0.5467, "step": 11229 }, { "epoch": 1.16, "learning_rate": 4.816096935016649e-05, "loss": 0.2012, "step": 11230 }, { "epoch": 1.16, "learning_rate": 4.815901012776299e-05, "loss": 0.7089, "step": 11231 }, { "epoch": 1.16, "learning_rate": 4.8157050783117295e-05, "loss": 0.6959, "step": 11232 }, { "epoch": 1.16, "learning_rate": 4.8155091316242574e-05, "loss": 0.5085, "step": 11233 }, { "epoch": 1.16, "learning_rate": 4.815313172715203e-05, "loss": 0.7572, "step": 11234 }, { "epoch": 1.16, "learning_rate": 4.815117201585886e-05, "loss": 0.7097, "step": 11235 }, { "epoch": 1.16, "learning_rate": 4.814921218237624e-05, "loss": 0.8018, "step": 11236 }, { "epoch": 1.16, "learning_rate": 4.814725222671736e-05, "loss": 1.0135, "step": 11237 }, { "epoch": 1.16, "learning_rate": 4.814529214889543e-05, "loss": 0.4499, "step": 11238 }, { "epoch": 1.16, "learning_rate": 4.814333194892365e-05, "loss": 0.6219, "step": 11239 }, { "epoch": 1.16, "learning_rate": 4.814137162681519e-05, "loss": 0.7992, "step": 11240 }, { "epoch": 1.16, "learning_rate": 4.813941118258327e-05, "loss": 0.2143, "step": 11241 }, { "epoch": 1.16, "learning_rate": 4.813745061624107e-05, "loss": 0.7134, "step": 11242 }, { "epoch": 1.16, "learning_rate": 4.8135489927801805e-05, "loss": 0.3449, "step": 11243 }, { "epoch": 1.16, "learning_rate": 4.813352911727866e-05, "loss": 0.493, "step": 11244 }, { "epoch": 1.16, "learning_rate": 4.813156818468484e-05, "loss": 0.3229, "step": 11245 }, { "epoch": 1.16, "learning_rate": 4.8129607130033544e-05, "loss": 0.3909, "step": 11246 }, { "epoch": 1.16, "learning_rate": 4.8127645953337974e-05, "loss": 0.8284, "step": 11247 }, { "epoch": 1.16, "learning_rate": 4.8125684654611336e-05, "loss": 0.7046, "step": 11248 }, { "epoch": 1.16, "learning_rate": 4.812372323386683e-05, "loss": 0.9684, "step": 11249 }, { "epoch": 1.16, "learning_rate": 4.8121761691117665e-05, "loss": 1.2539, "step": 11250 }, { "epoch": 1.16, "learning_rate": 4.811980002637704e-05, "loss": 0.3435, "step": 11251 }, { "epoch": 1.16, "learning_rate": 4.811783823965815e-05, "loss": 0.9017, "step": 11252 }, { "epoch": 1.16, "learning_rate": 4.811587633097422e-05, "loss": 0.1637, "step": 11253 }, { "epoch": 1.16, "learning_rate": 4.811391430033845e-05, "loss": 0.9334, "step": 11254 }, { "epoch": 1.16, "learning_rate": 4.811195214776404e-05, "loss": 0.5841, "step": 11255 }, { "epoch": 1.16, "learning_rate": 4.810998987326421e-05, "loss": 0.5155, "step": 11256 }, { "epoch": 1.16, "learning_rate": 4.8108027476852175e-05, "loss": 0.6909, "step": 11257 }, { "epoch": 1.16, "learning_rate": 4.8106064958541126e-05, "loss": 0.9995, "step": 11258 }, { "epoch": 1.16, "learning_rate": 4.810410231834428e-05, "loss": 1.8143, "step": 11259 }, { "epoch": 1.16, "learning_rate": 4.810213955627486e-05, "loss": 0.6822, "step": 11260 }, { "epoch": 1.16, "learning_rate": 4.810017667234607e-05, "loss": 0.3264, "step": 11261 }, { "epoch": 1.16, "learning_rate": 4.809821366657112e-05, "loss": 0.2971, "step": 11262 }, { "epoch": 1.16, "learning_rate": 4.809625053896325e-05, "loss": 0.3497, "step": 11263 }, { "epoch": 1.16, "learning_rate": 4.809428728953564e-05, "loss": 1.1482, "step": 11264 }, { "epoch": 1.16, "learning_rate": 4.8092323918301523e-05, "loss": 0.8596, "step": 11265 }, { "epoch": 1.16, "learning_rate": 4.8090360425274115e-05, "loss": 0.0111, "step": 11266 }, { "epoch": 1.16, "learning_rate": 4.808839681046663e-05, "loss": 1.0094, "step": 11267 }, { "epoch": 1.16, "learning_rate": 4.808643307389229e-05, "loss": 0.6495, "step": 11268 }, { "epoch": 1.16, "learning_rate": 4.8084469215564334e-05, "loss": 0.7992, "step": 11269 }, { "epoch": 1.16, "learning_rate": 4.808250523549594e-05, "loss": 1.2831, "step": 11270 }, { "epoch": 1.16, "learning_rate": 4.8080541133700356e-05, "loss": 0.2627, "step": 11271 }, { "epoch": 1.16, "learning_rate": 4.80785769101908e-05, "loss": 1.4367, "step": 11272 }, { "epoch": 1.16, "learning_rate": 4.80766125649805e-05, "loss": 0.1317, "step": 11273 }, { "epoch": 1.16, "learning_rate": 4.8074648098082675e-05, "loss": 0.4529, "step": 11274 }, { "epoch": 1.16, "learning_rate": 4.807268350951054e-05, "loss": 0.6371, "step": 11275 }, { "epoch": 1.16, "learning_rate": 4.8070718799277334e-05, "loss": 1.097, "step": 11276 }, { "epoch": 1.16, "learning_rate": 4.8068753967396276e-05, "loss": 0.2984, "step": 11277 }, { "epoch": 1.16, "learning_rate": 4.8066789013880595e-05, "loss": 0.4033, "step": 11278 }, { "epoch": 1.16, "learning_rate": 4.806482393874352e-05, "loss": 0.789, "step": 11279 }, { "epoch": 1.16, "learning_rate": 4.806285874199827e-05, "loss": 0.2508, "step": 11280 }, { "epoch": 1.16, "learning_rate": 4.806089342365808e-05, "loss": 0.6267, "step": 11281 }, { "epoch": 1.16, "learning_rate": 4.805892798373619e-05, "loss": 1.0476, "step": 11282 }, { "epoch": 1.16, "learning_rate": 4.8056962422245816e-05, "loss": 1.0312, "step": 11283 }, { "epoch": 1.16, "learning_rate": 4.80549967392002e-05, "loss": 0.9715, "step": 11284 }, { "epoch": 1.16, "learning_rate": 4.805303093461257e-05, "loss": 0.5216, "step": 11285 }, { "epoch": 1.16, "learning_rate": 4.805106500849615e-05, "loss": 0.6746, "step": 11286 }, { "epoch": 1.16, "learning_rate": 4.8049098960864196e-05, "loss": 0.1185, "step": 11287 }, { "epoch": 1.16, "learning_rate": 4.804713279172993e-05, "loss": 0.8718, "step": 11288 }, { "epoch": 1.16, "learning_rate": 4.804516650110659e-05, "loss": 0.3479, "step": 11289 }, { "epoch": 1.16, "learning_rate": 4.80432000890074e-05, "loss": 0.7059, "step": 11290 }, { "epoch": 1.16, "learning_rate": 4.8041233555445616e-05, "loss": 0.4724, "step": 11291 }, { "epoch": 1.16, "learning_rate": 4.803926690043447e-05, "loss": 0.3548, "step": 11292 }, { "epoch": 1.16, "learning_rate": 4.80373001239872e-05, "loss": 0.7648, "step": 11293 }, { "epoch": 1.16, "learning_rate": 4.8035333226117045e-05, "loss": 0.2161, "step": 11294 }, { "epoch": 1.16, "learning_rate": 4.803336620683725e-05, "loss": 0.8577, "step": 11295 }, { "epoch": 1.16, "learning_rate": 4.803139906616105e-05, "loss": 0.5091, "step": 11296 }, { "epoch": 1.16, "learning_rate": 4.802943180410169e-05, "loss": 0.627, "step": 11297 }, { "epoch": 1.16, "learning_rate": 4.802746442067242e-05, "loss": 0.683, "step": 11298 }, { "epoch": 1.16, "learning_rate": 4.8025496915886465e-05, "loss": 0.9649, "step": 11299 }, { "epoch": 1.16, "learning_rate": 4.8023529289757094e-05, "loss": 0.976, "step": 11300 }, { "epoch": 1.16, "learning_rate": 4.802156154229754e-05, "loss": 0.8339, "step": 11301 }, { "epoch": 1.16, "learning_rate": 4.8019593673521044e-05, "loss": 0.6335, "step": 11302 }, { "epoch": 1.17, "learning_rate": 4.801762568344086e-05, "loss": 1.1843, "step": 11303 }, { "epoch": 1.17, "learning_rate": 4.801565757207023e-05, "loss": 0.3891, "step": 11304 }, { "epoch": 1.17, "learning_rate": 4.801368933942242e-05, "loss": 0.6467, "step": 11305 }, { "epoch": 1.17, "learning_rate": 4.8011720985510665e-05, "loss": 0.0974, "step": 11306 }, { "epoch": 1.17, "learning_rate": 4.8009752510348224e-05, "loss": 0.273, "step": 11307 }, { "epoch": 1.17, "learning_rate": 4.800778391394834e-05, "loss": 0.3472, "step": 11308 }, { "epoch": 1.17, "learning_rate": 4.800581519632426e-05, "loss": 0.2622, "step": 11309 }, { "epoch": 1.17, "learning_rate": 4.800384635748925e-05, "loss": 0.5492, "step": 11310 }, { "epoch": 1.17, "learning_rate": 4.800187739745656e-05, "loss": 0.8333, "step": 11311 }, { "epoch": 1.17, "learning_rate": 4.799990831623945e-05, "loss": 1.1559, "step": 11312 }, { "epoch": 1.17, "learning_rate": 4.799793911385116e-05, "loss": 0.3163, "step": 11313 }, { "epoch": 1.17, "learning_rate": 4.799596979030495e-05, "loss": 0.2196, "step": 11314 }, { "epoch": 1.17, "learning_rate": 4.7994000345614096e-05, "loss": 0.2674, "step": 11315 }, { "epoch": 1.17, "learning_rate": 4.7992030779791826e-05, "loss": 0.6871, "step": 11316 }, { "epoch": 1.17, "learning_rate": 4.7990061092851426e-05, "loss": 0.4949, "step": 11317 }, { "epoch": 1.17, "learning_rate": 4.7988091284806146e-05, "loss": 0.5775, "step": 11318 }, { "epoch": 1.17, "learning_rate": 4.798612135566924e-05, "loss": 1.214, "step": 11319 }, { "epoch": 1.17, "learning_rate": 4.798415130545397e-05, "loss": 0.5728, "step": 11320 }, { "epoch": 1.17, "learning_rate": 4.7982181134173614e-05, "loss": 0.6202, "step": 11321 }, { "epoch": 1.17, "learning_rate": 4.798021084184141e-05, "loss": 0.2429, "step": 11322 }, { "epoch": 1.17, "learning_rate": 4.797824042847063e-05, "loss": 0.7544, "step": 11323 }, { "epoch": 1.17, "learning_rate": 4.797626989407455e-05, "loss": 0.5052, "step": 11324 }, { "epoch": 1.17, "learning_rate": 4.7974299238666423e-05, "loss": 0.3384, "step": 11325 }, { "epoch": 1.17, "learning_rate": 4.797232846225953e-05, "loss": 0.4571, "step": 11326 }, { "epoch": 1.17, "learning_rate": 4.797035756486712e-05, "loss": 1.1011, "step": 11327 }, { "epoch": 1.17, "learning_rate": 4.796838654650247e-05, "loss": 0.0987, "step": 11328 }, { "epoch": 1.17, "learning_rate": 4.796641540717884e-05, "loss": 1.0598, "step": 11329 }, { "epoch": 1.17, "learning_rate": 4.79644441469095e-05, "loss": 0.3041, "step": 11330 }, { "epoch": 1.17, "learning_rate": 4.796247276570774e-05, "loss": 0.5334, "step": 11331 }, { "epoch": 1.17, "learning_rate": 4.7960501263586806e-05, "loss": 0.4593, "step": 11332 }, { "epoch": 1.17, "learning_rate": 4.7958529640559984e-05, "loss": 0.4891, "step": 11333 }, { "epoch": 1.17, "learning_rate": 4.7956557896640546e-05, "loss": 0.3343, "step": 11334 }, { "epoch": 1.17, "learning_rate": 4.7954586031841756e-05, "loss": 0.3185, "step": 11335 }, { "epoch": 1.17, "learning_rate": 4.795261404617689e-05, "loss": 0.9814, "step": 11336 }, { "epoch": 1.17, "learning_rate": 4.795064193965924e-05, "loss": 0.5585, "step": 11337 }, { "epoch": 1.17, "learning_rate": 4.7948669712302056e-05, "loss": 0.7672, "step": 11338 }, { "epoch": 1.17, "learning_rate": 4.794669736411863e-05, "loss": 0.8317, "step": 11339 }, { "epoch": 1.17, "learning_rate": 4.794472489512225e-05, "loss": 0.7353, "step": 11340 }, { "epoch": 1.17, "learning_rate": 4.7942752305326165e-05, "loss": 0.2312, "step": 11341 }, { "epoch": 1.17, "learning_rate": 4.794077959474367e-05, "loss": 0.8874, "step": 11342 }, { "epoch": 1.17, "learning_rate": 4.793880676338805e-05, "loss": 0.101, "step": 11343 }, { "epoch": 1.17, "learning_rate": 4.793683381127257e-05, "loss": 0.9833, "step": 11344 }, { "epoch": 1.17, "learning_rate": 4.7934860738410545e-05, "loss": 0.6881, "step": 11345 }, { "epoch": 1.17, "learning_rate": 4.793288754481521e-05, "loss": 0.3991, "step": 11346 }, { "epoch": 1.17, "learning_rate": 4.793091423049989e-05, "loss": 1.0355, "step": 11347 }, { "epoch": 1.17, "learning_rate": 4.792894079547784e-05, "loss": 0.6963, "step": 11348 }, { "epoch": 1.17, "learning_rate": 4.792696723976235e-05, "loss": 0.3281, "step": 11349 }, { "epoch": 1.17, "learning_rate": 4.792499356336672e-05, "loss": 0.8193, "step": 11350 }, { "epoch": 1.17, "learning_rate": 4.792301976630423e-05, "loss": 1.2747, "step": 11351 }, { "epoch": 1.17, "learning_rate": 4.792104584858816e-05, "loss": 0.9056, "step": 11352 }, { "epoch": 1.17, "learning_rate": 4.7919071810231806e-05, "loss": 0.9703, "step": 11353 }, { "epoch": 1.17, "learning_rate": 4.791709765124845e-05, "loss": 0.6878, "step": 11354 }, { "epoch": 1.17, "learning_rate": 4.791512337165138e-05, "loss": 0.9189, "step": 11355 }, { "epoch": 1.17, "learning_rate": 4.79131489714539e-05, "loss": 0.362, "step": 11356 }, { "epoch": 1.17, "learning_rate": 4.791117445066929e-05, "loss": 1.2972, "step": 11357 }, { "epoch": 1.17, "learning_rate": 4.790919980931085e-05, "loss": 0.3382, "step": 11358 }, { "epoch": 1.17, "learning_rate": 4.790722504739186e-05, "loss": 0.3047, "step": 11359 }, { "epoch": 1.17, "learning_rate": 4.790525016492562e-05, "loss": 0.5538, "step": 11360 }, { "epoch": 1.17, "learning_rate": 4.790327516192543e-05, "loss": 0.5622, "step": 11361 }, { "epoch": 1.17, "learning_rate": 4.790130003840458e-05, "loss": 0.4797, "step": 11362 }, { "epoch": 1.17, "learning_rate": 4.789932479437636e-05, "loss": 1.0599, "step": 11363 }, { "epoch": 1.17, "learning_rate": 4.789734942985408e-05, "loss": 0.9942, "step": 11364 }, { "epoch": 1.17, "learning_rate": 4.789537394485103e-05, "loss": 1.2316, "step": 11365 }, { "epoch": 1.17, "learning_rate": 4.789339833938051e-05, "loss": 0.1563, "step": 11366 }, { "epoch": 1.17, "learning_rate": 4.789142261345582e-05, "loss": 0.48, "step": 11367 }, { "epoch": 1.17, "learning_rate": 4.788944676709025e-05, "loss": 0.7133, "step": 11368 }, { "epoch": 1.17, "learning_rate": 4.788747080029712e-05, "loss": 0.5293, "step": 11369 }, { "epoch": 1.17, "learning_rate": 4.788549471308973e-05, "loss": 0.3253, "step": 11370 }, { "epoch": 1.17, "learning_rate": 4.788351850548136e-05, "loss": 0.5616, "step": 11371 }, { "epoch": 1.17, "learning_rate": 4.788154217748533e-05, "loss": 0.4728, "step": 11372 }, { "epoch": 1.17, "learning_rate": 4.787956572911494e-05, "loss": 0.1728, "step": 11373 }, { "epoch": 1.17, "learning_rate": 4.787758916038351e-05, "loss": 0.4944, "step": 11374 }, { "epoch": 1.17, "learning_rate": 4.787561247130432e-05, "loss": 0.9762, "step": 11375 }, { "epoch": 1.17, "learning_rate": 4.787363566189069e-05, "loss": 0.4536, "step": 11376 }, { "epoch": 1.17, "learning_rate": 4.787165873215594e-05, "loss": 0.8501, "step": 11377 }, { "epoch": 1.17, "learning_rate": 4.786968168211336e-05, "loss": 0.6083, "step": 11378 }, { "epoch": 1.17, "learning_rate": 4.786770451177625e-05, "loss": 0.5126, "step": 11379 }, { "epoch": 1.17, "learning_rate": 4.786572722115794e-05, "loss": 0.4267, "step": 11380 }, { "epoch": 1.17, "learning_rate": 4.786374981027174e-05, "loss": 0.3877, "step": 11381 }, { "epoch": 1.17, "learning_rate": 4.786177227913095e-05, "loss": 0.288, "step": 11382 }, { "epoch": 1.17, "learning_rate": 4.78597946277489e-05, "loss": 0.4114, "step": 11383 }, { "epoch": 1.17, "learning_rate": 4.785781685613889e-05, "loss": 0.5607, "step": 11384 }, { "epoch": 1.17, "learning_rate": 4.785583896431422e-05, "loss": 1.2929, "step": 11385 }, { "epoch": 1.17, "learning_rate": 4.785386095228823e-05, "loss": 0.5461, "step": 11386 }, { "epoch": 1.17, "learning_rate": 4.7851882820074225e-05, "loss": 0.6222, "step": 11387 }, { "epoch": 1.17, "learning_rate": 4.784990456768552e-05, "loss": 0.3233, "step": 11388 }, { "epoch": 1.17, "learning_rate": 4.784792619513544e-05, "loss": 0.7686, "step": 11389 }, { "epoch": 1.17, "learning_rate": 4.7845947702437285e-05, "loss": 0.2826, "step": 11390 }, { "epoch": 1.17, "learning_rate": 4.7843969089604403e-05, "loss": 0.6884, "step": 11391 }, { "epoch": 1.17, "learning_rate": 4.784199035665008e-05, "loss": 0.6845, "step": 11392 }, { "epoch": 1.17, "learning_rate": 4.784001150358766e-05, "loss": 0.337, "step": 11393 }, { "epoch": 1.17, "learning_rate": 4.783803253043045e-05, "loss": 0.2707, "step": 11394 }, { "epoch": 1.17, "learning_rate": 4.783605343719179e-05, "loss": 0.6369, "step": 11395 }, { "epoch": 1.17, "learning_rate": 4.7834074223884996e-05, "loss": 0.7156, "step": 11396 }, { "epoch": 1.17, "learning_rate": 4.783209489052337e-05, "loss": 0.3464, "step": 11397 }, { "epoch": 1.17, "learning_rate": 4.783011543712026e-05, "loss": 0.379, "step": 11398 }, { "epoch": 1.18, "learning_rate": 4.782813586368899e-05, "loss": 1.307, "step": 11399 }, { "epoch": 1.18, "learning_rate": 4.7826156170242874e-05, "loss": 0.6705, "step": 11400 }, { "epoch": 1.18, "learning_rate": 4.782417635679525e-05, "loss": 1.1004, "step": 11401 }, { "epoch": 1.18, "learning_rate": 4.782219642335944e-05, "loss": 0.3213, "step": 11402 }, { "epoch": 1.18, "learning_rate": 4.7820216369948765e-05, "loss": 0.8506, "step": 11403 }, { "epoch": 1.18, "learning_rate": 4.7818236196576575e-05, "loss": 1.1176, "step": 11404 }, { "epoch": 1.18, "learning_rate": 4.7816255903256184e-05, "loss": 0.8617, "step": 11405 }, { "epoch": 1.18, "learning_rate": 4.781427549000093e-05, "loss": 0.506, "step": 11406 }, { "epoch": 1.18, "learning_rate": 4.781229495682414e-05, "loss": 0.5284, "step": 11407 }, { "epoch": 1.18, "learning_rate": 4.781031430373915e-05, "loss": 0.2063, "step": 11408 }, { "epoch": 1.18, "learning_rate": 4.7808333530759286e-05, "loss": 1.0965, "step": 11409 }, { "epoch": 1.18, "learning_rate": 4.780635263789789e-05, "loss": 0.5302, "step": 11410 }, { "epoch": 1.18, "learning_rate": 4.780437162516829e-05, "loss": 0.9403, "step": 11411 }, { "epoch": 1.18, "learning_rate": 4.7802390492583826e-05, "loss": 1.2007, "step": 11412 }, { "epoch": 1.18, "learning_rate": 4.7800409240157844e-05, "loss": 0.6262, "step": 11413 }, { "epoch": 1.18, "learning_rate": 4.779842786790367e-05, "loss": 0.8796, "step": 11414 }, { "epoch": 1.18, "learning_rate": 4.779644637583464e-05, "loss": 0.5366, "step": 11415 }, { "epoch": 1.18, "learning_rate": 4.77944647639641e-05, "loss": 0.5225, "step": 11416 }, { "epoch": 1.18, "learning_rate": 4.779248303230538e-05, "loss": 1.2319, "step": 11417 }, { "epoch": 1.18, "learning_rate": 4.779050118087183e-05, "loss": 0.4705, "step": 11418 }, { "epoch": 1.18, "learning_rate": 4.7788519209676796e-05, "loss": 0.2019, "step": 11419 }, { "epoch": 1.18, "learning_rate": 4.778653711873361e-05, "loss": 1.2276, "step": 11420 }, { "epoch": 1.18, "learning_rate": 4.778455490805561e-05, "loss": 0.4383, "step": 11421 }, { "epoch": 1.18, "learning_rate": 4.778257257765616e-05, "loss": 0.3911, "step": 11422 }, { "epoch": 1.18, "learning_rate": 4.778059012754857e-05, "loss": 0.4453, "step": 11423 }, { "epoch": 1.18, "learning_rate": 4.7778607557746235e-05, "loss": 0.5145, "step": 11424 }, { "epoch": 1.18, "learning_rate": 4.777662486826246e-05, "loss": 0.6612, "step": 11425 }, { "epoch": 1.18, "learning_rate": 4.7774642059110604e-05, "loss": 0.5294, "step": 11426 }, { "epoch": 1.18, "learning_rate": 4.7772659130304024e-05, "loss": 0.5692, "step": 11427 }, { "epoch": 1.18, "learning_rate": 4.777067608185606e-05, "loss": 1.494, "step": 11428 }, { "epoch": 1.18, "learning_rate": 4.7768692913780065e-05, "loss": 0.8296, "step": 11429 }, { "epoch": 1.18, "learning_rate": 4.776670962608937e-05, "loss": 0.1688, "step": 11430 }, { "epoch": 1.18, "learning_rate": 4.7764726218797356e-05, "loss": 0.77, "step": 11431 }, { "epoch": 1.18, "learning_rate": 4.776274269191735e-05, "loss": 0.4643, "step": 11432 }, { "epoch": 1.18, "learning_rate": 4.776075904546274e-05, "loss": 0.6686, "step": 11433 }, { "epoch": 1.18, "learning_rate": 4.775877527944685e-05, "loss": 0.645, "step": 11434 }, { "epoch": 1.18, "learning_rate": 4.7756791393883025e-05, "loss": 0.7663, "step": 11435 }, { "epoch": 1.18, "learning_rate": 4.7754807388784645e-05, "loss": 1.0488, "step": 11436 }, { "epoch": 1.18, "learning_rate": 4.775282326416505e-05, "loss": 0.1846, "step": 11437 }, { "epoch": 1.18, "learning_rate": 4.775083902003761e-05, "loss": 0.9394, "step": 11438 }, { "epoch": 1.18, "learning_rate": 4.774885465641567e-05, "loss": 0.9584, "step": 11439 }, { "epoch": 1.18, "learning_rate": 4.77468701733126e-05, "loss": 0.2759, "step": 11440 }, { "epoch": 1.18, "learning_rate": 4.774488557074174e-05, "loss": 0.4329, "step": 11441 }, { "epoch": 1.18, "learning_rate": 4.774290084871647e-05, "loss": 0.4706, "step": 11442 }, { "epoch": 1.18, "learning_rate": 4.774091600725014e-05, "loss": 0.7631, "step": 11443 }, { "epoch": 1.18, "learning_rate": 4.773893104635612e-05, "loss": 1.3144, "step": 11444 }, { "epoch": 1.18, "learning_rate": 4.773694596604777e-05, "loss": 0.2381, "step": 11445 }, { "epoch": 1.18, "learning_rate": 4.773496076633844e-05, "loss": 0.7173, "step": 11446 }, { "epoch": 1.18, "learning_rate": 4.7732975447241504e-05, "loss": 1.4101, "step": 11447 }, { "epoch": 1.18, "learning_rate": 4.773099000877033e-05, "loss": 0.9974, "step": 11448 }, { "epoch": 1.18, "learning_rate": 4.772900445093828e-05, "loss": 0.2803, "step": 11449 }, { "epoch": 1.18, "learning_rate": 4.7727018773758724e-05, "loss": 0.3187, "step": 11450 }, { "epoch": 1.18, "learning_rate": 4.7725032977245014e-05, "loss": 0.788, "step": 11451 }, { "epoch": 1.18, "learning_rate": 4.7723047061410545e-05, "loss": 0.5946, "step": 11452 }, { "epoch": 1.18, "learning_rate": 4.772106102626865e-05, "loss": 0.8571, "step": 11453 }, { "epoch": 1.18, "learning_rate": 4.7719074871832734e-05, "loss": 0.2129, "step": 11454 }, { "epoch": 1.18, "learning_rate": 4.771708859811615e-05, "loss": 0.7425, "step": 11455 }, { "epoch": 1.18, "learning_rate": 4.771510220513226e-05, "loss": 0.248, "step": 11456 }, { "epoch": 1.18, "learning_rate": 4.771311569289447e-05, "loss": 0.3851, "step": 11457 }, { "epoch": 1.18, "learning_rate": 4.771112906141611e-05, "loss": 1.161, "step": 11458 }, { "epoch": 1.18, "learning_rate": 4.770914231071058e-05, "loss": 0.4984, "step": 11459 }, { "epoch": 1.18, "learning_rate": 4.770715544079125e-05, "loss": 0.7118, "step": 11460 }, { "epoch": 1.18, "learning_rate": 4.770516845167149e-05, "loss": 1.145, "step": 11461 }, { "epoch": 1.18, "learning_rate": 4.7703181343364675e-05, "loss": 0.7109, "step": 11462 }, { "epoch": 1.18, "learning_rate": 4.7701194115884184e-05, "loss": 0.5823, "step": 11463 }, { "epoch": 1.18, "learning_rate": 4.769920676924341e-05, "loss": 0.8533, "step": 11464 }, { "epoch": 1.18, "learning_rate": 4.7697219303455706e-05, "loss": 0.8397, "step": 11465 }, { "epoch": 1.18, "learning_rate": 4.769523171853446e-05, "loss": 1.2882, "step": 11466 }, { "epoch": 1.18, "learning_rate": 4.7693244014493065e-05, "loss": 0.401, "step": 11467 }, { "epoch": 1.18, "learning_rate": 4.769125619134487e-05, "loss": 0.7576, "step": 11468 }, { "epoch": 1.18, "learning_rate": 4.76892682491033e-05, "loss": 0.4022, "step": 11469 }, { "epoch": 1.18, "learning_rate": 4.7687280187781714e-05, "loss": 0.3785, "step": 11470 }, { "epoch": 1.18, "learning_rate": 4.768529200739349e-05, "loss": 0.513, "step": 11471 }, { "epoch": 1.18, "learning_rate": 4.768330370795202e-05, "loss": 0.3229, "step": 11472 }, { "epoch": 1.18, "learning_rate": 4.7681315289470685e-05, "loss": 0.3496, "step": 11473 }, { "epoch": 1.18, "learning_rate": 4.7679326751962864e-05, "loss": 0.2834, "step": 11474 }, { "epoch": 1.18, "learning_rate": 4.767733809544196e-05, "loss": 0.8723, "step": 11475 }, { "epoch": 1.18, "learning_rate": 4.767534931992136e-05, "loss": 0.3549, "step": 11476 }, { "epoch": 1.18, "learning_rate": 4.767336042541444e-05, "loss": 0.9017, "step": 11477 }, { "epoch": 1.18, "learning_rate": 4.7671371411934585e-05, "loss": 0.346, "step": 11478 }, { "epoch": 1.18, "learning_rate": 4.7669382279495196e-05, "loss": 0.9896, "step": 11479 }, { "epoch": 1.18, "learning_rate": 4.766739302810967e-05, "loss": 0.8201, "step": 11480 }, { "epoch": 1.18, "learning_rate": 4.7665403657791375e-05, "loss": 0.6773, "step": 11481 }, { "epoch": 1.18, "learning_rate": 4.7663414168553715e-05, "loss": 0.4751, "step": 11482 }, { "epoch": 1.18, "learning_rate": 4.766142456041009e-05, "loss": 0.8425, "step": 11483 }, { "epoch": 1.18, "learning_rate": 4.765943483337389e-05, "loss": 0.1759, "step": 11484 }, { "epoch": 1.18, "learning_rate": 4.76574449874585e-05, "loss": 0.7614, "step": 11485 }, { "epoch": 1.18, "learning_rate": 4.7655455022677317e-05, "loss": 1.2871, "step": 11486 }, { "epoch": 1.18, "learning_rate": 4.765346493904375e-05, "loss": 0.4987, "step": 11487 }, { "epoch": 1.18, "learning_rate": 4.765147473657119e-05, "loss": 0.7702, "step": 11488 }, { "epoch": 1.18, "learning_rate": 4.7649484415273035e-05, "loss": 1.2133, "step": 11489 }, { "epoch": 1.18, "learning_rate": 4.764749397516267e-05, "loss": 0.4934, "step": 11490 }, { "epoch": 1.18, "learning_rate": 4.764550341625351e-05, "loss": 0.7309, "step": 11491 }, { "epoch": 1.18, "learning_rate": 4.7643512738558945e-05, "loss": 0.3857, "step": 11492 }, { "epoch": 1.18, "learning_rate": 4.764152194209239e-05, "loss": 0.8499, "step": 11493 }, { "epoch": 1.18, "learning_rate": 4.763953102686723e-05, "loss": 0.962, "step": 11494 }, { "epoch": 1.19, "learning_rate": 4.763753999289687e-05, "loss": 0.4789, "step": 11495 }, { "epoch": 1.19, "learning_rate": 4.763554884019473e-05, "loss": 0.4838, "step": 11496 }, { "epoch": 1.19, "learning_rate": 4.7633557568774194e-05, "loss": 1.132, "step": 11497 }, { "epoch": 1.19, "learning_rate": 4.763156617864868e-05, "loss": 0.9946, "step": 11498 }, { "epoch": 1.19, "learning_rate": 4.762957466983158e-05, "loss": 0.3806, "step": 11499 }, { "epoch": 1.19, "learning_rate": 4.762758304233632e-05, "loss": 0.2437, "step": 11500 }, { "epoch": 1.19, "learning_rate": 4.7625591296176273e-05, "loss": 0.7212, "step": 11501 }, { "epoch": 1.19, "learning_rate": 4.762359943136489e-05, "loss": 1.1946, "step": 11502 }, { "epoch": 1.19, "learning_rate": 4.762160744791555e-05, "loss": 0.8052, "step": 11503 }, { "epoch": 1.19, "learning_rate": 4.7619615345841684e-05, "loss": 0.6852, "step": 11504 }, { "epoch": 1.19, "learning_rate": 4.761762312515667e-05, "loss": 0.366, "step": 11505 }, { "epoch": 1.19, "learning_rate": 4.761563078587396e-05, "loss": 0.6493, "step": 11506 }, { "epoch": 1.19, "learning_rate": 4.7613638328006934e-05, "loss": 1.2423, "step": 11507 }, { "epoch": 1.19, "learning_rate": 4.761164575156902e-05, "loss": 1.097, "step": 11508 }, { "epoch": 1.19, "learning_rate": 4.760965305657362e-05, "loss": 0.6162, "step": 11509 }, { "epoch": 1.19, "learning_rate": 4.760766024303417e-05, "loss": 0.2149, "step": 11510 }, { "epoch": 1.19, "learning_rate": 4.760566731096405e-05, "loss": 0.8532, "step": 11511 }, { "epoch": 1.19, "learning_rate": 4.760367426037672e-05, "loss": 0.5086, "step": 11512 }, { "epoch": 1.19, "learning_rate": 4.760168109128556e-05, "loss": 0.4071, "step": 11513 }, { "epoch": 1.19, "learning_rate": 4.759968780370401e-05, "loss": 0.9581, "step": 11514 }, { "epoch": 1.19, "learning_rate": 4.759769439764547e-05, "loss": 0.9681, "step": 11515 }, { "epoch": 1.19, "learning_rate": 4.759570087312337e-05, "loss": 0.737, "step": 11516 }, { "epoch": 1.19, "learning_rate": 4.759370723015114e-05, "loss": 0.42, "step": 11517 }, { "epoch": 1.19, "learning_rate": 4.759171346874217e-05, "loss": 0.2591, "step": 11518 }, { "epoch": 1.19, "learning_rate": 4.758971958890992e-05, "loss": 0.9181, "step": 11519 }, { "epoch": 1.19, "learning_rate": 4.758772559066778e-05, "loss": 0.5689, "step": 11520 }, { "epoch": 1.19, "learning_rate": 4.7585731474029194e-05, "loss": 0.3624, "step": 11521 }, { "epoch": 1.19, "learning_rate": 4.7583737239007576e-05, "loss": 0.5071, "step": 11522 }, { "epoch": 1.19, "learning_rate": 4.758174288561636e-05, "loss": 0.1676, "step": 11523 }, { "epoch": 1.19, "learning_rate": 4.757974841386896e-05, "loss": 0.6134, "step": 11524 }, { "epoch": 1.19, "learning_rate": 4.7577753823778805e-05, "loss": 0.446, "step": 11525 }, { "epoch": 1.19, "learning_rate": 4.757575911535932e-05, "loss": 0.2119, "step": 11526 }, { "epoch": 1.19, "learning_rate": 4.7573764288623946e-05, "loss": 1.2385, "step": 11527 }, { "epoch": 1.19, "learning_rate": 4.7571769343586094e-05, "loss": 0.689, "step": 11528 }, { "epoch": 1.19, "learning_rate": 4.75697742802592e-05, "loss": 0.7075, "step": 11529 }, { "epoch": 1.19, "learning_rate": 4.756777909865672e-05, "loss": 0.7354, "step": 11530 }, { "epoch": 1.19, "learning_rate": 4.7565783798792034e-05, "loss": 0.3018, "step": 11531 }, { "epoch": 1.19, "learning_rate": 4.756378838067861e-05, "loss": 0.5982, "step": 11532 }, { "epoch": 1.19, "learning_rate": 4.756179284432988e-05, "loss": 0.3084, "step": 11533 }, { "epoch": 1.19, "learning_rate": 4.755979718975926e-05, "loss": 0.3279, "step": 11534 }, { "epoch": 1.19, "learning_rate": 4.75578014169802e-05, "loss": 1.1356, "step": 11535 }, { "epoch": 1.19, "learning_rate": 4.755580552600613e-05, "loss": 1.3483, "step": 11536 }, { "epoch": 1.19, "learning_rate": 4.755380951685048e-05, "loss": 0.17, "step": 11537 }, { "epoch": 1.19, "learning_rate": 4.75518133895267e-05, "loss": 0.6753, "step": 11538 }, { "epoch": 1.19, "learning_rate": 4.754981714404821e-05, "loss": 0.3876, "step": 11539 }, { "epoch": 1.19, "learning_rate": 4.7547820780428463e-05, "loss": 0.8386, "step": 11540 }, { "epoch": 1.19, "learning_rate": 4.7545824298680894e-05, "loss": 0.88, "step": 11541 }, { "epoch": 1.19, "learning_rate": 4.754382769881893e-05, "loss": 0.2838, "step": 11542 }, { "epoch": 1.19, "learning_rate": 4.7541830980856036e-05, "loss": 0.6065, "step": 11543 }, { "epoch": 1.19, "learning_rate": 4.753983414480563e-05, "loss": 1.1795, "step": 11544 }, { "epoch": 1.19, "learning_rate": 4.753783719068117e-05, "loss": 0.7499, "step": 11545 }, { "epoch": 1.19, "learning_rate": 4.753584011849609e-05, "loss": 1.3502, "step": 11546 }, { "epoch": 1.19, "learning_rate": 4.753384292826384e-05, "loss": 1.0612, "step": 11547 }, { "epoch": 1.19, "learning_rate": 4.753184561999786e-05, "loss": 0.6733, "step": 11548 }, { "epoch": 1.19, "learning_rate": 4.75298481937116e-05, "loss": 0.6124, "step": 11549 }, { "epoch": 1.19, "learning_rate": 4.75278506494185e-05, "loss": 0.8036, "step": 11550 }, { "epoch": 1.19, "learning_rate": 4.7525852987132014e-05, "loss": 0.1453, "step": 11551 }, { "epoch": 1.19, "learning_rate": 4.752385520686559e-05, "loss": 1.1261, "step": 11552 }, { "epoch": 1.19, "learning_rate": 4.752185730863267e-05, "loss": 0.3086, "step": 11553 }, { "epoch": 1.19, "learning_rate": 4.7519859292446704e-05, "loss": 0.2436, "step": 11554 }, { "epoch": 1.19, "learning_rate": 4.7517861158321145e-05, "loss": 0.871, "step": 11555 }, { "epoch": 1.19, "learning_rate": 4.7515862906269444e-05, "loss": 0.7328, "step": 11556 }, { "epoch": 1.19, "learning_rate": 4.751386453630505e-05, "loss": 0.3014, "step": 11557 }, { "epoch": 1.19, "learning_rate": 4.751186604844142e-05, "loss": 0.2356, "step": 11558 }, { "epoch": 1.19, "learning_rate": 4.7509867442692006e-05, "loss": 1.1472, "step": 11559 }, { "epoch": 1.19, "learning_rate": 4.750786871907026e-05, "loss": 0.4494, "step": 11560 }, { "epoch": 1.19, "learning_rate": 4.7505869877589644e-05, "loss": 0.376, "step": 11561 }, { "epoch": 1.19, "learning_rate": 4.7503870918263596e-05, "loss": 0.8551, "step": 11562 }, { "epoch": 1.19, "learning_rate": 4.75018718411056e-05, "loss": 0.6884, "step": 11563 }, { "epoch": 1.19, "learning_rate": 4.749987264612909e-05, "loss": 0.1896, "step": 11564 }, { "epoch": 1.19, "learning_rate": 4.749787333334753e-05, "loss": 0.6313, "step": 11565 }, { "epoch": 1.19, "learning_rate": 4.749587390277438e-05, "loss": 0.6334, "step": 11566 }, { "epoch": 1.19, "learning_rate": 4.749387435442311e-05, "loss": 1.0717, "step": 11567 }, { "epoch": 1.19, "learning_rate": 4.749187468830716e-05, "loss": 0.6354, "step": 11568 }, { "epoch": 1.19, "learning_rate": 4.7489874904440005e-05, "loss": 0.9747, "step": 11569 }, { "epoch": 1.19, "learning_rate": 4.748787500283511e-05, "loss": 0.4984, "step": 11570 }, { "epoch": 1.19, "learning_rate": 4.748587498350593e-05, "loss": 0.962, "step": 11571 }, { "epoch": 1.19, "learning_rate": 4.7483874846465927e-05, "loss": 0.236, "step": 11572 }, { "epoch": 1.19, "learning_rate": 4.7481874591728576e-05, "loss": 0.9464, "step": 11573 }, { "epoch": 1.19, "learning_rate": 4.747987421930733e-05, "loss": 0.8911, "step": 11574 }, { "epoch": 1.19, "learning_rate": 4.747787372921566e-05, "loss": 0.4825, "step": 11575 }, { "epoch": 1.19, "learning_rate": 4.7475873121467035e-05, "loss": 1.2038, "step": 11576 }, { "epoch": 1.19, "learning_rate": 4.747387239607493e-05, "loss": 0.5733, "step": 11577 }, { "epoch": 1.19, "learning_rate": 4.747187155305279e-05, "loss": 0.8097, "step": 11578 }, { "epoch": 1.19, "learning_rate": 4.746987059241411e-05, "loss": 1.0324, "step": 11579 }, { "epoch": 1.19, "learning_rate": 4.746786951417234e-05, "loss": 0.6019, "step": 11580 }, { "epoch": 1.19, "learning_rate": 4.746586831834096e-05, "loss": 0.6216, "step": 11581 }, { "epoch": 1.19, "learning_rate": 4.746386700493345e-05, "loss": 0.3145, "step": 11582 }, { "epoch": 1.19, "learning_rate": 4.7461865573963265e-05, "loss": 0.2023, "step": 11583 }, { "epoch": 1.19, "learning_rate": 4.745986402544389e-05, "loss": 0.3941, "step": 11584 }, { "epoch": 1.19, "learning_rate": 4.7457862359388796e-05, "loss": 0.7732, "step": 11585 }, { "epoch": 1.19, "learning_rate": 4.745586057581146e-05, "loss": 0.2472, "step": 11586 }, { "epoch": 1.19, "learning_rate": 4.7453858674725346e-05, "loss": 1.4194, "step": 11587 }, { "epoch": 1.19, "learning_rate": 4.745185665614395e-05, "loss": 0.4277, "step": 11588 }, { "epoch": 1.19, "learning_rate": 4.7449854520080725e-05, "loss": 0.4464, "step": 11589 }, { "epoch": 1.19, "learning_rate": 4.744785226654918e-05, "loss": 0.2925, "step": 11590 }, { "epoch": 1.2, "learning_rate": 4.744584989556277e-05, "loss": 0.5726, "step": 11591 }, { "epoch": 1.2, "learning_rate": 4.7443847407134975e-05, "loss": 0.6532, "step": 11592 }, { "epoch": 1.2, "learning_rate": 4.744184480127928e-05, "loss": 0.2225, "step": 11593 }, { "epoch": 1.2, "learning_rate": 4.743984207800917e-05, "loss": 0.4943, "step": 11594 }, { "epoch": 1.2, "learning_rate": 4.7437839237338125e-05, "loss": 0.3406, "step": 11595 }, { "epoch": 1.2, "learning_rate": 4.7435836279279633e-05, "loss": 0.6736, "step": 11596 }, { "epoch": 1.2, "learning_rate": 4.743383320384716e-05, "loss": 0.9491, "step": 11597 }, { "epoch": 1.2, "learning_rate": 4.74318300110542e-05, "loss": 0.5337, "step": 11598 }, { "epoch": 1.2, "learning_rate": 4.7429826700914243e-05, "loss": 0.5964, "step": 11599 }, { "epoch": 1.2, "learning_rate": 4.742782327344077e-05, "loss": 0.9471, "step": 11600 }, { "epoch": 1.2, "learning_rate": 4.742581972864727e-05, "loss": 0.8053, "step": 11601 }, { "epoch": 1.2, "learning_rate": 4.742381606654724e-05, "loss": 0.44, "step": 11602 }, { "epoch": 1.2, "learning_rate": 4.742181228715414e-05, "loss": 0.7972, "step": 11603 }, { "epoch": 1.2, "learning_rate": 4.741980839048148e-05, "loss": 0.5389, "step": 11604 }, { "epoch": 1.2, "learning_rate": 4.7417804376542747e-05, "loss": 0.6435, "step": 11605 }, { "epoch": 1.2, "learning_rate": 4.7415800245351436e-05, "loss": 0.3233, "step": 11606 }, { "epoch": 1.2, "learning_rate": 4.741379599692103e-05, "loss": 0.3444, "step": 11607 }, { "epoch": 1.2, "learning_rate": 4.741179163126501e-05, "loss": 0.7076, "step": 11608 }, { "epoch": 1.2, "learning_rate": 4.74097871483969e-05, "loss": 0.9558, "step": 11609 }, { "epoch": 1.2, "learning_rate": 4.7407782548330174e-05, "loss": 0.9237, "step": 11610 }, { "epoch": 1.2, "learning_rate": 4.7405777831078326e-05, "loss": 0.5927, "step": 11611 }, { "epoch": 1.2, "learning_rate": 4.740377299665486e-05, "loss": 0.7148, "step": 11612 }, { "epoch": 1.2, "learning_rate": 4.740176804507325e-05, "loss": 0.3348, "step": 11613 }, { "epoch": 1.2, "learning_rate": 4.739976297634702e-05, "loss": 0.5239, "step": 11614 }, { "epoch": 1.2, "learning_rate": 4.739775779048966e-05, "loss": 0.38, "step": 11615 }, { "epoch": 1.2, "learning_rate": 4.7395752487514664e-05, "loss": 1.1739, "step": 11616 }, { "epoch": 1.2, "learning_rate": 4.739374706743553e-05, "loss": 0.326, "step": 11617 }, { "epoch": 1.2, "learning_rate": 4.739174153026576e-05, "loss": 0.8983, "step": 11618 }, { "epoch": 1.2, "learning_rate": 4.738973587601885e-05, "loss": 0.315, "step": 11619 }, { "epoch": 1.2, "learning_rate": 4.738773010470832e-05, "loss": 0.5433, "step": 11620 }, { "epoch": 1.2, "learning_rate": 4.738572421634765e-05, "loss": 0.5863, "step": 11621 }, { "epoch": 1.2, "learning_rate": 4.738371821095036e-05, "loss": 0.1936, "step": 11622 }, { "epoch": 1.2, "learning_rate": 4.738171208852994e-05, "loss": 0.961, "step": 11623 }, { "epoch": 1.2, "learning_rate": 4.73797058490999e-05, "loss": 0.7153, "step": 11624 }, { "epoch": 1.2, "learning_rate": 4.737769949267375e-05, "loss": 0.5577, "step": 11625 }, { "epoch": 1.2, "learning_rate": 4.7375693019264994e-05, "loss": 0.3333, "step": 11626 }, { "epoch": 1.2, "learning_rate": 4.737368642888713e-05, "loss": 0.2088, "step": 11627 }, { "epoch": 1.2, "learning_rate": 4.737167972155369e-05, "loss": 0.3908, "step": 11628 }, { "epoch": 1.2, "learning_rate": 4.7369672897278154e-05, "loss": 0.2324, "step": 11629 }, { "epoch": 1.2, "learning_rate": 4.736766595607405e-05, "loss": 0.6002, "step": 11630 }, { "epoch": 1.2, "learning_rate": 4.736565889795488e-05, "loss": 0.8712, "step": 11631 }, { "epoch": 1.2, "learning_rate": 4.736365172293415e-05, "loss": 0.9214, "step": 11632 }, { "epoch": 1.2, "learning_rate": 4.73616444310254e-05, "loss": 0.392, "step": 11633 }, { "epoch": 1.2, "learning_rate": 4.7359637022242104e-05, "loss": 0.8988, "step": 11634 }, { "epoch": 1.2, "learning_rate": 4.735762949659779e-05, "loss": 0.4444, "step": 11635 }, { "epoch": 1.2, "learning_rate": 4.7355621854106e-05, "loss": 1.9836, "step": 11636 }, { "epoch": 1.2, "learning_rate": 4.73536140947802e-05, "loss": 1.1804, "step": 11637 }, { "epoch": 1.2, "learning_rate": 4.735160621863394e-05, "loss": 0.5271, "step": 11638 }, { "epoch": 1.2, "learning_rate": 4.734959822568073e-05, "loss": 0.2215, "step": 11639 }, { "epoch": 1.2, "learning_rate": 4.7347590115934084e-05, "loss": 0.4358, "step": 11640 }, { "epoch": 1.2, "learning_rate": 4.7345581889407516e-05, "loss": 0.5689, "step": 11641 }, { "epoch": 1.2, "learning_rate": 4.734357354611454e-05, "loss": 1.145, "step": 11642 }, { "epoch": 1.2, "learning_rate": 4.7341565086068703e-05, "loss": 0.7816, "step": 11643 }, { "epoch": 1.2, "learning_rate": 4.7339556509283496e-05, "loss": 0.6287, "step": 11644 }, { "epoch": 1.2, "learning_rate": 4.733754781577246e-05, "loss": 0.7812, "step": 11645 }, { "epoch": 1.2, "learning_rate": 4.7335539005549104e-05, "loss": 0.9862, "step": 11646 }, { "epoch": 1.2, "learning_rate": 4.733353007862695e-05, "loss": 1.1484, "step": 11647 }, { "epoch": 1.2, "learning_rate": 4.733152103501954e-05, "loss": 0.1976, "step": 11648 }, { "epoch": 1.2, "learning_rate": 4.732951187474037e-05, "loss": 0.1003, "step": 11649 }, { "epoch": 1.2, "learning_rate": 4.732750259780299e-05, "loss": 0.9017, "step": 11650 }, { "epoch": 1.2, "learning_rate": 4.732549320422092e-05, "loss": 0.6211, "step": 11651 }, { "epoch": 1.2, "learning_rate": 4.732348369400768e-05, "loss": 0.3013, "step": 11652 }, { "epoch": 1.2, "learning_rate": 4.73214740671768e-05, "loss": 0.2311, "step": 11653 }, { "epoch": 1.2, "learning_rate": 4.731946432374181e-05, "loss": 1.0192, "step": 11654 }, { "epoch": 1.2, "learning_rate": 4.731745446371624e-05, "loss": 1.5279, "step": 11655 }, { "epoch": 1.2, "learning_rate": 4.7315444487113616e-05, "loss": 0.3746, "step": 11656 }, { "epoch": 1.2, "learning_rate": 4.7313434393947475e-05, "loss": 0.5678, "step": 11657 }, { "epoch": 1.2, "learning_rate": 4.7311424184231334e-05, "loss": 0.499, "step": 11658 }, { "epoch": 1.2, "learning_rate": 4.730941385797875e-05, "loss": 0.3566, "step": 11659 }, { "epoch": 1.2, "learning_rate": 4.7307403415203233e-05, "loss": 0.2377, "step": 11660 }, { "epoch": 1.2, "learning_rate": 4.7305392855918324e-05, "loss": 0.5854, "step": 11661 }, { "epoch": 1.2, "learning_rate": 4.730338218013757e-05, "loss": 0.505, "step": 11662 }, { "epoch": 1.2, "learning_rate": 4.730137138787449e-05, "loss": 0.5716, "step": 11663 }, { "epoch": 1.2, "learning_rate": 4.7299360479142625e-05, "loss": 0.4094, "step": 11664 }, { "epoch": 1.2, "learning_rate": 4.729734945395552e-05, "loss": 0.5213, "step": 11665 }, { "epoch": 1.2, "learning_rate": 4.7295338312326694e-05, "loss": 0.413, "step": 11666 }, { "epoch": 1.2, "learning_rate": 4.72933270542697e-05, "loss": 1.3126, "step": 11667 }, { "epoch": 1.2, "learning_rate": 4.729131567979808e-05, "loss": 0.2172, "step": 11668 }, { "epoch": 1.2, "learning_rate": 4.7289304188925365e-05, "loss": 0.7355, "step": 11669 }, { "epoch": 1.2, "learning_rate": 4.72872925816651e-05, "loss": 0.8107, "step": 11670 }, { "epoch": 1.2, "learning_rate": 4.728528085803083e-05, "loss": 0.7122, "step": 11671 }, { "epoch": 1.2, "learning_rate": 4.72832690180361e-05, "loss": 0.9351, "step": 11672 }, { "epoch": 1.2, "learning_rate": 4.728125706169443e-05, "loss": 0.0642, "step": 11673 }, { "epoch": 1.2, "learning_rate": 4.7279244989019386e-05, "loss": 0.3201, "step": 11674 }, { "epoch": 1.2, "learning_rate": 4.7277232800024505e-05, "loss": 0.2838, "step": 11675 }, { "epoch": 1.2, "learning_rate": 4.727522049472335e-05, "loss": 0.2167, "step": 11676 }, { "epoch": 1.2, "learning_rate": 4.7273208073129445e-05, "loss": 0.9951, "step": 11677 }, { "epoch": 1.2, "learning_rate": 4.7271195535256344e-05, "loss": 0.3754, "step": 11678 }, { "epoch": 1.2, "learning_rate": 4.7269182881117596e-05, "loss": 1.335, "step": 11679 }, { "epoch": 1.2, "learning_rate": 4.726717011072674e-05, "loss": 0.6658, "step": 11680 }, { "epoch": 1.2, "learning_rate": 4.726515722409735e-05, "loss": 0.876, "step": 11681 }, { "epoch": 1.2, "learning_rate": 4.726314422124295e-05, "loss": 0.4146, "step": 11682 }, { "epoch": 1.2, "learning_rate": 4.7261131102177116e-05, "loss": 0.6862, "step": 11683 }, { "epoch": 1.2, "learning_rate": 4.7259117866913376e-05, "loss": 0.2035, "step": 11684 }, { "epoch": 1.2, "learning_rate": 4.72571045154653e-05, "loss": 0.6365, "step": 11685 }, { "epoch": 1.21, "learning_rate": 4.725509104784642e-05, "loss": 0.5537, "step": 11686 }, { "epoch": 1.21, "learning_rate": 4.725307746407032e-05, "loss": 0.7194, "step": 11687 }, { "epoch": 1.21, "learning_rate": 4.725106376415053e-05, "loss": 0.4219, "step": 11688 }, { "epoch": 1.21, "learning_rate": 4.7249049948100615e-05, "loss": 1.1792, "step": 11689 }, { "epoch": 1.21, "learning_rate": 4.7247036015934135e-05, "loss": 0.4719, "step": 11690 }, { "epoch": 1.21, "learning_rate": 4.7245021967664645e-05, "loss": 0.7831, "step": 11691 }, { "epoch": 1.21, "learning_rate": 4.72430078033057e-05, "loss": 0.2806, "step": 11692 }, { "epoch": 1.21, "learning_rate": 4.724099352287086e-05, "loss": 0.2623, "step": 11693 }, { "epoch": 1.21, "learning_rate": 4.7238979126373686e-05, "loss": 0.7541, "step": 11694 }, { "epoch": 1.21, "learning_rate": 4.7236964613827745e-05, "loss": 0.9986, "step": 11695 }, { "epoch": 1.21, "learning_rate": 4.7234949985246584e-05, "loss": 0.7259, "step": 11696 }, { "epoch": 1.21, "learning_rate": 4.723293524064378e-05, "loss": 0.5631, "step": 11697 }, { "epoch": 1.21, "learning_rate": 4.7230920380032875e-05, "loss": 1.3479, "step": 11698 }, { "epoch": 1.21, "learning_rate": 4.722890540342745e-05, "loss": 0.3687, "step": 11699 }, { "epoch": 1.21, "learning_rate": 4.7226890310841074e-05, "loss": 0.1673, "step": 11700 }, { "epoch": 1.21, "learning_rate": 4.72248751022873e-05, "loss": 0.8364, "step": 11701 }, { "epoch": 1.21, "learning_rate": 4.722285977777969e-05, "loss": 0.3707, "step": 11702 }, { "epoch": 1.21, "learning_rate": 4.722084433733182e-05, "loss": 0.7527, "step": 11703 }, { "epoch": 1.21, "learning_rate": 4.721882878095726e-05, "loss": 0.6604, "step": 11704 }, { "epoch": 1.21, "learning_rate": 4.7216813108669564e-05, "loss": 0.2451, "step": 11705 }, { "epoch": 1.21, "learning_rate": 4.721479732048232e-05, "loss": 0.8517, "step": 11706 }, { "epoch": 1.21, "learning_rate": 4.721278141640909e-05, "loss": 0.9839, "step": 11707 }, { "epoch": 1.21, "learning_rate": 4.721076539646343e-05, "loss": 1.1377, "step": 11708 }, { "epoch": 1.21, "learning_rate": 4.720874926065894e-05, "loss": 0.6211, "step": 11709 }, { "epoch": 1.21, "learning_rate": 4.720673300900917e-05, "loss": 0.2027, "step": 11710 }, { "epoch": 1.21, "learning_rate": 4.72047166415277e-05, "loss": 0.329, "step": 11711 }, { "epoch": 1.21, "learning_rate": 4.72027001582281e-05, "loss": 0.3945, "step": 11712 }, { "epoch": 1.21, "learning_rate": 4.720068355912395e-05, "loss": 0.5012, "step": 11713 }, { "epoch": 1.21, "learning_rate": 4.719866684422882e-05, "loss": 0.7684, "step": 11714 }, { "epoch": 1.21, "learning_rate": 4.71966500135563e-05, "loss": 0.3998, "step": 11715 }, { "epoch": 1.21, "learning_rate": 4.719463306711995e-05, "loss": 0.5319, "step": 11716 }, { "epoch": 1.21, "learning_rate": 4.719261600493335e-05, "loss": 0.3289, "step": 11717 }, { "epoch": 1.21, "learning_rate": 4.719059882701008e-05, "loss": 0.3404, "step": 11718 }, { "epoch": 1.21, "learning_rate": 4.718858153336372e-05, "loss": 1.2181, "step": 11719 }, { "epoch": 1.21, "learning_rate": 4.7186564124007854e-05, "loss": 0.9088, "step": 11720 }, { "epoch": 1.21, "learning_rate": 4.7184546598956067e-05, "loss": 0.5032, "step": 11721 }, { "epoch": 1.21, "learning_rate": 4.7182528958221926e-05, "loss": 0.9257, "step": 11722 }, { "epoch": 1.21, "learning_rate": 4.718051120181902e-05, "loss": 0.2367, "step": 11723 }, { "epoch": 1.21, "learning_rate": 4.7178493329760934e-05, "loss": 0.9597, "step": 11724 }, { "epoch": 1.21, "learning_rate": 4.717647534206125e-05, "loss": 1.0004, "step": 11725 }, { "epoch": 1.21, "learning_rate": 4.717445723873355e-05, "loss": 0.6465, "step": 11726 }, { "epoch": 1.21, "learning_rate": 4.717243901979142e-05, "loss": 0.3768, "step": 11727 }, { "epoch": 1.21, "learning_rate": 4.717042068524847e-05, "loss": 0.6976, "step": 11728 }, { "epoch": 1.21, "learning_rate": 4.716840223511824e-05, "loss": 0.4736, "step": 11729 }, { "epoch": 1.21, "learning_rate": 4.716638366941435e-05, "loss": 1.1777, "step": 11730 }, { "epoch": 1.21, "learning_rate": 4.7164364988150385e-05, "loss": 0.4807, "step": 11731 }, { "epoch": 1.21, "learning_rate": 4.716234619133993e-05, "loss": 0.4668, "step": 11732 }, { "epoch": 1.21, "learning_rate": 4.7160327278996576e-05, "loss": 0.635, "step": 11733 }, { "epoch": 1.21, "learning_rate": 4.715830825113391e-05, "loss": 0.419, "step": 11734 }, { "epoch": 1.21, "learning_rate": 4.715628910776554e-05, "loss": 0.9414, "step": 11735 }, { "epoch": 1.21, "learning_rate": 4.715426984890504e-05, "loss": 1.0638, "step": 11736 }, { "epoch": 1.21, "learning_rate": 4.7152250474566e-05, "loss": 0.9121, "step": 11737 }, { "epoch": 1.21, "learning_rate": 4.715023098476203e-05, "loss": 0.6788, "step": 11738 }, { "epoch": 1.21, "learning_rate": 4.7148211379506716e-05, "loss": 1.1429, "step": 11739 }, { "epoch": 1.21, "learning_rate": 4.714619165881366e-05, "loss": 1.0938, "step": 11740 }, { "epoch": 1.21, "learning_rate": 4.714417182269645e-05, "loss": 0.62, "step": 11741 }, { "epoch": 1.21, "learning_rate": 4.7142151871168694e-05, "loss": 0.5471, "step": 11742 }, { "epoch": 1.21, "learning_rate": 4.714013180424398e-05, "loss": 1.0879, "step": 11743 }, { "epoch": 1.21, "learning_rate": 4.71381116219359e-05, "loss": 0.285, "step": 11744 }, { "epoch": 1.21, "learning_rate": 4.713609132425807e-05, "loss": 0.8035, "step": 11745 }, { "epoch": 1.21, "learning_rate": 4.713407091122407e-05, "loss": 0.3732, "step": 11746 }, { "epoch": 1.21, "learning_rate": 4.7132050382847534e-05, "loss": 0.2473, "step": 11747 }, { "epoch": 1.21, "learning_rate": 4.7130029739142035e-05, "loss": 0.9607, "step": 11748 }, { "epoch": 1.21, "learning_rate": 4.712800898012118e-05, "loss": 0.8053, "step": 11749 }, { "epoch": 1.21, "learning_rate": 4.7125988105798585e-05, "loss": 0.2908, "step": 11750 }, { "epoch": 1.21, "learning_rate": 4.7123967116187836e-05, "loss": 0.3885, "step": 11751 }, { "epoch": 1.21, "learning_rate": 4.712194601130255e-05, "loss": 0.7833, "step": 11752 }, { "epoch": 1.21, "learning_rate": 4.7119924791156333e-05, "loss": 0.3446, "step": 11753 }, { "epoch": 1.21, "learning_rate": 4.7117903455762774e-05, "loss": 1.1089, "step": 11754 }, { "epoch": 1.21, "learning_rate": 4.711588200513552e-05, "loss": 0.8708, "step": 11755 }, { "epoch": 1.21, "learning_rate": 4.7113860439288126e-05, "loss": 0.9136, "step": 11756 }, { "epoch": 1.21, "learning_rate": 4.7111838758234244e-05, "loss": 0.3142, "step": 11757 }, { "epoch": 1.21, "learning_rate": 4.710981696198746e-05, "loss": 0.8613, "step": 11758 }, { "epoch": 1.21, "learning_rate": 4.7107795050561405e-05, "loss": 0.9884, "step": 11759 }, { "epoch": 1.21, "learning_rate": 4.710577302396966e-05, "loss": 0.7956, "step": 11760 }, { "epoch": 1.21, "learning_rate": 4.710375088222586e-05, "loss": 0.3905, "step": 11761 }, { "epoch": 1.21, "learning_rate": 4.710172862534361e-05, "loss": 0.7762, "step": 11762 }, { "epoch": 1.21, "learning_rate": 4.7099706253336525e-05, "loss": 0.7239, "step": 11763 }, { "epoch": 1.21, "learning_rate": 4.709768376621822e-05, "loss": 1.0926, "step": 11764 }, { "epoch": 1.21, "learning_rate": 4.709566116400231e-05, "loss": 0.5713, "step": 11765 }, { "epoch": 1.21, "learning_rate": 4.7093638446702405e-05, "loss": 0.3455, "step": 11766 }, { "epoch": 1.21, "learning_rate": 4.709161561433213e-05, "loss": 1.2868, "step": 11767 }, { "epoch": 1.21, "learning_rate": 4.708959266690509e-05, "loss": 0.39, "step": 11768 }, { "epoch": 1.21, "learning_rate": 4.708756960443492e-05, "loss": 1.0123, "step": 11769 }, { "epoch": 1.21, "learning_rate": 4.708554642693522e-05, "loss": 0.6877, "step": 11770 }, { "epoch": 1.21, "learning_rate": 4.708352313441963e-05, "loss": 0.8956, "step": 11771 }, { "epoch": 1.21, "learning_rate": 4.708149972690176e-05, "loss": 0.489, "step": 11772 }, { "epoch": 1.21, "learning_rate": 4.707947620439521e-05, "loss": 0.802, "step": 11773 }, { "epoch": 1.21, "learning_rate": 4.707745256691364e-05, "loss": 1.2131, "step": 11774 }, { "epoch": 1.21, "learning_rate": 4.707542881447066e-05, "loss": 0.3555, "step": 11775 }, { "epoch": 1.21, "learning_rate": 4.707340494707987e-05, "loss": 0.5189, "step": 11776 }, { "epoch": 1.21, "learning_rate": 4.707138096475493e-05, "loss": 0.187, "step": 11777 }, { "epoch": 1.21, "learning_rate": 4.706935686750944e-05, "loss": 0.1347, "step": 11778 }, { "epoch": 1.21, "learning_rate": 4.706733265535704e-05, "loss": 1.3214, "step": 11779 }, { "epoch": 1.21, "learning_rate": 4.7065308328311346e-05, "loss": 0.7645, "step": 11780 }, { "epoch": 1.21, "learning_rate": 4.706328388638598e-05, "loss": 0.17, "step": 11781 }, { "epoch": 1.22, "learning_rate": 4.706125932959459e-05, "loss": 0.7496, "step": 11782 }, { "epoch": 1.22, "learning_rate": 4.705923465795079e-05, "loss": 0.4349, "step": 11783 }, { "epoch": 1.22, "learning_rate": 4.705720987146821e-05, "loss": 0.9639, "step": 11784 }, { "epoch": 1.22, "learning_rate": 4.7055184970160495e-05, "loss": 0.3562, "step": 11785 }, { "epoch": 1.22, "learning_rate": 4.705315995404126e-05, "loss": 1.5885, "step": 11786 }, { "epoch": 1.22, "learning_rate": 4.705113482312414e-05, "loss": 0.2744, "step": 11787 }, { "epoch": 1.22, "learning_rate": 4.7049109577422775e-05, "loss": 0.7658, "step": 11788 }, { "epoch": 1.22, "learning_rate": 4.704708421695079e-05, "loss": 0.8468, "step": 11789 }, { "epoch": 1.22, "learning_rate": 4.704505874172183e-05, "loss": 0.3554, "step": 11790 }, { "epoch": 1.22, "learning_rate": 4.704303315174952e-05, "loss": 1.3692, "step": 11791 }, { "epoch": 1.22, "learning_rate": 4.704100744704749e-05, "loss": 0.6682, "step": 11792 }, { "epoch": 1.22, "learning_rate": 4.703898162762939e-05, "loss": 0.1138, "step": 11793 }, { "epoch": 1.22, "learning_rate": 4.703695569350885e-05, "loss": 1.0678, "step": 11794 }, { "epoch": 1.22, "learning_rate": 4.703492964469951e-05, "loss": 0.7994, "step": 11795 }, { "epoch": 1.22, "learning_rate": 4.703290348121501e-05, "loss": 0.9365, "step": 11796 }, { "epoch": 1.22, "learning_rate": 4.7030877203069e-05, "loss": 1.1716, "step": 11797 }, { "epoch": 1.22, "learning_rate": 4.70288508102751e-05, "loss": 0.4573, "step": 11798 }, { "epoch": 1.22, "learning_rate": 4.702682430284696e-05, "loss": 0.4658, "step": 11799 }, { "epoch": 1.22, "learning_rate": 4.702479768079823e-05, "loss": 0.1656, "step": 11800 }, { "epoch": 1.22, "learning_rate": 4.702277094414253e-05, "loss": 0.9804, "step": 11801 }, { "epoch": 1.22, "learning_rate": 4.702074409289353e-05, "loss": 0.2989, "step": 11802 }, { "epoch": 1.22, "learning_rate": 4.701871712706487e-05, "loss": 0.9491, "step": 11803 }, { "epoch": 1.22, "learning_rate": 4.7016690046670187e-05, "loss": 0.3416, "step": 11804 }, { "epoch": 1.22, "learning_rate": 4.7014662851723116e-05, "loss": 0.8071, "step": 11805 }, { "epoch": 1.22, "learning_rate": 4.701263554223732e-05, "loss": 1.0034, "step": 11806 }, { "epoch": 1.22, "learning_rate": 4.701060811822645e-05, "loss": 0.969, "step": 11807 }, { "epoch": 1.22, "learning_rate": 4.700858057970413e-05, "loss": 0.5226, "step": 11808 }, { "epoch": 1.22, "learning_rate": 4.7006552926684036e-05, "loss": 0.4931, "step": 11809 }, { "epoch": 1.22, "learning_rate": 4.7004525159179814e-05, "loss": 0.7651, "step": 11810 }, { "epoch": 1.22, "learning_rate": 4.70024972772051e-05, "loss": 1.2846, "step": 11811 }, { "epoch": 1.22, "learning_rate": 4.700046928077355e-05, "loss": 0.5484, "step": 11812 }, { "epoch": 1.22, "learning_rate": 4.699844116989882e-05, "loss": 0.1837, "step": 11813 }, { "epoch": 1.22, "learning_rate": 4.6996412944594564e-05, "loss": 0.3833, "step": 11814 }, { "epoch": 1.22, "learning_rate": 4.699438460487443e-05, "loss": 0.1804, "step": 11815 }, { "epoch": 1.22, "learning_rate": 4.699235615075208e-05, "loss": 0.7228, "step": 11816 }, { "epoch": 1.22, "learning_rate": 4.699032758224116e-05, "loss": 0.8832, "step": 11817 }, { "epoch": 1.22, "learning_rate": 4.698829889935533e-05, "loss": 1.2268, "step": 11818 }, { "epoch": 1.22, "learning_rate": 4.698627010210825e-05, "loss": 1.0697, "step": 11819 }, { "epoch": 1.22, "learning_rate": 4.698424119051357e-05, "loss": 0.4076, "step": 11820 }, { "epoch": 1.22, "learning_rate": 4.698221216458495e-05, "loss": 1.2933, "step": 11821 }, { "epoch": 1.22, "learning_rate": 4.6980183024336064e-05, "loss": 0.3668, "step": 11822 }, { "epoch": 1.22, "learning_rate": 4.697815376978054e-05, "loss": 0.2686, "step": 11823 }, { "epoch": 1.22, "learning_rate": 4.697612440093207e-05, "loss": 1.179, "step": 11824 }, { "epoch": 1.22, "learning_rate": 4.69740949178043e-05, "loss": 1.0794, "step": 11825 }, { "epoch": 1.22, "learning_rate": 4.69720653204109e-05, "loss": 0.9068, "step": 11826 }, { "epoch": 1.22, "learning_rate": 4.697003560876552e-05, "loss": 0.317, "step": 11827 }, { "epoch": 1.22, "learning_rate": 4.696800578288184e-05, "loss": 0.456, "step": 11828 }, { "epoch": 1.22, "learning_rate": 4.6965975842773505e-05, "loss": 0.748, "step": 11829 }, { "epoch": 1.22, "learning_rate": 4.696394578845419e-05, "loss": 0.6828, "step": 11830 }, { "epoch": 1.22, "learning_rate": 4.696191561993757e-05, "loss": 0.5039, "step": 11831 }, { "epoch": 1.22, "learning_rate": 4.69598853372373e-05, "loss": 0.8851, "step": 11832 }, { "epoch": 1.22, "learning_rate": 4.695785494036704e-05, "loss": 0.4339, "step": 11833 }, { "epoch": 1.22, "learning_rate": 4.695582442934048e-05, "loss": 1.3628, "step": 11834 }, { "epoch": 1.22, "learning_rate": 4.695379380417128e-05, "loss": 1.0082, "step": 11835 }, { "epoch": 1.22, "learning_rate": 4.6951763064873093e-05, "loss": 0.4372, "step": 11836 }, { "epoch": 1.22, "learning_rate": 4.694973221145962e-05, "loss": 0.1403, "step": 11837 }, { "epoch": 1.22, "learning_rate": 4.694770124394451e-05, "loss": 0.8374, "step": 11838 }, { "epoch": 1.22, "learning_rate": 4.6945670162341425e-05, "loss": 0.7639, "step": 11839 }, { "epoch": 1.22, "learning_rate": 4.6943638966664065e-05, "loss": 0.8186, "step": 11840 }, { "epoch": 1.22, "learning_rate": 4.69416076569261e-05, "loss": 0.6664, "step": 11841 }, { "epoch": 1.22, "learning_rate": 4.693957623314119e-05, "loss": 0.4341, "step": 11842 }, { "epoch": 1.22, "learning_rate": 4.6937544695323013e-05, "loss": 0.1033, "step": 11843 }, { "epoch": 1.22, "learning_rate": 4.693551304348525e-05, "loss": 0.5517, "step": 11844 }, { "epoch": 1.22, "learning_rate": 4.693348127764157e-05, "loss": 1.6887, "step": 11845 }, { "epoch": 1.22, "learning_rate": 4.693144939780567e-05, "loss": 1.3037, "step": 11846 }, { "epoch": 1.22, "learning_rate": 4.6929417403991204e-05, "loss": 0.1917, "step": 11847 }, { "epoch": 1.22, "learning_rate": 4.692738529621187e-05, "loss": 0.52, "step": 11848 }, { "epoch": 1.22, "learning_rate": 4.6925353074481336e-05, "loss": 0.9102, "step": 11849 }, { "epoch": 1.22, "learning_rate": 4.692332073881328e-05, "loss": 0.8459, "step": 11850 }, { "epoch": 1.22, "learning_rate": 4.692128828922139e-05, "loss": 0.5483, "step": 11851 }, { "epoch": 1.22, "learning_rate": 4.6919255725719344e-05, "loss": 0.1944, "step": 11852 }, { "epoch": 1.22, "learning_rate": 4.691722304832084e-05, "loss": 0.182, "step": 11853 }, { "epoch": 1.22, "learning_rate": 4.6915190257039534e-05, "loss": 1.0008, "step": 11854 }, { "epoch": 1.22, "learning_rate": 4.691315735188913e-05, "loss": 0.3696, "step": 11855 }, { "epoch": 1.22, "learning_rate": 4.691112433288331e-05, "loss": 0.925, "step": 11856 }, { "epoch": 1.22, "learning_rate": 4.690909120003576e-05, "loss": 0.5401, "step": 11857 }, { "epoch": 1.22, "learning_rate": 4.690705795336015e-05, "loss": 1.2947, "step": 11858 }, { "epoch": 1.22, "learning_rate": 4.69050245928702e-05, "loss": 0.3113, "step": 11859 }, { "epoch": 1.22, "learning_rate": 4.6902991118579575e-05, "loss": 0.3954, "step": 11860 }, { "epoch": 1.22, "learning_rate": 4.6900957530501966e-05, "loss": 0.7272, "step": 11861 }, { "epoch": 1.22, "learning_rate": 4.689892382865106e-05, "loss": 0.174, "step": 11862 }, { "epoch": 1.22, "learning_rate": 4.689689001304057e-05, "loss": 1.0164, "step": 11863 }, { "epoch": 1.22, "learning_rate": 4.689485608368415e-05, "loss": 1.1054, "step": 11864 }, { "epoch": 1.22, "learning_rate": 4.689282204059552e-05, "loss": 0.1896, "step": 11865 }, { "epoch": 1.22, "learning_rate": 4.689078788378837e-05, "loss": 0.2043, "step": 11866 }, { "epoch": 1.22, "learning_rate": 4.6888753613276386e-05, "loss": 0.1934, "step": 11867 }, { "epoch": 1.22, "learning_rate": 4.6886719229073255e-05, "loss": 0.418, "step": 11868 }, { "epoch": 1.22, "learning_rate": 4.6884684731192686e-05, "loss": 0.934, "step": 11869 }, { "epoch": 1.22, "learning_rate": 4.688265011964838e-05, "loss": 0.8064, "step": 11870 }, { "epoch": 1.22, "learning_rate": 4.6880615394454015e-05, "loss": 0.3234, "step": 11871 }, { "epoch": 1.22, "learning_rate": 4.68785805556233e-05, "loss": 0.6694, "step": 11872 }, { "epoch": 1.22, "learning_rate": 4.687654560316992e-05, "loss": 0.3586, "step": 11873 }, { "epoch": 1.22, "learning_rate": 4.68745105371076e-05, "loss": 0.5331, "step": 11874 }, { "epoch": 1.22, "learning_rate": 4.687247535745001e-05, "loss": 0.5895, "step": 11875 }, { "epoch": 1.22, "learning_rate": 4.6870440064210864e-05, "loss": 1.5233, "step": 11876 }, { "epoch": 1.22, "learning_rate": 4.6868404657403864e-05, "loss": 0.7289, "step": 11877 }, { "epoch": 1.23, "learning_rate": 4.686636913704271e-05, "loss": 0.5133, "step": 11878 }, { "epoch": 1.23, "learning_rate": 4.686433350314112e-05, "loss": 0.3582, "step": 11879 }, { "epoch": 1.23, "learning_rate": 4.686229775571277e-05, "loss": 1.1007, "step": 11880 }, { "epoch": 1.23, "learning_rate": 4.686026189477139e-05, "loss": 0.3689, "step": 11881 }, { "epoch": 1.23, "learning_rate": 4.685822592033065e-05, "loss": 0.872, "step": 11882 }, { "epoch": 1.23, "learning_rate": 4.685618983240429e-05, "loss": 0.7327, "step": 11883 }, { "epoch": 1.23, "learning_rate": 4.6854153631006004e-05, "loss": 0.7415, "step": 11884 }, { "epoch": 1.23, "learning_rate": 4.6852117316149496e-05, "loss": 0.6732, "step": 11885 }, { "epoch": 1.23, "learning_rate": 4.685008088784848e-05, "loss": 0.5168, "step": 11886 }, { "epoch": 1.23, "learning_rate": 4.6848044346116666e-05, "loss": 1.2852, "step": 11887 }, { "epoch": 1.23, "learning_rate": 4.6846007690967754e-05, "loss": 0.507, "step": 11888 }, { "epoch": 1.23, "learning_rate": 4.6843970922415456e-05, "loss": 0.3407, "step": 11889 }, { "epoch": 1.23, "learning_rate": 4.684193404047349e-05, "loss": 0.5912, "step": 11890 }, { "epoch": 1.23, "learning_rate": 4.6839897045155574e-05, "loss": 0.6281, "step": 11891 }, { "epoch": 1.23, "learning_rate": 4.683785993647541e-05, "loss": 0.5035, "step": 11892 }, { "epoch": 1.23, "learning_rate": 4.68358227144467e-05, "loss": 0.384, "step": 11893 }, { "epoch": 1.23, "learning_rate": 4.6833785379083176e-05, "loss": 0.3329, "step": 11894 }, { "epoch": 1.23, "learning_rate": 4.6831747930398556e-05, "loss": 0.5051, "step": 11895 }, { "epoch": 1.23, "learning_rate": 4.682971036840653e-05, "loss": 0.2793, "step": 11896 }, { "epoch": 1.23, "learning_rate": 4.682767269312085e-05, "loss": 1.4316, "step": 11897 }, { "epoch": 1.23, "learning_rate": 4.6825634904555206e-05, "loss": 0.5609, "step": 11898 }, { "epoch": 1.23, "learning_rate": 4.682359700272332e-05, "loss": 0.5856, "step": 11899 }, { "epoch": 1.23, "learning_rate": 4.682155898763892e-05, "loss": 0.7399, "step": 11900 }, { "epoch": 1.23, "learning_rate": 4.681952085931573e-05, "loss": 0.4509, "step": 11901 }, { "epoch": 1.23, "learning_rate": 4.681748261776745e-05, "loss": 0.6725, "step": 11902 }, { "epoch": 1.23, "learning_rate": 4.681544426300781e-05, "loss": 0.7046, "step": 11903 }, { "epoch": 1.23, "learning_rate": 4.6813405795050544e-05, "loss": 0.9641, "step": 11904 }, { "epoch": 1.23, "learning_rate": 4.6811367213909355e-05, "loss": 0.3022, "step": 11905 }, { "epoch": 1.23, "learning_rate": 4.680932851959798e-05, "loss": 0.7069, "step": 11906 }, { "epoch": 1.23, "learning_rate": 4.6807289712130135e-05, "loss": 0.6177, "step": 11907 }, { "epoch": 1.23, "learning_rate": 4.680525079151955e-05, "loss": 0.7315, "step": 11908 }, { "epoch": 1.23, "learning_rate": 4.680321175777995e-05, "loss": 0.4535, "step": 11909 }, { "epoch": 1.23, "learning_rate": 4.6801172610925066e-05, "loss": 0.8004, "step": 11910 }, { "epoch": 1.23, "learning_rate": 4.679913335096861e-05, "loss": 0.4867, "step": 11911 }, { "epoch": 1.23, "learning_rate": 4.679709397792432e-05, "loss": 0.3898, "step": 11912 }, { "epoch": 1.23, "learning_rate": 4.6795054491805926e-05, "loss": 0.5869, "step": 11913 }, { "epoch": 1.23, "learning_rate": 4.6793014892627154e-05, "loss": 0.433, "step": 11914 }, { "epoch": 1.23, "learning_rate": 4.679097518040173e-05, "loss": 0.6961, "step": 11915 }, { "epoch": 1.23, "learning_rate": 4.67889353551434e-05, "loss": 1.1615, "step": 11916 }, { "epoch": 1.23, "learning_rate": 4.6786895416865887e-05, "loss": 0.5235, "step": 11917 }, { "epoch": 1.23, "learning_rate": 4.6784855365582907e-05, "loss": 0.3491, "step": 11918 }, { "epoch": 1.23, "learning_rate": 4.6782815201308225e-05, "loss": 0.1505, "step": 11919 }, { "epoch": 1.23, "learning_rate": 4.6780774924055544e-05, "loss": 0.8616, "step": 11920 }, { "epoch": 1.23, "learning_rate": 4.677873453383861e-05, "loss": 1.0462, "step": 11921 }, { "epoch": 1.23, "learning_rate": 4.677669403067117e-05, "loss": 0.3824, "step": 11922 }, { "epoch": 1.23, "learning_rate": 4.677465341456695e-05, "loss": 0.716, "step": 11923 }, { "epoch": 1.23, "learning_rate": 4.67726126855397e-05, "loss": 0.3259, "step": 11924 }, { "epoch": 1.23, "learning_rate": 4.677057184360313e-05, "loss": 0.9796, "step": 11925 }, { "epoch": 1.23, "learning_rate": 4.6768530888771e-05, "loss": 0.3672, "step": 11926 }, { "epoch": 1.23, "learning_rate": 4.676648982105704e-05, "loss": 0.7342, "step": 11927 }, { "epoch": 1.23, "learning_rate": 4.676444864047499e-05, "loss": 0.9444, "step": 11928 }, { "epoch": 1.23, "learning_rate": 4.6762407347038605e-05, "loss": 0.2451, "step": 11929 }, { "epoch": 1.23, "learning_rate": 4.6760365940761606e-05, "loss": 0.7254, "step": 11930 }, { "epoch": 1.23, "learning_rate": 4.675832442165776e-05, "loss": 0.5315, "step": 11931 }, { "epoch": 1.23, "learning_rate": 4.675628278974078e-05, "loss": 1.0909, "step": 11932 }, { "epoch": 1.23, "learning_rate": 4.675424104502443e-05, "loss": 0.5689, "step": 11933 }, { "epoch": 1.23, "learning_rate": 4.675219918752244e-05, "loss": 0.6128, "step": 11934 }, { "epoch": 1.23, "learning_rate": 4.675015721724859e-05, "loss": 0.4161, "step": 11935 }, { "epoch": 1.23, "learning_rate": 4.674811513421658e-05, "loss": 1.1457, "step": 11936 }, { "epoch": 1.23, "learning_rate": 4.674607293844018e-05, "loss": 1.0961, "step": 11937 }, { "epoch": 1.23, "learning_rate": 4.6744030629933145e-05, "loss": 0.9529, "step": 11938 }, { "epoch": 1.23, "learning_rate": 4.67419882087092e-05, "loss": 0.7807, "step": 11939 }, { "epoch": 1.23, "learning_rate": 4.673994567478211e-05, "loss": 0.1941, "step": 11940 }, { "epoch": 1.23, "learning_rate": 4.6737903028165635e-05, "loss": 0.9075, "step": 11941 }, { "epoch": 1.23, "learning_rate": 4.67358602688735e-05, "loss": 0.8871, "step": 11942 }, { "epoch": 1.23, "learning_rate": 4.673381739691948e-05, "loss": 0.6344, "step": 11943 }, { "epoch": 1.23, "learning_rate": 4.673177441231731e-05, "loss": 0.337, "step": 11944 }, { "epoch": 1.23, "learning_rate": 4.672973131508075e-05, "loss": 0.9122, "step": 11945 }, { "epoch": 1.23, "learning_rate": 4.672768810522355e-05, "loss": 0.5405, "step": 11946 }, { "epoch": 1.23, "learning_rate": 4.6725644782759476e-05, "loss": 0.6062, "step": 11947 }, { "epoch": 1.23, "learning_rate": 4.672360134770227e-05, "loss": 1.3725, "step": 11948 }, { "epoch": 1.23, "learning_rate": 4.6721557800065694e-05, "loss": 1.0137, "step": 11949 }, { "epoch": 1.23, "learning_rate": 4.67195141398635e-05, "loss": 0.4294, "step": 11950 }, { "epoch": 1.23, "learning_rate": 4.671747036710945e-05, "loss": 0.3457, "step": 11951 }, { "epoch": 1.23, "learning_rate": 4.6715426481817305e-05, "loss": 0.6549, "step": 11952 }, { "epoch": 1.23, "learning_rate": 4.6713382484000816e-05, "loss": 0.3552, "step": 11953 }, { "epoch": 1.23, "learning_rate": 4.671133837367375e-05, "loss": 0.6225, "step": 11954 }, { "epoch": 1.23, "learning_rate": 4.6709294150849873e-05, "loss": 0.5901, "step": 11955 }, { "epoch": 1.23, "learning_rate": 4.670724981554292e-05, "loss": 0.7864, "step": 11956 }, { "epoch": 1.23, "learning_rate": 4.670520536776668e-05, "loss": 0.262, "step": 11957 }, { "epoch": 1.23, "learning_rate": 4.6703160807534894e-05, "loss": 0.393, "step": 11958 }, { "epoch": 1.23, "learning_rate": 4.670111613486135e-05, "loss": 0.3873, "step": 11959 }, { "epoch": 1.23, "learning_rate": 4.66990713497598e-05, "loss": 0.737, "step": 11960 }, { "epoch": 1.23, "learning_rate": 4.6697026452244004e-05, "loss": 0.3733, "step": 11961 }, { "epoch": 1.23, "learning_rate": 4.669498144232774e-05, "loss": 0.2, "step": 11962 }, { "epoch": 1.23, "learning_rate": 4.6692936320024755e-05, "loss": 0.3119, "step": 11963 }, { "epoch": 1.23, "learning_rate": 4.669089108534884e-05, "loss": 0.8721, "step": 11964 }, { "epoch": 1.23, "learning_rate": 4.668884573831374e-05, "loss": 1.3718, "step": 11965 }, { "epoch": 1.23, "learning_rate": 4.6686800278933243e-05, "loss": 0.3094, "step": 11966 }, { "epoch": 1.23, "learning_rate": 4.668475470722112e-05, "loss": 0.6485, "step": 11967 }, { "epoch": 1.23, "learning_rate": 4.668270902319112e-05, "loss": 0.5401, "step": 11968 }, { "epoch": 1.23, "learning_rate": 4.6680663226857027e-05, "loss": 0.5263, "step": 11969 }, { "epoch": 1.23, "learning_rate": 4.667861731823261e-05, "loss": 0.4619, "step": 11970 }, { "epoch": 1.23, "learning_rate": 4.667657129733165e-05, "loss": 0.5399, "step": 11971 }, { "epoch": 1.23, "learning_rate": 4.667452516416792e-05, "loss": 0.7215, "step": 11972 }, { "epoch": 1.23, "learning_rate": 4.6672478918755175e-05, "loss": 0.6141, "step": 11973 }, { "epoch": 1.24, "learning_rate": 4.667043256110721e-05, "loss": 0.3683, "step": 11974 }, { "epoch": 1.24, "learning_rate": 4.666838609123779e-05, "loss": 0.7527, "step": 11975 }, { "epoch": 1.24, "learning_rate": 4.66663395091607e-05, "loss": 0.6862, "step": 11976 }, { "epoch": 1.24, "learning_rate": 4.66642928148897e-05, "loss": 0.3643, "step": 11977 }, { "epoch": 1.24, "learning_rate": 4.666224600843859e-05, "loss": 0.8453, "step": 11978 }, { "epoch": 1.24, "learning_rate": 4.6660199089821144e-05, "loss": 0.3808, "step": 11979 }, { "epoch": 1.24, "learning_rate": 4.665815205905113e-05, "loss": 0.331, "step": 11980 }, { "epoch": 1.24, "learning_rate": 4.6656104916142324e-05, "loss": 1.1017, "step": 11981 }, { "epoch": 1.24, "learning_rate": 4.665405766110853e-05, "loss": 0.4575, "step": 11982 }, { "epoch": 1.24, "learning_rate": 4.665201029396352e-05, "loss": 0.2255, "step": 11983 }, { "epoch": 1.24, "learning_rate": 4.664996281472106e-05, "loss": 1.4026, "step": 11984 }, { "epoch": 1.24, "learning_rate": 4.664791522339495e-05, "loss": 0.5065, "step": 11985 }, { "epoch": 1.24, "learning_rate": 4.664586751999897e-05, "loss": 0.5102, "step": 11986 }, { "epoch": 1.24, "learning_rate": 4.664381970454691e-05, "loss": 0.9848, "step": 11987 }, { "epoch": 1.24, "learning_rate": 4.664177177705255e-05, "loss": 0.3607, "step": 11988 }, { "epoch": 1.24, "learning_rate": 4.663972373752967e-05, "loss": 0.7655, "step": 11989 }, { "epoch": 1.24, "learning_rate": 4.663767558599206e-05, "loss": 0.703, "step": 11990 }, { "epoch": 1.24, "learning_rate": 4.6635627322453515e-05, "loss": 1.0679, "step": 11991 }, { "epoch": 1.24, "learning_rate": 4.6633578946927825e-05, "loss": 0.6799, "step": 11992 }, { "epoch": 1.24, "learning_rate": 4.663153045942877e-05, "loss": 0.5343, "step": 11993 }, { "epoch": 1.24, "learning_rate": 4.662948185997014e-05, "loss": 0.7519, "step": 11994 }, { "epoch": 1.24, "learning_rate": 4.6627433148565734e-05, "loss": 0.6376, "step": 11995 }, { "epoch": 1.24, "learning_rate": 4.662538432522934e-05, "loss": 0.4602, "step": 11996 }, { "epoch": 1.24, "learning_rate": 4.662333538997474e-05, "loss": 1.2653, "step": 11997 }, { "epoch": 1.24, "learning_rate": 4.6621286342815744e-05, "loss": 0.7566, "step": 11998 }, { "epoch": 1.24, "learning_rate": 4.661923718376613e-05, "loss": 0.2267, "step": 11999 }, { "epoch": 1.24, "learning_rate": 4.6617187912839705e-05, "loss": 0.409, "step": 12000 }, { "epoch": 1.24, "learning_rate": 4.661513853005026e-05, "loss": 1.1843, "step": 12001 }, { "epoch": 1.24, "learning_rate": 4.661308903541159e-05, "loss": 0.1866, "step": 12002 }, { "epoch": 1.24, "learning_rate": 4.661103942893748e-05, "loss": 0.5347, "step": 12003 }, { "epoch": 1.24, "learning_rate": 4.660898971064176e-05, "loss": 0.6106, "step": 12004 }, { "epoch": 1.24, "learning_rate": 4.660693988053819e-05, "loss": 0.7137, "step": 12005 }, { "epoch": 1.24, "learning_rate": 4.660488993864059e-05, "loss": 0.9695, "step": 12006 }, { "epoch": 1.24, "learning_rate": 4.6602839884962765e-05, "loss": 0.8369, "step": 12007 }, { "epoch": 1.24, "learning_rate": 4.66007897195185e-05, "loss": 0.7763, "step": 12008 }, { "epoch": 1.24, "learning_rate": 4.6598739442321606e-05, "loss": 1.1982, "step": 12009 }, { "epoch": 1.24, "learning_rate": 4.6596689053385875e-05, "loss": 1.03, "step": 12010 }, { "epoch": 1.24, "learning_rate": 4.659463855272512e-05, "loss": 0.2505, "step": 12011 }, { "epoch": 1.24, "learning_rate": 4.659258794035315e-05, "loss": 0.5123, "step": 12012 }, { "epoch": 1.24, "learning_rate": 4.659053721628375e-05, "loss": 0.2668, "step": 12013 }, { "epoch": 1.24, "learning_rate": 4.6588486380530735e-05, "loss": 0.7688, "step": 12014 }, { "epoch": 1.24, "learning_rate": 4.6586435433107916e-05, "loss": 1.1602, "step": 12015 }, { "epoch": 1.24, "learning_rate": 4.658438437402909e-05, "loss": 0.3565, "step": 12016 }, { "epoch": 1.24, "learning_rate": 4.658233320330808e-05, "loss": 0.8613, "step": 12017 }, { "epoch": 1.24, "learning_rate": 4.658028192095867e-05, "loss": 0.8504, "step": 12018 }, { "epoch": 1.24, "learning_rate": 4.657823052699469e-05, "loss": 0.602, "step": 12019 }, { "epoch": 1.24, "learning_rate": 4.657617902142993e-05, "loss": 0.472, "step": 12020 }, { "epoch": 1.24, "learning_rate": 4.657412740427822e-05, "loss": 0.5744, "step": 12021 }, { "epoch": 1.24, "learning_rate": 4.6572075675553356e-05, "loss": 0.7267, "step": 12022 }, { "epoch": 1.24, "learning_rate": 4.657002383526917e-05, "loss": 0.9138, "step": 12023 }, { "epoch": 1.24, "learning_rate": 4.656797188343945e-05, "loss": 0.5407, "step": 12024 }, { "epoch": 1.24, "learning_rate": 4.656591982007802e-05, "loss": 1.0575, "step": 12025 }, { "epoch": 1.24, "learning_rate": 4.65638676451987e-05, "loss": 0.4059, "step": 12026 }, { "epoch": 1.24, "learning_rate": 4.6561815358815286e-05, "loss": 0.9546, "step": 12027 }, { "epoch": 1.24, "learning_rate": 4.655976296094162e-05, "loss": 0.5992, "step": 12028 }, { "epoch": 1.24, "learning_rate": 4.655771045159149e-05, "loss": 1.0529, "step": 12029 }, { "epoch": 1.24, "learning_rate": 4.655565783077874e-05, "loss": 0.767, "step": 12030 }, { "epoch": 1.24, "learning_rate": 4.655360509851718e-05, "loss": 0.5369, "step": 12031 }, { "epoch": 1.24, "learning_rate": 4.6551552254820604e-05, "loss": 0.253, "step": 12032 }, { "epoch": 1.24, "learning_rate": 4.6549499299702865e-05, "loss": 0.5172, "step": 12033 }, { "epoch": 1.24, "learning_rate": 4.654744623317777e-05, "loss": 1.011, "step": 12034 }, { "epoch": 1.24, "learning_rate": 4.654539305525913e-05, "loss": 0.4701, "step": 12035 }, { "epoch": 1.24, "learning_rate": 4.654333976596079e-05, "loss": 0.1786, "step": 12036 }, { "epoch": 1.24, "learning_rate": 4.654128636529654e-05, "loss": 0.1687, "step": 12037 }, { "epoch": 1.24, "learning_rate": 4.6539232853280226e-05, "loss": 0.4653, "step": 12038 }, { "epoch": 1.24, "learning_rate": 4.6537179229925665e-05, "loss": 1.0519, "step": 12039 }, { "epoch": 1.24, "learning_rate": 4.653512549524669e-05, "loss": 0.8645, "step": 12040 }, { "epoch": 1.24, "learning_rate": 4.6533071649257114e-05, "loss": 0.663, "step": 12041 }, { "epoch": 1.24, "learning_rate": 4.6531017691970765e-05, "loss": 0.2408, "step": 12042 }, { "epoch": 1.24, "learning_rate": 4.652896362340148e-05, "loss": 1.1935, "step": 12043 }, { "epoch": 1.24, "learning_rate": 4.6526909443563075e-05, "loss": 1.0416, "step": 12044 }, { "epoch": 1.24, "learning_rate": 4.652485515246939e-05, "loss": 0.6502, "step": 12045 }, { "epoch": 1.24, "learning_rate": 4.6522800750134236e-05, "loss": 0.7318, "step": 12046 }, { "epoch": 1.24, "learning_rate": 4.6520746236571466e-05, "loss": 0.4277, "step": 12047 }, { "epoch": 1.24, "learning_rate": 4.6518691611794884e-05, "loss": 0.2163, "step": 12048 }, { "epoch": 1.24, "learning_rate": 4.6516636875818345e-05, "loss": 0.7839, "step": 12049 }, { "epoch": 1.24, "learning_rate": 4.651458202865567e-05, "loss": 0.6951, "step": 12050 }, { "epoch": 1.24, "learning_rate": 4.651252707032068e-05, "loss": 0.4522, "step": 12051 }, { "epoch": 1.24, "learning_rate": 4.651047200082723e-05, "loss": 0.6101, "step": 12052 }, { "epoch": 1.24, "learning_rate": 4.6508416820189146e-05, "loss": 0.4214, "step": 12053 }, { "epoch": 1.24, "learning_rate": 4.650636152842027e-05, "loss": 1.0207, "step": 12054 }, { "epoch": 1.24, "learning_rate": 4.650430612553442e-05, "loss": 0.5158, "step": 12055 }, { "epoch": 1.24, "learning_rate": 4.650225061154545e-05, "loss": 0.562, "step": 12056 }, { "epoch": 1.24, "learning_rate": 4.650019498646719e-05, "loss": 1.0993, "step": 12057 }, { "epoch": 1.24, "learning_rate": 4.649813925031347e-05, "loss": 0.7641, "step": 12058 }, { "epoch": 1.24, "learning_rate": 4.6496083403098145e-05, "loss": 0.629, "step": 12059 }, { "epoch": 1.24, "learning_rate": 4.649402744483504e-05, "loss": 0.6494, "step": 12060 }, { "epoch": 1.24, "learning_rate": 4.649197137553801e-05, "loss": 1.1662, "step": 12061 }, { "epoch": 1.24, "learning_rate": 4.648991519522088e-05, "loss": 1.1499, "step": 12062 }, { "epoch": 1.24, "learning_rate": 4.64878589038975e-05, "loss": 1.3087, "step": 12063 }, { "epoch": 1.24, "learning_rate": 4.6485802501581714e-05, "loss": 0.2905, "step": 12064 }, { "epoch": 1.24, "learning_rate": 4.648374598828736e-05, "loss": 0.244, "step": 12065 }, { "epoch": 1.24, "learning_rate": 4.6481689364028295e-05, "loss": 0.2398, "step": 12066 }, { "epoch": 1.24, "learning_rate": 4.647963262881834e-05, "loss": 0.8505, "step": 12067 }, { "epoch": 1.24, "learning_rate": 4.647757578267136e-05, "loss": 0.7624, "step": 12068 }, { "epoch": 1.25, "learning_rate": 4.647551882560119e-05, "loss": 0.6025, "step": 12069 }, { "epoch": 1.25, "learning_rate": 4.647346175762169e-05, "loss": 0.3008, "step": 12070 }, { "epoch": 1.25, "learning_rate": 4.6471404578746694e-05, "loss": 0.3499, "step": 12071 }, { "epoch": 1.25, "learning_rate": 4.6469347288990065e-05, "loss": 0.3521, "step": 12072 }, { "epoch": 1.25, "learning_rate": 4.646728988836564e-05, "loss": 0.5715, "step": 12073 }, { "epoch": 1.25, "learning_rate": 4.6465232376887274e-05, "loss": 0.591, "step": 12074 }, { "epoch": 1.25, "learning_rate": 4.6463174754568806e-05, "loss": 0.8276, "step": 12075 }, { "epoch": 1.25, "learning_rate": 4.646111702142411e-05, "loss": 0.2978, "step": 12076 }, { "epoch": 1.25, "learning_rate": 4.6459059177467014e-05, "loss": 0.7134, "step": 12077 }, { "epoch": 1.25, "learning_rate": 4.645700122271139e-05, "loss": 0.839, "step": 12078 }, { "epoch": 1.25, "learning_rate": 4.645494315717108e-05, "loss": 0.5808, "step": 12079 }, { "epoch": 1.25, "learning_rate": 4.645288498085996e-05, "loss": 0.8137, "step": 12080 }, { "epoch": 1.25, "learning_rate": 4.645082669379185e-05, "loss": 0.9114, "step": 12081 }, { "epoch": 1.25, "learning_rate": 4.644876829598062e-05, "loss": 0.8796, "step": 12082 }, { "epoch": 1.25, "learning_rate": 4.644670978744014e-05, "loss": 0.6652, "step": 12083 }, { "epoch": 1.25, "learning_rate": 4.644465116818426e-05, "loss": 0.9247, "step": 12084 }, { "epoch": 1.25, "learning_rate": 4.644259243822683e-05, "loss": 1.0228, "step": 12085 }, { "epoch": 1.25, "learning_rate": 4.6440533597581725e-05, "loss": 0.4823, "step": 12086 }, { "epoch": 1.25, "learning_rate": 4.6438474646262776e-05, "loss": 0.1735, "step": 12087 }, { "epoch": 1.25, "learning_rate": 4.6436415584283885e-05, "loss": 0.2397, "step": 12088 }, { "epoch": 1.25, "learning_rate": 4.643435641165887e-05, "loss": 0.4993, "step": 12089 }, { "epoch": 1.25, "learning_rate": 4.643229712840162e-05, "loss": 0.144, "step": 12090 }, { "epoch": 1.25, "learning_rate": 4.643023773452599e-05, "loss": 0.7432, "step": 12091 }, { "epoch": 1.25, "learning_rate": 4.6428178230045844e-05, "loss": 0.3269, "step": 12092 }, { "epoch": 1.25, "learning_rate": 4.642611861497506e-05, "loss": 0.2485, "step": 12093 }, { "epoch": 1.25, "learning_rate": 4.642405888932747e-05, "loss": 0.3585, "step": 12094 }, { "epoch": 1.25, "learning_rate": 4.642199905311696e-05, "loss": 0.4343, "step": 12095 }, { "epoch": 1.25, "learning_rate": 4.64199391063574e-05, "loss": 1.0339, "step": 12096 }, { "epoch": 1.25, "learning_rate": 4.641787904906266e-05, "loss": 0.7658, "step": 12097 }, { "epoch": 1.25, "learning_rate": 4.6415818881246586e-05, "loss": 0.9075, "step": 12098 }, { "epoch": 1.25, "learning_rate": 4.6413758602923064e-05, "loss": 0.5299, "step": 12099 }, { "epoch": 1.25, "learning_rate": 4.641169821410597e-05, "loss": 0.8742, "step": 12100 }, { "epoch": 1.25, "learning_rate": 4.6409637714809144e-05, "loss": 0.3992, "step": 12101 }, { "epoch": 1.25, "learning_rate": 4.6407577105046495e-05, "loss": 0.8714, "step": 12102 }, { "epoch": 1.25, "learning_rate": 4.640551638483187e-05, "loss": 0.5685, "step": 12103 }, { "epoch": 1.25, "learning_rate": 4.640345555417914e-05, "loss": 1.1175, "step": 12104 }, { "epoch": 1.25, "learning_rate": 4.6401394613102196e-05, "loss": 0.35, "step": 12105 }, { "epoch": 1.25, "learning_rate": 4.6399333561614894e-05, "loss": 0.1792, "step": 12106 }, { "epoch": 1.25, "learning_rate": 4.639727239973112e-05, "loss": 0.9407, "step": 12107 }, { "epoch": 1.25, "learning_rate": 4.639521112746474e-05, "loss": 0.6928, "step": 12108 }, { "epoch": 1.25, "learning_rate": 4.6393149744829635e-05, "loss": 0.873, "step": 12109 }, { "epoch": 1.25, "learning_rate": 4.639108825183969e-05, "loss": 0.1558, "step": 12110 }, { "epoch": 1.25, "learning_rate": 4.638902664850877e-05, "loss": 0.9136, "step": 12111 }, { "epoch": 1.25, "learning_rate": 4.638696493485075e-05, "loss": 0.7655, "step": 12112 }, { "epoch": 1.25, "learning_rate": 4.6384903110879525e-05, "loss": 0.4422, "step": 12113 }, { "epoch": 1.25, "learning_rate": 4.638284117660896e-05, "loss": 0.9532, "step": 12114 }, { "epoch": 1.25, "learning_rate": 4.638077913205294e-05, "loss": 0.5868, "step": 12115 }, { "epoch": 1.25, "learning_rate": 4.637871697722536e-05, "loss": 0.7387, "step": 12116 }, { "epoch": 1.25, "learning_rate": 4.6376654712140084e-05, "loss": 0.4743, "step": 12117 }, { "epoch": 1.25, "learning_rate": 4.6374592336810996e-05, "loss": 0.332, "step": 12118 }, { "epoch": 1.25, "learning_rate": 4.637252985125198e-05, "loss": 0.1128, "step": 12119 }, { "epoch": 1.25, "learning_rate": 4.637046725547694e-05, "loss": 0.9454, "step": 12120 }, { "epoch": 1.25, "learning_rate": 4.636840454949972e-05, "loss": 0.2158, "step": 12121 }, { "epoch": 1.25, "learning_rate": 4.6366341733334254e-05, "loss": 0.5313, "step": 12122 }, { "epoch": 1.25, "learning_rate": 4.63642788069944e-05, "loss": 0.3086, "step": 12123 }, { "epoch": 1.25, "learning_rate": 4.636221577049404e-05, "loss": 0.3011, "step": 12124 }, { "epoch": 1.25, "learning_rate": 4.6360152623847076e-05, "loss": 1.2435, "step": 12125 }, { "epoch": 1.25, "learning_rate": 4.6358089367067395e-05, "loss": 0.2454, "step": 12126 }, { "epoch": 1.25, "learning_rate": 4.6356026000168884e-05, "loss": 0.6413, "step": 12127 }, { "epoch": 1.25, "learning_rate": 4.635396252316543e-05, "loss": 0.3636, "step": 12128 }, { "epoch": 1.25, "learning_rate": 4.635189893607093e-05, "loss": 0.6781, "step": 12129 }, { "epoch": 1.25, "learning_rate": 4.634983523889928e-05, "loss": 0.5385, "step": 12130 }, { "epoch": 1.25, "learning_rate": 4.634777143166435e-05, "loss": 0.8109, "step": 12131 }, { "epoch": 1.25, "learning_rate": 4.634570751438005e-05, "loss": 0.2521, "step": 12132 }, { "epoch": 1.25, "learning_rate": 4.634364348706028e-05, "loss": 0.9286, "step": 12133 }, { "epoch": 1.25, "learning_rate": 4.634157934971892e-05, "loss": 0.718, "step": 12134 }, { "epoch": 1.25, "learning_rate": 4.6339515102369875e-05, "loss": 1.0089, "step": 12135 }, { "epoch": 1.25, "learning_rate": 4.633745074502704e-05, "loss": 0.7107, "step": 12136 }, { "epoch": 1.25, "learning_rate": 4.63353862777043e-05, "loss": 0.8681, "step": 12137 }, { "epoch": 1.25, "learning_rate": 4.6333321700415566e-05, "loss": 1.1558, "step": 12138 }, { "epoch": 1.25, "learning_rate": 4.633125701317473e-05, "loss": 0.6528, "step": 12139 }, { "epoch": 1.25, "learning_rate": 4.6329192215995694e-05, "loss": 0.8975, "step": 12140 }, { "epoch": 1.25, "learning_rate": 4.6327127308892356e-05, "loss": 0.6326, "step": 12141 }, { "epoch": 1.25, "learning_rate": 4.632506229187863e-05, "loss": 0.1833, "step": 12142 }, { "epoch": 1.25, "learning_rate": 4.632299716496838e-05, "loss": 0.2113, "step": 12143 }, { "epoch": 1.25, "learning_rate": 4.632093192817555e-05, "loss": 0.8474, "step": 12144 }, { "epoch": 1.25, "learning_rate": 4.631886658151403e-05, "loss": 0.999, "step": 12145 }, { "epoch": 1.25, "learning_rate": 4.63168011249977e-05, "loss": 0.9344, "step": 12146 }, { "epoch": 1.25, "learning_rate": 4.631473555864049e-05, "loss": 1.0896, "step": 12147 }, { "epoch": 1.25, "learning_rate": 4.631266988245631e-05, "loss": 0.4645, "step": 12148 }, { "epoch": 1.25, "learning_rate": 4.631060409645904e-05, "loss": 0.3561, "step": 12149 }, { "epoch": 1.25, "learning_rate": 4.63085382006626e-05, "loss": 0.2591, "step": 12150 }, { "epoch": 1.25, "learning_rate": 4.63064721950809e-05, "loss": 0.4612, "step": 12151 }, { "epoch": 1.25, "learning_rate": 4.630440607972785e-05, "loss": 0.6574, "step": 12152 }, { "epoch": 1.25, "learning_rate": 4.630233985461735e-05, "loss": 0.1413, "step": 12153 }, { "epoch": 1.25, "learning_rate": 4.6300273519763303e-05, "loss": 0.4059, "step": 12154 }, { "epoch": 1.25, "learning_rate": 4.629820707517964e-05, "loss": 0.856, "step": 12155 }, { "epoch": 1.25, "learning_rate": 4.629614052088025e-05, "loss": 0.8663, "step": 12156 }, { "epoch": 1.25, "learning_rate": 4.629407385687906e-05, "loss": 1.1155, "step": 12157 }, { "epoch": 1.25, "learning_rate": 4.629200708318998e-05, "loss": 0.7067, "step": 12158 }, { "epoch": 1.25, "learning_rate": 4.628994019982692e-05, "loss": 0.5522, "step": 12159 }, { "epoch": 1.25, "learning_rate": 4.628787320680379e-05, "loss": 0.2634, "step": 12160 }, { "epoch": 1.25, "learning_rate": 4.628580610413451e-05, "loss": 0.4712, "step": 12161 }, { "epoch": 1.25, "learning_rate": 4.6283738891833e-05, "loss": 0.2036, "step": 12162 }, { "epoch": 1.25, "learning_rate": 4.628167156991316e-05, "loss": 0.6717, "step": 12163 }, { "epoch": 1.25, "learning_rate": 4.627960413838893e-05, "loss": 0.7305, "step": 12164 }, { "epoch": 1.26, "learning_rate": 4.62775365972742e-05, "loss": 1.0744, "step": 12165 }, { "epoch": 1.26, "learning_rate": 4.627546894658291e-05, "loss": 0.2719, "step": 12166 }, { "epoch": 1.26, "learning_rate": 4.6273401186328965e-05, "loss": 0.261, "step": 12167 }, { "epoch": 1.26, "learning_rate": 4.62713333165263e-05, "loss": 1.1709, "step": 12168 }, { "epoch": 1.26, "learning_rate": 4.626926533718882e-05, "loss": 0.4377, "step": 12169 }, { "epoch": 1.26, "learning_rate": 4.626719724833045e-05, "loss": 0.6034, "step": 12170 }, { "epoch": 1.26, "learning_rate": 4.6265129049965116e-05, "loss": 1.0425, "step": 12171 }, { "epoch": 1.26, "learning_rate": 4.6263060742106746e-05, "loss": 1.0226, "step": 12172 }, { "epoch": 1.26, "learning_rate": 4.626099232476925e-05, "loss": 1.2831, "step": 12173 }, { "epoch": 1.26, "learning_rate": 4.625892379796656e-05, "loss": 0.9729, "step": 12174 }, { "epoch": 1.26, "learning_rate": 4.62568551617126e-05, "loss": 1.0027, "step": 12175 }, { "epoch": 1.26, "learning_rate": 4.62547864160213e-05, "loss": 0.414, "step": 12176 }, { "epoch": 1.26, "learning_rate": 4.625271756090657e-05, "loss": 0.2307, "step": 12177 }, { "epoch": 1.26, "learning_rate": 4.625064859638235e-05, "loss": 0.9856, "step": 12178 }, { "epoch": 1.26, "learning_rate": 4.6248579522462565e-05, "loss": 0.404, "step": 12179 }, { "epoch": 1.26, "learning_rate": 4.624651033916115e-05, "loss": 0.8271, "step": 12180 }, { "epoch": 1.26, "learning_rate": 4.624444104649203e-05, "loss": 0.2376, "step": 12181 }, { "epoch": 1.26, "learning_rate": 4.624237164446913e-05, "loss": 1.2092, "step": 12182 }, { "epoch": 1.26, "learning_rate": 4.624030213310639e-05, "loss": 0.4634, "step": 12183 }, { "epoch": 1.26, "learning_rate": 4.623823251241773e-05, "loss": 0.5777, "step": 12184 }, { "epoch": 1.26, "learning_rate": 4.623616278241709e-05, "loss": 0.9364, "step": 12185 }, { "epoch": 1.26, "learning_rate": 4.62340929431184e-05, "loss": 0.6772, "step": 12186 }, { "epoch": 1.26, "learning_rate": 4.6232022994535604e-05, "loss": 0.4893, "step": 12187 }, { "epoch": 1.26, "learning_rate": 4.6229952936682614e-05, "loss": 0.4944, "step": 12188 }, { "epoch": 1.26, "learning_rate": 4.622788276957338e-05, "loss": 0.2494, "step": 12189 }, { "epoch": 1.26, "learning_rate": 4.6225812493221844e-05, "loss": 0.9879, "step": 12190 }, { "epoch": 1.26, "learning_rate": 4.6223742107641934e-05, "loss": 0.2733, "step": 12191 }, { "epoch": 1.26, "learning_rate": 4.622167161284758e-05, "loss": 0.5479, "step": 12192 }, { "epoch": 1.26, "learning_rate": 4.621960100885273e-05, "loss": 0.9655, "step": 12193 }, { "epoch": 1.26, "learning_rate": 4.621753029567133e-05, "loss": 1.2897, "step": 12194 }, { "epoch": 1.26, "learning_rate": 4.621545947331731e-05, "loss": 0.4953, "step": 12195 }, { "epoch": 1.26, "learning_rate": 4.62133885418046e-05, "loss": 0.4399, "step": 12196 }, { "epoch": 1.26, "learning_rate": 4.6211317501147154e-05, "loss": 0.2193, "step": 12197 }, { "epoch": 1.26, "learning_rate": 4.620924635135892e-05, "loss": 0.5708, "step": 12198 }, { "epoch": 1.26, "learning_rate": 4.620717509245383e-05, "loss": 0.9647, "step": 12199 }, { "epoch": 1.26, "learning_rate": 4.620510372444583e-05, "loss": 0.5224, "step": 12200 }, { "epoch": 1.26, "learning_rate": 4.620303224734885e-05, "loss": 1.2724, "step": 12201 }, { "epoch": 1.26, "learning_rate": 4.620096066117686e-05, "loss": 1.1843, "step": 12202 }, { "epoch": 1.26, "learning_rate": 4.619888896594379e-05, "loss": 0.3559, "step": 12203 }, { "epoch": 1.26, "learning_rate": 4.6196817161663595e-05, "loss": 0.4761, "step": 12204 }, { "epoch": 1.26, "learning_rate": 4.619474524835022e-05, "loss": 0.683, "step": 12205 }, { "epoch": 1.26, "learning_rate": 4.61926732260176e-05, "loss": 0.3514, "step": 12206 }, { "epoch": 1.26, "learning_rate": 4.619060109467969e-05, "loss": 0.8119, "step": 12207 }, { "epoch": 1.26, "learning_rate": 4.618852885435045e-05, "loss": 0.6235, "step": 12208 }, { "epoch": 1.26, "learning_rate": 4.618645650504382e-05, "loss": 0.3321, "step": 12209 }, { "epoch": 1.26, "learning_rate": 4.618438404677375e-05, "loss": 0.852, "step": 12210 }, { "epoch": 1.26, "learning_rate": 4.618231147955419e-05, "loss": 0.4213, "step": 12211 }, { "epoch": 1.26, "learning_rate": 4.61802388033991e-05, "loss": 0.844, "step": 12212 }, { "epoch": 1.26, "learning_rate": 4.617816601832243e-05, "loss": 1.5892, "step": 12213 }, { "epoch": 1.26, "learning_rate": 4.6176093124338136e-05, "loss": 0.8198, "step": 12214 }, { "epoch": 1.26, "learning_rate": 4.617402012146016e-05, "loss": 0.9827, "step": 12215 }, { "epoch": 1.26, "learning_rate": 4.617194700970246e-05, "loss": 0.8966, "step": 12216 }, { "epoch": 1.26, "learning_rate": 4.6169873789079007e-05, "loss": 0.5236, "step": 12217 }, { "epoch": 1.26, "learning_rate": 4.6167800459603745e-05, "loss": 0.4114, "step": 12218 }, { "epoch": 1.26, "learning_rate": 4.616572702129063e-05, "loss": 0.799, "step": 12219 }, { "epoch": 1.26, "learning_rate": 4.616365347415362e-05, "loss": 0.4361, "step": 12220 }, { "epoch": 1.26, "learning_rate": 4.6161579818206686e-05, "loss": 0.6724, "step": 12221 }, { "epoch": 1.26, "learning_rate": 4.615950605346377e-05, "loss": 0.4957, "step": 12222 }, { "epoch": 1.26, "learning_rate": 4.615743217993885e-05, "loss": 0.4182, "step": 12223 }, { "epoch": 1.26, "learning_rate": 4.615535819764587e-05, "loss": 0.3004, "step": 12224 }, { "epoch": 1.26, "learning_rate": 4.615328410659879e-05, "loss": 0.5801, "step": 12225 }, { "epoch": 1.26, "learning_rate": 4.61512099068116e-05, "loss": 0.3621, "step": 12226 }, { "epoch": 1.26, "learning_rate": 4.614913559829823e-05, "loss": 0.7118, "step": 12227 }, { "epoch": 1.26, "learning_rate": 4.614706118107266e-05, "loss": 0.9786, "step": 12228 }, { "epoch": 1.26, "learning_rate": 4.6144986655148856e-05, "loss": 0.595, "step": 12229 }, { "epoch": 1.26, "learning_rate": 4.614291202054078e-05, "loss": 0.9234, "step": 12230 }, { "epoch": 1.26, "learning_rate": 4.6140837277262396e-05, "loss": 0.2801, "step": 12231 }, { "epoch": 1.26, "learning_rate": 4.613876242532768e-05, "loss": 0.2797, "step": 12232 }, { "epoch": 1.26, "learning_rate": 4.613668746475057e-05, "loss": 0.2591, "step": 12233 }, { "epoch": 1.26, "learning_rate": 4.613461239554508e-05, "loss": 0.4715, "step": 12234 }, { "epoch": 1.26, "learning_rate": 4.6132537217725144e-05, "loss": 0.7473, "step": 12235 }, { "epoch": 1.26, "learning_rate": 4.613046193130474e-05, "loss": 0.5614, "step": 12236 }, { "epoch": 1.26, "learning_rate": 4.6128386536297845e-05, "loss": 0.4551, "step": 12237 }, { "epoch": 1.26, "learning_rate": 4.612631103271842e-05, "loss": 0.7604, "step": 12238 }, { "epoch": 1.26, "learning_rate": 4.612423542058044e-05, "loss": 0.2432, "step": 12239 }, { "epoch": 1.26, "learning_rate": 4.612215969989789e-05, "loss": 0.8146, "step": 12240 }, { "epoch": 1.26, "learning_rate": 4.612008387068473e-05, "loss": 0.1675, "step": 12241 }, { "epoch": 1.26, "learning_rate": 4.611800793295493e-05, "loss": 0.9529, "step": 12242 }, { "epoch": 1.26, "learning_rate": 4.611593188672249e-05, "loss": 0.9384, "step": 12243 }, { "epoch": 1.26, "learning_rate": 4.611385573200134e-05, "loss": 0.4437, "step": 12244 }, { "epoch": 1.26, "learning_rate": 4.6111779468805506e-05, "loss": 0.5513, "step": 12245 }, { "epoch": 1.26, "learning_rate": 4.610970309714893e-05, "loss": 0.7985, "step": 12246 }, { "epoch": 1.26, "learning_rate": 4.61076266170456e-05, "loss": 0.3392, "step": 12247 }, { "epoch": 1.26, "learning_rate": 4.6105550028509514e-05, "loss": 0.2429, "step": 12248 }, { "epoch": 1.26, "learning_rate": 4.6103473331554624e-05, "loss": 0.4666, "step": 12249 }, { "epoch": 1.26, "learning_rate": 4.610139652619491e-05, "loss": 1.0438, "step": 12250 }, { "epoch": 1.26, "learning_rate": 4.609931961244438e-05, "loss": 1.1206, "step": 12251 }, { "epoch": 1.26, "learning_rate": 4.609724259031699e-05, "loss": 0.4709, "step": 12252 }, { "epoch": 1.26, "learning_rate": 4.609516545982672e-05, "loss": 0.8173, "step": 12253 }, { "epoch": 1.26, "learning_rate": 4.6093088220987563e-05, "loss": 0.2959, "step": 12254 }, { "epoch": 1.26, "learning_rate": 4.609101087381351e-05, "loss": 0.851, "step": 12255 }, { "epoch": 1.26, "learning_rate": 4.6088933418318546e-05, "loss": 0.6826, "step": 12256 }, { "epoch": 1.26, "learning_rate": 4.608685585451663e-05, "loss": 0.6007, "step": 12257 }, { "epoch": 1.26, "learning_rate": 4.608477818242177e-05, "loss": 0.777, "step": 12258 }, { "epoch": 1.26, "learning_rate": 4.608270040204794e-05, "loss": 0.4612, "step": 12259 }, { "epoch": 1.26, "learning_rate": 4.608062251340914e-05, "loss": 1.2341, "step": 12260 }, { "epoch": 1.27, "learning_rate": 4.607854451651936e-05, "loss": 0.6385, "step": 12261 }, { "epoch": 1.27, "learning_rate": 4.607646641139258e-05, "loss": 0.6466, "step": 12262 }, { "epoch": 1.27, "learning_rate": 4.6074388198042775e-05, "loss": 0.5033, "step": 12263 }, { "epoch": 1.27, "learning_rate": 4.607230987648396e-05, "loss": 0.5639, "step": 12264 }, { "epoch": 1.27, "learning_rate": 4.6070231446730106e-05, "loss": 0.283, "step": 12265 }, { "epoch": 1.27, "learning_rate": 4.6068152908795225e-05, "loss": 0.8431, "step": 12266 }, { "epoch": 1.27, "learning_rate": 4.6066074262693294e-05, "loss": 0.8732, "step": 12267 }, { "epoch": 1.27, "learning_rate": 4.606399550843831e-05, "loss": 0.7003, "step": 12268 }, { "epoch": 1.27, "learning_rate": 4.6061916646044264e-05, "loss": 0.6532, "step": 12269 }, { "epoch": 1.27, "learning_rate": 4.605983767552516e-05, "loss": 0.5086, "step": 12270 }, { "epoch": 1.27, "learning_rate": 4.605775859689498e-05, "loss": 0.1994, "step": 12271 }, { "epoch": 1.27, "learning_rate": 4.6055679410167725e-05, "loss": 0.4536, "step": 12272 }, { "epoch": 1.27, "learning_rate": 4.605360011535739e-05, "loss": 0.541, "step": 12273 }, { "epoch": 1.27, "learning_rate": 4.605152071247799e-05, "loss": 1.0212, "step": 12274 }, { "epoch": 1.27, "learning_rate": 4.604944120154351e-05, "loss": 0.8356, "step": 12275 }, { "epoch": 1.27, "learning_rate": 4.604736158256793e-05, "loss": 0.7507, "step": 12276 }, { "epoch": 1.27, "learning_rate": 4.604528185556527e-05, "loss": 0.3568, "step": 12277 }, { "epoch": 1.27, "learning_rate": 4.604320202054953e-05, "loss": 0.6386, "step": 12278 }, { "epoch": 1.27, "learning_rate": 4.604112207753472e-05, "loss": 0.3987, "step": 12279 }, { "epoch": 1.27, "learning_rate": 4.603904202653481e-05, "loss": 1.1545, "step": 12280 }, { "epoch": 1.27, "learning_rate": 4.603696186756384e-05, "loss": 0.225, "step": 12281 }, { "epoch": 1.27, "learning_rate": 4.603488160063578e-05, "loss": 0.905, "step": 12282 }, { "epoch": 1.27, "learning_rate": 4.603280122576465e-05, "loss": 0.4704, "step": 12283 }, { "epoch": 1.27, "learning_rate": 4.603072074296447e-05, "loss": 0.7437, "step": 12284 }, { "epoch": 1.27, "learning_rate": 4.602864015224921e-05, "loss": 0.1233, "step": 12285 }, { "epoch": 1.27, "learning_rate": 4.60265594536329e-05, "loss": 1.3745, "step": 12286 }, { "epoch": 1.27, "learning_rate": 4.6024478647129546e-05, "loss": 0.6757, "step": 12287 }, { "epoch": 1.27, "learning_rate": 4.6022397732753146e-05, "loss": 1.0149, "step": 12288 }, { "epoch": 1.27, "learning_rate": 4.602031671051772e-05, "loss": 0.5512, "step": 12289 }, { "epoch": 1.27, "learning_rate": 4.6018235580437264e-05, "loss": 0.6749, "step": 12290 }, { "epoch": 1.27, "learning_rate": 4.6016154342525795e-05, "loss": 1.1446, "step": 12291 }, { "epoch": 1.27, "learning_rate": 4.6014072996797324e-05, "loss": 0.5539, "step": 12292 }, { "epoch": 1.27, "learning_rate": 4.6011991543265866e-05, "loss": 0.3166, "step": 12293 }, { "epoch": 1.27, "learning_rate": 4.600990998194543e-05, "loss": 0.8943, "step": 12294 }, { "epoch": 1.27, "learning_rate": 4.6007828312850006e-05, "loss": 0.9366, "step": 12295 }, { "epoch": 1.27, "learning_rate": 4.600574653599364e-05, "loss": 0.5333, "step": 12296 }, { "epoch": 1.27, "learning_rate": 4.600366465139033e-05, "loss": 0.6639, "step": 12297 }, { "epoch": 1.27, "learning_rate": 4.6001582659054104e-05, "loss": 0.2472, "step": 12298 }, { "epoch": 1.27, "learning_rate": 4.599950055899895e-05, "loss": 1.2056, "step": 12299 }, { "epoch": 1.27, "learning_rate": 4.5997418351238926e-05, "loss": 0.2845, "step": 12300 }, { "epoch": 1.27, "learning_rate": 4.5995336035788e-05, "loss": 0.3647, "step": 12301 }, { "epoch": 1.27, "learning_rate": 4.599325361266022e-05, "loss": 0.7777, "step": 12302 }, { "epoch": 1.27, "learning_rate": 4.599117108186961e-05, "loss": 1.0632, "step": 12303 }, { "epoch": 1.27, "learning_rate": 4.5989088443430175e-05, "loss": 0.6991, "step": 12304 }, { "epoch": 1.27, "learning_rate": 4.598700569735593e-05, "loss": 1.094, "step": 12305 }, { "epoch": 1.27, "learning_rate": 4.598492284366092e-05, "loss": 0.2201, "step": 12306 }, { "epoch": 1.27, "learning_rate": 4.5982839882359123e-05, "loss": 0.2121, "step": 12307 }, { "epoch": 1.27, "learning_rate": 4.598075681346461e-05, "loss": 0.8759, "step": 12308 }, { "epoch": 1.27, "learning_rate": 4.5978673636991375e-05, "loss": 0.8824, "step": 12309 }, { "epoch": 1.27, "learning_rate": 4.597659035295344e-05, "loss": 0.612, "step": 12310 }, { "epoch": 1.27, "learning_rate": 4.597450696136484e-05, "loss": 0.5974, "step": 12311 }, { "epoch": 1.27, "learning_rate": 4.5972423462239605e-05, "loss": 0.6612, "step": 12312 }, { "epoch": 1.27, "learning_rate": 4.597033985559175e-05, "loss": 0.7034, "step": 12313 }, { "epoch": 1.27, "learning_rate": 4.59682561414353e-05, "loss": 0.8624, "step": 12314 }, { "epoch": 1.27, "learning_rate": 4.596617231978429e-05, "loss": 0.3694, "step": 12315 }, { "epoch": 1.27, "learning_rate": 4.596408839065274e-05, "loss": 0.6222, "step": 12316 }, { "epoch": 1.27, "learning_rate": 4.5962004354054685e-05, "loss": 0.4024, "step": 12317 }, { "epoch": 1.27, "learning_rate": 4.595992021000415e-05, "loss": 0.5592, "step": 12318 }, { "epoch": 1.27, "learning_rate": 4.5957835958515166e-05, "loss": 0.3422, "step": 12319 }, { "epoch": 1.27, "learning_rate": 4.5955751599601766e-05, "loss": 0.4142, "step": 12320 }, { "epoch": 1.27, "learning_rate": 4.5953667133277974e-05, "loss": 1.1476, "step": 12321 }, { "epoch": 1.27, "learning_rate": 4.595158255955784e-05, "loss": 0.3763, "step": 12322 }, { "epoch": 1.27, "learning_rate": 4.5949497878455365e-05, "loss": 0.3046, "step": 12323 }, { "epoch": 1.27, "learning_rate": 4.5947413089984615e-05, "loss": 0.1763, "step": 12324 }, { "epoch": 1.27, "learning_rate": 4.594532819415962e-05, "loss": 0.349, "step": 12325 }, { "epoch": 1.27, "learning_rate": 4.594324319099439e-05, "loss": 1.6371, "step": 12326 }, { "epoch": 1.27, "learning_rate": 4.594115808050299e-05, "loss": 1.086, "step": 12327 }, { "epoch": 1.27, "learning_rate": 4.593907286269943e-05, "loss": 0.6546, "step": 12328 }, { "epoch": 1.27, "learning_rate": 4.5936987537597766e-05, "loss": 0.816, "step": 12329 }, { "epoch": 1.27, "learning_rate": 4.5934902105212034e-05, "loss": 0.7345, "step": 12330 }, { "epoch": 1.27, "learning_rate": 4.593281656555627e-05, "loss": 0.6347, "step": 12331 }, { "epoch": 1.27, "learning_rate": 4.593073091864452e-05, "loss": 0.5322, "step": 12332 }, { "epoch": 1.27, "learning_rate": 4.592864516449081e-05, "loss": 0.7193, "step": 12333 }, { "epoch": 1.27, "learning_rate": 4.592655930310918e-05, "loss": 0.8918, "step": 12334 }, { "epoch": 1.27, "learning_rate": 4.5924473334513684e-05, "loss": 0.7689, "step": 12335 }, { "epoch": 1.27, "learning_rate": 4.5922387258718364e-05, "loss": 0.6628, "step": 12336 }, { "epoch": 1.27, "learning_rate": 4.592030107573726e-05, "loss": 0.8291, "step": 12337 }, { "epoch": 1.27, "learning_rate": 4.59182147855844e-05, "loss": 0.4813, "step": 12338 }, { "epoch": 1.27, "learning_rate": 4.591612838827386e-05, "loss": 0.4739, "step": 12339 }, { "epoch": 1.27, "learning_rate": 4.591404188381966e-05, "loss": 0.6557, "step": 12340 }, { "epoch": 1.27, "learning_rate": 4.5911955272235854e-05, "loss": 0.3673, "step": 12341 }, { "epoch": 1.27, "learning_rate": 4.590986855353649e-05, "loss": 0.4246, "step": 12342 }, { "epoch": 1.27, "learning_rate": 4.590778172773561e-05, "loss": 0.2592, "step": 12343 }, { "epoch": 1.27, "learning_rate": 4.590569479484727e-05, "loss": 0.7379, "step": 12344 }, { "epoch": 1.27, "learning_rate": 4.5903607754885505e-05, "loss": 0.7796, "step": 12345 }, { "epoch": 1.27, "learning_rate": 4.590152060786439e-05, "loss": 0.6002, "step": 12346 }, { "epoch": 1.27, "learning_rate": 4.5899433353797944e-05, "loss": 0.8155, "step": 12347 }, { "epoch": 1.27, "learning_rate": 4.589734599270024e-05, "loss": 0.5469, "step": 12348 }, { "epoch": 1.27, "learning_rate": 4.589525852458532e-05, "loss": 0.8688, "step": 12349 }, { "epoch": 1.27, "learning_rate": 4.589317094946724e-05, "loss": 0.3105, "step": 12350 }, { "epoch": 1.27, "learning_rate": 4.589108326736005e-05, "loss": 0.9076, "step": 12351 }, { "epoch": 1.27, "learning_rate": 4.5888995478277804e-05, "loss": 0.5151, "step": 12352 }, { "epoch": 1.27, "learning_rate": 4.5886907582234566e-05, "loss": 1.1858, "step": 12353 }, { "epoch": 1.27, "learning_rate": 4.588481957924437e-05, "loss": 0.4789, "step": 12354 }, { "epoch": 1.27, "learning_rate": 4.5882731469321297e-05, "loss": 0.4583, "step": 12355 }, { "epoch": 1.27, "learning_rate": 4.58806432524794e-05, "loss": 0.4726, "step": 12356 }, { "epoch": 1.28, "learning_rate": 4.587855492873271e-05, "loss": 0.678, "step": 12357 }, { "epoch": 1.28, "learning_rate": 4.587646649809531e-05, "loss": 1.2137, "step": 12358 }, { "epoch": 1.28, "learning_rate": 4.587437796058125e-05, "loss": 1.2586, "step": 12359 }, { "epoch": 1.28, "learning_rate": 4.58722893162046e-05, "loss": 0.635, "step": 12360 }, { "epoch": 1.28, "learning_rate": 4.5870200564979406e-05, "loss": 0.4693, "step": 12361 }, { "epoch": 1.28, "learning_rate": 4.586811170691974e-05, "loss": 0.1856, "step": 12362 }, { "epoch": 1.28, "learning_rate": 4.5866022742039654e-05, "loss": 1.0644, "step": 12363 }, { "epoch": 1.28, "learning_rate": 4.586393367035321e-05, "loss": 0.9462, "step": 12364 }, { "epoch": 1.28, "learning_rate": 4.586184449187449e-05, "loss": 0.5312, "step": 12365 }, { "epoch": 1.28, "learning_rate": 4.585975520661753e-05, "loss": 0.7167, "step": 12366 }, { "epoch": 1.28, "learning_rate": 4.585766581459641e-05, "loss": 0.4051, "step": 12367 }, { "epoch": 1.28, "learning_rate": 4.58555763158252e-05, "loss": 1.0823, "step": 12368 }, { "epoch": 1.28, "learning_rate": 4.585348671031796e-05, "loss": 1.3825, "step": 12369 }, { "epoch": 1.28, "learning_rate": 4.585139699808877e-05, "loss": 0.8766, "step": 12370 }, { "epoch": 1.28, "learning_rate": 4.584930717915167e-05, "loss": 0.9467, "step": 12371 }, { "epoch": 1.28, "learning_rate": 4.584721725352074e-05, "loss": 0.6991, "step": 12372 }, { "epoch": 1.28, "learning_rate": 4.5845127221210054e-05, "loss": 0.1845, "step": 12373 }, { "epoch": 1.28, "learning_rate": 4.584303708223368e-05, "loss": 0.1566, "step": 12374 }, { "epoch": 1.28, "learning_rate": 4.5840946836605697e-05, "loss": 0.5011, "step": 12375 }, { "epoch": 1.28, "learning_rate": 4.5838856484340147e-05, "loss": 0.7542, "step": 12376 }, { "epoch": 1.28, "learning_rate": 4.583676602545113e-05, "loss": 0.9729, "step": 12377 }, { "epoch": 1.28, "learning_rate": 4.583467545995271e-05, "loss": 0.1357, "step": 12378 }, { "epoch": 1.28, "learning_rate": 4.583258478785897e-05, "loss": 0.4081, "step": 12379 }, { "epoch": 1.28, "learning_rate": 4.5830494009183956e-05, "loss": 1.4523, "step": 12380 }, { "epoch": 1.28, "learning_rate": 4.5828403123941776e-05, "loss": 1.1864, "step": 12381 }, { "epoch": 1.28, "learning_rate": 4.582631213214648e-05, "loss": 0.293, "step": 12382 }, { "epoch": 1.28, "learning_rate": 4.582422103381215e-05, "loss": 0.2963, "step": 12383 }, { "epoch": 1.28, "learning_rate": 4.582212982895288e-05, "loss": 0.8483, "step": 12384 }, { "epoch": 1.28, "learning_rate": 4.5820038517582724e-05, "loss": 1.1219, "step": 12385 }, { "epoch": 1.28, "learning_rate": 4.581794709971577e-05, "loss": 0.3149, "step": 12386 }, { "epoch": 1.28, "learning_rate": 4.5815855575366096e-05, "loss": 0.2325, "step": 12387 }, { "epoch": 1.28, "learning_rate": 4.5813763944547795e-05, "loss": 0.8552, "step": 12388 }, { "epoch": 1.28, "learning_rate": 4.581167220727493e-05, "loss": 0.6001, "step": 12389 }, { "epoch": 1.28, "learning_rate": 4.580958036356158e-05, "loss": 0.3349, "step": 12390 }, { "epoch": 1.28, "learning_rate": 4.580748841342184e-05, "loss": 0.2933, "step": 12391 }, { "epoch": 1.28, "learning_rate": 4.5805396356869776e-05, "loss": 0.7207, "step": 12392 }, { "epoch": 1.28, "learning_rate": 4.58033041939195e-05, "loss": 0.3281, "step": 12393 }, { "epoch": 1.28, "learning_rate": 4.5801211924585076e-05, "loss": 1.0177, "step": 12394 }, { "epoch": 1.28, "learning_rate": 4.579911954888058e-05, "loss": 0.7038, "step": 12395 }, { "epoch": 1.28, "learning_rate": 4.579702706682011e-05, "loss": 0.5112, "step": 12396 }, { "epoch": 1.28, "learning_rate": 4.5794934478417745e-05, "loss": 0.8936, "step": 12397 }, { "epoch": 1.28, "learning_rate": 4.579284178368759e-05, "loss": 0.2472, "step": 12398 }, { "epoch": 1.28, "learning_rate": 4.5790748982643715e-05, "loss": 0.6117, "step": 12399 }, { "epoch": 1.28, "learning_rate": 4.5788656075300216e-05, "loss": 0.3634, "step": 12400 }, { "epoch": 1.28, "learning_rate": 4.578656306167116e-05, "loss": 0.7225, "step": 12401 }, { "epoch": 1.28, "learning_rate": 4.578446994177067e-05, "loss": 0.3241, "step": 12402 }, { "epoch": 1.28, "learning_rate": 4.578237671561282e-05, "loss": 0.8187, "step": 12403 }, { "epoch": 1.28, "learning_rate": 4.5780283383211705e-05, "loss": 0.2849, "step": 12404 }, { "epoch": 1.28, "learning_rate": 4.577818994458141e-05, "loss": 0.5885, "step": 12405 }, { "epoch": 1.28, "learning_rate": 4.577609639973604e-05, "loss": 0.6157, "step": 12406 }, { "epoch": 1.28, "learning_rate": 4.577400274868967e-05, "loss": 0.3048, "step": 12407 }, { "epoch": 1.28, "learning_rate": 4.577190899145641e-05, "loss": 1.1438, "step": 12408 }, { "epoch": 1.28, "learning_rate": 4.5769815128050346e-05, "loss": 0.6324, "step": 12409 }, { "epoch": 1.28, "learning_rate": 4.576772115848558e-05, "loss": 0.6372, "step": 12410 }, { "epoch": 1.28, "learning_rate": 4.57656270827762e-05, "loss": 0.3806, "step": 12411 }, { "epoch": 1.28, "learning_rate": 4.576353290093632e-05, "loss": 0.4855, "step": 12412 }, { "epoch": 1.28, "learning_rate": 4.576143861298001e-05, "loss": 0.3371, "step": 12413 }, { "epoch": 1.28, "learning_rate": 4.5759344218921386e-05, "loss": 0.89, "step": 12414 }, { "epoch": 1.28, "learning_rate": 4.5757249718774544e-05, "loss": 0.2954, "step": 12415 }, { "epoch": 1.28, "learning_rate": 4.5755155112553584e-05, "loss": 0.487, "step": 12416 }, { "epoch": 1.28, "learning_rate": 4.5753060400272614e-05, "loss": 0.4938, "step": 12417 }, { "epoch": 1.28, "learning_rate": 4.575096558194572e-05, "loss": 0.37, "step": 12418 }, { "epoch": 1.28, "learning_rate": 4.574887065758701e-05, "loss": 0.3516, "step": 12419 }, { "epoch": 1.28, "learning_rate": 4.574677562721059e-05, "loss": 0.8519, "step": 12420 }, { "epoch": 1.28, "learning_rate": 4.574468049083056e-05, "loss": 0.9783, "step": 12421 }, { "epoch": 1.28, "learning_rate": 4.574258524846103e-05, "loss": 0.6565, "step": 12422 }, { "epoch": 1.28, "learning_rate": 4.57404899001161e-05, "loss": 0.6164, "step": 12423 }, { "epoch": 1.28, "learning_rate": 4.5738394445809865e-05, "loss": 0.5553, "step": 12424 }, { "epoch": 1.28, "learning_rate": 4.573629888555645e-05, "loss": 0.5975, "step": 12425 }, { "epoch": 1.28, "learning_rate": 4.573420321936995e-05, "loss": 0.817, "step": 12426 }, { "epoch": 1.28, "learning_rate": 4.573210744726448e-05, "loss": 0.9727, "step": 12427 }, { "epoch": 1.28, "learning_rate": 4.573001156925414e-05, "loss": 0.1551, "step": 12428 }, { "epoch": 1.28, "learning_rate": 4.5727915585353036e-05, "loss": 0.579, "step": 12429 }, { "epoch": 1.28, "learning_rate": 4.572581949557529e-05, "loss": 0.798, "step": 12430 }, { "epoch": 1.28, "learning_rate": 4.572372329993501e-05, "loss": 0.6761, "step": 12431 }, { "epoch": 1.28, "learning_rate": 4.5721626998446304e-05, "loss": 0.6991, "step": 12432 }, { "epoch": 1.28, "learning_rate": 4.571953059112327e-05, "loss": 0.2757, "step": 12433 }, { "epoch": 1.28, "learning_rate": 4.571743407798005e-05, "loss": 0.294, "step": 12434 }, { "epoch": 1.28, "learning_rate": 4.571533745903073e-05, "loss": 0.8936, "step": 12435 }, { "epoch": 1.28, "learning_rate": 4.5713240734289444e-05, "loss": 0.9809, "step": 12436 }, { "epoch": 1.28, "learning_rate": 4.57111439037703e-05, "loss": 0.6782, "step": 12437 }, { "epoch": 1.28, "learning_rate": 4.57090469674874e-05, "loss": 0.3738, "step": 12438 }, { "epoch": 1.28, "learning_rate": 4.570694992545487e-05, "loss": 0.9687, "step": 12439 }, { "epoch": 1.28, "learning_rate": 4.570485277768684e-05, "loss": 0.8326, "step": 12440 }, { "epoch": 1.28, "learning_rate": 4.570275552419742e-05, "loss": 0.7404, "step": 12441 }, { "epoch": 1.28, "learning_rate": 4.570065816500071e-05, "loss": 0.3276, "step": 12442 }, { "epoch": 1.28, "learning_rate": 4.569856070011085e-05, "loss": 0.4525, "step": 12443 }, { "epoch": 1.28, "learning_rate": 4.569646312954195e-05, "loss": 0.3143, "step": 12444 }, { "epoch": 1.28, "learning_rate": 4.569436545330813e-05, "loss": 0.402, "step": 12445 }, { "epoch": 1.28, "learning_rate": 4.569226767142352e-05, "loss": 0.5255, "step": 12446 }, { "epoch": 1.28, "learning_rate": 4.5690169783902233e-05, "loss": 0.3534, "step": 12447 }, { "epoch": 1.28, "learning_rate": 4.56880717907584e-05, "loss": 0.8494, "step": 12448 }, { "epoch": 1.28, "learning_rate": 4.5685973692006127e-05, "loss": 0.5972, "step": 12449 }, { "epoch": 1.28, "learning_rate": 4.568387548765955e-05, "loss": 0.7002, "step": 12450 }, { "epoch": 1.28, "learning_rate": 4.5681777177732805e-05, "loss": 0.1291, "step": 12451 }, { "epoch": 1.29, "learning_rate": 4.567967876224e-05, "loss": 0.8074, "step": 12452 }, { "epoch": 1.29, "learning_rate": 4.5677580241195264e-05, "loss": 0.2161, "step": 12453 }, { "epoch": 1.29, "learning_rate": 4.5675481614612725e-05, "loss": 0.7973, "step": 12454 }, { "epoch": 1.29, "learning_rate": 4.567338288250651e-05, "loss": 0.6555, "step": 12455 }, { "epoch": 1.29, "learning_rate": 4.5671284044890755e-05, "loss": 0.9097, "step": 12456 }, { "epoch": 1.29, "learning_rate": 4.5669185101779586e-05, "loss": 0.5947, "step": 12457 }, { "epoch": 1.29, "learning_rate": 4.566708605318713e-05, "loss": 0.9239, "step": 12458 }, { "epoch": 1.29, "learning_rate": 4.56649868991275e-05, "loss": 0.7233, "step": 12459 }, { "epoch": 1.29, "learning_rate": 4.566288763961486e-05, "loss": 0.8751, "step": 12460 }, { "epoch": 1.29, "learning_rate": 4.566078827466331e-05, "loss": 0.2241, "step": 12461 }, { "epoch": 1.29, "learning_rate": 4.5658688804287e-05, "loss": 0.6392, "step": 12462 }, { "epoch": 1.29, "learning_rate": 4.565658922850008e-05, "loss": 0.7572, "step": 12463 }, { "epoch": 1.29, "learning_rate": 4.5654489547316645e-05, "loss": 0.3805, "step": 12464 }, { "epoch": 1.29, "learning_rate": 4.565238976075086e-05, "loss": 0.838, "step": 12465 }, { "epoch": 1.29, "learning_rate": 4.565028986881685e-05, "loss": 1.0966, "step": 12466 }, { "epoch": 1.29, "learning_rate": 4.564818987152874e-05, "loss": 1.1811, "step": 12467 }, { "epoch": 1.29, "learning_rate": 4.564608976890068e-05, "loss": 0.7679, "step": 12468 }, { "epoch": 1.29, "learning_rate": 4.564398956094681e-05, "loss": 1.0715, "step": 12469 }, { "epoch": 1.29, "learning_rate": 4.564188924768127e-05, "loss": 0.7177, "step": 12470 }, { "epoch": 1.29, "learning_rate": 4.563978882911818e-05, "loss": 0.3653, "step": 12471 }, { "epoch": 1.29, "learning_rate": 4.563768830527168e-05, "loss": 0.7239, "step": 12472 }, { "epoch": 1.29, "learning_rate": 4.5635587676155934e-05, "loss": 0.7097, "step": 12473 }, { "epoch": 1.29, "learning_rate": 4.563348694178507e-05, "loss": 0.8077, "step": 12474 }, { "epoch": 1.29, "learning_rate": 4.563138610217324e-05, "loss": 0.8935, "step": 12475 }, { "epoch": 1.29, "learning_rate": 4.562928515733457e-05, "loss": 0.7318, "step": 12476 }, { "epoch": 1.29, "learning_rate": 4.56271841072832e-05, "loss": 0.9929, "step": 12477 }, { "epoch": 1.29, "learning_rate": 4.562508295203329e-05, "loss": 0.2838, "step": 12478 }, { "epoch": 1.29, "learning_rate": 4.562298169159897e-05, "loss": 0.6789, "step": 12479 }, { "epoch": 1.29, "learning_rate": 4.56208803259944e-05, "loss": 0.5973, "step": 12480 }, { "epoch": 1.29, "learning_rate": 4.5618778855233714e-05, "loss": 0.1912, "step": 12481 }, { "epoch": 1.29, "learning_rate": 4.561667727933107e-05, "loss": 0.0772, "step": 12482 }, { "epoch": 1.29, "learning_rate": 4.5614575598300604e-05, "loss": 0.6538, "step": 12483 }, { "epoch": 1.29, "learning_rate": 4.561247381215647e-05, "loss": 0.6117, "step": 12484 }, { "epoch": 1.29, "learning_rate": 4.5610371920912815e-05, "loss": 0.7895, "step": 12485 }, { "epoch": 1.29, "learning_rate": 4.560826992458379e-05, "loss": 0.9856, "step": 12486 }, { "epoch": 1.29, "learning_rate": 4.5606167823183535e-05, "loss": 0.6566, "step": 12487 }, { "epoch": 1.29, "learning_rate": 4.5604065616726224e-05, "loss": 0.818, "step": 12488 }, { "epoch": 1.29, "learning_rate": 4.5601963305226e-05, "loss": 0.6178, "step": 12489 }, { "epoch": 1.29, "learning_rate": 4.5599860888696994e-05, "loss": 0.4649, "step": 12490 }, { "epoch": 1.29, "learning_rate": 4.559775836715337e-05, "loss": 0.7442, "step": 12491 }, { "epoch": 1.29, "learning_rate": 4.559565574060929e-05, "loss": 0.7898, "step": 12492 }, { "epoch": 1.29, "learning_rate": 4.5593553009078914e-05, "loss": 0.6419, "step": 12493 }, { "epoch": 1.29, "learning_rate": 4.5591450172576385e-05, "loss": 0.1445, "step": 12494 }, { "epoch": 1.29, "learning_rate": 4.5589347231115856e-05, "loss": 1.0962, "step": 12495 }, { "epoch": 1.29, "learning_rate": 4.5587244184711496e-05, "loss": 0.1882, "step": 12496 }, { "epoch": 1.29, "learning_rate": 4.558514103337745e-05, "loss": 2.3843, "step": 12497 }, { "epoch": 1.29, "learning_rate": 4.558303777712789e-05, "loss": 0.4723, "step": 12498 }, { "epoch": 1.29, "learning_rate": 4.5580934415976954e-05, "loss": 0.9391, "step": 12499 }, { "epoch": 1.29, "learning_rate": 4.557883094993882e-05, "loss": 1.2723, "step": 12500 }, { "epoch": 1.29, "learning_rate": 4.557672737902764e-05, "loss": 0.1847, "step": 12501 }, { "epoch": 1.29, "learning_rate": 4.5574623703257567e-05, "loss": 1.0731, "step": 12502 }, { "epoch": 1.29, "learning_rate": 4.5572519922642784e-05, "loss": 0.2955, "step": 12503 }, { "epoch": 1.29, "learning_rate": 4.557041603719743e-05, "loss": 0.4435, "step": 12504 }, { "epoch": 1.29, "learning_rate": 4.556831204693568e-05, "loss": 0.5665, "step": 12505 }, { "epoch": 1.29, "learning_rate": 4.55662079518717e-05, "loss": 0.189, "step": 12506 }, { "epoch": 1.29, "learning_rate": 4.5564103752019664e-05, "loss": 0.9062, "step": 12507 }, { "epoch": 1.29, "learning_rate": 4.5561999447393715e-05, "loss": 0.5525, "step": 12508 }, { "epoch": 1.29, "learning_rate": 4.555989503800801e-05, "loss": 0.5086, "step": 12509 }, { "epoch": 1.29, "learning_rate": 4.5557790523876744e-05, "loss": 0.7428, "step": 12510 }, { "epoch": 1.29, "learning_rate": 4.5555685905014074e-05, "loss": 0.2765, "step": 12511 }, { "epoch": 1.29, "learning_rate": 4.5553581181434164e-05, "loss": 0.5386, "step": 12512 }, { "epoch": 1.29, "learning_rate": 4.555147635315119e-05, "loss": 0.372, "step": 12513 }, { "epoch": 1.29, "learning_rate": 4.554937142017931e-05, "loss": 0.5133, "step": 12514 }, { "epoch": 1.29, "learning_rate": 4.55472663825327e-05, "loss": 0.433, "step": 12515 }, { "epoch": 1.29, "learning_rate": 4.554516124022553e-05, "loss": 0.2679, "step": 12516 }, { "epoch": 1.29, "learning_rate": 4.554305599327198e-05, "loss": 1.2887, "step": 12517 }, { "epoch": 1.29, "learning_rate": 4.554095064168621e-05, "loss": 0.4259, "step": 12518 }, { "epoch": 1.29, "learning_rate": 4.553884518548239e-05, "loss": 0.4212, "step": 12519 }, { "epoch": 1.29, "learning_rate": 4.5536739624674704e-05, "loss": 0.6586, "step": 12520 }, { "epoch": 1.29, "learning_rate": 4.553463395927732e-05, "loss": 0.8679, "step": 12521 }, { "epoch": 1.29, "learning_rate": 4.5532528189304425e-05, "loss": 0.9767, "step": 12522 }, { "epoch": 1.29, "learning_rate": 4.553042231477016e-05, "loss": 0.4639, "step": 12523 }, { "epoch": 1.29, "learning_rate": 4.5528316335688745e-05, "loss": 1.0194, "step": 12524 }, { "epoch": 1.29, "learning_rate": 4.5526210252074334e-05, "loss": 0.3995, "step": 12525 }, { "epoch": 1.29, "learning_rate": 4.552410406394111e-05, "loss": 0.4405, "step": 12526 }, { "epoch": 1.29, "learning_rate": 4.552199777130326e-05, "loss": 0.4783, "step": 12527 }, { "epoch": 1.29, "learning_rate": 4.551989137417494e-05, "loss": 0.2392, "step": 12528 }, { "epoch": 1.29, "learning_rate": 4.551778487257034e-05, "loss": 1.1232, "step": 12529 }, { "epoch": 1.29, "learning_rate": 4.5515678266503645e-05, "loss": 0.1572, "step": 12530 }, { "epoch": 1.29, "learning_rate": 4.551357155598905e-05, "loss": 0.9916, "step": 12531 }, { "epoch": 1.29, "learning_rate": 4.551146474104071e-05, "loss": 0.2546, "step": 12532 }, { "epoch": 1.29, "learning_rate": 4.550935782167282e-05, "loss": 0.8401, "step": 12533 }, { "epoch": 1.29, "learning_rate": 4.550725079789955e-05, "loss": 0.545, "step": 12534 }, { "epoch": 1.29, "learning_rate": 4.5505143669735104e-05, "loss": 1.1694, "step": 12535 }, { "epoch": 1.29, "learning_rate": 4.550303643719367e-05, "loss": 0.6262, "step": 12536 }, { "epoch": 1.29, "learning_rate": 4.55009291002894e-05, "loss": 0.85, "step": 12537 }, { "epoch": 1.29, "learning_rate": 4.549882165903652e-05, "loss": 0.2604, "step": 12538 }, { "epoch": 1.29, "learning_rate": 4.5496714113449193e-05, "loss": 0.901, "step": 12539 }, { "epoch": 1.29, "learning_rate": 4.549460646354161e-05, "loss": 0.7748, "step": 12540 }, { "epoch": 1.29, "learning_rate": 4.549249870932797e-05, "loss": 0.272, "step": 12541 }, { "epoch": 1.29, "learning_rate": 4.5490390850822444e-05, "loss": 0.9252, "step": 12542 }, { "epoch": 1.29, "learning_rate": 4.5488282888039236e-05, "loss": 0.3562, "step": 12543 }, { "epoch": 1.29, "learning_rate": 4.548617482099253e-05, "loss": 0.9036, "step": 12544 }, { "epoch": 1.29, "learning_rate": 4.5484066649696525e-05, "loss": 1.5532, "step": 12545 }, { "epoch": 1.29, "learning_rate": 4.54819583741654e-05, "loss": 0.6853, "step": 12546 }, { "epoch": 1.29, "learning_rate": 4.547984999441336e-05, "loss": 1.2483, "step": 12547 }, { "epoch": 1.3, "learning_rate": 4.547774151045458e-05, "loss": 1.1077, "step": 12548 }, { "epoch": 1.3, "learning_rate": 4.547563292230327e-05, "loss": 0.4486, "step": 12549 }, { "epoch": 1.3, "learning_rate": 4.547352422997363e-05, "loss": 1.069, "step": 12550 }, { "epoch": 1.3, "learning_rate": 4.547141543347984e-05, "loss": 0.7234, "step": 12551 }, { "epoch": 1.3, "learning_rate": 4.5469306532836096e-05, "loss": 0.5749, "step": 12552 }, { "epoch": 1.3, "learning_rate": 4.54671975280566e-05, "loss": 0.1135, "step": 12553 }, { "epoch": 1.3, "learning_rate": 4.5465088419155555e-05, "loss": 0.4021, "step": 12554 }, { "epoch": 1.3, "learning_rate": 4.546297920614715e-05, "loss": 1.0394, "step": 12555 }, { "epoch": 1.3, "learning_rate": 4.5460869889045584e-05, "loss": 0.4966, "step": 12556 }, { "epoch": 1.3, "learning_rate": 4.545876046786507e-05, "loss": 0.3522, "step": 12557 }, { "epoch": 1.3, "learning_rate": 4.5456650942619787e-05, "loss": 0.6411, "step": 12558 }, { "epoch": 1.3, "learning_rate": 4.545454131332394e-05, "loss": 1.2237, "step": 12559 }, { "epoch": 1.3, "learning_rate": 4.545243157999176e-05, "loss": 1.0945, "step": 12560 }, { "epoch": 1.3, "learning_rate": 4.545032174263741e-05, "loss": 0.4079, "step": 12561 }, { "epoch": 1.3, "learning_rate": 4.544821180127511e-05, "loss": 0.475, "step": 12562 }, { "epoch": 1.3, "learning_rate": 4.544610175591907e-05, "loss": 0.5368, "step": 12563 }, { "epoch": 1.3, "learning_rate": 4.5443991606583485e-05, "loss": 0.41, "step": 12564 }, { "epoch": 1.3, "learning_rate": 4.544188135328255e-05, "loss": 0.301, "step": 12565 }, { "epoch": 1.3, "learning_rate": 4.54397709960305e-05, "loss": 1.0935, "step": 12566 }, { "epoch": 1.3, "learning_rate": 4.543766053484152e-05, "loss": 0.367, "step": 12567 }, { "epoch": 1.3, "learning_rate": 4.543554996972981e-05, "loss": 0.5061, "step": 12568 }, { "epoch": 1.3, "learning_rate": 4.5433439300709596e-05, "loss": 1.2262, "step": 12569 }, { "epoch": 1.3, "learning_rate": 4.543132852779509e-05, "loss": 0.6692, "step": 12570 }, { "epoch": 1.3, "learning_rate": 4.5429217651000475e-05, "loss": 0.68, "step": 12571 }, { "epoch": 1.3, "learning_rate": 4.542710667033999e-05, "loss": 0.6527, "step": 12572 }, { "epoch": 1.3, "learning_rate": 4.542499558582782e-05, "loss": 0.9833, "step": 12573 }, { "epoch": 1.3, "learning_rate": 4.54228843974782e-05, "loss": 0.2754, "step": 12574 }, { "epoch": 1.3, "learning_rate": 4.542077310530532e-05, "loss": 0.8655, "step": 12575 }, { "epoch": 1.3, "learning_rate": 4.5418661709323414e-05, "loss": 0.1275, "step": 12576 }, { "epoch": 1.3, "learning_rate": 4.541655020954668e-05, "loss": 0.6477, "step": 12577 }, { "epoch": 1.3, "learning_rate": 4.541443860598934e-05, "loss": 0.7981, "step": 12578 }, { "epoch": 1.3, "learning_rate": 4.54123268986656e-05, "loss": 0.7761, "step": 12579 }, { "epoch": 1.3, "learning_rate": 4.541021508758969e-05, "loss": 1.187, "step": 12580 }, { "epoch": 1.3, "learning_rate": 4.5408103172775813e-05, "loss": 0.7874, "step": 12581 }, { "epoch": 1.3, "learning_rate": 4.5405991154238195e-05, "loss": 0.8364, "step": 12582 }, { "epoch": 1.3, "learning_rate": 4.540387903199104e-05, "loss": 0.3091, "step": 12583 }, { "epoch": 1.3, "learning_rate": 4.5401766806048584e-05, "loss": 0.2535, "step": 12584 }, { "epoch": 1.3, "learning_rate": 4.539965447642503e-05, "loss": 0.3274, "step": 12585 }, { "epoch": 1.3, "learning_rate": 4.5397542043134606e-05, "loss": 0.8593, "step": 12586 }, { "epoch": 1.3, "learning_rate": 4.539542950619154e-05, "loss": 0.4281, "step": 12587 }, { "epoch": 1.3, "learning_rate": 4.539331686561004e-05, "loss": 1.5963, "step": 12588 }, { "epoch": 1.3, "learning_rate": 4.5391204121404335e-05, "loss": 0.6627, "step": 12589 }, { "epoch": 1.3, "learning_rate": 4.538909127358864e-05, "loss": 1.0642, "step": 12590 }, { "epoch": 1.3, "learning_rate": 4.5386978322177185e-05, "loss": 0.2911, "step": 12591 }, { "epoch": 1.3, "learning_rate": 4.5384865267184194e-05, "loss": 0.5187, "step": 12592 }, { "epoch": 1.3, "learning_rate": 4.5382752108623895e-05, "loss": 0.5973, "step": 12593 }, { "epoch": 1.3, "learning_rate": 4.53806388465105e-05, "loss": 0.0598, "step": 12594 }, { "epoch": 1.3, "learning_rate": 4.537852548085825e-05, "loss": 0.7136, "step": 12595 }, { "epoch": 1.3, "learning_rate": 4.537641201168136e-05, "loss": 0.1398, "step": 12596 }, { "epoch": 1.3, "learning_rate": 4.5374298438994054e-05, "loss": 0.4592, "step": 12597 }, { "epoch": 1.3, "learning_rate": 4.5372184762810586e-05, "loss": 0.4602, "step": 12598 }, { "epoch": 1.3, "learning_rate": 4.537007098314515e-05, "loss": 0.689, "step": 12599 }, { "epoch": 1.3, "learning_rate": 4.5367957100012e-05, "loss": 0.5772, "step": 12600 }, { "epoch": 1.3, "learning_rate": 4.5365843113425365e-05, "loss": 0.5661, "step": 12601 }, { "epoch": 1.3, "learning_rate": 4.5363729023399455e-05, "loss": 0.2412, "step": 12602 }, { "epoch": 1.3, "learning_rate": 4.536161482994853e-05, "loss": 0.3368, "step": 12603 }, { "epoch": 1.3, "learning_rate": 4.53595005330868e-05, "loss": 0.8799, "step": 12604 }, { "epoch": 1.3, "learning_rate": 4.5357386132828504e-05, "loss": 1.1296, "step": 12605 }, { "epoch": 1.3, "learning_rate": 4.535527162918789e-05, "loss": 0.2403, "step": 12606 }, { "epoch": 1.3, "learning_rate": 4.5353157022179165e-05, "loss": 0.4063, "step": 12607 }, { "epoch": 1.3, "learning_rate": 4.5351042311816594e-05, "loss": 0.8237, "step": 12608 }, { "epoch": 1.3, "learning_rate": 4.534892749811439e-05, "loss": 0.5503, "step": 12609 }, { "epoch": 1.3, "learning_rate": 4.534681258108679e-05, "loss": 0.6039, "step": 12610 }, { "epoch": 1.3, "learning_rate": 4.534469756074805e-05, "loss": 0.8071, "step": 12611 }, { "epoch": 1.3, "learning_rate": 4.5342582437112395e-05, "loss": 0.7798, "step": 12612 }, { "epoch": 1.3, "learning_rate": 4.5340467210194066e-05, "loss": 0.815, "step": 12613 }, { "epoch": 1.3, "learning_rate": 4.5338351880007295e-05, "loss": 0.5167, "step": 12614 }, { "epoch": 1.3, "learning_rate": 4.533623644656633e-05, "loss": 0.5656, "step": 12615 }, { "epoch": 1.3, "learning_rate": 4.533412090988541e-05, "loss": 0.5889, "step": 12616 }, { "epoch": 1.3, "learning_rate": 4.5332005269978775e-05, "loss": 0.6993, "step": 12617 }, { "epoch": 1.3, "learning_rate": 4.532988952686068e-05, "loss": 0.2443, "step": 12618 }, { "epoch": 1.3, "learning_rate": 4.532777368054534e-05, "loss": 0.6117, "step": 12619 }, { "epoch": 1.3, "learning_rate": 4.5325657731047026e-05, "loss": 0.7462, "step": 12620 }, { "epoch": 1.3, "learning_rate": 4.5323541678379966e-05, "loss": 0.5159, "step": 12621 }, { "epoch": 1.3, "learning_rate": 4.53214255225584e-05, "loss": 0.3728, "step": 12622 }, { "epoch": 1.3, "learning_rate": 4.531930926359658e-05, "loss": 0.5366, "step": 12623 }, { "epoch": 1.3, "learning_rate": 4.5317192901508765e-05, "loss": 0.3828, "step": 12624 }, { "epoch": 1.3, "learning_rate": 4.531507643630919e-05, "loss": 0.4866, "step": 12625 }, { "epoch": 1.3, "learning_rate": 4.531295986801211e-05, "loss": 0.276, "step": 12626 }, { "epoch": 1.3, "learning_rate": 4.5310843196631764e-05, "loss": 0.61, "step": 12627 }, { "epoch": 1.3, "learning_rate": 4.53087264221824e-05, "loss": 0.5955, "step": 12628 }, { "epoch": 1.3, "learning_rate": 4.530660954467827e-05, "loss": 0.5328, "step": 12629 }, { "epoch": 1.3, "learning_rate": 4.5304492564133624e-05, "loss": 0.3555, "step": 12630 }, { "epoch": 1.3, "learning_rate": 4.5302375480562726e-05, "loss": 0.3022, "step": 12631 }, { "epoch": 1.3, "learning_rate": 4.530025829397981e-05, "loss": 0.6986, "step": 12632 }, { "epoch": 1.3, "learning_rate": 4.529814100439914e-05, "loss": 0.5805, "step": 12633 }, { "epoch": 1.3, "learning_rate": 4.529602361183495e-05, "loss": 0.607, "step": 12634 }, { "epoch": 1.3, "learning_rate": 4.529390611630151e-05, "loss": 0.7155, "step": 12635 }, { "epoch": 1.3, "learning_rate": 4.529178851781309e-05, "loss": 0.2083, "step": 12636 }, { "epoch": 1.3, "learning_rate": 4.528967081638392e-05, "loss": 1.1902, "step": 12637 }, { "epoch": 1.3, "learning_rate": 4.5287553012028255e-05, "loss": 0.5208, "step": 12638 }, { "epoch": 1.3, "learning_rate": 4.528543510476037e-05, "loss": 0.7211, "step": 12639 }, { "epoch": 1.3, "learning_rate": 4.5283317094594505e-05, "loss": 1.0103, "step": 12640 }, { "epoch": 1.3, "learning_rate": 4.528119898154494e-05, "loss": 1.9132, "step": 12641 }, { "epoch": 1.3, "learning_rate": 4.5279080765625895e-05, "loss": 0.9702, "step": 12642 }, { "epoch": 1.3, "learning_rate": 4.5276962446851665e-05, "loss": 0.7892, "step": 12643 }, { "epoch": 1.31, "learning_rate": 4.52748440252365e-05, "loss": 0.1675, "step": 12644 }, { "epoch": 1.31, "learning_rate": 4.5272725500794663e-05, "loss": 0.4087, "step": 12645 }, { "epoch": 1.31, "learning_rate": 4.527060687354041e-05, "loss": 0.7499, "step": 12646 }, { "epoch": 1.31, "learning_rate": 4.5268488143488e-05, "loss": 0.4455, "step": 12647 }, { "epoch": 1.31, "learning_rate": 4.52663693106517e-05, "loss": 0.4218, "step": 12648 }, { "epoch": 1.31, "learning_rate": 4.5264250375045775e-05, "loss": 0.9621, "step": 12649 }, { "epoch": 1.31, "learning_rate": 4.52621313366845e-05, "loss": 0.7528, "step": 12650 }, { "epoch": 1.31, "learning_rate": 4.526001219558212e-05, "loss": 0.4609, "step": 12651 }, { "epoch": 1.31, "learning_rate": 4.52578929517529e-05, "loss": 1.2267, "step": 12652 }, { "epoch": 1.31, "learning_rate": 4.525577360521112e-05, "loss": 0.8087, "step": 12653 }, { "epoch": 1.31, "learning_rate": 4.525365415597104e-05, "loss": 0.253, "step": 12654 }, { "epoch": 1.31, "learning_rate": 4.525153460404693e-05, "loss": 0.3903, "step": 12655 }, { "epoch": 1.31, "learning_rate": 4.524941494945306e-05, "loss": 0.5455, "step": 12656 }, { "epoch": 1.31, "learning_rate": 4.524729519220369e-05, "loss": 0.1876, "step": 12657 }, { "epoch": 1.31, "learning_rate": 4.5245175332313106e-05, "loss": 0.4297, "step": 12658 }, { "epoch": 1.31, "learning_rate": 4.524305536979556e-05, "loss": 0.729, "step": 12659 }, { "epoch": 1.31, "learning_rate": 4.524093530466534e-05, "loss": 0.5271, "step": 12660 }, { "epoch": 1.31, "learning_rate": 4.5238815136936705e-05, "loss": 0.5163, "step": 12661 }, { "epoch": 1.31, "learning_rate": 4.5236694866623935e-05, "loss": 1.3266, "step": 12662 }, { "epoch": 1.31, "learning_rate": 4.5234574493741296e-05, "loss": 1.4583, "step": 12663 }, { "epoch": 1.31, "learning_rate": 4.5232454018303066e-05, "loss": 0.4691, "step": 12664 }, { "epoch": 1.31, "learning_rate": 4.523033344032353e-05, "loss": 0.1947, "step": 12665 }, { "epoch": 1.31, "learning_rate": 4.522821275981694e-05, "loss": 0.5541, "step": 12666 }, { "epoch": 1.31, "learning_rate": 4.522609197679759e-05, "loss": 1.2004, "step": 12667 }, { "epoch": 1.31, "learning_rate": 4.522397109127975e-05, "loss": 0.6436, "step": 12668 }, { "epoch": 1.31, "learning_rate": 4.5221850103277695e-05, "loss": 0.7581, "step": 12669 }, { "epoch": 1.31, "learning_rate": 4.521972901280572e-05, "loss": 0.2648, "step": 12670 }, { "epoch": 1.31, "learning_rate": 4.521760781987808e-05, "loss": 0.6932, "step": 12671 }, { "epoch": 1.31, "learning_rate": 4.521548652450906e-05, "loss": 0.4159, "step": 12672 }, { "epoch": 1.31, "learning_rate": 4.521336512671296e-05, "loss": 0.6788, "step": 12673 }, { "epoch": 1.31, "learning_rate": 4.521124362650403e-05, "loss": 0.5022, "step": 12674 }, { "epoch": 1.31, "learning_rate": 4.5209122023896576e-05, "loss": 0.6701, "step": 12675 }, { "epoch": 1.31, "learning_rate": 4.520700031890487e-05, "loss": 0.382, "step": 12676 }, { "epoch": 1.31, "learning_rate": 4.52048785115432e-05, "loss": 0.5534, "step": 12677 }, { "epoch": 1.31, "learning_rate": 4.520275660182583e-05, "loss": 0.6866, "step": 12678 }, { "epoch": 1.31, "learning_rate": 4.520063458976709e-05, "loss": 1.2291, "step": 12679 }, { "epoch": 1.31, "learning_rate": 4.5198512475381206e-05, "loss": 0.4292, "step": 12680 }, { "epoch": 1.31, "learning_rate": 4.51963902586825e-05, "loss": 0.8903, "step": 12681 }, { "epoch": 1.31, "learning_rate": 4.5194267939685245e-05, "loss": 0.8048, "step": 12682 }, { "epoch": 1.31, "learning_rate": 4.519214551840375e-05, "loss": 0.6498, "step": 12683 }, { "epoch": 1.31, "learning_rate": 4.5190022994852285e-05, "loss": 0.3398, "step": 12684 }, { "epoch": 1.31, "learning_rate": 4.518790036904512e-05, "loss": 0.6197, "step": 12685 }, { "epoch": 1.31, "learning_rate": 4.518577764099657e-05, "loss": 0.4862, "step": 12686 }, { "epoch": 1.31, "learning_rate": 4.518365481072093e-05, "loss": 0.3666, "step": 12687 }, { "epoch": 1.31, "learning_rate": 4.5181531878232456e-05, "loss": 0.4194, "step": 12688 }, { "epoch": 1.31, "learning_rate": 4.5179408843545494e-05, "loss": 0.5427, "step": 12689 }, { "epoch": 1.31, "learning_rate": 4.517728570667428e-05, "loss": 0.8727, "step": 12690 }, { "epoch": 1.31, "learning_rate": 4.517516246763314e-05, "loss": 0.7905, "step": 12691 }, { "epoch": 1.31, "learning_rate": 4.5173039126436346e-05, "loss": 1.079, "step": 12692 }, { "epoch": 1.31, "learning_rate": 4.517091568309821e-05, "loss": 1.1025, "step": 12693 }, { "epoch": 1.31, "learning_rate": 4.516879213763302e-05, "loss": 0.7604, "step": 12694 }, { "epoch": 1.31, "learning_rate": 4.5166668490055074e-05, "loss": 0.8876, "step": 12695 }, { "epoch": 1.31, "learning_rate": 4.5164544740378656e-05, "loss": 0.3352, "step": 12696 }, { "epoch": 1.31, "learning_rate": 4.516242088861807e-05, "loss": 0.3041, "step": 12697 }, { "epoch": 1.31, "learning_rate": 4.516029693478762e-05, "loss": 0.6327, "step": 12698 }, { "epoch": 1.31, "learning_rate": 4.515817287890161e-05, "loss": 0.7542, "step": 12699 }, { "epoch": 1.31, "learning_rate": 4.51560487209743e-05, "loss": 0.9496, "step": 12700 }, { "epoch": 1.31, "learning_rate": 4.515392446102003e-05, "loss": 1.3224, "step": 12701 }, { "epoch": 1.31, "learning_rate": 4.5151800099053094e-05, "loss": 0.6648, "step": 12702 }, { "epoch": 1.31, "learning_rate": 4.514967563508778e-05, "loss": 0.5164, "step": 12703 }, { "epoch": 1.31, "learning_rate": 4.5147551069138394e-05, "loss": 0.9049, "step": 12704 }, { "epoch": 1.31, "learning_rate": 4.5145426401219233e-05, "loss": 0.7552, "step": 12705 }, { "epoch": 1.31, "learning_rate": 4.51433016313446e-05, "loss": 0.1654, "step": 12706 }, { "epoch": 1.31, "learning_rate": 4.5141176759528814e-05, "loss": 1.2999, "step": 12707 }, { "epoch": 1.31, "learning_rate": 4.513905178578617e-05, "loss": 0.5941, "step": 12708 }, { "epoch": 1.31, "learning_rate": 4.513692671013097e-05, "loss": 0.1886, "step": 12709 }, { "epoch": 1.31, "learning_rate": 4.5134801532577525e-05, "loss": 0.5665, "step": 12710 }, { "epoch": 1.31, "learning_rate": 4.513267625314012e-05, "loss": 0.2115, "step": 12711 }, { "epoch": 1.31, "learning_rate": 4.51305508718331e-05, "loss": 0.6291, "step": 12712 }, { "epoch": 1.31, "learning_rate": 4.512842538867074e-05, "loss": 1.0429, "step": 12713 }, { "epoch": 1.31, "learning_rate": 4.5126299803667374e-05, "loss": 0.5828, "step": 12714 }, { "epoch": 1.31, "learning_rate": 4.5124174116837285e-05, "loss": 0.6751, "step": 12715 }, { "epoch": 1.31, "learning_rate": 4.5122048328194785e-05, "loss": 0.5425, "step": 12716 }, { "epoch": 1.31, "learning_rate": 4.511992243775422e-05, "loss": 0.8169, "step": 12717 }, { "epoch": 1.31, "learning_rate": 4.511779644552985e-05, "loss": 0.6076, "step": 12718 }, { "epoch": 1.31, "learning_rate": 4.5115670351536024e-05, "loss": 0.2092, "step": 12719 }, { "epoch": 1.31, "learning_rate": 4.511354415578704e-05, "loss": 1.2129, "step": 12720 }, { "epoch": 1.31, "learning_rate": 4.5111417858297215e-05, "loss": 0.4481, "step": 12721 }, { "epoch": 1.31, "learning_rate": 4.5109291459080853e-05, "loss": 0.7481, "step": 12722 }, { "epoch": 1.31, "learning_rate": 4.510716495815228e-05, "loss": 0.4163, "step": 12723 }, { "epoch": 1.31, "learning_rate": 4.51050383555258e-05, "loss": 0.8016, "step": 12724 }, { "epoch": 1.31, "learning_rate": 4.510291165121575e-05, "loss": 0.5397, "step": 12725 }, { "epoch": 1.31, "learning_rate": 4.510078484523643e-05, "loss": 0.5597, "step": 12726 }, { "epoch": 1.31, "learning_rate": 4.509865793760217e-05, "loss": 0.7428, "step": 12727 }, { "epoch": 1.31, "learning_rate": 4.509653092832725e-05, "loss": 0.6047, "step": 12728 }, { "epoch": 1.31, "learning_rate": 4.509440381742604e-05, "loss": 0.3345, "step": 12729 }, { "epoch": 1.31, "learning_rate": 4.509227660491282e-05, "loss": 0.2827, "step": 12730 }, { "epoch": 1.31, "learning_rate": 4.5090149290801926e-05, "loss": 0.1433, "step": 12731 }, { "epoch": 1.31, "learning_rate": 4.508802187510769e-05, "loss": 0.6465, "step": 12732 }, { "epoch": 1.31, "learning_rate": 4.5085894357844415e-05, "loss": 0.5686, "step": 12733 }, { "epoch": 1.31, "learning_rate": 4.508376673902643e-05, "loss": 0.5948, "step": 12734 }, { "epoch": 1.31, "learning_rate": 4.508163901866805e-05, "loss": 0.5432, "step": 12735 }, { "epoch": 1.31, "learning_rate": 4.507951119678361e-05, "loss": 0.3379, "step": 12736 }, { "epoch": 1.31, "learning_rate": 4.507738327338743e-05, "loss": 0.6203, "step": 12737 }, { "epoch": 1.31, "learning_rate": 4.5075255248493834e-05, "loss": 0.5281, "step": 12738 }, { "epoch": 1.31, "learning_rate": 4.507312712211715e-05, "loss": 0.5091, "step": 12739 }, { "epoch": 1.32, "learning_rate": 4.5070998894271694e-05, "loss": 0.615, "step": 12740 }, { "epoch": 1.32, "learning_rate": 4.50688705649718e-05, "loss": 0.9529, "step": 12741 }, { "epoch": 1.32, "learning_rate": 4.506674213423179e-05, "loss": 0.2813, "step": 12742 }, { "epoch": 1.32, "learning_rate": 4.5064613602066e-05, "loss": 1.3762, "step": 12743 }, { "epoch": 1.32, "learning_rate": 4.5062484968488765e-05, "loss": 0.9464, "step": 12744 }, { "epoch": 1.32, "learning_rate": 4.50603562335144e-05, "loss": 1.2938, "step": 12745 }, { "epoch": 1.32, "learning_rate": 4.505822739715725e-05, "loss": 0.6476, "step": 12746 }, { "epoch": 1.32, "learning_rate": 4.505609845943163e-05, "loss": 0.5595, "step": 12747 }, { "epoch": 1.32, "learning_rate": 4.505396942035187e-05, "loss": 0.294, "step": 12748 }, { "epoch": 1.32, "learning_rate": 4.505184027993232e-05, "loss": 1.1183, "step": 12749 }, { "epoch": 1.32, "learning_rate": 4.504971103818729e-05, "loss": 0.1162, "step": 12750 }, { "epoch": 1.32, "learning_rate": 4.504758169513115e-05, "loss": 0.9554, "step": 12751 }, { "epoch": 1.32, "learning_rate": 4.5045452250778195e-05, "loss": 1.0405, "step": 12752 }, { "epoch": 1.32, "learning_rate": 4.5043322705142775e-05, "loss": 1.1691, "step": 12753 }, { "epoch": 1.32, "learning_rate": 4.504119305823923e-05, "loss": 0.5221, "step": 12754 }, { "epoch": 1.32, "learning_rate": 4.503906331008189e-05, "loss": 0.7447, "step": 12755 }, { "epoch": 1.32, "learning_rate": 4.50369334606851e-05, "loss": 0.4223, "step": 12756 }, { "epoch": 1.32, "learning_rate": 4.503480351006319e-05, "loss": 0.3757, "step": 12757 }, { "epoch": 1.32, "learning_rate": 4.50326734582305e-05, "loss": 0.36, "step": 12758 }, { "epoch": 1.32, "learning_rate": 4.5030543305201366e-05, "loss": 1.092, "step": 12759 }, { "epoch": 1.32, "learning_rate": 4.502841305099014e-05, "loss": 1.257, "step": 12760 }, { "epoch": 1.32, "learning_rate": 4.502628269561115e-05, "loss": 0.5613, "step": 12761 }, { "epoch": 1.32, "learning_rate": 4.5024152239078726e-05, "loss": 0.2713, "step": 12762 }, { "epoch": 1.32, "learning_rate": 4.502202168140724e-05, "loss": 0.2891, "step": 12763 }, { "epoch": 1.32, "learning_rate": 4.501989102261102e-05, "loss": 0.9413, "step": 12764 }, { "epoch": 1.32, "learning_rate": 4.501776026270441e-05, "loss": 1.1765, "step": 12765 }, { "epoch": 1.32, "learning_rate": 4.5015629401701746e-05, "loss": 0.6899, "step": 12766 }, { "epoch": 1.32, "learning_rate": 4.501349843961737e-05, "loss": 0.2718, "step": 12767 }, { "epoch": 1.32, "learning_rate": 4.501136737646564e-05, "loss": 0.9772, "step": 12768 }, { "epoch": 1.32, "learning_rate": 4.500923621226091e-05, "loss": 1.045, "step": 12769 }, { "epoch": 1.32, "learning_rate": 4.5007104947017505e-05, "loss": 0.2583, "step": 12770 }, { "epoch": 1.32, "learning_rate": 4.500497358074978e-05, "loss": 1.0144, "step": 12771 }, { "epoch": 1.32, "learning_rate": 4.500284211347209e-05, "loss": 2.0371, "step": 12772 }, { "epoch": 1.32, "learning_rate": 4.5000710545198765e-05, "loss": 0.3851, "step": 12773 }, { "epoch": 1.32, "learning_rate": 4.499857887594417e-05, "loss": 0.5835, "step": 12774 }, { "epoch": 1.32, "learning_rate": 4.499644710572266e-05, "loss": 0.925, "step": 12775 }, { "epoch": 1.32, "learning_rate": 4.499431523454857e-05, "loss": 0.3191, "step": 12776 }, { "epoch": 1.32, "learning_rate": 4.4992183262436265e-05, "loss": 0.3007, "step": 12777 }, { "epoch": 1.32, "learning_rate": 4.499005118940008e-05, "loss": 0.6338, "step": 12778 }, { "epoch": 1.32, "learning_rate": 4.498791901545439e-05, "loss": 0.1829, "step": 12779 }, { "epoch": 1.32, "learning_rate": 4.4985786740613524e-05, "loss": 0.6463, "step": 12780 }, { "epoch": 1.32, "learning_rate": 4.498365436489185e-05, "loss": 0.658, "step": 12781 }, { "epoch": 1.32, "learning_rate": 4.498152188830372e-05, "loss": 0.3373, "step": 12782 }, { "epoch": 1.32, "learning_rate": 4.4979389310863506e-05, "loss": 0.2352, "step": 12783 }, { "epoch": 1.32, "learning_rate": 4.497725663258553e-05, "loss": 0.7558, "step": 12784 }, { "epoch": 1.32, "learning_rate": 4.497512385348417e-05, "loss": 0.4915, "step": 12785 }, { "epoch": 1.32, "learning_rate": 4.4972990973573785e-05, "loss": 0.7795, "step": 12786 }, { "epoch": 1.32, "learning_rate": 4.4970857992868726e-05, "loss": 0.5984, "step": 12787 }, { "epoch": 1.32, "learning_rate": 4.496872491138335e-05, "loss": 0.7537, "step": 12788 }, { "epoch": 1.32, "learning_rate": 4.4966591729132025e-05, "loss": 0.7455, "step": 12789 }, { "epoch": 1.32, "learning_rate": 4.4964458446129114e-05, "loss": 0.4911, "step": 12790 }, { "epoch": 1.32, "learning_rate": 4.496232506238896e-05, "loss": 0.2679, "step": 12791 }, { "epoch": 1.32, "learning_rate": 4.496019157792593e-05, "loss": 0.4981, "step": 12792 }, { "epoch": 1.32, "learning_rate": 4.495805799275439e-05, "loss": 0.3414, "step": 12793 }, { "epoch": 1.32, "learning_rate": 4.495592430688872e-05, "loss": 0.8891, "step": 12794 }, { "epoch": 1.32, "learning_rate": 4.495379052034325e-05, "loss": 0.9125, "step": 12795 }, { "epoch": 1.32, "learning_rate": 4.495165663313237e-05, "loss": 0.579, "step": 12796 }, { "epoch": 1.32, "learning_rate": 4.494952264527044e-05, "loss": 0.4659, "step": 12797 }, { "epoch": 1.32, "learning_rate": 4.494738855677181e-05, "loss": 1.0462, "step": 12798 }, { "epoch": 1.32, "learning_rate": 4.494525436765087e-05, "loss": 0.3702, "step": 12799 }, { "epoch": 1.32, "learning_rate": 4.4943120077921964e-05, "loss": 0.1665, "step": 12800 }, { "epoch": 1.32, "learning_rate": 4.494098568759947e-05, "loss": 0.3207, "step": 12801 }, { "epoch": 1.32, "learning_rate": 4.493885119669777e-05, "loss": 0.2322, "step": 12802 }, { "epoch": 1.32, "learning_rate": 4.493671660523121e-05, "loss": 0.1956, "step": 12803 }, { "epoch": 1.32, "learning_rate": 4.493458191321416e-05, "loss": 0.6765, "step": 12804 }, { "epoch": 1.32, "learning_rate": 4.493244712066101e-05, "loss": 0.7841, "step": 12805 }, { "epoch": 1.32, "learning_rate": 4.493031222758612e-05, "loss": 1.3997, "step": 12806 }, { "epoch": 1.32, "learning_rate": 4.4928177234003855e-05, "loss": 0.5667, "step": 12807 }, { "epoch": 1.32, "learning_rate": 4.49260421399286e-05, "loss": 0.4152, "step": 12808 }, { "epoch": 1.32, "learning_rate": 4.492390694537472e-05, "loss": 0.4447, "step": 12809 }, { "epoch": 1.32, "learning_rate": 4.492177165035659e-05, "loss": 1.0175, "step": 12810 }, { "epoch": 1.32, "learning_rate": 4.491963625488858e-05, "loss": 0.0536, "step": 12811 }, { "epoch": 1.32, "learning_rate": 4.491750075898507e-05, "loss": 0.5468, "step": 12812 }, { "epoch": 1.32, "learning_rate": 4.4915365162660444e-05, "loss": 0.659, "step": 12813 }, { "epoch": 1.32, "learning_rate": 4.491322946592906e-05, "loss": 0.8736, "step": 12814 }, { "epoch": 1.32, "learning_rate": 4.4911093668805315e-05, "loss": 0.9174, "step": 12815 }, { "epoch": 1.32, "learning_rate": 4.490895777130357e-05, "loss": 0.2485, "step": 12816 }, { "epoch": 1.32, "learning_rate": 4.490682177343821e-05, "loss": 0.9664, "step": 12817 }, { "epoch": 1.32, "learning_rate": 4.490468567522361e-05, "loss": 0.3592, "step": 12818 }, { "epoch": 1.32, "learning_rate": 4.490254947667416e-05, "loss": 0.8118, "step": 12819 }, { "epoch": 1.32, "learning_rate": 4.490041317780423e-05, "loss": 0.4199, "step": 12820 }, { "epoch": 1.32, "learning_rate": 4.489827677862821e-05, "loss": 0.4227, "step": 12821 }, { "epoch": 1.32, "learning_rate": 4.489614027916047e-05, "loss": 1.0222, "step": 12822 }, { "epoch": 1.32, "learning_rate": 4.48940036794154e-05, "loss": 0.4543, "step": 12823 }, { "epoch": 1.32, "learning_rate": 4.489186697940739e-05, "loss": 0.8812, "step": 12824 }, { "epoch": 1.32, "learning_rate": 4.48897301791508e-05, "loss": 1.0329, "step": 12825 }, { "epoch": 1.32, "learning_rate": 4.4887593278660044e-05, "loss": 1.0112, "step": 12826 }, { "epoch": 1.32, "learning_rate": 4.48854562779495e-05, "loss": 0.6434, "step": 12827 }, { "epoch": 1.32, "learning_rate": 4.488331917703353e-05, "loss": 0.5576, "step": 12828 }, { "epoch": 1.32, "learning_rate": 4.488118197592655e-05, "loss": 0.7785, "step": 12829 }, { "epoch": 1.32, "learning_rate": 4.487904467464293e-05, "loss": 0.4547, "step": 12830 }, { "epoch": 1.32, "learning_rate": 4.487690727319705e-05, "loss": 0.5541, "step": 12831 }, { "epoch": 1.32, "learning_rate": 4.487476977160333e-05, "loss": 0.9132, "step": 12832 }, { "epoch": 1.32, "learning_rate": 4.487263216987613e-05, "loss": 0.7284, "step": 12833 }, { "epoch": 1.32, "learning_rate": 4.4870494468029865e-05, "loss": 0.7434, "step": 12834 }, { "epoch": 1.33, "learning_rate": 4.486835666607889e-05, "loss": 0.9095, "step": 12835 }, { "epoch": 1.33, "learning_rate": 4.486621876403763e-05, "loss": 0.2959, "step": 12836 }, { "epoch": 1.33, "learning_rate": 4.486408076192046e-05, "loss": 1.0946, "step": 12837 }, { "epoch": 1.33, "learning_rate": 4.486194265974177e-05, "loss": 1.2359, "step": 12838 }, { "epoch": 1.33, "learning_rate": 4.4859804457515974e-05, "loss": 0.2692, "step": 12839 }, { "epoch": 1.33, "learning_rate": 4.485766615525745e-05, "loss": 0.2639, "step": 12840 }, { "epoch": 1.33, "learning_rate": 4.485552775298058e-05, "loss": 0.3597, "step": 12841 }, { "epoch": 1.33, "learning_rate": 4.485338925069978e-05, "loss": 1.0428, "step": 12842 }, { "epoch": 1.33, "learning_rate": 4.4851250648429436e-05, "loss": 0.7503, "step": 12843 }, { "epoch": 1.33, "learning_rate": 4.484911194618395e-05, "loss": 0.3006, "step": 12844 }, { "epoch": 1.33, "learning_rate": 4.484697314397772e-05, "loss": 0.7802, "step": 12845 }, { "epoch": 1.33, "learning_rate": 4.484483424182514e-05, "loss": 1.0235, "step": 12846 }, { "epoch": 1.33, "learning_rate": 4.484269523974061e-05, "loss": 0.4444, "step": 12847 }, { "epoch": 1.33, "learning_rate": 4.484055613773852e-05, "loss": 0.5322, "step": 12848 }, { "epoch": 1.33, "learning_rate": 4.4838416935833284e-05, "loss": 0.9749, "step": 12849 }, { "epoch": 1.33, "learning_rate": 4.4836277634039296e-05, "loss": 0.8771, "step": 12850 }, { "epoch": 1.33, "learning_rate": 4.483413823237096e-05, "loss": 0.9426, "step": 12851 }, { "epoch": 1.33, "learning_rate": 4.483199873084268e-05, "loss": 0.6625, "step": 12852 }, { "epoch": 1.33, "learning_rate": 4.4829859129468854e-05, "loss": 0.5362, "step": 12853 }, { "epoch": 1.33, "learning_rate": 4.482771942826388e-05, "loss": 0.6242, "step": 12854 }, { "epoch": 1.33, "learning_rate": 4.482557962724217e-05, "loss": 0.9392, "step": 12855 }, { "epoch": 1.33, "learning_rate": 4.4823439726418134e-05, "loss": 0.3407, "step": 12856 }, { "epoch": 1.33, "learning_rate": 4.482129972580616e-05, "loss": 0.5909, "step": 12857 }, { "epoch": 1.33, "learning_rate": 4.481915962542067e-05, "loss": 0.8092, "step": 12858 }, { "epoch": 1.33, "learning_rate": 4.4817019425276066e-05, "loss": 1.2817, "step": 12859 }, { "epoch": 1.33, "learning_rate": 4.481487912538675e-05, "loss": 0.8179, "step": 12860 }, { "epoch": 1.33, "learning_rate": 4.481273872576714e-05, "loss": 0.3426, "step": 12861 }, { "epoch": 1.33, "learning_rate": 4.4810598226431635e-05, "loss": 1.002, "step": 12862 }, { "epoch": 1.33, "learning_rate": 4.480845762739464e-05, "loss": 0.5017, "step": 12863 }, { "epoch": 1.33, "learning_rate": 4.480631692867058e-05, "loss": 0.8287, "step": 12864 }, { "epoch": 1.33, "learning_rate": 4.4804176130273865e-05, "loss": 0.5596, "step": 12865 }, { "epoch": 1.33, "learning_rate": 4.48020352322189e-05, "loss": 0.1672, "step": 12866 }, { "epoch": 1.33, "learning_rate": 4.479989423452009e-05, "loss": 0.2876, "step": 12867 }, { "epoch": 1.33, "learning_rate": 4.479775313719186e-05, "loss": 0.2893, "step": 12868 }, { "epoch": 1.33, "learning_rate": 4.479561194024861e-05, "loss": 0.8959, "step": 12869 }, { "epoch": 1.33, "learning_rate": 4.479347064370477e-05, "loss": 1.0915, "step": 12870 }, { "epoch": 1.33, "learning_rate": 4.4791329247574746e-05, "loss": 1.0854, "step": 12871 }, { "epoch": 1.33, "learning_rate": 4.4789187751872955e-05, "loss": 0.2759, "step": 12872 }, { "epoch": 1.33, "learning_rate": 4.47870461566138e-05, "loss": 0.9445, "step": 12873 }, { "epoch": 1.33, "learning_rate": 4.4784904461811724e-05, "loss": 1.0459, "step": 12874 }, { "epoch": 1.33, "learning_rate": 4.4782762667481136e-05, "loss": 0.4446, "step": 12875 }, { "epoch": 1.33, "learning_rate": 4.4780620773636435e-05, "loss": 0.2729, "step": 12876 }, { "epoch": 1.33, "learning_rate": 4.477847878029206e-05, "loss": 0.2336, "step": 12877 }, { "epoch": 1.33, "learning_rate": 4.477633668746242e-05, "loss": 0.2377, "step": 12878 }, { "epoch": 1.33, "learning_rate": 4.477419449516194e-05, "loss": 0.4622, "step": 12879 }, { "epoch": 1.33, "learning_rate": 4.477205220340504e-05, "loss": 0.3778, "step": 12880 }, { "epoch": 1.33, "learning_rate": 4.476990981220614e-05, "loss": 0.8407, "step": 12881 }, { "epoch": 1.33, "learning_rate": 4.476776732157967e-05, "loss": 0.3624, "step": 12882 }, { "epoch": 1.33, "learning_rate": 4.4765624731540035e-05, "loss": 0.5493, "step": 12883 }, { "epoch": 1.33, "learning_rate": 4.476348204210168e-05, "loss": 0.4062, "step": 12884 }, { "epoch": 1.33, "learning_rate": 4.476133925327901e-05, "loss": 0.8575, "step": 12885 }, { "epoch": 1.33, "learning_rate": 4.475919636508646e-05, "loss": 1.3613, "step": 12886 }, { "epoch": 1.33, "learning_rate": 4.475705337753845e-05, "loss": 0.3285, "step": 12887 }, { "epoch": 1.33, "learning_rate": 4.4754910290649416e-05, "loss": 1.0158, "step": 12888 }, { "epoch": 1.33, "learning_rate": 4.4752767104433775e-05, "loss": 1.1598, "step": 12889 }, { "epoch": 1.33, "learning_rate": 4.4750623818905966e-05, "loss": 1.0859, "step": 12890 }, { "epoch": 1.33, "learning_rate": 4.47484804340804e-05, "loss": 0.6551, "step": 12891 }, { "epoch": 1.33, "learning_rate": 4.4746336949971514e-05, "loss": 0.8285, "step": 12892 }, { "epoch": 1.33, "learning_rate": 4.474419336659374e-05, "loss": 1.1672, "step": 12893 }, { "epoch": 1.33, "learning_rate": 4.474204968396151e-05, "loss": 0.2549, "step": 12894 }, { "epoch": 1.33, "learning_rate": 4.473990590208924e-05, "loss": 1.2649, "step": 12895 }, { "epoch": 1.33, "learning_rate": 4.473776202099139e-05, "loss": 0.4923, "step": 12896 }, { "epoch": 1.33, "learning_rate": 4.473561804068236e-05, "loss": 1.175, "step": 12897 }, { "epoch": 1.33, "learning_rate": 4.4733473961176594e-05, "loss": 1.1496, "step": 12898 }, { "epoch": 1.33, "learning_rate": 4.4731329782488545e-05, "loss": 0.8657, "step": 12899 }, { "epoch": 1.33, "learning_rate": 4.4729185504632615e-05, "loss": 0.9186, "step": 12900 }, { "epoch": 1.33, "learning_rate": 4.472704112762326e-05, "loss": 0.1217, "step": 12901 }, { "epoch": 1.33, "learning_rate": 4.4724896651474905e-05, "loss": 0.4969, "step": 12902 }, { "epoch": 1.33, "learning_rate": 4.4722752076202e-05, "loss": 0.7518, "step": 12903 }, { "epoch": 1.33, "learning_rate": 4.472060740181898e-05, "loss": 0.5407, "step": 12904 }, { "epoch": 1.33, "learning_rate": 4.471846262834026e-05, "loss": 0.4961, "step": 12905 }, { "epoch": 1.33, "learning_rate": 4.47163177557803e-05, "loss": 0.2994, "step": 12906 }, { "epoch": 1.33, "learning_rate": 4.471417278415353e-05, "loss": 0.5543, "step": 12907 }, { "epoch": 1.33, "learning_rate": 4.4712027713474393e-05, "loss": 0.7792, "step": 12908 }, { "epoch": 1.33, "learning_rate": 4.470988254375733e-05, "loss": 0.6331, "step": 12909 }, { "epoch": 1.33, "learning_rate": 4.470773727501678e-05, "loss": 0.8577, "step": 12910 }, { "epoch": 1.33, "learning_rate": 4.470559190726717e-05, "loss": 1.1157, "step": 12911 }, { "epoch": 1.33, "learning_rate": 4.470344644052297e-05, "loss": 1.3202, "step": 12912 }, { "epoch": 1.33, "learning_rate": 4.470130087479861e-05, "loss": 1.3771, "step": 12913 }, { "epoch": 1.33, "learning_rate": 4.4699155210108526e-05, "loss": 1.0034, "step": 12914 }, { "epoch": 1.33, "learning_rate": 4.469700944646718e-05, "loss": 0.3975, "step": 12915 }, { "epoch": 1.33, "learning_rate": 4.4694863583888986e-05, "loss": 0.4345, "step": 12916 }, { "epoch": 1.33, "learning_rate": 4.469271762238841e-05, "loss": 0.623, "step": 12917 }, { "epoch": 1.33, "learning_rate": 4.4690571561979915e-05, "loss": 0.9301, "step": 12918 }, { "epoch": 1.33, "learning_rate": 4.468842540267792e-05, "loss": 0.7134, "step": 12919 }, { "epoch": 1.33, "learning_rate": 4.468627914449688e-05, "loss": 0.0413, "step": 12920 }, { "epoch": 1.33, "learning_rate": 4.468413278745125e-05, "loss": 0.3015, "step": 12921 }, { "epoch": 1.33, "learning_rate": 4.4681986331555474e-05, "loss": 0.9784, "step": 12922 }, { "epoch": 1.33, "learning_rate": 4.4679839776824e-05, "loss": 0.457, "step": 12923 }, { "epoch": 1.33, "learning_rate": 4.4677693123271275e-05, "loss": 0.3792, "step": 12924 }, { "epoch": 1.33, "learning_rate": 4.467554637091176e-05, "loss": 0.4783, "step": 12925 }, { "epoch": 1.33, "learning_rate": 4.467339951975989e-05, "loss": 0.5642, "step": 12926 }, { "epoch": 1.33, "learning_rate": 4.4671252569830135e-05, "loss": 1.0042, "step": 12927 }, { "epoch": 1.33, "learning_rate": 4.466910552113695e-05, "loss": 0.8289, "step": 12928 }, { "epoch": 1.33, "learning_rate": 4.4666958373694764e-05, "loss": 0.2571, "step": 12929 }, { "epoch": 1.33, "learning_rate": 4.466481112751805e-05, "loss": 0.2255, "step": 12930 }, { "epoch": 1.34, "learning_rate": 4.466266378262126e-05, "loss": 0.6805, "step": 12931 }, { "epoch": 1.34, "learning_rate": 4.466051633901885e-05, "loss": 0.6111, "step": 12932 }, { "epoch": 1.34, "learning_rate": 4.4658368796725274e-05, "loss": 0.7368, "step": 12933 }, { "epoch": 1.34, "learning_rate": 4.4656221155754985e-05, "loss": 0.6479, "step": 12934 }, { "epoch": 1.34, "learning_rate": 4.465407341612245e-05, "loss": 0.754, "step": 12935 }, { "epoch": 1.34, "learning_rate": 4.4651925577842116e-05, "loss": 0.7692, "step": 12936 }, { "epoch": 1.34, "learning_rate": 4.4649777640928457e-05, "loss": 0.7689, "step": 12937 }, { "epoch": 1.34, "learning_rate": 4.464762960539591e-05, "loss": 0.7206, "step": 12938 }, { "epoch": 1.34, "learning_rate": 4.4645481471258957e-05, "loss": 0.6855, "step": 12939 }, { "epoch": 1.34, "learning_rate": 4.464333323853205e-05, "loss": 1.1036, "step": 12940 }, { "epoch": 1.34, "learning_rate": 4.464118490722964e-05, "loss": 0.7408, "step": 12941 }, { "epoch": 1.34, "learning_rate": 4.4639036477366216e-05, "loss": 0.9664, "step": 12942 }, { "epoch": 1.34, "learning_rate": 4.46368879489562e-05, "loss": 1.0897, "step": 12943 }, { "epoch": 1.34, "learning_rate": 4.46347393220141e-05, "loss": 0.4652, "step": 12944 }, { "epoch": 1.34, "learning_rate": 4.4632590596554344e-05, "loss": 1.104, "step": 12945 }, { "epoch": 1.34, "learning_rate": 4.463044177259142e-05, "loss": 1.091, "step": 12946 }, { "epoch": 1.34, "learning_rate": 4.4628292850139795e-05, "loss": 0.5614, "step": 12947 }, { "epoch": 1.34, "learning_rate": 4.462614382921391e-05, "loss": 0.5521, "step": 12948 }, { "epoch": 1.34, "learning_rate": 4.462399470982825e-05, "loss": 1.3667, "step": 12949 }, { "epoch": 1.34, "learning_rate": 4.4621845491997275e-05, "loss": 0.7382, "step": 12950 }, { "epoch": 1.34, "learning_rate": 4.461969617573547e-05, "loss": 0.2472, "step": 12951 }, { "epoch": 1.34, "learning_rate": 4.461754676105728e-05, "loss": 0.1365, "step": 12952 }, { "epoch": 1.34, "learning_rate": 4.4615397247977194e-05, "loss": 0.4023, "step": 12953 }, { "epoch": 1.34, "learning_rate": 4.461324763650967e-05, "loss": 0.7295, "step": 12954 }, { "epoch": 1.34, "learning_rate": 4.461109792666918e-05, "loss": 0.431, "step": 12955 }, { "epoch": 1.34, "learning_rate": 4.46089481184702e-05, "loss": 0.4762, "step": 12956 }, { "epoch": 1.34, "learning_rate": 4.46067982119272e-05, "loss": 0.218, "step": 12957 }, { "epoch": 1.34, "learning_rate": 4.4604648207054644e-05, "loss": 0.7359, "step": 12958 }, { "epoch": 1.34, "learning_rate": 4.460249810386702e-05, "loss": 0.9035, "step": 12959 }, { "epoch": 1.34, "learning_rate": 4.46003479023788e-05, "loss": 0.8716, "step": 12960 }, { "epoch": 1.34, "learning_rate": 4.4598197602604454e-05, "loss": 0.1991, "step": 12961 }, { "epoch": 1.34, "learning_rate": 4.459604720455845e-05, "loss": 0.6019, "step": 12962 }, { "epoch": 1.34, "learning_rate": 4.459389670825527e-05, "loss": 0.3601, "step": 12963 }, { "epoch": 1.34, "learning_rate": 4.45917461137094e-05, "loss": 0.743, "step": 12964 }, { "epoch": 1.34, "learning_rate": 4.4589595420935304e-05, "loss": 0.4022, "step": 12965 }, { "epoch": 1.34, "learning_rate": 4.4587444629947476e-05, "loss": 0.7314, "step": 12966 }, { "epoch": 1.34, "learning_rate": 4.458529374076037e-05, "loss": 1.1173, "step": 12967 }, { "epoch": 1.34, "learning_rate": 4.4583142753388486e-05, "loss": 0.3209, "step": 12968 }, { "epoch": 1.34, "learning_rate": 4.4580991667846286e-05, "loss": 0.9416, "step": 12969 }, { "epoch": 1.34, "learning_rate": 4.4578840484148275e-05, "loss": 1.0735, "step": 12970 }, { "epoch": 1.34, "learning_rate": 4.457668920230892e-05, "loss": 1.0798, "step": 12971 }, { "epoch": 1.34, "learning_rate": 4.45745378223427e-05, "loss": 0.7187, "step": 12972 }, { "epoch": 1.34, "learning_rate": 4.45723863442641e-05, "loss": 0.3379, "step": 12973 }, { "epoch": 1.34, "learning_rate": 4.457023476808761e-05, "loss": 0.7364, "step": 12974 }, { "epoch": 1.34, "learning_rate": 4.456808309382771e-05, "loss": 0.6073, "step": 12975 }, { "epoch": 1.34, "learning_rate": 4.4565931321498875e-05, "loss": 0.2309, "step": 12976 }, { "epoch": 1.34, "learning_rate": 4.45637794511156e-05, "loss": 0.5354, "step": 12977 }, { "epoch": 1.34, "learning_rate": 4.4561627482692375e-05, "loss": 0.7631, "step": 12978 }, { "epoch": 1.34, "learning_rate": 4.4559475416243684e-05, "loss": 0.9286, "step": 12979 }, { "epoch": 1.34, "learning_rate": 4.4557323251784004e-05, "loss": 0.1934, "step": 12980 }, { "epoch": 1.34, "learning_rate": 4.4555170989327836e-05, "loss": 0.2511, "step": 12981 }, { "epoch": 1.34, "learning_rate": 4.455301862888965e-05, "loss": 1.1053, "step": 12982 }, { "epoch": 1.34, "learning_rate": 4.4550866170483955e-05, "loss": 0.5905, "step": 12983 }, { "epoch": 1.34, "learning_rate": 4.454871361412524e-05, "loss": 0.8984, "step": 12984 }, { "epoch": 1.34, "learning_rate": 4.454656095982798e-05, "loss": 0.4852, "step": 12985 }, { "epoch": 1.34, "learning_rate": 4.454440820760668e-05, "loss": 0.8023, "step": 12986 }, { "epoch": 1.34, "learning_rate": 4.454225535747582e-05, "loss": 0.5834, "step": 12987 }, { "epoch": 1.34, "learning_rate": 4.454010240944991e-05, "loss": 0.7645, "step": 12988 }, { "epoch": 1.34, "learning_rate": 4.453794936354343e-05, "loss": 0.8613, "step": 12989 }, { "epoch": 1.34, "learning_rate": 4.453579621977087e-05, "loss": 0.4913, "step": 12990 }, { "epoch": 1.34, "learning_rate": 4.453364297814675e-05, "loss": 1.0638, "step": 12991 }, { "epoch": 1.34, "learning_rate": 4.4531489638685524e-05, "loss": 0.41, "step": 12992 }, { "epoch": 1.34, "learning_rate": 4.452933620140171e-05, "loss": 0.6808, "step": 12993 }, { "epoch": 1.34, "learning_rate": 4.4527182666309815e-05, "loss": 0.5704, "step": 12994 }, { "epoch": 1.34, "learning_rate": 4.4525029033424324e-05, "loss": 1.3841, "step": 12995 }, { "epoch": 1.34, "learning_rate": 4.452287530275973e-05, "loss": 0.4426, "step": 12996 }, { "epoch": 1.34, "learning_rate": 4.452072147433055e-05, "loss": 0.6164, "step": 12997 }, { "epoch": 1.34, "learning_rate": 4.4518567548151254e-05, "loss": 1.3438, "step": 12998 }, { "epoch": 1.34, "learning_rate": 4.451641352423638e-05, "loss": 0.6608, "step": 12999 }, { "epoch": 1.34, "learning_rate": 4.451425940260038e-05, "loss": 0.9908, "step": 13000 }, { "epoch": 1.34, "learning_rate": 4.45121051832578e-05, "loss": 0.5161, "step": 13001 }, { "epoch": 1.34, "learning_rate": 4.450995086622312e-05, "loss": 0.0016, "step": 13002 }, { "epoch": 1.34, "learning_rate": 4.450779645151085e-05, "loss": 0.8444, "step": 13003 }, { "epoch": 1.34, "learning_rate": 4.4505641939135476e-05, "loss": 0.9747, "step": 13004 }, { "epoch": 1.34, "learning_rate": 4.4503487329111524e-05, "loss": 0.8776, "step": 13005 }, { "epoch": 1.34, "learning_rate": 4.4501332621453485e-05, "loss": 1.0987, "step": 13006 }, { "epoch": 1.34, "learning_rate": 4.449917781617587e-05, "loss": 0.4968, "step": 13007 }, { "epoch": 1.34, "learning_rate": 4.449702291329318e-05, "loss": 0.4574, "step": 13008 }, { "epoch": 1.34, "learning_rate": 4.449486791281993e-05, "loss": 0.5792, "step": 13009 }, { "epoch": 1.34, "learning_rate": 4.4492712814770616e-05, "loss": 0.2316, "step": 13010 }, { "epoch": 1.34, "learning_rate": 4.4490557619159746e-05, "loss": 0.4252, "step": 13011 }, { "epoch": 1.34, "learning_rate": 4.448840232600184e-05, "loss": 0.9135, "step": 13012 }, { "epoch": 1.34, "learning_rate": 4.44862469353114e-05, "loss": 0.7343, "step": 13013 }, { "epoch": 1.34, "learning_rate": 4.448409144710293e-05, "loss": 0.8924, "step": 13014 }, { "epoch": 1.34, "learning_rate": 4.448193586139094e-05, "loss": 0.5417, "step": 13015 }, { "epoch": 1.34, "learning_rate": 4.4479780178189954e-05, "loss": 0.3495, "step": 13016 }, { "epoch": 1.34, "learning_rate": 4.447762439751448e-05, "loss": 0.3695, "step": 13017 }, { "epoch": 1.34, "learning_rate": 4.4475468519379025e-05, "loss": 0.2586, "step": 13018 }, { "epoch": 1.34, "learning_rate": 4.447331254379809e-05, "loss": 0.6457, "step": 13019 }, { "epoch": 1.34, "learning_rate": 4.447115647078622e-05, "loss": 0.6202, "step": 13020 }, { "epoch": 1.34, "learning_rate": 4.44690003003579e-05, "loss": 1.3491, "step": 13021 }, { "epoch": 1.34, "learning_rate": 4.446684403252766e-05, "loss": 0.3068, "step": 13022 }, { "epoch": 1.34, "learning_rate": 4.446468766731001e-05, "loss": 0.2971, "step": 13023 }, { "epoch": 1.34, "learning_rate": 4.4462531204719455e-05, "loss": 0.8329, "step": 13024 }, { "epoch": 1.34, "learning_rate": 4.446037464477054e-05, "loss": 0.9254, "step": 13025 }, { "epoch": 1.34, "learning_rate": 4.445821798747775e-05, "loss": 0.3619, "step": 13026 }, { "epoch": 1.35, "learning_rate": 4.445606123285564e-05, "loss": 0.7055, "step": 13027 }, { "epoch": 1.35, "learning_rate": 4.44539043809187e-05, "loss": 0.6544, "step": 13028 }, { "epoch": 1.35, "learning_rate": 4.445174743168145e-05, "loss": 0.6191, "step": 13029 }, { "epoch": 1.35, "learning_rate": 4.444959038515843e-05, "loss": 0.3872, "step": 13030 }, { "epoch": 1.35, "learning_rate": 4.444743324136414e-05, "loss": 1.0566, "step": 13031 }, { "epoch": 1.35, "learning_rate": 4.444527600031311e-05, "loss": 1.2136, "step": 13032 }, { "epoch": 1.35, "learning_rate": 4.444311866201987e-05, "loss": 0.9334, "step": 13033 }, { "epoch": 1.35, "learning_rate": 4.444096122649892e-05, "loss": 0.8352, "step": 13034 }, { "epoch": 1.35, "learning_rate": 4.4438803693764816e-05, "loss": 1.1062, "step": 13035 }, { "epoch": 1.35, "learning_rate": 4.443664606383205e-05, "loss": 0.2682, "step": 13036 }, { "epoch": 1.35, "learning_rate": 4.443448833671517e-05, "loss": 1.1119, "step": 13037 }, { "epoch": 1.35, "learning_rate": 4.443233051242868e-05, "loss": 0.351, "step": 13038 }, { "epoch": 1.35, "learning_rate": 4.443017259098713e-05, "loss": 0.4634, "step": 13039 }, { "epoch": 1.35, "learning_rate": 4.442801457240502e-05, "loss": 0.3885, "step": 13040 }, { "epoch": 1.35, "learning_rate": 4.4425856456696906e-05, "loss": 0.6162, "step": 13041 }, { "epoch": 1.35, "learning_rate": 4.44236982438773e-05, "loss": 0.2185, "step": 13042 }, { "epoch": 1.35, "learning_rate": 4.442153993396073e-05, "loss": 0.4356, "step": 13043 }, { "epoch": 1.35, "learning_rate": 4.441938152696172e-05, "loss": 0.8786, "step": 13044 }, { "epoch": 1.35, "learning_rate": 4.441722302289481e-05, "loss": 0.3792, "step": 13045 }, { "epoch": 1.35, "learning_rate": 4.441506442177454e-05, "loss": 0.2623, "step": 13046 }, { "epoch": 1.35, "learning_rate": 4.4412905723615425e-05, "loss": 0.9521, "step": 13047 }, { "epoch": 1.35, "learning_rate": 4.441074692843199e-05, "loss": 1.0239, "step": 13048 }, { "epoch": 1.35, "learning_rate": 4.440858803623879e-05, "loss": 0.8982, "step": 13049 }, { "epoch": 1.35, "learning_rate": 4.440642904705034e-05, "loss": 0.6045, "step": 13050 }, { "epoch": 1.35, "learning_rate": 4.440426996088118e-05, "loss": 0.8101, "step": 13051 }, { "epoch": 1.35, "learning_rate": 4.440211077774585e-05, "loss": 0.2787, "step": 13052 }, { "epoch": 1.35, "learning_rate": 4.439995149765888e-05, "loss": 1.0692, "step": 13053 }, { "epoch": 1.35, "learning_rate": 4.43977921206348e-05, "loss": 0.4228, "step": 13054 }, { "epoch": 1.35, "learning_rate": 4.4395632646688154e-05, "loss": 0.5842, "step": 13055 }, { "epoch": 1.35, "learning_rate": 4.4393473075833485e-05, "loss": 0.1495, "step": 13056 }, { "epoch": 1.35, "learning_rate": 4.4391313408085314e-05, "loss": 1.0292, "step": 13057 }, { "epoch": 1.35, "learning_rate": 4.438915364345819e-05, "loss": 0.408, "step": 13058 }, { "epoch": 1.35, "learning_rate": 4.4386993781966656e-05, "loss": 0.9638, "step": 13059 }, { "epoch": 1.35, "learning_rate": 4.438483382362524e-05, "loss": 0.5877, "step": 13060 }, { "epoch": 1.35, "learning_rate": 4.4382673768448494e-05, "loss": 0.3189, "step": 13061 }, { "epoch": 1.35, "learning_rate": 4.438051361645094e-05, "loss": 0.1326, "step": 13062 }, { "epoch": 1.35, "learning_rate": 4.437835336764715e-05, "loss": 0.4583, "step": 13063 }, { "epoch": 1.35, "learning_rate": 4.437619302205164e-05, "loss": 0.3372, "step": 13064 }, { "epoch": 1.35, "learning_rate": 4.437403257967897e-05, "loss": 0.5964, "step": 13065 }, { "epoch": 1.35, "learning_rate": 4.437187204054367e-05, "loss": 0.9435, "step": 13066 }, { "epoch": 1.35, "learning_rate": 4.4369711404660294e-05, "loss": 0.2119, "step": 13067 }, { "epoch": 1.35, "learning_rate": 4.4367550672043376e-05, "loss": 0.4698, "step": 13068 }, { "epoch": 1.35, "learning_rate": 4.436538984270747e-05, "loss": 0.3891, "step": 13069 }, { "epoch": 1.35, "learning_rate": 4.4363228916667125e-05, "loss": 2.2358, "step": 13070 }, { "epoch": 1.35, "learning_rate": 4.436106789393688e-05, "loss": 0.5825, "step": 13071 }, { "epoch": 1.35, "learning_rate": 4.4358906774531294e-05, "loss": 0.7455, "step": 13072 }, { "epoch": 1.35, "learning_rate": 4.435674555846491e-05, "loss": 0.2917, "step": 13073 }, { "epoch": 1.35, "learning_rate": 4.435458424575226e-05, "loss": 0.3993, "step": 13074 }, { "epoch": 1.35, "learning_rate": 4.4352422836407914e-05, "loss": 0.8087, "step": 13075 }, { "epoch": 1.35, "learning_rate": 4.4350261330446406e-05, "loss": 0.4565, "step": 13076 }, { "epoch": 1.35, "learning_rate": 4.4348099727882314e-05, "loss": 0.3814, "step": 13077 }, { "epoch": 1.35, "learning_rate": 4.434593802873017e-05, "loss": 0.742, "step": 13078 }, { "epoch": 1.35, "learning_rate": 4.4343776233004515e-05, "loss": 1.0549, "step": 13079 }, { "epoch": 1.35, "learning_rate": 4.4341614340719925e-05, "loss": 0.2413, "step": 13080 }, { "epoch": 1.35, "learning_rate": 4.4339452351890934e-05, "loss": 0.5842, "step": 13081 }, { "epoch": 1.35, "learning_rate": 4.43372902665321e-05, "loss": 0.64, "step": 13082 }, { "epoch": 1.35, "learning_rate": 4.4335128084657994e-05, "loss": 0.4154, "step": 13083 }, { "epoch": 1.35, "learning_rate": 4.433296580628316e-05, "loss": 0.2671, "step": 13084 }, { "epoch": 1.35, "learning_rate": 4.433080343142216e-05, "loss": 0.1583, "step": 13085 }, { "epoch": 1.35, "learning_rate": 4.4328640960089535e-05, "loss": 0.2473, "step": 13086 }, { "epoch": 1.35, "learning_rate": 4.432647839229985e-05, "loss": 1.0906, "step": 13087 }, { "epoch": 1.35, "learning_rate": 4.4324315728067664e-05, "loss": 0.3582, "step": 13088 }, { "epoch": 1.35, "learning_rate": 4.4322152967407546e-05, "loss": 0.3148, "step": 13089 }, { "epoch": 1.35, "learning_rate": 4.431999011033404e-05, "loss": 0.9597, "step": 13090 }, { "epoch": 1.35, "learning_rate": 4.431782715686171e-05, "loss": 0.7524, "step": 13091 }, { "epoch": 1.35, "learning_rate": 4.431566410700512e-05, "loss": 0.5673, "step": 13092 }, { "epoch": 1.35, "learning_rate": 4.431350096077882e-05, "loss": 0.6159, "step": 13093 }, { "epoch": 1.35, "learning_rate": 4.431133771819739e-05, "loss": 0.8121, "step": 13094 }, { "epoch": 1.35, "learning_rate": 4.430917437927538e-05, "loss": 0.6943, "step": 13095 }, { "epoch": 1.35, "learning_rate": 4.4307010944027356e-05, "loss": 0.8109, "step": 13096 }, { "epoch": 1.35, "learning_rate": 4.430484741246789e-05, "loss": 0.4752, "step": 13097 }, { "epoch": 1.35, "learning_rate": 4.4302683784611524e-05, "loss": 0.2899, "step": 13098 }, { "epoch": 1.35, "learning_rate": 4.430052006047285e-05, "loss": 0.8015, "step": 13099 }, { "epoch": 1.35, "learning_rate": 4.429835624006641e-05, "loss": 0.6455, "step": 13100 }, { "epoch": 1.35, "learning_rate": 4.4296192323406784e-05, "loss": 1.3233, "step": 13101 }, { "epoch": 1.35, "learning_rate": 4.429402831050854e-05, "loss": 0.2765, "step": 13102 }, { "epoch": 1.35, "learning_rate": 4.429186420138625e-05, "loss": 1.2138, "step": 13103 }, { "epoch": 1.35, "learning_rate": 4.4289699996054463e-05, "loss": 0.8515, "step": 13104 }, { "epoch": 1.35, "learning_rate": 4.4287535694527757e-05, "loss": 1.3382, "step": 13105 }, { "epoch": 1.35, "learning_rate": 4.428537129682071e-05, "loss": 0.8628, "step": 13106 }, { "epoch": 1.35, "learning_rate": 4.428320680294788e-05, "loss": 2.2922, "step": 13107 }, { "epoch": 1.35, "learning_rate": 4.428104221292385e-05, "loss": 1.3294, "step": 13108 }, { "epoch": 1.35, "learning_rate": 4.427887752676318e-05, "loss": 0.6791, "step": 13109 }, { "epoch": 1.35, "learning_rate": 4.4276712744480454e-05, "loss": 0.3616, "step": 13110 }, { "epoch": 1.35, "learning_rate": 4.427454786609023e-05, "loss": 0.555, "step": 13111 }, { "epoch": 1.35, "learning_rate": 4.42723828916071e-05, "loss": 0.4579, "step": 13112 }, { "epoch": 1.35, "learning_rate": 4.427021782104562e-05, "loss": 0.6271, "step": 13113 }, { "epoch": 1.35, "learning_rate": 4.426805265442037e-05, "loss": 0.666, "step": 13114 }, { "epoch": 1.35, "learning_rate": 4.426588739174593e-05, "loss": 0.3551, "step": 13115 }, { "epoch": 1.35, "learning_rate": 4.426372203303689e-05, "loss": 0.4229, "step": 13116 }, { "epoch": 1.35, "learning_rate": 4.426155657830779e-05, "loss": 0.7068, "step": 13117 }, { "epoch": 1.35, "learning_rate": 4.425939102757325e-05, "loss": 0.7721, "step": 13118 }, { "epoch": 1.35, "learning_rate": 4.42572253808478e-05, "loss": 0.5209, "step": 13119 }, { "epoch": 1.35, "learning_rate": 4.425505963814606e-05, "loss": 0.3617, "step": 13120 }, { "epoch": 1.35, "learning_rate": 4.425289379948258e-05, "loss": 1.0593, "step": 13121 }, { "epoch": 1.35, "learning_rate": 4.425072786487197e-05, "loss": 0.3051, "step": 13122 }, { "epoch": 1.36, "learning_rate": 4.4248561834328797e-05, "loss": 0.8471, "step": 13123 }, { "epoch": 1.36, "learning_rate": 4.4246395707867626e-05, "loss": 0.3392, "step": 13124 }, { "epoch": 1.36, "learning_rate": 4.424422948550306e-05, "loss": 0.341, "step": 13125 }, { "epoch": 1.36, "learning_rate": 4.424206316724967e-05, "loss": 0.8573, "step": 13126 }, { "epoch": 1.36, "learning_rate": 4.4239896753122044e-05, "loss": 0.7186, "step": 13127 }, { "epoch": 1.36, "learning_rate": 4.4237730243134763e-05, "loss": 0.4314, "step": 13128 }, { "epoch": 1.36, "learning_rate": 4.4235563637302425e-05, "loss": 0.6899, "step": 13129 }, { "epoch": 1.36, "learning_rate": 4.423339693563959e-05, "loss": 0.7409, "step": 13130 }, { "epoch": 1.36, "learning_rate": 4.423123013816086e-05, "loss": 0.1855, "step": 13131 }, { "epoch": 1.36, "learning_rate": 4.422906324488083e-05, "loss": 0.5586, "step": 13132 }, { "epoch": 1.36, "learning_rate": 4.422689625581406e-05, "loss": 0.5952, "step": 13133 }, { "epoch": 1.36, "learning_rate": 4.4224729170975155e-05, "loss": 0.3758, "step": 13134 }, { "epoch": 1.36, "learning_rate": 4.422256199037871e-05, "loss": 1.4305, "step": 13135 }, { "epoch": 1.36, "learning_rate": 4.42203947140393e-05, "loss": 0.7502, "step": 13136 }, { "epoch": 1.36, "learning_rate": 4.4218227341971524e-05, "loss": 0.7406, "step": 13137 }, { "epoch": 1.36, "learning_rate": 4.421605987418996e-05, "loss": 0.6608, "step": 13138 }, { "epoch": 1.36, "learning_rate": 4.4213892310709214e-05, "loss": 1.0192, "step": 13139 }, { "epoch": 1.36, "learning_rate": 4.421172465154386e-05, "loss": 0.5144, "step": 13140 }, { "epoch": 1.36, "learning_rate": 4.420955689670852e-05, "loss": 0.5366, "step": 13141 }, { "epoch": 1.36, "learning_rate": 4.4207389046217766e-05, "loss": 0.8327, "step": 13142 }, { "epoch": 1.36, "learning_rate": 4.4205221100086185e-05, "loss": 0.5117, "step": 13143 }, { "epoch": 1.36, "learning_rate": 4.420305305832837e-05, "loss": 0.3143, "step": 13144 }, { "epoch": 1.36, "learning_rate": 4.4200884920958935e-05, "loss": 0.9277, "step": 13145 }, { "epoch": 1.36, "learning_rate": 4.419871668799247e-05, "loss": 0.9152, "step": 13146 }, { "epoch": 1.36, "learning_rate": 4.4196548359443555e-05, "loss": 0.192, "step": 13147 }, { "epoch": 1.36, "learning_rate": 4.419437993532681e-05, "loss": 0.1525, "step": 13148 }, { "epoch": 1.36, "learning_rate": 4.419221141565682e-05, "loss": 0.2974, "step": 13149 }, { "epoch": 1.36, "learning_rate": 4.419004280044817e-05, "loss": 1.0238, "step": 13150 }, { "epoch": 1.36, "learning_rate": 4.4187874089715487e-05, "loss": 0.5155, "step": 13151 }, { "epoch": 1.36, "learning_rate": 4.4185705283473345e-05, "loss": 0.618, "step": 13152 }, { "epoch": 1.36, "learning_rate": 4.4183536381736357e-05, "loss": 0.2899, "step": 13153 }, { "epoch": 1.36, "learning_rate": 4.418136738451913e-05, "loss": 0.1536, "step": 13154 }, { "epoch": 1.36, "learning_rate": 4.4179198291836244e-05, "loss": 0.0743, "step": 13155 }, { "epoch": 1.36, "learning_rate": 4.417702910370232e-05, "loss": 0.8676, "step": 13156 }, { "epoch": 1.36, "learning_rate": 4.4174859820131944e-05, "loss": 0.7479, "step": 13157 }, { "epoch": 1.36, "learning_rate": 4.4172690441139744e-05, "loss": 0.2644, "step": 13158 }, { "epoch": 1.36, "learning_rate": 4.417052096674029e-05, "loss": 0.2179, "step": 13159 }, { "epoch": 1.36, "learning_rate": 4.4168351396948216e-05, "loss": 0.2097, "step": 13160 }, { "epoch": 1.36, "learning_rate": 4.416618173177812e-05, "loss": 0.8219, "step": 13161 }, { "epoch": 1.36, "learning_rate": 4.416401197124459e-05, "loss": 1.1496, "step": 13162 }, { "epoch": 1.36, "learning_rate": 4.416184211536225e-05, "loss": 0.7916, "step": 13163 }, { "epoch": 1.36, "learning_rate": 4.41596721641457e-05, "loss": 0.59, "step": 13164 }, { "epoch": 1.36, "learning_rate": 4.415750211760956e-05, "loss": 0.638, "step": 13165 }, { "epoch": 1.36, "learning_rate": 4.415533197576843e-05, "loss": 0.3087, "step": 13166 }, { "epoch": 1.36, "learning_rate": 4.4153161738636904e-05, "loss": 0.6682, "step": 13167 }, { "epoch": 1.36, "learning_rate": 4.415099140622961e-05, "loss": 0.2133, "step": 13168 }, { "epoch": 1.36, "learning_rate": 4.4148820978561146e-05, "loss": 0.0879, "step": 13169 }, { "epoch": 1.36, "learning_rate": 4.4146650455646146e-05, "loss": 0.2845, "step": 13170 }, { "epoch": 1.36, "learning_rate": 4.414447983749919e-05, "loss": 0.3024, "step": 13171 }, { "epoch": 1.36, "learning_rate": 4.414230912413491e-05, "loss": 0.5456, "step": 13172 }, { "epoch": 1.36, "learning_rate": 4.414013831556792e-05, "loss": 0.7752, "step": 13173 }, { "epoch": 1.36, "learning_rate": 4.413796741181282e-05, "loss": 1.2548, "step": 13174 }, { "epoch": 1.36, "learning_rate": 4.413579641288424e-05, "loss": 1.0315, "step": 13175 }, { "epoch": 1.36, "learning_rate": 4.413362531879678e-05, "loss": 0.6035, "step": 13176 }, { "epoch": 1.36, "learning_rate": 4.4131454129565057e-05, "loss": 0.4882, "step": 13177 }, { "epoch": 1.36, "learning_rate": 4.41292828452037e-05, "loss": 0.4041, "step": 13178 }, { "epoch": 1.36, "learning_rate": 4.4127111465727313e-05, "loss": 0.2912, "step": 13179 }, { "epoch": 1.36, "learning_rate": 4.412493999115052e-05, "loss": 1.265, "step": 13180 }, { "epoch": 1.36, "learning_rate": 4.412276842148793e-05, "loss": 0.3414, "step": 13181 }, { "epoch": 1.36, "learning_rate": 4.4120596756754165e-05, "loss": 0.6433, "step": 13182 }, { "epoch": 1.36, "learning_rate": 4.411842499696385e-05, "loss": 0.9077, "step": 13183 }, { "epoch": 1.36, "learning_rate": 4.411625314213161e-05, "loss": 0.9165, "step": 13184 }, { "epoch": 1.36, "learning_rate": 4.4114081192272054e-05, "loss": 0.928, "step": 13185 }, { "epoch": 1.36, "learning_rate": 4.41119091473998e-05, "loss": 1.6111, "step": 13186 }, { "epoch": 1.36, "learning_rate": 4.410973700752947e-05, "loss": 1.0949, "step": 13187 }, { "epoch": 1.36, "learning_rate": 4.41075647726757e-05, "loss": 0.6029, "step": 13188 }, { "epoch": 1.36, "learning_rate": 4.410539244285311e-05, "loss": 0.4154, "step": 13189 }, { "epoch": 1.36, "learning_rate": 4.410322001807631e-05, "loss": 1.0569, "step": 13190 }, { "epoch": 1.36, "learning_rate": 4.410104749835993e-05, "loss": 0.3004, "step": 13191 }, { "epoch": 1.36, "learning_rate": 4.409887488371861e-05, "loss": 0.804, "step": 13192 }, { "epoch": 1.36, "learning_rate": 4.4096702174166955e-05, "loss": 1.3918, "step": 13193 }, { "epoch": 1.36, "learning_rate": 4.40945293697196e-05, "loss": 0.748, "step": 13194 }, { "epoch": 1.36, "learning_rate": 4.409235647039117e-05, "loss": 0.4924, "step": 13195 }, { "epoch": 1.36, "learning_rate": 4.40901834761963e-05, "loss": 0.314, "step": 13196 }, { "epoch": 1.36, "learning_rate": 4.4088010387149605e-05, "loss": 0.2362, "step": 13197 }, { "epoch": 1.36, "learning_rate": 4.408583720326572e-05, "loss": 0.1141, "step": 13198 }, { "epoch": 1.36, "learning_rate": 4.4083663924559276e-05, "loss": 0.3672, "step": 13199 }, { "epoch": 1.36, "learning_rate": 4.408149055104491e-05, "loss": 0.7521, "step": 13200 }, { "epoch": 1.36, "learning_rate": 4.4079317082737226e-05, "loss": 0.3805, "step": 13201 }, { "epoch": 1.36, "learning_rate": 4.407714351965089e-05, "loss": 0.9041, "step": 13202 }, { "epoch": 1.36, "learning_rate": 4.4074969861800515e-05, "loss": 0.4785, "step": 13203 }, { "epoch": 1.36, "learning_rate": 4.407279610920074e-05, "loss": 0.5039, "step": 13204 }, { "epoch": 1.36, "learning_rate": 4.407062226186618e-05, "loss": 0.62, "step": 13205 }, { "epoch": 1.36, "learning_rate": 4.406844831981149e-05, "loss": 0.2337, "step": 13206 }, { "epoch": 1.36, "learning_rate": 4.4066274283051297e-05, "loss": 0.509, "step": 13207 }, { "epoch": 1.36, "learning_rate": 4.4064100151600246e-05, "loss": 0.6473, "step": 13208 }, { "epoch": 1.36, "learning_rate": 4.406192592547295e-05, "loss": 0.8943, "step": 13209 }, { "epoch": 1.36, "learning_rate": 4.405975160468407e-05, "loss": 1.199, "step": 13210 }, { "epoch": 1.36, "learning_rate": 4.405757718924823e-05, "loss": 1.0163, "step": 13211 }, { "epoch": 1.36, "learning_rate": 4.405540267918006e-05, "loss": 1.2941, "step": 13212 }, { "epoch": 1.36, "learning_rate": 4.4053228074494214e-05, "loss": 1.0088, "step": 13213 }, { "epoch": 1.36, "learning_rate": 4.405105337520533e-05, "loss": 0.8582, "step": 13214 }, { "epoch": 1.36, "learning_rate": 4.404887858132803e-05, "loss": 0.5723, "step": 13215 }, { "epoch": 1.36, "learning_rate": 4.404670369287698e-05, "loss": 0.6012, "step": 13216 }, { "epoch": 1.36, "learning_rate": 4.40445287098668e-05, "loss": 0.2913, "step": 13217 }, { "epoch": 1.37, "learning_rate": 4.404235363231214e-05, "loss": 0.9777, "step": 13218 }, { "epoch": 1.37, "learning_rate": 4.404017846022764e-05, "loss": 1.2395, "step": 13219 }, { "epoch": 1.37, "learning_rate": 4.4038003193627946e-05, "loss": 1.489, "step": 13220 }, { "epoch": 1.37, "learning_rate": 4.4035827832527694e-05, "loss": 0.7469, "step": 13221 }, { "epoch": 1.37, "learning_rate": 4.403365237694154e-05, "loss": 0.7294, "step": 13222 }, { "epoch": 1.37, "learning_rate": 4.403147682688413e-05, "loss": 0.6991, "step": 13223 }, { "epoch": 1.37, "learning_rate": 4.402930118237009e-05, "loss": 0.5847, "step": 13224 }, { "epoch": 1.37, "learning_rate": 4.402712544341407e-05, "loss": 0.9002, "step": 13225 }, { "epoch": 1.37, "learning_rate": 4.402494961003073e-05, "loss": 0.8313, "step": 13226 }, { "epoch": 1.37, "learning_rate": 4.402277368223472e-05, "loss": 0.2895, "step": 13227 }, { "epoch": 1.37, "learning_rate": 4.402059766004067e-05, "loss": 0.9535, "step": 13228 }, { "epoch": 1.37, "learning_rate": 4.401842154346324e-05, "loss": 1.0092, "step": 13229 }, { "epoch": 1.37, "learning_rate": 4.4016245332517065e-05, "loss": 0.8357, "step": 13230 }, { "epoch": 1.37, "learning_rate": 4.401406902721681e-05, "loss": 0.2221, "step": 13231 }, { "epoch": 1.37, "learning_rate": 4.4011892627577135e-05, "loss": 0.4099, "step": 13232 }, { "epoch": 1.37, "learning_rate": 4.400971613361267e-05, "loss": 0.2077, "step": 13233 }, { "epoch": 1.37, "learning_rate": 4.400753954533807e-05, "loss": 0.6406, "step": 13234 }, { "epoch": 1.37, "learning_rate": 4.4005362862768e-05, "loss": 0.9438, "step": 13235 }, { "epoch": 1.37, "learning_rate": 4.4003186085917094e-05, "loss": 0.9756, "step": 13236 }, { "epoch": 1.37, "learning_rate": 4.400100921480002e-05, "loss": 1.0646, "step": 13237 }, { "epoch": 1.37, "learning_rate": 4.399883224943143e-05, "loss": 0.4481, "step": 13238 }, { "epoch": 1.37, "learning_rate": 4.3996655189825974e-05, "loss": 1.0927, "step": 13239 }, { "epoch": 1.37, "learning_rate": 4.3994478035998304e-05, "loss": 0.275, "step": 13240 }, { "epoch": 1.37, "learning_rate": 4.399230078796309e-05, "loss": 0.6508, "step": 13241 }, { "epoch": 1.37, "learning_rate": 4.399012344573499e-05, "loss": 0.5507, "step": 13242 }, { "epoch": 1.37, "learning_rate": 4.398794600932863e-05, "loss": 0.4384, "step": 13243 }, { "epoch": 1.37, "learning_rate": 4.398576847875871e-05, "loss": 0.2311, "step": 13244 }, { "epoch": 1.37, "learning_rate": 4.398359085403986e-05, "loss": 1.0606, "step": 13245 }, { "epoch": 1.37, "learning_rate": 4.398141313518675e-05, "loss": 0.9258, "step": 13246 }, { "epoch": 1.37, "learning_rate": 4.397923532221403e-05, "loss": 0.7694, "step": 13247 }, { "epoch": 1.37, "learning_rate": 4.397705741513639e-05, "loss": 0.3698, "step": 13248 }, { "epoch": 1.37, "learning_rate": 4.3974879413968455e-05, "loss": 1.137, "step": 13249 }, { "epoch": 1.37, "learning_rate": 4.39727013187249e-05, "loss": 0.4722, "step": 13250 }, { "epoch": 1.37, "learning_rate": 4.397052312942039e-05, "loss": 0.9131, "step": 13251 }, { "epoch": 1.37, "learning_rate": 4.396834484606959e-05, "loss": 0.7854, "step": 13252 }, { "epoch": 1.37, "learning_rate": 4.396616646868716e-05, "loss": 0.665, "step": 13253 }, { "epoch": 1.37, "learning_rate": 4.396398799728777e-05, "loss": 0.7192, "step": 13254 }, { "epoch": 1.37, "learning_rate": 4.3961809431886075e-05, "loss": 0.3093, "step": 13255 }, { "epoch": 1.37, "learning_rate": 4.3959630772496745e-05, "loss": 0.7627, "step": 13256 }, { "epoch": 1.37, "learning_rate": 4.395745201913444e-05, "loss": 0.3866, "step": 13257 }, { "epoch": 1.37, "learning_rate": 4.395527317181385e-05, "loss": 0.6945, "step": 13258 }, { "epoch": 1.37, "learning_rate": 4.395309423054961e-05, "loss": 0.7082, "step": 13259 }, { "epoch": 1.37, "learning_rate": 4.395091519535642e-05, "loss": 0.7523, "step": 13260 }, { "epoch": 1.37, "learning_rate": 4.3948736066248934e-05, "loss": 1.0203, "step": 13261 }, { "epoch": 1.37, "learning_rate": 4.394655684324181e-05, "loss": 1.1189, "step": 13262 }, { "epoch": 1.37, "learning_rate": 4.3944377526349725e-05, "loss": 1.4682, "step": 13263 }, { "epoch": 1.37, "learning_rate": 4.394219811558736e-05, "loss": 0.621, "step": 13264 }, { "epoch": 1.37, "learning_rate": 4.394001861096938e-05, "loss": 0.3129, "step": 13265 }, { "epoch": 1.37, "learning_rate": 4.393783901251046e-05, "loss": 0.1613, "step": 13266 }, { "epoch": 1.37, "learning_rate": 4.393565932022526e-05, "loss": 0.3604, "step": 13267 }, { "epoch": 1.37, "learning_rate": 4.3933479534128464e-05, "loss": 0.3883, "step": 13268 }, { "epoch": 1.37, "learning_rate": 4.393129965423475e-05, "loss": 0.4559, "step": 13269 }, { "epoch": 1.37, "learning_rate": 4.3929119680558784e-05, "loss": 0.2617, "step": 13270 }, { "epoch": 1.37, "learning_rate": 4.392693961311524e-05, "loss": 0.3953, "step": 13271 }, { "epoch": 1.37, "learning_rate": 4.39247594519188e-05, "loss": 0.8508, "step": 13272 }, { "epoch": 1.37, "learning_rate": 4.3922579196984133e-05, "loss": 0.9487, "step": 13273 }, { "epoch": 1.37, "learning_rate": 4.392039884832593e-05, "loss": 0.6454, "step": 13274 }, { "epoch": 1.37, "learning_rate": 4.3918218405958855e-05, "loss": 1.3337, "step": 13275 }, { "epoch": 1.37, "learning_rate": 4.3916037869897584e-05, "loss": 0.9403, "step": 13276 }, { "epoch": 1.37, "learning_rate": 4.391385724015681e-05, "loss": 0.6593, "step": 13277 }, { "epoch": 1.37, "learning_rate": 4.3911676516751195e-05, "loss": 1.6238, "step": 13278 }, { "epoch": 1.37, "learning_rate": 4.390949569969544e-05, "loss": 0.8391, "step": 13279 }, { "epoch": 1.37, "learning_rate": 4.3907314789004215e-05, "loss": 0.6872, "step": 13280 }, { "epoch": 1.37, "learning_rate": 4.3905133784692194e-05, "loss": 0.936, "step": 13281 }, { "epoch": 1.37, "learning_rate": 4.390295268677406e-05, "loss": 0.4006, "step": 13282 }, { "epoch": 1.37, "learning_rate": 4.390077149526451e-05, "loss": 0.6222, "step": 13283 }, { "epoch": 1.37, "learning_rate": 4.389859021017822e-05, "loss": 0.6242, "step": 13284 }, { "epoch": 1.37, "learning_rate": 4.389640883152987e-05, "loss": 1.1104, "step": 13285 }, { "epoch": 1.37, "learning_rate": 4.389422735933415e-05, "loss": 1.0112, "step": 13286 }, { "epoch": 1.37, "learning_rate": 4.3892045793605737e-05, "loss": 1.0329, "step": 13287 }, { "epoch": 1.37, "learning_rate": 4.3889864134359323e-05, "loss": 0.1863, "step": 13288 }, { "epoch": 1.37, "learning_rate": 4.38876823816096e-05, "loss": 0.7351, "step": 13289 }, { "epoch": 1.37, "learning_rate": 4.388550053537124e-05, "loss": 0.793, "step": 13290 }, { "epoch": 1.37, "learning_rate": 4.388331859565895e-05, "loss": 0.2297, "step": 13291 }, { "epoch": 1.37, "learning_rate": 4.38811365624874e-05, "loss": 0.5107, "step": 13292 }, { "epoch": 1.37, "learning_rate": 4.387895443587129e-05, "loss": 0.948, "step": 13293 }, { "epoch": 1.37, "learning_rate": 4.3876772215825306e-05, "loss": 0.5739, "step": 13294 }, { "epoch": 1.37, "learning_rate": 4.387458990236413e-05, "loss": 0.9845, "step": 13295 }, { "epoch": 1.37, "learning_rate": 4.3872407495502466e-05, "loss": 0.8093, "step": 13296 }, { "epoch": 1.37, "learning_rate": 4.3870224995255e-05, "loss": 1.0035, "step": 13297 }, { "epoch": 1.37, "learning_rate": 4.3868042401636436e-05, "loss": 0.1751, "step": 13298 }, { "epoch": 1.37, "learning_rate": 4.386585971466145e-05, "loss": 0.7464, "step": 13299 }, { "epoch": 1.37, "learning_rate": 4.386367693434474e-05, "loss": 0.5674, "step": 13300 }, { "epoch": 1.37, "learning_rate": 4.3861494060700994e-05, "loss": 0.5671, "step": 13301 }, { "epoch": 1.37, "learning_rate": 4.385931109374491e-05, "loss": 1.0144, "step": 13302 }, { "epoch": 1.37, "learning_rate": 4.38571280334912e-05, "loss": 0.4062, "step": 13303 }, { "epoch": 1.37, "learning_rate": 4.385494487995454e-05, "loss": 0.3175, "step": 13304 }, { "epoch": 1.37, "learning_rate": 4.3852761633149635e-05, "loss": 0.9485, "step": 13305 }, { "epoch": 1.37, "learning_rate": 4.3850578293091184e-05, "loss": 0.3154, "step": 13306 }, { "epoch": 1.37, "learning_rate": 4.384839485979387e-05, "loss": 0.2195, "step": 13307 }, { "epoch": 1.37, "learning_rate": 4.384621133327241e-05, "loss": 0.8854, "step": 13308 }, { "epoch": 1.37, "learning_rate": 4.3844027713541486e-05, "loss": 0.3521, "step": 13309 }, { "epoch": 1.37, "learning_rate": 4.384184400061581e-05, "loss": 1.431, "step": 13310 }, { "epoch": 1.37, "learning_rate": 4.383966019451009e-05, "loss": 1.2463, "step": 13311 }, { "epoch": 1.37, "learning_rate": 4.383747629523901e-05, "loss": 0.092, "step": 13312 }, { "epoch": 1.37, "learning_rate": 4.3835292302817275e-05, "loss": 0.9086, "step": 13313 }, { "epoch": 1.38, "learning_rate": 4.383310821725958e-05, "loss": 0.4799, "step": 13314 }, { "epoch": 1.38, "learning_rate": 4.383092403858065e-05, "loss": 0.4034, "step": 13315 }, { "epoch": 1.38, "learning_rate": 4.3828739766795166e-05, "loss": 1.0251, "step": 13316 }, { "epoch": 1.38, "learning_rate": 4.3826555401917853e-05, "loss": 0.2593, "step": 13317 }, { "epoch": 1.38, "learning_rate": 4.3824370943963404e-05, "loss": 0.6964, "step": 13318 }, { "epoch": 1.38, "learning_rate": 4.3822186392946516e-05, "loss": 0.4582, "step": 13319 }, { "epoch": 1.38, "learning_rate": 4.382000174888191e-05, "loss": 0.9126, "step": 13320 }, { "epoch": 1.38, "learning_rate": 4.3817817011784274e-05, "loss": 0.8565, "step": 13321 }, { "epoch": 1.38, "learning_rate": 4.381563218166834e-05, "loss": 0.3668, "step": 13322 }, { "epoch": 1.38, "learning_rate": 4.38134472585488e-05, "loss": 0.2211, "step": 13323 }, { "epoch": 1.38, "learning_rate": 4.381126224244036e-05, "loss": 0.4158, "step": 13324 }, { "epoch": 1.38, "learning_rate": 4.380907713335773e-05, "loss": 0.5635, "step": 13325 }, { "epoch": 1.38, "learning_rate": 4.380689193131564e-05, "loss": 0.8519, "step": 13326 }, { "epoch": 1.38, "learning_rate": 4.380470663632877e-05, "loss": 0.6205, "step": 13327 }, { "epoch": 1.38, "learning_rate": 4.380252124841185e-05, "loss": 0.2272, "step": 13328 }, { "epoch": 1.38, "learning_rate": 4.380033576757959e-05, "loss": 0.8703, "step": 13329 }, { "epoch": 1.38, "learning_rate": 4.3798150193846695e-05, "loss": 0.9444, "step": 13330 }, { "epoch": 1.38, "learning_rate": 4.379596452722788e-05, "loss": 0.2198, "step": 13331 }, { "epoch": 1.38, "learning_rate": 4.3793778767737864e-05, "loss": 0.1088, "step": 13332 }, { "epoch": 1.38, "learning_rate": 4.379159291539135e-05, "loss": 0.8216, "step": 13333 }, { "epoch": 1.38, "learning_rate": 4.3789406970203065e-05, "loss": 1.2132, "step": 13334 }, { "epoch": 1.38, "learning_rate": 4.378722093218771e-05, "loss": 0.3879, "step": 13335 }, { "epoch": 1.38, "learning_rate": 4.378503480136003e-05, "loss": 0.3019, "step": 13336 }, { "epoch": 1.38, "learning_rate": 4.378284857773471e-05, "loss": 1.0891, "step": 13337 }, { "epoch": 1.38, "learning_rate": 4.378066226132647e-05, "loss": 0.7973, "step": 13338 }, { "epoch": 1.38, "learning_rate": 4.377847585215005e-05, "loss": 0.2747, "step": 13339 }, { "epoch": 1.38, "learning_rate": 4.377628935022014e-05, "loss": 0.5795, "step": 13340 }, { "epoch": 1.38, "learning_rate": 4.377410275555148e-05, "loss": 0.8856, "step": 13341 }, { "epoch": 1.38, "learning_rate": 4.3771916068158794e-05, "loss": 0.9301, "step": 13342 }, { "epoch": 1.38, "learning_rate": 4.376972928805678e-05, "loss": 0.2171, "step": 13343 }, { "epoch": 1.38, "learning_rate": 4.3767542415260165e-05, "loss": 0.9741, "step": 13344 }, { "epoch": 1.38, "learning_rate": 4.376535544978369e-05, "loss": 0.2811, "step": 13345 }, { "epoch": 1.38, "learning_rate": 4.376316839164205e-05, "loss": 1.3706, "step": 13346 }, { "epoch": 1.38, "learning_rate": 4.376098124084998e-05, "loss": 0.5225, "step": 13347 }, { "epoch": 1.38, "learning_rate": 4.375879399742222e-05, "loss": 0.5707, "step": 13348 }, { "epoch": 1.38, "learning_rate": 4.375660666137346e-05, "loss": 0.2697, "step": 13349 }, { "epoch": 1.38, "learning_rate": 4.3754419232718454e-05, "loss": 0.441, "step": 13350 }, { "epoch": 1.38, "learning_rate": 4.375223171147191e-05, "loss": 0.2487, "step": 13351 }, { "epoch": 1.38, "learning_rate": 4.375004409764857e-05, "loss": 0.3364, "step": 13352 }, { "epoch": 1.38, "learning_rate": 4.3747856391263134e-05, "loss": 0.5759, "step": 13353 }, { "epoch": 1.38, "learning_rate": 4.374566859233035e-05, "loss": 0.2153, "step": 13354 }, { "epoch": 1.38, "learning_rate": 4.3743480700864955e-05, "loss": 0.7986, "step": 13355 }, { "epoch": 1.38, "learning_rate": 4.374129271688166e-05, "loss": 0.4218, "step": 13356 }, { "epoch": 1.38, "learning_rate": 4.373910464039519e-05, "loss": 1.341, "step": 13357 }, { "epoch": 1.38, "learning_rate": 4.373691647142028e-05, "loss": 0.657, "step": 13358 }, { "epoch": 1.38, "learning_rate": 4.3734728209971665e-05, "loss": 0.6415, "step": 13359 }, { "epoch": 1.38, "learning_rate": 4.373253985606408e-05, "loss": 0.4814, "step": 13360 }, { "epoch": 1.38, "learning_rate": 4.373035140971225e-05, "loss": 0.647, "step": 13361 }, { "epoch": 1.38, "learning_rate": 4.37281628709309e-05, "loss": 0.5055, "step": 13362 }, { "epoch": 1.38, "learning_rate": 4.372597423973477e-05, "loss": 0.5574, "step": 13363 }, { "epoch": 1.38, "learning_rate": 4.372378551613859e-05, "loss": 0.4088, "step": 13364 }, { "epoch": 1.38, "learning_rate": 4.372159670015711e-05, "loss": 0.874, "step": 13365 }, { "epoch": 1.38, "learning_rate": 4.3719407791805036e-05, "loss": 0.8015, "step": 13366 }, { "epoch": 1.38, "learning_rate": 4.371721879109713e-05, "loss": 1.7586, "step": 13367 }, { "epoch": 1.38, "learning_rate": 4.3715029698048116e-05, "loss": 0.0723, "step": 13368 }, { "epoch": 1.38, "learning_rate": 4.3712840512672724e-05, "loss": 0.6484, "step": 13369 }, { "epoch": 1.38, "learning_rate": 4.371065123498571e-05, "loss": 0.8366, "step": 13370 }, { "epoch": 1.38, "learning_rate": 4.3708461865001786e-05, "loss": 0.6137, "step": 13371 }, { "epoch": 1.38, "learning_rate": 4.3706272402735715e-05, "loss": 0.3272, "step": 13372 }, { "epoch": 1.38, "learning_rate": 4.3704082848202217e-05, "loss": 0.7114, "step": 13373 }, { "epoch": 1.38, "learning_rate": 4.370189320141604e-05, "loss": 0.2169, "step": 13374 }, { "epoch": 1.38, "learning_rate": 4.369970346239194e-05, "loss": 1.1996, "step": 13375 }, { "epoch": 1.38, "learning_rate": 4.369751363114462e-05, "loss": 0.3035, "step": 13376 }, { "epoch": 1.38, "learning_rate": 4.369532370768885e-05, "loss": 0.2431, "step": 13377 }, { "epoch": 1.38, "learning_rate": 4.369313369203936e-05, "loss": 1.0472, "step": 13378 }, { "epoch": 1.38, "learning_rate": 4.369094358421091e-05, "loss": 0.32, "step": 13379 }, { "epoch": 1.38, "learning_rate": 4.368875338421824e-05, "loss": 0.4795, "step": 13380 }, { "epoch": 1.38, "learning_rate": 4.368656309207607e-05, "loss": 0.3426, "step": 13381 }, { "epoch": 1.38, "learning_rate": 4.368437270779916e-05, "loss": 0.8482, "step": 13382 }, { "epoch": 1.38, "learning_rate": 4.368218223140225e-05, "loss": 0.5539, "step": 13383 }, { "epoch": 1.38, "learning_rate": 4.3679991662900104e-05, "loss": 0.5134, "step": 13384 }, { "epoch": 1.38, "learning_rate": 4.367780100230744e-05, "loss": 0.5027, "step": 13385 }, { "epoch": 1.38, "learning_rate": 4.3675610249639034e-05, "loss": 1.0478, "step": 13386 }, { "epoch": 1.38, "learning_rate": 4.367341940490961e-05, "loss": 0.314, "step": 13387 }, { "epoch": 1.38, "learning_rate": 4.367122846813393e-05, "loss": 0.7421, "step": 13388 }, { "epoch": 1.38, "learning_rate": 4.366903743932674e-05, "loss": 0.414, "step": 13389 }, { "epoch": 1.38, "learning_rate": 4.3666846318502785e-05, "loss": 0.3483, "step": 13390 }, { "epoch": 1.38, "learning_rate": 4.366465510567682e-05, "loss": 0.1372, "step": 13391 }, { "epoch": 1.38, "learning_rate": 4.366246380086359e-05, "loss": 0.6343, "step": 13392 }, { "epoch": 1.38, "learning_rate": 4.3660272404077856e-05, "loss": 0.771, "step": 13393 }, { "epoch": 1.38, "learning_rate": 4.365808091533437e-05, "loss": 1.0575, "step": 13394 }, { "epoch": 1.38, "learning_rate": 4.365588933464787e-05, "loss": 0.6372, "step": 13395 }, { "epoch": 1.38, "learning_rate": 4.3653697662033115e-05, "loss": 0.9309, "step": 13396 }, { "epoch": 1.38, "learning_rate": 4.365150589750487e-05, "loss": 0.3644, "step": 13397 }, { "epoch": 1.38, "learning_rate": 4.364931404107788e-05, "loss": 1.1425, "step": 13398 }, { "epoch": 1.38, "learning_rate": 4.3647122092766904e-05, "loss": 0.8681, "step": 13399 }, { "epoch": 1.38, "learning_rate": 4.364493005258669e-05, "loss": 0.8132, "step": 13400 }, { "epoch": 1.38, "learning_rate": 4.3642737920552e-05, "loss": 0.1885, "step": 13401 }, { "epoch": 1.38, "learning_rate": 4.364054569667759e-05, "loss": 1.2067, "step": 13402 }, { "epoch": 1.38, "learning_rate": 4.363835338097822e-05, "loss": 0.6689, "step": 13403 }, { "epoch": 1.38, "learning_rate": 4.3636160973468645e-05, "loss": 0.8052, "step": 13404 }, { "epoch": 1.38, "learning_rate": 4.363396847416363e-05, "loss": 0.2672, "step": 13405 }, { "epoch": 1.38, "learning_rate": 4.3631775883077915e-05, "loss": 0.3635, "step": 13406 }, { "epoch": 1.38, "learning_rate": 4.362958320022629e-05, "loss": 0.3045, "step": 13407 }, { "epoch": 1.38, "learning_rate": 4.36273904256235e-05, "loss": 0.7725, "step": 13408 }, { "epoch": 1.38, "learning_rate": 4.362519755928429e-05, "loss": 0.2189, "step": 13409 }, { "epoch": 1.39, "learning_rate": 4.3623004601223454e-05, "loss": 0.8537, "step": 13410 }, { "epoch": 1.39, "learning_rate": 4.3620811551455736e-05, "loss": 0.6457, "step": 13411 }, { "epoch": 1.39, "learning_rate": 4.3618618409995895e-05, "loss": 0.247, "step": 13412 }, { "epoch": 1.39, "learning_rate": 4.361642517685871e-05, "loss": 0.8718, "step": 13413 }, { "epoch": 1.39, "learning_rate": 4.361423185205893e-05, "loss": 0.4297, "step": 13414 }, { "epoch": 1.39, "learning_rate": 4.361203843561133e-05, "loss": 0.6366, "step": 13415 }, { "epoch": 1.39, "learning_rate": 4.360984492753066e-05, "loss": 1.1313, "step": 13416 }, { "epoch": 1.39, "learning_rate": 4.3607651327831714e-05, "loss": 1.1895, "step": 13417 }, { "epoch": 1.39, "learning_rate": 4.3605457636529244e-05, "loss": 0.2793, "step": 13418 }, { "epoch": 1.39, "learning_rate": 4.3603263853638004e-05, "loss": 0.3986, "step": 13419 }, { "epoch": 1.39, "learning_rate": 4.3601069979172784e-05, "loss": 1.0807, "step": 13420 }, { "epoch": 1.39, "learning_rate": 4.359887601314834e-05, "loss": 1.3255, "step": 13421 }, { "epoch": 1.39, "learning_rate": 4.359668195557945e-05, "loss": 0.1595, "step": 13422 }, { "epoch": 1.39, "learning_rate": 4.3594487806480874e-05, "loss": 0.5993, "step": 13423 }, { "epoch": 1.39, "learning_rate": 4.359229356586739e-05, "loss": 1.3974, "step": 13424 }, { "epoch": 1.39, "learning_rate": 4.3590099233753757e-05, "loss": 0.6899, "step": 13425 }, { "epoch": 1.39, "learning_rate": 4.358790481015476e-05, "loss": 0.376, "step": 13426 }, { "epoch": 1.39, "learning_rate": 4.358571029508517e-05, "loss": 0.7383, "step": 13427 }, { "epoch": 1.39, "learning_rate": 4.358351568855975e-05, "loss": 0.5231, "step": 13428 }, { "epoch": 1.39, "learning_rate": 4.358132099059329e-05, "loss": 0.8354, "step": 13429 }, { "epoch": 1.39, "learning_rate": 4.3579126201200556e-05, "loss": 0.608, "step": 13430 }, { "epoch": 1.39, "learning_rate": 4.3576931320396315e-05, "loss": 0.2565, "step": 13431 }, { "epoch": 1.39, "learning_rate": 4.357473634819535e-05, "loss": 0.8537, "step": 13432 }, { "epoch": 1.39, "learning_rate": 4.357254128461244e-05, "loss": 1.0896, "step": 13433 }, { "epoch": 1.39, "learning_rate": 4.357034612966235e-05, "loss": 0.9239, "step": 13434 }, { "epoch": 1.39, "learning_rate": 4.356815088335987e-05, "loss": 0.2557, "step": 13435 }, { "epoch": 1.39, "learning_rate": 4.356595554571977e-05, "loss": 0.3097, "step": 13436 }, { "epoch": 1.39, "learning_rate": 4.3563760116756846e-05, "loss": 1.1554, "step": 13437 }, { "epoch": 1.39, "learning_rate": 4.356156459648585e-05, "loss": 0.4295, "step": 13438 }, { "epoch": 1.39, "learning_rate": 4.355936898492157e-05, "loss": 0.2793, "step": 13439 }, { "epoch": 1.39, "learning_rate": 4.355717328207879e-05, "loss": 0.3235, "step": 13440 }, { "epoch": 1.39, "learning_rate": 4.35549774879723e-05, "loss": 0.4823, "step": 13441 }, { "epoch": 1.39, "learning_rate": 4.3552781602616876e-05, "loss": 0.9966, "step": 13442 }, { "epoch": 1.39, "learning_rate": 4.3550585626027294e-05, "loss": 0.2584, "step": 13443 }, { "epoch": 1.39, "learning_rate": 4.3548389558218336e-05, "loss": 0.6234, "step": 13444 }, { "epoch": 1.39, "learning_rate": 4.3546193399204786e-05, "loss": 0.4236, "step": 13445 }, { "epoch": 1.39, "learning_rate": 4.3543997149001444e-05, "loss": 0.3331, "step": 13446 }, { "epoch": 1.39, "learning_rate": 4.3541800807623076e-05, "loss": 0.5884, "step": 13447 }, { "epoch": 1.39, "learning_rate": 4.353960437508447e-05, "loss": 0.2308, "step": 13448 }, { "epoch": 1.39, "learning_rate": 4.353740785140042e-05, "loss": 0.8224, "step": 13449 }, { "epoch": 1.39, "learning_rate": 4.353521123658571e-05, "loss": 0.4673, "step": 13450 }, { "epoch": 1.39, "learning_rate": 4.353301453065512e-05, "loss": 0.7984, "step": 13451 }, { "epoch": 1.39, "learning_rate": 4.353081773362344e-05, "loss": 0.8468, "step": 13452 }, { "epoch": 1.39, "learning_rate": 4.3528620845505466e-05, "loss": 0.6274, "step": 13453 }, { "epoch": 1.39, "learning_rate": 4.352642386631598e-05, "loss": 0.4931, "step": 13454 }, { "epoch": 1.39, "learning_rate": 4.352422679606978e-05, "loss": 0.9716, "step": 13455 }, { "epoch": 1.39, "learning_rate": 4.352202963478164e-05, "loss": 1.0632, "step": 13456 }, { "epoch": 1.39, "learning_rate": 4.351983238246637e-05, "loss": 0.3551, "step": 13457 }, { "epoch": 1.39, "learning_rate": 4.351763503913874e-05, "loss": 0.62, "step": 13458 }, { "epoch": 1.39, "learning_rate": 4.351543760481356e-05, "loss": 0.7052, "step": 13459 }, { "epoch": 1.39, "learning_rate": 4.351324007950562e-05, "loss": 0.2801, "step": 13460 }, { "epoch": 1.39, "learning_rate": 4.351104246322971e-05, "loss": 0.3498, "step": 13461 }, { "epoch": 1.39, "learning_rate": 4.3508844756000615e-05, "loss": 2.1125, "step": 13462 }, { "epoch": 1.39, "learning_rate": 4.350664695783314e-05, "loss": 0.7555, "step": 13463 }, { "epoch": 1.39, "learning_rate": 4.350444906874208e-05, "loss": 0.9913, "step": 13464 }, { "epoch": 1.39, "learning_rate": 4.350225108874223e-05, "loss": 0.4955, "step": 13465 }, { "epoch": 1.39, "learning_rate": 4.3500053017848385e-05, "loss": 0.193, "step": 13466 }, { "epoch": 1.39, "learning_rate": 4.349785485607533e-05, "loss": 1.0084, "step": 13467 }, { "epoch": 1.39, "learning_rate": 4.34956566034379e-05, "loss": 0.6885, "step": 13468 }, { "epoch": 1.39, "learning_rate": 4.3493458259950845e-05, "loss": 0.6434, "step": 13469 }, { "epoch": 1.39, "learning_rate": 4.3491259825628997e-05, "loss": 0.7971, "step": 13470 }, { "epoch": 1.39, "learning_rate": 4.3489061300487144e-05, "loss": 0.3177, "step": 13471 }, { "epoch": 1.39, "learning_rate": 4.348686268454008e-05, "loss": 0.6848, "step": 13472 }, { "epoch": 1.39, "learning_rate": 4.3484663977802616e-05, "loss": 0.8232, "step": 13473 }, { "epoch": 1.39, "learning_rate": 4.3482465180289553e-05, "loss": 1.0089, "step": 13474 }, { "epoch": 1.39, "learning_rate": 4.3480266292015684e-05, "loss": 1.0604, "step": 13475 }, { "epoch": 1.39, "learning_rate": 4.3478067312995814e-05, "loss": 1.0052, "step": 13476 }, { "epoch": 1.39, "learning_rate": 4.3475868243244756e-05, "loss": 0.2621, "step": 13477 }, { "epoch": 1.39, "learning_rate": 4.3473669082777296e-05, "loss": 0.2233, "step": 13478 }, { "epoch": 1.39, "learning_rate": 4.347146983160826e-05, "loss": 0.4702, "step": 13479 }, { "epoch": 1.39, "learning_rate": 4.346927048975244e-05, "loss": 0.613, "step": 13480 }, { "epoch": 1.39, "learning_rate": 4.3467071057224634e-05, "loss": 0.7342, "step": 13481 }, { "epoch": 1.39, "learning_rate": 4.346487153403966e-05, "loss": 0.387, "step": 13482 }, { "epoch": 1.39, "learning_rate": 4.346267192021232e-05, "loss": 0.7225, "step": 13483 }, { "epoch": 1.39, "learning_rate": 4.346047221575743e-05, "loss": 0.316, "step": 13484 }, { "epoch": 1.39, "learning_rate": 4.345827242068978e-05, "loss": 1.0319, "step": 13485 }, { "epoch": 1.39, "learning_rate": 4.3456072535024196e-05, "loss": 0.5848, "step": 13486 }, { "epoch": 1.39, "learning_rate": 4.3453872558775476e-05, "loss": 0.68, "step": 13487 }, { "epoch": 1.39, "learning_rate": 4.3451672491958436e-05, "loss": 0.2129, "step": 13488 }, { "epoch": 1.39, "learning_rate": 4.3449472334587895e-05, "loss": 0.6556, "step": 13489 }, { "epoch": 1.39, "learning_rate": 4.344727208667864e-05, "loss": 0.5889, "step": 13490 }, { "epoch": 1.39, "learning_rate": 4.3445071748245496e-05, "loss": 0.7335, "step": 13491 }, { "epoch": 1.39, "learning_rate": 4.344287131930328e-05, "loss": 0.1778, "step": 13492 }, { "epoch": 1.39, "learning_rate": 4.34406707998668e-05, "loss": 1.144, "step": 13493 }, { "epoch": 1.39, "learning_rate": 4.3438470189950864e-05, "loss": 0.7797, "step": 13494 }, { "epoch": 1.39, "learning_rate": 4.343626948957029e-05, "loss": 0.1664, "step": 13495 }, { "epoch": 1.39, "learning_rate": 4.34340686987399e-05, "loss": 0.4926, "step": 13496 }, { "epoch": 1.39, "learning_rate": 4.3431867817474496e-05, "loss": 0.8079, "step": 13497 }, { "epoch": 1.39, "learning_rate": 4.342966684578891e-05, "loss": 0.7618, "step": 13498 }, { "epoch": 1.39, "learning_rate": 4.3427465783697944e-05, "loss": 0.3986, "step": 13499 }, { "epoch": 1.39, "learning_rate": 4.342526463121642e-05, "loss": 0.7561, "step": 13500 }, { "epoch": 1.39, "learning_rate": 4.342306338835916e-05, "loss": 0.4213, "step": 13501 }, { "epoch": 1.39, "learning_rate": 4.342086205514096e-05, "loss": 0.2553, "step": 13502 }, { "epoch": 1.39, "learning_rate": 4.341866063157668e-05, "loss": 0.2782, "step": 13503 }, { "epoch": 1.39, "learning_rate": 4.34164591176811e-05, "loss": 0.5926, "step": 13504 }, { "epoch": 1.39, "learning_rate": 4.3414257513469066e-05, "loss": 0.6787, "step": 13505 }, { "epoch": 1.4, "learning_rate": 4.3412055818955385e-05, "loss": 0.6312, "step": 13506 }, { "epoch": 1.4, "learning_rate": 4.340985403415488e-05, "loss": 0.172, "step": 13507 }, { "epoch": 1.4, "learning_rate": 4.3407652159082386e-05, "loss": 1.0721, "step": 13508 }, { "epoch": 1.4, "learning_rate": 4.340545019375271e-05, "loss": 0.1382, "step": 13509 }, { "epoch": 1.4, "learning_rate": 4.3403248138180675e-05, "loss": 0.6621, "step": 13510 }, { "epoch": 1.4, "learning_rate": 4.340104599238111e-05, "loss": 0.7044, "step": 13511 }, { "epoch": 1.4, "learning_rate": 4.3398843756368844e-05, "loss": 0.4872, "step": 13512 }, { "epoch": 1.4, "learning_rate": 4.33966414301587e-05, "loss": 0.8613, "step": 13513 }, { "epoch": 1.4, "learning_rate": 4.339443901376549e-05, "loss": 1.1758, "step": 13514 }, { "epoch": 1.4, "learning_rate": 4.339223650720405e-05, "loss": 0.4492, "step": 13515 }, { "epoch": 1.4, "learning_rate": 4.339003391048921e-05, "loss": 0.9794, "step": 13516 }, { "epoch": 1.4, "learning_rate": 4.3387831223635803e-05, "loss": 0.4842, "step": 13517 }, { "epoch": 1.4, "learning_rate": 4.3385628446658644e-05, "loss": 0.2793, "step": 13518 }, { "epoch": 1.4, "learning_rate": 4.338342557957257e-05, "loss": 0.4723, "step": 13519 }, { "epoch": 1.4, "learning_rate": 4.33812226223924e-05, "loss": 0.187, "step": 13520 }, { "epoch": 1.4, "learning_rate": 4.337901957513297e-05, "loss": 0.4259, "step": 13521 }, { "epoch": 1.4, "learning_rate": 4.337681643780911e-05, "loss": 1.0202, "step": 13522 }, { "epoch": 1.4, "learning_rate": 4.337461321043566e-05, "loss": 0.2351, "step": 13523 }, { "epoch": 1.4, "learning_rate": 4.3372409893027436e-05, "loss": 1.4141, "step": 13524 }, { "epoch": 1.4, "learning_rate": 4.337020648559928e-05, "loss": 0.4924, "step": 13525 }, { "epoch": 1.4, "learning_rate": 4.336800298816602e-05, "loss": 0.2146, "step": 13526 }, { "epoch": 1.4, "learning_rate": 4.33657994007425e-05, "loss": 0.4515, "step": 13527 }, { "epoch": 1.4, "learning_rate": 4.336359572334353e-05, "loss": 2.4222, "step": 13528 }, { "epoch": 1.4, "learning_rate": 4.336139195598397e-05, "loss": 0.4882, "step": 13529 }, { "epoch": 1.4, "learning_rate": 4.3359188098678644e-05, "loss": 0.5916, "step": 13530 }, { "epoch": 1.4, "learning_rate": 4.335698415144239e-05, "loss": 0.9142, "step": 13531 }, { "epoch": 1.4, "learning_rate": 4.3354780114290046e-05, "loss": 0.285, "step": 13532 }, { "epoch": 1.4, "learning_rate": 4.335257598723645e-05, "loss": 0.7608, "step": 13533 }, { "epoch": 1.4, "learning_rate": 4.335037177029643e-05, "loss": 0.7175, "step": 13534 }, { "epoch": 1.4, "learning_rate": 4.334816746348483e-05, "loss": 1.0158, "step": 13535 }, { "epoch": 1.4, "learning_rate": 4.33459630668165e-05, "loss": 0.4787, "step": 13536 }, { "epoch": 1.4, "learning_rate": 4.3343758580306265e-05, "loss": 0.9015, "step": 13537 }, { "epoch": 1.4, "learning_rate": 4.3341554003968965e-05, "loss": 0.7539, "step": 13538 }, { "epoch": 1.4, "learning_rate": 4.333934933781944e-05, "loss": 0.9326, "step": 13539 }, { "epoch": 1.4, "learning_rate": 4.3337144581872545e-05, "loss": 0.4047, "step": 13540 }, { "epoch": 1.4, "learning_rate": 4.333493973614311e-05, "loss": 0.8166, "step": 13541 }, { "epoch": 1.4, "learning_rate": 4.3332734800645986e-05, "loss": 0.2792, "step": 13542 }, { "epoch": 1.4, "learning_rate": 4.333052977539601e-05, "loss": 2.2102, "step": 13543 }, { "epoch": 1.4, "learning_rate": 4.332832466040802e-05, "loss": 0.6727, "step": 13544 }, { "epoch": 1.4, "learning_rate": 4.332611945569687e-05, "loss": 0.77, "step": 13545 }, { "epoch": 1.4, "learning_rate": 4.332391416127741e-05, "loss": 0.5826, "step": 13546 }, { "epoch": 1.4, "learning_rate": 4.332170877716446e-05, "loss": 1.0537, "step": 13547 }, { "epoch": 1.4, "learning_rate": 4.33195033033729e-05, "loss": 0.6484, "step": 13548 }, { "epoch": 1.4, "learning_rate": 4.331729773991755e-05, "loss": 0.788, "step": 13549 }, { "epoch": 1.4, "learning_rate": 4.3315092086813274e-05, "loss": 0.7496, "step": 13550 }, { "epoch": 1.4, "learning_rate": 4.3312886344074926e-05, "loss": 0.8675, "step": 13551 }, { "epoch": 1.4, "learning_rate": 4.331068051171732e-05, "loss": 0.8811, "step": 13552 }, { "epoch": 1.4, "learning_rate": 4.330847458975533e-05, "loss": 0.4337, "step": 13553 }, { "epoch": 1.4, "learning_rate": 4.330626857820381e-05, "loss": 0.3297, "step": 13554 }, { "epoch": 1.4, "learning_rate": 4.3304062477077606e-05, "loss": 0.4743, "step": 13555 }, { "epoch": 1.4, "learning_rate": 4.330185628639156e-05, "loss": 1.3417, "step": 13556 }, { "epoch": 1.4, "learning_rate": 4.329965000616053e-05, "loss": 0.8734, "step": 13557 }, { "epoch": 1.4, "learning_rate": 4.329744363639937e-05, "loss": 1.1719, "step": 13558 }, { "epoch": 1.4, "learning_rate": 4.3295237177122924e-05, "loss": 0.6556, "step": 13559 }, { "epoch": 1.4, "learning_rate": 4.3293030628346065e-05, "loss": 0.6559, "step": 13560 }, { "epoch": 1.4, "learning_rate": 4.3290823990083624e-05, "loss": 0.1734, "step": 13561 }, { "epoch": 1.4, "learning_rate": 4.3288617262350475e-05, "loss": 0.354, "step": 13562 }, { "epoch": 1.4, "learning_rate": 4.328641044516145e-05, "loss": 1.1174, "step": 13563 }, { "epoch": 1.4, "learning_rate": 4.3284203538531435e-05, "loss": 1.0449, "step": 13564 }, { "epoch": 1.4, "learning_rate": 4.328199654247526e-05, "loss": 0.4223, "step": 13565 }, { "epoch": 1.4, "learning_rate": 4.327978945700779e-05, "loss": 0.8423, "step": 13566 }, { "epoch": 1.4, "learning_rate": 4.327758228214389e-05, "loss": 0.2481, "step": 13567 }, { "epoch": 1.4, "learning_rate": 4.327537501789841e-05, "loss": 0.3754, "step": 13568 }, { "epoch": 1.4, "learning_rate": 4.327316766428621e-05, "loss": 0.4617, "step": 13569 }, { "epoch": 1.4, "learning_rate": 4.327096022132216e-05, "loss": 0.3651, "step": 13570 }, { "epoch": 1.4, "learning_rate": 4.32687526890211e-05, "loss": 0.8365, "step": 13571 }, { "epoch": 1.4, "learning_rate": 4.326654506739791e-05, "loss": 0.7496, "step": 13572 }, { "epoch": 1.4, "learning_rate": 4.326433735646743e-05, "loss": 0.6366, "step": 13573 }, { "epoch": 1.4, "learning_rate": 4.326212955624455e-05, "loss": 0.4088, "step": 13574 }, { "epoch": 1.4, "learning_rate": 4.325992166674411e-05, "loss": 1.1386, "step": 13575 }, { "epoch": 1.4, "learning_rate": 4.325771368798098e-05, "loss": 0.6196, "step": 13576 }, { "epoch": 1.4, "learning_rate": 4.3255505619970024e-05, "loss": 0.7896, "step": 13577 }, { "epoch": 1.4, "learning_rate": 4.32532974627261e-05, "loss": 0.9345, "step": 13578 }, { "epoch": 1.4, "learning_rate": 4.325108921626409e-05, "loss": 0.9409, "step": 13579 }, { "epoch": 1.4, "learning_rate": 4.324888088059885e-05, "loss": 0.559, "step": 13580 }, { "epoch": 1.4, "learning_rate": 4.324667245574523e-05, "loss": 0.4842, "step": 13581 }, { "epoch": 1.4, "learning_rate": 4.324446394171812e-05, "loss": 0.8791, "step": 13582 }, { "epoch": 1.4, "learning_rate": 4.3242255338532365e-05, "loss": 1.5783, "step": 13583 }, { "epoch": 1.4, "learning_rate": 4.3240046646202864e-05, "loss": 0.5469, "step": 13584 }, { "epoch": 1.4, "learning_rate": 4.3237837864744456e-05, "loss": 0.2309, "step": 13585 }, { "epoch": 1.4, "learning_rate": 4.3235628994172016e-05, "loss": 0.5635, "step": 13586 }, { "epoch": 1.4, "learning_rate": 4.323342003450043e-05, "loss": 0.3445, "step": 13587 }, { "epoch": 1.4, "learning_rate": 4.323121098574455e-05, "loss": 0.2623, "step": 13588 }, { "epoch": 1.4, "learning_rate": 4.3229001847919254e-05, "loss": 0.471, "step": 13589 }, { "epoch": 1.4, "learning_rate": 4.322679262103941e-05, "loss": 1.0279, "step": 13590 }, { "epoch": 1.4, "learning_rate": 4.3224583305119895e-05, "loss": 0.2288, "step": 13591 }, { "epoch": 1.4, "learning_rate": 4.3222373900175585e-05, "loss": 0.6206, "step": 13592 }, { "epoch": 1.4, "learning_rate": 4.3220164406221346e-05, "loss": 0.3467, "step": 13593 }, { "epoch": 1.4, "learning_rate": 4.321795482327205e-05, "loss": 0.9497, "step": 13594 }, { "epoch": 1.4, "learning_rate": 4.321574515134258e-05, "loss": 0.5884, "step": 13595 }, { "epoch": 1.4, "learning_rate": 4.32135353904478e-05, "loss": 0.7844, "step": 13596 }, { "epoch": 1.4, "learning_rate": 4.3211325540602596e-05, "loss": 0.2481, "step": 13597 }, { "epoch": 1.4, "learning_rate": 4.3209115601821836e-05, "loss": 1.0578, "step": 13598 }, { "epoch": 1.4, "learning_rate": 4.32069055741204e-05, "loss": 0.2106, "step": 13599 }, { "epoch": 1.4, "learning_rate": 4.320469545751317e-05, "loss": 1.0996, "step": 13600 }, { "epoch": 1.41, "learning_rate": 4.320248525201501e-05, "loss": 1.1747, "step": 13601 }, { "epoch": 1.41, "learning_rate": 4.320027495764082e-05, "loss": 0.8852, "step": 13602 }, { "epoch": 1.41, "learning_rate": 4.319806457440547e-05, "loss": 0.5852, "step": 13603 }, { "epoch": 1.41, "learning_rate": 4.319585410232383e-05, "loss": 0.6028, "step": 13604 }, { "epoch": 1.41, "learning_rate": 4.319364354141078e-05, "loss": 1.0554, "step": 13605 }, { "epoch": 1.41, "learning_rate": 4.319143289168122e-05, "loss": 0.3783, "step": 13606 }, { "epoch": 1.41, "learning_rate": 4.318922215315002e-05, "loss": 0.9348, "step": 13607 }, { "epoch": 1.41, "learning_rate": 4.318701132583207e-05, "loss": 0.3465, "step": 13608 }, { "epoch": 1.41, "learning_rate": 4.318480040974223e-05, "loss": 0.4702, "step": 13609 }, { "epoch": 1.41, "learning_rate": 4.31825894048954e-05, "loss": 0.9826, "step": 13610 }, { "epoch": 1.41, "learning_rate": 4.3180378311306466e-05, "loss": 1.1286, "step": 13611 }, { "epoch": 1.41, "learning_rate": 4.3178167128990306e-05, "loss": 0.3218, "step": 13612 }, { "epoch": 1.41, "learning_rate": 4.317595585796182e-05, "loss": 0.8779, "step": 13613 }, { "epoch": 1.41, "learning_rate": 4.317374449823587e-05, "loss": 0.9449, "step": 13614 }, { "epoch": 1.41, "learning_rate": 4.317153304982736e-05, "loss": 0.7114, "step": 13615 }, { "epoch": 1.41, "learning_rate": 4.316932151275117e-05, "loss": 0.2986, "step": 13616 }, { "epoch": 1.41, "learning_rate": 4.316710988702219e-05, "loss": 0.4167, "step": 13617 }, { "epoch": 1.41, "learning_rate": 4.3164898172655305e-05, "loss": 0.4099, "step": 13618 }, { "epoch": 1.41, "learning_rate": 4.316268636966541e-05, "loss": 1.0456, "step": 13619 }, { "epoch": 1.41, "learning_rate": 4.3160474478067384e-05, "loss": 1.2001, "step": 13620 }, { "epoch": 1.41, "learning_rate": 4.3158262497876126e-05, "loss": 1.0259, "step": 13621 }, { "epoch": 1.41, "learning_rate": 4.315605042910653e-05, "loss": 0.3724, "step": 13622 }, { "epoch": 1.41, "learning_rate": 4.315383827177347e-05, "loss": 0.9209, "step": 13623 }, { "epoch": 1.41, "learning_rate": 4.315162602589185e-05, "loss": 0.5198, "step": 13624 }, { "epoch": 1.41, "learning_rate": 4.3149413691476566e-05, "loss": 0.9329, "step": 13625 }, { "epoch": 1.41, "learning_rate": 4.314720126854251e-05, "loss": 0.2558, "step": 13626 }, { "epoch": 1.41, "learning_rate": 4.3144988757104565e-05, "loss": 1.0267, "step": 13627 }, { "epoch": 1.41, "learning_rate": 4.3142776157177627e-05, "loss": 1.6027, "step": 13628 }, { "epoch": 1.41, "learning_rate": 4.31405634687766e-05, "loss": 0.6757, "step": 13629 }, { "epoch": 1.41, "learning_rate": 4.313835069191637e-05, "loss": 0.2554, "step": 13630 }, { "epoch": 1.41, "learning_rate": 4.313613782661185e-05, "loss": 1.3149, "step": 13631 }, { "epoch": 1.41, "learning_rate": 4.313392487287792e-05, "loss": 0.5065, "step": 13632 }, { "epoch": 1.41, "learning_rate": 4.313171183072947e-05, "loss": 0.7677, "step": 13633 }, { "epoch": 1.41, "learning_rate": 4.312949870018141e-05, "loss": 0.482, "step": 13634 }, { "epoch": 1.41, "learning_rate": 4.312728548124864e-05, "loss": 0.4119, "step": 13635 }, { "epoch": 1.41, "learning_rate": 4.312507217394606e-05, "loss": 0.6332, "step": 13636 }, { "epoch": 1.41, "learning_rate": 4.312285877828857e-05, "loss": 0.6965, "step": 13637 }, { "epoch": 1.41, "learning_rate": 4.312064529429105e-05, "loss": 0.991, "step": 13638 }, { "epoch": 1.41, "learning_rate": 4.3118431721968426e-05, "loss": 0.6662, "step": 13639 }, { "epoch": 1.41, "learning_rate": 4.311621806133558e-05, "loss": 1.0714, "step": 13640 }, { "epoch": 1.41, "learning_rate": 4.3114004312407433e-05, "loss": 0.9796, "step": 13641 }, { "epoch": 1.41, "learning_rate": 4.3111790475198875e-05, "loss": 0.2411, "step": 13642 }, { "epoch": 1.41, "learning_rate": 4.31095765497248e-05, "loss": 0.1902, "step": 13643 }, { "epoch": 1.41, "learning_rate": 4.310736253600014e-05, "loss": 0.6506, "step": 13644 }, { "epoch": 1.41, "learning_rate": 4.3105148434039773e-05, "loss": 0.9224, "step": 13645 }, { "epoch": 1.41, "learning_rate": 4.310293424385862e-05, "loss": 0.5462, "step": 13646 }, { "epoch": 1.41, "learning_rate": 4.310071996547157e-05, "loss": 0.6382, "step": 13647 }, { "epoch": 1.41, "learning_rate": 4.309850559889355e-05, "loss": 1.3328, "step": 13648 }, { "epoch": 1.41, "learning_rate": 4.309629114413944e-05, "loss": 0.7308, "step": 13649 }, { "epoch": 1.41, "learning_rate": 4.3094076601224175e-05, "loss": 0.5533, "step": 13650 }, { "epoch": 1.41, "learning_rate": 4.309186197016265e-05, "loss": 1.1199, "step": 13651 }, { "epoch": 1.41, "learning_rate": 4.3089647250969774e-05, "loss": 0.1273, "step": 13652 }, { "epoch": 1.41, "learning_rate": 4.3087432443660446e-05, "loss": 0.4475, "step": 13653 }, { "epoch": 1.41, "learning_rate": 4.308521754824959e-05, "loss": 0.3025, "step": 13654 }, { "epoch": 1.41, "learning_rate": 4.308300256475211e-05, "loss": 0.4512, "step": 13655 }, { "epoch": 1.41, "learning_rate": 4.308078749318292e-05, "loss": 0.1849, "step": 13656 }, { "epoch": 1.41, "learning_rate": 4.307857233355693e-05, "loss": 0.2059, "step": 13657 }, { "epoch": 1.41, "learning_rate": 4.307635708588904e-05, "loss": 0.3806, "step": 13658 }, { "epoch": 1.41, "learning_rate": 4.307414175019419e-05, "loss": 0.2718, "step": 13659 }, { "epoch": 1.41, "learning_rate": 4.307192632648727e-05, "loss": 0.6917, "step": 13660 }, { "epoch": 1.41, "learning_rate": 4.3069710814783204e-05, "loss": 0.3016, "step": 13661 }, { "epoch": 1.41, "learning_rate": 4.306749521509689e-05, "loss": 1.1942, "step": 13662 }, { "epoch": 1.41, "learning_rate": 4.3065279527443274e-05, "loss": 0.1161, "step": 13663 }, { "epoch": 1.41, "learning_rate": 4.306306375183724e-05, "loss": 0.0931, "step": 13664 }, { "epoch": 1.41, "learning_rate": 4.3060847888293726e-05, "loss": 0.2403, "step": 13665 }, { "epoch": 1.41, "learning_rate": 4.3058631936827635e-05, "loss": 0.3578, "step": 13666 }, { "epoch": 1.41, "learning_rate": 4.305641589745388e-05, "loss": 1.2023, "step": 13667 }, { "epoch": 1.41, "learning_rate": 4.3054199770187404e-05, "loss": 0.6107, "step": 13668 }, { "epoch": 1.41, "learning_rate": 4.3051983555043107e-05, "loss": 0.6084, "step": 13669 }, { "epoch": 1.41, "learning_rate": 4.3049767252035906e-05, "loss": 0.3131, "step": 13670 }, { "epoch": 1.41, "learning_rate": 4.304755086118073e-05, "loss": 0.63, "step": 13671 }, { "epoch": 1.41, "learning_rate": 4.3045334382492485e-05, "loss": 0.6496, "step": 13672 }, { "epoch": 1.41, "learning_rate": 4.304311781598611e-05, "loss": 0.5224, "step": 13673 }, { "epoch": 1.41, "learning_rate": 4.304090116167652e-05, "loss": 0.3279, "step": 13674 }, { "epoch": 1.41, "learning_rate": 4.3038684419578626e-05, "loss": 0.7976, "step": 13675 }, { "epoch": 1.41, "learning_rate": 4.3036467589707364e-05, "loss": 0.895, "step": 13676 }, { "epoch": 1.41, "learning_rate": 4.3034250672077654e-05, "loss": 0.3504, "step": 13677 }, { "epoch": 1.41, "learning_rate": 4.303203366670441e-05, "loss": 0.6446, "step": 13678 }, { "epoch": 1.41, "learning_rate": 4.302981657360258e-05, "loss": 0.9457, "step": 13679 }, { "epoch": 1.41, "learning_rate": 4.302759939278706e-05, "loss": 0.1868, "step": 13680 }, { "epoch": 1.41, "learning_rate": 4.30253821242728e-05, "loss": 0.3468, "step": 13681 }, { "epoch": 1.41, "learning_rate": 4.3023164768074704e-05, "loss": 0.6702, "step": 13682 }, { "epoch": 1.41, "learning_rate": 4.302094732420771e-05, "loss": 0.3951, "step": 13683 }, { "epoch": 1.41, "learning_rate": 4.301872979268675e-05, "loss": 0.7855, "step": 13684 }, { "epoch": 1.41, "learning_rate": 4.3016512173526744e-05, "loss": 1.0756, "step": 13685 }, { "epoch": 1.41, "learning_rate": 4.301429446674262e-05, "loss": 0.7994, "step": 13686 }, { "epoch": 1.41, "learning_rate": 4.3012076672349324e-05, "loss": 0.9369, "step": 13687 }, { "epoch": 1.41, "learning_rate": 4.300985879036176e-05, "loss": 0.9543, "step": 13688 }, { "epoch": 1.41, "learning_rate": 4.300764082079489e-05, "loss": 0.7877, "step": 13689 }, { "epoch": 1.41, "learning_rate": 4.300542276366361e-05, "loss": 0.8305, "step": 13690 }, { "epoch": 1.41, "learning_rate": 4.3003204618982865e-05, "loss": 1.3272, "step": 13691 }, { "epoch": 1.41, "learning_rate": 4.300098638676759e-05, "loss": 0.5214, "step": 13692 }, { "epoch": 1.41, "learning_rate": 4.299876806703272e-05, "loss": 0.6678, "step": 13693 }, { "epoch": 1.41, "learning_rate": 4.299654965979319e-05, "loss": 1.0087, "step": 13694 }, { "epoch": 1.41, "learning_rate": 4.299433116506392e-05, "loss": 0.3002, "step": 13695 }, { "epoch": 1.41, "learning_rate": 4.2992112582859855e-05, "loss": 0.1886, "step": 13696 }, { "epoch": 1.42, "learning_rate": 4.2989893913195925e-05, "loss": 0.8289, "step": 13697 }, { "epoch": 1.42, "learning_rate": 4.298767515608708e-05, "loss": 1.0107, "step": 13698 }, { "epoch": 1.42, "learning_rate": 4.298545631154823e-05, "loss": 0.5137, "step": 13699 }, { "epoch": 1.42, "learning_rate": 4.2983237379594346e-05, "loss": 0.3896, "step": 13700 }, { "epoch": 1.42, "learning_rate": 4.2981018360240324e-05, "loss": 0.508, "step": 13701 }, { "epoch": 1.42, "learning_rate": 4.297879925350113e-05, "loss": 0.56, "step": 13702 }, { "epoch": 1.42, "learning_rate": 4.2976580059391704e-05, "loss": 0.5269, "step": 13703 }, { "epoch": 1.42, "learning_rate": 4.297436077792697e-05, "loss": 0.3396, "step": 13704 }, { "epoch": 1.42, "learning_rate": 4.2972141409121874e-05, "loss": 0.3603, "step": 13705 }, { "epoch": 1.42, "learning_rate": 4.296992195299136e-05, "loss": 0.9643, "step": 13706 }, { "epoch": 1.42, "learning_rate": 4.2967702409550375e-05, "loss": 0.3084, "step": 13707 }, { "epoch": 1.42, "learning_rate": 4.2965482778813854e-05, "loss": 0.9796, "step": 13708 }, { "epoch": 1.42, "learning_rate": 4.296326306079672e-05, "loss": 0.7573, "step": 13709 }, { "epoch": 1.42, "learning_rate": 4.2961043255513935e-05, "loss": 0.9136, "step": 13710 }, { "epoch": 1.42, "learning_rate": 4.295882336298044e-05, "loss": 0.6412, "step": 13711 }, { "epoch": 1.42, "learning_rate": 4.295660338321118e-05, "loss": 0.143, "step": 13712 }, { "epoch": 1.42, "learning_rate": 4.295438331622111e-05, "loss": 2.1116, "step": 13713 }, { "epoch": 1.42, "learning_rate": 4.295216316202514e-05, "loss": 0.6548, "step": 13714 }, { "epoch": 1.42, "learning_rate": 4.2949942920638255e-05, "loss": 0.4042, "step": 13715 }, { "epoch": 1.42, "learning_rate": 4.2947722592075375e-05, "loss": 0.8785, "step": 13716 }, { "epoch": 1.42, "learning_rate": 4.294550217635147e-05, "loss": 0.2482, "step": 13717 }, { "epoch": 1.42, "learning_rate": 4.2943281673481454e-05, "loss": 1.1035, "step": 13718 }, { "epoch": 1.42, "learning_rate": 4.294106108348031e-05, "loss": 0.9963, "step": 13719 }, { "epoch": 1.42, "learning_rate": 4.2938840406362964e-05, "loss": 0.7668, "step": 13720 }, { "epoch": 1.42, "learning_rate": 4.293661964214437e-05, "loss": 0.513, "step": 13721 }, { "epoch": 1.42, "learning_rate": 4.293439879083949e-05, "loss": 0.1155, "step": 13722 }, { "epoch": 1.42, "learning_rate": 4.293217785246325e-05, "loss": 0.3444, "step": 13723 }, { "epoch": 1.42, "learning_rate": 4.2929956827030613e-05, "loss": 0.9683, "step": 13724 }, { "epoch": 1.42, "learning_rate": 4.292773571455655e-05, "loss": 0.6792, "step": 13725 }, { "epoch": 1.42, "learning_rate": 4.2925514515055974e-05, "loss": 0.6271, "step": 13726 }, { "epoch": 1.42, "learning_rate": 4.292329322854388e-05, "loss": 0.5452, "step": 13727 }, { "epoch": 1.42, "learning_rate": 4.2921071855035185e-05, "loss": 0.3365, "step": 13728 }, { "epoch": 1.42, "learning_rate": 4.291885039454485e-05, "loss": 0.4847, "step": 13729 }, { "epoch": 1.42, "learning_rate": 4.2916628847087854e-05, "loss": 0.3573, "step": 13730 }, { "epoch": 1.42, "learning_rate": 4.291440721267913e-05, "loss": 1.1039, "step": 13731 }, { "epoch": 1.42, "learning_rate": 4.291218549133364e-05, "loss": 0.6939, "step": 13732 }, { "epoch": 1.42, "learning_rate": 4.2909963683066335e-05, "loss": 0.9914, "step": 13733 }, { "epoch": 1.42, "learning_rate": 4.290774178789217e-05, "loss": 0.3333, "step": 13734 }, { "epoch": 1.42, "learning_rate": 4.290551980582611e-05, "loss": 1.1592, "step": 13735 }, { "epoch": 1.42, "learning_rate": 4.290329773688312e-05, "loss": 0.8133, "step": 13736 }, { "epoch": 1.42, "learning_rate": 4.290107558107814e-05, "loss": 0.4354, "step": 13737 }, { "epoch": 1.42, "learning_rate": 4.289885333842615e-05, "loss": 0.7853, "step": 13738 }, { "epoch": 1.42, "learning_rate": 4.289663100894208e-05, "loss": 0.3634, "step": 13739 }, { "epoch": 1.42, "learning_rate": 4.2894408592640915e-05, "loss": 0.6872, "step": 13740 }, { "epoch": 1.42, "learning_rate": 4.2892186089537614e-05, "loss": 0.2652, "step": 13741 }, { "epoch": 1.42, "learning_rate": 4.288996349964712e-05, "loss": 0.5636, "step": 13742 }, { "epoch": 1.42, "learning_rate": 4.288774082298442e-05, "loss": 0.1986, "step": 13743 }, { "epoch": 1.42, "learning_rate": 4.288551805956446e-05, "loss": 0.7175, "step": 13744 }, { "epoch": 1.42, "learning_rate": 4.288329520940221e-05, "loss": 1.1462, "step": 13745 }, { "epoch": 1.42, "learning_rate": 4.288107227251263e-05, "loss": 0.7169, "step": 13746 }, { "epoch": 1.42, "learning_rate": 4.2878849248910675e-05, "loss": 0.3316, "step": 13747 }, { "epoch": 1.42, "learning_rate": 4.287662613861133e-05, "loss": 0.826, "step": 13748 }, { "epoch": 1.42, "learning_rate": 4.287440294162954e-05, "loss": 0.5448, "step": 13749 }, { "epoch": 1.42, "learning_rate": 4.2872179657980305e-05, "loss": 1.1608, "step": 13750 }, { "epoch": 1.42, "learning_rate": 4.286995628767855e-05, "loss": 0.6763, "step": 13751 }, { "epoch": 1.42, "learning_rate": 4.286773283073927e-05, "loss": 0.5163, "step": 13752 }, { "epoch": 1.42, "learning_rate": 4.2865509287177416e-05, "loss": 0.8513, "step": 13753 }, { "epoch": 1.42, "learning_rate": 4.2863285657007955e-05, "loss": 0.1287, "step": 13754 }, { "epoch": 1.42, "learning_rate": 4.286106194024589e-05, "loss": 0.2531, "step": 13755 }, { "epoch": 1.42, "learning_rate": 4.285883813690614e-05, "loss": 1.0279, "step": 13756 }, { "epoch": 1.42, "learning_rate": 4.285661424700372e-05, "loss": 0.8692, "step": 13757 }, { "epoch": 1.42, "learning_rate": 4.285439027055357e-05, "loss": 0.8849, "step": 13758 }, { "epoch": 1.42, "learning_rate": 4.285216620757068e-05, "loss": 0.6986, "step": 13759 }, { "epoch": 1.42, "learning_rate": 4.284994205807001e-05, "loss": 0.3086, "step": 13760 }, { "epoch": 1.42, "learning_rate": 4.2847717822066536e-05, "loss": 0.1884, "step": 13761 }, { "epoch": 1.42, "learning_rate": 4.2845493499575224e-05, "loss": 1.0316, "step": 13762 }, { "epoch": 1.42, "learning_rate": 4.2843269090611064e-05, "loss": 0.4546, "step": 13763 }, { "epoch": 1.42, "learning_rate": 4.284104459518902e-05, "loss": 0.3181, "step": 13764 }, { "epoch": 1.42, "learning_rate": 4.2838820013324074e-05, "loss": 0.2086, "step": 13765 }, { "epoch": 1.42, "learning_rate": 4.2836595345031196e-05, "loss": 1.0971, "step": 13766 }, { "epoch": 1.42, "learning_rate": 4.2834370590325344e-05, "loss": 0.0883, "step": 13767 }, { "epoch": 1.42, "learning_rate": 4.283214574922152e-05, "loss": 0.7869, "step": 13768 }, { "epoch": 1.42, "learning_rate": 4.282992082173471e-05, "loss": 0.9096, "step": 13769 }, { "epoch": 1.42, "learning_rate": 4.282769580787987e-05, "loss": 0.2454, "step": 13770 }, { "epoch": 1.42, "learning_rate": 4.282547070767197e-05, "loss": 1.6009, "step": 13771 }, { "epoch": 1.42, "learning_rate": 4.2823245521126006e-05, "loss": 0.1536, "step": 13772 }, { "epoch": 1.42, "learning_rate": 4.282102024825695e-05, "loss": 0.2438, "step": 13773 }, { "epoch": 1.42, "learning_rate": 4.28187948890798e-05, "loss": 1.0354, "step": 13774 }, { "epoch": 1.42, "learning_rate": 4.2816569443609506e-05, "loss": 0.6668, "step": 13775 }, { "epoch": 1.42, "learning_rate": 4.281434391186109e-05, "loss": 0.7257, "step": 13776 }, { "epoch": 1.42, "learning_rate": 4.281211829384948e-05, "loss": 0.3665, "step": 13777 }, { "epoch": 1.42, "learning_rate": 4.280989258958969e-05, "loss": 0.6433, "step": 13778 }, { "epoch": 1.42, "learning_rate": 4.2807666799096724e-05, "loss": 0.5662, "step": 13779 }, { "epoch": 1.42, "learning_rate": 4.2805440922385525e-05, "loss": 0.5677, "step": 13780 }, { "epoch": 1.42, "learning_rate": 4.280321495947109e-05, "loss": 0.8445, "step": 13781 }, { "epoch": 1.42, "learning_rate": 4.2800988910368426e-05, "loss": 1.1668, "step": 13782 }, { "epoch": 1.42, "learning_rate": 4.279876277509248e-05, "loss": 0.7423, "step": 13783 }, { "epoch": 1.42, "learning_rate": 4.279653655365827e-05, "loss": 0.982, "step": 13784 }, { "epoch": 1.42, "learning_rate": 4.279431024608076e-05, "loss": 0.4661, "step": 13785 }, { "epoch": 1.42, "learning_rate": 4.2792083852374954e-05, "loss": 0.819, "step": 13786 }, { "epoch": 1.42, "learning_rate": 4.278985737255583e-05, "loss": 0.4629, "step": 13787 }, { "epoch": 1.42, "learning_rate": 4.2787630806638385e-05, "loss": 1.1981, "step": 13788 }, { "epoch": 1.42, "learning_rate": 4.2785404154637604e-05, "loss": 1.0771, "step": 13789 }, { "epoch": 1.42, "learning_rate": 4.278317741656846e-05, "loss": 0.5708, "step": 13790 }, { "epoch": 1.42, "learning_rate": 4.2780950592445966e-05, "loss": 1.1191, "step": 13791 }, { "epoch": 1.42, "learning_rate": 4.277872368228509e-05, "loss": 1.006, "step": 13792 }, { "epoch": 1.43, "learning_rate": 4.2776496686100857e-05, "loss": 0.6657, "step": 13793 }, { "epoch": 1.43, "learning_rate": 4.2774269603908216e-05, "loss": 1.4788, "step": 13794 }, { "epoch": 1.43, "learning_rate": 4.27720424357222e-05, "loss": 1.0318, "step": 13795 }, { "epoch": 1.43, "learning_rate": 4.276981518155778e-05, "loss": 0.399, "step": 13796 }, { "epoch": 1.43, "learning_rate": 4.276758784142994e-05, "loss": 0.7096, "step": 13797 }, { "epoch": 1.43, "learning_rate": 4.27653604153537e-05, "loss": 0.2413, "step": 13798 }, { "epoch": 1.43, "learning_rate": 4.276313290334403e-05, "loss": 0.2743, "step": 13799 }, { "epoch": 1.43, "learning_rate": 4.2760905305415936e-05, "loss": 1.0762, "step": 13800 }, { "epoch": 1.43, "learning_rate": 4.275867762158442e-05, "loss": 1.0253, "step": 13801 }, { "epoch": 1.43, "learning_rate": 4.275644985186447e-05, "loss": 0.5937, "step": 13802 }, { "epoch": 1.43, "learning_rate": 4.275422199627108e-05, "loss": 1.2109, "step": 13803 }, { "epoch": 1.43, "learning_rate": 4.275199405481926e-05, "loss": 0.8027, "step": 13804 }, { "epoch": 1.43, "learning_rate": 4.2749766027523986e-05, "loss": 0.5486, "step": 13805 }, { "epoch": 1.43, "learning_rate": 4.274753791440028e-05, "loss": 0.7379, "step": 13806 }, { "epoch": 1.43, "learning_rate": 4.274530971546313e-05, "loss": 0.3832, "step": 13807 }, { "epoch": 1.43, "learning_rate": 4.274308143072754e-05, "loss": 0.1437, "step": 13808 }, { "epoch": 1.43, "learning_rate": 4.2740853060208505e-05, "loss": 0.2435, "step": 13809 }, { "epoch": 1.43, "learning_rate": 4.2738624603921024e-05, "loss": 0.4685, "step": 13810 }, { "epoch": 1.43, "learning_rate": 4.27363960618801e-05, "loss": 0.3544, "step": 13811 }, { "epoch": 1.43, "learning_rate": 4.273416743410075e-05, "loss": 0.2096, "step": 13812 }, { "epoch": 1.43, "learning_rate": 4.2731938720597966e-05, "loss": 0.6713, "step": 13813 }, { "epoch": 1.43, "learning_rate": 4.2729709921386736e-05, "loss": 0.6941, "step": 13814 }, { "epoch": 1.43, "learning_rate": 4.272748103648208e-05, "loss": 0.2712, "step": 13815 }, { "epoch": 1.43, "learning_rate": 4.2725252065899005e-05, "loss": 0.3319, "step": 13816 }, { "epoch": 1.43, "learning_rate": 4.272302300965251e-05, "loss": 0.2104, "step": 13817 }, { "epoch": 1.43, "learning_rate": 4.2720793867757594e-05, "loss": 0.7293, "step": 13818 }, { "epoch": 1.43, "learning_rate": 4.271856464022928e-05, "loss": 0.2079, "step": 13819 }, { "epoch": 1.43, "learning_rate": 4.2716335327082564e-05, "loss": 1.212, "step": 13820 }, { "epoch": 1.43, "learning_rate": 4.271410592833244e-05, "loss": 0.5398, "step": 13821 }, { "epoch": 1.43, "learning_rate": 4.271187644399395e-05, "loss": 0.85, "step": 13822 }, { "epoch": 1.43, "learning_rate": 4.270964687408207e-05, "loss": 0.3929, "step": 13823 }, { "epoch": 1.43, "learning_rate": 4.270741721861183e-05, "loss": 0.6354, "step": 13824 }, { "epoch": 1.43, "learning_rate": 4.2705187477598216e-05, "loss": 0.8948, "step": 13825 }, { "epoch": 1.43, "learning_rate": 4.2702957651056266e-05, "loss": 0.6642, "step": 13826 }, { "epoch": 1.43, "learning_rate": 4.270072773900097e-05, "loss": 0.7526, "step": 13827 }, { "epoch": 1.43, "learning_rate": 4.269849774144735e-05, "loss": 1.1206, "step": 13828 }, { "epoch": 1.43, "learning_rate": 4.269626765841041e-05, "loss": 0.3078, "step": 13829 }, { "epoch": 1.43, "learning_rate": 4.269403748990517e-05, "loss": 0.9162, "step": 13830 }, { "epoch": 1.43, "learning_rate": 4.269180723594664e-05, "loss": 0.4659, "step": 13831 }, { "epoch": 1.43, "learning_rate": 4.268957689654984e-05, "loss": 0.8957, "step": 13832 }, { "epoch": 1.43, "learning_rate": 4.268734647172977e-05, "loss": 1.1179, "step": 13833 }, { "epoch": 1.43, "learning_rate": 4.268511596150145e-05, "loss": 0.4384, "step": 13834 }, { "epoch": 1.43, "learning_rate": 4.2682885365879896e-05, "loss": 0.7447, "step": 13835 }, { "epoch": 1.43, "learning_rate": 4.268065468488013e-05, "loss": 0.4381, "step": 13836 }, { "epoch": 1.43, "learning_rate": 4.267842391851716e-05, "loss": 1.0309, "step": 13837 }, { "epoch": 1.43, "learning_rate": 4.267619306680602e-05, "loss": 0.3778, "step": 13838 }, { "epoch": 1.43, "learning_rate": 4.26739621297617e-05, "loss": 0.2904, "step": 13839 }, { "epoch": 1.43, "learning_rate": 4.267173110739923e-05, "loss": 0.6718, "step": 13840 }, { "epoch": 1.43, "learning_rate": 4.266949999973364e-05, "loss": 0.5056, "step": 13841 }, { "epoch": 1.43, "learning_rate": 4.266726880677994e-05, "loss": 0.5362, "step": 13842 }, { "epoch": 1.43, "learning_rate": 4.266503752855315e-05, "loss": 0.7016, "step": 13843 }, { "epoch": 1.43, "learning_rate": 4.266280616506828e-05, "loss": 0.8117, "step": 13844 }, { "epoch": 1.43, "learning_rate": 4.2660574716340376e-05, "loss": 1.0022, "step": 13845 }, { "epoch": 1.43, "learning_rate": 4.265834318238445e-05, "loss": 1.2944, "step": 13846 }, { "epoch": 1.43, "learning_rate": 4.265611156321551e-05, "loss": 1.2568, "step": 13847 }, { "epoch": 1.43, "learning_rate": 4.265387985884858e-05, "loss": 0.7014, "step": 13848 }, { "epoch": 1.43, "learning_rate": 4.26516480692987e-05, "loss": 0.6782, "step": 13849 }, { "epoch": 1.43, "learning_rate": 4.2649416194580885e-05, "loss": 0.2305, "step": 13850 }, { "epoch": 1.43, "learning_rate": 4.264718423471017e-05, "loss": 0.8193, "step": 13851 }, { "epoch": 1.43, "learning_rate": 4.264495218970155e-05, "loss": 0.9231, "step": 13852 }, { "epoch": 1.43, "learning_rate": 4.2642720059570085e-05, "loss": 0.293, "step": 13853 }, { "epoch": 1.43, "learning_rate": 4.2640487844330774e-05, "loss": 0.6049, "step": 13854 }, { "epoch": 1.43, "learning_rate": 4.263825554399867e-05, "loss": 0.3566, "step": 13855 }, { "epoch": 1.43, "learning_rate": 4.263602315858877e-05, "loss": 0.2579, "step": 13856 }, { "epoch": 1.43, "learning_rate": 4.263379068811614e-05, "loss": 0.4553, "step": 13857 }, { "epoch": 1.43, "learning_rate": 4.2631558132595776e-05, "loss": 0.7961, "step": 13858 }, { "epoch": 1.43, "learning_rate": 4.2629325492042723e-05, "loss": 0.4832, "step": 13859 }, { "epoch": 1.43, "learning_rate": 4.2627092766472e-05, "loss": 1.1463, "step": 13860 }, { "epoch": 1.43, "learning_rate": 4.262485995589865e-05, "loss": 1.2299, "step": 13861 }, { "epoch": 1.43, "learning_rate": 4.262262706033768e-05, "loss": 1.2298, "step": 13862 }, { "epoch": 1.43, "learning_rate": 4.262039407980415e-05, "loss": 0.4149, "step": 13863 }, { "epoch": 1.43, "learning_rate": 4.2618161014313086e-05, "loss": 0.7215, "step": 13864 }, { "epoch": 1.43, "learning_rate": 4.261592786387951e-05, "loss": 0.394, "step": 13865 }, { "epoch": 1.43, "learning_rate": 4.2613694628518456e-05, "loss": 0.3969, "step": 13866 }, { "epoch": 1.43, "learning_rate": 4.261146130824496e-05, "loss": 1.4391, "step": 13867 }, { "epoch": 1.43, "learning_rate": 4.2609227903074056e-05, "loss": 1.4787, "step": 13868 }, { "epoch": 1.43, "learning_rate": 4.260699441302079e-05, "loss": 0.4705, "step": 13869 }, { "epoch": 1.43, "learning_rate": 4.2604760838100185e-05, "loss": 0.2949, "step": 13870 }, { "epoch": 1.43, "learning_rate": 4.2602527178327274e-05, "loss": 0.9457, "step": 13871 }, { "epoch": 1.43, "learning_rate": 4.26002934337171e-05, "loss": 0.7567, "step": 13872 }, { "epoch": 1.43, "learning_rate": 4.25980596042847e-05, "loss": 0.6026, "step": 13873 }, { "epoch": 1.43, "learning_rate": 4.259582569004512e-05, "loss": 1.3429, "step": 13874 }, { "epoch": 1.43, "learning_rate": 4.259359169101337e-05, "loss": 0.6777, "step": 13875 }, { "epoch": 1.43, "learning_rate": 4.259135760720452e-05, "loss": 0.3227, "step": 13876 }, { "epoch": 1.43, "learning_rate": 4.25891234386336e-05, "loss": 0.6291, "step": 13877 }, { "epoch": 1.43, "learning_rate": 4.258688918531563e-05, "loss": 0.7622, "step": 13878 }, { "epoch": 1.43, "learning_rate": 4.258465484726568e-05, "loss": 1.3574, "step": 13879 }, { "epoch": 1.43, "learning_rate": 4.2582420424498777e-05, "loss": 0.6178, "step": 13880 }, { "epoch": 1.43, "learning_rate": 4.2580185917029956e-05, "loss": 0.8496, "step": 13881 }, { "epoch": 1.43, "learning_rate": 4.257795132487428e-05, "loss": 0.2769, "step": 13882 }, { "epoch": 1.43, "learning_rate": 4.2575716648046765e-05, "loss": 0.1186, "step": 13883 }, { "epoch": 1.43, "learning_rate": 4.2573481886562474e-05, "loss": 1.2212, "step": 13884 }, { "epoch": 1.43, "learning_rate": 4.257124704043645e-05, "loss": 0.5547, "step": 13885 }, { "epoch": 1.43, "learning_rate": 4.2569012109683724e-05, "loss": 0.3747, "step": 13886 }, { "epoch": 1.43, "learning_rate": 4.256677709431935e-05, "loss": 0.4434, "step": 13887 }, { "epoch": 1.43, "learning_rate": 4.2564541994358385e-05, "loss": 0.3519, "step": 13888 }, { "epoch": 1.44, "learning_rate": 4.256230680981585e-05, "loss": 0.8623, "step": 13889 }, { "epoch": 1.44, "learning_rate": 4.256007154070681e-05, "loss": 0.4459, "step": 13890 }, { "epoch": 1.44, "learning_rate": 4.25578361870463e-05, "loss": 0.9924, "step": 13891 }, { "epoch": 1.44, "learning_rate": 4.2555600748849384e-05, "loss": 0.375, "step": 13892 }, { "epoch": 1.44, "learning_rate": 4.25533652261311e-05, "loss": 0.3257, "step": 13893 }, { "epoch": 1.44, "learning_rate": 4.2551129618906496e-05, "loss": 0.4661, "step": 13894 }, { "epoch": 1.44, "learning_rate": 4.2548893927190625e-05, "loss": 0.311, "step": 13895 }, { "epoch": 1.44, "learning_rate": 4.254665815099854e-05, "loss": 0.4912, "step": 13896 }, { "epoch": 1.44, "learning_rate": 4.254442229034528e-05, "loss": 0.4765, "step": 13897 }, { "epoch": 1.44, "learning_rate": 4.254218634524592e-05, "loss": 1.2119, "step": 13898 }, { "epoch": 1.44, "learning_rate": 4.253995031571548e-05, "loss": 0.5577, "step": 13899 }, { "epoch": 1.44, "learning_rate": 4.253771420176903e-05, "loss": 0.2131, "step": 13900 }, { "epoch": 1.44, "learning_rate": 4.253547800342163e-05, "loss": 0.9313, "step": 13901 }, { "epoch": 1.44, "learning_rate": 4.253324172068832e-05, "loss": 0.8442, "step": 13902 }, { "epoch": 1.44, "learning_rate": 4.253100535358416e-05, "loss": 0.955, "step": 13903 }, { "epoch": 1.44, "learning_rate": 4.2528768902124205e-05, "loss": 0.7776, "step": 13904 }, { "epoch": 1.44, "learning_rate": 4.25265323663235e-05, "loss": 0.5829, "step": 13905 }, { "epoch": 1.44, "learning_rate": 4.252429574619712e-05, "loss": 0.3297, "step": 13906 }, { "epoch": 1.44, "learning_rate": 4.252205904176011e-05, "loss": 0.7464, "step": 13907 }, { "epoch": 1.44, "learning_rate": 4.251982225302753e-05, "loss": 0.7176, "step": 13908 }, { "epoch": 1.44, "learning_rate": 4.2517585380014433e-05, "loss": 0.425, "step": 13909 }, { "epoch": 1.44, "learning_rate": 4.251534842273588e-05, "loss": 0.0705, "step": 13910 }, { "epoch": 1.44, "learning_rate": 4.2513111381206935e-05, "loss": 0.5587, "step": 13911 }, { "epoch": 1.44, "learning_rate": 4.251087425544265e-05, "loss": 1.1258, "step": 13912 }, { "epoch": 1.44, "learning_rate": 4.2508637045458086e-05, "loss": 0.6516, "step": 13913 }, { "epoch": 1.44, "learning_rate": 4.2506399751268304e-05, "loss": 0.7875, "step": 13914 }, { "epoch": 1.44, "learning_rate": 4.250416237288837e-05, "loss": 0.4152, "step": 13915 }, { "epoch": 1.44, "learning_rate": 4.2501924910333325e-05, "loss": 0.3725, "step": 13916 }, { "epoch": 1.44, "learning_rate": 4.249968736361826e-05, "loss": 0.7688, "step": 13917 }, { "epoch": 1.44, "learning_rate": 4.249744973275822e-05, "loss": 0.313, "step": 13918 }, { "epoch": 1.44, "learning_rate": 4.249521201776827e-05, "loss": 0.7116, "step": 13919 }, { "epoch": 1.44, "learning_rate": 4.249297421866348e-05, "loss": 0.9639, "step": 13920 }, { "epoch": 1.44, "learning_rate": 4.249073633545891e-05, "loss": 0.4853, "step": 13921 }, { "epoch": 1.44, "learning_rate": 4.248849836816964e-05, "loss": 0.242, "step": 13922 }, { "epoch": 1.44, "learning_rate": 4.2486260316810696e-05, "loss": 0.0567, "step": 13923 }, { "epoch": 1.44, "learning_rate": 4.248402218139718e-05, "loss": 1.2767, "step": 13924 }, { "epoch": 1.44, "learning_rate": 4.2481783961944145e-05, "loss": 0.5387, "step": 13925 }, { "epoch": 1.44, "learning_rate": 4.247954565846667e-05, "loss": 0.4997, "step": 13926 }, { "epoch": 1.44, "learning_rate": 4.247730727097981e-05, "loss": 0.7129, "step": 13927 }, { "epoch": 1.44, "learning_rate": 4.247506879949863e-05, "loss": 0.6685, "step": 13928 }, { "epoch": 1.44, "learning_rate": 4.2472830244038215e-05, "loss": 0.278, "step": 13929 }, { "epoch": 1.44, "learning_rate": 4.247059160461361e-05, "loss": 0.351, "step": 13930 }, { "epoch": 1.44, "learning_rate": 4.2468352881239916e-05, "loss": 0.5447, "step": 13931 }, { "epoch": 1.44, "learning_rate": 4.246611407393218e-05, "loss": 0.7257, "step": 13932 }, { "epoch": 1.44, "learning_rate": 4.2463875182705485e-05, "loss": 0.4229, "step": 13933 }, { "epoch": 1.44, "learning_rate": 4.2461636207574896e-05, "loss": 0.5485, "step": 13934 }, { "epoch": 1.44, "learning_rate": 4.245939714855549e-05, "loss": 0.8815, "step": 13935 }, { "epoch": 1.44, "learning_rate": 4.245715800566234e-05, "loss": 0.6131, "step": 13936 }, { "epoch": 1.44, "learning_rate": 4.2454918778910514e-05, "loss": 0.1475, "step": 13937 }, { "epoch": 1.44, "learning_rate": 4.245267946831508e-05, "loss": 0.2543, "step": 13938 }, { "epoch": 1.44, "learning_rate": 4.2450440073891135e-05, "loss": 1.239, "step": 13939 }, { "epoch": 1.44, "learning_rate": 4.244820059565374e-05, "loss": 0.19, "step": 13940 }, { "epoch": 1.44, "learning_rate": 4.244596103361796e-05, "loss": 0.5478, "step": 13941 }, { "epoch": 1.44, "learning_rate": 4.2443721387798896e-05, "loss": 0.3403, "step": 13942 }, { "epoch": 1.44, "learning_rate": 4.2441481658211595e-05, "loss": 0.8406, "step": 13943 }, { "epoch": 1.44, "learning_rate": 4.243924184487116e-05, "loss": 0.4421, "step": 13944 }, { "epoch": 1.44, "learning_rate": 4.243700194779266e-05, "loss": 1.0195, "step": 13945 }, { "epoch": 1.44, "learning_rate": 4.243476196699118e-05, "loss": 0.3161, "step": 13946 }, { "epoch": 1.44, "learning_rate": 4.2432521902481785e-05, "loss": 1.0053, "step": 13947 }, { "epoch": 1.44, "learning_rate": 4.243028175427956e-05, "loss": 0.7695, "step": 13948 }, { "epoch": 1.44, "learning_rate": 4.242804152239958e-05, "loss": 0.923, "step": 13949 }, { "epoch": 1.44, "learning_rate": 4.2425801206856944e-05, "loss": 0.7744, "step": 13950 }, { "epoch": 1.44, "learning_rate": 4.2423560807666714e-05, "loss": 0.8353, "step": 13951 }, { "epoch": 1.44, "learning_rate": 4.242132032484399e-05, "loss": 0.9017, "step": 13952 }, { "epoch": 1.44, "learning_rate": 4.2419079758403835e-05, "loss": 1.177, "step": 13953 }, { "epoch": 1.44, "learning_rate": 4.241683910836134e-05, "loss": 0.3292, "step": 13954 }, { "epoch": 1.44, "learning_rate": 4.24145983747316e-05, "loss": 0.1953, "step": 13955 }, { "epoch": 1.44, "learning_rate": 4.241235755752968e-05, "loss": 0.2593, "step": 13956 }, { "epoch": 1.44, "learning_rate": 4.2410116656770665e-05, "loss": 0.4918, "step": 13957 }, { "epoch": 1.44, "learning_rate": 4.2407875672469665e-05, "loss": 0.3761, "step": 13958 }, { "epoch": 1.44, "learning_rate": 4.2405634604641736e-05, "loss": 0.499, "step": 13959 }, { "epoch": 1.44, "learning_rate": 4.240339345330199e-05, "loss": 0.3979, "step": 13960 }, { "epoch": 1.44, "learning_rate": 4.2401152218465496e-05, "loss": 0.2801, "step": 13961 }, { "epoch": 1.44, "learning_rate": 4.239891090014734e-05, "loss": 0.1878, "step": 13962 }, { "epoch": 1.44, "learning_rate": 4.239666949836262e-05, "loss": 0.3413, "step": 13963 }, { "epoch": 1.44, "learning_rate": 4.239442801312643e-05, "loss": 0.7801, "step": 13964 }, { "epoch": 1.44, "learning_rate": 4.2392186444453846e-05, "loss": 1.0299, "step": 13965 }, { "epoch": 1.44, "learning_rate": 4.238994479235996e-05, "loss": 0.9357, "step": 13966 }, { "epoch": 1.44, "learning_rate": 4.238770305685986e-05, "loss": 0.4092, "step": 13967 }, { "epoch": 1.44, "learning_rate": 4.238546123796865e-05, "loss": 0.4614, "step": 13968 }, { "epoch": 1.44, "learning_rate": 4.238321933570141e-05, "loss": 0.9916, "step": 13969 }, { "epoch": 1.44, "learning_rate": 4.2380977350073245e-05, "loss": 1.4918, "step": 13970 }, { "epoch": 1.44, "learning_rate": 4.237873528109922e-05, "loss": 0.4471, "step": 13971 }, { "epoch": 1.44, "learning_rate": 4.237649312879445e-05, "loss": 0.5721, "step": 13972 }, { "epoch": 1.44, "learning_rate": 4.2374250893174025e-05, "loss": 0.699, "step": 13973 }, { "epoch": 1.44, "learning_rate": 4.237200857425304e-05, "loss": 0.4105, "step": 13974 }, { "epoch": 1.44, "learning_rate": 4.236976617204659e-05, "loss": 0.1721, "step": 13975 }, { "epoch": 1.44, "learning_rate": 4.2367523686569765e-05, "loss": 0.1071, "step": 13976 }, { "epoch": 1.44, "learning_rate": 4.2365281117837666e-05, "loss": 0.5387, "step": 13977 }, { "epoch": 1.44, "learning_rate": 4.236303846586539e-05, "loss": 0.3219, "step": 13978 }, { "epoch": 1.44, "learning_rate": 4.236079573066804e-05, "loss": 0.7073, "step": 13979 }, { "epoch": 1.44, "learning_rate": 4.2358552912260684e-05, "loss": 1.0621, "step": 13980 }, { "epoch": 1.44, "learning_rate": 4.235631001065845e-05, "loss": 0.5803, "step": 13981 }, { "epoch": 1.44, "learning_rate": 4.235406702587643e-05, "loss": 0.494, "step": 13982 }, { "epoch": 1.44, "learning_rate": 4.2351823957929726e-05, "loss": 1.0852, "step": 13983 }, { "epoch": 1.45, "learning_rate": 4.234958080683343e-05, "loss": 0.9305, "step": 13984 }, { "epoch": 1.45, "learning_rate": 4.234733757260264e-05, "loss": 0.72, "step": 13985 }, { "epoch": 1.45, "learning_rate": 4.234509425525247e-05, "loss": 0.2281, "step": 13986 }, { "epoch": 1.45, "learning_rate": 4.234285085479801e-05, "loss": 0.8381, "step": 13987 }, { "epoch": 1.45, "learning_rate": 4.234060737125437e-05, "loss": 0.6342, "step": 13988 }, { "epoch": 1.45, "learning_rate": 4.233836380463665e-05, "loss": 0.7256, "step": 13989 }, { "epoch": 1.45, "learning_rate": 4.233612015495995e-05, "loss": 0.1867, "step": 13990 }, { "epoch": 1.45, "learning_rate": 4.233387642223937e-05, "loss": 1.2945, "step": 13991 }, { "epoch": 1.45, "learning_rate": 4.233163260649002e-05, "loss": 1.2861, "step": 13992 }, { "epoch": 1.45, "learning_rate": 4.232938870772702e-05, "loss": 0.7521, "step": 13993 }, { "epoch": 1.45, "learning_rate": 4.2327144725965434e-05, "loss": 0.328, "step": 13994 }, { "epoch": 1.45, "learning_rate": 4.2324900661220406e-05, "loss": 0.2057, "step": 13995 }, { "epoch": 1.45, "learning_rate": 4.232265651350704e-05, "loss": 0.3429, "step": 13996 }, { "epoch": 1.45, "learning_rate": 4.2320412282840424e-05, "loss": 0.561, "step": 13997 }, { "epoch": 1.45, "learning_rate": 4.2318167969235685e-05, "loss": 0.2758, "step": 13998 }, { "epoch": 1.45, "learning_rate": 4.2315923572707906e-05, "loss": 0.5434, "step": 13999 }, { "epoch": 1.45, "learning_rate": 4.231367909327222e-05, "loss": 0.421, "step": 14000 }, { "epoch": 1.45, "learning_rate": 4.231143453094373e-05, "loss": 0.2888, "step": 14001 }, { "epoch": 1.45, "learning_rate": 4.2309189885737536e-05, "loss": 0.6687, "step": 14002 }, { "epoch": 1.45, "learning_rate": 4.230694515766876e-05, "loss": 0.1732, "step": 14003 }, { "epoch": 1.45, "learning_rate": 4.230470034675251e-05, "loss": 0.5334, "step": 14004 }, { "epoch": 1.45, "learning_rate": 4.2302455453003887e-05, "loss": 0.4175, "step": 14005 }, { "epoch": 1.45, "learning_rate": 4.230021047643802e-05, "loss": 1.1331, "step": 14006 }, { "epoch": 1.45, "learning_rate": 4.229796541707002e-05, "loss": 0.6509, "step": 14007 }, { "epoch": 1.45, "learning_rate": 4.229572027491499e-05, "loss": 0.5187, "step": 14008 }, { "epoch": 1.45, "learning_rate": 4.229347504998805e-05, "loss": 0.4861, "step": 14009 }, { "epoch": 1.45, "learning_rate": 4.229122974230431e-05, "loss": 0.3441, "step": 14010 }, { "epoch": 1.45, "learning_rate": 4.228898435187888e-05, "loss": 0.5545, "step": 14011 }, { "epoch": 1.45, "learning_rate": 4.228673887872689e-05, "loss": 0.5214, "step": 14012 }, { "epoch": 1.45, "learning_rate": 4.2284493322863446e-05, "loss": 0.6017, "step": 14013 }, { "epoch": 1.45, "learning_rate": 4.2282247684303676e-05, "loss": 0.862, "step": 14014 }, { "epoch": 1.45, "learning_rate": 4.228000196306268e-05, "loss": 0.7644, "step": 14015 }, { "epoch": 1.45, "learning_rate": 4.227775615915558e-05, "loss": 0.429, "step": 14016 }, { "epoch": 1.45, "learning_rate": 4.227551027259751e-05, "loss": 0.5516, "step": 14017 }, { "epoch": 1.45, "learning_rate": 4.227326430340357e-05, "loss": 0.6753, "step": 14018 }, { "epoch": 1.45, "learning_rate": 4.227101825158888e-05, "loss": 0.442, "step": 14019 }, { "epoch": 1.45, "learning_rate": 4.226877211716858e-05, "loss": 1.0436, "step": 14020 }, { "epoch": 1.45, "learning_rate": 4.226652590015777e-05, "loss": 0.6411, "step": 14021 }, { "epoch": 1.45, "learning_rate": 4.2264279600571585e-05, "loss": 0.3491, "step": 14022 }, { "epoch": 1.45, "learning_rate": 4.2262033218425125e-05, "loss": 1.0152, "step": 14023 }, { "epoch": 1.45, "learning_rate": 4.2259786753733535e-05, "loss": 1.22, "step": 14024 }, { "epoch": 1.45, "learning_rate": 4.225754020651193e-05, "loss": 0.9344, "step": 14025 }, { "epoch": 1.45, "learning_rate": 4.225529357677543e-05, "loss": 0.2597, "step": 14026 }, { "epoch": 1.45, "learning_rate": 4.225304686453917e-05, "loss": 1.0514, "step": 14027 }, { "epoch": 1.45, "learning_rate": 4.2250800069818255e-05, "loss": 0.7578, "step": 14028 }, { "epoch": 1.45, "learning_rate": 4.2248553192627813e-05, "loss": 0.1556, "step": 14029 }, { "epoch": 1.45, "learning_rate": 4.224630623298299e-05, "loss": 0.8203, "step": 14030 }, { "epoch": 1.45, "learning_rate": 4.22440591908989e-05, "loss": 0.228, "step": 14031 }, { "epoch": 1.45, "learning_rate": 4.224181206639066e-05, "loss": 0.553, "step": 14032 }, { "epoch": 1.45, "learning_rate": 4.223956485947341e-05, "loss": 0.1801, "step": 14033 }, { "epoch": 1.45, "learning_rate": 4.2237317570162274e-05, "loss": 0.6514, "step": 14034 }, { "epoch": 1.45, "learning_rate": 4.223507019847237e-05, "loss": 0.6146, "step": 14035 }, { "epoch": 1.45, "learning_rate": 4.2232822744418856e-05, "loss": 1.3827, "step": 14036 }, { "epoch": 1.45, "learning_rate": 4.2230575208016825e-05, "loss": 0.9967, "step": 14037 }, { "epoch": 1.45, "learning_rate": 4.222832758928143e-05, "loss": 0.3213, "step": 14038 }, { "epoch": 1.45, "learning_rate": 4.222607988822779e-05, "loss": 0.7754, "step": 14039 }, { "epoch": 1.45, "learning_rate": 4.222383210487105e-05, "loss": 0.9125, "step": 14040 }, { "epoch": 1.45, "learning_rate": 4.222158423922633e-05, "loss": 0.8905, "step": 14041 }, { "epoch": 1.45, "learning_rate": 4.2219336291308755e-05, "loss": 0.7678, "step": 14042 }, { "epoch": 1.45, "learning_rate": 4.221708826113347e-05, "loss": 1.171, "step": 14043 }, { "epoch": 1.45, "learning_rate": 4.2214840148715605e-05, "loss": 1.0819, "step": 14044 }, { "epoch": 1.45, "learning_rate": 4.2212591954070304e-05, "loss": 0.6458, "step": 14045 }, { "epoch": 1.45, "learning_rate": 4.221034367721269e-05, "loss": 1.1778, "step": 14046 }, { "epoch": 1.45, "learning_rate": 4.220809531815789e-05, "loss": 0.3168, "step": 14047 }, { "epoch": 1.45, "learning_rate": 4.2205846876921055e-05, "loss": 1.0807, "step": 14048 }, { "epoch": 1.45, "learning_rate": 4.22035983535173e-05, "loss": 1.0717, "step": 14049 }, { "epoch": 1.45, "learning_rate": 4.2201349747961805e-05, "loss": 0.3078, "step": 14050 }, { "epoch": 1.45, "learning_rate": 4.219910106026966e-05, "loss": 0.8288, "step": 14051 }, { "epoch": 1.45, "learning_rate": 4.219685229045602e-05, "loss": 0.4781, "step": 14052 }, { "epoch": 1.45, "learning_rate": 4.219460343853603e-05, "loss": 1.0688, "step": 14053 }, { "epoch": 1.45, "learning_rate": 4.219235450452482e-05, "loss": 0.6117, "step": 14054 }, { "epoch": 1.45, "learning_rate": 4.219010548843754e-05, "loss": 0.9703, "step": 14055 }, { "epoch": 1.45, "learning_rate": 4.2187856390289305e-05, "loss": 0.2661, "step": 14056 }, { "epoch": 1.45, "learning_rate": 4.2185607210095285e-05, "loss": 1.1144, "step": 14057 }, { "epoch": 1.45, "learning_rate": 4.218335794787061e-05, "loss": 1.0568, "step": 14058 }, { "epoch": 1.45, "learning_rate": 4.218110860363042e-05, "loss": 0.6423, "step": 14059 }, { "epoch": 1.45, "learning_rate": 4.217885917738985e-05, "loss": 0.3445, "step": 14060 }, { "epoch": 1.45, "learning_rate": 4.2176609669164055e-05, "loss": 0.8173, "step": 14061 }, { "epoch": 1.45, "learning_rate": 4.217436007896816e-05, "loss": 0.7827, "step": 14062 }, { "epoch": 1.45, "learning_rate": 4.217211040681734e-05, "loss": 0.4598, "step": 14063 }, { "epoch": 1.45, "learning_rate": 4.2169860652726715e-05, "loss": 0.2995, "step": 14064 }, { "epoch": 1.45, "learning_rate": 4.216761081671143e-05, "loss": 0.7881, "step": 14065 }, { "epoch": 1.45, "learning_rate": 4.2165360898786645e-05, "loss": 0.2647, "step": 14066 }, { "epoch": 1.45, "learning_rate": 4.216311089896748e-05, "loss": 0.7331, "step": 14067 }, { "epoch": 1.45, "learning_rate": 4.216086081726911e-05, "loss": 0.2512, "step": 14068 }, { "epoch": 1.45, "learning_rate": 4.2158610653706676e-05, "loss": 0.9536, "step": 14069 }, { "epoch": 1.45, "learning_rate": 4.215636040829531e-05, "loss": 0.1476, "step": 14070 }, { "epoch": 1.45, "learning_rate": 4.2154110081050175e-05, "loss": 0.3803, "step": 14071 }, { "epoch": 1.45, "learning_rate": 4.215185967198641e-05, "loss": 0.425, "step": 14072 }, { "epoch": 1.45, "learning_rate": 4.2149609181119176e-05, "loss": 0.4958, "step": 14073 }, { "epoch": 1.45, "learning_rate": 4.214735860846362e-05, "loss": 0.2701, "step": 14074 }, { "epoch": 1.45, "learning_rate": 4.2145107954034875e-05, "loss": 0.3361, "step": 14075 }, { "epoch": 1.45, "learning_rate": 4.214285721784811e-05, "loss": 0.8944, "step": 14076 }, { "epoch": 1.45, "learning_rate": 4.214060639991848e-05, "loss": 0.4482, "step": 14077 }, { "epoch": 1.45, "learning_rate": 4.2138355500261114e-05, "loss": 1.135, "step": 14078 }, { "epoch": 1.45, "learning_rate": 4.213610451889119e-05, "loss": 1.6124, "step": 14079 }, { "epoch": 1.46, "learning_rate": 4.2133853455823846e-05, "loss": 0.9066, "step": 14080 }, { "epoch": 1.46, "learning_rate": 4.213160231107423e-05, "loss": 0.8816, "step": 14081 }, { "epoch": 1.46, "learning_rate": 4.212935108465752e-05, "loss": 0.5318, "step": 14082 }, { "epoch": 1.46, "learning_rate": 4.212709977658885e-05, "loss": 1.5733, "step": 14083 }, { "epoch": 1.46, "learning_rate": 4.212484838688339e-05, "loss": 0.6942, "step": 14084 }, { "epoch": 1.46, "learning_rate": 4.212259691555628e-05, "loss": 0.8588, "step": 14085 }, { "epoch": 1.46, "learning_rate": 4.2120345362622686e-05, "loss": 1.3227, "step": 14086 }, { "epoch": 1.46, "learning_rate": 4.211809372809776e-05, "loss": 0.4088, "step": 14087 }, { "epoch": 1.46, "learning_rate": 4.2115842011996666e-05, "loss": 0.6771, "step": 14088 }, { "epoch": 1.46, "learning_rate": 4.2113590214334554e-05, "loss": 1.1087, "step": 14089 }, { "epoch": 1.46, "learning_rate": 4.211133833512659e-05, "loss": 0.4289, "step": 14090 }, { "epoch": 1.46, "learning_rate": 4.210908637438793e-05, "loss": 0.551, "step": 14091 }, { "epoch": 1.46, "learning_rate": 4.210683433213373e-05, "loss": 0.3465, "step": 14092 }, { "epoch": 1.46, "learning_rate": 4.210458220837916e-05, "loss": 1.009, "step": 14093 }, { "epoch": 1.46, "learning_rate": 4.2102330003139376e-05, "loss": 0.521, "step": 14094 }, { "epoch": 1.46, "learning_rate": 4.210007771642953e-05, "loss": 0.2759, "step": 14095 }, { "epoch": 1.46, "learning_rate": 4.20978253482648e-05, "loss": 1.3865, "step": 14096 }, { "epoch": 1.46, "learning_rate": 4.209557289866033e-05, "loss": 1.3163, "step": 14097 }, { "epoch": 1.46, "learning_rate": 4.2093320367631314e-05, "loss": 1.4155, "step": 14098 }, { "epoch": 1.46, "learning_rate": 4.2091067755192875e-05, "loss": 1.0519, "step": 14099 }, { "epoch": 1.46, "learning_rate": 4.208881506136021e-05, "loss": 0.529, "step": 14100 }, { "epoch": 1.46, "learning_rate": 4.208656228614846e-05, "loss": 0.4252, "step": 14101 }, { "epoch": 1.46, "learning_rate": 4.2084309429572804e-05, "loss": 0.381, "step": 14102 }, { "epoch": 1.46, "learning_rate": 4.208205649164841e-05, "loss": 0.8797, "step": 14103 }, { "epoch": 1.46, "learning_rate": 4.207980347239043e-05, "loss": 0.2321, "step": 14104 }, { "epoch": 1.46, "learning_rate": 4.207755037181404e-05, "loss": 0.2165, "step": 14105 }, { "epoch": 1.46, "learning_rate": 4.207529718993441e-05, "loss": 0.3275, "step": 14106 }, { "epoch": 1.46, "learning_rate": 4.2073043926766706e-05, "loss": 0.7376, "step": 14107 }, { "epoch": 1.46, "learning_rate": 4.2070790582326095e-05, "loss": 0.8101, "step": 14108 }, { "epoch": 1.46, "learning_rate": 4.206853715662775e-05, "loss": 1.4551, "step": 14109 }, { "epoch": 1.46, "learning_rate": 4.206628364968683e-05, "loss": 0.4533, "step": 14110 }, { "epoch": 1.46, "learning_rate": 4.206403006151851e-05, "loss": 0.6002, "step": 14111 }, { "epoch": 1.46, "learning_rate": 4.2061776392137974e-05, "loss": 0.4161, "step": 14112 }, { "epoch": 1.46, "learning_rate": 4.2059522641560365e-05, "loss": 1.0656, "step": 14113 }, { "epoch": 1.46, "learning_rate": 4.205726880980088e-05, "loss": 0.5378, "step": 14114 }, { "epoch": 1.46, "learning_rate": 4.2055014896874685e-05, "loss": 0.3097, "step": 14115 }, { "epoch": 1.46, "learning_rate": 4.205276090279695e-05, "loss": 0.4509, "step": 14116 }, { "epoch": 1.46, "learning_rate": 4.205050682758285e-05, "loss": 0.7587, "step": 14117 }, { "epoch": 1.46, "learning_rate": 4.204825267124755e-05, "loss": 0.9144, "step": 14118 }, { "epoch": 1.46, "learning_rate": 4.2045998433806235e-05, "loss": 0.7791, "step": 14119 }, { "epoch": 1.46, "learning_rate": 4.204374411527408e-05, "loss": 1.0192, "step": 14120 }, { "epoch": 1.46, "learning_rate": 4.204148971566625e-05, "loss": 0.416, "step": 14121 }, { "epoch": 1.46, "learning_rate": 4.203923523499794e-05, "loss": 0.8186, "step": 14122 }, { "epoch": 1.46, "learning_rate": 4.2036980673284314e-05, "loss": 1.1651, "step": 14123 }, { "epoch": 1.46, "learning_rate": 4.203472603054054e-05, "loss": 0.4323, "step": 14124 }, { "epoch": 1.46, "learning_rate": 4.203247130678181e-05, "loss": 0.8433, "step": 14125 }, { "epoch": 1.46, "learning_rate": 4.20302165020233e-05, "loss": 0.5159, "step": 14126 }, { "epoch": 1.46, "learning_rate": 4.20279616162802e-05, "loss": 0.2954, "step": 14127 }, { "epoch": 1.46, "learning_rate": 4.202570664956766e-05, "loss": 0.2875, "step": 14128 }, { "epoch": 1.46, "learning_rate": 4.202345160190088e-05, "loss": 0.499, "step": 14129 }, { "epoch": 1.46, "learning_rate": 4.2021196473295034e-05, "loss": 0.7262, "step": 14130 }, { "epoch": 1.46, "learning_rate": 4.201894126376532e-05, "loss": 0.8738, "step": 14131 }, { "epoch": 1.46, "learning_rate": 4.20166859733269e-05, "loss": 1.1298, "step": 14132 }, { "epoch": 1.46, "learning_rate": 4.201443060199495e-05, "loss": 0.6282, "step": 14133 }, { "epoch": 1.46, "learning_rate": 4.201217514978468e-05, "loss": 0.6222, "step": 14134 }, { "epoch": 1.46, "learning_rate": 4.200991961671124e-05, "loss": 0.5567, "step": 14135 }, { "epoch": 1.46, "learning_rate": 4.200766400278985e-05, "loss": 1.2845, "step": 14136 }, { "epoch": 1.46, "learning_rate": 4.2005408308035665e-05, "loss": 1.001, "step": 14137 }, { "epoch": 1.46, "learning_rate": 4.2003152532463884e-05, "loss": 0.6835, "step": 14138 }, { "epoch": 1.46, "learning_rate": 4.2000896676089686e-05, "loss": 0.1989, "step": 14139 }, { "epoch": 1.46, "learning_rate": 4.199864073892826e-05, "loss": 0.2965, "step": 14140 }, { "epoch": 1.46, "learning_rate": 4.19963847209948e-05, "loss": 0.1276, "step": 14141 }, { "epoch": 1.46, "learning_rate": 4.1994128622304476e-05, "loss": 0.5284, "step": 14142 }, { "epoch": 1.46, "learning_rate": 4.199187244287249e-05, "loss": 0.3786, "step": 14143 }, { "epoch": 1.46, "learning_rate": 4.198961618271401e-05, "loss": 0.1409, "step": 14144 }, { "epoch": 1.46, "learning_rate": 4.198735984184425e-05, "loss": 0.5528, "step": 14145 }, { "epoch": 1.46, "learning_rate": 4.1985103420278396e-05, "loss": 0.3415, "step": 14146 }, { "epoch": 1.46, "learning_rate": 4.198284691803162e-05, "loss": 0.1992, "step": 14147 }, { "epoch": 1.46, "learning_rate": 4.198059033511912e-05, "loss": 0.2312, "step": 14148 }, { "epoch": 1.46, "learning_rate": 4.1978333671556094e-05, "loss": 0.4151, "step": 14149 }, { "epoch": 1.46, "learning_rate": 4.197607692735773e-05, "loss": 0.3288, "step": 14150 }, { "epoch": 1.46, "learning_rate": 4.197382010253922e-05, "loss": 1.0237, "step": 14151 }, { "epoch": 1.46, "learning_rate": 4.197156319711574e-05, "loss": 0.3393, "step": 14152 }, { "epoch": 1.46, "learning_rate": 4.196930621110252e-05, "loss": 0.3246, "step": 14153 }, { "epoch": 1.46, "learning_rate": 4.196704914451472e-05, "loss": 0.3585, "step": 14154 }, { "epoch": 1.46, "learning_rate": 4.1964791997367543e-05, "loss": 0.4338, "step": 14155 }, { "epoch": 1.46, "learning_rate": 4.196253476967619e-05, "loss": 0.5097, "step": 14156 }, { "epoch": 1.46, "learning_rate": 4.196027746145585e-05, "loss": 0.5547, "step": 14157 }, { "epoch": 1.46, "learning_rate": 4.195802007272172e-05, "loss": 0.8225, "step": 14158 }, { "epoch": 1.46, "learning_rate": 4.195576260348899e-05, "loss": 0.7151, "step": 14159 }, { "epoch": 1.46, "learning_rate": 4.195350505377288e-05, "loss": 0.7118, "step": 14160 }, { "epoch": 1.46, "learning_rate": 4.195124742358856e-05, "loss": 1.0712, "step": 14161 }, { "epoch": 1.46, "learning_rate": 4.194898971295124e-05, "loss": 0.7638, "step": 14162 }, { "epoch": 1.46, "learning_rate": 4.1946731921876115e-05, "loss": 1.0652, "step": 14163 }, { "epoch": 1.46, "learning_rate": 4.1944474050378396e-05, "loss": 0.31, "step": 14164 }, { "epoch": 1.46, "learning_rate": 4.194221609847327e-05, "loss": 0.8384, "step": 14165 }, { "epoch": 1.46, "learning_rate": 4.193995806617593e-05, "loss": 0.9909, "step": 14166 }, { "epoch": 1.46, "learning_rate": 4.193769995350159e-05, "loss": 0.4799, "step": 14167 }, { "epoch": 1.46, "learning_rate": 4.1935441760465445e-05, "loss": 1.0458, "step": 14168 }, { "epoch": 1.46, "learning_rate": 4.1933183487082705e-05, "loss": 0.5438, "step": 14169 }, { "epoch": 1.46, "learning_rate": 4.193092513336856e-05, "loss": 0.8159, "step": 14170 }, { "epoch": 1.46, "learning_rate": 4.192866669933822e-05, "loss": 0.2769, "step": 14171 }, { "epoch": 1.46, "learning_rate": 4.19264081850069e-05, "loss": 0.3353, "step": 14172 }, { "epoch": 1.46, "learning_rate": 4.192414959038977e-05, "loss": 1.1509, "step": 14173 }, { "epoch": 1.46, "learning_rate": 4.1921890915502065e-05, "loss": 0.2521, "step": 14174 }, { "epoch": 1.46, "learning_rate": 4.191963216035898e-05, "loss": 0.918, "step": 14175 }, { "epoch": 1.47, "learning_rate": 4.1917373324975716e-05, "loss": 0.9909, "step": 14176 }, { "epoch": 1.47, "learning_rate": 4.191511440936748e-05, "loss": 0.3799, "step": 14177 }, { "epoch": 1.47, "learning_rate": 4.19128554135495e-05, "loss": 0.3988, "step": 14178 }, { "epoch": 1.47, "learning_rate": 4.1910596337536945e-05, "loss": 0.4258, "step": 14179 }, { "epoch": 1.47, "learning_rate": 4.190833718134505e-05, "loss": 0.2183, "step": 14180 }, { "epoch": 1.47, "learning_rate": 4.190607794498902e-05, "loss": 0.7679, "step": 14181 }, { "epoch": 1.47, "learning_rate": 4.190381862848404e-05, "loss": 0.5352, "step": 14182 }, { "epoch": 1.47, "learning_rate": 4.190155923184536e-05, "loss": 1.1198, "step": 14183 }, { "epoch": 1.47, "learning_rate": 4.189929975508817e-05, "loss": 0.6291, "step": 14184 }, { "epoch": 1.47, "learning_rate": 4.189704019822766e-05, "loss": 0.9268, "step": 14185 }, { "epoch": 1.47, "learning_rate": 4.189478056127907e-05, "loss": 0.6438, "step": 14186 }, { "epoch": 1.47, "learning_rate": 4.1892520844257594e-05, "loss": 1.0201, "step": 14187 }, { "epoch": 1.47, "learning_rate": 4.1890261047178456e-05, "loss": 0.2741, "step": 14188 }, { "epoch": 1.47, "learning_rate": 4.188800117005686e-05, "loss": 0.9381, "step": 14189 }, { "epoch": 1.47, "learning_rate": 4.188574121290803e-05, "loss": 0.3182, "step": 14190 }, { "epoch": 1.47, "learning_rate": 4.188348117574717e-05, "loss": 0.1998, "step": 14191 }, { "epoch": 1.47, "learning_rate": 4.188122105858948e-05, "loss": 0.858, "step": 14192 }, { "epoch": 1.47, "learning_rate": 4.18789608614502e-05, "loss": 0.6611, "step": 14193 }, { "epoch": 1.47, "learning_rate": 4.1876700584344535e-05, "loss": 0.9202, "step": 14194 }, { "epoch": 1.47, "learning_rate": 4.1874440227287694e-05, "loss": 0.2544, "step": 14195 }, { "epoch": 1.47, "learning_rate": 4.1872179790294904e-05, "loss": 0.9872, "step": 14196 }, { "epoch": 1.47, "learning_rate": 4.1869919273381384e-05, "loss": 0.7068, "step": 14197 }, { "epoch": 1.47, "learning_rate": 4.1867658676562346e-05, "loss": 1.2679, "step": 14198 }, { "epoch": 1.47, "learning_rate": 4.1865397999853004e-05, "loss": 0.3629, "step": 14199 }, { "epoch": 1.47, "learning_rate": 4.186313724326857e-05, "loss": 0.9479, "step": 14200 }, { "epoch": 1.47, "learning_rate": 4.186087640682427e-05, "loss": 0.6449, "step": 14201 }, { "epoch": 1.47, "learning_rate": 4.185861549053534e-05, "loss": 0.4059, "step": 14202 }, { "epoch": 1.47, "learning_rate": 4.185635449441698e-05, "loss": 0.7481, "step": 14203 }, { "epoch": 1.47, "learning_rate": 4.185409341848441e-05, "loss": 0.3852, "step": 14204 }, { "epoch": 1.47, "learning_rate": 4.185183226275286e-05, "loss": 0.8319, "step": 14205 }, { "epoch": 1.47, "learning_rate": 4.1849571027237544e-05, "loss": 0.2197, "step": 14206 }, { "epoch": 1.47, "learning_rate": 4.1847309711953696e-05, "loss": 0.3587, "step": 14207 }, { "epoch": 1.47, "learning_rate": 4.184504831691653e-05, "loss": 0.1578, "step": 14208 }, { "epoch": 1.47, "learning_rate": 4.184278684214127e-05, "loss": 0.2556, "step": 14209 }, { "epoch": 1.47, "learning_rate": 4.184052528764314e-05, "loss": 0.2515, "step": 14210 }, { "epoch": 1.47, "learning_rate": 4.183826365343736e-05, "loss": 0.3662, "step": 14211 }, { "epoch": 1.47, "learning_rate": 4.183600193953916e-05, "loss": 1.2262, "step": 14212 }, { "epoch": 1.47, "learning_rate": 4.183374014596377e-05, "loss": 0.286, "step": 14213 }, { "epoch": 1.47, "learning_rate": 4.18314782727264e-05, "loss": 0.5696, "step": 14214 }, { "epoch": 1.47, "learning_rate": 4.18292163198423e-05, "loss": 0.5952, "step": 14215 }, { "epoch": 1.47, "learning_rate": 4.182695428732667e-05, "loss": 0.5833, "step": 14216 }, { "epoch": 1.47, "learning_rate": 4.182469217519476e-05, "loss": 0.2721, "step": 14217 }, { "epoch": 1.47, "learning_rate": 4.182242998346179e-05, "loss": 1.0426, "step": 14218 }, { "epoch": 1.47, "learning_rate": 4.182016771214298e-05, "loss": 0.2344, "step": 14219 }, { "epoch": 1.47, "learning_rate": 4.181790536125357e-05, "loss": 0.3879, "step": 14220 }, { "epoch": 1.47, "learning_rate": 4.181564293080878e-05, "loss": 0.3402, "step": 14221 }, { "epoch": 1.47, "learning_rate": 4.181338042082386e-05, "loss": 0.5978, "step": 14222 }, { "epoch": 1.47, "learning_rate": 4.181111783131402e-05, "loss": 1.0305, "step": 14223 }, { "epoch": 1.47, "learning_rate": 4.18088551622945e-05, "loss": 1.0807, "step": 14224 }, { "epoch": 1.47, "learning_rate": 4.1806592413780526e-05, "loss": 0.8529, "step": 14225 }, { "epoch": 1.47, "learning_rate": 4.1804329585787334e-05, "loss": 0.9456, "step": 14226 }, { "epoch": 1.47, "learning_rate": 4.1802066678330165e-05, "loss": 1.2191, "step": 14227 }, { "epoch": 1.47, "learning_rate": 4.179980369142424e-05, "loss": 0.2646, "step": 14228 }, { "epoch": 1.47, "learning_rate": 4.179754062508479e-05, "loss": 0.8552, "step": 14229 }, { "epoch": 1.47, "learning_rate": 4.179527747932707e-05, "loss": 0.1901, "step": 14230 }, { "epoch": 1.47, "learning_rate": 4.17930142541663e-05, "loss": 0.2722, "step": 14231 }, { "epoch": 1.47, "learning_rate": 4.1790750949617714e-05, "loss": 0.4933, "step": 14232 }, { "epoch": 1.47, "learning_rate": 4.1788487565696543e-05, "loss": 1.0748, "step": 14233 }, { "epoch": 1.47, "learning_rate": 4.178622410241804e-05, "loss": 0.689, "step": 14234 }, { "epoch": 1.47, "learning_rate": 4.1783960559797436e-05, "loss": 0.9785, "step": 14235 }, { "epoch": 1.47, "learning_rate": 4.178169693784997e-05, "loss": 0.7187, "step": 14236 }, { "epoch": 1.47, "learning_rate": 4.177943323659087e-05, "loss": 0.4703, "step": 14237 }, { "epoch": 1.47, "learning_rate": 4.177716945603538e-05, "loss": 0.195, "step": 14238 }, { "epoch": 1.47, "learning_rate": 4.177490559619875e-05, "loss": 0.7748, "step": 14239 }, { "epoch": 1.47, "learning_rate": 4.17726416570962e-05, "loss": 1.0573, "step": 14240 }, { "epoch": 1.47, "learning_rate": 4.177037763874299e-05, "loss": 0.7576, "step": 14241 }, { "epoch": 1.47, "learning_rate": 4.176811354115435e-05, "loss": 0.3585, "step": 14242 }, { "epoch": 1.47, "learning_rate": 4.1765849364345517e-05, "loss": 0.3937, "step": 14243 }, { "epoch": 1.47, "learning_rate": 4.176358510833174e-05, "loss": 0.5202, "step": 14244 }, { "epoch": 1.47, "learning_rate": 4.176132077312826e-05, "loss": 0.1385, "step": 14245 }, { "epoch": 1.47, "learning_rate": 4.1759056358750326e-05, "loss": 0.4425, "step": 14246 }, { "epoch": 1.47, "learning_rate": 4.1756791865213176e-05, "loss": 0.3101, "step": 14247 }, { "epoch": 1.47, "learning_rate": 4.175452729253206e-05, "loss": 0.6253, "step": 14248 }, { "epoch": 1.47, "learning_rate": 4.17522626407222e-05, "loss": 0.3137, "step": 14249 }, { "epoch": 1.47, "learning_rate": 4.1749997909798865e-05, "loss": 1.2814, "step": 14250 }, { "epoch": 1.47, "learning_rate": 4.174773309977729e-05, "loss": 0.6003, "step": 14251 }, { "epoch": 1.47, "learning_rate": 4.1745468210672735e-05, "loss": 0.6762, "step": 14252 }, { "epoch": 1.47, "learning_rate": 4.1743203242500436e-05, "loss": 0.3016, "step": 14253 }, { "epoch": 1.47, "learning_rate": 4.174093819527563e-05, "loss": 0.4868, "step": 14254 }, { "epoch": 1.47, "learning_rate": 4.173867306901358e-05, "loss": 1.0662, "step": 14255 }, { "epoch": 1.47, "learning_rate": 4.1736407863729534e-05, "loss": 0.996, "step": 14256 }, { "epoch": 1.47, "learning_rate": 4.173414257943873e-05, "loss": 0.3763, "step": 14257 }, { "epoch": 1.47, "learning_rate": 4.173187721615642e-05, "loss": 0.6585, "step": 14258 }, { "epoch": 1.47, "learning_rate": 4.172961177389787e-05, "loss": 1.1555, "step": 14259 }, { "epoch": 1.47, "learning_rate": 4.172734625267831e-05, "loss": 0.9057, "step": 14260 }, { "epoch": 1.47, "learning_rate": 4.1725080652513007e-05, "loss": 0.6238, "step": 14261 }, { "epoch": 1.47, "learning_rate": 4.172281497341719e-05, "loss": 1.1129, "step": 14262 }, { "epoch": 1.47, "learning_rate": 4.172054921540613e-05, "loss": 0.8686, "step": 14263 }, { "epoch": 1.47, "learning_rate": 4.1718283378495084e-05, "loss": 0.2771, "step": 14264 }, { "epoch": 1.47, "learning_rate": 4.171601746269929e-05, "loss": 0.7006, "step": 14265 }, { "epoch": 1.47, "learning_rate": 4.171375146803401e-05, "loss": 0.8252, "step": 14266 }, { "epoch": 1.47, "learning_rate": 4.1711485394514493e-05, "loss": 0.8922, "step": 14267 }, { "epoch": 1.47, "learning_rate": 4.1709219242155995e-05, "loss": 0.63, "step": 14268 }, { "epoch": 1.47, "learning_rate": 4.170695301097378e-05, "loss": 0.5084, "step": 14269 }, { "epoch": 1.47, "learning_rate": 4.170468670098309e-05, "loss": 0.871, "step": 14270 }, { "epoch": 1.47, "learning_rate": 4.1702420312199184e-05, "loss": 0.3787, "step": 14271 }, { "epoch": 1.48, "learning_rate": 4.170015384463734e-05, "loss": 1.3718, "step": 14272 }, { "epoch": 1.48, "learning_rate": 4.1697887298312783e-05, "loss": 0.2983, "step": 14273 }, { "epoch": 1.48, "learning_rate": 4.169562067324079e-05, "loss": 0.3158, "step": 14274 }, { "epoch": 1.48, "learning_rate": 4.1693353969436616e-05, "loss": 0.2079, "step": 14275 }, { "epoch": 1.48, "learning_rate": 4.1691087186915515e-05, "loss": 1.161, "step": 14276 }, { "epoch": 1.48, "learning_rate": 4.168882032569275e-05, "loss": 0.7975, "step": 14277 }, { "epoch": 1.48, "learning_rate": 4.168655338578358e-05, "loss": 0.7195, "step": 14278 }, { "epoch": 1.48, "learning_rate": 4.1684286367203275e-05, "loss": 0.3943, "step": 14279 }, { "epoch": 1.48, "learning_rate": 4.168201926996709e-05, "loss": 0.3237, "step": 14280 }, { "epoch": 1.48, "learning_rate": 4.167975209409027e-05, "loss": 0.1862, "step": 14281 }, { "epoch": 1.48, "learning_rate": 4.16774848395881e-05, "loss": 0.1825, "step": 14282 }, { "epoch": 1.48, "learning_rate": 4.1675217506475846e-05, "loss": 0.9798, "step": 14283 }, { "epoch": 1.48, "learning_rate": 4.1672950094768745e-05, "loss": 0.4641, "step": 14284 }, { "epoch": 1.48, "learning_rate": 4.167068260448208e-05, "loss": 0.7484, "step": 14285 }, { "epoch": 1.48, "learning_rate": 4.166841503563111e-05, "loss": 0.3358, "step": 14286 }, { "epoch": 1.48, "learning_rate": 4.166614738823109e-05, "loss": 1.2666, "step": 14287 }, { "epoch": 1.48, "learning_rate": 4.166387966229732e-05, "loss": 1.3852, "step": 14288 }, { "epoch": 1.48, "learning_rate": 4.1661611857845016e-05, "loss": 0.8657, "step": 14289 }, { "epoch": 1.48, "learning_rate": 4.165934397488948e-05, "loss": 1.1528, "step": 14290 }, { "epoch": 1.48, "learning_rate": 4.1657076013445976e-05, "loss": 0.7314, "step": 14291 }, { "epoch": 1.48, "learning_rate": 4.165480797352976e-05, "loss": 0.2785, "step": 14292 }, { "epoch": 1.48, "learning_rate": 4.1652539855156096e-05, "loss": 0.2774, "step": 14293 }, { "epoch": 1.48, "learning_rate": 4.165027165834027e-05, "loss": 0.7979, "step": 14294 }, { "epoch": 1.48, "learning_rate": 4.1648003383097525e-05, "loss": 1.1323, "step": 14295 }, { "epoch": 1.48, "learning_rate": 4.164573502944315e-05, "loss": 0.4668, "step": 14296 }, { "epoch": 1.48, "learning_rate": 4.1643466597392425e-05, "loss": 0.5303, "step": 14297 }, { "epoch": 1.48, "learning_rate": 4.1641198086960615e-05, "loss": 0.5568, "step": 14298 }, { "epoch": 1.48, "learning_rate": 4.1638929498162965e-05, "loss": 0.8264, "step": 14299 }, { "epoch": 1.48, "learning_rate": 4.163666083101476e-05, "loss": 0.8148, "step": 14300 }, { "epoch": 1.48, "learning_rate": 4.163439208553129e-05, "loss": 0.5058, "step": 14301 }, { "epoch": 1.48, "learning_rate": 4.163212326172781e-05, "loss": 0.2314, "step": 14302 }, { "epoch": 1.48, "learning_rate": 4.16298543596196e-05, "loss": 0.7375, "step": 14303 }, { "epoch": 1.48, "learning_rate": 4.1627585379221936e-05, "loss": 0.755, "step": 14304 }, { "epoch": 1.48, "learning_rate": 4.162531632055008e-05, "loss": 0.2576, "step": 14305 }, { "epoch": 1.48, "learning_rate": 4.1623047183619314e-05, "loss": 0.6591, "step": 14306 }, { "epoch": 1.48, "learning_rate": 4.162077796844492e-05, "loss": 0.1004, "step": 14307 }, { "epoch": 1.48, "learning_rate": 4.161850867504216e-05, "loss": 0.192, "step": 14308 }, { "epoch": 1.48, "learning_rate": 4.161623930342632e-05, "loss": 0.2603, "step": 14309 }, { "epoch": 1.48, "learning_rate": 4.161396985361268e-05, "loss": 0.43, "step": 14310 }, { "epoch": 1.48, "learning_rate": 4.1611700325616514e-05, "loss": 0.6759, "step": 14311 }, { "epoch": 1.48, "learning_rate": 4.16094307194531e-05, "loss": 0.7533, "step": 14312 }, { "epoch": 1.48, "learning_rate": 4.16071610351377e-05, "loss": 0.3861, "step": 14313 }, { "epoch": 1.48, "learning_rate": 4.1604891272685616e-05, "loss": 0.5948, "step": 14314 }, { "epoch": 1.48, "learning_rate": 4.160262143211212e-05, "loss": 1.0312, "step": 14315 }, { "epoch": 1.48, "learning_rate": 4.16003515134325e-05, "loss": 0.7884, "step": 14316 }, { "epoch": 1.48, "learning_rate": 4.1598081516662016e-05, "loss": 1.5993, "step": 14317 }, { "epoch": 1.48, "learning_rate": 4.1595811441815966e-05, "loss": 0.8407, "step": 14318 }, { "epoch": 1.48, "learning_rate": 4.159354128890962e-05, "loss": 0.3857, "step": 14319 }, { "epoch": 1.48, "learning_rate": 4.1591271057958274e-05, "loss": 0.4999, "step": 14320 }, { "epoch": 1.48, "learning_rate": 4.1589000748977206e-05, "loss": 0.4302, "step": 14321 }, { "epoch": 1.48, "learning_rate": 4.158673036198169e-05, "loss": 2.2808, "step": 14322 }, { "epoch": 1.48, "learning_rate": 4.1584459896987016e-05, "loss": 0.2856, "step": 14323 }, { "epoch": 1.48, "learning_rate": 4.158218935400847e-05, "loss": 0.3512, "step": 14324 }, { "epoch": 1.48, "learning_rate": 4.157991873306133e-05, "loss": 0.8458, "step": 14325 }, { "epoch": 1.48, "learning_rate": 4.15776480341609e-05, "loss": 0.2654, "step": 14326 }, { "epoch": 1.48, "learning_rate": 4.157537725732245e-05, "loss": 1.0833, "step": 14327 }, { "epoch": 1.48, "learning_rate": 4.157310640256126e-05, "loss": 0.2383, "step": 14328 }, { "epoch": 1.48, "learning_rate": 4.1570835469892626e-05, "loss": 0.9823, "step": 14329 }, { "epoch": 1.48, "learning_rate": 4.1568564459331834e-05, "loss": 0.3292, "step": 14330 }, { "epoch": 1.48, "learning_rate": 4.156629337089418e-05, "loss": 0.7142, "step": 14331 }, { "epoch": 1.48, "learning_rate": 4.1564022204594944e-05, "loss": 0.8012, "step": 14332 }, { "epoch": 1.48, "learning_rate": 4.1561750960449414e-05, "loss": 0.6486, "step": 14333 }, { "epoch": 1.48, "learning_rate": 4.155947963847288e-05, "loss": 0.4152, "step": 14334 }, { "epoch": 1.48, "learning_rate": 4.155720823868064e-05, "loss": 0.8239, "step": 14335 }, { "epoch": 1.48, "learning_rate": 4.155493676108797e-05, "loss": 0.5713, "step": 14336 }, { "epoch": 1.48, "learning_rate": 4.1552665205710175e-05, "loss": 1.0796, "step": 14337 }, { "epoch": 1.48, "learning_rate": 4.155039357256254e-05, "loss": 0.4014, "step": 14338 }, { "epoch": 1.48, "learning_rate": 4.154812186166036e-05, "loss": 0.6559, "step": 14339 }, { "epoch": 1.48, "learning_rate": 4.1545850073018924e-05, "loss": 0.7925, "step": 14340 }, { "epoch": 1.48, "learning_rate": 4.1543578206653535e-05, "loss": 0.324, "step": 14341 }, { "epoch": 1.48, "learning_rate": 4.154130626257946e-05, "loss": 0.2475, "step": 14342 }, { "epoch": 1.48, "learning_rate": 4.153903424081202e-05, "loss": 0.7767, "step": 14343 }, { "epoch": 1.48, "learning_rate": 4.15367621413665e-05, "loss": 0.6819, "step": 14344 }, { "epoch": 1.48, "learning_rate": 4.1534489964258204e-05, "loss": 0.5871, "step": 14345 }, { "epoch": 1.48, "learning_rate": 4.153221770950242e-05, "loss": 0.4037, "step": 14346 }, { "epoch": 1.48, "learning_rate": 4.152994537711444e-05, "loss": 1.11, "step": 14347 }, { "epoch": 1.48, "learning_rate": 4.152767296710956e-05, "loss": 0.9372, "step": 14348 }, { "epoch": 1.48, "learning_rate": 4.152540047950309e-05, "loss": 0.7364, "step": 14349 }, { "epoch": 1.48, "learning_rate": 4.1523127914310324e-05, "loss": 0.7479, "step": 14350 }, { "epoch": 1.48, "learning_rate": 4.1520855271546546e-05, "loss": 0.1392, "step": 14351 }, { "epoch": 1.48, "learning_rate": 4.151858255122708e-05, "loss": 0.3328, "step": 14352 }, { "epoch": 1.48, "learning_rate": 4.1516309753367196e-05, "loss": 0.7509, "step": 14353 }, { "epoch": 1.48, "learning_rate": 4.151403687798222e-05, "loss": 0.6491, "step": 14354 }, { "epoch": 1.48, "learning_rate": 4.151176392508745e-05, "loss": 0.625, "step": 14355 }, { "epoch": 1.48, "learning_rate": 4.150949089469816e-05, "loss": 0.8528, "step": 14356 }, { "epoch": 1.48, "learning_rate": 4.1507217786829675e-05, "loss": 0.4863, "step": 14357 }, { "epoch": 1.48, "learning_rate": 4.1504944601497296e-05, "loss": 0.667, "step": 14358 }, { "epoch": 1.48, "learning_rate": 4.1502671338716325e-05, "loss": 0.2821, "step": 14359 }, { "epoch": 1.48, "learning_rate": 4.150039799850206e-05, "loss": 0.4066, "step": 14360 }, { "epoch": 1.48, "learning_rate": 4.1498124580869805e-05, "loss": 0.3154, "step": 14361 }, { "epoch": 1.48, "learning_rate": 4.149585108583486e-05, "loss": 0.7039, "step": 14362 }, { "epoch": 1.48, "learning_rate": 4.149357751341254e-05, "loss": 0.9983, "step": 14363 }, { "epoch": 1.48, "learning_rate": 4.149130386361815e-05, "loss": 0.3726, "step": 14364 }, { "epoch": 1.48, "learning_rate": 4.148903013646699e-05, "loss": 0.9487, "step": 14365 }, { "epoch": 1.48, "learning_rate": 4.148675633197437e-05, "loss": 0.3035, "step": 14366 }, { "epoch": 1.49, "learning_rate": 4.148448245015559e-05, "loss": 0.7547, "step": 14367 }, { "epoch": 1.49, "learning_rate": 4.148220849102595e-05, "loss": 0.8295, "step": 14368 }, { "epoch": 1.49, "learning_rate": 4.147993445460079e-05, "loss": 0.8109, "step": 14369 }, { "epoch": 1.49, "learning_rate": 4.147766034089538e-05, "loss": 1.0158, "step": 14370 }, { "epoch": 1.49, "learning_rate": 4.147538614992505e-05, "loss": 0.6063, "step": 14371 }, { "epoch": 1.49, "learning_rate": 4.147311188170511e-05, "loss": 0.2892, "step": 14372 }, { "epoch": 1.49, "learning_rate": 4.1470837536250864e-05, "loss": 1.4243, "step": 14373 }, { "epoch": 1.49, "learning_rate": 4.146856311357762e-05, "loss": 0.4093, "step": 14374 }, { "epoch": 1.49, "learning_rate": 4.14662886137007e-05, "loss": 0.413, "step": 14375 }, { "epoch": 1.49, "learning_rate": 4.146401403663541e-05, "loss": 0.7888, "step": 14376 }, { "epoch": 1.49, "learning_rate": 4.1461739382397046e-05, "loss": 0.8487, "step": 14377 }, { "epoch": 1.49, "learning_rate": 4.145946465100095e-05, "loss": 0.8452, "step": 14378 }, { "epoch": 1.49, "learning_rate": 4.145718984246241e-05, "loss": 1.0973, "step": 14379 }, { "epoch": 1.49, "learning_rate": 4.145491495679675e-05, "loss": 0.4085, "step": 14380 }, { "epoch": 1.49, "learning_rate": 4.1452639994019276e-05, "loss": 0.3604, "step": 14381 }, { "epoch": 1.49, "learning_rate": 4.145036495414532e-05, "loss": 0.2644, "step": 14382 }, { "epoch": 1.49, "learning_rate": 4.1448089837190184e-05, "loss": 0.3188, "step": 14383 }, { "epoch": 1.49, "learning_rate": 4.144581464316918e-05, "loss": 0.4348, "step": 14384 }, { "epoch": 1.49, "learning_rate": 4.144353937209764e-05, "loss": 0.218, "step": 14385 }, { "epoch": 1.49, "learning_rate": 4.1441264023990854e-05, "loss": 0.2487, "step": 14386 }, { "epoch": 1.49, "learning_rate": 4.1438988598864174e-05, "loss": 0.1336, "step": 14387 }, { "epoch": 1.49, "learning_rate": 4.143671309673289e-05, "loss": 0.3177, "step": 14388 }, { "epoch": 1.49, "learning_rate": 4.143443751761233e-05, "loss": 0.4275, "step": 14389 }, { "epoch": 1.49, "learning_rate": 4.143216186151781e-05, "loss": 1.0705, "step": 14390 }, { "epoch": 1.49, "learning_rate": 4.142988612846466e-05, "loss": 0.3099, "step": 14391 }, { "epoch": 1.49, "learning_rate": 4.142761031846818e-05, "loss": 0.8081, "step": 14392 }, { "epoch": 1.49, "learning_rate": 4.1425334431543714e-05, "loss": 0.5904, "step": 14393 }, { "epoch": 1.49, "learning_rate": 4.142305846770656e-05, "loss": 0.5495, "step": 14394 }, { "epoch": 1.49, "learning_rate": 4.142078242697205e-05, "loss": 0.7221, "step": 14395 }, { "epoch": 1.49, "learning_rate": 4.141850630935551e-05, "loss": 0.89, "step": 14396 }, { "epoch": 1.49, "learning_rate": 4.1416230114872244e-05, "loss": 0.9603, "step": 14397 }, { "epoch": 1.49, "learning_rate": 4.141395384353761e-05, "loss": 0.8944, "step": 14398 }, { "epoch": 1.49, "learning_rate": 4.141167749536689e-05, "loss": 0.4257, "step": 14399 }, { "epoch": 1.49, "learning_rate": 4.1409401070375434e-05, "loss": 0.5797, "step": 14400 }, { "epoch": 1.49, "learning_rate": 4.140712456857856e-05, "loss": 0.2453, "step": 14401 }, { "epoch": 1.49, "learning_rate": 4.14048479899916e-05, "loss": 1.2008, "step": 14402 }, { "epoch": 1.49, "learning_rate": 4.1402571334629856e-05, "loss": 0.5799, "step": 14403 }, { "epoch": 1.49, "learning_rate": 4.1400294602508685e-05, "loss": 1.517, "step": 14404 }, { "epoch": 1.49, "learning_rate": 4.139801779364339e-05, "loss": 0.3017, "step": 14405 }, { "epoch": 1.49, "learning_rate": 4.13957409080493e-05, "loss": 0.2941, "step": 14406 }, { "epoch": 1.49, "learning_rate": 4.1393463945741764e-05, "loss": 0.2172, "step": 14407 }, { "epoch": 1.49, "learning_rate": 4.139118690673608e-05, "loss": 0.7443, "step": 14408 }, { "epoch": 1.49, "learning_rate": 4.13889097910476e-05, "loss": 0.2796, "step": 14409 }, { "epoch": 1.49, "learning_rate": 4.1386632598691644e-05, "loss": 0.194, "step": 14410 }, { "epoch": 1.49, "learning_rate": 4.138435532968354e-05, "loss": 0.4317, "step": 14411 }, { "epoch": 1.49, "learning_rate": 4.138207798403862e-05, "loss": 0.2905, "step": 14412 }, { "epoch": 1.49, "learning_rate": 4.137980056177221e-05, "loss": 0.6584, "step": 14413 }, { "epoch": 1.49, "learning_rate": 4.137752306289965e-05, "loss": 1.0674, "step": 14414 }, { "epoch": 1.49, "learning_rate": 4.137524548743626e-05, "loss": 0.6951, "step": 14415 }, { "epoch": 1.49, "learning_rate": 4.1372967835397386e-05, "loss": 1.0873, "step": 14416 }, { "epoch": 1.49, "learning_rate": 4.137069010679836e-05, "loss": 0.2787, "step": 14417 }, { "epoch": 1.49, "learning_rate": 4.13684123016545e-05, "loss": 0.7596, "step": 14418 }, { "epoch": 1.49, "learning_rate": 4.136613441998114e-05, "loss": 0.8799, "step": 14419 }, { "epoch": 1.49, "learning_rate": 4.1363856461793633e-05, "loss": 0.7553, "step": 14420 }, { "epoch": 1.49, "learning_rate": 4.1361578427107304e-05, "loss": 0.6026, "step": 14421 }, { "epoch": 1.49, "learning_rate": 4.1359300315937484e-05, "loss": 0.365, "step": 14422 }, { "epoch": 1.49, "learning_rate": 4.1357022128299515e-05, "loss": 0.2225, "step": 14423 }, { "epoch": 1.49, "learning_rate": 4.1354743864208726e-05, "loss": 0.2361, "step": 14424 }, { "epoch": 1.49, "learning_rate": 4.135246552368046e-05, "loss": 0.3534, "step": 14425 }, { "epoch": 1.49, "learning_rate": 4.135018710673006e-05, "loss": 0.7395, "step": 14426 }, { "epoch": 1.49, "learning_rate": 4.1347908613372844e-05, "loss": 1.0052, "step": 14427 }, { "epoch": 1.49, "learning_rate": 4.134563004362416e-05, "loss": 0.4577, "step": 14428 }, { "epoch": 1.49, "learning_rate": 4.134335139749937e-05, "loss": 1.1337, "step": 14429 }, { "epoch": 1.49, "learning_rate": 4.1341072675013776e-05, "loss": 0.5158, "step": 14430 }, { "epoch": 1.49, "learning_rate": 4.1338793876182745e-05, "loss": 0.5881, "step": 14431 }, { "epoch": 1.49, "learning_rate": 4.133651500102159e-05, "loss": 1.1194, "step": 14432 }, { "epoch": 1.49, "learning_rate": 4.1334236049545674e-05, "loss": 0.2237, "step": 14433 }, { "epoch": 1.49, "learning_rate": 4.133195702177034e-05, "loss": 0.6916, "step": 14434 }, { "epoch": 1.49, "learning_rate": 4.1329677917710924e-05, "loss": 0.2344, "step": 14435 }, { "epoch": 1.49, "learning_rate": 4.1327398737382765e-05, "loss": 0.4905, "step": 14436 }, { "epoch": 1.49, "learning_rate": 4.1325119480801204e-05, "loss": 0.6511, "step": 14437 }, { "epoch": 1.49, "learning_rate": 4.132284014798159e-05, "loss": 0.4315, "step": 14438 }, { "epoch": 1.49, "learning_rate": 4.132056073893926e-05, "loss": 1.2512, "step": 14439 }, { "epoch": 1.49, "learning_rate": 4.131828125368958e-05, "loss": 0.737, "step": 14440 }, { "epoch": 1.49, "learning_rate": 4.1316001692247866e-05, "loss": 0.1934, "step": 14441 }, { "epoch": 1.49, "learning_rate": 4.131372205462948e-05, "loss": 0.5663, "step": 14442 }, { "epoch": 1.49, "learning_rate": 4.1311442340849765e-05, "loss": 0.6725, "step": 14443 }, { "epoch": 1.49, "learning_rate": 4.1309162550924056e-05, "loss": 1.1599, "step": 14444 }, { "epoch": 1.49, "learning_rate": 4.1306882684867724e-05, "loss": 1.0511, "step": 14445 }, { "epoch": 1.49, "learning_rate": 4.13046027426961e-05, "loss": 1.5881, "step": 14446 }, { "epoch": 1.49, "learning_rate": 4.1302322724424535e-05, "loss": 0.334, "step": 14447 }, { "epoch": 1.49, "learning_rate": 4.1300042630068374e-05, "loss": 0.4092, "step": 14448 }, { "epoch": 1.49, "learning_rate": 4.129776245964298e-05, "loss": 0.9569, "step": 14449 }, { "epoch": 1.49, "learning_rate": 4.129548221316369e-05, "loss": 0.6386, "step": 14450 }, { "epoch": 1.49, "learning_rate": 4.129320189064585e-05, "loss": 0.7409, "step": 14451 }, { "epoch": 1.49, "learning_rate": 4.1290921492104816e-05, "loss": 1.0563, "step": 14452 }, { "epoch": 1.49, "learning_rate": 4.1288641017555946e-05, "loss": 1.3263, "step": 14453 }, { "epoch": 1.49, "learning_rate": 4.1286360467014586e-05, "loss": 0.5283, "step": 14454 }, { "epoch": 1.49, "learning_rate": 4.128407984049609e-05, "loss": 0.5226, "step": 14455 }, { "epoch": 1.49, "learning_rate": 4.1281799138015805e-05, "loss": 0.542, "step": 14456 }, { "epoch": 1.49, "learning_rate": 4.127951835958908e-05, "loss": 0.7642, "step": 14457 }, { "epoch": 1.49, "learning_rate": 4.127723750523129e-05, "loss": 0.3607, "step": 14458 }, { "epoch": 1.49, "learning_rate": 4.127495657495778e-05, "loss": 0.8829, "step": 14459 }, { "epoch": 1.49, "learning_rate": 4.1272675568783885e-05, "loss": 0.2851, "step": 14460 }, { "epoch": 1.49, "learning_rate": 4.127039448672499e-05, "loss": 1.0794, "step": 14461 }, { "epoch": 1.49, "learning_rate": 4.126811332879642e-05, "loss": 0.8685, "step": 14462 }, { "epoch": 1.5, "learning_rate": 4.126583209501355e-05, "loss": 0.3215, "step": 14463 }, { "epoch": 1.5, "learning_rate": 4.126355078539175e-05, "loss": 0.4463, "step": 14464 }, { "epoch": 1.5, "learning_rate": 4.1261269399946345e-05, "loss": 0.7144, "step": 14465 }, { "epoch": 1.5, "learning_rate": 4.125898793869272e-05, "loss": 0.8054, "step": 14466 }, { "epoch": 1.5, "learning_rate": 4.1256706401646216e-05, "loss": 0.249, "step": 14467 }, { "epoch": 1.5, "learning_rate": 4.12544247888222e-05, "loss": 0.7235, "step": 14468 }, { "epoch": 1.5, "learning_rate": 4.125214310023604e-05, "loss": 0.2824, "step": 14469 }, { "epoch": 1.5, "learning_rate": 4.124986133590307e-05, "loss": 0.7989, "step": 14470 }, { "epoch": 1.5, "learning_rate": 4.124757949583867e-05, "loss": 1.1798, "step": 14471 }, { "epoch": 1.5, "learning_rate": 4.12452975800582e-05, "loss": 0.3542, "step": 14472 }, { "epoch": 1.5, "learning_rate": 4.124301558857702e-05, "loss": 0.6663, "step": 14473 }, { "epoch": 1.5, "learning_rate": 4.124073352141049e-05, "loss": 0.5592, "step": 14474 }, { "epoch": 1.5, "learning_rate": 4.123845137857397e-05, "loss": 0.3809, "step": 14475 }, { "epoch": 1.5, "learning_rate": 4.1236169160082815e-05, "loss": 0.8958, "step": 14476 }, { "epoch": 1.5, "learning_rate": 4.1233886865952404e-05, "loss": 0.2954, "step": 14477 }, { "epoch": 1.5, "learning_rate": 4.12316044961981e-05, "loss": 0.5476, "step": 14478 }, { "epoch": 1.5, "learning_rate": 4.122932205083526e-05, "loss": 0.6404, "step": 14479 }, { "epoch": 1.5, "learning_rate": 4.122703952987925e-05, "loss": 1.3377, "step": 14480 }, { "epoch": 1.5, "learning_rate": 4.122475693334544e-05, "loss": 0.6684, "step": 14481 }, { "epoch": 1.5, "learning_rate": 4.122247426124919e-05, "loss": 0.2731, "step": 14482 }, { "epoch": 1.5, "learning_rate": 4.122019151360587e-05, "loss": 0.2221, "step": 14483 }, { "epoch": 1.5, "learning_rate": 4.1217908690430846e-05, "loss": 0.1535, "step": 14484 }, { "epoch": 1.5, "learning_rate": 4.121562579173949e-05, "loss": 0.9223, "step": 14485 }, { "epoch": 1.5, "learning_rate": 4.121334281754716e-05, "loss": 0.736, "step": 14486 }, { "epoch": 1.5, "learning_rate": 4.121105976786923e-05, "loss": 0.6812, "step": 14487 }, { "epoch": 1.5, "learning_rate": 4.120877664272107e-05, "loss": 0.9964, "step": 14488 }, { "epoch": 1.5, "learning_rate": 4.120649344211805e-05, "loss": 1.2702, "step": 14489 }, { "epoch": 1.5, "learning_rate": 4.1204210166075536e-05, "loss": 0.1007, "step": 14490 }, { "epoch": 1.5, "learning_rate": 4.12019268146089e-05, "loss": 0.8018, "step": 14491 }, { "epoch": 1.5, "learning_rate": 4.1199643387733515e-05, "loss": 0.6223, "step": 14492 }, { "epoch": 1.5, "learning_rate": 4.119735988546476e-05, "loss": 0.9751, "step": 14493 }, { "epoch": 1.5, "learning_rate": 4.119507630781798e-05, "loss": 0.5566, "step": 14494 }, { "epoch": 1.5, "learning_rate": 4.119279265480857e-05, "loss": 0.3197, "step": 14495 }, { "epoch": 1.5, "learning_rate": 4.119050892645191e-05, "loss": 1.1009, "step": 14496 }, { "epoch": 1.5, "learning_rate": 4.118822512276336e-05, "loss": 0.9566, "step": 14497 }, { "epoch": 1.5, "learning_rate": 4.1185941243758295e-05, "loss": 1.0231, "step": 14498 }, { "epoch": 1.5, "learning_rate": 4.118365728945208e-05, "loss": 0.5335, "step": 14499 }, { "epoch": 1.5, "learning_rate": 4.118137325986011e-05, "loss": 0.8973, "step": 14500 }, { "epoch": 1.5, "learning_rate": 4.117908915499775e-05, "loss": 1.0737, "step": 14501 }, { "epoch": 1.5, "learning_rate": 4.117680497488038e-05, "loss": 1.0537, "step": 14502 }, { "epoch": 1.5, "learning_rate": 4.117452071952337e-05, "loss": 0.96, "step": 14503 }, { "epoch": 1.5, "learning_rate": 4.117223638894211e-05, "loss": 0.5361, "step": 14504 }, { "epoch": 1.5, "learning_rate": 4.116995198315196e-05, "loss": 0.0897, "step": 14505 }, { "epoch": 1.5, "learning_rate": 4.11676675021683e-05, "loss": 0.2266, "step": 14506 }, { "epoch": 1.5, "learning_rate": 4.116538294600653e-05, "loss": 0.8756, "step": 14507 }, { "epoch": 1.5, "learning_rate": 4.1163098314682e-05, "loss": 0.7603, "step": 14508 }, { "epoch": 1.5, "learning_rate": 4.116081360821012e-05, "loss": 0.3795, "step": 14509 }, { "epoch": 1.5, "learning_rate": 4.1158528826606245e-05, "loss": 1.0359, "step": 14510 }, { "epoch": 1.5, "learning_rate": 4.1156243969885774e-05, "loss": 0.69, "step": 14511 }, { "epoch": 1.5, "learning_rate": 4.1153959038064076e-05, "loss": 1.1722, "step": 14512 }, { "epoch": 1.5, "learning_rate": 4.115167403115652e-05, "loss": 0.9432, "step": 14513 }, { "epoch": 1.5, "learning_rate": 4.114938894917852e-05, "loss": 0.3279, "step": 14514 }, { "epoch": 1.5, "learning_rate": 4.114710379214543e-05, "loss": 1.3405, "step": 14515 }, { "epoch": 1.5, "learning_rate": 4.1144818560072654e-05, "loss": 0.4249, "step": 14516 }, { "epoch": 1.5, "learning_rate": 4.1142533252975576e-05, "loss": 1.0248, "step": 14517 }, { "epoch": 1.5, "learning_rate": 4.114024787086956e-05, "loss": 0.3096, "step": 14518 }, { "epoch": 1.5, "learning_rate": 4.113796241377e-05, "loss": 0.652, "step": 14519 }, { "epoch": 1.5, "learning_rate": 4.113567688169228e-05, "loss": 1.1556, "step": 14520 }, { "epoch": 1.5, "learning_rate": 4.1133391274651806e-05, "loss": 0.6953, "step": 14521 }, { "epoch": 1.5, "learning_rate": 4.113110559266393e-05, "loss": 0.2354, "step": 14522 }, { "epoch": 1.5, "learning_rate": 4.112881983574407e-05, "loss": 0.8328, "step": 14523 }, { "epoch": 1.5, "learning_rate": 4.112653400390758e-05, "loss": 1.0322, "step": 14524 }, { "epoch": 1.5, "learning_rate": 4.112424809716988e-05, "loss": 0.7153, "step": 14525 }, { "epoch": 1.5, "learning_rate": 4.1121962115546346e-05, "loss": 0.1284, "step": 14526 }, { "epoch": 1.5, "learning_rate": 4.111967605905236e-05, "loss": 1.7482, "step": 14527 }, { "epoch": 1.5, "learning_rate": 4.1117389927703305e-05, "loss": 0.2043, "step": 14528 }, { "epoch": 1.5, "learning_rate": 4.111510372151461e-05, "loss": 0.5903, "step": 14529 }, { "epoch": 1.5, "learning_rate": 4.1112817440501616e-05, "loss": 0.3373, "step": 14530 }, { "epoch": 1.5, "learning_rate": 4.1110531084679744e-05, "loss": 0.6726, "step": 14531 }, { "epoch": 1.5, "learning_rate": 4.110824465406437e-05, "loss": 0.6134, "step": 14532 }, { "epoch": 1.5, "learning_rate": 4.110595814867089e-05, "loss": 1.2699, "step": 14533 }, { "epoch": 1.5, "learning_rate": 4.110367156851471e-05, "loss": 0.7026, "step": 14534 }, { "epoch": 1.5, "learning_rate": 4.110138491361121e-05, "loss": 0.7098, "step": 14535 }, { "epoch": 1.5, "learning_rate": 4.109909818397578e-05, "loss": 0.3879, "step": 14536 }, { "epoch": 1.5, "learning_rate": 4.1096811379623815e-05, "loss": 0.8154, "step": 14537 }, { "epoch": 1.5, "learning_rate": 4.109452450057071e-05, "loss": 0.2965, "step": 14538 }, { "epoch": 1.5, "learning_rate": 4.109223754683187e-05, "loss": 0.9113, "step": 14539 }, { "epoch": 1.5, "learning_rate": 4.1089950518422685e-05, "loss": 0.5843, "step": 14540 }, { "epoch": 1.5, "learning_rate": 4.108766341535854e-05, "loss": 0.5533, "step": 14541 }, { "epoch": 1.5, "learning_rate": 4.1085376237654845e-05, "loss": 0.5092, "step": 14542 }, { "epoch": 1.5, "learning_rate": 4.108308898532698e-05, "loss": 0.5089, "step": 14543 }, { "epoch": 1.5, "learning_rate": 4.108080165839036e-05, "loss": 0.4478, "step": 14544 }, { "epoch": 1.5, "learning_rate": 4.1078514256860385e-05, "loss": 0.3342, "step": 14545 }, { "epoch": 1.5, "learning_rate": 4.107622678075243e-05, "loss": 0.49, "step": 14546 }, { "epoch": 1.5, "learning_rate": 4.107393923008191e-05, "loss": 0.311, "step": 14547 }, { "epoch": 1.5, "learning_rate": 4.107165160486423e-05, "loss": 0.227, "step": 14548 }, { "epoch": 1.5, "learning_rate": 4.106936390511478e-05, "loss": 0.2464, "step": 14549 }, { "epoch": 1.5, "learning_rate": 4.106707613084896e-05, "loss": 0.3388, "step": 14550 }, { "epoch": 1.5, "learning_rate": 4.106478828208217e-05, "loss": 1.1112, "step": 14551 }, { "epoch": 1.5, "learning_rate": 4.1062500358829816e-05, "loss": 0.8193, "step": 14552 }, { "epoch": 1.5, "learning_rate": 4.106021236110729e-05, "loss": 0.7852, "step": 14553 }, { "epoch": 1.5, "learning_rate": 4.1057924288930024e-05, "loss": 0.5059, "step": 14554 }, { "epoch": 1.5, "learning_rate": 4.1055636142313386e-05, "loss": 0.8447, "step": 14555 }, { "epoch": 1.5, "learning_rate": 4.105334792127279e-05, "loss": 1.1657, "step": 14556 }, { "epoch": 1.5, "learning_rate": 4.1051059625823644e-05, "loss": 1.1604, "step": 14557 }, { "epoch": 1.5, "learning_rate": 4.104877125598135e-05, "loss": 0.1142, "step": 14558 }, { "epoch": 1.51, "learning_rate": 4.1046482811761314e-05, "loss": 0.7218, "step": 14559 }, { "epoch": 1.51, "learning_rate": 4.104419429317894e-05, "loss": 0.5873, "step": 14560 }, { "epoch": 1.51, "learning_rate": 4.1041905700249643e-05, "loss": 0.4121, "step": 14561 }, { "epoch": 1.51, "learning_rate": 4.103961703298881e-05, "loss": 0.2559, "step": 14562 }, { "epoch": 1.51, "learning_rate": 4.103732829141185e-05, "loss": 0.8949, "step": 14563 }, { "epoch": 1.51, "learning_rate": 4.103503947553419e-05, "loss": 0.1912, "step": 14564 }, { "epoch": 1.51, "learning_rate": 4.103275058537123e-05, "loss": 0.3315, "step": 14565 }, { "epoch": 1.51, "learning_rate": 4.1030461620938366e-05, "loss": 0.17, "step": 14566 }, { "epoch": 1.51, "learning_rate": 4.102817258225102e-05, "loss": 0.5596, "step": 14567 }, { "epoch": 1.51, "learning_rate": 4.1025883469324595e-05, "loss": 0.7507, "step": 14568 }, { "epoch": 1.51, "learning_rate": 4.1023594282174507e-05, "loss": 0.6137, "step": 14569 }, { "epoch": 1.51, "learning_rate": 4.102130502081616e-05, "loss": 0.4934, "step": 14570 }, { "epoch": 1.51, "learning_rate": 4.101901568526496e-05, "loss": 0.4124, "step": 14571 }, { "epoch": 1.51, "learning_rate": 4.101672627553632e-05, "loss": 0.9907, "step": 14572 }, { "epoch": 1.51, "learning_rate": 4.1014436791645676e-05, "loss": 0.3102, "step": 14573 }, { "epoch": 1.51, "learning_rate": 4.1012147233608416e-05, "loss": 0.407, "step": 14574 }, { "epoch": 1.51, "learning_rate": 4.100985760143994e-05, "loss": 1.0613, "step": 14575 }, { "epoch": 1.51, "learning_rate": 4.100756789515569e-05, "loss": 1.1287, "step": 14576 }, { "epoch": 1.51, "learning_rate": 4.1005278114771065e-05, "loss": 1.1948, "step": 14577 }, { "epoch": 1.51, "learning_rate": 4.1002988260301486e-05, "loss": 1.0196, "step": 14578 }, { "epoch": 1.51, "learning_rate": 4.1000698331762365e-05, "loss": 0.2481, "step": 14579 }, { "epoch": 1.51, "learning_rate": 4.099840832916912e-05, "loss": 0.3706, "step": 14580 }, { "epoch": 1.51, "learning_rate": 4.0996118252537155e-05, "loss": 0.7962, "step": 14581 }, { "epoch": 1.51, "learning_rate": 4.0993828101881906e-05, "loss": 0.529, "step": 14582 }, { "epoch": 1.51, "learning_rate": 4.099153787721877e-05, "loss": 0.899, "step": 14583 }, { "epoch": 1.51, "learning_rate": 4.098924757856318e-05, "loss": 0.352, "step": 14584 }, { "epoch": 1.51, "learning_rate": 4.0986957205930536e-05, "loss": 1.2908, "step": 14585 }, { "epoch": 1.51, "learning_rate": 4.098466675933627e-05, "loss": 0.3634, "step": 14586 }, { "epoch": 1.51, "learning_rate": 4.09823762387958e-05, "loss": 0.7514, "step": 14587 }, { "epoch": 1.51, "learning_rate": 4.098008564432455e-05, "loss": 0.324, "step": 14588 }, { "epoch": 1.51, "learning_rate": 4.097779497593792e-05, "loss": 0.3572, "step": 14589 }, { "epoch": 1.51, "learning_rate": 4.0975504233651354e-05, "loss": 0.9337, "step": 14590 }, { "epoch": 1.51, "learning_rate": 4.0973213417480244e-05, "loss": 1.2184, "step": 14591 }, { "epoch": 1.51, "learning_rate": 4.097092252744004e-05, "loss": 0.6182, "step": 14592 }, { "epoch": 1.51, "learning_rate": 4.096863156354617e-05, "loss": 0.1624, "step": 14593 }, { "epoch": 1.51, "learning_rate": 4.0966340525814015e-05, "loss": 0.7043, "step": 14594 }, { "epoch": 1.51, "learning_rate": 4.0964049414259024e-05, "loss": 0.8982, "step": 14595 }, { "epoch": 1.51, "learning_rate": 4.0961758228896624e-05, "loss": 0.2801, "step": 14596 }, { "epoch": 1.51, "learning_rate": 4.095946696974223e-05, "loss": 0.4733, "step": 14597 }, { "epoch": 1.51, "learning_rate": 4.095717563681127e-05, "loss": 0.986, "step": 14598 }, { "epoch": 1.51, "learning_rate": 4.095488423011917e-05, "loss": 0.4369, "step": 14599 }, { "epoch": 1.51, "learning_rate": 4.0952592749681346e-05, "loss": 0.3084, "step": 14600 }, { "epoch": 1.51, "learning_rate": 4.095030119551323e-05, "loss": 1.4172, "step": 14601 }, { "epoch": 1.51, "learning_rate": 4.094800956763026e-05, "loss": 0.1809, "step": 14602 }, { "epoch": 1.51, "learning_rate": 4.094571786604784e-05, "loss": 1.2353, "step": 14603 }, { "epoch": 1.51, "learning_rate": 4.094342609078141e-05, "loss": 0.272, "step": 14604 }, { "epoch": 1.51, "learning_rate": 4.09411342418464e-05, "loss": 0.3684, "step": 14605 }, { "epoch": 1.51, "learning_rate": 4.0938842319258234e-05, "loss": 0.1905, "step": 14606 }, { "epoch": 1.51, "learning_rate": 4.093655032303234e-05, "loss": 0.3132, "step": 14607 }, { "epoch": 1.51, "learning_rate": 4.093425825318414e-05, "loss": 0.3231, "step": 14608 }, { "epoch": 1.51, "learning_rate": 4.093196610972908e-05, "loss": 0.6503, "step": 14609 }, { "epoch": 1.51, "learning_rate": 4.092967389268258e-05, "loss": 0.5732, "step": 14610 }, { "epoch": 1.51, "learning_rate": 4.0927381602060074e-05, "loss": 1.0068, "step": 14611 }, { "epoch": 1.51, "learning_rate": 4.0925089237877e-05, "loss": 0.3798, "step": 14612 }, { "epoch": 1.51, "learning_rate": 4.092279680014877e-05, "loss": 0.3608, "step": 14613 }, { "epoch": 1.51, "learning_rate": 4.092050428889083e-05, "loss": 1.1389, "step": 14614 }, { "epoch": 1.51, "learning_rate": 4.091821170411861e-05, "loss": 0.4749, "step": 14615 }, { "epoch": 1.51, "learning_rate": 4.0915919045847546e-05, "loss": 0.51, "step": 14616 }, { "epoch": 1.51, "learning_rate": 4.0913626314093077e-05, "loss": 0.576, "step": 14617 }, { "epoch": 1.51, "learning_rate": 4.0911333508870614e-05, "loss": 0.2748, "step": 14618 }, { "epoch": 1.51, "learning_rate": 4.090904063019561e-05, "loss": 0.7723, "step": 14619 }, { "epoch": 1.51, "learning_rate": 4.09067476780835e-05, "loss": 0.3937, "step": 14620 }, { "epoch": 1.51, "learning_rate": 4.090445465254971e-05, "loss": 0.8065, "step": 14621 }, { "epoch": 1.51, "learning_rate": 4.090216155360969e-05, "loss": 0.2497, "step": 14622 }, { "epoch": 1.51, "learning_rate": 4.089986838127887e-05, "loss": 0.3159, "step": 14623 }, { "epoch": 1.51, "learning_rate": 4.089757513557269e-05, "loss": 0.892, "step": 14624 }, { "epoch": 1.51, "learning_rate": 4.089528181650658e-05, "loss": 0.3498, "step": 14625 }, { "epoch": 1.51, "learning_rate": 4.0892988424095976e-05, "loss": 0.4068, "step": 14626 }, { "epoch": 1.51, "learning_rate": 4.089069495835633e-05, "loss": 0.9426, "step": 14627 }, { "epoch": 1.51, "learning_rate": 4.088840141930306e-05, "loss": 0.6866, "step": 14628 }, { "epoch": 1.51, "learning_rate": 4.088610780695163e-05, "loss": 0.9486, "step": 14629 }, { "epoch": 1.51, "learning_rate": 4.088381412131748e-05, "loss": 0.5877, "step": 14630 }, { "epoch": 1.51, "learning_rate": 4.0881520362416036e-05, "loss": 0.2392, "step": 14631 }, { "epoch": 1.51, "learning_rate": 4.0879226530262735e-05, "loss": 0.172, "step": 14632 }, { "epoch": 1.51, "learning_rate": 4.0876932624873026e-05, "loss": 1.066, "step": 14633 }, { "epoch": 1.51, "learning_rate": 4.087463864626235e-05, "loss": 0.7894, "step": 14634 }, { "epoch": 1.51, "learning_rate": 4.087234459444616e-05, "loss": 0.0873, "step": 14635 }, { "epoch": 1.51, "learning_rate": 4.087005046943989e-05, "loss": 0.4165, "step": 14636 }, { "epoch": 1.51, "learning_rate": 4.0867756271258976e-05, "loss": 0.8902, "step": 14637 }, { "epoch": 1.51, "learning_rate": 4.0865461999918874e-05, "loss": 0.9719, "step": 14638 }, { "epoch": 1.51, "learning_rate": 4.086316765543502e-05, "loss": 0.4862, "step": 14639 }, { "epoch": 1.51, "learning_rate": 4.086087323782287e-05, "loss": 0.2905, "step": 14640 }, { "epoch": 1.51, "learning_rate": 4.0858578747097854e-05, "loss": 0.5111, "step": 14641 }, { "epoch": 1.51, "learning_rate": 4.085628418327544e-05, "loss": 0.6967, "step": 14642 }, { "epoch": 1.51, "learning_rate": 4.085398954637104e-05, "loss": 0.6312, "step": 14643 }, { "epoch": 1.51, "learning_rate": 4.085169483640013e-05, "loss": 1.3531, "step": 14644 }, { "epoch": 1.51, "learning_rate": 4.084940005337817e-05, "loss": 1.017, "step": 14645 }, { "epoch": 1.51, "learning_rate": 4.0847105197320565e-05, "loss": 1.019, "step": 14646 }, { "epoch": 1.51, "learning_rate": 4.0844810268242784e-05, "loss": 1.2747, "step": 14647 }, { "epoch": 1.51, "learning_rate": 4.084251526616029e-05, "loss": 0.5013, "step": 14648 }, { "epoch": 1.51, "learning_rate": 4.0840220191088516e-05, "loss": 0.7241, "step": 14649 }, { "epoch": 1.51, "learning_rate": 4.083792504304293e-05, "loss": 0.8475, "step": 14650 }, { "epoch": 1.51, "learning_rate": 4.083562982203894e-05, "loss": 0.3343, "step": 14651 }, { "epoch": 1.51, "learning_rate": 4.083333452809204e-05, "loss": 0.1667, "step": 14652 }, { "epoch": 1.51, "learning_rate": 4.083103916121767e-05, "loss": 0.4058, "step": 14653 }, { "epoch": 1.51, "learning_rate": 4.082874372143127e-05, "loss": 0.9027, "step": 14654 }, { "epoch": 1.52, "learning_rate": 4.082644820874832e-05, "loss": 0.2478, "step": 14655 }, { "epoch": 1.52, "learning_rate": 4.082415262318423e-05, "loss": 0.4635, "step": 14656 }, { "epoch": 1.52, "learning_rate": 4.082185696475449e-05, "loss": 0.3701, "step": 14657 }, { "epoch": 1.52, "learning_rate": 4.081956123347453e-05, "loss": 0.481, "step": 14658 }, { "epoch": 1.52, "learning_rate": 4.081726542935983e-05, "loss": 0.5754, "step": 14659 }, { "epoch": 1.52, "learning_rate": 4.081496955242582e-05, "loss": 0.3882, "step": 14660 }, { "epoch": 1.52, "learning_rate": 4.081267360268798e-05, "loss": 0.8057, "step": 14661 }, { "epoch": 1.52, "learning_rate": 4.0810377580161734e-05, "loss": 0.2742, "step": 14662 }, { "epoch": 1.52, "learning_rate": 4.080808148486256e-05, "loss": 1.0291, "step": 14663 }, { "epoch": 1.52, "learning_rate": 4.0805785316805915e-05, "loss": 0.6196, "step": 14664 }, { "epoch": 1.52, "learning_rate": 4.080348907600725e-05, "loss": 0.4197, "step": 14665 }, { "epoch": 1.52, "learning_rate": 4.080119276248202e-05, "loss": 0.5013, "step": 14666 }, { "epoch": 1.52, "learning_rate": 4.079889637624569e-05, "loss": 0.3127, "step": 14667 }, { "epoch": 1.52, "learning_rate": 4.0796599917313716e-05, "loss": 0.6683, "step": 14668 }, { "epoch": 1.52, "learning_rate": 4.079430338570157e-05, "loss": 0.2136, "step": 14669 }, { "epoch": 1.52, "learning_rate": 4.079200678142468e-05, "loss": 0.2547, "step": 14670 }, { "epoch": 1.52, "learning_rate": 4.078971010449854e-05, "loss": 0.9456, "step": 14671 }, { "epoch": 1.52, "learning_rate": 4.0787413354938576e-05, "loss": 0.8638, "step": 14672 }, { "epoch": 1.52, "learning_rate": 4.0785116532760285e-05, "loss": 1.5428, "step": 14673 }, { "epoch": 1.52, "learning_rate": 4.078281963797912e-05, "loss": 0.3707, "step": 14674 }, { "epoch": 1.52, "learning_rate": 4.078052267061053e-05, "loss": 0.8183, "step": 14675 }, { "epoch": 1.52, "learning_rate": 4.077822563066997e-05, "loss": 0.0991, "step": 14676 }, { "epoch": 1.52, "learning_rate": 4.077592851817293e-05, "loss": 0.3866, "step": 14677 }, { "epoch": 1.52, "learning_rate": 4.077363133313486e-05, "loss": 1.7873, "step": 14678 }, { "epoch": 1.52, "learning_rate": 4.077133407557123e-05, "loss": 0.8778, "step": 14679 }, { "epoch": 1.52, "learning_rate": 4.076903674549749e-05, "loss": 0.6512, "step": 14680 }, { "epoch": 1.52, "learning_rate": 4.076673934292911e-05, "loss": 1.0734, "step": 14681 }, { "epoch": 1.52, "learning_rate": 4.076444186788157e-05, "loss": 0.8848, "step": 14682 }, { "epoch": 1.52, "learning_rate": 4.076214432037033e-05, "loss": 0.7644, "step": 14683 }, { "epoch": 1.52, "learning_rate": 4.075984670041084e-05, "loss": 0.2534, "step": 14684 }, { "epoch": 1.52, "learning_rate": 4.075754900801858e-05, "loss": 0.3337, "step": 14685 }, { "epoch": 1.52, "learning_rate": 4.0755251243209036e-05, "loss": 1.1667, "step": 14686 }, { "epoch": 1.52, "learning_rate": 4.0752953405997636e-05, "loss": 1.0383, "step": 14687 }, { "epoch": 1.52, "learning_rate": 4.0750655496399884e-05, "loss": 0.564, "step": 14688 }, { "epoch": 1.52, "learning_rate": 4.0748357514431224e-05, "loss": 0.1105, "step": 14689 }, { "epoch": 1.52, "learning_rate": 4.074605946010713e-05, "loss": 0.3559, "step": 14690 }, { "epoch": 1.52, "learning_rate": 4.074376133344309e-05, "loss": 0.9287, "step": 14691 }, { "epoch": 1.52, "learning_rate": 4.074146313445456e-05, "loss": 1.0548, "step": 14692 }, { "epoch": 1.52, "learning_rate": 4.0739164863157025e-05, "loss": 1.218, "step": 14693 }, { "epoch": 1.52, "learning_rate": 4.073686651956593e-05, "loss": 0.6213, "step": 14694 }, { "epoch": 1.52, "learning_rate": 4.073456810369676e-05, "loss": 1.0144, "step": 14695 }, { "epoch": 1.52, "learning_rate": 4.0732269615565e-05, "loss": 0.4806, "step": 14696 }, { "epoch": 1.52, "learning_rate": 4.0729971055186106e-05, "loss": 0.2857, "step": 14697 }, { "epoch": 1.52, "learning_rate": 4.072767242257556e-05, "loss": 1.8651, "step": 14698 }, { "epoch": 1.52, "learning_rate": 4.072537371774883e-05, "loss": 0.2779, "step": 14699 }, { "epoch": 1.52, "learning_rate": 4.07230749407214e-05, "loss": 0.2574, "step": 14700 }, { "epoch": 1.52, "learning_rate": 4.072077609150873e-05, "loss": 0.3801, "step": 14701 }, { "epoch": 1.52, "learning_rate": 4.0718477170126314e-05, "loss": 0.3459, "step": 14702 }, { "epoch": 1.52, "learning_rate": 4.0716178176589606e-05, "loss": 0.4175, "step": 14703 }, { "epoch": 1.52, "learning_rate": 4.07138791109141e-05, "loss": 0.8027, "step": 14704 }, { "epoch": 1.52, "learning_rate": 4.0711579973115266e-05, "loss": 0.4925, "step": 14705 }, { "epoch": 1.52, "learning_rate": 4.070928076320857e-05, "loss": 0.6905, "step": 14706 }, { "epoch": 1.52, "learning_rate": 4.0706981481209524e-05, "loss": 0.3384, "step": 14707 }, { "epoch": 1.52, "learning_rate": 4.0704682127133575e-05, "loss": 1.1136, "step": 14708 }, { "epoch": 1.52, "learning_rate": 4.070238270099621e-05, "loss": 1.1165, "step": 14709 }, { "epoch": 1.52, "learning_rate": 4.07000832028129e-05, "loss": 0.8789, "step": 14710 }, { "epoch": 1.52, "learning_rate": 4.0697783632599146e-05, "loss": 0.7699, "step": 14711 }, { "epoch": 1.52, "learning_rate": 4.0695483990370414e-05, "loss": 0.8614, "step": 14712 }, { "epoch": 1.52, "learning_rate": 4.069318427614219e-05, "loss": 0.5557, "step": 14713 }, { "epoch": 1.52, "learning_rate": 4.069088448992993e-05, "loss": 1.1062, "step": 14714 }, { "epoch": 1.52, "learning_rate": 4.0688584631749154e-05, "loss": 0.2147, "step": 14715 }, { "epoch": 1.52, "learning_rate": 4.0686284701615334e-05, "loss": 1.0833, "step": 14716 }, { "epoch": 1.52, "learning_rate": 4.068398469954393e-05, "loss": 0.0696, "step": 14717 }, { "epoch": 1.52, "learning_rate": 4.068168462555046e-05, "loss": 0.3975, "step": 14718 }, { "epoch": 1.52, "learning_rate": 4.067938447965037e-05, "loss": 1.5666, "step": 14719 }, { "epoch": 1.52, "learning_rate": 4.067708426185916e-05, "loss": 0.7433, "step": 14720 }, { "epoch": 1.52, "learning_rate": 4.067478397219234e-05, "loss": 1.5877, "step": 14721 }, { "epoch": 1.52, "learning_rate": 4.0672483610665355e-05, "loss": 0.5392, "step": 14722 }, { "epoch": 1.52, "learning_rate": 4.06701831772937e-05, "loss": 0.4778, "step": 14723 }, { "epoch": 1.52, "learning_rate": 4.066788267209289e-05, "loss": 0.5075, "step": 14724 }, { "epoch": 1.52, "learning_rate": 4.0665582095078374e-05, "loss": 0.2589, "step": 14725 }, { "epoch": 1.52, "learning_rate": 4.066328144626566e-05, "loss": 0.8529, "step": 14726 }, { "epoch": 1.52, "learning_rate": 4.066098072567023e-05, "loss": 0.2627, "step": 14727 }, { "epoch": 1.52, "learning_rate": 4.065867993330757e-05, "loss": 0.9375, "step": 14728 }, { "epoch": 1.52, "learning_rate": 4.065637906919317e-05, "loss": 0.3832, "step": 14729 }, { "epoch": 1.52, "learning_rate": 4.065407813334252e-05, "loss": 0.3093, "step": 14730 }, { "epoch": 1.52, "learning_rate": 4.0651777125771124e-05, "loss": 0.3978, "step": 14731 }, { "epoch": 1.52, "learning_rate": 4.064947604649444e-05, "loss": 0.9458, "step": 14732 }, { "epoch": 1.52, "learning_rate": 4.064717489552798e-05, "loss": 0.4849, "step": 14733 }, { "epoch": 1.52, "learning_rate": 4.0644873672887216e-05, "loss": 1.0067, "step": 14734 }, { "epoch": 1.52, "learning_rate": 4.064257237858768e-05, "loss": 0.5896, "step": 14735 }, { "epoch": 1.52, "learning_rate": 4.064027101264481e-05, "loss": 0.8662, "step": 14736 }, { "epoch": 1.52, "learning_rate": 4.063796957507414e-05, "loss": 0.7302, "step": 14737 }, { "epoch": 1.52, "learning_rate": 4.063566806589115e-05, "loss": 1.427, "step": 14738 }, { "epoch": 1.52, "learning_rate": 4.0633366485111324e-05, "loss": 0.5597, "step": 14739 }, { "epoch": 1.52, "learning_rate": 4.063106483275017e-05, "loss": 0.2892, "step": 14740 }, { "epoch": 1.52, "learning_rate": 4.0628763108823166e-05, "loss": 0.9863, "step": 14741 }, { "epoch": 1.52, "learning_rate": 4.062646131334582e-05, "loss": 0.754, "step": 14742 }, { "epoch": 1.52, "learning_rate": 4.062415944633363e-05, "loss": 0.3804, "step": 14743 }, { "epoch": 1.52, "learning_rate": 4.0621857507802066e-05, "loss": 0.3214, "step": 14744 }, { "epoch": 1.52, "learning_rate": 4.061955549776666e-05, "loss": 0.3169, "step": 14745 }, { "epoch": 1.52, "learning_rate": 4.0617253416242886e-05, "loss": 1.1259, "step": 14746 }, { "epoch": 1.52, "learning_rate": 4.061495126324624e-05, "loss": 1.2325, "step": 14747 }, { "epoch": 1.52, "learning_rate": 4.061264903879224e-05, "loss": 0.9936, "step": 14748 }, { "epoch": 1.52, "learning_rate": 4.061034674289636e-05, "loss": 0.7191, "step": 14749 }, { "epoch": 1.53, "learning_rate": 4.060804437557411e-05, "loss": 0.6088, "step": 14750 }, { "epoch": 1.53, "learning_rate": 4.060574193684099e-05, "loss": 0.446, "step": 14751 }, { "epoch": 1.53, "learning_rate": 4.060343942671249e-05, "loss": 0.4811, "step": 14752 }, { "epoch": 1.53, "learning_rate": 4.0601136845204113e-05, "loss": 0.4356, "step": 14753 }, { "epoch": 1.53, "learning_rate": 4.0598834192331375e-05, "loss": 0.4428, "step": 14754 }, { "epoch": 1.53, "learning_rate": 4.059653146810976e-05, "loss": 0.1636, "step": 14755 }, { "epoch": 1.53, "learning_rate": 4.059422867255478e-05, "loss": 0.5988, "step": 14756 }, { "epoch": 1.53, "learning_rate": 4.0591925805681925e-05, "loss": 1.1889, "step": 14757 }, { "epoch": 1.53, "learning_rate": 4.0589622867506705e-05, "loss": 0.8233, "step": 14758 }, { "epoch": 1.53, "learning_rate": 4.058731985804462e-05, "loss": 1.002, "step": 14759 }, { "epoch": 1.53, "learning_rate": 4.058501677731118e-05, "loss": 0.6324, "step": 14760 }, { "epoch": 1.53, "learning_rate": 4.0582713625321886e-05, "loss": 0.3212, "step": 14761 }, { "epoch": 1.53, "learning_rate": 4.0580410402092235e-05, "loss": 0.7198, "step": 14762 }, { "epoch": 1.53, "learning_rate": 4.057810710763773e-05, "loss": 0.976, "step": 14763 }, { "epoch": 1.53, "learning_rate": 4.05758037419739e-05, "loss": 0.3433, "step": 14764 }, { "epoch": 1.53, "learning_rate": 4.0573500305116216e-05, "loss": 0.6419, "step": 14765 }, { "epoch": 1.53, "learning_rate": 4.057119679708021e-05, "loss": 0.236, "step": 14766 }, { "epoch": 1.53, "learning_rate": 4.0568893217881375e-05, "loss": 1.0251, "step": 14767 }, { "epoch": 1.53, "learning_rate": 4.056658956753524e-05, "loss": 0.4142, "step": 14768 }, { "epoch": 1.53, "learning_rate": 4.056428584605729e-05, "loss": 0.5144, "step": 14769 }, { "epoch": 1.53, "learning_rate": 4.0561982053463025e-05, "loss": 0.1648, "step": 14770 }, { "epoch": 1.53, "learning_rate": 4.055967818976798e-05, "loss": 0.484, "step": 14771 }, { "epoch": 1.53, "learning_rate": 4.055737425498764e-05, "loss": 0.3322, "step": 14772 }, { "epoch": 1.53, "learning_rate": 4.055507024913754e-05, "loss": 0.0897, "step": 14773 }, { "epoch": 1.53, "learning_rate": 4.055276617223318e-05, "loss": 0.2635, "step": 14774 }, { "epoch": 1.53, "learning_rate": 4.0550462024290066e-05, "loss": 0.6366, "step": 14775 }, { "epoch": 1.53, "learning_rate": 4.05481578053237e-05, "loss": 1.4488, "step": 14776 }, { "epoch": 1.53, "learning_rate": 4.05458535153496e-05, "loss": 0.3825, "step": 14777 }, { "epoch": 1.53, "learning_rate": 4.05435491543833e-05, "loss": 0.1247, "step": 14778 }, { "epoch": 1.53, "learning_rate": 4.054124472244029e-05, "loss": 1.0014, "step": 14779 }, { "epoch": 1.53, "learning_rate": 4.0538940219536075e-05, "loss": 0.6081, "step": 14780 }, { "epoch": 1.53, "learning_rate": 4.053663564568619e-05, "loss": 0.808, "step": 14781 }, { "epoch": 1.53, "learning_rate": 4.053433100090613e-05, "loss": 0.7508, "step": 14782 }, { "epoch": 1.53, "learning_rate": 4.053202628521143e-05, "loss": 0.6373, "step": 14783 }, { "epoch": 1.53, "learning_rate": 4.052972149861759e-05, "loss": 0.6379, "step": 14784 }, { "epoch": 1.53, "learning_rate": 4.052741664114012e-05, "loss": 0.9295, "step": 14785 }, { "epoch": 1.53, "learning_rate": 4.052511171279455e-05, "loss": 0.3507, "step": 14786 }, { "epoch": 1.53, "learning_rate": 4.05228067135964e-05, "loss": 0.4591, "step": 14787 }, { "epoch": 1.53, "learning_rate": 4.0520501643561175e-05, "loss": 0.5399, "step": 14788 }, { "epoch": 1.53, "learning_rate": 4.051819650270438e-05, "loss": 0.7702, "step": 14789 }, { "epoch": 1.53, "learning_rate": 4.0515891291041556e-05, "loss": 0.9529, "step": 14790 }, { "epoch": 1.53, "learning_rate": 4.051358600858821e-05, "loss": 0.0736, "step": 14791 }, { "epoch": 1.53, "learning_rate": 4.051128065535986e-05, "loss": 0.6816, "step": 14792 }, { "epoch": 1.53, "learning_rate": 4.050897523137205e-05, "loss": 0.746, "step": 14793 }, { "epoch": 1.53, "learning_rate": 4.050666973664026e-05, "loss": 0.9845, "step": 14794 }, { "epoch": 1.53, "learning_rate": 4.050436417118002e-05, "loss": 0.4096, "step": 14795 }, { "epoch": 1.53, "learning_rate": 4.0502058535006865e-05, "loss": 0.2396, "step": 14796 }, { "epoch": 1.53, "learning_rate": 4.0499752828136317e-05, "loss": 0.1554, "step": 14797 }, { "epoch": 1.53, "learning_rate": 4.0497447050583884e-05, "loss": 0.7721, "step": 14798 }, { "epoch": 1.53, "learning_rate": 4.049514120236509e-05, "loss": 0.4617, "step": 14799 }, { "epoch": 1.53, "learning_rate": 4.049283528349547e-05, "loss": 0.5695, "step": 14800 }, { "epoch": 1.53, "learning_rate": 4.049052929399053e-05, "loss": 0.7732, "step": 14801 }, { "epoch": 1.53, "learning_rate": 4.04882232338658e-05, "loss": 0.4168, "step": 14802 }, { "epoch": 1.53, "learning_rate": 4.048591710313681e-05, "loss": 0.8111, "step": 14803 }, { "epoch": 1.53, "learning_rate": 4.048361090181908e-05, "loss": 0.8113, "step": 14804 }, { "epoch": 1.53, "learning_rate": 4.048130462992813e-05, "loss": 0.5249, "step": 14805 }, { "epoch": 1.53, "learning_rate": 4.04789982874795e-05, "loss": 0.5103, "step": 14806 }, { "epoch": 1.53, "learning_rate": 4.0476691874488694e-05, "loss": 0.8446, "step": 14807 }, { "epoch": 1.53, "learning_rate": 4.047438539097126e-05, "loss": 0.4546, "step": 14808 }, { "epoch": 1.53, "learning_rate": 4.04720788369427e-05, "loss": 0.4056, "step": 14809 }, { "epoch": 1.53, "learning_rate": 4.0469772212418563e-05, "loss": 0.6234, "step": 14810 }, { "epoch": 1.53, "learning_rate": 4.046746551741436e-05, "loss": 0.5223, "step": 14811 }, { "epoch": 1.53, "learning_rate": 4.046515875194566e-05, "loss": 0.5848, "step": 14812 }, { "epoch": 1.53, "learning_rate": 4.046285191602793e-05, "loss": 0.3078, "step": 14813 }, { "epoch": 1.53, "learning_rate": 4.0460545009676735e-05, "loss": 0.6794, "step": 14814 }, { "epoch": 1.53, "learning_rate": 4.04582380329076e-05, "loss": 0.859, "step": 14815 }, { "epoch": 1.53, "learning_rate": 4.045593098573606e-05, "loss": 0.3806, "step": 14816 }, { "epoch": 1.53, "learning_rate": 4.0453623868177625e-05, "loss": 1.0143, "step": 14817 }, { "epoch": 1.53, "learning_rate": 4.045131668024785e-05, "loss": 0.283, "step": 14818 }, { "epoch": 1.53, "learning_rate": 4.044900942196226e-05, "loss": 0.7289, "step": 14819 }, { "epoch": 1.53, "learning_rate": 4.0446702093336375e-05, "loss": 0.7263, "step": 14820 }, { "epoch": 1.53, "learning_rate": 4.0444394694385744e-05, "loss": 1.1192, "step": 14821 }, { "epoch": 1.53, "learning_rate": 4.044208722512589e-05, "loss": 0.4687, "step": 14822 }, { "epoch": 1.53, "learning_rate": 4.0439779685572334e-05, "loss": 0.197, "step": 14823 }, { "epoch": 1.53, "learning_rate": 4.043747207574063e-05, "loss": 0.7639, "step": 14824 }, { "epoch": 1.53, "learning_rate": 4.043516439564633e-05, "loss": 0.5072, "step": 14825 }, { "epoch": 1.53, "learning_rate": 4.0432856645304926e-05, "loss": 0.5006, "step": 14826 }, { "epoch": 1.53, "learning_rate": 4.043054882473196e-05, "loss": 0.2261, "step": 14827 }, { "epoch": 1.53, "learning_rate": 4.0428240933942996e-05, "loss": 1.1748, "step": 14828 }, { "epoch": 1.53, "learning_rate": 4.042593297295355e-05, "loss": 0.5282, "step": 14829 }, { "epoch": 1.53, "learning_rate": 4.0423624941779155e-05, "loss": 0.2408, "step": 14830 }, { "epoch": 1.53, "learning_rate": 4.042131684043538e-05, "loss": 0.9544, "step": 14831 }, { "epoch": 1.53, "learning_rate": 4.041900866893772e-05, "loss": 0.5259, "step": 14832 }, { "epoch": 1.53, "learning_rate": 4.041670042730173e-05, "loss": 0.6025, "step": 14833 }, { "epoch": 1.53, "learning_rate": 4.041439211554295e-05, "loss": 0.6387, "step": 14834 }, { "epoch": 1.53, "learning_rate": 4.0412083733676926e-05, "loss": 0.8678, "step": 14835 }, { "epoch": 1.53, "learning_rate": 4.040977528171919e-05, "loss": 0.5994, "step": 14836 }, { "epoch": 1.53, "learning_rate": 4.0407466759685286e-05, "loss": 0.2958, "step": 14837 }, { "epoch": 1.53, "learning_rate": 4.0405158167590736e-05, "loss": 0.5553, "step": 14838 }, { "epoch": 1.53, "learning_rate": 4.04028495054511e-05, "loss": 0.4819, "step": 14839 }, { "epoch": 1.53, "learning_rate": 4.040054077328193e-05, "loss": 0.9824, "step": 14840 }, { "epoch": 1.53, "learning_rate": 4.039823197109874e-05, "loss": 0.2962, "step": 14841 }, { "epoch": 1.53, "learning_rate": 4.0395923098917084e-05, "loss": 1.2085, "step": 14842 }, { "epoch": 1.53, "learning_rate": 4.039361415675252e-05, "loss": 0.6908, "step": 14843 }, { "epoch": 1.53, "learning_rate": 4.039130514462056e-05, "loss": 0.1993, "step": 14844 }, { "epoch": 1.53, "learning_rate": 4.038899606253677e-05, "loss": 0.9344, "step": 14845 }, { "epoch": 1.54, "learning_rate": 4.03866869105167e-05, "loss": 0.8842, "step": 14846 }, { "epoch": 1.54, "learning_rate": 4.038437768857587e-05, "loss": 0.5707, "step": 14847 }, { "epoch": 1.54, "learning_rate": 4.038206839672984e-05, "loss": 0.5294, "step": 14848 }, { "epoch": 1.54, "learning_rate": 4.037975903499416e-05, "loss": 0.1007, "step": 14849 }, { "epoch": 1.54, "learning_rate": 4.037744960338438e-05, "loss": 0.663, "step": 14850 }, { "epoch": 1.54, "learning_rate": 4.037514010191603e-05, "loss": 0.8012, "step": 14851 }, { "epoch": 1.54, "learning_rate": 4.0372830530604657e-05, "loss": 0.2607, "step": 14852 }, { "epoch": 1.54, "learning_rate": 4.037052088946582e-05, "loss": 0.4576, "step": 14853 }, { "epoch": 1.54, "learning_rate": 4.036821117851507e-05, "loss": 1.0014, "step": 14854 }, { "epoch": 1.54, "learning_rate": 4.036590139776794e-05, "loss": 0.6972, "step": 14855 }, { "epoch": 1.54, "learning_rate": 4.036359154723999e-05, "loss": 1.2911, "step": 14856 }, { "epoch": 1.54, "learning_rate": 4.036128162694677e-05, "loss": 0.6036, "step": 14857 }, { "epoch": 1.54, "learning_rate": 4.0358971636903824e-05, "loss": 0.239, "step": 14858 }, { "epoch": 1.54, "learning_rate": 4.0356661577126704e-05, "loss": 0.733, "step": 14859 }, { "epoch": 1.54, "learning_rate": 4.035435144763096e-05, "loss": 0.2717, "step": 14860 }, { "epoch": 1.54, "learning_rate": 4.035204124843214e-05, "loss": 0.9058, "step": 14861 }, { "epoch": 1.54, "learning_rate": 4.034973097954582e-05, "loss": 1.0072, "step": 14862 }, { "epoch": 1.54, "learning_rate": 4.034742064098751e-05, "loss": 0.5888, "step": 14863 }, { "epoch": 1.54, "learning_rate": 4.0345110232772794e-05, "loss": 0.8478, "step": 14864 }, { "epoch": 1.54, "learning_rate": 4.034279975491722e-05, "loss": 0.2297, "step": 14865 }, { "epoch": 1.54, "learning_rate": 4.034048920743633e-05, "loss": 0.5091, "step": 14866 }, { "epoch": 1.54, "learning_rate": 4.033817859034569e-05, "loss": 0.7732, "step": 14867 }, { "epoch": 1.54, "learning_rate": 4.033586790366085e-05, "loss": 1.3875, "step": 14868 }, { "epoch": 1.54, "learning_rate": 4.033355714739738e-05, "loss": 0.7272, "step": 14869 }, { "epoch": 1.54, "learning_rate": 4.0331246321570804e-05, "loss": 0.9555, "step": 14870 }, { "epoch": 1.54, "learning_rate": 4.03289354261967e-05, "loss": 0.6877, "step": 14871 }, { "epoch": 1.54, "learning_rate": 4.032662446129062e-05, "loss": 0.6923, "step": 14872 }, { "epoch": 1.54, "learning_rate": 4.0324313426868114e-05, "loss": 0.8904, "step": 14873 }, { "epoch": 1.54, "learning_rate": 4.032200232294475e-05, "loss": 0.2218, "step": 14874 }, { "epoch": 1.54, "learning_rate": 4.031969114953609e-05, "loss": 0.3876, "step": 14875 }, { "epoch": 1.54, "learning_rate": 4.0317379906657674e-05, "loss": 0.3302, "step": 14876 }, { "epoch": 1.54, "learning_rate": 4.0315068594325074e-05, "loss": 0.357, "step": 14877 }, { "epoch": 1.54, "learning_rate": 4.0312757212553845e-05, "loss": 0.4108, "step": 14878 }, { "epoch": 1.54, "learning_rate": 4.031044576135955e-05, "loss": 0.632, "step": 14879 }, { "epoch": 1.54, "learning_rate": 4.030813424075774e-05, "loss": 0.869, "step": 14880 }, { "epoch": 1.54, "learning_rate": 4.030582265076399e-05, "loss": 0.51, "step": 14881 }, { "epoch": 1.54, "learning_rate": 4.030351099139385e-05, "loss": 0.4884, "step": 14882 }, { "epoch": 1.54, "learning_rate": 4.030119926266289e-05, "loss": 0.246, "step": 14883 }, { "epoch": 1.54, "learning_rate": 4.029888746458666e-05, "loss": 1.0815, "step": 14884 }, { "epoch": 1.54, "learning_rate": 4.0296575597180726e-05, "loss": 0.6606, "step": 14885 }, { "epoch": 1.54, "learning_rate": 4.029426366046065e-05, "loss": 1.1613, "step": 14886 }, { "epoch": 1.54, "learning_rate": 4.0291951654442015e-05, "loss": 0.4238, "step": 14887 }, { "epoch": 1.54, "learning_rate": 4.028963957914037e-05, "loss": 0.6507, "step": 14888 }, { "epoch": 1.54, "learning_rate": 4.0287327434571274e-05, "loss": 0.4159, "step": 14889 }, { "epoch": 1.54, "learning_rate": 4.028501522075029e-05, "loss": 0.3662, "step": 14890 }, { "epoch": 1.54, "learning_rate": 4.028270293769299e-05, "loss": 1.2238, "step": 14891 }, { "epoch": 1.54, "learning_rate": 4.0280390585414943e-05, "loss": 0.8066, "step": 14892 }, { "epoch": 1.54, "learning_rate": 4.027807816393172e-05, "loss": 0.5852, "step": 14893 }, { "epoch": 1.54, "learning_rate": 4.0275765673258875e-05, "loss": 0.8334, "step": 14894 }, { "epoch": 1.54, "learning_rate": 4.027345311341197e-05, "loss": 0.7445, "step": 14895 }, { "epoch": 1.54, "learning_rate": 4.02711404844066e-05, "loss": 1.137, "step": 14896 }, { "epoch": 1.54, "learning_rate": 4.02688277862583e-05, "loss": 0.2293, "step": 14897 }, { "epoch": 1.54, "learning_rate": 4.0266515018982665e-05, "loss": 0.2005, "step": 14898 }, { "epoch": 1.54, "learning_rate": 4.0264202182595243e-05, "loss": 0.2767, "step": 14899 }, { "epoch": 1.54, "learning_rate": 4.0261889277111624e-05, "loss": 0.578, "step": 14900 }, { "epoch": 1.54, "learning_rate": 4.025957630254736e-05, "loss": 0.251, "step": 14901 }, { "epoch": 1.54, "learning_rate": 4.025726325891803e-05, "loss": 0.7854, "step": 14902 }, { "epoch": 1.54, "learning_rate": 4.025495014623921e-05, "loss": 0.4578, "step": 14903 }, { "epoch": 1.54, "learning_rate": 4.025263696452645e-05, "loss": 0.5143, "step": 14904 }, { "epoch": 1.54, "learning_rate": 4.025032371379535e-05, "loss": 1.0788, "step": 14905 }, { "epoch": 1.54, "learning_rate": 4.024801039406147e-05, "loss": 0.8419, "step": 14906 }, { "epoch": 1.54, "learning_rate": 4.0245697005340385e-05, "loss": 0.4395, "step": 14907 }, { "epoch": 1.54, "learning_rate": 4.0243383547647656e-05, "loss": 1.1592, "step": 14908 }, { "epoch": 1.54, "learning_rate": 4.024107002099886e-05, "loss": 0.5444, "step": 14909 }, { "epoch": 1.54, "learning_rate": 4.0238756425409583e-05, "loss": 0.6273, "step": 14910 }, { "epoch": 1.54, "learning_rate": 4.023644276089539e-05, "loss": 0.2309, "step": 14911 }, { "epoch": 1.54, "learning_rate": 4.0234129027471875e-05, "loss": 1.0891, "step": 14912 }, { "epoch": 1.54, "learning_rate": 4.023181522515459e-05, "loss": 1.1138, "step": 14913 }, { "epoch": 1.54, "learning_rate": 4.022950135395911e-05, "loss": 0.9147, "step": 14914 }, { "epoch": 1.54, "learning_rate": 4.022718741390103e-05, "loss": 1.1507, "step": 14915 }, { "epoch": 1.54, "learning_rate": 4.022487340499591e-05, "loss": 0.247, "step": 14916 }, { "epoch": 1.54, "learning_rate": 4.022255932725934e-05, "loss": 0.4495, "step": 14917 }, { "epoch": 1.54, "learning_rate": 4.022024518070689e-05, "loss": 0.9335, "step": 14918 }, { "epoch": 1.54, "learning_rate": 4.021793096535415e-05, "loss": 0.4339, "step": 14919 }, { "epoch": 1.54, "learning_rate": 4.021561668121667e-05, "loss": 0.3677, "step": 14920 }, { "epoch": 1.54, "learning_rate": 4.021330232831007e-05, "loss": 0.4818, "step": 14921 }, { "epoch": 1.54, "learning_rate": 4.0210987906649894e-05, "loss": 0.1626, "step": 14922 }, { "epoch": 1.54, "learning_rate": 4.020867341625174e-05, "loss": 0.8718, "step": 14923 }, { "epoch": 1.54, "learning_rate": 4.0206358857131186e-05, "loss": 0.836, "step": 14924 }, { "epoch": 1.54, "learning_rate": 4.0204044229303817e-05, "loss": 1.0813, "step": 14925 }, { "epoch": 1.54, "learning_rate": 4.020172953278521e-05, "loss": 1.0028, "step": 14926 }, { "epoch": 1.54, "learning_rate": 4.0199414767590945e-05, "loss": 0.8639, "step": 14927 }, { "epoch": 1.54, "learning_rate": 4.0197099933736594e-05, "loss": 0.6084, "step": 14928 }, { "epoch": 1.54, "learning_rate": 4.019478503123776e-05, "loss": 0.6903, "step": 14929 }, { "epoch": 1.54, "learning_rate": 4.019247006011002e-05, "loss": 0.2245, "step": 14930 }, { "epoch": 1.54, "learning_rate": 4.0190155020368964e-05, "loss": 0.8774, "step": 14931 }, { "epoch": 1.54, "learning_rate": 4.0187839912030164e-05, "loss": 0.3784, "step": 14932 }, { "epoch": 1.54, "learning_rate": 4.01855247351092e-05, "loss": 1.1233, "step": 14933 }, { "epoch": 1.54, "learning_rate": 4.018320948962167e-05, "loss": 0.2713, "step": 14934 }, { "epoch": 1.54, "learning_rate": 4.0180894175583164e-05, "loss": 0.6519, "step": 14935 }, { "epoch": 1.54, "learning_rate": 4.017857879300926e-05, "loss": 0.2284, "step": 14936 }, { "epoch": 1.54, "learning_rate": 4.017626334191554e-05, "loss": 0.7116, "step": 14937 }, { "epoch": 1.54, "learning_rate": 4.01739478223176e-05, "loss": 0.3365, "step": 14938 }, { "epoch": 1.54, "learning_rate": 4.0171632234231025e-05, "loss": 0.7827, "step": 14939 }, { "epoch": 1.54, "learning_rate": 4.0169316577671393e-05, "loss": 0.4111, "step": 14940 }, { "epoch": 1.54, "learning_rate": 4.016700085265431e-05, "loss": 0.9155, "step": 14941 }, { "epoch": 1.55, "learning_rate": 4.016468505919535e-05, "loss": 1.3645, "step": 14942 }, { "epoch": 1.55, "learning_rate": 4.016236919731011e-05, "loss": 0.367, "step": 14943 }, { "epoch": 1.55, "learning_rate": 4.016005326701419e-05, "loss": 0.7151, "step": 14944 }, { "epoch": 1.55, "learning_rate": 4.0157737268323156e-05, "loss": 0.7482, "step": 14945 }, { "epoch": 1.55, "learning_rate": 4.015542120125261e-05, "loss": 0.9177, "step": 14946 }, { "epoch": 1.55, "learning_rate": 4.015310506581815e-05, "loss": 1.1058, "step": 14947 }, { "epoch": 1.55, "learning_rate": 4.015078886203536e-05, "loss": 0.8865, "step": 14948 }, { "epoch": 1.55, "learning_rate": 4.014847258991983e-05, "loss": 0.6616, "step": 14949 }, { "epoch": 1.55, "learning_rate": 4.014615624948717e-05, "loss": 0.6048, "step": 14950 }, { "epoch": 1.55, "learning_rate": 4.014383984075295e-05, "loss": 1.1096, "step": 14951 }, { "epoch": 1.55, "learning_rate": 4.014152336373278e-05, "loss": 0.247, "step": 14952 }, { "epoch": 1.55, "learning_rate": 4.013920681844224e-05, "loss": 0.3603, "step": 14953 }, { "epoch": 1.55, "learning_rate": 4.0136890204896936e-05, "loss": 1.0564, "step": 14954 }, { "epoch": 1.55, "learning_rate": 4.013457352311246e-05, "loss": 0.206, "step": 14955 }, { "epoch": 1.55, "learning_rate": 4.01322567731044e-05, "loss": 0.8606, "step": 14956 }, { "epoch": 1.55, "learning_rate": 4.0129939954888366e-05, "loss": 0.7379, "step": 14957 }, { "epoch": 1.55, "learning_rate": 4.0127623068479945e-05, "loss": 0.5894, "step": 14958 }, { "epoch": 1.55, "learning_rate": 4.012530611389473e-05, "loss": 0.6742, "step": 14959 }, { "epoch": 1.55, "learning_rate": 4.012298909114834e-05, "loss": 0.2513, "step": 14960 }, { "epoch": 1.55, "learning_rate": 4.012067200025633e-05, "loss": 1.1196, "step": 14961 }, { "epoch": 1.55, "learning_rate": 4.011835484123434e-05, "loss": 0.5682, "step": 14962 }, { "epoch": 1.55, "learning_rate": 4.011603761409795e-05, "loss": 0.8828, "step": 14963 }, { "epoch": 1.55, "learning_rate": 4.011372031886277e-05, "loss": 0.5749, "step": 14964 }, { "epoch": 1.55, "learning_rate": 4.011140295554438e-05, "loss": 0.4, "step": 14965 }, { "epoch": 1.55, "learning_rate": 4.01090855241584e-05, "loss": 0.4895, "step": 14966 }, { "epoch": 1.55, "learning_rate": 4.0106768024720405e-05, "loss": 1.0455, "step": 14967 }, { "epoch": 1.55, "learning_rate": 4.010445045724602e-05, "loss": 0.3283, "step": 14968 }, { "epoch": 1.55, "learning_rate": 4.010213282175086e-05, "loss": 0.3036, "step": 14969 }, { "epoch": 1.55, "learning_rate": 4.009981511825048e-05, "loss": 0.197, "step": 14970 }, { "epoch": 1.55, "learning_rate": 4.0097497346760515e-05, "loss": 0.9963, "step": 14971 }, { "epoch": 1.55, "learning_rate": 4.009517950729656e-05, "loss": 1.2128, "step": 14972 }, { "epoch": 1.55, "learning_rate": 4.009286159987421e-05, "loss": 0.8043, "step": 14973 }, { "epoch": 1.55, "learning_rate": 4.00905436245091e-05, "loss": 0.6899, "step": 14974 }, { "epoch": 1.55, "learning_rate": 4.00882255812168e-05, "loss": 0.9288, "step": 14975 }, { "epoch": 1.55, "learning_rate": 4.0085907470012925e-05, "loss": 0.5889, "step": 14976 }, { "epoch": 1.55, "learning_rate": 4.008358929091308e-05, "loss": 0.4568, "step": 14977 }, { "epoch": 1.55, "learning_rate": 4.008127104393287e-05, "loss": 0.8128, "step": 14978 }, { "epoch": 1.55, "learning_rate": 4.007895272908791e-05, "loss": 0.265, "step": 14979 }, { "epoch": 1.55, "learning_rate": 4.0076634346393786e-05, "loss": 0.2249, "step": 14980 }, { "epoch": 1.55, "learning_rate": 4.0074315895866116e-05, "loss": 1.0186, "step": 14981 }, { "epoch": 1.55, "learning_rate": 4.007199737752052e-05, "loss": 0.9327, "step": 14982 }, { "epoch": 1.55, "learning_rate": 4.006967879137259e-05, "loss": 1.5946, "step": 14983 }, { "epoch": 1.55, "learning_rate": 4.0067360137437946e-05, "loss": 0.3398, "step": 14984 }, { "epoch": 1.55, "learning_rate": 4.0065041415732175e-05, "loss": 0.3533, "step": 14985 }, { "epoch": 1.55, "learning_rate": 4.0062722626270905e-05, "loss": 0.4952, "step": 14986 }, { "epoch": 1.55, "learning_rate": 4.0060403769069746e-05, "loss": 0.5528, "step": 14987 }, { "epoch": 1.55, "learning_rate": 4.005808484414431e-05, "loss": 0.6758, "step": 14988 }, { "epoch": 1.55, "learning_rate": 4.005576585151018e-05, "loss": 0.7626, "step": 14989 }, { "epoch": 1.55, "learning_rate": 4.0053446791182994e-05, "loss": 0.7046, "step": 14990 }, { "epoch": 1.55, "learning_rate": 4.005112766317836e-05, "loss": 1.1635, "step": 14991 }, { "epoch": 1.55, "learning_rate": 4.0048808467511884e-05, "loss": 1.0911, "step": 14992 }, { "epoch": 1.55, "learning_rate": 4.004648920419919e-05, "loss": 0.3268, "step": 14993 }, { "epoch": 1.55, "learning_rate": 4.0044169873255875e-05, "loss": 0.4361, "step": 14994 }, { "epoch": 1.55, "learning_rate": 4.0041850474697555e-05, "loss": 0.3813, "step": 14995 }, { "epoch": 1.55, "learning_rate": 4.003953100853985e-05, "loss": 0.0697, "step": 14996 }, { "epoch": 1.55, "learning_rate": 4.003721147479837e-05, "loss": 0.1072, "step": 14997 }, { "epoch": 1.55, "learning_rate": 4.003489187348873e-05, "loss": 1.4041, "step": 14998 }, { "epoch": 1.55, "learning_rate": 4.003257220462655e-05, "loss": 0.3971, "step": 14999 }, { "epoch": 1.55, "learning_rate": 4.003025246822743e-05, "loss": 0.7811, "step": 15000 }, { "epoch": 1.55, "learning_rate": 4.002793266430701e-05, "loss": 0.5691, "step": 15001 }, { "epoch": 1.55, "learning_rate": 4.0025612792880896e-05, "loss": 0.3682, "step": 15002 }, { "epoch": 1.55, "learning_rate": 4.0023292853964704e-05, "loss": 0.487, "step": 15003 }, { "epoch": 1.55, "learning_rate": 4.002097284757404e-05, "loss": 0.5372, "step": 15004 }, { "epoch": 1.55, "learning_rate": 4.001865277372452e-05, "loss": 0.6065, "step": 15005 }, { "epoch": 1.55, "learning_rate": 4.001633263243179e-05, "loss": 0.2965, "step": 15006 }, { "epoch": 1.55, "learning_rate": 4.001401242371146e-05, "loss": 0.2162, "step": 15007 }, { "epoch": 1.55, "learning_rate": 4.0011692147579126e-05, "loss": 1.4541, "step": 15008 }, { "epoch": 1.55, "learning_rate": 4.0009371804050434e-05, "loss": 1.2478, "step": 15009 }, { "epoch": 1.55, "learning_rate": 4.000705139314098e-05, "loss": 1.0487, "step": 15010 }, { "epoch": 1.55, "learning_rate": 4.0004730914866404e-05, "loss": 0.419, "step": 15011 }, { "epoch": 1.55, "learning_rate": 4.0002410369242325e-05, "loss": 0.1212, "step": 15012 }, { "epoch": 1.55, "learning_rate": 4.000008975628436e-05, "loss": 0.294, "step": 15013 }, { "epoch": 1.55, "learning_rate": 3.999776907600812e-05, "loss": 0.1018, "step": 15014 }, { "epoch": 1.55, "learning_rate": 3.9995448328429244e-05, "loss": 0.9118, "step": 15015 }, { "epoch": 1.55, "learning_rate": 3.999312751356335e-05, "loss": 0.2621, "step": 15016 }, { "epoch": 1.55, "learning_rate": 3.999080663142606e-05, "loss": 0.0838, "step": 15017 }, { "epoch": 1.55, "learning_rate": 3.9988485682033e-05, "loss": 0.4894, "step": 15018 }, { "epoch": 1.55, "learning_rate": 3.998616466539979e-05, "loss": 0.8215, "step": 15019 }, { "epoch": 1.55, "learning_rate": 3.998384358154205e-05, "loss": 0.3758, "step": 15020 }, { "epoch": 1.55, "learning_rate": 3.9981522430475415e-05, "loss": 0.4657, "step": 15021 }, { "epoch": 1.55, "learning_rate": 3.997920121221551e-05, "loss": 0.5099, "step": 15022 }, { "epoch": 1.55, "learning_rate": 3.997687992677795e-05, "loss": 0.5807, "step": 15023 }, { "epoch": 1.55, "learning_rate": 3.997455857417837e-05, "loss": 0.5464, "step": 15024 }, { "epoch": 1.55, "learning_rate": 3.997223715443239e-05, "loss": 0.256, "step": 15025 }, { "epoch": 1.55, "learning_rate": 3.9969915667555664e-05, "loss": 0.9621, "step": 15026 }, { "epoch": 1.55, "learning_rate": 3.996759411356379e-05, "loss": 0.4749, "step": 15027 }, { "epoch": 1.55, "learning_rate": 3.9965272492472394e-05, "loss": 0.3061, "step": 15028 }, { "epoch": 1.55, "learning_rate": 3.996295080429712e-05, "loss": 1.0195, "step": 15029 }, { "epoch": 1.55, "learning_rate": 3.9960629049053586e-05, "loss": 0.6609, "step": 15030 }, { "epoch": 1.55, "learning_rate": 3.995830722675745e-05, "loss": 0.412, "step": 15031 }, { "epoch": 1.55, "learning_rate": 3.99559853374243e-05, "loss": 0.437, "step": 15032 }, { "epoch": 1.55, "learning_rate": 3.9953663381069785e-05, "loss": 0.1868, "step": 15033 }, { "epoch": 1.55, "learning_rate": 3.995134135770955e-05, "loss": 1.0555, "step": 15034 }, { "epoch": 1.55, "learning_rate": 3.9949019267359206e-05, "loss": 0.5181, "step": 15035 }, { "epoch": 1.55, "learning_rate": 3.994669711003439e-05, "loss": 0.3281, "step": 15036 }, { "epoch": 1.55, "learning_rate": 3.994437488575074e-05, "loss": 0.54, "step": 15037 }, { "epoch": 1.56, "learning_rate": 3.994205259452389e-05, "loss": 0.1497, "step": 15038 }, { "epoch": 1.56, "learning_rate": 3.993973023636946e-05, "loss": 0.5672, "step": 15039 }, { "epoch": 1.56, "learning_rate": 3.993740781130309e-05, "loss": 0.4037, "step": 15040 }, { "epoch": 1.56, "learning_rate": 3.993508531934042e-05, "loss": 0.8425, "step": 15041 }, { "epoch": 1.56, "learning_rate": 3.993276276049708e-05, "loss": 1.1356, "step": 15042 }, { "epoch": 1.56, "learning_rate": 3.993044013478871e-05, "loss": 0.3115, "step": 15043 }, { "epoch": 1.56, "learning_rate": 3.992811744223094e-05, "loss": 0.3539, "step": 15044 }, { "epoch": 1.56, "learning_rate": 3.99257946828394e-05, "loss": 0.6505, "step": 15045 }, { "epoch": 1.56, "learning_rate": 3.9923471856629735e-05, "loss": 0.5468, "step": 15046 }, { "epoch": 1.56, "learning_rate": 3.9921148963617584e-05, "loss": 0.8204, "step": 15047 }, { "epoch": 1.56, "learning_rate": 3.991882600381857e-05, "loss": 0.5568, "step": 15048 }, { "epoch": 1.56, "learning_rate": 3.991650297724835e-05, "loss": 0.3215, "step": 15049 }, { "epoch": 1.56, "learning_rate": 3.991417988392255e-05, "loss": 1.1326, "step": 15050 }, { "epoch": 1.56, "learning_rate": 3.9911856723856806e-05, "loss": 1.3786, "step": 15051 }, { "epoch": 1.56, "learning_rate": 3.990953349706676e-05, "loss": 0.2092, "step": 15052 }, { "epoch": 1.56, "learning_rate": 3.9907210203568055e-05, "loss": 0.3727, "step": 15053 }, { "epoch": 1.56, "learning_rate": 3.990488684337633e-05, "loss": 1.0477, "step": 15054 }, { "epoch": 1.56, "learning_rate": 3.9902563416507234e-05, "loss": 0.2754, "step": 15055 }, { "epoch": 1.56, "learning_rate": 3.9900239922976384e-05, "loss": 0.6411, "step": 15056 }, { "epoch": 1.56, "learning_rate": 3.9897916362799444e-05, "loss": 0.2908, "step": 15057 }, { "epoch": 1.56, "learning_rate": 3.989559273599205e-05, "loss": 0.5569, "step": 15058 }, { "epoch": 1.56, "learning_rate": 3.9893269042569824e-05, "loss": 0.5753, "step": 15059 }, { "epoch": 1.56, "learning_rate": 3.9890945282548444e-05, "loss": 0.332, "step": 15060 }, { "epoch": 1.56, "learning_rate": 3.9888621455943524e-05, "loss": 0.3381, "step": 15061 }, { "epoch": 1.56, "learning_rate": 3.988629756277072e-05, "loss": 0.4137, "step": 15062 }, { "epoch": 1.56, "learning_rate": 3.9883973603045675e-05, "loss": 0.5239, "step": 15063 }, { "epoch": 1.56, "learning_rate": 3.988164957678403e-05, "loss": 0.4363, "step": 15064 }, { "epoch": 1.56, "learning_rate": 3.987932548400144e-05, "loss": 0.558, "step": 15065 }, { "epoch": 1.56, "learning_rate": 3.987700132471354e-05, "loss": 0.2903, "step": 15066 }, { "epoch": 1.56, "learning_rate": 3.987467709893597e-05, "loss": 1.5276, "step": 15067 }, { "epoch": 1.56, "learning_rate": 3.987235280668439e-05, "loss": 0.8701, "step": 15068 }, { "epoch": 1.56, "learning_rate": 3.987002844797444e-05, "loss": 0.6416, "step": 15069 }, { "epoch": 1.56, "learning_rate": 3.9867704022821773e-05, "loss": 0.5614, "step": 15070 }, { "epoch": 1.56, "learning_rate": 3.986537953124203e-05, "loss": 1.0922, "step": 15071 }, { "epoch": 1.56, "learning_rate": 3.9863054973250853e-05, "loss": 0.3425, "step": 15072 }, { "epoch": 1.56, "learning_rate": 3.98607303488639e-05, "loss": 0.4803, "step": 15073 }, { "epoch": 1.56, "learning_rate": 3.985840565809682e-05, "loss": 0.6891, "step": 15074 }, { "epoch": 1.56, "learning_rate": 3.985608090096526e-05, "loss": 0.9728, "step": 15075 }, { "epoch": 1.56, "learning_rate": 3.9853756077484874e-05, "loss": 0.7428, "step": 15076 }, { "epoch": 1.56, "learning_rate": 3.98514311876713e-05, "loss": 0.5459, "step": 15077 }, { "epoch": 1.56, "learning_rate": 3.984910623154019e-05, "loss": 0.3859, "step": 15078 }, { "epoch": 1.56, "learning_rate": 3.984678120910722e-05, "loss": 1.3196, "step": 15079 }, { "epoch": 1.56, "learning_rate": 3.984445612038801e-05, "loss": 0.9347, "step": 15080 }, { "epoch": 1.56, "learning_rate": 3.984213096539823e-05, "loss": 0.8848, "step": 15081 }, { "epoch": 1.56, "learning_rate": 3.983980574415352e-05, "loss": 0.1152, "step": 15082 }, { "epoch": 1.56, "learning_rate": 3.9837480456669544e-05, "loss": 0.9204, "step": 15083 }, { "epoch": 1.56, "learning_rate": 3.9835155102961955e-05, "loss": 0.7625, "step": 15084 }, { "epoch": 1.56, "learning_rate": 3.98328296830464e-05, "loss": 0.1867, "step": 15085 }, { "epoch": 1.56, "learning_rate": 3.983050419693854e-05, "loss": 0.2162, "step": 15086 }, { "epoch": 1.56, "learning_rate": 3.9828178644654014e-05, "loss": 0.2019, "step": 15087 }, { "epoch": 1.56, "learning_rate": 3.98258530262085e-05, "loss": 0.4773, "step": 15088 }, { "epoch": 1.56, "learning_rate": 3.982352734161765e-05, "loss": 1.0087, "step": 15089 }, { "epoch": 1.56, "learning_rate": 3.98212015908971e-05, "loss": 0.9114, "step": 15090 }, { "epoch": 1.56, "learning_rate": 3.981887577406252e-05, "loss": 1.1262, "step": 15091 }, { "epoch": 1.56, "learning_rate": 3.9816549891129565e-05, "loss": 0.2876, "step": 15092 }, { "epoch": 1.56, "learning_rate": 3.98142239421139e-05, "loss": 0.9587, "step": 15093 }, { "epoch": 1.56, "learning_rate": 3.981189792703116e-05, "loss": 1.4887, "step": 15094 }, { "epoch": 1.56, "learning_rate": 3.980957184589704e-05, "loss": 0.5245, "step": 15095 }, { "epoch": 1.56, "learning_rate": 3.9807245698727165e-05, "loss": 0.8524, "step": 15096 }, { "epoch": 1.56, "learning_rate": 3.98049194855372e-05, "loss": 0.7544, "step": 15097 }, { "epoch": 1.56, "learning_rate": 3.9802593206342835e-05, "loss": 1.434, "step": 15098 }, { "epoch": 1.56, "learning_rate": 3.980026686115969e-05, "loss": 0.5302, "step": 15099 }, { "epoch": 1.56, "learning_rate": 3.979794045000344e-05, "loss": 0.4657, "step": 15100 }, { "epoch": 1.56, "learning_rate": 3.979561397288975e-05, "loss": 0.7425, "step": 15101 }, { "epoch": 1.56, "learning_rate": 3.979328742983428e-05, "loss": 0.3693, "step": 15102 }, { "epoch": 1.56, "learning_rate": 3.979096082085269e-05, "loss": 0.7529, "step": 15103 }, { "epoch": 1.56, "learning_rate": 3.9788634145960636e-05, "loss": 0.829, "step": 15104 }, { "epoch": 1.56, "learning_rate": 3.97863074051738e-05, "loss": 0.9712, "step": 15105 }, { "epoch": 1.56, "learning_rate": 3.9783980598507825e-05, "loss": 0.7949, "step": 15106 }, { "epoch": 1.56, "learning_rate": 3.9781653725978384e-05, "loss": 0.6573, "step": 15107 }, { "epoch": 1.56, "learning_rate": 3.977932678760114e-05, "loss": 1.0864, "step": 15108 }, { "epoch": 1.56, "learning_rate": 3.977699978339175e-05, "loss": 0.2385, "step": 15109 }, { "epoch": 1.56, "learning_rate": 3.977467271336589e-05, "loss": 0.7576, "step": 15110 }, { "epoch": 1.56, "learning_rate": 3.977234557753922e-05, "loss": 0.519, "step": 15111 }, { "epoch": 1.56, "learning_rate": 3.97700183759274e-05, "loss": 0.8898, "step": 15112 }, { "epoch": 1.56, "learning_rate": 3.976769110854611e-05, "loss": 0.9406, "step": 15113 }, { "epoch": 1.56, "learning_rate": 3.9765363775411015e-05, "loss": 0.5876, "step": 15114 }, { "epoch": 1.56, "learning_rate": 3.976303637653777e-05, "loss": 1.0566, "step": 15115 }, { "epoch": 1.56, "learning_rate": 3.976070891194203e-05, "loss": 0.9203, "step": 15116 }, { "epoch": 1.56, "learning_rate": 3.975838138163951e-05, "loss": 1.0715, "step": 15117 }, { "epoch": 1.56, "learning_rate": 3.9756053785645835e-05, "loss": 1.0129, "step": 15118 }, { "epoch": 1.56, "learning_rate": 3.9753726123976685e-05, "loss": 1.0407, "step": 15119 }, { "epoch": 1.56, "learning_rate": 3.975139839664774e-05, "loss": 0.9021, "step": 15120 }, { "epoch": 1.56, "learning_rate": 3.974907060367467e-05, "loss": 1.0322, "step": 15121 }, { "epoch": 1.56, "learning_rate": 3.974674274507313e-05, "loss": 1.049, "step": 15122 }, { "epoch": 1.56, "learning_rate": 3.97444148208588e-05, "loss": 0.9704, "step": 15123 }, { "epoch": 1.56, "learning_rate": 3.974208683104734e-05, "loss": 1.3027, "step": 15124 }, { "epoch": 1.56, "learning_rate": 3.9739758775654437e-05, "loss": 0.4759, "step": 15125 }, { "epoch": 1.56, "learning_rate": 3.973743065469576e-05, "loss": 0.577, "step": 15126 }, { "epoch": 1.56, "learning_rate": 3.9735102468186976e-05, "loss": 0.581, "step": 15127 }, { "epoch": 1.56, "learning_rate": 3.973277421614376e-05, "loss": 0.1714, "step": 15128 }, { "epoch": 1.56, "learning_rate": 3.9730445898581785e-05, "loss": 0.4666, "step": 15129 }, { "epoch": 1.56, "learning_rate": 3.972811751551672e-05, "loss": 0.8038, "step": 15130 }, { "epoch": 1.56, "learning_rate": 3.972578906696426e-05, "loss": 0.9589, "step": 15131 }, { "epoch": 1.56, "learning_rate": 3.9723460552940056e-05, "loss": 0.4025, "step": 15132 }, { "epoch": 1.57, "learning_rate": 3.9721131973459784e-05, "loss": 0.9609, "step": 15133 }, { "epoch": 1.57, "learning_rate": 3.971880332853913e-05, "loss": 0.6359, "step": 15134 }, { "epoch": 1.57, "learning_rate": 3.971647461819376e-05, "loss": 0.7432, "step": 15135 }, { "epoch": 1.57, "learning_rate": 3.971414584243937e-05, "loss": 0.4289, "step": 15136 }, { "epoch": 1.57, "learning_rate": 3.971181700129161e-05, "loss": 0.7785, "step": 15137 }, { "epoch": 1.57, "learning_rate": 3.970948809476618e-05, "loss": 0.2777, "step": 15138 }, { "epoch": 1.57, "learning_rate": 3.9707159122878744e-05, "loss": 0.5922, "step": 15139 }, { "epoch": 1.57, "learning_rate": 3.970483008564498e-05, "loss": 0.238, "step": 15140 }, { "epoch": 1.57, "learning_rate": 3.9702500983080575e-05, "loss": 0.4721, "step": 15141 }, { "epoch": 1.57, "learning_rate": 3.97001718152012e-05, "loss": 0.2773, "step": 15142 }, { "epoch": 1.57, "learning_rate": 3.969784258202255e-05, "loss": 0.6107, "step": 15143 }, { "epoch": 1.57, "learning_rate": 3.969551328356027e-05, "loss": 0.4294, "step": 15144 }, { "epoch": 1.57, "learning_rate": 3.9693183919830084e-05, "loss": 0.3816, "step": 15145 }, { "epoch": 1.57, "learning_rate": 3.9690854490847655e-05, "loss": 0.541, "step": 15146 }, { "epoch": 1.57, "learning_rate": 3.968852499662865e-05, "loss": 0.4135, "step": 15147 }, { "epoch": 1.57, "learning_rate": 3.968619543718875e-05, "loss": 0.7227, "step": 15148 }, { "epoch": 1.57, "learning_rate": 3.968386581254367e-05, "loss": 0.7773, "step": 15149 }, { "epoch": 1.57, "learning_rate": 3.968153612270906e-05, "loss": 0.2459, "step": 15150 }, { "epoch": 1.57, "learning_rate": 3.967920636770062e-05, "loss": 0.7904, "step": 15151 }, { "epoch": 1.57, "learning_rate": 3.9676876547534024e-05, "loss": 0.9339, "step": 15152 }, { "epoch": 1.57, "learning_rate": 3.9674546662224954e-05, "loss": 0.325, "step": 15153 }, { "epoch": 1.57, "learning_rate": 3.967221671178911e-05, "loss": 1.1254, "step": 15154 }, { "epoch": 1.57, "learning_rate": 3.9669886696242156e-05, "loss": 0.686, "step": 15155 }, { "epoch": 1.57, "learning_rate": 3.9667556615599794e-05, "loss": 0.2359, "step": 15156 }, { "epoch": 1.57, "learning_rate": 3.96652264698777e-05, "loss": 0.2552, "step": 15157 }, { "epoch": 1.57, "learning_rate": 3.9662896259091575e-05, "loss": 0.5542, "step": 15158 }, { "epoch": 1.57, "learning_rate": 3.966056598325708e-05, "loss": 0.4528, "step": 15159 }, { "epoch": 1.57, "learning_rate": 3.965823564238992e-05, "loss": 0.411, "step": 15160 }, { "epoch": 1.57, "learning_rate": 3.965590523650577e-05, "loss": 0.803, "step": 15161 }, { "epoch": 1.57, "learning_rate": 3.9653574765620326e-05, "loss": 0.7836, "step": 15162 }, { "epoch": 1.57, "learning_rate": 3.965124422974928e-05, "loss": 0.9516, "step": 15163 }, { "epoch": 1.57, "learning_rate": 3.964891362890832e-05, "loss": 0.0821, "step": 15164 }, { "epoch": 1.57, "learning_rate": 3.964658296311313e-05, "loss": 0.34, "step": 15165 }, { "epoch": 1.57, "learning_rate": 3.96442522323794e-05, "loss": 0.5596, "step": 15166 }, { "epoch": 1.57, "learning_rate": 3.9641921436722817e-05, "loss": 0.1359, "step": 15167 }, { "epoch": 1.57, "learning_rate": 3.9639590576159066e-05, "loss": 0.4155, "step": 15168 }, { "epoch": 1.57, "learning_rate": 3.9637259650703865e-05, "loss": 0.8125, "step": 15169 }, { "epoch": 1.57, "learning_rate": 3.9634928660372885e-05, "loss": 0.6173, "step": 15170 }, { "epoch": 1.57, "learning_rate": 3.9632597605181814e-05, "loss": 1.0218, "step": 15171 }, { "epoch": 1.57, "learning_rate": 3.963026648514634e-05, "loss": 0.9626, "step": 15172 }, { "epoch": 1.57, "learning_rate": 3.9627935300282175e-05, "loss": 0.6807, "step": 15173 }, { "epoch": 1.57, "learning_rate": 3.962560405060501e-05, "loss": 0.5356, "step": 15174 }, { "epoch": 1.57, "learning_rate": 3.962327273613052e-05, "loss": 0.5527, "step": 15175 }, { "epoch": 1.57, "learning_rate": 3.962094135687441e-05, "loss": 0.4904, "step": 15176 }, { "epoch": 1.57, "learning_rate": 3.961860991285238e-05, "loss": 1.3076, "step": 15177 }, { "epoch": 1.57, "learning_rate": 3.961627840408011e-05, "loss": 0.257, "step": 15178 }, { "epoch": 1.57, "learning_rate": 3.961394683057331e-05, "loss": 1.0124, "step": 15179 }, { "epoch": 1.57, "learning_rate": 3.961161519234766e-05, "loss": 0.8868, "step": 15180 }, { "epoch": 1.57, "learning_rate": 3.960928348941888e-05, "loss": 0.3243, "step": 15181 }, { "epoch": 1.57, "learning_rate": 3.9606951721802644e-05, "loss": 0.4199, "step": 15182 }, { "epoch": 1.57, "learning_rate": 3.960461988951466e-05, "loss": 0.6647, "step": 15183 }, { "epoch": 1.57, "learning_rate": 3.9602287992570626e-05, "loss": 1.0571, "step": 15184 }, { "epoch": 1.57, "learning_rate": 3.9599956030986234e-05, "loss": 1.3657, "step": 15185 }, { "epoch": 1.57, "learning_rate": 3.959762400477717e-05, "loss": 0.6283, "step": 15186 }, { "epoch": 1.57, "learning_rate": 3.9595291913959154e-05, "loss": 0.4426, "step": 15187 }, { "epoch": 1.57, "learning_rate": 3.9592959758547894e-05, "loss": 0.5379, "step": 15188 }, { "epoch": 1.57, "learning_rate": 3.959062753855906e-05, "loss": 0.421, "step": 15189 }, { "epoch": 1.57, "learning_rate": 3.9588295254008364e-05, "loss": 0.3066, "step": 15190 }, { "epoch": 1.57, "learning_rate": 3.958596290491151e-05, "loss": 1.3476, "step": 15191 }, { "epoch": 1.57, "learning_rate": 3.958363049128419e-05, "loss": 0.5803, "step": 15192 }, { "epoch": 1.57, "learning_rate": 3.958129801314213e-05, "loss": 1.0622, "step": 15193 }, { "epoch": 1.57, "learning_rate": 3.9578965470501e-05, "loss": 0.2458, "step": 15194 }, { "epoch": 1.57, "learning_rate": 3.957663286337652e-05, "loss": 0.4328, "step": 15195 }, { "epoch": 1.57, "learning_rate": 3.957430019178439e-05, "loss": 0.8857, "step": 15196 }, { "epoch": 1.57, "learning_rate": 3.95719674557403e-05, "loss": 0.7553, "step": 15197 }, { "epoch": 1.57, "learning_rate": 3.956963465525998e-05, "loss": 0.3536, "step": 15198 }, { "epoch": 1.57, "learning_rate": 3.956730179035911e-05, "loss": 1.0401, "step": 15199 }, { "epoch": 1.57, "learning_rate": 3.95649688610534e-05, "loss": 0.5886, "step": 15200 }, { "epoch": 1.57, "learning_rate": 3.956263586735856e-05, "loss": 0.7434, "step": 15201 }, { "epoch": 1.57, "learning_rate": 3.95603028092903e-05, "loss": 0.1954, "step": 15202 }, { "epoch": 1.57, "learning_rate": 3.9557969686864314e-05, "loss": 0.7594, "step": 15203 }, { "epoch": 1.57, "learning_rate": 3.955563650009631e-05, "loss": 1.2354, "step": 15204 }, { "epoch": 1.57, "learning_rate": 3.9553303249002e-05, "loss": 0.7826, "step": 15205 }, { "epoch": 1.57, "learning_rate": 3.955096993359708e-05, "loss": 1.0011, "step": 15206 }, { "epoch": 1.57, "learning_rate": 3.9548636553897275e-05, "loss": 0.8823, "step": 15207 }, { "epoch": 1.57, "learning_rate": 3.9546303109918286e-05, "loss": 0.3029, "step": 15208 }, { "epoch": 1.57, "learning_rate": 3.954396960167581e-05, "loss": 0.2296, "step": 15209 }, { "epoch": 1.57, "learning_rate": 3.954163602918556e-05, "loss": 0.6906, "step": 15210 }, { "epoch": 1.57, "learning_rate": 3.953930239246325e-05, "loss": 0.6246, "step": 15211 }, { "epoch": 1.57, "learning_rate": 3.95369686915246e-05, "loss": 1.4936, "step": 15212 }, { "epoch": 1.57, "learning_rate": 3.9534634926385297e-05, "loss": 1.6412, "step": 15213 }, { "epoch": 1.57, "learning_rate": 3.9532301097061075e-05, "loss": 0.276, "step": 15214 }, { "epoch": 1.57, "learning_rate": 3.9529967203567624e-05, "loss": 0.185, "step": 15215 }, { "epoch": 1.57, "learning_rate": 3.9527633245920656e-05, "loss": 0.2628, "step": 15216 }, { "epoch": 1.57, "learning_rate": 3.95252992241359e-05, "loss": 0.7669, "step": 15217 }, { "epoch": 1.57, "learning_rate": 3.952296513822905e-05, "loss": 0.9304, "step": 15218 }, { "epoch": 1.57, "learning_rate": 3.9520630988215834e-05, "loss": 0.9416, "step": 15219 }, { "epoch": 1.57, "learning_rate": 3.9518296774111955e-05, "loss": 0.9557, "step": 15220 }, { "epoch": 1.57, "learning_rate": 3.951596249593313e-05, "loss": 0.5053, "step": 15221 }, { "epoch": 1.57, "learning_rate": 3.951362815369509e-05, "loss": 1.4855, "step": 15222 }, { "epoch": 1.57, "learning_rate": 3.95112937474135e-05, "loss": 0.5851, "step": 15223 }, { "epoch": 1.57, "learning_rate": 3.950895927710412e-05, "loss": 0.243, "step": 15224 }, { "epoch": 1.57, "learning_rate": 3.9506624742782644e-05, "loss": 0.6603, "step": 15225 }, { "epoch": 1.57, "learning_rate": 3.9504290144464816e-05, "loss": 0.3713, "step": 15226 }, { "epoch": 1.57, "learning_rate": 3.9501955482166315e-05, "loss": 0.7966, "step": 15227 }, { "epoch": 1.57, "learning_rate": 3.949962075590287e-05, "loss": 1.2605, "step": 15228 }, { "epoch": 1.58, "learning_rate": 3.9497285965690206e-05, "loss": 0.2884, "step": 15229 }, { "epoch": 1.58, "learning_rate": 3.9494951111544034e-05, "loss": 0.6601, "step": 15230 }, { "epoch": 1.58, "learning_rate": 3.949261619348008e-05, "loss": 0.6414, "step": 15231 }, { "epoch": 1.58, "learning_rate": 3.9490281211514046e-05, "loss": 0.6345, "step": 15232 }, { "epoch": 1.58, "learning_rate": 3.9487946165661665e-05, "loss": 1.2608, "step": 15233 }, { "epoch": 1.58, "learning_rate": 3.9485611055938636e-05, "loss": 1.3375, "step": 15234 }, { "epoch": 1.58, "learning_rate": 3.9483275882360706e-05, "loss": 0.6261, "step": 15235 }, { "epoch": 1.58, "learning_rate": 3.9480940644943584e-05, "loss": 0.8896, "step": 15236 }, { "epoch": 1.58, "learning_rate": 3.9478605343702984e-05, "loss": 0.7357, "step": 15237 }, { "epoch": 1.58, "learning_rate": 3.9476269978654625e-05, "loss": 1.0057, "step": 15238 }, { "epoch": 1.58, "learning_rate": 3.947393454981424e-05, "loss": 0.9051, "step": 15239 }, { "epoch": 1.58, "learning_rate": 3.947159905719755e-05, "loss": 0.2968, "step": 15240 }, { "epoch": 1.58, "learning_rate": 3.946926350082026e-05, "loss": 0.4197, "step": 15241 }, { "epoch": 1.58, "learning_rate": 3.946692788069811e-05, "loss": 0.3806, "step": 15242 }, { "epoch": 1.58, "learning_rate": 3.946459219684681e-05, "loss": 0.8618, "step": 15243 }, { "epoch": 1.58, "learning_rate": 3.946225644928209e-05, "loss": 0.2509, "step": 15244 }, { "epoch": 1.58, "learning_rate": 3.9459920638019686e-05, "loss": 0.1671, "step": 15245 }, { "epoch": 1.58, "learning_rate": 3.94575847630753e-05, "loss": 0.8811, "step": 15246 }, { "epoch": 1.58, "learning_rate": 3.945524882446466e-05, "loss": 0.204, "step": 15247 }, { "epoch": 1.58, "learning_rate": 3.9452912822203514e-05, "loss": 0.619, "step": 15248 }, { "epoch": 1.58, "learning_rate": 3.945057675630755e-05, "loss": 0.3711, "step": 15249 }, { "epoch": 1.58, "learning_rate": 3.944824062679254e-05, "loss": 0.6496, "step": 15250 }, { "epoch": 1.58, "learning_rate": 3.9445904433674166e-05, "loss": 0.3011, "step": 15251 }, { "epoch": 1.58, "learning_rate": 3.9443568176968183e-05, "loss": 0.213, "step": 15252 }, { "epoch": 1.58, "learning_rate": 3.94412318566903e-05, "loss": 1.2835, "step": 15253 }, { "epoch": 1.58, "learning_rate": 3.943889547285625e-05, "loss": 0.4689, "step": 15254 }, { "epoch": 1.58, "learning_rate": 3.943655902548179e-05, "loss": 0.2193, "step": 15255 }, { "epoch": 1.58, "learning_rate": 3.9434222514582595e-05, "loss": 0.7792, "step": 15256 }, { "epoch": 1.58, "learning_rate": 3.943188594017443e-05, "loss": 0.6741, "step": 15257 }, { "epoch": 1.58, "learning_rate": 3.9429549302273024e-05, "loss": 0.3094, "step": 15258 }, { "epoch": 1.58, "learning_rate": 3.9427212600894084e-05, "loss": 0.3704, "step": 15259 }, { "epoch": 1.58, "learning_rate": 3.942487583605336e-05, "loss": 0.2301, "step": 15260 }, { "epoch": 1.58, "learning_rate": 3.9422539007766585e-05, "loss": 0.7429, "step": 15261 }, { "epoch": 1.58, "learning_rate": 3.9420202116049475e-05, "loss": 0.3015, "step": 15262 }, { "epoch": 1.58, "learning_rate": 3.9417865160917766e-05, "loss": 1.0826, "step": 15263 }, { "epoch": 1.58, "learning_rate": 3.941552814238719e-05, "loss": 0.4213, "step": 15264 }, { "epoch": 1.58, "learning_rate": 3.9413191060473497e-05, "loss": 0.1873, "step": 15265 }, { "epoch": 1.58, "learning_rate": 3.941085391519239e-05, "loss": 0.5775, "step": 15266 }, { "epoch": 1.58, "learning_rate": 3.9408516706559625e-05, "loss": 0.8378, "step": 15267 }, { "epoch": 1.58, "learning_rate": 3.9406179434590926e-05, "loss": 0.3203, "step": 15268 }, { "epoch": 1.58, "learning_rate": 3.940384209930202e-05, "loss": 0.1284, "step": 15269 }, { "epoch": 1.58, "learning_rate": 3.940150470070865e-05, "loss": 0.534, "step": 15270 }, { "epoch": 1.58, "learning_rate": 3.939916723882656e-05, "loss": 0.9811, "step": 15271 }, { "epoch": 1.58, "learning_rate": 3.939682971367147e-05, "loss": 0.4035, "step": 15272 }, { "epoch": 1.58, "learning_rate": 3.939449212525912e-05, "loss": 0.9398, "step": 15273 }, { "epoch": 1.58, "learning_rate": 3.939215447360525e-05, "loss": 0.3533, "step": 15274 }, { "epoch": 1.58, "learning_rate": 3.938981675872559e-05, "loss": 0.5692, "step": 15275 }, { "epoch": 1.58, "learning_rate": 3.938747898063588e-05, "loss": 0.2959, "step": 15276 }, { "epoch": 1.58, "learning_rate": 3.9385141139351864e-05, "loss": 0.7463, "step": 15277 }, { "epoch": 1.58, "learning_rate": 3.938280323488927e-05, "loss": 0.1463, "step": 15278 }, { "epoch": 1.58, "learning_rate": 3.938046526726384e-05, "loss": 0.2885, "step": 15279 }, { "epoch": 1.58, "learning_rate": 3.937812723649131e-05, "loss": 0.1954, "step": 15280 }, { "epoch": 1.58, "learning_rate": 3.9375789142587426e-05, "loss": 0.9652, "step": 15281 }, { "epoch": 1.58, "learning_rate": 3.937345098556792e-05, "loss": 0.8408, "step": 15282 }, { "epoch": 1.58, "learning_rate": 3.937111276544854e-05, "loss": 0.4435, "step": 15283 }, { "epoch": 1.58, "learning_rate": 3.9368774482245026e-05, "loss": 0.8413, "step": 15284 }, { "epoch": 1.58, "learning_rate": 3.9366436135973096e-05, "loss": 1.1093, "step": 15285 }, { "epoch": 1.58, "learning_rate": 3.936409772664852e-05, "loss": 0.7323, "step": 15286 }, { "epoch": 1.58, "learning_rate": 3.936175925428703e-05, "loss": 1.1167, "step": 15287 }, { "epoch": 1.58, "learning_rate": 3.935942071890437e-05, "loss": 0.8587, "step": 15288 }, { "epoch": 1.58, "learning_rate": 3.935708212051628e-05, "loss": 0.7884, "step": 15289 }, { "epoch": 1.58, "learning_rate": 3.9354743459138494e-05, "loss": 1.1346, "step": 15290 }, { "epoch": 1.58, "learning_rate": 3.935240473478677e-05, "loss": 0.9074, "step": 15291 }, { "epoch": 1.58, "learning_rate": 3.9350065947476844e-05, "loss": 0.3696, "step": 15292 }, { "epoch": 1.58, "learning_rate": 3.934772709722446e-05, "loss": 0.9414, "step": 15293 }, { "epoch": 1.58, "learning_rate": 3.9345388184045363e-05, "loss": 0.2526, "step": 15294 }, { "epoch": 1.58, "learning_rate": 3.93430492079553e-05, "loss": 0.3707, "step": 15295 }, { "epoch": 1.58, "learning_rate": 3.934071016897003e-05, "loss": 0.3288, "step": 15296 }, { "epoch": 1.58, "learning_rate": 3.933837106710527e-05, "loss": 0.3788, "step": 15297 }, { "epoch": 1.58, "learning_rate": 3.933603190237679e-05, "loss": 0.3881, "step": 15298 }, { "epoch": 1.58, "learning_rate": 3.9333692674800314e-05, "loss": 0.4132, "step": 15299 }, { "epoch": 1.58, "learning_rate": 3.933135338439161e-05, "loss": 0.2605, "step": 15300 }, { "epoch": 1.58, "learning_rate": 3.932901403116642e-05, "loss": 0.9895, "step": 15301 }, { "epoch": 1.58, "learning_rate": 3.932667461514049e-05, "loss": 0.31, "step": 15302 }, { "epoch": 1.58, "learning_rate": 3.9324335136329575e-05, "loss": 1.1554, "step": 15303 }, { "epoch": 1.58, "learning_rate": 3.932199559474941e-05, "loss": 0.8703, "step": 15304 }, { "epoch": 1.58, "learning_rate": 3.931965599041575e-05, "loss": 1.1181, "step": 15305 }, { "epoch": 1.58, "learning_rate": 3.9317316323344346e-05, "loss": 0.7818, "step": 15306 }, { "epoch": 1.58, "learning_rate": 3.931497659355096e-05, "loss": 0.3037, "step": 15307 }, { "epoch": 1.58, "learning_rate": 3.9312636801051315e-05, "loss": 0.9548, "step": 15308 }, { "epoch": 1.58, "learning_rate": 3.9310296945861185e-05, "loss": 0.6011, "step": 15309 }, { "epoch": 1.58, "learning_rate": 3.930795702799632e-05, "loss": 0.1663, "step": 15310 }, { "epoch": 1.58, "learning_rate": 3.930561704747246e-05, "loss": 1.2761, "step": 15311 }, { "epoch": 1.58, "learning_rate": 3.930327700430537e-05, "loss": 0.4027, "step": 15312 }, { "epoch": 1.58, "learning_rate": 3.9300936898510785e-05, "loss": 0.7942, "step": 15313 }, { "epoch": 1.58, "learning_rate": 3.929859673010447e-05, "loss": 0.9236, "step": 15314 }, { "epoch": 1.58, "learning_rate": 3.929625649910219e-05, "loss": 0.1627, "step": 15315 }, { "epoch": 1.58, "learning_rate": 3.9293916205519674e-05, "loss": 1.0239, "step": 15316 }, { "epoch": 1.58, "learning_rate": 3.92915758493727e-05, "loss": 0.2454, "step": 15317 }, { "epoch": 1.58, "learning_rate": 3.9289235430677004e-05, "loss": 0.1747, "step": 15318 }, { "epoch": 1.58, "learning_rate": 3.9286894949448346e-05, "loss": 0.7908, "step": 15319 }, { "epoch": 1.58, "learning_rate": 3.9284554405702484e-05, "loss": 0.7929, "step": 15320 }, { "epoch": 1.58, "learning_rate": 3.9282213799455185e-05, "loss": 0.332, "step": 15321 }, { "epoch": 1.58, "learning_rate": 3.927987313072219e-05, "loss": 0.7959, "step": 15322 }, { "epoch": 1.58, "learning_rate": 3.9277532399519257e-05, "loss": 0.7591, "step": 15323 }, { "epoch": 1.58, "learning_rate": 3.9275191605862144e-05, "loss": 0.7285, "step": 15324 }, { "epoch": 1.59, "learning_rate": 3.9272850749766614e-05, "loss": 0.8184, "step": 15325 }, { "epoch": 1.59, "learning_rate": 3.927050983124842e-05, "loss": 0.6215, "step": 15326 }, { "epoch": 1.59, "learning_rate": 3.9268168850323335e-05, "loss": 0.3005, "step": 15327 }, { "epoch": 1.59, "learning_rate": 3.9265827807007096e-05, "loss": 0.8141, "step": 15328 }, { "epoch": 1.59, "learning_rate": 3.9263486701315467e-05, "loss": 0.4316, "step": 15329 }, { "epoch": 1.59, "learning_rate": 3.926114553326422e-05, "loss": 0.7848, "step": 15330 }, { "epoch": 1.59, "learning_rate": 3.9258804302869114e-05, "loss": 1.0359, "step": 15331 }, { "epoch": 1.59, "learning_rate": 3.9256463010145894e-05, "loss": 0.7934, "step": 15332 }, { "epoch": 1.59, "learning_rate": 3.925412165511034e-05, "loss": 1.2202, "step": 15333 }, { "epoch": 1.59, "learning_rate": 3.9251780237778196e-05, "loss": 1.2873, "step": 15334 }, { "epoch": 1.59, "learning_rate": 3.9249438758165227e-05, "loss": 0.3606, "step": 15335 }, { "epoch": 1.59, "learning_rate": 3.924709721628722e-05, "loss": 0.2445, "step": 15336 }, { "epoch": 1.59, "learning_rate": 3.92447556121599e-05, "loss": 0.7056, "step": 15337 }, { "epoch": 1.59, "learning_rate": 3.9242413945799053e-05, "loss": 1.0646, "step": 15338 }, { "epoch": 1.59, "learning_rate": 3.924007221722044e-05, "loss": 0.7025, "step": 15339 }, { "epoch": 1.59, "learning_rate": 3.923773042643982e-05, "loss": 0.2973, "step": 15340 }, { "epoch": 1.59, "learning_rate": 3.9235388573472965e-05, "loss": 0.8389, "step": 15341 }, { "epoch": 1.59, "learning_rate": 3.923304665833564e-05, "loss": 1.0085, "step": 15342 }, { "epoch": 1.59, "learning_rate": 3.923070468104359e-05, "loss": 1.1716, "step": 15343 }, { "epoch": 1.59, "learning_rate": 3.9228362641612605e-05, "loss": 1.0805, "step": 15344 }, { "epoch": 1.59, "learning_rate": 3.922602054005844e-05, "loss": 0.4575, "step": 15345 }, { "epoch": 1.59, "learning_rate": 3.9223678376396866e-05, "loss": 0.2987, "step": 15346 }, { "epoch": 1.59, "learning_rate": 3.9221336150643646e-05, "loss": 0.2007, "step": 15347 }, { "epoch": 1.59, "learning_rate": 3.9218993862814546e-05, "loss": 1.6448, "step": 15348 }, { "epoch": 1.59, "learning_rate": 3.921665151292534e-05, "loss": 0.6996, "step": 15349 }, { "epoch": 1.59, "learning_rate": 3.921430910099179e-05, "loss": 0.7012, "step": 15350 }, { "epoch": 1.59, "learning_rate": 3.921196662702967e-05, "loss": 0.7161, "step": 15351 }, { "epoch": 1.59, "learning_rate": 3.9209624091054745e-05, "loss": 1.0573, "step": 15352 }, { "epoch": 1.59, "learning_rate": 3.920728149308278e-05, "loss": 0.9725, "step": 15353 }, { "epoch": 1.59, "learning_rate": 3.920493883312956e-05, "loss": 1.1219, "step": 15354 }, { "epoch": 1.59, "learning_rate": 3.920259611121084e-05, "loss": 0.2031, "step": 15355 }, { "epoch": 1.59, "learning_rate": 3.92002533273424e-05, "loss": 0.6249, "step": 15356 }, { "epoch": 1.59, "learning_rate": 3.919791048153999e-05, "loss": 0.5551, "step": 15357 }, { "epoch": 1.59, "learning_rate": 3.919556757381942e-05, "loss": 0.3479, "step": 15358 }, { "epoch": 1.59, "learning_rate": 3.9193224604196436e-05, "loss": 0.7426, "step": 15359 }, { "epoch": 1.59, "learning_rate": 3.919088157268682e-05, "loss": 0.6366, "step": 15360 }, { "epoch": 1.59, "learning_rate": 3.918853847930633e-05, "loss": 0.4813, "step": 15361 }, { "epoch": 1.59, "learning_rate": 3.918619532407075e-05, "loss": 0.4782, "step": 15362 }, { "epoch": 1.59, "learning_rate": 3.918385210699585e-05, "loss": 0.365, "step": 15363 }, { "epoch": 1.59, "learning_rate": 3.918150882809741e-05, "loss": 0.6543, "step": 15364 }, { "epoch": 1.59, "learning_rate": 3.9179165487391205e-05, "loss": 0.3167, "step": 15365 }, { "epoch": 1.59, "learning_rate": 3.9176822084893004e-05, "loss": 0.5111, "step": 15366 }, { "epoch": 1.59, "learning_rate": 3.917447862061858e-05, "loss": 0.2271, "step": 15367 }, { "epoch": 1.59, "learning_rate": 3.917213509458371e-05, "loss": 1.1427, "step": 15368 }, { "epoch": 1.59, "learning_rate": 3.916979150680418e-05, "loss": 0.4459, "step": 15369 }, { "epoch": 1.59, "learning_rate": 3.916744785729575e-05, "loss": 0.5983, "step": 15370 }, { "epoch": 1.59, "learning_rate": 3.916510414607422e-05, "loss": 0.6469, "step": 15371 }, { "epoch": 1.59, "learning_rate": 3.9162760373155345e-05, "loss": 0.6361, "step": 15372 }, { "epoch": 1.59, "learning_rate": 3.916041653855491e-05, "loss": 0.2293, "step": 15373 }, { "epoch": 1.59, "learning_rate": 3.9158072642288696e-05, "loss": 0.4695, "step": 15374 }, { "epoch": 1.59, "learning_rate": 3.915572868437248e-05, "loss": 0.3226, "step": 15375 }, { "epoch": 1.59, "learning_rate": 3.915338466482204e-05, "loss": 0.8006, "step": 15376 }, { "epoch": 1.59, "learning_rate": 3.9151040583653156e-05, "loss": 0.7914, "step": 15377 }, { "epoch": 1.59, "learning_rate": 3.9148696440881614e-05, "loss": 0.356, "step": 15378 }, { "epoch": 1.59, "learning_rate": 3.914635223652318e-05, "loss": 0.3104, "step": 15379 }, { "epoch": 1.59, "learning_rate": 3.9144007970593644e-05, "loss": 0.9192, "step": 15380 }, { "epoch": 1.59, "learning_rate": 3.9141663643108786e-05, "loss": 0.592, "step": 15381 }, { "epoch": 1.59, "learning_rate": 3.9139319254084384e-05, "loss": 0.6522, "step": 15382 }, { "epoch": 1.59, "learning_rate": 3.913697480353623e-05, "loss": 0.3193, "step": 15383 }, { "epoch": 1.59, "learning_rate": 3.913463029148011e-05, "loss": 0.1429, "step": 15384 }, { "epoch": 1.59, "learning_rate": 3.9132285717931774e-05, "loss": 1.4248, "step": 15385 }, { "epoch": 1.59, "learning_rate": 3.9129941082907034e-05, "loss": 0.7569, "step": 15386 }, { "epoch": 1.59, "learning_rate": 3.912759638642168e-05, "loss": 0.9132, "step": 15387 }, { "epoch": 1.59, "learning_rate": 3.912525162849147e-05, "loss": 0.5019, "step": 15388 }, { "epoch": 1.59, "learning_rate": 3.91229068091322e-05, "loss": 0.2311, "step": 15389 }, { "epoch": 1.59, "learning_rate": 3.912056192835966e-05, "loss": 0.5964, "step": 15390 }, { "epoch": 1.59, "learning_rate": 3.911821698618963e-05, "loss": 0.5372, "step": 15391 }, { "epoch": 1.59, "learning_rate": 3.91158719826379e-05, "loss": 0.5994, "step": 15392 }, { "epoch": 1.59, "learning_rate": 3.9113526917720254e-05, "loss": 0.583, "step": 15393 }, { "epoch": 1.59, "learning_rate": 3.911118179145247e-05, "loss": 0.8742, "step": 15394 }, { "epoch": 1.59, "learning_rate": 3.910883660385034e-05, "loss": 0.2676, "step": 15395 }, { "epoch": 1.59, "learning_rate": 3.9106491354929666e-05, "loss": 0.5688, "step": 15396 }, { "epoch": 1.59, "learning_rate": 3.910414604470621e-05, "loss": 0.2476, "step": 15397 }, { "epoch": 1.59, "learning_rate": 3.910180067319579e-05, "loss": 0.7913, "step": 15398 }, { "epoch": 1.59, "learning_rate": 3.909945524041416e-05, "loss": 0.1416, "step": 15399 }, { "epoch": 1.59, "learning_rate": 3.909710974637713e-05, "loss": 1.1601, "step": 15400 }, { "epoch": 1.59, "learning_rate": 3.909476419110049e-05, "loss": 0.5856, "step": 15401 }, { "epoch": 1.59, "learning_rate": 3.909241857460003e-05, "loss": 0.5733, "step": 15402 }, { "epoch": 1.59, "learning_rate": 3.9090072896891526e-05, "loss": 1.8126, "step": 15403 }, { "epoch": 1.59, "learning_rate": 3.908772715799079e-05, "loss": 1.0553, "step": 15404 }, { "epoch": 1.59, "learning_rate": 3.908538135791359e-05, "loss": 0.3765, "step": 15405 }, { "epoch": 1.59, "learning_rate": 3.9083035496675726e-05, "loss": 0.3725, "step": 15406 }, { "epoch": 1.59, "learning_rate": 3.9080689574293004e-05, "loss": 0.5693, "step": 15407 }, { "epoch": 1.59, "learning_rate": 3.907834359078119e-05, "loss": 0.8615, "step": 15408 }, { "epoch": 1.59, "learning_rate": 3.9075997546156106e-05, "loss": 1.0904, "step": 15409 }, { "epoch": 1.59, "learning_rate": 3.9073651440433515e-05, "loss": 0.7675, "step": 15410 }, { "epoch": 1.59, "learning_rate": 3.907130527362924e-05, "loss": 0.4438, "step": 15411 }, { "epoch": 1.59, "learning_rate": 3.9068959045759054e-05, "loss": 0.6319, "step": 15412 }, { "epoch": 1.59, "learning_rate": 3.906661275683876e-05, "loss": 0.5316, "step": 15413 }, { "epoch": 1.59, "learning_rate": 3.906426640688414e-05, "loss": 2.0762, "step": 15414 }, { "epoch": 1.59, "learning_rate": 3.9061919995911014e-05, "loss": 1.0144, "step": 15415 }, { "epoch": 1.59, "learning_rate": 3.905957352393515e-05, "loss": 0.6625, "step": 15416 }, { "epoch": 1.59, "learning_rate": 3.9057226990972374e-05, "loss": 0.6359, "step": 15417 }, { "epoch": 1.59, "learning_rate": 3.905488039703845e-05, "loss": 0.5527, "step": 15418 }, { "epoch": 1.59, "learning_rate": 3.90525337421492e-05, "loss": 0.6897, "step": 15419 }, { "epoch": 1.59, "learning_rate": 3.9050187026320396e-05, "loss": 0.9522, "step": 15420 }, { "epoch": 1.6, "learning_rate": 3.904784024956787e-05, "loss": 0.3125, "step": 15421 }, { "epoch": 1.6, "learning_rate": 3.904549341190739e-05, "loss": 1.5173, "step": 15422 }, { "epoch": 1.6, "learning_rate": 3.9043146513354765e-05, "loss": 1.2731, "step": 15423 }, { "epoch": 1.6, "learning_rate": 3.904079955392579e-05, "loss": 0.7596, "step": 15424 }, { "epoch": 1.6, "learning_rate": 3.903845253363627e-05, "loss": 0.4176, "step": 15425 }, { "epoch": 1.6, "learning_rate": 3.9036105452502006e-05, "loss": 0.8137, "step": 15426 }, { "epoch": 1.6, "learning_rate": 3.903375831053879e-05, "loss": 0.273, "step": 15427 }, { "epoch": 1.6, "learning_rate": 3.903141110776243e-05, "loss": 0.4234, "step": 15428 }, { "epoch": 1.6, "learning_rate": 3.902906384418872e-05, "loss": 0.6247, "step": 15429 }, { "epoch": 1.6, "learning_rate": 3.902671651983347e-05, "loss": 0.395, "step": 15430 }, { "epoch": 1.6, "learning_rate": 3.902436913471248e-05, "loss": 0.5736, "step": 15431 }, { "epoch": 1.6, "learning_rate": 3.9022021688841546e-05, "loss": 0.4248, "step": 15432 }, { "epoch": 1.6, "learning_rate": 3.9019674182236466e-05, "loss": 0.7129, "step": 15433 }, { "epoch": 1.6, "learning_rate": 3.901732661491305e-05, "loss": 0.2453, "step": 15434 }, { "epoch": 1.6, "learning_rate": 3.9014978986887104e-05, "loss": 0.4784, "step": 15435 }, { "epoch": 1.6, "learning_rate": 3.901263129817444e-05, "loss": 1.0763, "step": 15436 }, { "epoch": 1.6, "learning_rate": 3.9010283548790825e-05, "loss": 0.6585, "step": 15437 }, { "epoch": 1.6, "learning_rate": 3.900793573875211e-05, "loss": 0.2912, "step": 15438 }, { "epoch": 1.6, "learning_rate": 3.9005587868074075e-05, "loss": 1.0871, "step": 15439 }, { "epoch": 1.6, "learning_rate": 3.900323993677253e-05, "loss": 0.2293, "step": 15440 }, { "epoch": 1.6, "learning_rate": 3.9000891944863284e-05, "loss": 0.1734, "step": 15441 }, { "epoch": 1.6, "learning_rate": 3.899854389236213e-05, "loss": 0.1502, "step": 15442 }, { "epoch": 1.6, "learning_rate": 3.8996195779284885e-05, "loss": 1.3126, "step": 15443 }, { "epoch": 1.6, "learning_rate": 3.8993847605647356e-05, "loss": 1.0109, "step": 15444 }, { "epoch": 1.6, "learning_rate": 3.8991499371465355e-05, "loss": 1.2347, "step": 15445 }, { "epoch": 1.6, "learning_rate": 3.8989151076754685e-05, "loss": 0.2419, "step": 15446 }, { "epoch": 1.6, "learning_rate": 3.898680272153115e-05, "loss": 0.7663, "step": 15447 }, { "epoch": 1.6, "learning_rate": 3.898445430581055e-05, "loss": 0.8234, "step": 15448 }, { "epoch": 1.6, "learning_rate": 3.898210582960872e-05, "loss": 0.7698, "step": 15449 }, { "epoch": 1.6, "learning_rate": 3.897975729294145e-05, "loss": 1.0061, "step": 15450 }, { "epoch": 1.6, "learning_rate": 3.897740869582456e-05, "loss": 0.6993, "step": 15451 }, { "epoch": 1.6, "learning_rate": 3.897506003827385e-05, "loss": 1.0298, "step": 15452 }, { "epoch": 1.6, "learning_rate": 3.897271132030514e-05, "loss": 0.5045, "step": 15453 }, { "epoch": 1.6, "learning_rate": 3.897036254193423e-05, "loss": 0.9239, "step": 15454 }, { "epoch": 1.6, "learning_rate": 3.896801370317695e-05, "loss": 0.7271, "step": 15455 }, { "epoch": 1.6, "learning_rate": 3.8965664804049095e-05, "loss": 1.156, "step": 15456 }, { "epoch": 1.6, "learning_rate": 3.896331584456648e-05, "loss": 0.9036, "step": 15457 }, { "epoch": 1.6, "learning_rate": 3.896096682474492e-05, "loss": 0.1804, "step": 15458 }, { "epoch": 1.6, "learning_rate": 3.8958617744600226e-05, "loss": 0.2886, "step": 15459 }, { "epoch": 1.6, "learning_rate": 3.895626860414822e-05, "loss": 0.8568, "step": 15460 }, { "epoch": 1.6, "learning_rate": 3.8953919403404705e-05, "loss": 0.7965, "step": 15461 }, { "epoch": 1.6, "learning_rate": 3.89515701423855e-05, "loss": 0.6004, "step": 15462 }, { "epoch": 1.6, "learning_rate": 3.894922082110641e-05, "loss": 0.794, "step": 15463 }, { "epoch": 1.6, "learning_rate": 3.894687143958328e-05, "loss": 0.4424, "step": 15464 }, { "epoch": 1.6, "learning_rate": 3.8944521997831894e-05, "loss": 0.3928, "step": 15465 }, { "epoch": 1.6, "learning_rate": 3.894217249586808e-05, "loss": 0.4201, "step": 15466 }, { "epoch": 1.6, "learning_rate": 3.893982293370765e-05, "loss": 0.9, "step": 15467 }, { "epoch": 1.6, "learning_rate": 3.8937473311366415e-05, "loss": 0.6772, "step": 15468 }, { "epoch": 1.6, "learning_rate": 3.893512362886022e-05, "loss": 0.3164, "step": 15469 }, { "epoch": 1.6, "learning_rate": 3.893277388620485e-05, "loss": 0.7895, "step": 15470 }, { "epoch": 1.6, "learning_rate": 3.893042408341613e-05, "loss": 0.2863, "step": 15471 }, { "epoch": 1.6, "learning_rate": 3.892807422050991e-05, "loss": 0.0845, "step": 15472 }, { "epoch": 1.6, "learning_rate": 3.8925724297501964e-05, "loss": 0.4368, "step": 15473 }, { "epoch": 1.6, "learning_rate": 3.892337431440814e-05, "loss": 1.0629, "step": 15474 }, { "epoch": 1.6, "learning_rate": 3.892102427124424e-05, "loss": 0.5883, "step": 15475 }, { "epoch": 1.6, "learning_rate": 3.891867416802609e-05, "loss": 0.7667, "step": 15476 }, { "epoch": 1.6, "learning_rate": 3.891632400476952e-05, "loss": 1.5638, "step": 15477 }, { "epoch": 1.6, "learning_rate": 3.8913973781490344e-05, "loss": 0.3926, "step": 15478 }, { "epoch": 1.6, "learning_rate": 3.891162349820438e-05, "loss": 0.2719, "step": 15479 }, { "epoch": 1.6, "learning_rate": 3.890927315492745e-05, "loss": 0.8738, "step": 15480 }, { "epoch": 1.6, "learning_rate": 3.890692275167537e-05, "loss": 0.3388, "step": 15481 }, { "epoch": 1.6, "learning_rate": 3.890457228846398e-05, "loss": 0.2989, "step": 15482 }, { "epoch": 1.6, "learning_rate": 3.8902221765309093e-05, "loss": 0.6575, "step": 15483 }, { "epoch": 1.6, "learning_rate": 3.889987118222653e-05, "loss": 0.5765, "step": 15484 }, { "epoch": 1.6, "learning_rate": 3.889752053923212e-05, "loss": 0.3923, "step": 15485 }, { "epoch": 1.6, "learning_rate": 3.889516983634168e-05, "loss": 1.1181, "step": 15486 }, { "epoch": 1.6, "learning_rate": 3.889281907357103e-05, "loss": 0.2493, "step": 15487 }, { "epoch": 1.6, "learning_rate": 3.889046825093602e-05, "loss": 0.6067, "step": 15488 }, { "epoch": 1.6, "learning_rate": 3.8888117368452445e-05, "loss": 0.7734, "step": 15489 }, { "epoch": 1.6, "learning_rate": 3.888576642613615e-05, "loss": 0.2405, "step": 15490 }, { "epoch": 1.6, "learning_rate": 3.888341542400295e-05, "loss": 1.078, "step": 15491 }, { "epoch": 1.6, "learning_rate": 3.8881064362068676e-05, "loss": 0.3824, "step": 15492 }, { "epoch": 1.6, "learning_rate": 3.887871324034916e-05, "loss": 0.3607, "step": 15493 }, { "epoch": 1.6, "learning_rate": 3.887636205886022e-05, "loss": 0.5231, "step": 15494 }, { "epoch": 1.6, "learning_rate": 3.887401081761768e-05, "loss": 0.9562, "step": 15495 }, { "epoch": 1.6, "learning_rate": 3.887165951663738e-05, "loss": 0.6144, "step": 15496 }, { "epoch": 1.6, "learning_rate": 3.886930815593516e-05, "loss": 0.4783, "step": 15497 }, { "epoch": 1.6, "learning_rate": 3.8866956735526825e-05, "loss": 0.4267, "step": 15498 }, { "epoch": 1.6, "learning_rate": 3.88646052554282e-05, "loss": 0.4661, "step": 15499 }, { "epoch": 1.6, "learning_rate": 3.8862253715655136e-05, "loss": 0.0984, "step": 15500 }, { "epoch": 1.6, "learning_rate": 3.8859902116223446e-05, "loss": 0.188, "step": 15501 }, { "epoch": 1.6, "learning_rate": 3.885755045714898e-05, "loss": 1.4029, "step": 15502 }, { "epoch": 1.6, "learning_rate": 3.885519873844756e-05, "loss": 1.2078, "step": 15503 }, { "epoch": 1.6, "learning_rate": 3.8852846960135e-05, "loss": 0.288, "step": 15504 }, { "epoch": 1.6, "learning_rate": 3.885049512222714e-05, "loss": 0.6328, "step": 15505 }, { "epoch": 1.6, "learning_rate": 3.8848143224739826e-05, "loss": 0.4131, "step": 15506 }, { "epoch": 1.6, "learning_rate": 3.884579126768889e-05, "loss": 0.177, "step": 15507 }, { "epoch": 1.6, "learning_rate": 3.8843439251090145e-05, "loss": 1.0503, "step": 15508 }, { "epoch": 1.6, "learning_rate": 3.8841087174959454e-05, "loss": 0.3227, "step": 15509 }, { "epoch": 1.6, "learning_rate": 3.8838735039312616e-05, "loss": 0.5019, "step": 15510 }, { "epoch": 1.6, "learning_rate": 3.883638284416548e-05, "loss": 0.9055, "step": 15511 }, { "epoch": 1.6, "learning_rate": 3.883403058953389e-05, "loss": 1.0751, "step": 15512 }, { "epoch": 1.6, "learning_rate": 3.883167827543367e-05, "loss": 0.8553, "step": 15513 }, { "epoch": 1.6, "learning_rate": 3.882932590188066e-05, "loss": 0.9462, "step": 15514 }, { "epoch": 1.6, "learning_rate": 3.882697346889069e-05, "loss": 0.3355, "step": 15515 }, { "epoch": 1.61, "learning_rate": 3.8824620976479615e-05, "loss": 1.5227, "step": 15516 }, { "epoch": 1.61, "learning_rate": 3.882226842466324e-05, "loss": 0.8016, "step": 15517 }, { "epoch": 1.61, "learning_rate": 3.881991581345743e-05, "loss": 0.5251, "step": 15518 }, { "epoch": 1.61, "learning_rate": 3.8817563142877997e-05, "loss": 1.0332, "step": 15519 }, { "epoch": 1.61, "learning_rate": 3.88152104129408e-05, "loss": 0.9927, "step": 15520 }, { "epoch": 1.61, "learning_rate": 3.881285762366167e-05, "loss": 0.2843, "step": 15521 }, { "epoch": 1.61, "learning_rate": 3.881050477505645e-05, "loss": 0.6669, "step": 15522 }, { "epoch": 1.61, "learning_rate": 3.880815186714096e-05, "loss": 0.8462, "step": 15523 }, { "epoch": 1.61, "learning_rate": 3.880579889993106e-05, "loss": 1.0771, "step": 15524 }, { "epoch": 1.61, "learning_rate": 3.880344587344258e-05, "loss": 0.6235, "step": 15525 }, { "epoch": 1.61, "learning_rate": 3.880109278769136e-05, "loss": 0.6486, "step": 15526 }, { "epoch": 1.61, "learning_rate": 3.879873964269325e-05, "loss": 0.3384, "step": 15527 }, { "epoch": 1.61, "learning_rate": 3.879638643846408e-05, "loss": 0.2054, "step": 15528 }, { "epoch": 1.61, "learning_rate": 3.879403317501969e-05, "loss": 0.433, "step": 15529 }, { "epoch": 1.61, "learning_rate": 3.879167985237593e-05, "loss": 0.2099, "step": 15530 }, { "epoch": 1.61, "learning_rate": 3.878932647054864e-05, "loss": 1.3273, "step": 15531 }, { "epoch": 1.61, "learning_rate": 3.8786973029553656e-05, "loss": 0.3779, "step": 15532 }, { "epoch": 1.61, "learning_rate": 3.878461952940683e-05, "loss": 0.3461, "step": 15533 }, { "epoch": 1.61, "learning_rate": 3.8782265970124e-05, "loss": 0.5804, "step": 15534 }, { "epoch": 1.61, "learning_rate": 3.8779912351721e-05, "loss": 0.1889, "step": 15535 }, { "epoch": 1.61, "learning_rate": 3.877755867421371e-05, "loss": 1.1906, "step": 15536 }, { "epoch": 1.61, "learning_rate": 3.877520493761793e-05, "loss": 0.9949, "step": 15537 }, { "epoch": 1.61, "learning_rate": 3.8772851141949524e-05, "loss": 0.6373, "step": 15538 }, { "epoch": 1.61, "learning_rate": 3.8770497287224335e-05, "loss": 0.5247, "step": 15539 }, { "epoch": 1.61, "learning_rate": 3.876814337345822e-05, "loss": 0.4179, "step": 15540 }, { "epoch": 1.61, "learning_rate": 3.876578940066701e-05, "loss": 0.8017, "step": 15541 }, { "epoch": 1.61, "learning_rate": 3.876343536886656e-05, "loss": 0.6927, "step": 15542 }, { "epoch": 1.61, "learning_rate": 3.87610812780727e-05, "loss": 0.3375, "step": 15543 }, { "epoch": 1.61, "learning_rate": 3.8758727128301306e-05, "loss": 0.3328, "step": 15544 }, { "epoch": 1.61, "learning_rate": 3.875637291956821e-05, "loss": 0.4911, "step": 15545 }, { "epoch": 1.61, "learning_rate": 3.8754018651889246e-05, "loss": 0.9655, "step": 15546 }, { "epoch": 1.61, "learning_rate": 3.8751664325280295e-05, "loss": 0.3897, "step": 15547 }, { "epoch": 1.61, "learning_rate": 3.874930993975717e-05, "loss": 1.0042, "step": 15548 }, { "epoch": 1.61, "learning_rate": 3.874695549533575e-05, "loss": 0.7051, "step": 15549 }, { "epoch": 1.61, "learning_rate": 3.874460099203187e-05, "loss": 0.7203, "step": 15550 }, { "epoch": 1.61, "learning_rate": 3.874224642986138e-05, "loss": 0.7414, "step": 15551 }, { "epoch": 1.61, "learning_rate": 3.873989180884013e-05, "loss": 0.5642, "step": 15552 }, { "epoch": 1.61, "learning_rate": 3.873753712898399e-05, "loss": 1.2549, "step": 15553 }, { "epoch": 1.61, "learning_rate": 3.8735182390308776e-05, "loss": 0.2339, "step": 15554 }, { "epoch": 1.61, "learning_rate": 3.8732827592830374e-05, "loss": 0.6002, "step": 15555 }, { "epoch": 1.61, "learning_rate": 3.87304727365646e-05, "loss": 1.091, "step": 15556 }, { "epoch": 1.61, "learning_rate": 3.8728117821527334e-05, "loss": 0.5898, "step": 15557 }, { "epoch": 1.61, "learning_rate": 3.872576284773442e-05, "loss": 1.0967, "step": 15558 }, { "epoch": 1.61, "learning_rate": 3.872340781520172e-05, "loss": 0.391, "step": 15559 }, { "epoch": 1.61, "learning_rate": 3.872105272394508e-05, "loss": 0.5038, "step": 15560 }, { "epoch": 1.61, "learning_rate": 3.871869757398035e-05, "loss": 0.4587, "step": 15561 }, { "epoch": 1.61, "learning_rate": 3.871634236532338e-05, "loss": 0.3448, "step": 15562 }, { "epoch": 1.61, "learning_rate": 3.8713987097990037e-05, "loss": 0.991, "step": 15563 }, { "epoch": 1.61, "learning_rate": 3.871163177199618e-05, "loss": 0.7103, "step": 15564 }, { "epoch": 1.61, "learning_rate": 3.870927638735765e-05, "loss": 1.0354, "step": 15565 }, { "epoch": 1.61, "learning_rate": 3.870692094409031e-05, "loss": 1.4452, "step": 15566 }, { "epoch": 1.61, "learning_rate": 3.870456544221001e-05, "loss": 1.172, "step": 15567 }, { "epoch": 1.61, "learning_rate": 3.8702209881732616e-05, "loss": 0.7423, "step": 15568 }, { "epoch": 1.61, "learning_rate": 3.869985426267398e-05, "loss": 0.4756, "step": 15569 }, { "epoch": 1.61, "learning_rate": 3.869749858504996e-05, "loss": 0.8464, "step": 15570 }, { "epoch": 1.61, "learning_rate": 3.869514284887642e-05, "loss": 0.7243, "step": 15571 }, { "epoch": 1.61, "learning_rate": 3.8692787054169215e-05, "loss": 0.9454, "step": 15572 }, { "epoch": 1.61, "learning_rate": 3.86904312009442e-05, "loss": 0.6716, "step": 15573 }, { "epoch": 1.61, "learning_rate": 3.868807528921723e-05, "loss": 0.4054, "step": 15574 }, { "epoch": 1.61, "learning_rate": 3.868571931900417e-05, "loss": 0.966, "step": 15575 }, { "epoch": 1.61, "learning_rate": 3.8683363290320877e-05, "loss": 0.5943, "step": 15576 }, { "epoch": 1.61, "learning_rate": 3.868100720318322e-05, "loss": 0.4818, "step": 15577 }, { "epoch": 1.61, "learning_rate": 3.8678651057607055e-05, "loss": 0.7406, "step": 15578 }, { "epoch": 1.61, "learning_rate": 3.8676294853608245e-05, "loss": 0.6743, "step": 15579 }, { "epoch": 1.61, "learning_rate": 3.867393859120264e-05, "loss": 0.7129, "step": 15580 }, { "epoch": 1.61, "learning_rate": 3.867158227040611e-05, "loss": 0.3393, "step": 15581 }, { "epoch": 1.61, "learning_rate": 3.8669225891234515e-05, "loss": 1.844, "step": 15582 }, { "epoch": 1.61, "learning_rate": 3.8666869453703734e-05, "loss": 0.9326, "step": 15583 }, { "epoch": 1.61, "learning_rate": 3.866451295782961e-05, "loss": 0.2003, "step": 15584 }, { "epoch": 1.61, "learning_rate": 3.8662156403628e-05, "loss": 1.2899, "step": 15585 }, { "epoch": 1.61, "learning_rate": 3.865979979111479e-05, "loss": 0.9638, "step": 15586 }, { "epoch": 1.61, "learning_rate": 3.865744312030583e-05, "loss": 0.2071, "step": 15587 }, { "epoch": 1.61, "learning_rate": 3.8655086391217e-05, "loss": 0.9744, "step": 15588 }, { "epoch": 1.61, "learning_rate": 3.8652729603864136e-05, "loss": 0.8051, "step": 15589 }, { "epoch": 1.61, "learning_rate": 3.865037275826313e-05, "loss": 0.4374, "step": 15590 }, { "epoch": 1.61, "learning_rate": 3.864801585442984e-05, "loss": 1.1248, "step": 15591 }, { "epoch": 1.61, "learning_rate": 3.864565889238012e-05, "loss": 1.1119, "step": 15592 }, { "epoch": 1.61, "learning_rate": 3.864330187212987e-05, "loss": 0.4403, "step": 15593 }, { "epoch": 1.61, "learning_rate": 3.8640944793694915e-05, "loss": 1.0021, "step": 15594 }, { "epoch": 1.61, "learning_rate": 3.8638587657091146e-05, "loss": 1.1039, "step": 15595 }, { "epoch": 1.61, "learning_rate": 3.8636230462334425e-05, "loss": 0.9556, "step": 15596 }, { "epoch": 1.61, "learning_rate": 3.863387320944062e-05, "loss": 0.5453, "step": 15597 }, { "epoch": 1.61, "learning_rate": 3.8631515898425606e-05, "loss": 0.2164, "step": 15598 }, { "epoch": 1.61, "learning_rate": 3.862915852930525e-05, "loss": 0.5163, "step": 15599 }, { "epoch": 1.61, "learning_rate": 3.8626801102095406e-05, "loss": 1.2046, "step": 15600 }, { "epoch": 1.61, "learning_rate": 3.862444361681196e-05, "loss": 0.8187, "step": 15601 }, { "epoch": 1.61, "learning_rate": 3.862208607347078e-05, "loss": 0.4311, "step": 15602 }, { "epoch": 1.61, "learning_rate": 3.8619728472087744e-05, "loss": 0.7479, "step": 15603 }, { "epoch": 1.61, "learning_rate": 3.86173708126787e-05, "loss": 0.8734, "step": 15604 }, { "epoch": 1.61, "learning_rate": 3.8615013095259525e-05, "loss": 1.2658, "step": 15605 }, { "epoch": 1.61, "learning_rate": 3.8612655319846106e-05, "loss": 0.4028, "step": 15606 }, { "epoch": 1.61, "learning_rate": 3.861029748645431e-05, "loss": 0.3263, "step": 15607 }, { "epoch": 1.61, "learning_rate": 3.8607939595100006e-05, "loss": 1.0909, "step": 15608 }, { "epoch": 1.61, "learning_rate": 3.860558164579906e-05, "loss": 0.888, "step": 15609 }, { "epoch": 1.61, "learning_rate": 3.860322363856736e-05, "loss": 0.9126, "step": 15610 }, { "epoch": 1.61, "learning_rate": 3.860086557342076e-05, "loss": 1.4042, "step": 15611 }, { "epoch": 1.62, "learning_rate": 3.859850745037516e-05, "loss": 1.0605, "step": 15612 }, { "epoch": 1.62, "learning_rate": 3.859614926944641e-05, "loss": 0.913, "step": 15613 }, { "epoch": 1.62, "learning_rate": 3.8593791030650394e-05, "loss": 1.2289, "step": 15614 }, { "epoch": 1.62, "learning_rate": 3.8591432734002986e-05, "loss": 1.0815, "step": 15615 }, { "epoch": 1.62, "learning_rate": 3.858907437952007e-05, "loss": 0.3161, "step": 15616 }, { "epoch": 1.62, "learning_rate": 3.858671596721752e-05, "loss": 1.0187, "step": 15617 }, { "epoch": 1.62, "learning_rate": 3.8584357497111194e-05, "loss": 1.0679, "step": 15618 }, { "epoch": 1.62, "learning_rate": 3.858199896921699e-05, "loss": 0.4512, "step": 15619 }, { "epoch": 1.62, "learning_rate": 3.857964038355077e-05, "loss": 1.1247, "step": 15620 }, { "epoch": 1.62, "learning_rate": 3.8577281740128426e-05, "loss": 0.1675, "step": 15621 }, { "epoch": 1.62, "learning_rate": 3.857492303896583e-05, "loss": 0.4397, "step": 15622 }, { "epoch": 1.62, "learning_rate": 3.857256428007885e-05, "loss": 0.8326, "step": 15623 }, { "epoch": 1.62, "learning_rate": 3.857020546348338e-05, "loss": 0.4296, "step": 15624 }, { "epoch": 1.62, "learning_rate": 3.8567846589195284e-05, "loss": 0.8229, "step": 15625 }, { "epoch": 1.62, "learning_rate": 3.856548765723046e-05, "loss": 0.5677, "step": 15626 }, { "epoch": 1.62, "learning_rate": 3.8563128667604765e-05, "loss": 0.4065, "step": 15627 }, { "epoch": 1.62, "learning_rate": 3.856076962033409e-05, "loss": 0.2279, "step": 15628 }, { "epoch": 1.62, "learning_rate": 3.8558410515434335e-05, "loss": 1.4002, "step": 15629 }, { "epoch": 1.62, "learning_rate": 3.855605135292135e-05, "loss": 0.9585, "step": 15630 }, { "epoch": 1.62, "learning_rate": 3.8553692132811034e-05, "loss": 0.4032, "step": 15631 }, { "epoch": 1.62, "learning_rate": 3.8551332855119265e-05, "loss": 0.0711, "step": 15632 }, { "epoch": 1.62, "learning_rate": 3.854897351986192e-05, "loss": 0.394, "step": 15633 }, { "epoch": 1.62, "learning_rate": 3.854661412705488e-05, "loss": 0.6127, "step": 15634 }, { "epoch": 1.62, "learning_rate": 3.8544254676714046e-05, "loss": 0.6754, "step": 15635 }, { "epoch": 1.62, "learning_rate": 3.854189516885529e-05, "loss": 0.2312, "step": 15636 }, { "epoch": 1.62, "learning_rate": 3.853953560349448e-05, "loss": 0.5053, "step": 15637 }, { "epoch": 1.62, "learning_rate": 3.853717598064752e-05, "loss": 0.3031, "step": 15638 }, { "epoch": 1.62, "learning_rate": 3.853481630033028e-05, "loss": 0.6984, "step": 15639 }, { "epoch": 1.62, "learning_rate": 3.853245656255867e-05, "loss": 0.3907, "step": 15640 }, { "epoch": 1.62, "learning_rate": 3.853009676734856e-05, "loss": 1.8145, "step": 15641 }, { "epoch": 1.62, "learning_rate": 3.852773691471583e-05, "loss": 0.0731, "step": 15642 }, { "epoch": 1.62, "learning_rate": 3.8525377004676366e-05, "loss": 0.6459, "step": 15643 }, { "epoch": 1.62, "learning_rate": 3.8523017037246055e-05, "loss": 0.6494, "step": 15644 }, { "epoch": 1.62, "learning_rate": 3.8520657012440794e-05, "loss": 0.3225, "step": 15645 }, { "epoch": 1.62, "learning_rate": 3.851829693027646e-05, "loss": 0.3093, "step": 15646 }, { "epoch": 1.62, "learning_rate": 3.851593679076895e-05, "loss": 0.3565, "step": 15647 }, { "epoch": 1.62, "learning_rate": 3.851357659393415e-05, "loss": 0.2885, "step": 15648 }, { "epoch": 1.62, "learning_rate": 3.851121633978793e-05, "loss": 0.6991, "step": 15649 }, { "epoch": 1.62, "learning_rate": 3.85088560283462e-05, "loss": 1.0995, "step": 15650 }, { "epoch": 1.62, "learning_rate": 3.850649565962485e-05, "loss": 0.8301, "step": 15651 }, { "epoch": 1.62, "learning_rate": 3.850413523363975e-05, "loss": 0.4128, "step": 15652 }, { "epoch": 1.62, "learning_rate": 3.8501774750406797e-05, "loss": 1.0872, "step": 15653 }, { "epoch": 1.62, "learning_rate": 3.8499414209941904e-05, "loss": 1.1457, "step": 15654 }, { "epoch": 1.62, "learning_rate": 3.849705361226093e-05, "loss": 0.266, "step": 15655 }, { "epoch": 1.62, "learning_rate": 3.8494692957379786e-05, "loss": 0.6934, "step": 15656 }, { "epoch": 1.62, "learning_rate": 3.849233224531435e-05, "loss": 0.2379, "step": 15657 }, { "epoch": 1.62, "learning_rate": 3.8489971476080526e-05, "loss": 0.3945, "step": 15658 }, { "epoch": 1.62, "learning_rate": 3.84876106496942e-05, "loss": 0.8116, "step": 15659 }, { "epoch": 1.62, "learning_rate": 3.848524976617126e-05, "loss": 0.2676, "step": 15660 }, { "epoch": 1.62, "learning_rate": 3.848288882552761e-05, "loss": 0.831, "step": 15661 }, { "epoch": 1.62, "learning_rate": 3.8480527827779135e-05, "loss": 0.3154, "step": 15662 }, { "epoch": 1.62, "learning_rate": 3.847816677294172e-05, "loss": 1.8653, "step": 15663 }, { "epoch": 1.62, "learning_rate": 3.847580566103129e-05, "loss": 0.6186, "step": 15664 }, { "epoch": 1.62, "learning_rate": 3.847344449206371e-05, "loss": 1.0911, "step": 15665 }, { "epoch": 1.62, "learning_rate": 3.8471083266054884e-05, "loss": 0.2856, "step": 15666 }, { "epoch": 1.62, "learning_rate": 3.846872198302071e-05, "loss": 0.6513, "step": 15667 }, { "epoch": 1.62, "learning_rate": 3.846636064297708e-05, "loss": 0.2585, "step": 15668 }, { "epoch": 1.62, "learning_rate": 3.846399924593989e-05, "loss": 1.1046, "step": 15669 }, { "epoch": 1.62, "learning_rate": 3.846163779192504e-05, "loss": 0.7299, "step": 15670 }, { "epoch": 1.62, "learning_rate": 3.8459276280948415e-05, "loss": 1.331, "step": 15671 }, { "epoch": 1.62, "learning_rate": 3.8456914713025924e-05, "loss": 0.9057, "step": 15672 }, { "epoch": 1.62, "learning_rate": 3.845455308817347e-05, "loss": 0.248, "step": 15673 }, { "epoch": 1.62, "learning_rate": 3.845219140640695e-05, "loss": 0.3955, "step": 15674 }, { "epoch": 1.62, "learning_rate": 3.844982966774224e-05, "loss": 0.1695, "step": 15675 }, { "epoch": 1.62, "learning_rate": 3.844746787219525e-05, "loss": 0.5169, "step": 15676 }, { "epoch": 1.62, "learning_rate": 3.844510601978189e-05, "loss": 0.5899, "step": 15677 }, { "epoch": 1.62, "learning_rate": 3.844274411051806e-05, "loss": 0.564, "step": 15678 }, { "epoch": 1.62, "learning_rate": 3.844038214441965e-05, "loss": 0.387, "step": 15679 }, { "epoch": 1.62, "learning_rate": 3.843802012150255e-05, "loss": 0.2208, "step": 15680 }, { "epoch": 1.62, "learning_rate": 3.843565804178267e-05, "loss": 0.635, "step": 15681 }, { "epoch": 1.62, "learning_rate": 3.843329590527593e-05, "loss": 1.1856, "step": 15682 }, { "epoch": 1.62, "learning_rate": 3.843093371199822e-05, "loss": 0.5624, "step": 15683 }, { "epoch": 1.62, "learning_rate": 3.842857146196542e-05, "loss": 0.2878, "step": 15684 }, { "epoch": 1.62, "learning_rate": 3.842620915519346e-05, "loss": 1.0071, "step": 15685 }, { "epoch": 1.62, "learning_rate": 3.842384679169822e-05, "loss": 0.6072, "step": 15686 }, { "epoch": 1.62, "learning_rate": 3.842148437149562e-05, "loss": 0.6907, "step": 15687 }, { "epoch": 1.62, "learning_rate": 3.841912189460156e-05, "loss": 0.5909, "step": 15688 }, { "epoch": 1.62, "learning_rate": 3.841675936103194e-05, "loss": 1.2564, "step": 15689 }, { "epoch": 1.62, "learning_rate": 3.841439677080266e-05, "loss": 0.2909, "step": 15690 }, { "epoch": 1.62, "learning_rate": 3.8412034123929645e-05, "loss": 0.4312, "step": 15691 }, { "epoch": 1.62, "learning_rate": 3.840967142042877e-05, "loss": 0.1982, "step": 15692 }, { "epoch": 1.62, "learning_rate": 3.840730866031596e-05, "loss": 0.4042, "step": 15693 }, { "epoch": 1.62, "learning_rate": 3.840494584360712e-05, "loss": 0.7746, "step": 15694 }, { "epoch": 1.62, "learning_rate": 3.840258297031814e-05, "loss": 0.288, "step": 15695 }, { "epoch": 1.62, "learning_rate": 3.8400220040464946e-05, "loss": 0.9221, "step": 15696 }, { "epoch": 1.62, "learning_rate": 3.839785705406343e-05, "loss": 0.2272, "step": 15697 }, { "epoch": 1.62, "learning_rate": 3.839549401112951e-05, "loss": 0.665, "step": 15698 }, { "epoch": 1.62, "learning_rate": 3.839313091167909e-05, "loss": 0.0717, "step": 15699 }, { "epoch": 1.62, "learning_rate": 3.8390767755728076e-05, "loss": 1.2049, "step": 15700 }, { "epoch": 1.62, "learning_rate": 3.838840454329237e-05, "loss": 0.4383, "step": 15701 }, { "epoch": 1.62, "learning_rate": 3.83860412743879e-05, "loss": 0.9553, "step": 15702 }, { "epoch": 1.62, "learning_rate": 3.838367794903055e-05, "loss": 0.3577, "step": 15703 }, { "epoch": 1.62, "learning_rate": 3.838131456723625e-05, "loss": 0.4425, "step": 15704 }, { "epoch": 1.62, "learning_rate": 3.837895112902089e-05, "loss": 0.3039, "step": 15705 }, { "epoch": 1.62, "learning_rate": 3.83765876344004e-05, "loss": 0.8579, "step": 15706 }, { "epoch": 1.62, "learning_rate": 3.837422408339069e-05, "loss": 0.4545, "step": 15707 }, { "epoch": 1.63, "learning_rate": 3.837186047600764e-05, "loss": 0.9354, "step": 15708 }, { "epoch": 1.63, "learning_rate": 3.8369496812267207e-05, "loss": 0.6918, "step": 15709 }, { "epoch": 1.63, "learning_rate": 3.836713309218527e-05, "loss": 0.3684, "step": 15710 }, { "epoch": 1.63, "learning_rate": 3.836476931577775e-05, "loss": 0.9422, "step": 15711 }, { "epoch": 1.63, "learning_rate": 3.836240548306057e-05, "loss": 1.0187, "step": 15712 }, { "epoch": 1.63, "learning_rate": 3.8360041594049614e-05, "loss": 1.2783, "step": 15713 }, { "epoch": 1.63, "learning_rate": 3.8357677648760834e-05, "loss": 1.5205, "step": 15714 }, { "epoch": 1.63, "learning_rate": 3.835531364721011e-05, "loss": 0.1557, "step": 15715 }, { "epoch": 1.63, "learning_rate": 3.8352949589413376e-05, "loss": 1.026, "step": 15716 }, { "epoch": 1.63, "learning_rate": 3.835058547538654e-05, "loss": 0.5805, "step": 15717 }, { "epoch": 1.63, "learning_rate": 3.834822130514551e-05, "loss": 0.2859, "step": 15718 }, { "epoch": 1.63, "learning_rate": 3.834585707870622e-05, "loss": 1.2168, "step": 15719 }, { "epoch": 1.63, "learning_rate": 3.8343492796084554e-05, "loss": 0.8811, "step": 15720 }, { "epoch": 1.63, "learning_rate": 3.834112845729646e-05, "loss": 0.5865, "step": 15721 }, { "epoch": 1.63, "learning_rate": 3.833876406235784e-05, "loss": 0.4354, "step": 15722 }, { "epoch": 1.63, "learning_rate": 3.8336399611284615e-05, "loss": 0.7751, "step": 15723 }, { "epoch": 1.63, "learning_rate": 3.833403510409269e-05, "loss": 0.6206, "step": 15724 }, { "epoch": 1.63, "learning_rate": 3.833167054079799e-05, "loss": 0.2708, "step": 15725 }, { "epoch": 1.63, "learning_rate": 3.832930592141643e-05, "loss": 0.3524, "step": 15726 }, { "epoch": 1.63, "learning_rate": 3.832694124596395e-05, "loss": 0.7064, "step": 15727 }, { "epoch": 1.63, "learning_rate": 3.832457651445643e-05, "loss": 1.1209, "step": 15728 }, { "epoch": 1.63, "learning_rate": 3.832221172690982e-05, "loss": 0.2183, "step": 15729 }, { "epoch": 1.63, "learning_rate": 3.831984688334002e-05, "loss": 1.0, "step": 15730 }, { "epoch": 1.63, "learning_rate": 3.831748198376297e-05, "loss": 0.7648, "step": 15731 }, { "epoch": 1.63, "learning_rate": 3.831511702819457e-05, "loss": 0.3352, "step": 15732 }, { "epoch": 1.63, "learning_rate": 3.831275201665074e-05, "loss": 1.0058, "step": 15733 }, { "epoch": 1.63, "learning_rate": 3.8310386949147424e-05, "loss": 0.277, "step": 15734 }, { "epoch": 1.63, "learning_rate": 3.830802182570052e-05, "loss": 0.2432, "step": 15735 }, { "epoch": 1.63, "learning_rate": 3.830565664632596e-05, "loss": 0.5414, "step": 15736 }, { "epoch": 1.63, "learning_rate": 3.8303291411039656e-05, "loss": 0.3074, "step": 15737 }, { "epoch": 1.63, "learning_rate": 3.830092611985754e-05, "loss": 0.2239, "step": 15738 }, { "epoch": 1.63, "learning_rate": 3.829856077279553e-05, "loss": 0.4492, "step": 15739 }, { "epoch": 1.63, "learning_rate": 3.8296195369869555e-05, "loss": 0.9638, "step": 15740 }, { "epoch": 1.63, "learning_rate": 3.8293829911095535e-05, "loss": 0.734, "step": 15741 }, { "epoch": 1.63, "learning_rate": 3.829146439648939e-05, "loss": 0.7034, "step": 15742 }, { "epoch": 1.63, "learning_rate": 3.8289098826067046e-05, "loss": 0.4693, "step": 15743 }, { "epoch": 1.63, "learning_rate": 3.828673319984442e-05, "loss": 1.4433, "step": 15744 }, { "epoch": 1.63, "learning_rate": 3.828436751783746e-05, "loss": 0.2158, "step": 15745 }, { "epoch": 1.63, "learning_rate": 3.8282001780062063e-05, "loss": 1.154, "step": 15746 }, { "epoch": 1.63, "learning_rate": 3.827963598653418e-05, "loss": 0.1709, "step": 15747 }, { "epoch": 1.63, "learning_rate": 3.827727013726972e-05, "loss": 0.3901, "step": 15748 }, { "epoch": 1.63, "learning_rate": 3.827490423228461e-05, "loss": 0.5042, "step": 15749 }, { "epoch": 1.63, "learning_rate": 3.8272538271594786e-05, "loss": 0.688, "step": 15750 }, { "epoch": 1.63, "learning_rate": 3.8270172255216165e-05, "loss": 0.4075, "step": 15751 }, { "epoch": 1.63, "learning_rate": 3.8267806183164684e-05, "loss": 0.4241, "step": 15752 }, { "epoch": 1.63, "learning_rate": 3.826544005545627e-05, "loss": 0.383, "step": 15753 }, { "epoch": 1.63, "learning_rate": 3.826307387210683e-05, "loss": 0.4798, "step": 15754 }, { "epoch": 1.63, "learning_rate": 3.826070763313234e-05, "loss": 0.7858, "step": 15755 }, { "epoch": 1.63, "learning_rate": 3.825834133854868e-05, "loss": 1.2124, "step": 15756 }, { "epoch": 1.63, "learning_rate": 3.825597498837179e-05, "loss": 0.4691, "step": 15757 }, { "epoch": 1.63, "learning_rate": 3.825360858261762e-05, "loss": 0.4093, "step": 15758 }, { "epoch": 1.63, "learning_rate": 3.825124212130209e-05, "loss": 0.5642, "step": 15759 }, { "epoch": 1.63, "learning_rate": 3.824887560444112e-05, "loss": 0.7676, "step": 15760 }, { "epoch": 1.63, "learning_rate": 3.824650903205066e-05, "loss": 0.1762, "step": 15761 }, { "epoch": 1.63, "learning_rate": 3.824414240414662e-05, "loss": 0.707, "step": 15762 }, { "epoch": 1.63, "learning_rate": 3.824177572074495e-05, "loss": 1.549, "step": 15763 }, { "epoch": 1.63, "learning_rate": 3.823940898186158e-05, "loss": 0.4997, "step": 15764 }, { "epoch": 1.63, "learning_rate": 3.823704218751242e-05, "loss": 0.4539, "step": 15765 }, { "epoch": 1.63, "learning_rate": 3.8234675337713425e-05, "loss": 0.361, "step": 15766 }, { "epoch": 1.63, "learning_rate": 3.823230843248053e-05, "loss": 0.365, "step": 15767 }, { "epoch": 1.63, "learning_rate": 3.822994147182965e-05, "loss": 0.6127, "step": 15768 }, { "epoch": 1.63, "learning_rate": 3.822757445577674e-05, "loss": 0.7857, "step": 15769 }, { "epoch": 1.63, "learning_rate": 3.8225207384337724e-05, "loss": 0.2869, "step": 15770 }, { "epoch": 1.63, "learning_rate": 3.822284025752853e-05, "loss": 0.9474, "step": 15771 }, { "epoch": 1.63, "learning_rate": 3.82204730753651e-05, "loss": 1.1545, "step": 15772 }, { "epoch": 1.63, "learning_rate": 3.821810583786337e-05, "loss": 0.3571, "step": 15773 }, { "epoch": 1.63, "learning_rate": 3.821573854503928e-05, "loss": 0.7511, "step": 15774 }, { "epoch": 1.63, "learning_rate": 3.8213371196908756e-05, "loss": 0.4527, "step": 15775 }, { "epoch": 1.63, "learning_rate": 3.821100379348773e-05, "loss": 0.7132, "step": 15776 }, { "epoch": 1.63, "learning_rate": 3.8208636334792155e-05, "loss": 0.5397, "step": 15777 }, { "epoch": 1.63, "learning_rate": 3.8206268820837965e-05, "loss": 0.4377, "step": 15778 }, { "epoch": 1.63, "learning_rate": 3.820390125164109e-05, "loss": 0.469, "step": 15779 }, { "epoch": 1.63, "learning_rate": 3.820153362721747e-05, "loss": 1.1105, "step": 15780 }, { "epoch": 1.63, "learning_rate": 3.8199165947583045e-05, "loss": 0.2027, "step": 15781 }, { "epoch": 1.63, "learning_rate": 3.819679821275375e-05, "loss": 0.4766, "step": 15782 }, { "epoch": 1.63, "learning_rate": 3.8194430422745535e-05, "loss": 1.2514, "step": 15783 }, { "epoch": 1.63, "learning_rate": 3.8192062577574324e-05, "loss": 1.0093, "step": 15784 }, { "epoch": 1.63, "learning_rate": 3.818969467725607e-05, "loss": 0.6498, "step": 15785 }, { "epoch": 1.63, "learning_rate": 3.8187326721806706e-05, "loss": 0.6333, "step": 15786 }, { "epoch": 1.63, "learning_rate": 3.8184958711242174e-05, "loss": 0.2678, "step": 15787 }, { "epoch": 1.63, "learning_rate": 3.818259064557843e-05, "loss": 0.7495, "step": 15788 }, { "epoch": 1.63, "learning_rate": 3.8180222524831376e-05, "loss": 0.3828, "step": 15789 }, { "epoch": 1.63, "learning_rate": 3.8177854349016985e-05, "loss": 0.6362, "step": 15790 }, { "epoch": 1.63, "learning_rate": 3.817548611815119e-05, "loss": 0.4123, "step": 15791 }, { "epoch": 1.63, "learning_rate": 3.817311783224994e-05, "loss": 0.3801, "step": 15792 }, { "epoch": 1.63, "learning_rate": 3.817074949132918e-05, "loss": 0.4483, "step": 15793 }, { "epoch": 1.63, "learning_rate": 3.816838109540484e-05, "loss": 0.729, "step": 15794 }, { "epoch": 1.63, "learning_rate": 3.816601264449287e-05, "loss": 1.258, "step": 15795 }, { "epoch": 1.63, "learning_rate": 3.8163644138609205e-05, "loss": 0.7548, "step": 15796 }, { "epoch": 1.63, "learning_rate": 3.8161275577769815e-05, "loss": 0.5658, "step": 15797 }, { "epoch": 1.63, "learning_rate": 3.815890696199062e-05, "loss": 0.3207, "step": 15798 }, { "epoch": 1.63, "learning_rate": 3.8156538291287567e-05, "loss": 0.289, "step": 15799 }, { "epoch": 1.63, "learning_rate": 3.815416956567661e-05, "loss": 0.88, "step": 15800 }, { "epoch": 1.63, "learning_rate": 3.815180078517369e-05, "loss": 1.0038, "step": 15801 }, { "epoch": 1.63, "learning_rate": 3.814943194979476e-05, "loss": 0.1554, "step": 15802 }, { "epoch": 1.63, "learning_rate": 3.814706305955575e-05, "loss": 0.7166, "step": 15803 }, { "epoch": 1.64, "learning_rate": 3.814469411447263e-05, "loss": 0.6852, "step": 15804 }, { "epoch": 1.64, "learning_rate": 3.814232511456133e-05, "loss": 0.9506, "step": 15805 }, { "epoch": 1.64, "learning_rate": 3.813995605983779e-05, "loss": 0.7497, "step": 15806 }, { "epoch": 1.64, "learning_rate": 3.813758695031799e-05, "loss": 0.2193, "step": 15807 }, { "epoch": 1.64, "learning_rate": 3.813521778601785e-05, "loss": 0.1508, "step": 15808 }, { "epoch": 1.64, "learning_rate": 3.813284856695333e-05, "loss": 1.3693, "step": 15809 }, { "epoch": 1.64, "learning_rate": 3.8130479293140376e-05, "loss": 0.8975, "step": 15810 }, { "epoch": 1.64, "learning_rate": 3.812810996459494e-05, "loss": 0.2801, "step": 15811 }, { "epoch": 1.64, "learning_rate": 3.812574058133297e-05, "loss": 0.9841, "step": 15812 }, { "epoch": 1.64, "learning_rate": 3.812337114337041e-05, "loss": 0.6518, "step": 15813 }, { "epoch": 1.64, "learning_rate": 3.812100165072321e-05, "loss": 0.2825, "step": 15814 }, { "epoch": 1.64, "learning_rate": 3.811863210340733e-05, "loss": 0.3537, "step": 15815 }, { "epoch": 1.64, "learning_rate": 3.811626250143873e-05, "loss": 0.7314, "step": 15816 }, { "epoch": 1.64, "learning_rate": 3.811389284483335e-05, "loss": 0.7114, "step": 15817 }, { "epoch": 1.64, "learning_rate": 3.8111523133607134e-05, "loss": 0.3917, "step": 15818 }, { "epoch": 1.64, "learning_rate": 3.8109153367776036e-05, "loss": 0.2153, "step": 15819 }, { "epoch": 1.64, "learning_rate": 3.8106783547356024e-05, "loss": 0.7092, "step": 15820 }, { "epoch": 1.64, "learning_rate": 3.810441367236305e-05, "loss": 0.4744, "step": 15821 }, { "epoch": 1.64, "learning_rate": 3.810204374281304e-05, "loss": 0.8599, "step": 15822 }, { "epoch": 1.64, "learning_rate": 3.809967375872198e-05, "loss": 0.2699, "step": 15823 }, { "epoch": 1.64, "learning_rate": 3.8097303720105806e-05, "loss": 1.3069, "step": 15824 }, { "epoch": 1.64, "learning_rate": 3.809493362698048e-05, "loss": 0.6844, "step": 15825 }, { "epoch": 1.64, "learning_rate": 3.809256347936196e-05, "loss": 0.3092, "step": 15826 }, { "epoch": 1.64, "learning_rate": 3.809019327726619e-05, "loss": 0.4624, "step": 15827 }, { "epoch": 1.64, "learning_rate": 3.808782302070913e-05, "loss": 0.8502, "step": 15828 }, { "epoch": 1.64, "learning_rate": 3.8085452709706736e-05, "loss": 0.4288, "step": 15829 }, { "epoch": 1.64, "learning_rate": 3.808308234427498e-05, "loss": 0.696, "step": 15830 }, { "epoch": 1.64, "learning_rate": 3.8080711924429796e-05, "loss": 0.271, "step": 15831 }, { "epoch": 1.64, "learning_rate": 3.8078341450187146e-05, "loss": 0.756, "step": 15832 }, { "epoch": 1.64, "learning_rate": 3.8075970921563e-05, "loss": 0.6459, "step": 15833 }, { "epoch": 1.64, "learning_rate": 3.80736003385733e-05, "loss": 0.413, "step": 15834 }, { "epoch": 1.64, "learning_rate": 3.807122970123402e-05, "loss": 1.1061, "step": 15835 }, { "epoch": 1.64, "learning_rate": 3.806885900956111e-05, "loss": 0.5927, "step": 15836 }, { "epoch": 1.64, "learning_rate": 3.806648826357052e-05, "loss": 0.6666, "step": 15837 }, { "epoch": 1.64, "learning_rate": 3.806411746327822e-05, "loss": 1.0315, "step": 15838 }, { "epoch": 1.64, "learning_rate": 3.806174660870017e-05, "loss": 0.5486, "step": 15839 }, { "epoch": 1.64, "learning_rate": 3.805937569985233e-05, "loss": 0.8883, "step": 15840 }, { "epoch": 1.64, "learning_rate": 3.8057004736750656e-05, "loss": 0.8023, "step": 15841 }, { "epoch": 1.64, "learning_rate": 3.805463371941111e-05, "loss": 0.881, "step": 15842 }, { "epoch": 1.64, "learning_rate": 3.8052262647849664e-05, "loss": 0.1751, "step": 15843 }, { "epoch": 1.64, "learning_rate": 3.8049891522082254e-05, "loss": 0.6988, "step": 15844 }, { "epoch": 1.64, "learning_rate": 3.804752034212487e-05, "loss": 0.3514, "step": 15845 }, { "epoch": 1.64, "learning_rate": 3.804514910799346e-05, "loss": 0.2329, "step": 15846 }, { "epoch": 1.64, "learning_rate": 3.804277781970399e-05, "loss": 0.881, "step": 15847 }, { "epoch": 1.64, "learning_rate": 3.8040406477272416e-05, "loss": 1.4709, "step": 15848 }, { "epoch": 1.64, "learning_rate": 3.803803508071471e-05, "loss": 0.5513, "step": 15849 }, { "epoch": 1.64, "learning_rate": 3.8035663630046834e-05, "loss": 0.4475, "step": 15850 }, { "epoch": 1.64, "learning_rate": 3.803329212528475e-05, "loss": 0.8873, "step": 15851 }, { "epoch": 1.64, "learning_rate": 3.803092056644442e-05, "loss": 1.2056, "step": 15852 }, { "epoch": 1.64, "learning_rate": 3.8028548953541806e-05, "loss": 0.894, "step": 15853 }, { "epoch": 1.64, "learning_rate": 3.8026177286592895e-05, "loss": 0.813, "step": 15854 }, { "epoch": 1.64, "learning_rate": 3.802380556561363e-05, "loss": 0.4028, "step": 15855 }, { "epoch": 1.64, "learning_rate": 3.802143379061998e-05, "loss": 1.444, "step": 15856 }, { "epoch": 1.64, "learning_rate": 3.801906196162791e-05, "loss": 0.5695, "step": 15857 }, { "epoch": 1.64, "learning_rate": 3.8016690078653397e-05, "loss": 1.1029, "step": 15858 }, { "epoch": 1.64, "learning_rate": 3.801431814171241e-05, "loss": 0.2641, "step": 15859 }, { "epoch": 1.64, "learning_rate": 3.80119461508209e-05, "loss": 0.4261, "step": 15860 }, { "epoch": 1.64, "learning_rate": 3.800957410599485e-05, "loss": 0.5151, "step": 15861 }, { "epoch": 1.64, "learning_rate": 3.800720200725021e-05, "loss": 0.356, "step": 15862 }, { "epoch": 1.64, "learning_rate": 3.800482985460296e-05, "loss": 0.5355, "step": 15863 }, { "epoch": 1.64, "learning_rate": 3.800245764806908e-05, "loss": 0.3828, "step": 15864 }, { "epoch": 1.64, "learning_rate": 3.800008538766452e-05, "loss": 0.32, "step": 15865 }, { "epoch": 1.64, "learning_rate": 3.799771307340526e-05, "loss": 0.4218, "step": 15866 }, { "epoch": 1.64, "learning_rate": 3.799534070530727e-05, "loss": 0.389, "step": 15867 }, { "epoch": 1.64, "learning_rate": 3.799296828338651e-05, "loss": 0.6062, "step": 15868 }, { "epoch": 1.64, "learning_rate": 3.7990595807658954e-05, "loss": 1.4515, "step": 15869 }, { "epoch": 1.64, "learning_rate": 3.798822327814058e-05, "loss": 1.3724, "step": 15870 }, { "epoch": 1.64, "learning_rate": 3.798585069484736e-05, "loss": 0.7567, "step": 15871 }, { "epoch": 1.64, "learning_rate": 3.798347805779526e-05, "loss": 0.7449, "step": 15872 }, { "epoch": 1.64, "learning_rate": 3.7981105367000255e-05, "loss": 0.2637, "step": 15873 }, { "epoch": 1.64, "learning_rate": 3.7978732622478316e-05, "loss": 0.3261, "step": 15874 }, { "epoch": 1.64, "learning_rate": 3.7976359824245415e-05, "loss": 0.5912, "step": 15875 }, { "epoch": 1.64, "learning_rate": 3.7973986972317524e-05, "loss": 0.519, "step": 15876 }, { "epoch": 1.64, "learning_rate": 3.797161406671062e-05, "loss": 0.7679, "step": 15877 }, { "epoch": 1.64, "learning_rate": 3.7969241107440676e-05, "loss": 0.912, "step": 15878 }, { "epoch": 1.64, "learning_rate": 3.796686809452367e-05, "loss": 0.4198, "step": 15879 }, { "epoch": 1.64, "learning_rate": 3.7964495027975565e-05, "loss": 0.4818, "step": 15880 }, { "epoch": 1.64, "learning_rate": 3.796212190781234e-05, "loss": 0.2805, "step": 15881 }, { "epoch": 1.64, "learning_rate": 3.795974873404999e-05, "loss": 0.3832, "step": 15882 }, { "epoch": 1.64, "learning_rate": 3.7957375506704455e-05, "loss": 0.9202, "step": 15883 }, { "epoch": 1.64, "learning_rate": 3.7955002225791733e-05, "loss": 0.6392, "step": 15884 }, { "epoch": 1.64, "learning_rate": 3.7952628891327807e-05, "loss": 0.3223, "step": 15885 }, { "epoch": 1.64, "learning_rate": 3.795025550332864e-05, "loss": 0.7923, "step": 15886 }, { "epoch": 1.64, "learning_rate": 3.794788206181021e-05, "loss": 0.3259, "step": 15887 }, { "epoch": 1.64, "learning_rate": 3.79455085667885e-05, "loss": 0.1283, "step": 15888 }, { "epoch": 1.64, "learning_rate": 3.794313501827949e-05, "loss": 0.3799, "step": 15889 }, { "epoch": 1.64, "learning_rate": 3.7940761416299144e-05, "loss": 0.245, "step": 15890 }, { "epoch": 1.64, "learning_rate": 3.793838776086345e-05, "loss": 0.9422, "step": 15891 }, { "epoch": 1.64, "learning_rate": 3.793601405198839e-05, "loss": 0.4723, "step": 15892 }, { "epoch": 1.64, "learning_rate": 3.7933640289689945e-05, "loss": 0.2744, "step": 15893 }, { "epoch": 1.64, "learning_rate": 3.7931266473984093e-05, "loss": 0.3984, "step": 15894 }, { "epoch": 1.64, "learning_rate": 3.79288926048868e-05, "loss": 0.4079, "step": 15895 }, { "epoch": 1.64, "learning_rate": 3.792651868241406e-05, "loss": 0.8556, "step": 15896 }, { "epoch": 1.64, "learning_rate": 3.792414470658186e-05, "loss": 0.966, "step": 15897 }, { "epoch": 1.64, "learning_rate": 3.792177067740616e-05, "loss": 0.6029, "step": 15898 }, { "epoch": 1.65, "learning_rate": 3.791939659490296e-05, "loss": 1.1265, "step": 15899 }, { "epoch": 1.65, "learning_rate": 3.791702245908823e-05, "loss": 0.4686, "step": 15900 }, { "epoch": 1.65, "learning_rate": 3.791464826997796e-05, "loss": 0.4733, "step": 15901 }, { "epoch": 1.65, "learning_rate": 3.7912274027588136e-05, "loss": 0.1871, "step": 15902 }, { "epoch": 1.65, "learning_rate": 3.790989973193473e-05, "loss": 0.9496, "step": 15903 }, { "epoch": 1.65, "learning_rate": 3.7907525383033724e-05, "loss": 0.273, "step": 15904 }, { "epoch": 1.65, "learning_rate": 3.790515098090112e-05, "loss": 0.523, "step": 15905 }, { "epoch": 1.65, "learning_rate": 3.7902776525552875e-05, "loss": 0.3161, "step": 15906 }, { "epoch": 1.65, "learning_rate": 3.7900402017005e-05, "loss": 0.8505, "step": 15907 }, { "epoch": 1.65, "learning_rate": 3.789802745527345e-05, "loss": 0.6109, "step": 15908 }, { "epoch": 1.65, "learning_rate": 3.7895652840374244e-05, "loss": 0.3715, "step": 15909 }, { "epoch": 1.65, "learning_rate": 3.789327817232334e-05, "loss": 0.9225, "step": 15910 }, { "epoch": 1.65, "learning_rate": 3.789090345113674e-05, "loss": 0.9089, "step": 15911 }, { "epoch": 1.65, "learning_rate": 3.788852867683043e-05, "loss": 1.0996, "step": 15912 }, { "epoch": 1.65, "learning_rate": 3.788615384942038e-05, "loss": 0.3611, "step": 15913 }, { "epoch": 1.65, "learning_rate": 3.788377896892259e-05, "loss": 1.1408, "step": 15914 }, { "epoch": 1.65, "learning_rate": 3.7881404035353045e-05, "loss": 0.8036, "step": 15915 }, { "epoch": 1.65, "learning_rate": 3.787902904872774e-05, "loss": 0.6922, "step": 15916 }, { "epoch": 1.65, "learning_rate": 3.787665400906264e-05, "loss": 0.6976, "step": 15917 }, { "epoch": 1.65, "learning_rate": 3.7874278916373765e-05, "loss": 0.2003, "step": 15918 }, { "epoch": 1.65, "learning_rate": 3.7871903770677075e-05, "loss": 0.6831, "step": 15919 }, { "epoch": 1.65, "learning_rate": 3.786952857198857e-05, "loss": 0.2296, "step": 15920 }, { "epoch": 1.65, "learning_rate": 3.7867153320324246e-05, "loss": 0.8035, "step": 15921 }, { "epoch": 1.65, "learning_rate": 3.7864778015700085e-05, "loss": 0.297, "step": 15922 }, { "epoch": 1.65, "learning_rate": 3.7862402658132075e-05, "loss": 1.5639, "step": 15923 }, { "epoch": 1.65, "learning_rate": 3.7860027247636215e-05, "loss": 0.7688, "step": 15924 }, { "epoch": 1.65, "learning_rate": 3.785765178422849e-05, "loss": 0.6493, "step": 15925 }, { "epoch": 1.65, "learning_rate": 3.7855276267924894e-05, "loss": 0.457, "step": 15926 }, { "epoch": 1.65, "learning_rate": 3.78529006987414e-05, "loss": 0.7583, "step": 15927 }, { "epoch": 1.65, "learning_rate": 3.785052507669403e-05, "loss": 1.0265, "step": 15928 }, { "epoch": 1.65, "learning_rate": 3.784814940179876e-05, "loss": 1.2443, "step": 15929 }, { "epoch": 1.65, "learning_rate": 3.784577367407159e-05, "loss": 1.1235, "step": 15930 }, { "epoch": 1.65, "learning_rate": 3.7843397893528506e-05, "loss": 1.4084, "step": 15931 }, { "epoch": 1.65, "learning_rate": 3.78410220601855e-05, "loss": 0.8907, "step": 15932 }, { "epoch": 1.65, "learning_rate": 3.783864617405856e-05, "loss": 1.1786, "step": 15933 }, { "epoch": 1.65, "learning_rate": 3.7836270235163695e-05, "loss": 0.3158, "step": 15934 }, { "epoch": 1.65, "learning_rate": 3.78338942435169e-05, "loss": 0.4224, "step": 15935 }, { "epoch": 1.65, "learning_rate": 3.7831518199134154e-05, "loss": 0.8065, "step": 15936 }, { "epoch": 1.65, "learning_rate": 3.782914210203146e-05, "loss": 0.3271, "step": 15937 }, { "epoch": 1.65, "learning_rate": 3.782676595222481e-05, "loss": 0.3864, "step": 15938 }, { "epoch": 1.65, "learning_rate": 3.78243897497302e-05, "loss": 0.3693, "step": 15939 }, { "epoch": 1.65, "learning_rate": 3.7822013494563646e-05, "loss": 1.0576, "step": 15940 }, { "epoch": 1.65, "learning_rate": 3.781963718674112e-05, "loss": 0.257, "step": 15941 }, { "epoch": 1.65, "learning_rate": 3.781726082627861e-05, "loss": 0.8748, "step": 15942 }, { "epoch": 1.65, "learning_rate": 3.781488441319216e-05, "loss": 1.4871, "step": 15943 }, { "epoch": 1.65, "learning_rate": 3.781250794749771e-05, "loss": 0.84, "step": 15944 }, { "epoch": 1.65, "learning_rate": 3.78101314292113e-05, "loss": 1.1304, "step": 15945 }, { "epoch": 1.65, "learning_rate": 3.7807754858348896e-05, "loss": 0.8996, "step": 15946 }, { "epoch": 1.65, "learning_rate": 3.780537823492653e-05, "loss": 0.9134, "step": 15947 }, { "epoch": 1.65, "learning_rate": 3.7803001558960166e-05, "loss": 0.2634, "step": 15948 }, { "epoch": 1.65, "learning_rate": 3.780062483046584e-05, "loss": 1.2863, "step": 15949 }, { "epoch": 1.65, "learning_rate": 3.7798248049459526e-05, "loss": 0.8152, "step": 15950 }, { "epoch": 1.65, "learning_rate": 3.779587121595723e-05, "loss": 0.4908, "step": 15951 }, { "epoch": 1.65, "learning_rate": 3.7793494329974946e-05, "loss": 0.2693, "step": 15952 }, { "epoch": 1.65, "learning_rate": 3.7791117391528686e-05, "loss": 0.1687, "step": 15953 }, { "epoch": 1.65, "learning_rate": 3.778874040063446e-05, "loss": 0.9526, "step": 15954 }, { "epoch": 1.65, "learning_rate": 3.778636335730825e-05, "loss": 0.3585, "step": 15955 }, { "epoch": 1.65, "learning_rate": 3.7783986261566055e-05, "loss": 0.1784, "step": 15956 }, { "epoch": 1.65, "learning_rate": 3.778160911342389e-05, "loss": 0.8653, "step": 15957 }, { "epoch": 1.65, "learning_rate": 3.777923191289775e-05, "loss": 0.2041, "step": 15958 }, { "epoch": 1.65, "learning_rate": 3.777685466000364e-05, "loss": 0.4483, "step": 15959 }, { "epoch": 1.65, "learning_rate": 3.777447735475757e-05, "loss": 1.898, "step": 15960 }, { "epoch": 1.65, "learning_rate": 3.7772099997175535e-05, "loss": 0.3172, "step": 15961 }, { "epoch": 1.65, "learning_rate": 3.776972258727355e-05, "loss": 0.3217, "step": 15962 }, { "epoch": 1.65, "learning_rate": 3.7767345125067596e-05, "loss": 0.602, "step": 15963 }, { "epoch": 1.65, "learning_rate": 3.77649676105737e-05, "loss": 0.4567, "step": 15964 }, { "epoch": 1.65, "learning_rate": 3.7762590043807855e-05, "loss": 0.4937, "step": 15965 }, { "epoch": 1.65, "learning_rate": 3.776021242478607e-05, "loss": 0.2063, "step": 15966 }, { "epoch": 1.65, "learning_rate": 3.775783475352436e-05, "loss": 0.9844, "step": 15967 }, { "epoch": 1.65, "learning_rate": 3.775545703003871e-05, "loss": 0.4888, "step": 15968 }, { "epoch": 1.65, "learning_rate": 3.7753079254345155e-05, "loss": 0.4757, "step": 15969 }, { "epoch": 1.65, "learning_rate": 3.775070142645967e-05, "loss": 0.7807, "step": 15970 }, { "epoch": 1.65, "learning_rate": 3.7748323546398276e-05, "loss": 0.6406, "step": 15971 }, { "epoch": 1.65, "learning_rate": 3.774594561417698e-05, "loss": 0.6457, "step": 15972 }, { "epoch": 1.65, "learning_rate": 3.77435676298118e-05, "loss": 0.1333, "step": 15973 }, { "epoch": 1.65, "learning_rate": 3.774118959331873e-05, "loss": 0.4938, "step": 15974 }, { "epoch": 1.65, "learning_rate": 3.773881150471378e-05, "loss": 0.3775, "step": 15975 }, { "epoch": 1.65, "learning_rate": 3.773643336401296e-05, "loss": 0.9862, "step": 15976 }, { "epoch": 1.65, "learning_rate": 3.773405517123228e-05, "loss": 1.236, "step": 15977 }, { "epoch": 1.65, "learning_rate": 3.773167692638777e-05, "loss": 0.4495, "step": 15978 }, { "epoch": 1.65, "learning_rate": 3.77292986294954e-05, "loss": 1.7816, "step": 15979 }, { "epoch": 1.65, "learning_rate": 3.77269202805712e-05, "loss": 1.3913, "step": 15980 }, { "epoch": 1.65, "learning_rate": 3.7724541879631186e-05, "loss": 1.212, "step": 15981 }, { "epoch": 1.65, "learning_rate": 3.7722163426691365e-05, "loss": 0.6977, "step": 15982 }, { "epoch": 1.65, "learning_rate": 3.771978492176775e-05, "loss": 0.4649, "step": 15983 }, { "epoch": 1.65, "learning_rate": 3.771740636487633e-05, "loss": 1.5012, "step": 15984 }, { "epoch": 1.65, "learning_rate": 3.7715027756033156e-05, "loss": 0.837, "step": 15985 }, { "epoch": 1.65, "learning_rate": 3.771264909525421e-05, "loss": 0.193, "step": 15986 }, { "epoch": 1.65, "learning_rate": 3.771027038255552e-05, "loss": 0.7846, "step": 15987 }, { "epoch": 1.65, "learning_rate": 3.7707891617953104e-05, "loss": 0.8549, "step": 15988 }, { "epoch": 1.65, "learning_rate": 3.770551280146295e-05, "loss": 0.1673, "step": 15989 }, { "epoch": 1.65, "learning_rate": 3.770313393310109e-05, "loss": 0.1964, "step": 15990 }, { "epoch": 1.65, "learning_rate": 3.770075501288353e-05, "loss": 0.7932, "step": 15991 }, { "epoch": 1.65, "learning_rate": 3.7698376040826296e-05, "loss": 0.8641, "step": 15992 }, { "epoch": 1.65, "learning_rate": 3.76959970169454e-05, "loss": 0.5589, "step": 15993 }, { "epoch": 1.65, "learning_rate": 3.7693617941256846e-05, "loss": 0.852, "step": 15994 }, { "epoch": 1.66, "learning_rate": 3.769123881377665e-05, "loss": 0.9461, "step": 15995 }, { "epoch": 1.66, "learning_rate": 3.768885963452084e-05, "loss": 0.4195, "step": 15996 }, { "epoch": 1.66, "learning_rate": 3.768648040350544e-05, "loss": 0.4214, "step": 15997 }, { "epoch": 1.66, "learning_rate": 3.768410112074643e-05, "loss": 0.3984, "step": 15998 }, { "epoch": 1.66, "learning_rate": 3.7681721786259864e-05, "loss": 0.8633, "step": 15999 }, { "epoch": 1.66, "learning_rate": 3.767934240006174e-05, "loss": 0.8573, "step": 16000 }, { "epoch": 1.66, "learning_rate": 3.767696296216807e-05, "loss": 1.3786, "step": 16001 }, { "epoch": 1.66, "learning_rate": 3.7674583472594905e-05, "loss": 1.1409, "step": 16002 }, { "epoch": 1.66, "learning_rate": 3.767220393135822e-05, "loss": 0.3183, "step": 16003 }, { "epoch": 1.66, "learning_rate": 3.766982433847405e-05, "loss": 0.7536, "step": 16004 }, { "epoch": 1.66, "learning_rate": 3.766744469395843e-05, "loss": 0.612, "step": 16005 }, { "epoch": 1.66, "learning_rate": 3.7665064997827364e-05, "loss": 0.5433, "step": 16006 }, { "epoch": 1.66, "learning_rate": 3.766268525009687e-05, "loss": 1.0598, "step": 16007 }, { "epoch": 1.66, "learning_rate": 3.7660305450782975e-05, "loss": 0.3386, "step": 16008 }, { "epoch": 1.66, "learning_rate": 3.765792559990169e-05, "loss": 0.5155, "step": 16009 }, { "epoch": 1.66, "learning_rate": 3.765554569746905e-05, "loss": 0.4358, "step": 16010 }, { "epoch": 1.66, "learning_rate": 3.765316574350107e-05, "loss": 0.2387, "step": 16011 }, { "epoch": 1.66, "learning_rate": 3.7650785738013764e-05, "loss": 0.3237, "step": 16012 }, { "epoch": 1.66, "learning_rate": 3.7648405681023154e-05, "loss": 0.9468, "step": 16013 }, { "epoch": 1.66, "learning_rate": 3.7646025572545267e-05, "loss": 0.9034, "step": 16014 }, { "epoch": 1.66, "learning_rate": 3.764364541259613e-05, "loss": 0.7205, "step": 16015 }, { "epoch": 1.66, "learning_rate": 3.764126520119176e-05, "loss": 1.0061, "step": 16016 }, { "epoch": 1.66, "learning_rate": 3.7638884938348174e-05, "loss": 0.3421, "step": 16017 }, { "epoch": 1.66, "learning_rate": 3.7636504624081416e-05, "loss": 1.1642, "step": 16018 }, { "epoch": 1.66, "learning_rate": 3.7634124258407484e-05, "loss": 0.2582, "step": 16019 }, { "epoch": 1.66, "learning_rate": 3.7631743841342416e-05, "loss": 0.4285, "step": 16020 }, { "epoch": 1.66, "learning_rate": 3.762936337290223e-05, "loss": 0.8323, "step": 16021 }, { "epoch": 1.66, "learning_rate": 3.762698285310296e-05, "loss": 1.7502, "step": 16022 }, { "epoch": 1.66, "learning_rate": 3.762460228196062e-05, "loss": 0.6815, "step": 16023 }, { "epoch": 1.66, "learning_rate": 3.762222165949126e-05, "loss": 1.1344, "step": 16024 }, { "epoch": 1.66, "learning_rate": 3.7619840985710867e-05, "loss": 0.8748, "step": 16025 }, { "epoch": 1.66, "learning_rate": 3.76174602606355e-05, "loss": 0.8386, "step": 16026 }, { "epoch": 1.66, "learning_rate": 3.761507948428117e-05, "loss": 0.5285, "step": 16027 }, { "epoch": 1.66, "learning_rate": 3.76126986566639e-05, "loss": 0.5248, "step": 16028 }, { "epoch": 1.66, "learning_rate": 3.761031777779972e-05, "loss": 0.3004, "step": 16029 }, { "epoch": 1.66, "learning_rate": 3.760793684770467e-05, "loss": 0.6783, "step": 16030 }, { "epoch": 1.66, "learning_rate": 3.7605555866394773e-05, "loss": 0.4629, "step": 16031 }, { "epoch": 1.66, "learning_rate": 3.760317483388605e-05, "loss": 1.0881, "step": 16032 }, { "epoch": 1.66, "learning_rate": 3.760079375019454e-05, "loss": 0.8499, "step": 16033 }, { "epoch": 1.66, "learning_rate": 3.759841261533625e-05, "loss": 0.9599, "step": 16034 }, { "epoch": 1.66, "learning_rate": 3.759603142932724e-05, "loss": 0.3028, "step": 16035 }, { "epoch": 1.66, "learning_rate": 3.7593650192183504e-05, "loss": 0.6834, "step": 16036 }, { "epoch": 1.66, "learning_rate": 3.759126890392111e-05, "loss": 0.0141, "step": 16037 }, { "epoch": 1.66, "learning_rate": 3.758888756455606e-05, "loss": 0.4917, "step": 16038 }, { "epoch": 1.66, "learning_rate": 3.7586506174104397e-05, "loss": 0.3419, "step": 16039 }, { "epoch": 1.66, "learning_rate": 3.758412473258215e-05, "loss": 1.0126, "step": 16040 }, { "epoch": 1.66, "learning_rate": 3.758174324000535e-05, "loss": 0.225, "step": 16041 }, { "epoch": 1.66, "learning_rate": 3.757936169639002e-05, "loss": 0.832, "step": 16042 }, { "epoch": 1.66, "learning_rate": 3.757698010175222e-05, "loss": 0.4079, "step": 16043 }, { "epoch": 1.66, "learning_rate": 3.757459845610795e-05, "loss": 0.806, "step": 16044 }, { "epoch": 1.66, "learning_rate": 3.757221675947326e-05, "loss": 0.4062, "step": 16045 }, { "epoch": 1.66, "learning_rate": 3.756983501186417e-05, "loss": 0.8335, "step": 16046 }, { "epoch": 1.66, "learning_rate": 3.756745321329673e-05, "loss": 0.2861, "step": 16047 }, { "epoch": 1.66, "learning_rate": 3.756507136378696e-05, "loss": 1.1322, "step": 16048 }, { "epoch": 1.66, "learning_rate": 3.7562689463350904e-05, "loss": 0.5511, "step": 16049 }, { "epoch": 1.66, "learning_rate": 3.7560307512004595e-05, "loss": 1.2336, "step": 16050 }, { "epoch": 1.66, "learning_rate": 3.7557925509764055e-05, "loss": 0.2898, "step": 16051 }, { "epoch": 1.66, "learning_rate": 3.7555543456645336e-05, "loss": 0.3591, "step": 16052 }, { "epoch": 1.66, "learning_rate": 3.7553161352664456e-05, "loss": 1.3309, "step": 16053 }, { "epoch": 1.66, "learning_rate": 3.755077919783748e-05, "loss": 0.6676, "step": 16054 }, { "epoch": 1.66, "learning_rate": 3.754839699218041e-05, "loss": 0.8906, "step": 16055 }, { "epoch": 1.66, "learning_rate": 3.7546014735709305e-05, "loss": 0.1568, "step": 16056 }, { "epoch": 1.66, "learning_rate": 3.754363242844019e-05, "loss": 0.5395, "step": 16057 }, { "epoch": 1.66, "learning_rate": 3.7541250070389104e-05, "loss": 0.7782, "step": 16058 }, { "epoch": 1.66, "learning_rate": 3.75388676615721e-05, "loss": 0.4455, "step": 16059 }, { "epoch": 1.66, "learning_rate": 3.7536485202005196e-05, "loss": 1.0085, "step": 16060 }, { "epoch": 1.66, "learning_rate": 3.753410269170443e-05, "loss": 1.192, "step": 16061 }, { "epoch": 1.66, "learning_rate": 3.753172013068586e-05, "loss": 0.9851, "step": 16062 }, { "epoch": 1.66, "learning_rate": 3.75293375189655e-05, "loss": 0.3311, "step": 16063 }, { "epoch": 1.66, "learning_rate": 3.7526954856559414e-05, "loss": 1.0335, "step": 16064 }, { "epoch": 1.66, "learning_rate": 3.7524572143483614e-05, "loss": 0.951, "step": 16065 }, { "epoch": 1.66, "learning_rate": 3.7522189379754165e-05, "loss": 0.3076, "step": 16066 }, { "epoch": 1.66, "learning_rate": 3.751980656538709e-05, "loss": 0.4373, "step": 16067 }, { "epoch": 1.66, "learning_rate": 3.751742370039845e-05, "loss": 1.0886, "step": 16068 }, { "epoch": 1.66, "learning_rate": 3.751504078480427e-05, "loss": 1.6298, "step": 16069 }, { "epoch": 1.66, "learning_rate": 3.751265781862058e-05, "loss": 1.2043, "step": 16070 }, { "epoch": 1.66, "learning_rate": 3.7510274801863446e-05, "loss": 1.2404, "step": 16071 }, { "epoch": 1.66, "learning_rate": 3.750789173454889e-05, "loss": 0.4352, "step": 16072 }, { "epoch": 1.66, "learning_rate": 3.750550861669297e-05, "loss": 0.2722, "step": 16073 }, { "epoch": 1.66, "learning_rate": 3.750312544831173e-05, "loss": 0.3121, "step": 16074 }, { "epoch": 1.66, "learning_rate": 3.75007422294212e-05, "loss": 0.2842, "step": 16075 }, { "epoch": 1.66, "learning_rate": 3.749835896003742e-05, "loss": 0.9708, "step": 16076 }, { "epoch": 1.66, "learning_rate": 3.749597564017645e-05, "loss": 0.9156, "step": 16077 }, { "epoch": 1.66, "learning_rate": 3.749359226985433e-05, "loss": 0.3178, "step": 16078 }, { "epoch": 1.66, "learning_rate": 3.749120884908708e-05, "loss": 0.9747, "step": 16079 }, { "epoch": 1.66, "learning_rate": 3.748882537789078e-05, "loss": 0.36, "step": 16080 }, { "epoch": 1.66, "learning_rate": 3.748644185628147e-05, "loss": 0.3073, "step": 16081 }, { "epoch": 1.66, "learning_rate": 3.748405828427517e-05, "loss": 1.1325, "step": 16082 }, { "epoch": 1.66, "learning_rate": 3.748167466188794e-05, "loss": 0.645, "step": 16083 }, { "epoch": 1.66, "learning_rate": 3.7479290989135834e-05, "loss": 0.9657, "step": 16084 }, { "epoch": 1.66, "learning_rate": 3.747690726603489e-05, "loss": 0.3266, "step": 16085 }, { "epoch": 1.66, "learning_rate": 3.747452349260115e-05, "loss": 0.8043, "step": 16086 }, { "epoch": 1.66, "learning_rate": 3.747213966885067e-05, "loss": 0.5044, "step": 16087 }, { "epoch": 1.66, "learning_rate": 3.7469755794799504e-05, "loss": 0.3777, "step": 16088 }, { "epoch": 1.66, "learning_rate": 3.746737187046368e-05, "loss": 0.2085, "step": 16089 }, { "epoch": 1.66, "learning_rate": 3.746498789585925e-05, "loss": 0.3734, "step": 16090 }, { "epoch": 1.67, "learning_rate": 3.7462603871002276e-05, "loss": 0.5955, "step": 16091 }, { "epoch": 1.67, "learning_rate": 3.74602197959088e-05, "loss": 0.1389, "step": 16092 }, { "epoch": 1.67, "learning_rate": 3.7457835670594875e-05, "loss": 0.3887, "step": 16093 }, { "epoch": 1.67, "learning_rate": 3.7455451495076534e-05, "loss": 0.3244, "step": 16094 }, { "epoch": 1.67, "learning_rate": 3.745306726936984e-05, "loss": 0.7522, "step": 16095 }, { "epoch": 1.67, "learning_rate": 3.7450682993490834e-05, "loss": 0.4912, "step": 16096 }, { "epoch": 1.67, "learning_rate": 3.744829866745559e-05, "loss": 0.2348, "step": 16097 }, { "epoch": 1.67, "learning_rate": 3.7445914291280134e-05, "loss": 0.3636, "step": 16098 }, { "epoch": 1.67, "learning_rate": 3.744352986498052e-05, "loss": 0.4463, "step": 16099 }, { "epoch": 1.67, "learning_rate": 3.744114538857281e-05, "loss": 0.2038, "step": 16100 }, { "epoch": 1.67, "learning_rate": 3.7438760862073056e-05, "loss": 0.6377, "step": 16101 }, { "epoch": 1.67, "learning_rate": 3.74363762854973e-05, "loss": 0.2494, "step": 16102 }, { "epoch": 1.67, "learning_rate": 3.743399165886159e-05, "loss": 0.275, "step": 16103 }, { "epoch": 1.67, "learning_rate": 3.7431606982182e-05, "loss": 0.3377, "step": 16104 }, { "epoch": 1.67, "learning_rate": 3.742922225547456e-05, "loss": 0.6194, "step": 16105 }, { "epoch": 1.67, "learning_rate": 3.742683747875535e-05, "loss": 0.8633, "step": 16106 }, { "epoch": 1.67, "learning_rate": 3.74244526520404e-05, "loss": 1.3889, "step": 16107 }, { "epoch": 1.67, "learning_rate": 3.742206777534576e-05, "loss": 0.5743, "step": 16108 }, { "epoch": 1.67, "learning_rate": 3.741968284868751e-05, "loss": 1.2777, "step": 16109 }, { "epoch": 1.67, "learning_rate": 3.741729787208169e-05, "loss": 0.7832, "step": 16110 }, { "epoch": 1.67, "learning_rate": 3.741491284554435e-05, "loss": 0.8145, "step": 16111 }, { "epoch": 1.67, "learning_rate": 3.741252776909156e-05, "loss": 0.6978, "step": 16112 }, { "epoch": 1.67, "learning_rate": 3.7410142642739366e-05, "loss": 0.7388, "step": 16113 }, { "epoch": 1.67, "learning_rate": 3.7407757466503816e-05, "loss": 0.4558, "step": 16114 }, { "epoch": 1.67, "learning_rate": 3.7405372240400984e-05, "loss": 0.502, "step": 16115 }, { "epoch": 1.67, "learning_rate": 3.7402986964446915e-05, "loss": 0.7525, "step": 16116 }, { "epoch": 1.67, "learning_rate": 3.740060163865767e-05, "loss": 0.6653, "step": 16117 }, { "epoch": 1.67, "learning_rate": 3.739821626304931e-05, "loss": 0.499, "step": 16118 }, { "epoch": 1.67, "learning_rate": 3.739583083763789e-05, "loss": 0.6428, "step": 16119 }, { "epoch": 1.67, "learning_rate": 3.739344536243946e-05, "loss": 0.983, "step": 16120 }, { "epoch": 1.67, "learning_rate": 3.739105983747009e-05, "loss": 0.6899, "step": 16121 }, { "epoch": 1.67, "learning_rate": 3.738867426274584e-05, "loss": 1.1386, "step": 16122 }, { "epoch": 1.67, "learning_rate": 3.7386288638282753e-05, "loss": 0.4533, "step": 16123 }, { "epoch": 1.67, "learning_rate": 3.7383902964096904e-05, "loss": 0.2272, "step": 16124 }, { "epoch": 1.67, "learning_rate": 3.738151724020435e-05, "loss": 1.1683, "step": 16125 }, { "epoch": 1.67, "learning_rate": 3.737913146662115e-05, "loss": 0.2633, "step": 16126 }, { "epoch": 1.67, "learning_rate": 3.737674564336336e-05, "loss": 0.8939, "step": 16127 }, { "epoch": 1.67, "learning_rate": 3.737435977044705e-05, "loss": 0.6867, "step": 16128 }, { "epoch": 1.67, "learning_rate": 3.737197384788827e-05, "loss": 0.4136, "step": 16129 }, { "epoch": 1.67, "learning_rate": 3.736958787570309e-05, "loss": 0.9806, "step": 16130 }, { "epoch": 1.67, "learning_rate": 3.736720185390756e-05, "loss": 0.4289, "step": 16131 }, { "epoch": 1.67, "learning_rate": 3.736481578251776e-05, "loss": 1.0212, "step": 16132 }, { "epoch": 1.67, "learning_rate": 3.736242966154974e-05, "loss": 1.0823, "step": 16133 }, { "epoch": 1.67, "learning_rate": 3.736004349101957e-05, "loss": 0.4863, "step": 16134 }, { "epoch": 1.67, "learning_rate": 3.735765727094331e-05, "loss": 0.3466, "step": 16135 }, { "epoch": 1.67, "learning_rate": 3.7355271001337016e-05, "loss": 0.1689, "step": 16136 }, { "epoch": 1.67, "learning_rate": 3.7352884682216764e-05, "loss": 0.9128, "step": 16137 }, { "epoch": 1.67, "learning_rate": 3.735049831359861e-05, "loss": 0.1632, "step": 16138 }, { "epoch": 1.67, "learning_rate": 3.734811189549862e-05, "loss": 0.7792, "step": 16139 }, { "epoch": 1.67, "learning_rate": 3.734572542793286e-05, "loss": 0.3878, "step": 16140 }, { "epoch": 1.67, "learning_rate": 3.73433389109174e-05, "loss": 1.2225, "step": 16141 }, { "epoch": 1.67, "learning_rate": 3.73409523444683e-05, "loss": 0.187, "step": 16142 }, { "epoch": 1.67, "learning_rate": 3.733856572860162e-05, "loss": 0.1916, "step": 16143 }, { "epoch": 1.67, "learning_rate": 3.733617906333344e-05, "loss": 0.2895, "step": 16144 }, { "epoch": 1.67, "learning_rate": 3.733379234867982e-05, "loss": 0.5379, "step": 16145 }, { "epoch": 1.67, "learning_rate": 3.733140558465682e-05, "loss": 0.6918, "step": 16146 }, { "epoch": 1.67, "learning_rate": 3.732901877128051e-05, "loss": 0.3347, "step": 16147 }, { "epoch": 1.67, "learning_rate": 3.7326631908566964e-05, "loss": 0.8975, "step": 16148 }, { "epoch": 1.67, "learning_rate": 3.7324244996532255e-05, "loss": 0.7698, "step": 16149 }, { "epoch": 1.67, "learning_rate": 3.7321858035192434e-05, "loss": 0.9181, "step": 16150 }, { "epoch": 1.67, "learning_rate": 3.731947102456358e-05, "loss": 1.1844, "step": 16151 }, { "epoch": 1.67, "learning_rate": 3.731708396466176e-05, "loss": 0.7018, "step": 16152 }, { "epoch": 1.67, "learning_rate": 3.731469685550304e-05, "loss": 0.1044, "step": 16153 }, { "epoch": 1.67, "learning_rate": 3.731230969710349e-05, "loss": 0.7672, "step": 16154 }, { "epoch": 1.67, "learning_rate": 3.7309922489479185e-05, "loss": 1.1637, "step": 16155 }, { "epoch": 1.67, "learning_rate": 3.73075352326462e-05, "loss": 1.0448, "step": 16156 }, { "epoch": 1.67, "learning_rate": 3.730514792662059e-05, "loss": 0.3969, "step": 16157 }, { "epoch": 1.67, "learning_rate": 3.730276057141843e-05, "loss": 0.341, "step": 16158 }, { "epoch": 1.67, "learning_rate": 3.73003731670558e-05, "loss": 0.9918, "step": 16159 }, { "epoch": 1.67, "learning_rate": 3.729798571354876e-05, "loss": 0.729, "step": 16160 }, { "epoch": 1.67, "learning_rate": 3.7295598210913396e-05, "loss": 0.9943, "step": 16161 }, { "epoch": 1.67, "learning_rate": 3.729321065916576e-05, "loss": 0.3674, "step": 16162 }, { "epoch": 1.67, "learning_rate": 3.7290823058321954e-05, "loss": 0.9319, "step": 16163 }, { "epoch": 1.67, "learning_rate": 3.728843540839803e-05, "loss": 0.277, "step": 16164 }, { "epoch": 1.67, "learning_rate": 3.728604770941005e-05, "loss": 0.606, "step": 16165 }, { "epoch": 1.67, "learning_rate": 3.7283659961374106e-05, "loss": 0.483, "step": 16166 }, { "epoch": 1.67, "learning_rate": 3.728127216430627e-05, "loss": 0.3968, "step": 16167 }, { "epoch": 1.67, "learning_rate": 3.7278884318222624e-05, "loss": 1.6977, "step": 16168 }, { "epoch": 1.67, "learning_rate": 3.727649642313922e-05, "loss": 0.7936, "step": 16169 }, { "epoch": 1.67, "learning_rate": 3.727410847907214e-05, "loss": 0.3806, "step": 16170 }, { "epoch": 1.67, "learning_rate": 3.727172048603748e-05, "loss": 0.8087, "step": 16171 }, { "epoch": 1.67, "learning_rate": 3.726933244405128e-05, "loss": 0.5044, "step": 16172 }, { "epoch": 1.67, "learning_rate": 3.726694435312965e-05, "loss": 1.0326, "step": 16173 }, { "epoch": 1.67, "learning_rate": 3.726455621328864e-05, "loss": 0.9668, "step": 16174 }, { "epoch": 1.67, "learning_rate": 3.7262168024544355e-05, "loss": 0.2415, "step": 16175 }, { "epoch": 1.67, "learning_rate": 3.7259779786912833e-05, "loss": 0.666, "step": 16176 }, { "epoch": 1.67, "learning_rate": 3.725739150041018e-05, "loss": 0.7224, "step": 16177 }, { "epoch": 1.67, "learning_rate": 3.725500316505248e-05, "loss": 0.4256, "step": 16178 }, { "epoch": 1.67, "learning_rate": 3.725261478085577e-05, "loss": 0.6831, "step": 16179 }, { "epoch": 1.67, "learning_rate": 3.725022634783617e-05, "loss": 0.7515, "step": 16180 }, { "epoch": 1.67, "learning_rate": 3.724783786600974e-05, "loss": 0.5038, "step": 16181 }, { "epoch": 1.67, "learning_rate": 3.724544933539256e-05, "loss": 0.2881, "step": 16182 }, { "epoch": 1.67, "learning_rate": 3.724306075600071e-05, "loss": 0.3666, "step": 16183 }, { "epoch": 1.67, "learning_rate": 3.724067212785027e-05, "loss": 0.9159, "step": 16184 }, { "epoch": 1.67, "learning_rate": 3.7238283450957325e-05, "loss": 0.715, "step": 16185 }, { "epoch": 1.67, "learning_rate": 3.7235894725337946e-05, "loss": 0.9217, "step": 16186 }, { "epoch": 1.68, "learning_rate": 3.7233505951008216e-05, "loss": 0.8727, "step": 16187 }, { "epoch": 1.68, "learning_rate": 3.7231117127984215e-05, "loss": 0.286, "step": 16188 }, { "epoch": 1.68, "learning_rate": 3.722872825628203e-05, "loss": 1.0852, "step": 16189 }, { "epoch": 1.68, "learning_rate": 3.722633933591773e-05, "loss": 0.5373, "step": 16190 }, { "epoch": 1.68, "learning_rate": 3.722395036690741e-05, "loss": 0.8128, "step": 16191 }, { "epoch": 1.68, "learning_rate": 3.722156134926715e-05, "loss": 0.7156, "step": 16192 }, { "epoch": 1.68, "learning_rate": 3.721917228301302e-05, "loss": 1.0459, "step": 16193 }, { "epoch": 1.68, "learning_rate": 3.721678316816111e-05, "loss": 0.5042, "step": 16194 }, { "epoch": 1.68, "learning_rate": 3.721439400472752e-05, "loss": 0.9759, "step": 16195 }, { "epoch": 1.68, "learning_rate": 3.72120047927283e-05, "loss": 0.1351, "step": 16196 }, { "epoch": 1.68, "learning_rate": 3.720961553217956e-05, "loss": 0.8398, "step": 16197 }, { "epoch": 1.68, "learning_rate": 3.7207226223097375e-05, "loss": 1.039, "step": 16198 }, { "epoch": 1.68, "learning_rate": 3.7204836865497823e-05, "loss": 0.7593, "step": 16199 }, { "epoch": 1.68, "learning_rate": 3.720244745939701e-05, "loss": 1.2831, "step": 16200 }, { "epoch": 1.68, "learning_rate": 3.7200058004810986e-05, "loss": 1.0312, "step": 16201 }, { "epoch": 1.68, "learning_rate": 3.7197668501755874e-05, "loss": 0.6572, "step": 16202 }, { "epoch": 1.68, "learning_rate": 3.7195278950247726e-05, "loss": 0.467, "step": 16203 }, { "epoch": 1.68, "learning_rate": 3.7192889350302657e-05, "loss": 0.9144, "step": 16204 }, { "epoch": 1.68, "learning_rate": 3.719049970193673e-05, "loss": 0.7383, "step": 16205 }, { "epoch": 1.68, "learning_rate": 3.718811000516605e-05, "loss": 0.6735, "step": 16206 }, { "epoch": 1.68, "learning_rate": 3.718572026000669e-05, "loss": 1.19, "step": 16207 }, { "epoch": 1.68, "learning_rate": 3.718333046647475e-05, "loss": 0.7844, "step": 16208 }, { "epoch": 1.68, "learning_rate": 3.71809406245863e-05, "loss": 0.6238, "step": 16209 }, { "epoch": 1.68, "learning_rate": 3.717855073435743e-05, "loss": 0.4286, "step": 16210 }, { "epoch": 1.68, "learning_rate": 3.717616079580426e-05, "loss": 0.8918, "step": 16211 }, { "epoch": 1.68, "learning_rate": 3.7173770808942835e-05, "loss": 0.318, "step": 16212 }, { "epoch": 1.68, "learning_rate": 3.717138077378928e-05, "loss": 0.2089, "step": 16213 }, { "epoch": 1.68, "learning_rate": 3.716899069035966e-05, "loss": 0.8995, "step": 16214 }, { "epoch": 1.68, "learning_rate": 3.7166600558670065e-05, "loss": 0.8504, "step": 16215 }, { "epoch": 1.68, "learning_rate": 3.7164210378736606e-05, "loss": 1.0175, "step": 16216 }, { "epoch": 1.68, "learning_rate": 3.716182015057535e-05, "loss": 0.4724, "step": 16217 }, { "epoch": 1.68, "learning_rate": 3.7159429874202384e-05, "loss": 0.7052, "step": 16218 }, { "epoch": 1.68, "learning_rate": 3.7157039549633835e-05, "loss": 0.8987, "step": 16219 }, { "epoch": 1.68, "learning_rate": 3.7154649176885754e-05, "loss": 0.7811, "step": 16220 }, { "epoch": 1.68, "learning_rate": 3.715225875597426e-05, "loss": 0.5718, "step": 16221 }, { "epoch": 1.68, "learning_rate": 3.714986828691543e-05, "loss": 0.8886, "step": 16222 }, { "epoch": 1.68, "learning_rate": 3.714747776972535e-05, "loss": 0.7178, "step": 16223 }, { "epoch": 1.68, "learning_rate": 3.714508720442013e-05, "loss": 0.2894, "step": 16224 }, { "epoch": 1.68, "learning_rate": 3.714269659101586e-05, "loss": 0.3848, "step": 16225 }, { "epoch": 1.68, "learning_rate": 3.7140305929528625e-05, "loss": 0.656, "step": 16226 }, { "epoch": 1.68, "learning_rate": 3.7137915219974514e-05, "loss": 0.7931, "step": 16227 }, { "epoch": 1.68, "learning_rate": 3.713552446236963e-05, "loss": 0.4621, "step": 16228 }, { "epoch": 1.68, "learning_rate": 3.7133133656730064e-05, "loss": 0.1037, "step": 16229 }, { "epoch": 1.68, "learning_rate": 3.713074280307192e-05, "loss": 0.2966, "step": 16230 }, { "epoch": 1.68, "learning_rate": 3.712835190141128e-05, "loss": 0.5989, "step": 16231 }, { "epoch": 1.68, "learning_rate": 3.712596095176424e-05, "loss": 0.9563, "step": 16232 }, { "epoch": 1.68, "learning_rate": 3.7123569954146904e-05, "loss": 0.8328, "step": 16233 }, { "epoch": 1.68, "learning_rate": 3.712117890857535e-05, "loss": 1.0285, "step": 16234 }, { "epoch": 1.68, "learning_rate": 3.711878781506569e-05, "loss": 0.3921, "step": 16235 }, { "epoch": 1.68, "learning_rate": 3.7116396673634026e-05, "loss": 1.6413, "step": 16236 }, { "epoch": 1.68, "learning_rate": 3.711400548429643e-05, "loss": 0.8923, "step": 16237 }, { "epoch": 1.68, "learning_rate": 3.711161424706903e-05, "loss": 0.524, "step": 16238 }, { "epoch": 1.68, "learning_rate": 3.710922296196789e-05, "loss": 1.0731, "step": 16239 }, { "epoch": 1.68, "learning_rate": 3.710683162900914e-05, "loss": 0.9182, "step": 16240 }, { "epoch": 1.68, "learning_rate": 3.7104440248208856e-05, "loss": 0.5178, "step": 16241 }, { "epoch": 1.68, "learning_rate": 3.710204881958314e-05, "loss": 0.1836, "step": 16242 }, { "epoch": 1.68, "learning_rate": 3.709965734314809e-05, "loss": 0.3795, "step": 16243 }, { "epoch": 1.68, "learning_rate": 3.709726581891982e-05, "loss": 0.7335, "step": 16244 }, { "epoch": 1.68, "learning_rate": 3.709487424691441e-05, "loss": 1.535, "step": 16245 }, { "epoch": 1.68, "learning_rate": 3.7092482627147966e-05, "loss": 1.1388, "step": 16246 }, { "epoch": 1.68, "learning_rate": 3.709009095963659e-05, "loss": 0.9746, "step": 16247 }, { "epoch": 1.68, "learning_rate": 3.708769924439637e-05, "loss": 1.0316, "step": 16248 }, { "epoch": 1.68, "learning_rate": 3.708530748144343e-05, "loss": 0.8694, "step": 16249 }, { "epoch": 1.68, "learning_rate": 3.708291567079386e-05, "loss": 0.7443, "step": 16250 }, { "epoch": 1.68, "learning_rate": 3.708052381246376e-05, "loss": 0.62, "step": 16251 }, { "epoch": 1.68, "learning_rate": 3.707813190646921e-05, "loss": 1.103, "step": 16252 }, { "epoch": 1.68, "learning_rate": 3.7075739952826354e-05, "loss": 0.3325, "step": 16253 }, { "epoch": 1.68, "learning_rate": 3.707334795155127e-05, "loss": 0.7647, "step": 16254 }, { "epoch": 1.68, "learning_rate": 3.707095590266005e-05, "loss": 0.4381, "step": 16255 }, { "epoch": 1.68, "learning_rate": 3.706856380616882e-05, "loss": 0.3373, "step": 16256 }, { "epoch": 1.68, "learning_rate": 3.7066171662093666e-05, "loss": 0.276, "step": 16257 }, { "epoch": 1.68, "learning_rate": 3.70637794704507e-05, "loss": 0.5789, "step": 16258 }, { "epoch": 1.68, "learning_rate": 3.706138723125602e-05, "loss": 0.3318, "step": 16259 }, { "epoch": 1.68, "learning_rate": 3.7058994944525744e-05, "loss": 0.7834, "step": 16260 }, { "epoch": 1.68, "learning_rate": 3.7056602610275955e-05, "loss": 0.7611, "step": 16261 }, { "epoch": 1.68, "learning_rate": 3.705421022852277e-05, "loss": 0.4042, "step": 16262 }, { "epoch": 1.68, "learning_rate": 3.7051817799282295e-05, "loss": 1.1383, "step": 16263 }, { "epoch": 1.68, "learning_rate": 3.704942532257064e-05, "loss": 0.1892, "step": 16264 }, { "epoch": 1.68, "learning_rate": 3.704703279840389e-05, "loss": 1.1353, "step": 16265 }, { "epoch": 1.68, "learning_rate": 3.7044640226798164e-05, "loss": 0.3713, "step": 16266 }, { "epoch": 1.68, "learning_rate": 3.704224760776957e-05, "loss": 0.4529, "step": 16267 }, { "epoch": 1.68, "learning_rate": 3.703985494133423e-05, "loss": 1.5336, "step": 16268 }, { "epoch": 1.68, "learning_rate": 3.703746222750822e-05, "loss": 0.2995, "step": 16269 }, { "epoch": 1.68, "learning_rate": 3.703506946630766e-05, "loss": 0.4416, "step": 16270 }, { "epoch": 1.68, "learning_rate": 3.703267665774866e-05, "loss": 0.916, "step": 16271 }, { "epoch": 1.68, "learning_rate": 3.703028380184732e-05, "loss": 0.745, "step": 16272 }, { "epoch": 1.68, "learning_rate": 3.7027890898619765e-05, "loss": 0.4502, "step": 16273 }, { "epoch": 1.68, "learning_rate": 3.7025497948082086e-05, "loss": 1.966, "step": 16274 }, { "epoch": 1.68, "learning_rate": 3.7023104950250396e-05, "loss": 0.5177, "step": 16275 }, { "epoch": 1.68, "learning_rate": 3.7020711905140816e-05, "loss": 1.0048, "step": 16276 }, { "epoch": 1.68, "learning_rate": 3.701831881276944e-05, "loss": 0.7061, "step": 16277 }, { "epoch": 1.68, "learning_rate": 3.701592567315239e-05, "loss": 0.2063, "step": 16278 }, { "epoch": 1.68, "learning_rate": 3.701353248630576e-05, "loss": 1.0676, "step": 16279 }, { "epoch": 1.68, "learning_rate": 3.7011139252245675e-05, "loss": 0.6496, "step": 16280 }, { "epoch": 1.68, "learning_rate": 3.7008745970988243e-05, "loss": 0.3999, "step": 16281 }, { "epoch": 1.69, "learning_rate": 3.700635264254958e-05, "loss": 0.1698, "step": 16282 }, { "epoch": 1.69, "learning_rate": 3.700395926694578e-05, "loss": 0.5352, "step": 16283 }, { "epoch": 1.69, "learning_rate": 3.700156584419297e-05, "loss": 0.1429, "step": 16284 }, { "epoch": 1.69, "learning_rate": 3.6999172374307254e-05, "loss": 0.4453, "step": 16285 }, { "epoch": 1.69, "learning_rate": 3.699677885730474e-05, "loss": 0.18, "step": 16286 }, { "epoch": 1.69, "learning_rate": 3.699438529320156e-05, "loss": 0.8784, "step": 16287 }, { "epoch": 1.69, "learning_rate": 3.699199168201381e-05, "loss": 0.2651, "step": 16288 }, { "epoch": 1.69, "learning_rate": 3.698959802375761e-05, "loss": 0.7463, "step": 16289 }, { "epoch": 1.69, "learning_rate": 3.6987204318449065e-05, "loss": 0.2646, "step": 16290 }, { "epoch": 1.69, "learning_rate": 3.69848105661043e-05, "loss": 0.4361, "step": 16291 }, { "epoch": 1.69, "learning_rate": 3.698241676673943e-05, "loss": 0.5564, "step": 16292 }, { "epoch": 1.69, "learning_rate": 3.698002292037055e-05, "loss": 1.2105, "step": 16293 }, { "epoch": 1.69, "learning_rate": 3.697762902701381e-05, "loss": 1.1754, "step": 16294 }, { "epoch": 1.69, "learning_rate": 3.6975235086685284e-05, "loss": 0.7252, "step": 16295 }, { "epoch": 1.69, "learning_rate": 3.697284109940111e-05, "loss": 0.9553, "step": 16296 }, { "epoch": 1.69, "learning_rate": 3.697044706517742e-05, "loss": 0.9019, "step": 16297 }, { "epoch": 1.69, "learning_rate": 3.696805298403029e-05, "loss": 1.262, "step": 16298 }, { "epoch": 1.69, "learning_rate": 3.696565885597586e-05, "loss": 0.0491, "step": 16299 }, { "epoch": 1.69, "learning_rate": 3.696326468103025e-05, "loss": 0.4998, "step": 16300 }, { "epoch": 1.69, "learning_rate": 3.696087045920957e-05, "loss": 0.8782, "step": 16301 }, { "epoch": 1.69, "learning_rate": 3.695847619052995e-05, "loss": 0.519, "step": 16302 }, { "epoch": 1.69, "learning_rate": 3.695608187500748e-05, "loss": 0.4026, "step": 16303 }, { "epoch": 1.69, "learning_rate": 3.6953687512658305e-05, "loss": 0.6695, "step": 16304 }, { "epoch": 1.69, "learning_rate": 3.695129310349853e-05, "loss": 0.2902, "step": 16305 }, { "epoch": 1.69, "learning_rate": 3.6948898647544277e-05, "loss": 0.1491, "step": 16306 }, { "epoch": 1.69, "learning_rate": 3.694650414481167e-05, "loss": 0.4751, "step": 16307 }, { "epoch": 1.69, "learning_rate": 3.69441095953168e-05, "loss": 0.5021, "step": 16308 }, { "epoch": 1.69, "learning_rate": 3.6941714999075826e-05, "loss": 0.5727, "step": 16309 }, { "epoch": 1.69, "learning_rate": 3.6939320356104843e-05, "loss": 0.2904, "step": 16310 }, { "epoch": 1.69, "learning_rate": 3.6936925666419994e-05, "loss": 0.5337, "step": 16311 }, { "epoch": 1.69, "learning_rate": 3.693453093003737e-05, "loss": 0.4198, "step": 16312 }, { "epoch": 1.69, "learning_rate": 3.693213614697312e-05, "loss": 0.7267, "step": 16313 }, { "epoch": 1.69, "learning_rate": 3.6929741317243334e-05, "loss": 0.9523, "step": 16314 }, { "epoch": 1.69, "learning_rate": 3.692734644086416e-05, "loss": 0.3926, "step": 16315 }, { "epoch": 1.69, "learning_rate": 3.6924951517851714e-05, "loss": 1.2941, "step": 16316 }, { "epoch": 1.69, "learning_rate": 3.69225565482221e-05, "loss": 0.5156, "step": 16317 }, { "epoch": 1.69, "learning_rate": 3.6920161531991476e-05, "loss": 1.2672, "step": 16318 }, { "epoch": 1.69, "learning_rate": 3.691776646917593e-05, "loss": 0.436, "step": 16319 }, { "epoch": 1.69, "learning_rate": 3.6915371359791606e-05, "loss": 0.9641, "step": 16320 }, { "epoch": 1.69, "learning_rate": 3.691297620385462e-05, "loss": 0.6751, "step": 16321 }, { "epoch": 1.69, "learning_rate": 3.6910581001381084e-05, "loss": 0.6586, "step": 16322 }, { "epoch": 1.69, "learning_rate": 3.6908185752387145e-05, "loss": 0.2225, "step": 16323 }, { "epoch": 1.69, "learning_rate": 3.690579045688891e-05, "loss": 0.8904, "step": 16324 }, { "epoch": 1.69, "learning_rate": 3.6903395114902516e-05, "loss": 0.9766, "step": 16325 }, { "epoch": 1.69, "learning_rate": 3.690099972644408e-05, "loss": 0.9398, "step": 16326 }, { "epoch": 1.69, "learning_rate": 3.689860429152972e-05, "loss": 0.9132, "step": 16327 }, { "epoch": 1.69, "learning_rate": 3.689620881017558e-05, "loss": 0.7375, "step": 16328 }, { "epoch": 1.69, "learning_rate": 3.689381328239777e-05, "loss": 0.8326, "step": 16329 }, { "epoch": 1.69, "learning_rate": 3.689141770821243e-05, "loss": 0.7026, "step": 16330 }, { "epoch": 1.69, "learning_rate": 3.688902208763567e-05, "loss": 0.9731, "step": 16331 }, { "epoch": 1.69, "learning_rate": 3.6886626420683634e-05, "loss": 0.5161, "step": 16332 }, { "epoch": 1.69, "learning_rate": 3.688423070737243e-05, "loss": 0.5179, "step": 16333 }, { "epoch": 1.69, "learning_rate": 3.688183494771819e-05, "loss": 0.5041, "step": 16334 }, { "epoch": 1.69, "learning_rate": 3.6879439141737064e-05, "loss": 1.2377, "step": 16335 }, { "epoch": 1.69, "learning_rate": 3.6877043289445155e-05, "loss": 1.2494, "step": 16336 }, { "epoch": 1.69, "learning_rate": 3.68746473908586e-05, "loss": 0.5216, "step": 16337 }, { "epoch": 1.69, "learning_rate": 3.687225144599353e-05, "loss": 0.6656, "step": 16338 }, { "epoch": 1.69, "learning_rate": 3.6869855454866065e-05, "loss": 0.7215, "step": 16339 }, { "epoch": 1.69, "learning_rate": 3.6867459417492355e-05, "loss": 0.3615, "step": 16340 }, { "epoch": 1.69, "learning_rate": 3.68650633338885e-05, "loss": 0.3119, "step": 16341 }, { "epoch": 1.69, "learning_rate": 3.686266720407065e-05, "loss": 0.8281, "step": 16342 }, { "epoch": 1.69, "learning_rate": 3.686027102805493e-05, "loss": 0.7818, "step": 16343 }, { "epoch": 1.69, "learning_rate": 3.685787480585747e-05, "loss": 1.3558, "step": 16344 }, { "epoch": 1.69, "learning_rate": 3.685547853749441e-05, "loss": 0.8455, "step": 16345 }, { "epoch": 1.69, "learning_rate": 3.685308222298186e-05, "loss": 0.6772, "step": 16346 }, { "epoch": 1.69, "learning_rate": 3.685068586233597e-05, "loss": 0.3399, "step": 16347 }, { "epoch": 1.69, "learning_rate": 3.684828945557286e-05, "loss": 0.515, "step": 16348 }, { "epoch": 1.69, "learning_rate": 3.684589300270868e-05, "loss": 0.9039, "step": 16349 }, { "epoch": 1.69, "learning_rate": 3.684349650375954e-05, "loss": 1.025, "step": 16350 }, { "epoch": 1.69, "learning_rate": 3.684109995874158e-05, "loss": 0.6714, "step": 16351 }, { "epoch": 1.69, "learning_rate": 3.6838703367670946e-05, "loss": 0.4348, "step": 16352 }, { "epoch": 1.69, "learning_rate": 3.683630673056376e-05, "loss": 0.4545, "step": 16353 }, { "epoch": 1.69, "learning_rate": 3.683391004743615e-05, "loss": 0.0333, "step": 16354 }, { "epoch": 1.69, "learning_rate": 3.683151331830426e-05, "loss": 1.1376, "step": 16355 }, { "epoch": 1.69, "learning_rate": 3.682911654318421e-05, "loss": 0.5201, "step": 16356 }, { "epoch": 1.69, "learning_rate": 3.682671972209217e-05, "loss": 0.8088, "step": 16357 }, { "epoch": 1.69, "learning_rate": 3.6824322855044225e-05, "loss": 0.7451, "step": 16358 }, { "epoch": 1.69, "learning_rate": 3.682192594205655e-05, "loss": 0.6351, "step": 16359 }, { "epoch": 1.69, "learning_rate": 3.6819528983145266e-05, "loss": 0.6717, "step": 16360 }, { "epoch": 1.69, "learning_rate": 3.68171319783265e-05, "loss": 0.7859, "step": 16361 }, { "epoch": 1.69, "learning_rate": 3.6814734927616394e-05, "loss": 0.193, "step": 16362 }, { "epoch": 1.69, "learning_rate": 3.68123378310311e-05, "loss": 0.8785, "step": 16363 }, { "epoch": 1.69, "learning_rate": 3.680994068858674e-05, "loss": 0.5987, "step": 16364 }, { "epoch": 1.69, "learning_rate": 3.680754350029944e-05, "loss": 0.4244, "step": 16365 }, { "epoch": 1.69, "learning_rate": 3.680514626618536e-05, "loss": 0.821, "step": 16366 }, { "epoch": 1.69, "learning_rate": 3.6802748986260624e-05, "loss": 0.275, "step": 16367 }, { "epoch": 1.69, "learning_rate": 3.680035166054138e-05, "loss": 0.8759, "step": 16368 }, { "epoch": 1.69, "learning_rate": 3.679795428904375e-05, "loss": 0.8041, "step": 16369 }, { "epoch": 1.69, "learning_rate": 3.679555687178389e-05, "loss": 1.117, "step": 16370 }, { "epoch": 1.69, "learning_rate": 3.679315940877793e-05, "loss": 0.3882, "step": 16371 }, { "epoch": 1.69, "learning_rate": 3.6790761900042004e-05, "loss": 0.189, "step": 16372 }, { "epoch": 1.69, "learning_rate": 3.678836434559227e-05, "loss": 0.6343, "step": 16373 }, { "epoch": 1.69, "learning_rate": 3.678596674544484e-05, "loss": 1.0394, "step": 16374 }, { "epoch": 1.69, "learning_rate": 3.678356909961588e-05, "loss": 1.0656, "step": 16375 }, { "epoch": 1.69, "learning_rate": 3.678117140812152e-05, "loss": 0.2343, "step": 16376 }, { "epoch": 1.69, "learning_rate": 3.6778773670977895e-05, "loss": 0.7511, "step": 16377 }, { "epoch": 1.7, "learning_rate": 3.677637588820117e-05, "loss": 0.9176, "step": 16378 }, { "epoch": 1.7, "learning_rate": 3.677397805980744e-05, "loss": 0.6723, "step": 16379 }, { "epoch": 1.7, "learning_rate": 3.677158018581289e-05, "loss": 0.5374, "step": 16380 }, { "epoch": 1.7, "learning_rate": 3.676918226623364e-05, "loss": 0.8338, "step": 16381 }, { "epoch": 1.7, "learning_rate": 3.6766784301085855e-05, "loss": 0.8032, "step": 16382 }, { "epoch": 1.7, "learning_rate": 3.676438629038565e-05, "loss": 0.9523, "step": 16383 }, { "epoch": 1.7, "learning_rate": 3.676198823414918e-05, "loss": 0.4476, "step": 16384 }, { "epoch": 1.7, "learning_rate": 3.6759590132392583e-05, "loss": 1.1973, "step": 16385 }, { "epoch": 1.7, "learning_rate": 3.6757191985132e-05, "loss": 0.626, "step": 16386 }, { "epoch": 1.7, "learning_rate": 3.67547937923836e-05, "loss": 0.9455, "step": 16387 }, { "epoch": 1.7, "learning_rate": 3.67523955541635e-05, "loss": 0.3427, "step": 16388 }, { "epoch": 1.7, "learning_rate": 3.674999727048786e-05, "loss": 0.4969, "step": 16389 }, { "epoch": 1.7, "learning_rate": 3.674759894137281e-05, "loss": 0.7294, "step": 16390 }, { "epoch": 1.7, "learning_rate": 3.67452005668345e-05, "loss": 0.1554, "step": 16391 }, { "epoch": 1.7, "learning_rate": 3.674280214688908e-05, "loss": 0.3787, "step": 16392 }, { "epoch": 1.7, "learning_rate": 3.6740403681552694e-05, "loss": 1.1022, "step": 16393 }, { "epoch": 1.7, "learning_rate": 3.673800517084149e-05, "loss": 0.4344, "step": 16394 }, { "epoch": 1.7, "learning_rate": 3.673560661477162e-05, "loss": 1.4228, "step": 16395 }, { "epoch": 1.7, "learning_rate": 3.673320801335921e-05, "loss": 0.5242, "step": 16396 }, { "epoch": 1.7, "learning_rate": 3.6730809366620426e-05, "loss": 0.3858, "step": 16397 }, { "epoch": 1.7, "learning_rate": 3.67284106745714e-05, "loss": 0.545, "step": 16398 }, { "epoch": 1.7, "learning_rate": 3.672601193722829e-05, "loss": 0.1608, "step": 16399 }, { "epoch": 1.7, "learning_rate": 3.672361315460725e-05, "loss": 0.3529, "step": 16400 }, { "epoch": 1.7, "learning_rate": 3.672121432672441e-05, "loss": 0.6718, "step": 16401 }, { "epoch": 1.7, "learning_rate": 3.6718815453595934e-05, "loss": 0.2132, "step": 16402 }, { "epoch": 1.7, "learning_rate": 3.671641653523796e-05, "loss": 1.0602, "step": 16403 }, { "epoch": 1.7, "learning_rate": 3.6714017571666644e-05, "loss": 0.3339, "step": 16404 }, { "epoch": 1.7, "learning_rate": 3.671161856289813e-05, "loss": 0.8574, "step": 16405 }, { "epoch": 1.7, "learning_rate": 3.670921950894857e-05, "loss": 0.5649, "step": 16406 }, { "epoch": 1.7, "learning_rate": 3.670682040983412e-05, "loss": 0.7797, "step": 16407 }, { "epoch": 1.7, "learning_rate": 3.670442126557092e-05, "loss": 0.3205, "step": 16408 }, { "epoch": 1.7, "learning_rate": 3.6702022076175125e-05, "loss": 0.3016, "step": 16409 }, { "epoch": 1.7, "learning_rate": 3.6699622841662884e-05, "loss": 1.0373, "step": 16410 }, { "epoch": 1.7, "learning_rate": 3.669722356205035e-05, "loss": 0.8769, "step": 16411 }, { "epoch": 1.7, "learning_rate": 3.669482423735368e-05, "loss": 0.0932, "step": 16412 }, { "epoch": 1.7, "learning_rate": 3.669242486758901e-05, "loss": 0.0962, "step": 16413 }, { "epoch": 1.7, "learning_rate": 3.6690025452772515e-05, "loss": 0.1705, "step": 16414 }, { "epoch": 1.7, "learning_rate": 3.668762599292032e-05, "loss": 0.3859, "step": 16415 }, { "epoch": 1.7, "learning_rate": 3.66852264880486e-05, "loss": 0.3791, "step": 16416 }, { "epoch": 1.7, "learning_rate": 3.6682826938173494e-05, "loss": 0.1373, "step": 16417 }, { "epoch": 1.7, "learning_rate": 3.6680427343311164e-05, "loss": 0.3147, "step": 16418 }, { "epoch": 1.7, "learning_rate": 3.667802770347776e-05, "loss": 0.9416, "step": 16419 }, { "epoch": 1.7, "learning_rate": 3.667562801868943e-05, "loss": 0.574, "step": 16420 }, { "epoch": 1.7, "learning_rate": 3.6673228288962345e-05, "loss": 0.925, "step": 16421 }, { "epoch": 1.7, "learning_rate": 3.6670828514312644e-05, "loss": 0.2079, "step": 16422 }, { "epoch": 1.7, "learning_rate": 3.666842869475649e-05, "loss": 0.7617, "step": 16423 }, { "epoch": 1.7, "learning_rate": 3.666602883031002e-05, "loss": 0.6283, "step": 16424 }, { "epoch": 1.7, "learning_rate": 3.666362892098942e-05, "loss": 0.2828, "step": 16425 }, { "epoch": 1.7, "learning_rate": 3.666122896681082e-05, "loss": 0.8403, "step": 16426 }, { "epoch": 1.7, "learning_rate": 3.6658828967790395e-05, "loss": 0.2702, "step": 16427 }, { "epoch": 1.7, "learning_rate": 3.665642892394428e-05, "loss": 0.7328, "step": 16428 }, { "epoch": 1.7, "learning_rate": 3.665402883528865e-05, "loss": 0.6272, "step": 16429 }, { "epoch": 1.7, "learning_rate": 3.665162870183965e-05, "loss": 0.586, "step": 16430 }, { "epoch": 1.7, "learning_rate": 3.664922852361344e-05, "loss": 0.8658, "step": 16431 }, { "epoch": 1.7, "learning_rate": 3.664682830062618e-05, "loss": 0.7158, "step": 16432 }, { "epoch": 1.7, "learning_rate": 3.664442803289404e-05, "loss": 0.1711, "step": 16433 }, { "epoch": 1.7, "learning_rate": 3.664202772043315e-05, "loss": 0.5217, "step": 16434 }, { "epoch": 1.7, "learning_rate": 3.6639627363259696e-05, "loss": 0.6387, "step": 16435 }, { "epoch": 1.7, "learning_rate": 3.663722696138981e-05, "loss": 0.2644, "step": 16436 }, { "epoch": 1.7, "learning_rate": 3.663482651483967e-05, "loss": 0.3743, "step": 16437 }, { "epoch": 1.7, "learning_rate": 3.663242602362543e-05, "loss": 0.5494, "step": 16438 }, { "epoch": 1.7, "learning_rate": 3.663002548776326e-05, "loss": 0.1929, "step": 16439 }, { "epoch": 1.7, "learning_rate": 3.662762490726931e-05, "loss": 1.1384, "step": 16440 }, { "epoch": 1.7, "learning_rate": 3.662522428215973e-05, "loss": 0.7963, "step": 16441 }, { "epoch": 1.7, "learning_rate": 3.662282361245069e-05, "loss": 0.7497, "step": 16442 }, { "epoch": 1.7, "learning_rate": 3.662042289815835e-05, "loss": 0.8346, "step": 16443 }, { "epoch": 1.7, "learning_rate": 3.661802213929888e-05, "loss": 0.7879, "step": 16444 }, { "epoch": 1.7, "learning_rate": 3.6615621335888435e-05, "loss": 0.4904, "step": 16445 }, { "epoch": 1.7, "learning_rate": 3.661322048794317e-05, "loss": 1.6069, "step": 16446 }, { "epoch": 1.7, "learning_rate": 3.661081959547924e-05, "loss": 0.755, "step": 16447 }, { "epoch": 1.7, "learning_rate": 3.660841865851284e-05, "loss": 0.5465, "step": 16448 }, { "epoch": 1.7, "learning_rate": 3.66060176770601e-05, "loss": 0.3358, "step": 16449 }, { "epoch": 1.7, "learning_rate": 3.660361665113719e-05, "loss": 0.4906, "step": 16450 }, { "epoch": 1.7, "learning_rate": 3.66012155807603e-05, "loss": 0.5372, "step": 16451 }, { "epoch": 1.7, "learning_rate": 3.6598814465945544e-05, "loss": 0.2912, "step": 16452 }, { "epoch": 1.7, "learning_rate": 3.659641330670912e-05, "loss": 0.549, "step": 16453 }, { "epoch": 1.7, "learning_rate": 3.65940121030672e-05, "loss": 1.3672, "step": 16454 }, { "epoch": 1.7, "learning_rate": 3.659161085503592e-05, "loss": 0.4311, "step": 16455 }, { "epoch": 1.7, "learning_rate": 3.658920956263146e-05, "loss": 1.4673, "step": 16456 }, { "epoch": 1.7, "learning_rate": 3.6586808225869984e-05, "loss": 0.6916, "step": 16457 }, { "epoch": 1.7, "learning_rate": 3.658440684476766e-05, "loss": 0.4782, "step": 16458 }, { "epoch": 1.7, "learning_rate": 3.658200541934065e-05, "loss": 0.2728, "step": 16459 }, { "epoch": 1.7, "learning_rate": 3.657960394960511e-05, "loss": 0.9096, "step": 16460 }, { "epoch": 1.7, "learning_rate": 3.6577202435577225e-05, "loss": 0.5293, "step": 16461 }, { "epoch": 1.7, "learning_rate": 3.657480087727315e-05, "loss": 0.3474, "step": 16462 }, { "epoch": 1.7, "learning_rate": 3.657239927470906e-05, "loss": 0.4549, "step": 16463 }, { "epoch": 1.7, "learning_rate": 3.656999762790111e-05, "loss": 0.3569, "step": 16464 }, { "epoch": 1.7, "learning_rate": 3.656759593686547e-05, "loss": 0.2634, "step": 16465 }, { "epoch": 1.7, "learning_rate": 3.656519420161831e-05, "loss": 0.8567, "step": 16466 }, { "epoch": 1.7, "learning_rate": 3.65627924221758e-05, "loss": 0.4847, "step": 16467 }, { "epoch": 1.7, "learning_rate": 3.6560390598554114e-05, "loss": 0.467, "step": 16468 }, { "epoch": 1.7, "learning_rate": 3.65579887307694e-05, "loss": 0.7271, "step": 16469 }, { "epoch": 1.7, "learning_rate": 3.655558681883785e-05, "loss": 0.2649, "step": 16470 }, { "epoch": 1.7, "learning_rate": 3.655318486277563e-05, "loss": 0.936, "step": 16471 }, { "epoch": 1.7, "learning_rate": 3.6550782862598885e-05, "loss": 0.5262, "step": 16472 }, { "epoch": 1.7, "learning_rate": 3.654838081832382e-05, "loss": 0.5179, "step": 16473 }, { "epoch": 1.71, "learning_rate": 3.6545978729966575e-05, "loss": 0.3023, "step": 16474 }, { "epoch": 1.71, "learning_rate": 3.654357659754334e-05, "loss": 0.6411, "step": 16475 }, { "epoch": 1.71, "learning_rate": 3.6541174421070274e-05, "loss": 0.2245, "step": 16476 }, { "epoch": 1.71, "learning_rate": 3.653877220056356e-05, "loss": 0.4532, "step": 16477 }, { "epoch": 1.71, "learning_rate": 3.653636993603935e-05, "loss": 0.51, "step": 16478 }, { "epoch": 1.71, "learning_rate": 3.653396762751384e-05, "loss": 0.876, "step": 16479 }, { "epoch": 1.71, "learning_rate": 3.653156527500318e-05, "loss": 0.875, "step": 16480 }, { "epoch": 1.71, "learning_rate": 3.652916287852355e-05, "loss": 0.6027, "step": 16481 }, { "epoch": 1.71, "learning_rate": 3.652676043809113e-05, "loss": 0.2501, "step": 16482 }, { "epoch": 1.71, "learning_rate": 3.652435795372209e-05, "loss": 0.5684, "step": 16483 }, { "epoch": 1.71, "learning_rate": 3.652195542543258e-05, "loss": 0.3985, "step": 16484 }, { "epoch": 1.71, "learning_rate": 3.6519552853238796e-05, "loss": 0.5402, "step": 16485 }, { "epoch": 1.71, "learning_rate": 3.651715023715692e-05, "loss": 0.2314, "step": 16486 }, { "epoch": 1.71, "learning_rate": 3.6514747577203114e-05, "loss": 0.4409, "step": 16487 }, { "epoch": 1.71, "learning_rate": 3.651234487339354e-05, "loss": 1.4114, "step": 16488 }, { "epoch": 1.71, "learning_rate": 3.650994212574439e-05, "loss": 0.6406, "step": 16489 }, { "epoch": 1.71, "learning_rate": 3.650753933427183e-05, "loss": 0.1936, "step": 16490 }, { "epoch": 1.71, "learning_rate": 3.650513649899204e-05, "loss": 0.6506, "step": 16491 }, { "epoch": 1.71, "learning_rate": 3.65027336199212e-05, "loss": 0.1569, "step": 16492 }, { "epoch": 1.71, "learning_rate": 3.650033069707547e-05, "loss": 0.3861, "step": 16493 }, { "epoch": 1.71, "learning_rate": 3.649792773047103e-05, "loss": 0.4762, "step": 16494 }, { "epoch": 1.71, "learning_rate": 3.649552472012407e-05, "loss": 0.3287, "step": 16495 }, { "epoch": 1.71, "learning_rate": 3.6493121666050753e-05, "loss": 0.7312, "step": 16496 }, { "epoch": 1.71, "learning_rate": 3.6490718568267264e-05, "loss": 0.8436, "step": 16497 }, { "epoch": 1.71, "learning_rate": 3.648831542678977e-05, "loss": 0.2703, "step": 16498 }, { "epoch": 1.71, "learning_rate": 3.6485912241634466e-05, "loss": 0.9142, "step": 16499 }, { "epoch": 1.71, "learning_rate": 3.648350901281751e-05, "loss": 0.5479, "step": 16500 }, { "epoch": 1.71, "learning_rate": 3.648110574035509e-05, "loss": 1.3608, "step": 16501 }, { "epoch": 1.71, "learning_rate": 3.6478702424263386e-05, "loss": 0.5852, "step": 16502 }, { "epoch": 1.71, "learning_rate": 3.6476299064558573e-05, "loss": 0.3774, "step": 16503 }, { "epoch": 1.71, "learning_rate": 3.647389566125682e-05, "loss": 1.6183, "step": 16504 }, { "epoch": 1.71, "learning_rate": 3.647149221437432e-05, "loss": 0.8415, "step": 16505 }, { "epoch": 1.71, "learning_rate": 3.646908872392726e-05, "loss": 0.3731, "step": 16506 }, { "epoch": 1.71, "learning_rate": 3.64666851899318e-05, "loss": 1.0414, "step": 16507 }, { "epoch": 1.71, "learning_rate": 3.646428161240413e-05, "loss": 0.3642, "step": 16508 }, { "epoch": 1.71, "learning_rate": 3.6461877991360424e-05, "loss": 0.4611, "step": 16509 }, { "epoch": 1.71, "learning_rate": 3.645947432681687e-05, "loss": 0.3509, "step": 16510 }, { "epoch": 1.71, "learning_rate": 3.6457070618789645e-05, "loss": 0.2417, "step": 16511 }, { "epoch": 1.71, "learning_rate": 3.645466686729493e-05, "loss": 0.8741, "step": 16512 }, { "epoch": 1.71, "learning_rate": 3.645226307234891e-05, "loss": 0.7738, "step": 16513 }, { "epoch": 1.71, "learning_rate": 3.644985923396777e-05, "loss": 0.5673, "step": 16514 }, { "epoch": 1.71, "learning_rate": 3.644745535216768e-05, "loss": 0.3116, "step": 16515 }, { "epoch": 1.71, "learning_rate": 3.644505142696484e-05, "loss": 1.0215, "step": 16516 }, { "epoch": 1.71, "learning_rate": 3.644264745837541e-05, "loss": 0.5176, "step": 16517 }, { "epoch": 1.71, "learning_rate": 3.6440243446415584e-05, "loss": 0.9969, "step": 16518 }, { "epoch": 1.71, "learning_rate": 3.643783939110155e-05, "loss": 1.3307, "step": 16519 }, { "epoch": 1.71, "learning_rate": 3.643543529244949e-05, "loss": 0.7288, "step": 16520 }, { "epoch": 1.71, "learning_rate": 3.6433031150475586e-05, "loss": 0.6135, "step": 16521 }, { "epoch": 1.71, "learning_rate": 3.6430626965196016e-05, "loss": 0.4436, "step": 16522 }, { "epoch": 1.71, "learning_rate": 3.642822273662697e-05, "loss": 0.8525, "step": 16523 }, { "epoch": 1.71, "learning_rate": 3.642581846478463e-05, "loss": 1.0203, "step": 16524 }, { "epoch": 1.71, "learning_rate": 3.64234141496852e-05, "loss": 0.3823, "step": 16525 }, { "epoch": 1.71, "learning_rate": 3.6421009791344834e-05, "loss": 1.0884, "step": 16526 }, { "epoch": 1.71, "learning_rate": 3.641860538977974e-05, "loss": 0.5313, "step": 16527 }, { "epoch": 1.71, "learning_rate": 3.641620094500609e-05, "loss": 1.7644, "step": 16528 }, { "epoch": 1.71, "learning_rate": 3.6413796457040075e-05, "loss": 0.4984, "step": 16529 }, { "epoch": 1.71, "learning_rate": 3.641139192589789e-05, "loss": 0.57, "step": 16530 }, { "epoch": 1.71, "learning_rate": 3.6408987351595713e-05, "loss": 0.5781, "step": 16531 }, { "epoch": 1.71, "learning_rate": 3.640658273414973e-05, "loss": 0.4336, "step": 16532 }, { "epoch": 1.71, "learning_rate": 3.6404178073576134e-05, "loss": 0.2199, "step": 16533 }, { "epoch": 1.71, "learning_rate": 3.64017733698911e-05, "loss": 0.4599, "step": 16534 }, { "epoch": 1.71, "learning_rate": 3.6399368623110845e-05, "loss": 0.4351, "step": 16535 }, { "epoch": 1.71, "learning_rate": 3.639696383325152e-05, "loss": 0.2926, "step": 16536 }, { "epoch": 1.71, "learning_rate": 3.639455900032933e-05, "loss": 1.5483, "step": 16537 }, { "epoch": 1.71, "learning_rate": 3.639215412436048e-05, "loss": 0.7003, "step": 16538 }, { "epoch": 1.71, "learning_rate": 3.638974920536114e-05, "loss": 1.1643, "step": 16539 }, { "epoch": 1.71, "learning_rate": 3.638734424334751e-05, "loss": 0.5717, "step": 16540 }, { "epoch": 1.71, "learning_rate": 3.638493923833575e-05, "loss": 0.4091, "step": 16541 }, { "epoch": 1.71, "learning_rate": 3.6382534190342085e-05, "loss": 0.2291, "step": 16542 }, { "epoch": 1.71, "learning_rate": 3.6380129099382694e-05, "loss": 0.1807, "step": 16543 }, { "epoch": 1.71, "learning_rate": 3.637772396547377e-05, "loss": 1.5624, "step": 16544 }, { "epoch": 1.71, "learning_rate": 3.6375318788631496e-05, "loss": 0.2627, "step": 16545 }, { "epoch": 1.71, "learning_rate": 3.637291356887208e-05, "loss": 0.4585, "step": 16546 }, { "epoch": 1.71, "learning_rate": 3.637050830621169e-05, "loss": 1.1744, "step": 16547 }, { "epoch": 1.71, "learning_rate": 3.636810300066652e-05, "loss": 0.9621, "step": 16548 }, { "epoch": 1.71, "learning_rate": 3.636569765225278e-05, "loss": 0.4521, "step": 16549 }, { "epoch": 1.71, "learning_rate": 3.636329226098665e-05, "loss": 0.4207, "step": 16550 }, { "epoch": 1.71, "learning_rate": 3.6360886826884325e-05, "loss": 0.4599, "step": 16551 }, { "epoch": 1.71, "learning_rate": 3.635848134996201e-05, "loss": 1.0565, "step": 16552 }, { "epoch": 1.71, "learning_rate": 3.635607583023588e-05, "loss": 0.812, "step": 16553 }, { "epoch": 1.71, "learning_rate": 3.635367026772213e-05, "loss": 0.8874, "step": 16554 }, { "epoch": 1.71, "learning_rate": 3.6351264662436963e-05, "loss": 0.471, "step": 16555 }, { "epoch": 1.71, "learning_rate": 3.634885901439656e-05, "loss": 1.2007, "step": 16556 }, { "epoch": 1.71, "learning_rate": 3.634645332361712e-05, "loss": 0.3359, "step": 16557 }, { "epoch": 1.71, "learning_rate": 3.634404759011486e-05, "loss": 0.6492, "step": 16558 }, { "epoch": 1.71, "learning_rate": 3.634164181390596e-05, "loss": 0.6593, "step": 16559 }, { "epoch": 1.71, "learning_rate": 3.633923599500659e-05, "loss": 1.098, "step": 16560 }, { "epoch": 1.71, "learning_rate": 3.633683013343297e-05, "loss": 0.32, "step": 16561 }, { "epoch": 1.71, "learning_rate": 3.6334424229201296e-05, "loss": 0.7012, "step": 16562 }, { "epoch": 1.71, "learning_rate": 3.633201828232777e-05, "loss": 0.7277, "step": 16563 }, { "epoch": 1.71, "learning_rate": 3.6329612292828565e-05, "loss": 0.6171, "step": 16564 }, { "epoch": 1.71, "learning_rate": 3.6327206260719895e-05, "loss": 0.9809, "step": 16565 }, { "epoch": 1.71, "learning_rate": 3.632480018601795e-05, "loss": 0.0604, "step": 16566 }, { "epoch": 1.71, "learning_rate": 3.632239406873893e-05, "loss": 0.3356, "step": 16567 }, { "epoch": 1.71, "learning_rate": 3.6319987908899043e-05, "loss": 1.0633, "step": 16568 }, { "epoch": 1.71, "learning_rate": 3.631758170651447e-05, "loss": 0.6329, "step": 16569 }, { "epoch": 1.72, "learning_rate": 3.631517546160141e-05, "loss": 1.3612, "step": 16570 }, { "epoch": 1.72, "learning_rate": 3.6312769174176074e-05, "loss": 0.4865, "step": 16571 }, { "epoch": 1.72, "learning_rate": 3.6310362844254645e-05, "loss": 0.6755, "step": 16572 }, { "epoch": 1.72, "learning_rate": 3.6307956471853335e-05, "loss": 1.8857, "step": 16573 }, { "epoch": 1.72, "learning_rate": 3.6305550056988346e-05, "loss": 1.1025, "step": 16574 }, { "epoch": 1.72, "learning_rate": 3.630314359967585e-05, "loss": 0.3005, "step": 16575 }, { "epoch": 1.72, "learning_rate": 3.630073709993208e-05, "loss": 1.1228, "step": 16576 }, { "epoch": 1.72, "learning_rate": 3.629833055777322e-05, "loss": 0.2466, "step": 16577 }, { "epoch": 1.72, "learning_rate": 3.629592397321548e-05, "loss": 0.1965, "step": 16578 }, { "epoch": 1.72, "learning_rate": 3.629351734627504e-05, "loss": 0.2173, "step": 16579 }, { "epoch": 1.72, "learning_rate": 3.6291110676968115e-05, "loss": 0.7652, "step": 16580 }, { "epoch": 1.72, "learning_rate": 3.6288703965310905e-05, "loss": 0.7487, "step": 16581 }, { "epoch": 1.72, "learning_rate": 3.628629721131962e-05, "loss": 0.4192, "step": 16582 }, { "epoch": 1.72, "learning_rate": 3.628389041501045e-05, "loss": 0.6981, "step": 16583 }, { "epoch": 1.72, "learning_rate": 3.62814835763996e-05, "loss": 0.6687, "step": 16584 }, { "epoch": 1.72, "learning_rate": 3.627907669550326e-05, "loss": 0.0806, "step": 16585 }, { "epoch": 1.72, "learning_rate": 3.627666977233765e-05, "loss": 0.3318, "step": 16586 }, { "epoch": 1.72, "learning_rate": 3.627426280691898e-05, "loss": 0.2816, "step": 16587 }, { "epoch": 1.72, "learning_rate": 3.627185579926343e-05, "loss": 0.3864, "step": 16588 }, { "epoch": 1.72, "learning_rate": 3.626944874938722e-05, "loss": 0.5368, "step": 16589 }, { "epoch": 1.72, "learning_rate": 3.626704165730655e-05, "loss": 1.0224, "step": 16590 }, { "epoch": 1.72, "learning_rate": 3.626463452303761e-05, "loss": 0.4377, "step": 16591 }, { "epoch": 1.72, "learning_rate": 3.626222734659663e-05, "loss": 0.4082, "step": 16592 }, { "epoch": 1.72, "learning_rate": 3.625982012799979e-05, "loss": 0.2079, "step": 16593 }, { "epoch": 1.72, "learning_rate": 3.625741286726331e-05, "loss": 0.6388, "step": 16594 }, { "epoch": 1.72, "learning_rate": 3.625500556440339e-05, "loss": 0.4363, "step": 16595 }, { "epoch": 1.72, "learning_rate": 3.625259821943624e-05, "loss": 0.9457, "step": 16596 }, { "epoch": 1.72, "learning_rate": 3.625019083237806e-05, "loss": 0.7442, "step": 16597 }, { "epoch": 1.72, "learning_rate": 3.624778340324505e-05, "loss": 0.4794, "step": 16598 }, { "epoch": 1.72, "learning_rate": 3.6245375932053434e-05, "loss": 0.7955, "step": 16599 }, { "epoch": 1.72, "learning_rate": 3.62429684188194e-05, "loss": 0.9049, "step": 16600 }, { "epoch": 1.72, "learning_rate": 3.6240560863559174e-05, "loss": 0.4717, "step": 16601 }, { "epoch": 1.72, "learning_rate": 3.623815326628895e-05, "loss": 1.0221, "step": 16602 }, { "epoch": 1.72, "learning_rate": 3.623574562702493e-05, "loss": 0.8877, "step": 16603 }, { "epoch": 1.72, "learning_rate": 3.6233337945783333e-05, "loss": 0.304, "step": 16604 }, { "epoch": 1.72, "learning_rate": 3.6230930222580365e-05, "loss": 0.6782, "step": 16605 }, { "epoch": 1.72, "learning_rate": 3.622852245743223e-05, "loss": 0.1706, "step": 16606 }, { "epoch": 1.72, "learning_rate": 3.622611465035514e-05, "loss": 0.8962, "step": 16607 }, { "epoch": 1.72, "learning_rate": 3.622370680136531e-05, "loss": 0.401, "step": 16608 }, { "epoch": 1.72, "learning_rate": 3.622129891047894e-05, "loss": 0.6004, "step": 16609 }, { "epoch": 1.72, "learning_rate": 3.6218890977712234e-05, "loss": 1.0981, "step": 16610 }, { "epoch": 1.72, "learning_rate": 3.621648300308141e-05, "loss": 0.2481, "step": 16611 }, { "epoch": 1.72, "learning_rate": 3.621407498660268e-05, "loss": 0.3984, "step": 16612 }, { "epoch": 1.72, "learning_rate": 3.621166692829225e-05, "loss": 0.5769, "step": 16613 }, { "epoch": 1.72, "learning_rate": 3.620925882816633e-05, "loss": 0.4829, "step": 16614 }, { "epoch": 1.72, "learning_rate": 3.620685068624114e-05, "loss": 0.4972, "step": 16615 }, { "epoch": 1.72, "learning_rate": 3.620444250253288e-05, "loss": 0.3596, "step": 16616 }, { "epoch": 1.72, "learning_rate": 3.6202034277057765e-05, "loss": 0.6032, "step": 16617 }, { "epoch": 1.72, "learning_rate": 3.6199626009831995e-05, "loss": 0.3752, "step": 16618 }, { "epoch": 1.72, "learning_rate": 3.61972177008718e-05, "loss": 0.2309, "step": 16619 }, { "epoch": 1.72, "learning_rate": 3.61948093501934e-05, "loss": 0.527, "step": 16620 }, { "epoch": 1.72, "learning_rate": 3.6192400957812974e-05, "loss": 0.451, "step": 16621 }, { "epoch": 1.72, "learning_rate": 3.6189992523746765e-05, "loss": 0.7028, "step": 16622 }, { "epoch": 1.72, "learning_rate": 3.618758404801097e-05, "loss": 0.8793, "step": 16623 }, { "epoch": 1.72, "learning_rate": 3.61851755306218e-05, "loss": 0.6291, "step": 16624 }, { "epoch": 1.72, "learning_rate": 3.618276697159549e-05, "loss": 1.7549, "step": 16625 }, { "epoch": 1.72, "learning_rate": 3.618035837094823e-05, "loss": 0.9082, "step": 16626 }, { "epoch": 1.72, "learning_rate": 3.617794972869625e-05, "loss": 0.8371, "step": 16627 }, { "epoch": 1.72, "learning_rate": 3.617554104485575e-05, "loss": 0.6394, "step": 16628 }, { "epoch": 1.72, "learning_rate": 3.6173132319442954e-05, "loss": 0.5894, "step": 16629 }, { "epoch": 1.72, "learning_rate": 3.617072355247408e-05, "loss": 0.4502, "step": 16630 }, { "epoch": 1.72, "learning_rate": 3.616831474396534e-05, "loss": 0.6638, "step": 16631 }, { "epoch": 1.72, "learning_rate": 3.6165905893932944e-05, "loss": 0.4645, "step": 16632 }, { "epoch": 1.72, "learning_rate": 3.616349700239311e-05, "loss": 0.2215, "step": 16633 }, { "epoch": 1.72, "learning_rate": 3.616108806936207e-05, "loss": 0.2968, "step": 16634 }, { "epoch": 1.72, "learning_rate": 3.6158679094856014e-05, "loss": 0.2244, "step": 16635 }, { "epoch": 1.72, "learning_rate": 3.6156270078891174e-05, "loss": 0.4197, "step": 16636 }, { "epoch": 1.72, "learning_rate": 3.6153861021483764e-05, "loss": 0.7181, "step": 16637 }, { "epoch": 1.72, "learning_rate": 3.615145192264999e-05, "loss": 0.3975, "step": 16638 }, { "epoch": 1.72, "learning_rate": 3.61490427824061e-05, "loss": 0.4141, "step": 16639 }, { "epoch": 1.72, "learning_rate": 3.614663360076829e-05, "loss": 1.3283, "step": 16640 }, { "epoch": 1.72, "learning_rate": 3.614422437775277e-05, "loss": 0.8952, "step": 16641 }, { "epoch": 1.72, "learning_rate": 3.614181511337578e-05, "loss": 1.2051, "step": 16642 }, { "epoch": 1.72, "learning_rate": 3.613940580765352e-05, "loss": 0.6975, "step": 16643 }, { "epoch": 1.72, "learning_rate": 3.6136996460602224e-05, "loss": 0.7563, "step": 16644 }, { "epoch": 1.72, "learning_rate": 3.6134587072238096e-05, "loss": 1.1374, "step": 16645 }, { "epoch": 1.72, "learning_rate": 3.613217764257737e-05, "loss": 0.3196, "step": 16646 }, { "epoch": 1.72, "learning_rate": 3.612976817163625e-05, "loss": 0.2468, "step": 16647 }, { "epoch": 1.72, "learning_rate": 3.612735865943097e-05, "loss": 1.0374, "step": 16648 }, { "epoch": 1.72, "learning_rate": 3.612494910597775e-05, "loss": 0.344, "step": 16649 }, { "epoch": 1.72, "learning_rate": 3.61225395112928e-05, "loss": 0.5551, "step": 16650 }, { "epoch": 1.72, "learning_rate": 3.612012987539235e-05, "loss": 1.0, "step": 16651 }, { "epoch": 1.72, "learning_rate": 3.611772019829262e-05, "loss": 1.3245, "step": 16652 }, { "epoch": 1.72, "learning_rate": 3.611531048000982e-05, "loss": 1.134, "step": 16653 }, { "epoch": 1.72, "learning_rate": 3.611290072056019e-05, "loss": 0.8743, "step": 16654 }, { "epoch": 1.72, "learning_rate": 3.611049091995993e-05, "loss": 0.526, "step": 16655 }, { "epoch": 1.72, "learning_rate": 3.610808107822529e-05, "loss": 0.685, "step": 16656 }, { "epoch": 1.72, "learning_rate": 3.610567119537246e-05, "loss": 0.8003, "step": 16657 }, { "epoch": 1.72, "learning_rate": 3.61032612714177e-05, "loss": 1.0133, "step": 16658 }, { "epoch": 1.72, "learning_rate": 3.610085130637721e-05, "loss": 1.1952, "step": 16659 }, { "epoch": 1.72, "learning_rate": 3.60984413002672e-05, "loss": 0.8222, "step": 16660 }, { "epoch": 1.72, "learning_rate": 3.609603125310392e-05, "loss": 0.7655, "step": 16661 }, { "epoch": 1.72, "learning_rate": 3.609362116490358e-05, "loss": 0.7405, "step": 16662 }, { "epoch": 1.72, "learning_rate": 3.6091211035682416e-05, "loss": 1.0667, "step": 16663 }, { "epoch": 1.72, "learning_rate": 3.6088800865456645e-05, "loss": 0.3885, "step": 16664 }, { "epoch": 1.73, "learning_rate": 3.608639065424249e-05, "loss": 1.1788, "step": 16665 }, { "epoch": 1.73, "learning_rate": 3.608398040205617e-05, "loss": 1.247, "step": 16666 }, { "epoch": 1.73, "learning_rate": 3.6081570108913924e-05, "loss": 0.622, "step": 16667 }, { "epoch": 1.73, "learning_rate": 3.607915977483197e-05, "loss": 0.536, "step": 16668 }, { "epoch": 1.73, "learning_rate": 3.6076749399826535e-05, "loss": 0.5897, "step": 16669 }, { "epoch": 1.73, "learning_rate": 3.6074338983913836e-05, "loss": 0.5876, "step": 16670 }, { "epoch": 1.73, "learning_rate": 3.607192852711012e-05, "loss": 0.5152, "step": 16671 }, { "epoch": 1.73, "learning_rate": 3.6069518029431596e-05, "loss": 0.6186, "step": 16672 }, { "epoch": 1.73, "learning_rate": 3.606710749089451e-05, "loss": 0.9591, "step": 16673 }, { "epoch": 1.73, "learning_rate": 3.6064696911515054e-05, "loss": 0.2746, "step": 16674 }, { "epoch": 1.73, "learning_rate": 3.606228629130949e-05, "loss": 0.5759, "step": 16675 }, { "epoch": 1.73, "learning_rate": 3.605987563029402e-05, "loss": 0.6626, "step": 16676 }, { "epoch": 1.73, "learning_rate": 3.60574649284849e-05, "loss": 0.6042, "step": 16677 }, { "epoch": 1.73, "learning_rate": 3.605505418589835e-05, "loss": 1.0766, "step": 16678 }, { "epoch": 1.73, "learning_rate": 3.605264340255058e-05, "loss": 0.8596, "step": 16679 }, { "epoch": 1.73, "learning_rate": 3.6050232578457824e-05, "loss": 0.8436, "step": 16680 }, { "epoch": 1.73, "learning_rate": 3.604782171363633e-05, "loss": 0.6431, "step": 16681 }, { "epoch": 1.73, "learning_rate": 3.6045410808102316e-05, "loss": 0.5225, "step": 16682 }, { "epoch": 1.73, "learning_rate": 3.6042999861872e-05, "loss": 0.7814, "step": 16683 }, { "epoch": 1.73, "learning_rate": 3.604058887496164e-05, "loss": 0.5301, "step": 16684 }, { "epoch": 1.73, "learning_rate": 3.6038177847387434e-05, "loss": 0.2978, "step": 16685 }, { "epoch": 1.73, "learning_rate": 3.6035766779165623e-05, "loss": 1.3827, "step": 16686 }, { "epoch": 1.73, "learning_rate": 3.6033355670312455e-05, "loss": 0.3967, "step": 16687 }, { "epoch": 1.73, "learning_rate": 3.6030944520844145e-05, "loss": 1.1178, "step": 16688 }, { "epoch": 1.73, "learning_rate": 3.602853333077692e-05, "loss": 0.8229, "step": 16689 }, { "epoch": 1.73, "learning_rate": 3.602612210012704e-05, "loss": 0.3065, "step": 16690 }, { "epoch": 1.73, "learning_rate": 3.60237108289107e-05, "loss": 1.2709, "step": 16691 }, { "epoch": 1.73, "learning_rate": 3.602129951714415e-05, "loss": 0.204, "step": 16692 }, { "epoch": 1.73, "learning_rate": 3.6018888164843625e-05, "loss": 0.5916, "step": 16693 }, { "epoch": 1.73, "learning_rate": 3.6016476772025345e-05, "loss": 0.6991, "step": 16694 }, { "epoch": 1.73, "learning_rate": 3.6014065338705554e-05, "loss": 0.5053, "step": 16695 }, { "epoch": 1.73, "learning_rate": 3.601165386490048e-05, "loss": 1.1995, "step": 16696 }, { "epoch": 1.73, "learning_rate": 3.600924235062637e-05, "loss": 0.5126, "step": 16697 }, { "epoch": 1.73, "learning_rate": 3.600683079589944e-05, "loss": 0.7253, "step": 16698 }, { "epoch": 1.73, "learning_rate": 3.600441920073592e-05, "loss": 0.8396, "step": 16699 }, { "epoch": 1.73, "learning_rate": 3.600200756515207e-05, "loss": 0.4967, "step": 16700 }, { "epoch": 1.73, "learning_rate": 3.599959588916411e-05, "loss": 0.1825, "step": 16701 }, { "epoch": 1.73, "learning_rate": 3.599718417278826e-05, "loss": 0.2458, "step": 16702 }, { "epoch": 1.73, "learning_rate": 3.5994772416040784e-05, "loss": 0.1578, "step": 16703 }, { "epoch": 1.73, "learning_rate": 3.599236061893789e-05, "loss": 0.3713, "step": 16704 }, { "epoch": 1.73, "learning_rate": 3.598994878149584e-05, "loss": 0.8321, "step": 16705 }, { "epoch": 1.73, "learning_rate": 3.598753690373085e-05, "loss": 0.2294, "step": 16706 }, { "epoch": 1.73, "learning_rate": 3.598512498565916e-05, "loss": 0.6201, "step": 16707 }, { "epoch": 1.73, "learning_rate": 3.5982713027297005e-05, "loss": 0.1209, "step": 16708 }, { "epoch": 1.73, "learning_rate": 3.5980301028660635e-05, "loss": 0.9235, "step": 16709 }, { "epoch": 1.73, "learning_rate": 3.597788898976627e-05, "loss": 0.5498, "step": 16710 }, { "epoch": 1.73, "learning_rate": 3.597547691063016e-05, "loss": 0.6468, "step": 16711 }, { "epoch": 1.73, "learning_rate": 3.597306479126854e-05, "loss": 0.2187, "step": 16712 }, { "epoch": 1.73, "learning_rate": 3.597065263169764e-05, "loss": 0.3079, "step": 16713 }, { "epoch": 1.73, "learning_rate": 3.5968240431933706e-05, "loss": 0.6276, "step": 16714 }, { "epoch": 1.73, "learning_rate": 3.596582819199298e-05, "loss": 0.4109, "step": 16715 }, { "epoch": 1.73, "learning_rate": 3.5963415911891686e-05, "loss": 0.8793, "step": 16716 }, { "epoch": 1.73, "learning_rate": 3.5961003591646074e-05, "loss": 0.8032, "step": 16717 }, { "epoch": 1.73, "learning_rate": 3.595859123127238e-05, "loss": 0.37, "step": 16718 }, { "epoch": 1.73, "learning_rate": 3.5956178830786835e-05, "loss": 0.1208, "step": 16719 }, { "epoch": 1.73, "learning_rate": 3.5953766390205705e-05, "loss": 0.9253, "step": 16720 }, { "epoch": 1.73, "learning_rate": 3.59513539095452e-05, "loss": 0.4226, "step": 16721 }, { "epoch": 1.73, "learning_rate": 3.594894138882158e-05, "loss": 0.7912, "step": 16722 }, { "epoch": 1.73, "learning_rate": 3.5946528828051077e-05, "loss": 0.6492, "step": 16723 }, { "epoch": 1.73, "learning_rate": 3.594411622724993e-05, "loss": 0.2615, "step": 16724 }, { "epoch": 1.73, "learning_rate": 3.594170358643439e-05, "loss": 0.749, "step": 16725 }, { "epoch": 1.73, "learning_rate": 3.593929090562069e-05, "loss": 0.5242, "step": 16726 }, { "epoch": 1.73, "learning_rate": 3.593687818482506e-05, "loss": 0.875, "step": 16727 }, { "epoch": 1.73, "learning_rate": 3.593446542406377e-05, "loss": 0.7278, "step": 16728 }, { "epoch": 1.73, "learning_rate": 3.593205262335304e-05, "loss": 0.1875, "step": 16729 }, { "epoch": 1.73, "learning_rate": 3.592963978270913e-05, "loss": 1.1052, "step": 16730 }, { "epoch": 1.73, "learning_rate": 3.5927226902148265e-05, "loss": 0.2595, "step": 16731 }, { "epoch": 1.73, "learning_rate": 3.59248139816867e-05, "loss": 0.1609, "step": 16732 }, { "epoch": 1.73, "learning_rate": 3.592240102134066e-05, "loss": 0.2731, "step": 16733 }, { "epoch": 1.73, "learning_rate": 3.591998802112641e-05, "loss": 0.4025, "step": 16734 }, { "epoch": 1.73, "learning_rate": 3.5917574981060196e-05, "loss": 0.8282, "step": 16735 }, { "epoch": 1.73, "learning_rate": 3.591516190115824e-05, "loss": 0.2566, "step": 16736 }, { "epoch": 1.73, "learning_rate": 3.5912748781436795e-05, "loss": 0.6141, "step": 16737 }, { "epoch": 1.73, "learning_rate": 3.591033562191212e-05, "loss": 0.2279, "step": 16738 }, { "epoch": 1.73, "learning_rate": 3.590792242260045e-05, "loss": 1.1271, "step": 16739 }, { "epoch": 1.73, "learning_rate": 3.590550918351802e-05, "loss": 0.6404, "step": 16740 }, { "epoch": 1.73, "learning_rate": 3.590309590468108e-05, "loss": 0.7359, "step": 16741 }, { "epoch": 1.73, "learning_rate": 3.590068258610589e-05, "loss": 1.3402, "step": 16742 }, { "epoch": 1.73, "learning_rate": 3.5898269227808674e-05, "loss": 1.3108, "step": 16743 }, { "epoch": 1.73, "learning_rate": 3.58958558298057e-05, "loss": 0.3326, "step": 16744 }, { "epoch": 1.73, "learning_rate": 3.58934423921132e-05, "loss": 1.0887, "step": 16745 }, { "epoch": 1.73, "learning_rate": 3.589102891474742e-05, "loss": 0.5292, "step": 16746 }, { "epoch": 1.73, "learning_rate": 3.588861539772463e-05, "loss": 0.3931, "step": 16747 }, { "epoch": 1.73, "learning_rate": 3.588620184106104e-05, "loss": 0.3182, "step": 16748 }, { "epoch": 1.73, "learning_rate": 3.588378824477293e-05, "loss": 0.8019, "step": 16749 }, { "epoch": 1.73, "learning_rate": 3.5881374608876516e-05, "loss": 0.5261, "step": 16750 }, { "epoch": 1.73, "learning_rate": 3.5878960933388075e-05, "loss": 1.0464, "step": 16751 }, { "epoch": 1.73, "learning_rate": 3.587654721832384e-05, "loss": 0.4382, "step": 16752 }, { "epoch": 1.73, "learning_rate": 3.5874133463700075e-05, "loss": 0.8859, "step": 16753 }, { "epoch": 1.73, "learning_rate": 3.5871719669533015e-05, "loss": 0.3152, "step": 16754 }, { "epoch": 1.73, "learning_rate": 3.586930583583891e-05, "loss": 0.671, "step": 16755 }, { "epoch": 1.73, "learning_rate": 3.586689196263401e-05, "loss": 1.0298, "step": 16756 }, { "epoch": 1.73, "learning_rate": 3.586447804993456e-05, "loss": 0.1364, "step": 16757 }, { "epoch": 1.73, "learning_rate": 3.586206409775682e-05, "loss": 1.0532, "step": 16758 }, { "epoch": 1.73, "learning_rate": 3.585965010611705e-05, "loss": 1.0531, "step": 16759 }, { "epoch": 1.73, "learning_rate": 3.585723607503147e-05, "loss": 1.1837, "step": 16760 }, { "epoch": 1.74, "learning_rate": 3.585482200451635e-05, "loss": 0.7469, "step": 16761 }, { "epoch": 1.74, "learning_rate": 3.585240789458794e-05, "loss": 0.2717, "step": 16762 }, { "epoch": 1.74, "learning_rate": 3.584999374526249e-05, "loss": 0.6785, "step": 16763 }, { "epoch": 1.74, "learning_rate": 3.584757955655625e-05, "loss": 1.0375, "step": 16764 }, { "epoch": 1.74, "learning_rate": 3.5845165328485473e-05, "loss": 0.5337, "step": 16765 }, { "epoch": 1.74, "learning_rate": 3.5842751061066415e-05, "loss": 0.2262, "step": 16766 }, { "epoch": 1.74, "learning_rate": 3.584033675431532e-05, "loss": 1.271, "step": 16767 }, { "epoch": 1.74, "learning_rate": 3.5837922408248445e-05, "loss": 0.2261, "step": 16768 }, { "epoch": 1.74, "learning_rate": 3.583550802288204e-05, "loss": 0.2394, "step": 16769 }, { "epoch": 1.74, "learning_rate": 3.5833093598232355e-05, "loss": 1.0296, "step": 16770 }, { "epoch": 1.74, "learning_rate": 3.583067913431566e-05, "loss": 0.2568, "step": 16771 }, { "epoch": 1.74, "learning_rate": 3.5828264631148195e-05, "loss": 0.362, "step": 16772 }, { "epoch": 1.74, "learning_rate": 3.582585008874621e-05, "loss": 0.8867, "step": 16773 }, { "epoch": 1.74, "learning_rate": 3.582343550712597e-05, "loss": 0.9854, "step": 16774 }, { "epoch": 1.74, "learning_rate": 3.582102088630373e-05, "loss": 0.9664, "step": 16775 }, { "epoch": 1.74, "learning_rate": 3.581860622629573e-05, "loss": 1.0273, "step": 16776 }, { "epoch": 1.74, "learning_rate": 3.581619152711824e-05, "loss": 0.9982, "step": 16777 }, { "epoch": 1.74, "learning_rate": 3.581377678878751e-05, "loss": 0.7573, "step": 16778 }, { "epoch": 1.74, "learning_rate": 3.581136201131979e-05, "loss": 0.6993, "step": 16779 }, { "epoch": 1.74, "learning_rate": 3.580894719473134e-05, "loss": 0.1168, "step": 16780 }, { "epoch": 1.74, "learning_rate": 3.580653233903842e-05, "loss": 0.1975, "step": 16781 }, { "epoch": 1.74, "learning_rate": 3.5804117444257286e-05, "loss": 1.1556, "step": 16782 }, { "epoch": 1.74, "learning_rate": 3.580170251040419e-05, "loss": 0.548, "step": 16783 }, { "epoch": 1.74, "learning_rate": 3.579928753749539e-05, "loss": 0.6053, "step": 16784 }, { "epoch": 1.74, "learning_rate": 3.579687252554714e-05, "loss": 0.7977, "step": 16785 }, { "epoch": 1.74, "learning_rate": 3.5794457474575694e-05, "loss": 0.7242, "step": 16786 }, { "epoch": 1.74, "learning_rate": 3.5792042384597336e-05, "loss": 0.567, "step": 16787 }, { "epoch": 1.74, "learning_rate": 3.578962725562829e-05, "loss": 0.6605, "step": 16788 }, { "epoch": 1.74, "learning_rate": 3.5787212087684825e-05, "loss": 1.2174, "step": 16789 }, { "epoch": 1.74, "learning_rate": 3.5784796880783204e-05, "loss": 0.7423, "step": 16790 }, { "epoch": 1.74, "learning_rate": 3.578238163493969e-05, "loss": 0.9034, "step": 16791 }, { "epoch": 1.74, "learning_rate": 3.577996635017054e-05, "loss": 0.5568, "step": 16792 }, { "epoch": 1.74, "learning_rate": 3.5777551026492e-05, "loss": 0.6167, "step": 16793 }, { "epoch": 1.74, "learning_rate": 3.577513566392033e-05, "loss": 0.3675, "step": 16794 }, { "epoch": 1.74, "learning_rate": 3.577272026247182e-05, "loss": 0.6996, "step": 16795 }, { "epoch": 1.74, "learning_rate": 3.5770304822162694e-05, "loss": 0.6058, "step": 16796 }, { "epoch": 1.74, "learning_rate": 3.5767889343009235e-05, "loss": 0.7265, "step": 16797 }, { "epoch": 1.74, "learning_rate": 3.576547382502768e-05, "loss": 0.8412, "step": 16798 }, { "epoch": 1.74, "learning_rate": 3.5763058268234316e-05, "loss": 0.9921, "step": 16799 }, { "epoch": 1.74, "learning_rate": 3.576064267264539e-05, "loss": 0.8345, "step": 16800 }, { "epoch": 1.74, "learning_rate": 3.575822703827717e-05, "loss": 0.2099, "step": 16801 }, { "epoch": 1.74, "learning_rate": 3.57558113651459e-05, "loss": 0.3009, "step": 16802 }, { "epoch": 1.74, "learning_rate": 3.575339565326786e-05, "loss": 0.4905, "step": 16803 }, { "epoch": 1.74, "learning_rate": 3.575097990265931e-05, "loss": 0.2644, "step": 16804 }, { "epoch": 1.74, "learning_rate": 3.5748564113336505e-05, "loss": 0.5186, "step": 16805 }, { "epoch": 1.74, "learning_rate": 3.5746148285315714e-05, "loss": 0.7142, "step": 16806 }, { "epoch": 1.74, "learning_rate": 3.574373241861319e-05, "loss": 0.6362, "step": 16807 }, { "epoch": 1.74, "learning_rate": 3.5741316513245215e-05, "loss": 0.4146, "step": 16808 }, { "epoch": 1.74, "learning_rate": 3.573890056922804e-05, "loss": 0.1963, "step": 16809 }, { "epoch": 1.74, "learning_rate": 3.573648458657792e-05, "loss": 0.4569, "step": 16810 }, { "epoch": 1.74, "learning_rate": 3.573406856531114e-05, "loss": 0.6898, "step": 16811 }, { "epoch": 1.74, "learning_rate": 3.573165250544394e-05, "loss": 0.9836, "step": 16812 }, { "epoch": 1.74, "learning_rate": 3.57292364069926e-05, "loss": 0.7562, "step": 16813 }, { "epoch": 1.74, "learning_rate": 3.5726820269973374e-05, "loss": 0.1382, "step": 16814 }, { "epoch": 1.74, "learning_rate": 3.572440409440255e-05, "loss": 0.9947, "step": 16815 }, { "epoch": 1.74, "learning_rate": 3.572198788029637e-05, "loss": 0.3911, "step": 16816 }, { "epoch": 1.74, "learning_rate": 3.571957162767111e-05, "loss": 0.7601, "step": 16817 }, { "epoch": 1.74, "learning_rate": 3.5717155336543016e-05, "loss": 1.0544, "step": 16818 }, { "epoch": 1.74, "learning_rate": 3.571473900692838e-05, "loss": 0.9157, "step": 16819 }, { "epoch": 1.74, "learning_rate": 3.5712322638843464e-05, "loss": 0.2043, "step": 16820 }, { "epoch": 1.74, "learning_rate": 3.5709906232304523e-05, "loss": 0.839, "step": 16821 }, { "epoch": 1.74, "learning_rate": 3.570748978732784e-05, "loss": 0.8073, "step": 16822 }, { "epoch": 1.74, "learning_rate": 3.5705073303929656e-05, "loss": 1.033, "step": 16823 }, { "epoch": 1.74, "learning_rate": 3.5702656782126255e-05, "loss": 1.6868, "step": 16824 }, { "epoch": 1.74, "learning_rate": 3.570024022193392e-05, "loss": 0.7695, "step": 16825 }, { "epoch": 1.74, "learning_rate": 3.569782362336888e-05, "loss": 1.1824, "step": 16826 }, { "epoch": 1.74, "learning_rate": 3.569540698644743e-05, "loss": 0.175, "step": 16827 }, { "epoch": 1.74, "learning_rate": 3.569299031118584e-05, "loss": 0.5004, "step": 16828 }, { "epoch": 1.74, "learning_rate": 3.5690573597600365e-05, "loss": 0.5169, "step": 16829 }, { "epoch": 1.74, "learning_rate": 3.568815684570729e-05, "loss": 0.811, "step": 16830 }, { "epoch": 1.74, "learning_rate": 3.5685740055522866e-05, "loss": 1.0984, "step": 16831 }, { "epoch": 1.74, "learning_rate": 3.5683323227063364e-05, "loss": 1.5954, "step": 16832 }, { "epoch": 1.74, "learning_rate": 3.5680906360345064e-05, "loss": 0.5135, "step": 16833 }, { "epoch": 1.74, "learning_rate": 3.5678489455384236e-05, "loss": 0.8333, "step": 16834 }, { "epoch": 1.74, "learning_rate": 3.567607251219715e-05, "loss": 0.2433, "step": 16835 }, { "epoch": 1.74, "learning_rate": 3.567365553080006e-05, "loss": 0.2985, "step": 16836 }, { "epoch": 1.74, "learning_rate": 3.567123851120926e-05, "loss": 0.8848, "step": 16837 }, { "epoch": 1.74, "learning_rate": 3.5668821453441e-05, "loss": 0.9934, "step": 16838 }, { "epoch": 1.74, "learning_rate": 3.5666404357511566e-05, "loss": 0.8479, "step": 16839 }, { "epoch": 1.74, "learning_rate": 3.5663987223437216e-05, "loss": 0.5968, "step": 16840 }, { "epoch": 1.74, "learning_rate": 3.566157005123424e-05, "loss": 0.3597, "step": 16841 }, { "epoch": 1.74, "learning_rate": 3.565915284091889e-05, "loss": 0.8442, "step": 16842 }, { "epoch": 1.74, "learning_rate": 3.565673559250745e-05, "loss": 0.2579, "step": 16843 }, { "epoch": 1.74, "learning_rate": 3.5654318306016196e-05, "loss": 0.196, "step": 16844 }, { "epoch": 1.74, "learning_rate": 3.565190098146139e-05, "loss": 0.3795, "step": 16845 }, { "epoch": 1.74, "learning_rate": 3.5649483618859305e-05, "loss": 0.1922, "step": 16846 }, { "epoch": 1.74, "learning_rate": 3.564706621822622e-05, "loss": 1.3597, "step": 16847 }, { "epoch": 1.74, "learning_rate": 3.564464877957841e-05, "loss": 0.3601, "step": 16848 }, { "epoch": 1.74, "learning_rate": 3.5642231302932145e-05, "loss": 0.8317, "step": 16849 }, { "epoch": 1.74, "learning_rate": 3.563981378830369e-05, "loss": 0.9834, "step": 16850 }, { "epoch": 1.74, "learning_rate": 3.5637396235709335e-05, "loss": 0.4287, "step": 16851 }, { "epoch": 1.74, "learning_rate": 3.5634978645165345e-05, "loss": 0.7331, "step": 16852 }, { "epoch": 1.74, "learning_rate": 3.563256101668801e-05, "loss": 0.476, "step": 16853 }, { "epoch": 1.74, "learning_rate": 3.563014335029358e-05, "loss": 0.4323, "step": 16854 }, { "epoch": 1.74, "learning_rate": 3.5627725645998344e-05, "loss": 0.9899, "step": 16855 }, { "epoch": 1.74, "learning_rate": 3.5625307903818577e-05, "loss": 0.3487, "step": 16856 }, { "epoch": 1.75, "learning_rate": 3.562289012377055e-05, "loss": 1.0238, "step": 16857 }, { "epoch": 1.75, "learning_rate": 3.562047230587054e-05, "loss": 0.1293, "step": 16858 }, { "epoch": 1.75, "learning_rate": 3.5618054450134834e-05, "loss": 0.3745, "step": 16859 }, { "epoch": 1.75, "learning_rate": 3.56156365565797e-05, "loss": 0.6699, "step": 16860 }, { "epoch": 1.75, "learning_rate": 3.561321862522141e-05, "loss": 0.8624, "step": 16861 }, { "epoch": 1.75, "learning_rate": 3.561080065607624e-05, "loss": 0.2267, "step": 16862 }, { "epoch": 1.75, "learning_rate": 3.560838264916049e-05, "loss": 0.9029, "step": 16863 }, { "epoch": 1.75, "learning_rate": 3.56059646044904e-05, "loss": 1.226, "step": 16864 }, { "epoch": 1.75, "learning_rate": 3.560354652208228e-05, "loss": 0.9297, "step": 16865 }, { "epoch": 1.75, "learning_rate": 3.5601128401952394e-05, "loss": 0.7308, "step": 16866 }, { "epoch": 1.75, "learning_rate": 3.5598710244117026e-05, "loss": 0.7881, "step": 16867 }, { "epoch": 1.75, "learning_rate": 3.5596292048592445e-05, "loss": 0.6338, "step": 16868 }, { "epoch": 1.75, "learning_rate": 3.5593873815394936e-05, "loss": 0.8011, "step": 16869 }, { "epoch": 1.75, "learning_rate": 3.5591455544540775e-05, "loss": 0.146, "step": 16870 }, { "epoch": 1.75, "learning_rate": 3.5589037236046245e-05, "loss": 0.1719, "step": 16871 }, { "epoch": 1.75, "learning_rate": 3.558661888992762e-05, "loss": 0.1726, "step": 16872 }, { "epoch": 1.75, "learning_rate": 3.55842005062012e-05, "loss": 1.1518, "step": 16873 }, { "epoch": 1.75, "learning_rate": 3.5581782084883235e-05, "loss": 0.7356, "step": 16874 }, { "epoch": 1.75, "learning_rate": 3.5579363625990015e-05, "loss": 0.8988, "step": 16875 }, { "epoch": 1.75, "learning_rate": 3.557694512953783e-05, "loss": 0.694, "step": 16876 }, { "epoch": 1.75, "learning_rate": 3.5574526595542964e-05, "loss": 0.4109, "step": 16877 }, { "epoch": 1.75, "learning_rate": 3.557210802402168e-05, "loss": 0.3795, "step": 16878 }, { "epoch": 1.75, "learning_rate": 3.5569689414990276e-05, "loss": 0.2759, "step": 16879 }, { "epoch": 1.75, "learning_rate": 3.556727076846502e-05, "loss": 0.3973, "step": 16880 }, { "epoch": 1.75, "learning_rate": 3.556485208446219e-05, "loss": 0.3251, "step": 16881 }, { "epoch": 1.75, "learning_rate": 3.5562433362998105e-05, "loss": 0.1729, "step": 16882 }, { "epoch": 1.75, "learning_rate": 3.5560014604089e-05, "loss": 0.6639, "step": 16883 }, { "epoch": 1.75, "learning_rate": 3.555759580775118e-05, "loss": 0.9302, "step": 16884 }, { "epoch": 1.75, "learning_rate": 3.555517697400093e-05, "loss": 0.3714, "step": 16885 }, { "epoch": 1.75, "learning_rate": 3.555275810285452e-05, "loss": 0.7324, "step": 16886 }, { "epoch": 1.75, "learning_rate": 3.5550339194328255e-05, "loss": 1.3271, "step": 16887 }, { "epoch": 1.75, "learning_rate": 3.5547920248438396e-05, "loss": 0.982, "step": 16888 }, { "epoch": 1.75, "learning_rate": 3.554550126520123e-05, "loss": 1.3, "step": 16889 }, { "epoch": 1.75, "learning_rate": 3.5543082244633055e-05, "loss": 0.4754, "step": 16890 }, { "epoch": 1.75, "learning_rate": 3.554066318675015e-05, "loss": 1.1132, "step": 16891 }, { "epoch": 1.75, "learning_rate": 3.553824409156879e-05, "loss": 0.3637, "step": 16892 }, { "epoch": 1.75, "learning_rate": 3.553582495910527e-05, "loss": 1.0611, "step": 16893 }, { "epoch": 1.75, "learning_rate": 3.553340578937587e-05, "loss": 0.5389, "step": 16894 }, { "epoch": 1.75, "learning_rate": 3.5530986582396874e-05, "loss": 0.1657, "step": 16895 }, { "epoch": 1.75, "learning_rate": 3.5528567338184586e-05, "loss": 0.6486, "step": 16896 }, { "epoch": 1.75, "learning_rate": 3.552614805675527e-05, "loss": 0.7749, "step": 16897 }, { "epoch": 1.75, "learning_rate": 3.55237287381252e-05, "loss": 0.6962, "step": 16898 }, { "epoch": 1.75, "learning_rate": 3.55213093823107e-05, "loss": 0.4022, "step": 16899 }, { "epoch": 1.75, "learning_rate": 3.5518889989328024e-05, "loss": 0.4524, "step": 16900 }, { "epoch": 1.75, "learning_rate": 3.551647055919348e-05, "loss": 0.4179, "step": 16901 }, { "epoch": 1.75, "learning_rate": 3.5514051091923355e-05, "loss": 0.994, "step": 16902 }, { "epoch": 1.75, "learning_rate": 3.551163158753391e-05, "loss": 0.3579, "step": 16903 }, { "epoch": 1.75, "learning_rate": 3.550921204604147e-05, "loss": 0.4077, "step": 16904 }, { "epoch": 1.75, "learning_rate": 3.5506792467462286e-05, "loss": 1.4191, "step": 16905 }, { "epoch": 1.75, "learning_rate": 3.550437285181268e-05, "loss": 0.9238, "step": 16906 }, { "epoch": 1.75, "learning_rate": 3.550195319910891e-05, "loss": 0.2303, "step": 16907 }, { "epoch": 1.75, "learning_rate": 3.549953350936728e-05, "loss": 0.1077, "step": 16908 }, { "epoch": 1.75, "learning_rate": 3.549711378260408e-05, "loss": 0.3142, "step": 16909 }, { "epoch": 1.75, "learning_rate": 3.5494694018835595e-05, "loss": 1.3098, "step": 16910 }, { "epoch": 1.75, "learning_rate": 3.549227421807813e-05, "loss": 0.1959, "step": 16911 }, { "epoch": 1.75, "learning_rate": 3.548985438034794e-05, "loss": 0.9284, "step": 16912 }, { "epoch": 1.75, "learning_rate": 3.548743450566134e-05, "loss": 0.3005, "step": 16913 }, { "epoch": 1.75, "learning_rate": 3.5485014594034616e-05, "loss": 0.6219, "step": 16914 }, { "epoch": 1.75, "learning_rate": 3.5482594645484054e-05, "loss": 0.4168, "step": 16915 }, { "epoch": 1.75, "learning_rate": 3.5480174660025964e-05, "loss": 0.1285, "step": 16916 }, { "epoch": 1.75, "learning_rate": 3.5477754637676606e-05, "loss": 0.2138, "step": 16917 }, { "epoch": 1.75, "learning_rate": 3.5475334578452286e-05, "loss": 0.2978, "step": 16918 }, { "epoch": 1.75, "learning_rate": 3.5472914482369294e-05, "loss": 1.2306, "step": 16919 }, { "epoch": 1.75, "learning_rate": 3.547049434944393e-05, "loss": 0.6331, "step": 16920 }, { "epoch": 1.75, "learning_rate": 3.5468074179692465e-05, "loss": 0.6117, "step": 16921 }, { "epoch": 1.75, "learning_rate": 3.5465653973131216e-05, "loss": 0.7244, "step": 16922 }, { "epoch": 1.75, "learning_rate": 3.546323372977647e-05, "loss": 0.3037, "step": 16923 }, { "epoch": 1.75, "learning_rate": 3.54608134496445e-05, "loss": 0.5889, "step": 16924 }, { "epoch": 1.75, "learning_rate": 3.545839313275161e-05, "loss": 0.5104, "step": 16925 }, { "epoch": 1.75, "learning_rate": 3.5455972779114096e-05, "loss": 0.7375, "step": 16926 }, { "epoch": 1.75, "learning_rate": 3.545355238874825e-05, "loss": 0.3697, "step": 16927 }, { "epoch": 1.75, "learning_rate": 3.545113196167037e-05, "loss": 1.0415, "step": 16928 }, { "epoch": 1.75, "learning_rate": 3.544871149789674e-05, "loss": 0.8941, "step": 16929 }, { "epoch": 1.75, "learning_rate": 3.544629099744367e-05, "loss": 0.4078, "step": 16930 }, { "epoch": 1.75, "learning_rate": 3.5443870460327434e-05, "loss": 0.3425, "step": 16931 }, { "epoch": 1.75, "learning_rate": 3.544144988656433e-05, "loss": 0.6551, "step": 16932 }, { "epoch": 1.75, "learning_rate": 3.5439029276170664e-05, "loss": 0.3883, "step": 16933 }, { "epoch": 1.75, "learning_rate": 3.5436608629162725e-05, "loss": 0.4123, "step": 16934 }, { "epoch": 1.75, "learning_rate": 3.543418794555682e-05, "loss": 0.659, "step": 16935 }, { "epoch": 1.75, "learning_rate": 3.543176722536923e-05, "loss": 0.7043, "step": 16936 }, { "epoch": 1.75, "learning_rate": 3.542934646861624e-05, "loss": 0.2343, "step": 16937 }, { "epoch": 1.75, "learning_rate": 3.542692567531416e-05, "loss": 1.0655, "step": 16938 }, { "epoch": 1.75, "learning_rate": 3.54245048454793e-05, "loss": 0.6973, "step": 16939 }, { "epoch": 1.75, "learning_rate": 3.542208397912793e-05, "loss": 0.8314, "step": 16940 }, { "epoch": 1.75, "learning_rate": 3.5419663076276375e-05, "loss": 0.9766, "step": 16941 }, { "epoch": 1.75, "learning_rate": 3.5417242136940906e-05, "loss": 1.0904, "step": 16942 }, { "epoch": 1.75, "learning_rate": 3.541482116113783e-05, "loss": 0.8285, "step": 16943 }, { "epoch": 1.75, "learning_rate": 3.5412400148883455e-05, "loss": 0.7292, "step": 16944 }, { "epoch": 1.75, "learning_rate": 3.5409979100194054e-05, "loss": 0.4736, "step": 16945 }, { "epoch": 1.75, "learning_rate": 3.540755801508594e-05, "loss": 0.6887, "step": 16946 }, { "epoch": 1.75, "learning_rate": 3.540513689357542e-05, "loss": 0.4514, "step": 16947 }, { "epoch": 1.75, "learning_rate": 3.5402715735678784e-05, "loss": 0.6352, "step": 16948 }, { "epoch": 1.75, "learning_rate": 3.5400294541412324e-05, "loss": 0.8228, "step": 16949 }, { "epoch": 1.75, "learning_rate": 3.539787331079235e-05, "loss": 0.5087, "step": 16950 }, { "epoch": 1.75, "learning_rate": 3.5395452043835146e-05, "loss": 0.1407, "step": 16951 }, { "epoch": 1.75, "learning_rate": 3.539303074055702e-05, "loss": 0.3648, "step": 16952 }, { "epoch": 1.76, "learning_rate": 3.539060940097428e-05, "loss": 0.9871, "step": 16953 }, { "epoch": 1.76, "learning_rate": 3.538818802510323e-05, "loss": 1.0373, "step": 16954 }, { "epoch": 1.76, "learning_rate": 3.5385766612960145e-05, "loss": 0.7479, "step": 16955 }, { "epoch": 1.76, "learning_rate": 3.538334516456134e-05, "loss": 1.3281, "step": 16956 }, { "epoch": 1.76, "learning_rate": 3.538092367992311e-05, "loss": 0.2067, "step": 16957 }, { "epoch": 1.76, "learning_rate": 3.537850215906178e-05, "loss": 0.6675, "step": 16958 }, { "epoch": 1.76, "learning_rate": 3.537608060199362e-05, "loss": 1.5104, "step": 16959 }, { "epoch": 1.76, "learning_rate": 3.537365900873495e-05, "loss": 0.6335, "step": 16960 }, { "epoch": 1.76, "learning_rate": 3.537123737930205e-05, "loss": 0.31, "step": 16961 }, { "epoch": 1.76, "learning_rate": 3.536881571371124e-05, "loss": 0.6228, "step": 16962 }, { "epoch": 1.76, "learning_rate": 3.536639401197883e-05, "loss": 1.4451, "step": 16963 }, { "epoch": 1.76, "learning_rate": 3.5363972274121104e-05, "loss": 1.205, "step": 16964 }, { "epoch": 1.76, "learning_rate": 3.536155050015437e-05, "loss": 0.2927, "step": 16965 }, { "epoch": 1.76, "learning_rate": 3.535912869009493e-05, "loss": 0.4381, "step": 16966 }, { "epoch": 1.76, "learning_rate": 3.53567068439591e-05, "loss": 0.205, "step": 16967 }, { "epoch": 1.76, "learning_rate": 3.535428496176317e-05, "loss": 0.566, "step": 16968 }, { "epoch": 1.76, "learning_rate": 3.535186304352345e-05, "loss": 0.5635, "step": 16969 }, { "epoch": 1.76, "learning_rate": 3.534944108925622e-05, "loss": 0.2036, "step": 16970 }, { "epoch": 1.76, "learning_rate": 3.5347019098977823e-05, "loss": 0.6892, "step": 16971 }, { "epoch": 1.76, "learning_rate": 3.5344597072704537e-05, "loss": 0.1044, "step": 16972 }, { "epoch": 1.76, "learning_rate": 3.534217501045269e-05, "loss": 0.4381, "step": 16973 }, { "epoch": 1.76, "learning_rate": 3.533975291223855e-05, "loss": 0.5959, "step": 16974 }, { "epoch": 1.76, "learning_rate": 3.5337330778078455e-05, "loss": 0.4207, "step": 16975 }, { "epoch": 1.76, "learning_rate": 3.533490860798869e-05, "loss": 0.4834, "step": 16976 }, { "epoch": 1.76, "learning_rate": 3.533248640198557e-05, "loss": 0.5639, "step": 16977 }, { "epoch": 1.76, "learning_rate": 3.53300641600854e-05, "loss": 0.6854, "step": 16978 }, { "epoch": 1.76, "learning_rate": 3.532764188230449e-05, "loss": 0.7304, "step": 16979 }, { "epoch": 1.76, "learning_rate": 3.5325219568659144e-05, "loss": 1.1504, "step": 16980 }, { "epoch": 1.76, "learning_rate": 3.5322797219165654e-05, "loss": 1.0418, "step": 16981 }, { "epoch": 1.76, "learning_rate": 3.5320374833840345e-05, "loss": 0.8435, "step": 16982 }, { "epoch": 1.76, "learning_rate": 3.531795241269952e-05, "loss": 0.1688, "step": 16983 }, { "epoch": 1.76, "learning_rate": 3.5315529955759476e-05, "loss": 0.9475, "step": 16984 }, { "epoch": 1.76, "learning_rate": 3.5313107463036535e-05, "loss": 0.1736, "step": 16985 }, { "epoch": 1.76, "learning_rate": 3.531068493454699e-05, "loss": 0.1817, "step": 16986 }, { "epoch": 1.76, "learning_rate": 3.5308262370307177e-05, "loss": 0.4689, "step": 16987 }, { "epoch": 1.76, "learning_rate": 3.530583977033336e-05, "loss": 0.9315, "step": 16988 }, { "epoch": 1.76, "learning_rate": 3.5303417134641875e-05, "loss": 0.6264, "step": 16989 }, { "epoch": 1.76, "learning_rate": 3.530099446324903e-05, "loss": 0.1026, "step": 16990 }, { "epoch": 1.76, "learning_rate": 3.529857175617113e-05, "loss": 0.5523, "step": 16991 }, { "epoch": 1.76, "learning_rate": 3.5296149013424506e-05, "loss": 0.4577, "step": 16992 }, { "epoch": 1.76, "learning_rate": 3.529372623502542e-05, "loss": 1.0851, "step": 16993 }, { "epoch": 1.76, "learning_rate": 3.529130342099021e-05, "loss": 0.4705, "step": 16994 }, { "epoch": 1.76, "learning_rate": 3.528888057133519e-05, "loss": 0.265, "step": 16995 }, { "epoch": 1.76, "learning_rate": 3.5286457686076654e-05, "loss": 0.5287, "step": 16996 }, { "epoch": 1.76, "learning_rate": 3.528403476523094e-05, "loss": 0.583, "step": 16997 }, { "epoch": 1.76, "learning_rate": 3.5281611808814334e-05, "loss": 0.6558, "step": 16998 }, { "epoch": 1.76, "learning_rate": 3.527918881684315e-05, "loss": 0.9983, "step": 16999 }, { "epoch": 1.76, "learning_rate": 3.527676578933369e-05, "loss": 0.0206, "step": 17000 }, { "epoch": 1.76, "learning_rate": 3.5274342726302296e-05, "loss": 1.0518, "step": 17001 }, { "epoch": 1.76, "learning_rate": 3.527191962776525e-05, "loss": 0.6331, "step": 17002 }, { "epoch": 1.76, "learning_rate": 3.5269496493738874e-05, "loss": 1.1072, "step": 17003 }, { "epoch": 1.76, "learning_rate": 3.5267073324239495e-05, "loss": 0.8125, "step": 17004 }, { "epoch": 1.76, "learning_rate": 3.526465011928339e-05, "loss": 0.5037, "step": 17005 }, { "epoch": 1.76, "learning_rate": 3.526222687888691e-05, "loss": 0.9906, "step": 17006 }, { "epoch": 1.76, "learning_rate": 3.5259803603066336e-05, "loss": 1.2057, "step": 17007 }, { "epoch": 1.76, "learning_rate": 3.5257380291838e-05, "loss": 0.8511, "step": 17008 }, { "epoch": 1.76, "learning_rate": 3.5254956945218215e-05, "loss": 0.5114, "step": 17009 }, { "epoch": 1.76, "learning_rate": 3.525253356322329e-05, "loss": 0.417, "step": 17010 }, { "epoch": 1.76, "learning_rate": 3.525011014586953e-05, "loss": 0.9919, "step": 17011 }, { "epoch": 1.76, "learning_rate": 3.524768669317326e-05, "loss": 0.1412, "step": 17012 }, { "epoch": 1.76, "learning_rate": 3.524526320515079e-05, "loss": 1.1021, "step": 17013 }, { "epoch": 1.76, "learning_rate": 3.524283968181844e-05, "loss": 0.2737, "step": 17014 }, { "epoch": 1.76, "learning_rate": 3.524041612319251e-05, "loss": 0.9365, "step": 17015 }, { "epoch": 1.76, "learning_rate": 3.523799252928934e-05, "loss": 0.9257, "step": 17016 }, { "epoch": 1.76, "learning_rate": 3.523556890012523e-05, "loss": 0.4984, "step": 17017 }, { "epoch": 1.76, "learning_rate": 3.523314523571648e-05, "loss": 0.2957, "step": 17018 }, { "epoch": 1.76, "learning_rate": 3.523072153607942e-05, "loss": 1.125, "step": 17019 }, { "epoch": 1.76, "learning_rate": 3.5228297801230374e-05, "loss": 0.4981, "step": 17020 }, { "epoch": 1.76, "learning_rate": 3.522587403118566e-05, "loss": 0.949, "step": 17021 }, { "epoch": 1.76, "learning_rate": 3.522345022596157e-05, "loss": 0.8301, "step": 17022 }, { "epoch": 1.76, "learning_rate": 3.5221026385574444e-05, "loss": 0.6834, "step": 17023 }, { "epoch": 1.76, "learning_rate": 3.5218602510040585e-05, "loss": 1.1585, "step": 17024 }, { "epoch": 1.76, "learning_rate": 3.521617859937632e-05, "loss": 1.0692, "step": 17025 }, { "epoch": 1.76, "learning_rate": 3.5213754653597957e-05, "loss": 0.2312, "step": 17026 }, { "epoch": 1.76, "learning_rate": 3.521133067272181e-05, "loss": 0.9388, "step": 17027 }, { "epoch": 1.76, "learning_rate": 3.520890665676421e-05, "loss": 1.3252, "step": 17028 }, { "epoch": 1.76, "learning_rate": 3.5206482605741484e-05, "loss": 1.0451, "step": 17029 }, { "epoch": 1.76, "learning_rate": 3.520405851966992e-05, "loss": 0.3424, "step": 17030 }, { "epoch": 1.76, "learning_rate": 3.5201634398565856e-05, "loss": 0.5424, "step": 17031 }, { "epoch": 1.76, "learning_rate": 3.51992102424456e-05, "loss": 0.5464, "step": 17032 }, { "epoch": 1.76, "learning_rate": 3.5196786051325484e-05, "loss": 1.1757, "step": 17033 }, { "epoch": 1.76, "learning_rate": 3.519436182522181e-05, "loss": 0.1932, "step": 17034 }, { "epoch": 1.76, "learning_rate": 3.5191937564150925e-05, "loss": 0.4864, "step": 17035 }, { "epoch": 1.76, "learning_rate": 3.518951326812913e-05, "loss": 1.9686, "step": 17036 }, { "epoch": 1.76, "learning_rate": 3.518708893717274e-05, "loss": 0.4377, "step": 17037 }, { "epoch": 1.76, "learning_rate": 3.518466457129807e-05, "loss": 0.2405, "step": 17038 }, { "epoch": 1.76, "learning_rate": 3.518224017052146e-05, "loss": 1.1969, "step": 17039 }, { "epoch": 1.76, "learning_rate": 3.517981573485922e-05, "loss": 0.5791, "step": 17040 }, { "epoch": 1.76, "learning_rate": 3.517739126432767e-05, "loss": 0.7677, "step": 17041 }, { "epoch": 1.76, "learning_rate": 3.517496675894315e-05, "loss": 0.4709, "step": 17042 }, { "epoch": 1.76, "learning_rate": 3.517254221872195e-05, "loss": 0.9647, "step": 17043 }, { "epoch": 1.76, "learning_rate": 3.517011764368041e-05, "loss": 0.2928, "step": 17044 }, { "epoch": 1.76, "learning_rate": 3.5167693033834846e-05, "loss": 0.2868, "step": 17045 }, { "epoch": 1.76, "learning_rate": 3.516526838920158e-05, "loss": 1.3446, "step": 17046 }, { "epoch": 1.76, "learning_rate": 3.516284370979693e-05, "loss": 0.9412, "step": 17047 }, { "epoch": 1.77, "learning_rate": 3.516041899563725e-05, "loss": 1.093, "step": 17048 }, { "epoch": 1.77, "learning_rate": 3.515799424673881e-05, "loss": 0.6543, "step": 17049 }, { "epoch": 1.77, "learning_rate": 3.515556946311797e-05, "loss": 0.736, "step": 17050 }, { "epoch": 1.77, "learning_rate": 3.515314464479104e-05, "loss": 0.4665, "step": 17051 }, { "epoch": 1.77, "learning_rate": 3.515071979177434e-05, "loss": 0.4082, "step": 17052 }, { "epoch": 1.77, "learning_rate": 3.514829490408421e-05, "loss": 0.3207, "step": 17053 }, { "epoch": 1.77, "learning_rate": 3.514586998173697e-05, "loss": 1.0088, "step": 17054 }, { "epoch": 1.77, "learning_rate": 3.514344502474892e-05, "loss": 0.5287, "step": 17055 }, { "epoch": 1.77, "learning_rate": 3.514102003313641e-05, "loss": 0.6203, "step": 17056 }, { "epoch": 1.77, "learning_rate": 3.5138595006915755e-05, "loss": 1.0471, "step": 17057 }, { "epoch": 1.77, "learning_rate": 3.5136169946103275e-05, "loss": 0.7555, "step": 17058 }, { "epoch": 1.77, "learning_rate": 3.513374485071531e-05, "loss": 0.4528, "step": 17059 }, { "epoch": 1.77, "learning_rate": 3.513131972076817e-05, "loss": 0.6926, "step": 17060 }, { "epoch": 1.77, "learning_rate": 3.5128894556278194e-05, "loss": 0.6622, "step": 17061 }, { "epoch": 1.77, "learning_rate": 3.512646935726169e-05, "loss": 1.0731, "step": 17062 }, { "epoch": 1.77, "learning_rate": 3.5124044123735e-05, "loss": 0.9216, "step": 17063 }, { "epoch": 1.77, "learning_rate": 3.512161885571444e-05, "loss": 1.0299, "step": 17064 }, { "epoch": 1.77, "learning_rate": 3.511919355321634e-05, "loss": 0.5526, "step": 17065 }, { "epoch": 1.77, "learning_rate": 3.5116768216257034e-05, "loss": 0.7435, "step": 17066 }, { "epoch": 1.77, "learning_rate": 3.511434284485284e-05, "loss": 0.6584, "step": 17067 }, { "epoch": 1.77, "learning_rate": 3.511191743902008e-05, "loss": 0.8049, "step": 17068 }, { "epoch": 1.77, "learning_rate": 3.5109491998775093e-05, "loss": 0.2907, "step": 17069 }, { "epoch": 1.77, "learning_rate": 3.510706652413419e-05, "loss": 0.6285, "step": 17070 }, { "epoch": 1.77, "learning_rate": 3.510464101511372e-05, "loss": 0.41, "step": 17071 }, { "epoch": 1.77, "learning_rate": 3.510221547173e-05, "loss": 0.4508, "step": 17072 }, { "epoch": 1.77, "learning_rate": 3.5099789893999375e-05, "loss": 0.2741, "step": 17073 }, { "epoch": 1.77, "learning_rate": 3.509736428193814e-05, "loss": 0.9003, "step": 17074 }, { "epoch": 1.77, "learning_rate": 3.509493863556264e-05, "loss": 0.3979, "step": 17075 }, { "epoch": 1.77, "learning_rate": 3.5092512954889206e-05, "loss": 1.091, "step": 17076 }, { "epoch": 1.77, "learning_rate": 3.509008723993417e-05, "loss": 0.8289, "step": 17077 }, { "epoch": 1.77, "learning_rate": 3.508766149071386e-05, "loss": 0.2298, "step": 17078 }, { "epoch": 1.77, "learning_rate": 3.5085235707244596e-05, "loss": 1.0703, "step": 17079 }, { "epoch": 1.77, "learning_rate": 3.5082809889542726e-05, "loss": 0.2638, "step": 17080 }, { "epoch": 1.77, "learning_rate": 3.508038403762455e-05, "loss": 0.2898, "step": 17081 }, { "epoch": 1.77, "learning_rate": 3.5077958151506436e-05, "loss": 1.1053, "step": 17082 }, { "epoch": 1.77, "learning_rate": 3.50755322312047e-05, "loss": 0.4817, "step": 17083 }, { "epoch": 1.77, "learning_rate": 3.507310627673565e-05, "loss": 0.2126, "step": 17084 }, { "epoch": 1.77, "learning_rate": 3.507068028811564e-05, "loss": 0.5355, "step": 17085 }, { "epoch": 1.77, "learning_rate": 3.506825426536101e-05, "loss": 0.7071, "step": 17086 }, { "epoch": 1.77, "learning_rate": 3.5065828208488075e-05, "loss": 0.6069, "step": 17087 }, { "epoch": 1.77, "learning_rate": 3.5063402117513156e-05, "loss": 0.9999, "step": 17088 }, { "epoch": 1.77, "learning_rate": 3.506097599245261e-05, "loss": 0.4496, "step": 17089 }, { "epoch": 1.77, "learning_rate": 3.505854983332275e-05, "loss": 0.6928, "step": 17090 }, { "epoch": 1.77, "learning_rate": 3.505612364013992e-05, "loss": 0.3968, "step": 17091 }, { "epoch": 1.77, "learning_rate": 3.505369741292045e-05, "loss": 0.81, "step": 17092 }, { "epoch": 1.77, "learning_rate": 3.505127115168067e-05, "loss": 1.0887, "step": 17093 }, { "epoch": 1.77, "learning_rate": 3.504884485643691e-05, "loss": 0.8756, "step": 17094 }, { "epoch": 1.77, "learning_rate": 3.504641852720552e-05, "loss": 0.93, "step": 17095 }, { "epoch": 1.77, "learning_rate": 3.504399216400281e-05, "loss": 0.2078, "step": 17096 }, { "epoch": 1.77, "learning_rate": 3.504156576684513e-05, "loss": 0.5211, "step": 17097 }, { "epoch": 1.77, "learning_rate": 3.503913933574881e-05, "loss": 0.8395, "step": 17098 }, { "epoch": 1.77, "learning_rate": 3.503671287073019e-05, "loss": 0.5678, "step": 17099 }, { "epoch": 1.77, "learning_rate": 3.503428637180557e-05, "loss": 1.0136, "step": 17100 }, { "epoch": 1.77, "learning_rate": 3.503185983899133e-05, "loss": 0.6244, "step": 17101 }, { "epoch": 1.77, "learning_rate": 3.50294332723038e-05, "loss": 0.5429, "step": 17102 }, { "epoch": 1.77, "learning_rate": 3.5027006671759284e-05, "loss": 0.5357, "step": 17103 }, { "epoch": 1.77, "learning_rate": 3.502458003737414e-05, "loss": 0.7506, "step": 17104 }, { "epoch": 1.77, "learning_rate": 3.5022153369164703e-05, "loss": 0.3184, "step": 17105 }, { "epoch": 1.77, "learning_rate": 3.50197266671473e-05, "loss": 0.2685, "step": 17106 }, { "epoch": 1.77, "learning_rate": 3.501729993133827e-05, "loss": 0.2871, "step": 17107 }, { "epoch": 1.77, "learning_rate": 3.5014873161753954e-05, "loss": 0.3766, "step": 17108 }, { "epoch": 1.77, "learning_rate": 3.5012446358410685e-05, "loss": 0.8796, "step": 17109 }, { "epoch": 1.77, "learning_rate": 3.50100195213248e-05, "loss": 0.7515, "step": 17110 }, { "epoch": 1.77, "learning_rate": 3.500759265051264e-05, "loss": 0.158, "step": 17111 }, { "epoch": 1.77, "learning_rate": 3.5005165745990535e-05, "loss": 0.9022, "step": 17112 }, { "epoch": 1.77, "learning_rate": 3.500273880777482e-05, "loss": 0.2419, "step": 17113 }, { "epoch": 1.77, "learning_rate": 3.500031183588184e-05, "loss": 0.4983, "step": 17114 }, { "epoch": 1.77, "learning_rate": 3.499788483032793e-05, "loss": 0.4714, "step": 17115 }, { "epoch": 1.77, "learning_rate": 3.499545779112943e-05, "loss": 0.3679, "step": 17116 }, { "epoch": 1.77, "learning_rate": 3.499303071830269e-05, "loss": 0.6419, "step": 17117 }, { "epoch": 1.77, "learning_rate": 3.4990603611864026e-05, "loss": 0.8989, "step": 17118 }, { "epoch": 1.77, "learning_rate": 3.4988176471829775e-05, "loss": 0.5589, "step": 17119 }, { "epoch": 1.77, "learning_rate": 3.498574929821629e-05, "loss": 0.83, "step": 17120 }, { "epoch": 1.77, "learning_rate": 3.498332209103992e-05, "loss": 0.8303, "step": 17121 }, { "epoch": 1.77, "learning_rate": 3.498089485031698e-05, "loss": 1.1348, "step": 17122 }, { "epoch": 1.77, "learning_rate": 3.497846757606383e-05, "loss": 0.543, "step": 17123 }, { "epoch": 1.77, "learning_rate": 3.49760402682968e-05, "loss": 1.2387, "step": 17124 }, { "epoch": 1.77, "learning_rate": 3.497361292703223e-05, "loss": 0.6072, "step": 17125 }, { "epoch": 1.77, "learning_rate": 3.497118555228646e-05, "loss": 0.2789, "step": 17126 }, { "epoch": 1.77, "learning_rate": 3.4968758144075833e-05, "loss": 0.4019, "step": 17127 }, { "epoch": 1.77, "learning_rate": 3.496633070241669e-05, "loss": 0.8419, "step": 17128 }, { "epoch": 1.77, "learning_rate": 3.496390322732536e-05, "loss": 0.5614, "step": 17129 }, { "epoch": 1.77, "learning_rate": 3.496147571881821e-05, "loss": 1.1511, "step": 17130 }, { "epoch": 1.77, "learning_rate": 3.495904817691156e-05, "loss": 1.0581, "step": 17131 }, { "epoch": 1.77, "learning_rate": 3.495662060162176e-05, "loss": 0.2244, "step": 17132 }, { "epoch": 1.77, "learning_rate": 3.4954192992965146e-05, "loss": 0.6642, "step": 17133 }, { "epoch": 1.77, "learning_rate": 3.4951765350958056e-05, "loss": 0.0969, "step": 17134 }, { "epoch": 1.77, "learning_rate": 3.494933767561686e-05, "loss": 0.0684, "step": 17135 }, { "epoch": 1.77, "learning_rate": 3.494690996695787e-05, "loss": 1.392, "step": 17136 }, { "epoch": 1.77, "learning_rate": 3.4944482224997444e-05, "loss": 1.2325, "step": 17137 }, { "epoch": 1.77, "learning_rate": 3.494205444975191e-05, "loss": 0.7716, "step": 17138 }, { "epoch": 1.77, "learning_rate": 3.493962664123763e-05, "loss": 0.53, "step": 17139 }, { "epoch": 1.77, "learning_rate": 3.493719879947094e-05, "loss": 0.6641, "step": 17140 }, { "epoch": 1.77, "learning_rate": 3.493477092446818e-05, "loss": 0.2293, "step": 17141 }, { "epoch": 1.77, "learning_rate": 3.493234301624571e-05, "loss": 0.2804, "step": 17142 }, { "epoch": 1.77, "learning_rate": 3.492991507481984e-05, "loss": 0.9868, "step": 17143 }, { "epoch": 1.78, "learning_rate": 3.4927487100206944e-05, "loss": 0.3814, "step": 17144 }, { "epoch": 1.78, "learning_rate": 3.492505909242336e-05, "loss": 0.3461, "step": 17145 }, { "epoch": 1.78, "learning_rate": 3.492263105148542e-05, "loss": 0.1245, "step": 17146 }, { "epoch": 1.78, "learning_rate": 3.4920202977409495e-05, "loss": 0.3556, "step": 17147 }, { "epoch": 1.78, "learning_rate": 3.49177748702119e-05, "loss": 1.0385, "step": 17148 }, { "epoch": 1.78, "learning_rate": 3.491534672990901e-05, "loss": 0.6105, "step": 17149 }, { "epoch": 1.78, "learning_rate": 3.491291855651715e-05, "loss": 0.7418, "step": 17150 }, { "epoch": 1.78, "learning_rate": 3.4910490350052666e-05, "loss": 1.0786, "step": 17151 }, { "epoch": 1.78, "learning_rate": 3.490806211053192e-05, "loss": 0.9127, "step": 17152 }, { "epoch": 1.78, "learning_rate": 3.4905633837971237e-05, "loss": 0.8009, "step": 17153 }, { "epoch": 1.78, "learning_rate": 3.490320553238698e-05, "loss": 1.2296, "step": 17154 }, { "epoch": 1.78, "learning_rate": 3.49007771937955e-05, "loss": 0.3366, "step": 17155 }, { "epoch": 1.78, "learning_rate": 3.489834882221312e-05, "loss": 0.7675, "step": 17156 }, { "epoch": 1.78, "learning_rate": 3.4895920417656205e-05, "loss": 0.4361, "step": 17157 }, { "epoch": 1.78, "learning_rate": 3.48934919801411e-05, "loss": 0.7784, "step": 17158 }, { "epoch": 1.78, "learning_rate": 3.489106350968416e-05, "loss": 0.4869, "step": 17159 }, { "epoch": 1.78, "learning_rate": 3.4888635006301714e-05, "loss": 1.0376, "step": 17160 }, { "epoch": 1.78, "learning_rate": 3.4886206470010134e-05, "loss": 0.2903, "step": 17161 }, { "epoch": 1.78, "learning_rate": 3.488377790082574e-05, "loss": 0.9229, "step": 17162 }, { "epoch": 1.78, "learning_rate": 3.48813492987649e-05, "loss": 0.5954, "step": 17163 }, { "epoch": 1.78, "learning_rate": 3.487892066384397e-05, "loss": 0.4652, "step": 17164 }, { "epoch": 1.78, "learning_rate": 3.4876491996079276e-05, "loss": 0.6194, "step": 17165 }, { "epoch": 1.78, "learning_rate": 3.4874063295487184e-05, "loss": 0.3543, "step": 17166 }, { "epoch": 1.78, "learning_rate": 3.487163456208403e-05, "loss": 1.0039, "step": 17167 }, { "epoch": 1.78, "learning_rate": 3.4869205795886186e-05, "loss": 0.8601, "step": 17168 }, { "epoch": 1.78, "learning_rate": 3.486677699690998e-05, "loss": 0.5655, "step": 17169 }, { "epoch": 1.78, "learning_rate": 3.486434816517177e-05, "loss": 0.7788, "step": 17170 }, { "epoch": 1.78, "learning_rate": 3.4861919300687905e-05, "loss": 0.3274, "step": 17171 }, { "epoch": 1.78, "learning_rate": 3.485949040347473e-05, "loss": 1.3651, "step": 17172 }, { "epoch": 1.78, "learning_rate": 3.4857061473548625e-05, "loss": 1.5992, "step": 17173 }, { "epoch": 1.78, "learning_rate": 3.485463251092591e-05, "loss": 1.1088, "step": 17174 }, { "epoch": 1.78, "learning_rate": 3.485220351562293e-05, "loss": 0.1391, "step": 17175 }, { "epoch": 1.78, "learning_rate": 3.484977448765606e-05, "loss": 0.7879, "step": 17176 }, { "epoch": 1.78, "learning_rate": 3.4847345427041646e-05, "loss": 0.9491, "step": 17177 }, { "epoch": 1.78, "learning_rate": 3.484491633379603e-05, "loss": 0.7807, "step": 17178 }, { "epoch": 1.78, "learning_rate": 3.4842487207935586e-05, "loss": 0.9589, "step": 17179 }, { "epoch": 1.78, "learning_rate": 3.484005804947664e-05, "loss": 0.28, "step": 17180 }, { "epoch": 1.78, "learning_rate": 3.483762885843555e-05, "loss": 0.4363, "step": 17181 }, { "epoch": 1.78, "learning_rate": 3.483519963482868e-05, "loss": 0.9674, "step": 17182 }, { "epoch": 1.78, "learning_rate": 3.483277037867238e-05, "loss": 0.8223, "step": 17183 }, { "epoch": 1.78, "learning_rate": 3.4830341089983e-05, "loss": 0.1743, "step": 17184 }, { "epoch": 1.78, "learning_rate": 3.482791176877689e-05, "loss": 0.4212, "step": 17185 }, { "epoch": 1.78, "learning_rate": 3.4825482415070415e-05, "loss": 0.3776, "step": 17186 }, { "epoch": 1.78, "learning_rate": 3.4823053028879914e-05, "loss": 0.8532, "step": 17187 }, { "epoch": 1.78, "learning_rate": 3.482062361022176e-05, "loss": 0.4475, "step": 17188 }, { "epoch": 1.78, "learning_rate": 3.481819415911228e-05, "loss": 0.5449, "step": 17189 }, { "epoch": 1.78, "learning_rate": 3.481576467556785e-05, "loss": 0.3044, "step": 17190 }, { "epoch": 1.78, "learning_rate": 3.481333515960482e-05, "loss": 1.1904, "step": 17191 }, { "epoch": 1.78, "learning_rate": 3.481090561123955e-05, "loss": 0.302, "step": 17192 }, { "epoch": 1.78, "learning_rate": 3.480847603048839e-05, "loss": 0.2571, "step": 17193 }, { "epoch": 1.78, "learning_rate": 3.480604641736768e-05, "loss": 0.5367, "step": 17194 }, { "epoch": 1.78, "learning_rate": 3.48036167718938e-05, "loss": 0.5245, "step": 17195 }, { "epoch": 1.78, "learning_rate": 3.480118709408309e-05, "loss": 0.776, "step": 17196 }, { "epoch": 1.78, "learning_rate": 3.479875738395192e-05, "loss": 0.9491, "step": 17197 }, { "epoch": 1.78, "learning_rate": 3.479632764151663e-05, "loss": 1.2501, "step": 17198 }, { "epoch": 1.78, "learning_rate": 3.479389786679359e-05, "loss": 0.5471, "step": 17199 }, { "epoch": 1.78, "learning_rate": 3.479146805979915e-05, "loss": 0.4091, "step": 17200 }, { "epoch": 1.78, "learning_rate": 3.4789038220549665e-05, "loss": 0.6423, "step": 17201 }, { "epoch": 1.78, "learning_rate": 3.47866083490615e-05, "loss": 1.186, "step": 17202 }, { "epoch": 1.78, "learning_rate": 3.4784178445351e-05, "loss": 0.2465, "step": 17203 }, { "epoch": 1.78, "learning_rate": 3.478174850943454e-05, "loss": 0.6312, "step": 17204 }, { "epoch": 1.78, "learning_rate": 3.477931854132847e-05, "loss": 0.7867, "step": 17205 }, { "epoch": 1.78, "learning_rate": 3.4776888541049135e-05, "loss": 0.9506, "step": 17206 }, { "epoch": 1.78, "learning_rate": 3.477445850861291e-05, "loss": 0.5634, "step": 17207 }, { "epoch": 1.78, "learning_rate": 3.477202844403615e-05, "loss": 0.4762, "step": 17208 }, { "epoch": 1.78, "learning_rate": 3.47695983473352e-05, "loss": 0.5814, "step": 17209 }, { "epoch": 1.78, "learning_rate": 3.476716821852643e-05, "loss": 1.0026, "step": 17210 }, { "epoch": 1.78, "learning_rate": 3.476473805762621e-05, "loss": 0.2131, "step": 17211 }, { "epoch": 1.78, "learning_rate": 3.47623078646509e-05, "loss": 0.7699, "step": 17212 }, { "epoch": 1.78, "learning_rate": 3.475987763961682e-05, "loss": 0.2156, "step": 17213 }, { "epoch": 1.78, "learning_rate": 3.475744738254037e-05, "loss": 0.3917, "step": 17214 }, { "epoch": 1.78, "learning_rate": 3.475501709343789e-05, "loss": 0.9799, "step": 17215 }, { "epoch": 1.78, "learning_rate": 3.4752586772325764e-05, "loss": 0.2418, "step": 17216 }, { "epoch": 1.78, "learning_rate": 3.475015641922032e-05, "loss": 0.2733, "step": 17217 }, { "epoch": 1.78, "learning_rate": 3.4747726034137946e-05, "loss": 0.5651, "step": 17218 }, { "epoch": 1.78, "learning_rate": 3.474529561709498e-05, "loss": 0.5282, "step": 17219 }, { "epoch": 1.78, "learning_rate": 3.47428651681078e-05, "loss": 0.6904, "step": 17220 }, { "epoch": 1.78, "learning_rate": 3.4740434687192765e-05, "loss": 0.5846, "step": 17221 }, { "epoch": 1.78, "learning_rate": 3.4738004174366234e-05, "loss": 1.3054, "step": 17222 }, { "epoch": 1.78, "learning_rate": 3.4735573629644556e-05, "loss": 0.9521, "step": 17223 }, { "epoch": 1.78, "learning_rate": 3.4733143053044116e-05, "loss": 0.3405, "step": 17224 }, { "epoch": 1.78, "learning_rate": 3.473071244458126e-05, "loss": 0.4767, "step": 17225 }, { "epoch": 1.78, "learning_rate": 3.472828180427236e-05, "loss": 1.3266, "step": 17226 }, { "epoch": 1.78, "learning_rate": 3.472585113213376e-05, "loss": 0.3737, "step": 17227 }, { "epoch": 1.78, "learning_rate": 3.4723420428181845e-05, "loss": 0.4982, "step": 17228 }, { "epoch": 1.78, "learning_rate": 3.472098969243296e-05, "loss": 0.5661, "step": 17229 }, { "epoch": 1.78, "learning_rate": 3.4718558924903486e-05, "loss": 0.3301, "step": 17230 }, { "epoch": 1.78, "learning_rate": 3.471612812560978e-05, "loss": 0.5824, "step": 17231 }, { "epoch": 1.78, "learning_rate": 3.47136972945682e-05, "loss": 0.5624, "step": 17232 }, { "epoch": 1.78, "learning_rate": 3.471126643179511e-05, "loss": 0.56, "step": 17233 }, { "epoch": 1.78, "learning_rate": 3.470883553730688e-05, "loss": 0.7581, "step": 17234 }, { "epoch": 1.78, "learning_rate": 3.470640461111987e-05, "loss": 0.9913, "step": 17235 }, { "epoch": 1.78, "learning_rate": 3.470397365325044e-05, "loss": 0.252, "step": 17236 }, { "epoch": 1.78, "learning_rate": 3.470154266371497e-05, "loss": 0.1603, "step": 17237 }, { "epoch": 1.78, "learning_rate": 3.46991116425298e-05, "loss": 1.0985, "step": 17238 }, { "epoch": 1.78, "learning_rate": 3.4696680589711326e-05, "loss": 1.3715, "step": 17239 }, { "epoch": 1.79, "learning_rate": 3.469424950527589e-05, "loss": 0.9674, "step": 17240 }, { "epoch": 1.79, "learning_rate": 3.4691818389239864e-05, "loss": 0.8837, "step": 17241 }, { "epoch": 1.79, "learning_rate": 3.4689387241619616e-05, "loss": 0.3419, "step": 17242 }, { "epoch": 1.79, "learning_rate": 3.4686956062431513e-05, "loss": 1.0042, "step": 17243 }, { "epoch": 1.79, "learning_rate": 3.4684524851691915e-05, "loss": 1.026, "step": 17244 }, { "epoch": 1.79, "learning_rate": 3.46820936094172e-05, "loss": 1.007, "step": 17245 }, { "epoch": 1.79, "learning_rate": 3.46796623356237e-05, "loss": 0.7802, "step": 17246 }, { "epoch": 1.79, "learning_rate": 3.467723103032784e-05, "loss": 0.3659, "step": 17247 }, { "epoch": 1.79, "learning_rate": 3.4674799693545935e-05, "loss": 0.2372, "step": 17248 }, { "epoch": 1.79, "learning_rate": 3.467236832529438e-05, "loss": 0.7655, "step": 17249 }, { "epoch": 1.79, "learning_rate": 3.466993692558954e-05, "loss": 1.2805, "step": 17250 }, { "epoch": 1.79, "learning_rate": 3.4667505494447765e-05, "loss": 0.3267, "step": 17251 }, { "epoch": 1.79, "learning_rate": 3.466507403188544e-05, "loss": 0.2422, "step": 17252 }, { "epoch": 1.79, "learning_rate": 3.466264253791892e-05, "loss": 1.1114, "step": 17253 }, { "epoch": 1.79, "learning_rate": 3.46602110125646e-05, "loss": 0.6955, "step": 17254 }, { "epoch": 1.79, "learning_rate": 3.4657779455838825e-05, "loss": 0.4051, "step": 17255 }, { "epoch": 1.79, "learning_rate": 3.465534786775796e-05, "loss": 1.1102, "step": 17256 }, { "epoch": 1.79, "learning_rate": 3.465291624833839e-05, "loss": 0.8554, "step": 17257 }, { "epoch": 1.79, "learning_rate": 3.465048459759647e-05, "loss": 0.5391, "step": 17258 }, { "epoch": 1.79, "learning_rate": 3.4648052915548575e-05, "loss": 0.5273, "step": 17259 }, { "epoch": 1.79, "learning_rate": 3.464562120221108e-05, "loss": 0.4026, "step": 17260 }, { "epoch": 1.79, "learning_rate": 3.464318945760035e-05, "loss": 0.3198, "step": 17261 }, { "epoch": 1.79, "learning_rate": 3.464075768173276e-05, "loss": 0.3841, "step": 17262 }, { "epoch": 1.79, "learning_rate": 3.4638325874624664e-05, "loss": 1.0803, "step": 17263 }, { "epoch": 1.79, "learning_rate": 3.4635894036292457e-05, "loss": 1.8904, "step": 17264 }, { "epoch": 1.79, "learning_rate": 3.4633462166752485e-05, "loss": 0.5415, "step": 17265 }, { "epoch": 1.79, "learning_rate": 3.463103026602113e-05, "loss": 0.8261, "step": 17266 }, { "epoch": 1.79, "learning_rate": 3.462859833411476e-05, "loss": 0.1682, "step": 17267 }, { "epoch": 1.79, "learning_rate": 3.462616637104976e-05, "loss": 0.5393, "step": 17268 }, { "epoch": 1.79, "learning_rate": 3.462373437684249e-05, "loss": 2.786, "step": 17269 }, { "epoch": 1.79, "learning_rate": 3.4621302351509314e-05, "loss": 0.3322, "step": 17270 }, { "epoch": 1.79, "learning_rate": 3.461887029506661e-05, "loss": 0.9942, "step": 17271 }, { "epoch": 1.79, "learning_rate": 3.4616438207530756e-05, "loss": 0.6511, "step": 17272 }, { "epoch": 1.79, "learning_rate": 3.4614006088918125e-05, "loss": 0.4035, "step": 17273 }, { "epoch": 1.79, "learning_rate": 3.461157393924508e-05, "loss": 0.6632, "step": 17274 }, { "epoch": 1.79, "learning_rate": 3.4609141758528005e-05, "loss": 0.6377, "step": 17275 }, { "epoch": 1.79, "learning_rate": 3.4606709546783254e-05, "loss": 0.6737, "step": 17276 }, { "epoch": 1.79, "learning_rate": 3.460427730402722e-05, "loss": 1.0556, "step": 17277 }, { "epoch": 1.79, "learning_rate": 3.4601845030276265e-05, "loss": 0.4629, "step": 17278 }, { "epoch": 1.79, "learning_rate": 3.4599412725546765e-05, "loss": 0.4871, "step": 17279 }, { "epoch": 1.79, "learning_rate": 3.459698038985509e-05, "loss": 0.2018, "step": 17280 }, { "epoch": 1.79, "learning_rate": 3.4594548023217634e-05, "loss": 0.5463, "step": 17281 }, { "epoch": 1.79, "learning_rate": 3.459211562565075e-05, "loss": 0.2836, "step": 17282 }, { "epoch": 1.79, "learning_rate": 3.458968319717082e-05, "loss": 0.1264, "step": 17283 }, { "epoch": 1.79, "learning_rate": 3.458725073779421e-05, "loss": 0.9357, "step": 17284 }, { "epoch": 1.79, "learning_rate": 3.45848182475373e-05, "loss": 1.1622, "step": 17285 }, { "epoch": 1.79, "learning_rate": 3.458238572641646e-05, "loss": 0.486, "step": 17286 }, { "epoch": 1.79, "learning_rate": 3.457995317444809e-05, "loss": 0.7637, "step": 17287 }, { "epoch": 1.79, "learning_rate": 3.457752059164854e-05, "loss": 0.725, "step": 17288 }, { "epoch": 1.79, "learning_rate": 3.4575087978034194e-05, "loss": 1.1456, "step": 17289 }, { "epoch": 1.79, "learning_rate": 3.457265533362142e-05, "loss": 0.9671, "step": 17290 }, { "epoch": 1.79, "learning_rate": 3.45702226584266e-05, "loss": 1.0223, "step": 17291 }, { "epoch": 1.79, "learning_rate": 3.456778995246612e-05, "loss": 0.9243, "step": 17292 }, { "epoch": 1.79, "learning_rate": 3.456535721575634e-05, "loss": 0.8696, "step": 17293 }, { "epoch": 1.79, "learning_rate": 3.456292444831365e-05, "loss": 0.7895, "step": 17294 }, { "epoch": 1.79, "learning_rate": 3.456049165015441e-05, "loss": 0.2017, "step": 17295 }, { "epoch": 1.79, "learning_rate": 3.4558058821295006e-05, "loss": 1.1031, "step": 17296 }, { "epoch": 1.79, "learning_rate": 3.4555625961751823e-05, "loss": 0.517, "step": 17297 }, { "epoch": 1.79, "learning_rate": 3.455319307154124e-05, "loss": 0.3386, "step": 17298 }, { "epoch": 1.79, "learning_rate": 3.455076015067962e-05, "loss": 0.4176, "step": 17299 }, { "epoch": 1.79, "learning_rate": 3.454832719918333e-05, "loss": 0.3547, "step": 17300 }, { "epoch": 1.79, "learning_rate": 3.454589421706878e-05, "loss": 0.6819, "step": 17301 }, { "epoch": 1.79, "learning_rate": 3.454346120435234e-05, "loss": 0.7242, "step": 17302 }, { "epoch": 1.79, "learning_rate": 3.4541028161050376e-05, "loss": 0.5433, "step": 17303 }, { "epoch": 1.79, "learning_rate": 3.4538595087179275e-05, "loss": 1.0432, "step": 17304 }, { "epoch": 1.79, "learning_rate": 3.4536161982755404e-05, "loss": 0.5255, "step": 17305 }, { "epoch": 1.79, "learning_rate": 3.453372884779516e-05, "loss": 0.5372, "step": 17306 }, { "epoch": 1.79, "learning_rate": 3.453129568231493e-05, "loss": 0.2063, "step": 17307 }, { "epoch": 1.79, "learning_rate": 3.452886248633106e-05, "loss": 0.3916, "step": 17308 }, { "epoch": 1.79, "learning_rate": 3.452642925985994e-05, "loss": 0.1762, "step": 17309 }, { "epoch": 1.79, "learning_rate": 3.4523996002917973e-05, "loss": 0.2537, "step": 17310 }, { "epoch": 1.79, "learning_rate": 3.4521562715521515e-05, "loss": 0.2645, "step": 17311 }, { "epoch": 1.79, "learning_rate": 3.451912939768696e-05, "loss": 0.4678, "step": 17312 }, { "epoch": 1.79, "learning_rate": 3.451669604943068e-05, "loss": 0.4363, "step": 17313 }, { "epoch": 1.79, "learning_rate": 3.451426267076906e-05, "loss": 0.1547, "step": 17314 }, { "epoch": 1.79, "learning_rate": 3.451182926171847e-05, "loss": 1.0987, "step": 17315 }, { "epoch": 1.79, "learning_rate": 3.4509395822295314e-05, "loss": 0.2965, "step": 17316 }, { "epoch": 1.79, "learning_rate": 3.450696235251596e-05, "loss": 0.4185, "step": 17317 }, { "epoch": 1.79, "learning_rate": 3.450452885239679e-05, "loss": 0.4615, "step": 17318 }, { "epoch": 1.79, "learning_rate": 3.450209532195418e-05, "loss": 0.4064, "step": 17319 }, { "epoch": 1.79, "learning_rate": 3.449966176120452e-05, "loss": 0.9515, "step": 17320 }, { "epoch": 1.79, "learning_rate": 3.449722817016419e-05, "loss": 1.7516, "step": 17321 }, { "epoch": 1.79, "learning_rate": 3.449479454884957e-05, "loss": 0.3162, "step": 17322 }, { "epoch": 1.79, "learning_rate": 3.449236089727705e-05, "loss": 0.6121, "step": 17323 }, { "epoch": 1.79, "learning_rate": 3.4489927215463e-05, "loss": 1.3184, "step": 17324 }, { "epoch": 1.79, "learning_rate": 3.448749350342382e-05, "loss": 0.8585, "step": 17325 }, { "epoch": 1.79, "learning_rate": 3.448505976117588e-05, "loss": 1.0849, "step": 17326 }, { "epoch": 1.79, "learning_rate": 3.4482625988735565e-05, "loss": 0.579, "step": 17327 }, { "epoch": 1.79, "learning_rate": 3.448019218611926e-05, "loss": 0.7499, "step": 17328 }, { "epoch": 1.79, "learning_rate": 3.447775835334335e-05, "loss": 0.9652, "step": 17329 }, { "epoch": 1.79, "learning_rate": 3.447532449042423e-05, "loss": 0.8284, "step": 17330 }, { "epoch": 1.79, "learning_rate": 3.447289059737826e-05, "loss": 1.0192, "step": 17331 }, { "epoch": 1.79, "learning_rate": 3.447045667422184e-05, "loss": 1.4238, "step": 17332 }, { "epoch": 1.79, "learning_rate": 3.446802272097135e-05, "loss": 1.2617, "step": 17333 }, { "epoch": 1.79, "learning_rate": 3.4465588737643173e-05, "loss": 0.9855, "step": 17334 }, { "epoch": 1.79, "learning_rate": 3.4463154724253705e-05, "loss": 0.7547, "step": 17335 }, { "epoch": 1.8, "learning_rate": 3.4460720680819315e-05, "loss": 0.2877, "step": 17336 }, { "epoch": 1.8, "learning_rate": 3.445828660735641e-05, "loss": 0.3384, "step": 17337 }, { "epoch": 1.8, "learning_rate": 3.445585250388135e-05, "loss": 1.1541, "step": 17338 }, { "epoch": 1.8, "learning_rate": 3.445341837041054e-05, "loss": 0.4402, "step": 17339 }, { "epoch": 1.8, "learning_rate": 3.445098420696036e-05, "loss": 0.1554, "step": 17340 }, { "epoch": 1.8, "learning_rate": 3.444855001354719e-05, "loss": 0.3433, "step": 17341 }, { "epoch": 1.8, "learning_rate": 3.444611579018743e-05, "loss": 1.122, "step": 17342 }, { "epoch": 1.8, "learning_rate": 3.4443681536897456e-05, "loss": 0.2781, "step": 17343 }, { "epoch": 1.8, "learning_rate": 3.444124725369366e-05, "loss": 0.9884, "step": 17344 }, { "epoch": 1.8, "learning_rate": 3.4438812940592425e-05, "loss": 0.9954, "step": 17345 }, { "epoch": 1.8, "learning_rate": 3.443637859761013e-05, "loss": 0.2716, "step": 17346 }, { "epoch": 1.8, "learning_rate": 3.4433944224763184e-05, "loss": 1.3037, "step": 17347 }, { "epoch": 1.8, "learning_rate": 3.443150982206796e-05, "loss": 0.7736, "step": 17348 }, { "epoch": 1.8, "learning_rate": 3.442907538954085e-05, "loss": 0.0696, "step": 17349 }, { "epoch": 1.8, "learning_rate": 3.442664092719824e-05, "loss": 0.69, "step": 17350 }, { "epoch": 1.8, "learning_rate": 3.442420643505652e-05, "loss": 0.5588, "step": 17351 }, { "epoch": 1.8, "learning_rate": 3.4421771913132075e-05, "loss": 0.4741, "step": 17352 }, { "epoch": 1.8, "learning_rate": 3.44193373614413e-05, "loss": 0.3894, "step": 17353 }, { "epoch": 1.8, "learning_rate": 3.4416902780000574e-05, "loss": 1.0741, "step": 17354 }, { "epoch": 1.8, "learning_rate": 3.44144681688263e-05, "loss": 0.7617, "step": 17355 }, { "epoch": 1.8, "learning_rate": 3.441203352793485e-05, "loss": 0.7062, "step": 17356 }, { "epoch": 1.8, "learning_rate": 3.4409598857342624e-05, "loss": 0.2919, "step": 17357 }, { "epoch": 1.8, "learning_rate": 3.440716415706602e-05, "loss": 0.7666, "step": 17358 }, { "epoch": 1.8, "learning_rate": 3.4404729427121404e-05, "loss": 0.8894, "step": 17359 }, { "epoch": 1.8, "learning_rate": 3.440229466752519e-05, "loss": 0.8427, "step": 17360 }, { "epoch": 1.8, "learning_rate": 3.439985987829375e-05, "loss": 0.4583, "step": 17361 }, { "epoch": 1.8, "learning_rate": 3.439742505944349e-05, "loss": 0.4982, "step": 17362 }, { "epoch": 1.8, "learning_rate": 3.4394990210990796e-05, "loss": 0.1956, "step": 17363 }, { "epoch": 1.8, "learning_rate": 3.439255533295205e-05, "loss": 1.103, "step": 17364 }, { "epoch": 1.8, "learning_rate": 3.4390120425343655e-05, "loss": 1.1019, "step": 17365 }, { "epoch": 1.8, "learning_rate": 3.4387685488181986e-05, "loss": 0.376, "step": 17366 }, { "epoch": 1.8, "learning_rate": 3.438525052148345e-05, "loss": 1.15, "step": 17367 }, { "epoch": 1.8, "learning_rate": 3.438281552526444e-05, "loss": 1.0195, "step": 17368 }, { "epoch": 1.8, "learning_rate": 3.438038049954134e-05, "loss": 1.0417, "step": 17369 }, { "epoch": 1.8, "learning_rate": 3.437794544433054e-05, "loss": 0.401, "step": 17370 }, { "epoch": 1.8, "learning_rate": 3.437551035964843e-05, "loss": 0.1367, "step": 17371 }, { "epoch": 1.8, "learning_rate": 3.437307524551141e-05, "loss": 0.9967, "step": 17372 }, { "epoch": 1.8, "learning_rate": 3.4370640101935884e-05, "loss": 1.1952, "step": 17373 }, { "epoch": 1.8, "learning_rate": 3.436820492893822e-05, "loss": 0.1395, "step": 17374 }, { "epoch": 1.8, "learning_rate": 3.4365769726534824e-05, "loss": 0.6959, "step": 17375 }, { "epoch": 1.8, "learning_rate": 3.436333449474209e-05, "loss": 1.2357, "step": 17376 }, { "epoch": 1.8, "learning_rate": 3.43608992335764e-05, "loss": 0.9122, "step": 17377 }, { "epoch": 1.8, "learning_rate": 3.435846394305417e-05, "loss": 1.1269, "step": 17378 }, { "epoch": 1.8, "learning_rate": 3.4356028623191776e-05, "loss": 0.4781, "step": 17379 }, { "epoch": 1.8, "learning_rate": 3.435359327400561e-05, "loss": 0.457, "step": 17380 }, { "epoch": 1.8, "learning_rate": 3.435115789551208e-05, "loss": 1.1714, "step": 17381 }, { "epoch": 1.8, "learning_rate": 3.4348722487727574e-05, "loss": 0.3651, "step": 17382 }, { "epoch": 1.8, "learning_rate": 3.4346287050668486e-05, "loss": 0.5464, "step": 17383 }, { "epoch": 1.8, "learning_rate": 3.4343851584351206e-05, "loss": 0.1823, "step": 17384 }, { "epoch": 1.8, "learning_rate": 3.434141608879213e-05, "loss": 0.4676, "step": 17385 }, { "epoch": 1.8, "learning_rate": 3.433898056400766e-05, "loss": 0.6466, "step": 17386 }, { "epoch": 1.8, "learning_rate": 3.433654501001419e-05, "loss": 0.7233, "step": 17387 }, { "epoch": 1.8, "learning_rate": 3.433410942682812e-05, "loss": 0.9839, "step": 17388 }, { "epoch": 1.8, "learning_rate": 3.4331673814465826e-05, "loss": 0.4697, "step": 17389 }, { "epoch": 1.8, "learning_rate": 3.4329238172943725e-05, "loss": 0.6995, "step": 17390 }, { "epoch": 1.8, "learning_rate": 3.43268025022782e-05, "loss": 0.1499, "step": 17391 }, { "epoch": 1.8, "learning_rate": 3.432436680248566e-05, "loss": 1.1404, "step": 17392 }, { "epoch": 1.8, "learning_rate": 3.432193107358249e-05, "loss": 0.9358, "step": 17393 }, { "epoch": 1.8, "learning_rate": 3.43194953155851e-05, "loss": 0.3925, "step": 17394 }, { "epoch": 1.8, "learning_rate": 3.4317059528509865e-05, "loss": 0.1996, "step": 17395 }, { "epoch": 1.8, "learning_rate": 3.43146237123732e-05, "loss": 0.5945, "step": 17396 }, { "epoch": 1.8, "learning_rate": 3.431218786719151e-05, "loss": 0.3395, "step": 17397 }, { "epoch": 1.8, "learning_rate": 3.430975199298116e-05, "loss": 0.6003, "step": 17398 }, { "epoch": 1.8, "learning_rate": 3.4307316089758584e-05, "loss": 0.4911, "step": 17399 }, { "epoch": 1.8, "learning_rate": 3.4304880157540155e-05, "loss": 0.5243, "step": 17400 }, { "epoch": 1.8, "learning_rate": 3.4302444196342284e-05, "loss": 0.3644, "step": 17401 }, { "epoch": 1.8, "learning_rate": 3.430000820618137e-05, "loss": 1.0522, "step": 17402 }, { "epoch": 1.8, "learning_rate": 3.42975721870738e-05, "loss": 0.4594, "step": 17403 }, { "epoch": 1.8, "learning_rate": 3.4295136139035974e-05, "loss": 0.356, "step": 17404 }, { "epoch": 1.8, "learning_rate": 3.42927000620843e-05, "loss": 0.5367, "step": 17405 }, { "epoch": 1.8, "learning_rate": 3.4290263956235184e-05, "loss": 0.57, "step": 17406 }, { "epoch": 1.8, "learning_rate": 3.4287827821505015e-05, "loss": 0.1777, "step": 17407 }, { "epoch": 1.8, "learning_rate": 3.428539165791018e-05, "loss": 0.2959, "step": 17408 }, { "epoch": 1.8, "learning_rate": 3.428295546546709e-05, "loss": 0.1337, "step": 17409 }, { "epoch": 1.8, "learning_rate": 3.4280519244192155e-05, "loss": 0.3707, "step": 17410 }, { "epoch": 1.8, "learning_rate": 3.427808299410176e-05, "loss": 0.8402, "step": 17411 }, { "epoch": 1.8, "learning_rate": 3.427564671521232e-05, "loss": 0.796, "step": 17412 }, { "epoch": 1.8, "learning_rate": 3.427321040754022e-05, "loss": 0.3542, "step": 17413 }, { "epoch": 1.8, "learning_rate": 3.427077407110187e-05, "loss": 0.5424, "step": 17414 }, { "epoch": 1.8, "learning_rate": 3.426833770591366e-05, "loss": 0.2352, "step": 17415 }, { "epoch": 1.8, "learning_rate": 3.426590131199201e-05, "loss": 0.3351, "step": 17416 }, { "epoch": 1.8, "learning_rate": 3.4263464889353305e-05, "loss": 0.3193, "step": 17417 }, { "epoch": 1.8, "learning_rate": 3.426102843801395e-05, "loss": 0.9556, "step": 17418 }, { "epoch": 1.8, "learning_rate": 3.425859195799036e-05, "loss": 0.5079, "step": 17419 }, { "epoch": 1.8, "learning_rate": 3.425615544929891e-05, "loss": 0.847, "step": 17420 }, { "epoch": 1.8, "learning_rate": 3.4253718911956033e-05, "loss": 0.4358, "step": 17421 }, { "epoch": 1.8, "learning_rate": 3.42512823459781e-05, "loss": 0.7709, "step": 17422 }, { "epoch": 1.8, "learning_rate": 3.4248845751381534e-05, "loss": 0.4099, "step": 17423 }, { "epoch": 1.8, "learning_rate": 3.424640912818273e-05, "loss": 0.2232, "step": 17424 }, { "epoch": 1.8, "learning_rate": 3.4243972476398105e-05, "loss": 0.4114, "step": 17425 }, { "epoch": 1.8, "learning_rate": 3.424153579604405e-05, "loss": 0.6428, "step": 17426 }, { "epoch": 1.8, "learning_rate": 3.4239099087136955e-05, "loss": 0.4045, "step": 17427 }, { "epoch": 1.8, "learning_rate": 3.423666234969324e-05, "loss": 0.3291, "step": 17428 }, { "epoch": 1.8, "learning_rate": 3.423422558372931e-05, "loss": 0.3899, "step": 17429 }, { "epoch": 1.8, "learning_rate": 3.423178878926157e-05, "loss": 0.2009, "step": 17430 }, { "epoch": 1.81, "learning_rate": 3.4229351966306415e-05, "loss": 0.912, "step": 17431 }, { "epoch": 1.81, "learning_rate": 3.422691511488024e-05, "loss": 1.1644, "step": 17432 }, { "epoch": 1.81, "learning_rate": 3.422447823499947e-05, "loss": 1.4623, "step": 17433 }, { "epoch": 1.81, "learning_rate": 3.42220413266805e-05, "loss": 0.3164, "step": 17434 }, { "epoch": 1.81, "learning_rate": 3.421960438993974e-05, "loss": 0.1372, "step": 17435 }, { "epoch": 1.81, "learning_rate": 3.421716742479359e-05, "loss": 0.2434, "step": 17436 }, { "epoch": 1.81, "learning_rate": 3.421473043125844e-05, "loss": 1.061, "step": 17437 }, { "epoch": 1.81, "learning_rate": 3.4212293409350734e-05, "loss": 2.5862, "step": 17438 }, { "epoch": 1.81, "learning_rate": 3.420985635908684e-05, "loss": 0.9254, "step": 17439 }, { "epoch": 1.81, "learning_rate": 3.420741928048318e-05, "loss": 0.6632, "step": 17440 }, { "epoch": 1.81, "learning_rate": 3.420498217355616e-05, "loss": 0.2789, "step": 17441 }, { "epoch": 1.81, "learning_rate": 3.420254503832219e-05, "loss": 0.2601, "step": 17442 }, { "epoch": 1.81, "learning_rate": 3.4200107874797654e-05, "loss": 0.3821, "step": 17443 }, { "epoch": 1.81, "learning_rate": 3.419767068299899e-05, "loss": 0.3171, "step": 17444 }, { "epoch": 1.81, "learning_rate": 3.419523346294258e-05, "loss": 0.2496, "step": 17445 }, { "epoch": 1.81, "learning_rate": 3.419279621464485e-05, "loss": 0.976, "step": 17446 }, { "epoch": 1.81, "learning_rate": 3.419035893812218e-05, "loss": 0.6917, "step": 17447 }, { "epoch": 1.81, "learning_rate": 3.4187921633391007e-05, "loss": 0.8798, "step": 17448 }, { "epoch": 1.81, "learning_rate": 3.418548430046773e-05, "loss": 0.4136, "step": 17449 }, { "epoch": 1.81, "learning_rate": 3.418304693936875e-05, "loss": 0.5969, "step": 17450 }, { "epoch": 1.81, "learning_rate": 3.418060955011047e-05, "loss": 1.11, "step": 17451 }, { "epoch": 1.81, "learning_rate": 3.41781721327093e-05, "loss": 1.1348, "step": 17452 }, { "epoch": 1.81, "learning_rate": 3.4175734687181654e-05, "loss": 0.7287, "step": 17453 }, { "epoch": 1.81, "learning_rate": 3.417329721354395e-05, "loss": 0.5849, "step": 17454 }, { "epoch": 1.81, "learning_rate": 3.417085971181258e-05, "loss": 0.4551, "step": 17455 }, { "epoch": 1.81, "learning_rate": 3.416842218200396e-05, "loss": 0.244, "step": 17456 }, { "epoch": 1.81, "learning_rate": 3.416598462413449e-05, "loss": 1.0366, "step": 17457 }, { "epoch": 1.81, "learning_rate": 3.416354703822059e-05, "loss": 0.7847, "step": 17458 }, { "epoch": 1.81, "learning_rate": 3.416110942427867e-05, "loss": 0.362, "step": 17459 }, { "epoch": 1.81, "learning_rate": 3.415867178232513e-05, "loss": 0.6975, "step": 17460 }, { "epoch": 1.81, "learning_rate": 3.4156234112376386e-05, "loss": 0.483, "step": 17461 }, { "epoch": 1.81, "learning_rate": 3.4153796414448844e-05, "loss": 0.6807, "step": 17462 }, { "epoch": 1.81, "learning_rate": 3.415135868855892e-05, "loss": 0.1578, "step": 17463 }, { "epoch": 1.81, "learning_rate": 3.414892093472303e-05, "loss": 1.0395, "step": 17464 }, { "epoch": 1.81, "learning_rate": 3.414648315295756e-05, "loss": 0.3629, "step": 17465 }, { "epoch": 1.81, "learning_rate": 3.414404534327894e-05, "loss": 0.9232, "step": 17466 }, { "epoch": 1.81, "learning_rate": 3.414160750570358e-05, "loss": 1.0207, "step": 17467 }, { "epoch": 1.81, "learning_rate": 3.413916964024789e-05, "loss": 0.5269, "step": 17468 }, { "epoch": 1.81, "learning_rate": 3.4136731746928275e-05, "loss": 0.7217, "step": 17469 }, { "epoch": 1.81, "learning_rate": 3.413429382576114e-05, "loss": 0.4682, "step": 17470 }, { "epoch": 1.81, "learning_rate": 3.413185587676292e-05, "loss": 0.2347, "step": 17471 }, { "epoch": 1.81, "learning_rate": 3.4129417899949994e-05, "loss": 0.531, "step": 17472 }, { "epoch": 1.81, "learning_rate": 3.412697989533882e-05, "loss": 0.1889, "step": 17473 }, { "epoch": 1.81, "learning_rate": 3.412454186294576e-05, "loss": 0.5795, "step": 17474 }, { "epoch": 1.81, "learning_rate": 3.4122103802787264e-05, "loss": 0.6441, "step": 17475 }, { "epoch": 1.81, "learning_rate": 3.411966571487972e-05, "loss": 0.228, "step": 17476 }, { "epoch": 1.81, "learning_rate": 3.411722759923956e-05, "loss": 0.3766, "step": 17477 }, { "epoch": 1.81, "learning_rate": 3.411478945588318e-05, "loss": 1.5611, "step": 17478 }, { "epoch": 1.81, "learning_rate": 3.4112351284827e-05, "loss": 0.3252, "step": 17479 }, { "epoch": 1.81, "learning_rate": 3.4109913086087434e-05, "loss": 1.0852, "step": 17480 }, { "epoch": 1.81, "learning_rate": 3.410747485968089e-05, "loss": 0.212, "step": 17481 }, { "epoch": 1.81, "learning_rate": 3.41050366056238e-05, "loss": 0.6515, "step": 17482 }, { "epoch": 1.81, "learning_rate": 3.410259832393256e-05, "loss": 0.5811, "step": 17483 }, { "epoch": 1.81, "learning_rate": 3.410016001462358e-05, "loss": 0.5532, "step": 17484 }, { "epoch": 1.81, "learning_rate": 3.4097721677713285e-05, "loss": 0.3682, "step": 17485 }, { "epoch": 1.81, "learning_rate": 3.4095283313218085e-05, "loss": 0.1377, "step": 17486 }, { "epoch": 1.81, "learning_rate": 3.40928449211544e-05, "loss": 0.6318, "step": 17487 }, { "epoch": 1.81, "learning_rate": 3.409040650153865e-05, "loss": 0.4165, "step": 17488 }, { "epoch": 1.81, "learning_rate": 3.408796805438722e-05, "loss": 0.7388, "step": 17489 }, { "epoch": 1.81, "learning_rate": 3.4085529579716556e-05, "loss": 0.875, "step": 17490 }, { "epoch": 1.81, "learning_rate": 3.4083091077543055e-05, "loss": 0.676, "step": 17491 }, { "epoch": 1.81, "learning_rate": 3.408065254788316e-05, "loss": 0.3013, "step": 17492 }, { "epoch": 1.81, "learning_rate": 3.4078213990753246e-05, "loss": 0.9705, "step": 17493 }, { "epoch": 1.81, "learning_rate": 3.407577540616976e-05, "loss": 0.5623, "step": 17494 }, { "epoch": 1.81, "learning_rate": 3.4073336794149106e-05, "loss": 0.5384, "step": 17495 }, { "epoch": 1.81, "learning_rate": 3.40708981547077e-05, "loss": 1.0577, "step": 17496 }, { "epoch": 1.81, "learning_rate": 3.4068459487861964e-05, "loss": 0.3394, "step": 17497 }, { "epoch": 1.81, "learning_rate": 3.406602079362831e-05, "loss": 0.4189, "step": 17498 }, { "epoch": 1.81, "learning_rate": 3.406358207202315e-05, "loss": 0.9778, "step": 17499 }, { "epoch": 1.81, "learning_rate": 3.406114332306291e-05, "loss": 0.813, "step": 17500 }, { "epoch": 1.81, "learning_rate": 3.405870454676401e-05, "loss": 0.3532, "step": 17501 }, { "epoch": 1.81, "learning_rate": 3.405626574314285e-05, "loss": 0.6427, "step": 17502 }, { "epoch": 1.81, "learning_rate": 3.405382691221586e-05, "loss": 0.2341, "step": 17503 }, { "epoch": 1.81, "learning_rate": 3.405138805399945e-05, "loss": 1.2056, "step": 17504 }, { "epoch": 1.81, "learning_rate": 3.4048949168510055e-05, "loss": 0.2002, "step": 17505 }, { "epoch": 1.81, "learning_rate": 3.4046510255764086e-05, "loss": 0.8912, "step": 17506 }, { "epoch": 1.81, "learning_rate": 3.404407131577795e-05, "loss": 0.9134, "step": 17507 }, { "epoch": 1.81, "learning_rate": 3.404163234856807e-05, "loss": 0.9582, "step": 17508 }, { "epoch": 1.81, "learning_rate": 3.4039193354150864e-05, "loss": 0.4205, "step": 17509 }, { "epoch": 1.81, "learning_rate": 3.4036754332542757e-05, "loss": 1.1332, "step": 17510 }, { "epoch": 1.81, "learning_rate": 3.403431528376017e-05, "loss": 1.4403, "step": 17511 }, { "epoch": 1.81, "learning_rate": 3.40318762078195e-05, "loss": 1.1434, "step": 17512 }, { "epoch": 1.81, "learning_rate": 3.40294371047372e-05, "loss": 0.6764, "step": 17513 }, { "epoch": 1.81, "learning_rate": 3.4026997974529664e-05, "loss": 0.4415, "step": 17514 }, { "epoch": 1.81, "learning_rate": 3.402455881721332e-05, "loss": 0.9331, "step": 17515 }, { "epoch": 1.81, "learning_rate": 3.402211963280459e-05, "loss": 0.6126, "step": 17516 }, { "epoch": 1.81, "learning_rate": 3.4019680421319894e-05, "loss": 0.7887, "step": 17517 }, { "epoch": 1.81, "learning_rate": 3.4017241182775645e-05, "loss": 0.9136, "step": 17518 }, { "epoch": 1.81, "learning_rate": 3.401480191718827e-05, "loss": 0.6788, "step": 17519 }, { "epoch": 1.81, "learning_rate": 3.401236262457419e-05, "loss": 0.5807, "step": 17520 }, { "epoch": 1.81, "learning_rate": 3.400992330494983e-05, "loss": 0.5533, "step": 17521 }, { "epoch": 1.81, "learning_rate": 3.400748395833159e-05, "loss": 0.1802, "step": 17522 }, { "epoch": 1.81, "learning_rate": 3.400504458473591e-05, "loss": 0.7989, "step": 17523 }, { "epoch": 1.81, "learning_rate": 3.400260518417921e-05, "loss": 0.9522, "step": 17524 }, { "epoch": 1.81, "learning_rate": 3.400016575667791e-05, "loss": 0.2969, "step": 17525 }, { "epoch": 1.81, "learning_rate": 3.399772630224843e-05, "loss": 1.1064, "step": 17526 }, { "epoch": 1.82, "learning_rate": 3.399528682090718e-05, "loss": 0.3326, "step": 17527 }, { "epoch": 1.82, "learning_rate": 3.3992847312670604e-05, "loss": 0.7341, "step": 17528 }, { "epoch": 1.82, "learning_rate": 3.3990407777555115e-05, "loss": 0.886, "step": 17529 }, { "epoch": 1.82, "learning_rate": 3.3987968215577135e-05, "loss": 0.5873, "step": 17530 }, { "epoch": 1.82, "learning_rate": 3.398552862675308e-05, "loss": 0.5648, "step": 17531 }, { "epoch": 1.82, "learning_rate": 3.398308901109939e-05, "loss": 0.9712, "step": 17532 }, { "epoch": 1.82, "learning_rate": 3.398064936863247e-05, "loss": 0.8139, "step": 17533 }, { "epoch": 1.82, "learning_rate": 3.397820969936874e-05, "loss": 0.5645, "step": 17534 }, { "epoch": 1.82, "learning_rate": 3.397577000332465e-05, "loss": 1.3182, "step": 17535 }, { "epoch": 1.82, "learning_rate": 3.39733302805166e-05, "loss": 0.2946, "step": 17536 }, { "epoch": 1.82, "learning_rate": 3.3970890530961014e-05, "loss": 0.4587, "step": 17537 }, { "epoch": 1.82, "learning_rate": 3.396845075467433e-05, "loss": 0.4784, "step": 17538 }, { "epoch": 1.82, "learning_rate": 3.3966010951672965e-05, "loss": 0.6237, "step": 17539 }, { "epoch": 1.82, "learning_rate": 3.396357112197334e-05, "loss": 1.1636, "step": 17540 }, { "epoch": 1.82, "learning_rate": 3.396113126559188e-05, "loss": 0.5609, "step": 17541 }, { "epoch": 1.82, "learning_rate": 3.3958691382545015e-05, "loss": 0.4853, "step": 17542 }, { "epoch": 1.82, "learning_rate": 3.3956251472849155e-05, "loss": 0.9385, "step": 17543 }, { "epoch": 1.82, "learning_rate": 3.395381153652075e-05, "loss": 0.964, "step": 17544 }, { "epoch": 1.82, "learning_rate": 3.395137157357621e-05, "loss": 1.0829, "step": 17545 }, { "epoch": 1.82, "learning_rate": 3.394893158403195e-05, "loss": 0.3752, "step": 17546 }, { "epoch": 1.82, "learning_rate": 3.3946491567904406e-05, "loss": 0.1839, "step": 17547 }, { "epoch": 1.82, "learning_rate": 3.3944051525210014e-05, "loss": 0.9027, "step": 17548 }, { "epoch": 1.82, "learning_rate": 3.3941611455965185e-05, "loss": 0.8915, "step": 17549 }, { "epoch": 1.82, "learning_rate": 3.393917136018635e-05, "loss": 0.7955, "step": 17550 }, { "epoch": 1.82, "learning_rate": 3.3936731237889936e-05, "loss": 0.3238, "step": 17551 }, { "epoch": 1.82, "learning_rate": 3.393429108909237e-05, "loss": 0.8525, "step": 17552 }, { "epoch": 1.82, "learning_rate": 3.393185091381007e-05, "loss": 0.6, "step": 17553 }, { "epoch": 1.82, "learning_rate": 3.392941071205948e-05, "loss": 0.4149, "step": 17554 }, { "epoch": 1.82, "learning_rate": 3.3926970483857014e-05, "loss": 1.0068, "step": 17555 }, { "epoch": 1.82, "learning_rate": 3.3924530229219095e-05, "loss": 0.3498, "step": 17556 }, { "epoch": 1.82, "learning_rate": 3.3922089948162164e-05, "loss": 0.9617, "step": 17557 }, { "epoch": 1.82, "learning_rate": 3.3919649640702633e-05, "loss": 0.3037, "step": 17558 }, { "epoch": 1.82, "learning_rate": 3.391720930685695e-05, "loss": 1.1025, "step": 17559 }, { "epoch": 1.82, "learning_rate": 3.391476894664152e-05, "loss": 1.2536, "step": 17560 }, { "epoch": 1.82, "learning_rate": 3.391232856007279e-05, "loss": 1.0448, "step": 17561 }, { "epoch": 1.82, "learning_rate": 3.390988814716717e-05, "loss": 0.4623, "step": 17562 }, { "epoch": 1.82, "learning_rate": 3.3907447707941106e-05, "loss": 0.3099, "step": 17563 }, { "epoch": 1.82, "learning_rate": 3.390500724241101e-05, "loss": 0.5063, "step": 17564 }, { "epoch": 1.82, "learning_rate": 3.390256675059333e-05, "loss": 0.6175, "step": 17565 }, { "epoch": 1.82, "learning_rate": 3.390012623250447e-05, "loss": 1.2352, "step": 17566 }, { "epoch": 1.82, "learning_rate": 3.389768568816088e-05, "loss": 1.0604, "step": 17567 }, { "epoch": 1.82, "learning_rate": 3.3895245117578985e-05, "loss": 0.0616, "step": 17568 }, { "epoch": 1.82, "learning_rate": 3.38928045207752e-05, "loss": 0.7787, "step": 17569 }, { "epoch": 1.82, "learning_rate": 3.389036389776598e-05, "loss": 0.8966, "step": 17570 }, { "epoch": 1.82, "learning_rate": 3.3887923248567735e-05, "loss": 0.814, "step": 17571 }, { "epoch": 1.82, "learning_rate": 3.3885482573196896e-05, "loss": 0.4279, "step": 17572 }, { "epoch": 1.82, "learning_rate": 3.38830418716699e-05, "loss": 0.3729, "step": 17573 }, { "epoch": 1.82, "learning_rate": 3.3880601144003175e-05, "loss": 0.8632, "step": 17574 }, { "epoch": 1.82, "learning_rate": 3.387816039021315e-05, "loss": 1.1846, "step": 17575 }, { "epoch": 1.82, "learning_rate": 3.3875719610316265e-05, "loss": 0.8412, "step": 17576 }, { "epoch": 1.82, "learning_rate": 3.387327880432894e-05, "loss": 0.1464, "step": 17577 }, { "epoch": 1.82, "learning_rate": 3.3870837972267604e-05, "loss": 0.6726, "step": 17578 }, { "epoch": 1.82, "learning_rate": 3.386839711414869e-05, "loss": 1.0821, "step": 17579 }, { "epoch": 1.82, "learning_rate": 3.386595622998863e-05, "loss": 0.248, "step": 17580 }, { "epoch": 1.82, "learning_rate": 3.386351531980386e-05, "loss": 0.8962, "step": 17581 }, { "epoch": 1.82, "learning_rate": 3.3861074383610816e-05, "loss": 0.2094, "step": 17582 }, { "epoch": 1.82, "learning_rate": 3.385863342142592e-05, "loss": 0.1217, "step": 17583 }, { "epoch": 1.82, "learning_rate": 3.385619243326561e-05, "loss": 0.9652, "step": 17584 }, { "epoch": 1.82, "learning_rate": 3.38537514191463e-05, "loss": 1.3253, "step": 17585 }, { "epoch": 1.82, "learning_rate": 3.3851310379084445e-05, "loss": 0.8484, "step": 17586 }, { "epoch": 1.82, "learning_rate": 3.384886931309647e-05, "loss": 0.3379, "step": 17587 }, { "epoch": 1.82, "learning_rate": 3.384642822119882e-05, "loss": 0.4958, "step": 17588 }, { "epoch": 1.82, "learning_rate": 3.38439871034079e-05, "loss": 0.5299, "step": 17589 }, { "epoch": 1.82, "learning_rate": 3.3841545959740155e-05, "loss": 1.1554, "step": 17590 }, { "epoch": 1.82, "learning_rate": 3.383910479021202e-05, "loss": 0.3213, "step": 17591 }, { "epoch": 1.82, "learning_rate": 3.383666359483995e-05, "loss": 0.5547, "step": 17592 }, { "epoch": 1.82, "learning_rate": 3.383422237364034e-05, "loss": 1.78, "step": 17593 }, { "epoch": 1.82, "learning_rate": 3.383178112662965e-05, "loss": 0.908, "step": 17594 }, { "epoch": 1.82, "learning_rate": 3.3829339853824306e-05, "loss": 0.3743, "step": 17595 }, { "epoch": 1.82, "learning_rate": 3.3826898555240737e-05, "loss": 0.3584, "step": 17596 }, { "epoch": 1.82, "learning_rate": 3.382445723089539e-05, "loss": 0.5116, "step": 17597 }, { "epoch": 1.82, "learning_rate": 3.382201588080468e-05, "loss": 0.9477, "step": 17598 }, { "epoch": 1.82, "learning_rate": 3.3819574504985064e-05, "loss": 0.2123, "step": 17599 }, { "epoch": 1.82, "learning_rate": 3.381713310345296e-05, "loss": 0.6865, "step": 17600 }, { "epoch": 1.82, "learning_rate": 3.381469167622482e-05, "loss": 0.5566, "step": 17601 }, { "epoch": 1.82, "learning_rate": 3.381225022331706e-05, "loss": 0.8076, "step": 17602 }, { "epoch": 1.82, "learning_rate": 3.380980874474612e-05, "loss": 0.2319, "step": 17603 }, { "epoch": 1.82, "learning_rate": 3.380736724052844e-05, "loss": 0.1289, "step": 17604 }, { "epoch": 1.82, "learning_rate": 3.380492571068046e-05, "loss": 0.8417, "step": 17605 }, { "epoch": 1.82, "learning_rate": 3.3802484155218615e-05, "loss": 0.9358, "step": 17606 }, { "epoch": 1.82, "learning_rate": 3.380004257415933e-05, "loss": 0.1185, "step": 17607 }, { "epoch": 1.82, "learning_rate": 3.3797600967519044e-05, "loss": 0.3238, "step": 17608 }, { "epoch": 1.82, "learning_rate": 3.37951593353142e-05, "loss": 0.3247, "step": 17609 }, { "epoch": 1.82, "learning_rate": 3.379271767756122e-05, "loss": 0.829, "step": 17610 }, { "epoch": 1.82, "learning_rate": 3.379027599427657e-05, "loss": 0.2779, "step": 17611 }, { "epoch": 1.82, "learning_rate": 3.378783428547666e-05, "loss": 0.7097, "step": 17612 }, { "epoch": 1.82, "learning_rate": 3.378539255117794e-05, "loss": 0.2921, "step": 17613 }, { "epoch": 1.82, "learning_rate": 3.3782950791396836e-05, "loss": 0.5035, "step": 17614 }, { "epoch": 1.82, "learning_rate": 3.37805090061498e-05, "loss": 0.2534, "step": 17615 }, { "epoch": 1.82, "learning_rate": 3.377806719545326e-05, "loss": 0.4571, "step": 17616 }, { "epoch": 1.82, "learning_rate": 3.377562535932366e-05, "loss": 0.3598, "step": 17617 }, { "epoch": 1.82, "learning_rate": 3.377318349777742e-05, "loss": 0.9955, "step": 17618 }, { "epoch": 1.82, "learning_rate": 3.3770741610831e-05, "loss": 0.7311, "step": 17619 }, { "epoch": 1.82, "learning_rate": 3.376829969850084e-05, "loss": 0.2223, "step": 17620 }, { "epoch": 1.82, "learning_rate": 3.376585776080335e-05, "loss": 0.9293, "step": 17621 }, { "epoch": 1.82, "learning_rate": 3.376341579775499e-05, "loss": 0.5882, "step": 17622 }, { "epoch": 1.83, "learning_rate": 3.3760973809372206e-05, "loss": 0.1791, "step": 17623 }, { "epoch": 1.83, "learning_rate": 3.375853179567142e-05, "loss": 0.9346, "step": 17624 }, { "epoch": 1.83, "learning_rate": 3.375608975666908e-05, "loss": 0.7607, "step": 17625 }, { "epoch": 1.83, "learning_rate": 3.375364769238162e-05, "loss": 0.5062, "step": 17626 }, { "epoch": 1.83, "learning_rate": 3.375120560282548e-05, "loss": 1.0203, "step": 17627 }, { "epoch": 1.83, "learning_rate": 3.37487634880171e-05, "loss": 1.0982, "step": 17628 }, { "epoch": 1.83, "learning_rate": 3.374632134797293e-05, "loss": 0.4321, "step": 17629 }, { "epoch": 1.83, "learning_rate": 3.3743879182709394e-05, "loss": 0.2788, "step": 17630 }, { "epoch": 1.83, "learning_rate": 3.374143699224294e-05, "loss": 0.5527, "step": 17631 }, { "epoch": 1.83, "learning_rate": 3.373899477659001e-05, "loss": 1.1415, "step": 17632 }, { "epoch": 1.83, "learning_rate": 3.373655253576704e-05, "loss": 1.0423, "step": 17633 }, { "epoch": 1.83, "learning_rate": 3.373411026979048e-05, "loss": 0.8721, "step": 17634 }, { "epoch": 1.83, "learning_rate": 3.373166797867675e-05, "loss": 0.4564, "step": 17635 }, { "epoch": 1.83, "learning_rate": 3.3729225662442314e-05, "loss": 1.343, "step": 17636 }, { "epoch": 1.83, "learning_rate": 3.37267833211036e-05, "loss": 0.3115, "step": 17637 }, { "epoch": 1.83, "learning_rate": 3.3724340954677046e-05, "loss": 0.6604, "step": 17638 }, { "epoch": 1.83, "learning_rate": 3.3721898563179116e-05, "loss": 0.675, "step": 17639 }, { "epoch": 1.83, "learning_rate": 3.371945614662623e-05, "loss": 0.4068, "step": 17640 }, { "epoch": 1.83, "learning_rate": 3.371701370503483e-05, "loss": 0.1356, "step": 17641 }, { "epoch": 1.83, "learning_rate": 3.371457123842136e-05, "loss": 0.9987, "step": 17642 }, { "epoch": 1.83, "learning_rate": 3.371212874680228e-05, "loss": 0.3422, "step": 17643 }, { "epoch": 1.83, "learning_rate": 3.3709686230194006e-05, "loss": 0.8865, "step": 17644 }, { "epoch": 1.83, "learning_rate": 3.3707243688613e-05, "loss": 0.3233, "step": 17645 }, { "epoch": 1.83, "learning_rate": 3.3704801122075695e-05, "loss": 0.8129, "step": 17646 }, { "epoch": 1.83, "learning_rate": 3.370235853059853e-05, "loss": 0.2676, "step": 17647 }, { "epoch": 1.83, "learning_rate": 3.369991591419796e-05, "loss": 0.0907, "step": 17648 }, { "epoch": 1.83, "learning_rate": 3.369747327289042e-05, "loss": 1.3044, "step": 17649 }, { "epoch": 1.83, "learning_rate": 3.369503060669235e-05, "loss": 0.3969, "step": 17650 }, { "epoch": 1.83, "learning_rate": 3.369258791562021e-05, "loss": 1.1382, "step": 17651 }, { "epoch": 1.83, "learning_rate": 3.369014519969042e-05, "loss": 0.1931, "step": 17652 }, { "epoch": 1.83, "learning_rate": 3.368770245891944e-05, "loss": 0.4966, "step": 17653 }, { "epoch": 1.83, "learning_rate": 3.368525969332372e-05, "loss": 0.3873, "step": 17654 }, { "epoch": 1.83, "learning_rate": 3.368281690291968e-05, "loss": 0.642, "step": 17655 }, { "epoch": 1.83, "learning_rate": 3.368037408772378e-05, "loss": 0.8435, "step": 17656 }, { "epoch": 1.83, "learning_rate": 3.3677931247752465e-05, "loss": 0.6416, "step": 17657 }, { "epoch": 1.83, "learning_rate": 3.367548838302218e-05, "loss": 0.8575, "step": 17658 }, { "epoch": 1.83, "learning_rate": 3.367304549354937e-05, "loss": 0.7253, "step": 17659 }, { "epoch": 1.83, "learning_rate": 3.3670602579350466e-05, "loss": 0.3208, "step": 17660 }, { "epoch": 1.83, "learning_rate": 3.366815964044193e-05, "loss": 0.8729, "step": 17661 }, { "epoch": 1.83, "learning_rate": 3.36657166768402e-05, "loss": 0.6175, "step": 17662 }, { "epoch": 1.83, "learning_rate": 3.366327368856173e-05, "loss": 0.9545, "step": 17663 }, { "epoch": 1.83, "learning_rate": 3.366083067562296e-05, "loss": 0.9242, "step": 17664 }, { "epoch": 1.83, "learning_rate": 3.365838763804032e-05, "loss": 1.0359, "step": 17665 }, { "epoch": 1.83, "learning_rate": 3.3655944575830275e-05, "loss": 0.7079, "step": 17666 }, { "epoch": 1.83, "learning_rate": 3.365350148900927e-05, "loss": 0.576, "step": 17667 }, { "epoch": 1.83, "learning_rate": 3.3651058377593755e-05, "loss": 0.278, "step": 17668 }, { "epoch": 1.83, "learning_rate": 3.3648615241600166e-05, "loss": 0.7647, "step": 17669 }, { "epoch": 1.83, "learning_rate": 3.364617208104495e-05, "loss": 0.2218, "step": 17670 }, { "epoch": 1.83, "learning_rate": 3.3643728895944554e-05, "loss": 0.4032, "step": 17671 }, { "epoch": 1.83, "learning_rate": 3.364128568631542e-05, "loss": 0.2178, "step": 17672 }, { "epoch": 1.83, "learning_rate": 3.363884245217402e-05, "loss": 0.667, "step": 17673 }, { "epoch": 1.83, "learning_rate": 3.363639919353678e-05, "loss": 0.7826, "step": 17674 }, { "epoch": 1.83, "learning_rate": 3.363395591042014e-05, "loss": 1.0977, "step": 17675 }, { "epoch": 1.83, "learning_rate": 3.3631512602840576e-05, "loss": 0.7109, "step": 17676 }, { "epoch": 1.83, "learning_rate": 3.362906927081451e-05, "loss": 0.9862, "step": 17677 }, { "epoch": 1.83, "learning_rate": 3.3626625914358395e-05, "loss": 0.2914, "step": 17678 }, { "epoch": 1.83, "learning_rate": 3.362418253348869e-05, "loss": 1.1112, "step": 17679 }, { "epoch": 1.83, "learning_rate": 3.362173912822183e-05, "loss": 0.9943, "step": 17680 }, { "epoch": 1.83, "learning_rate": 3.361929569857427e-05, "loss": 0.2603, "step": 17681 }, { "epoch": 1.83, "learning_rate": 3.361685224456246e-05, "loss": 0.902, "step": 17682 }, { "epoch": 1.83, "learning_rate": 3.361440876620285e-05, "loss": 0.3913, "step": 17683 }, { "epoch": 1.83, "learning_rate": 3.361196526351189e-05, "loss": 0.2908, "step": 17684 }, { "epoch": 1.83, "learning_rate": 3.360952173650601e-05, "loss": 0.4372, "step": 17685 }, { "epoch": 1.83, "learning_rate": 3.3607078185201686e-05, "loss": 0.1798, "step": 17686 }, { "epoch": 1.83, "learning_rate": 3.360463460961535e-05, "loss": 1.0313, "step": 17687 }, { "epoch": 1.83, "learning_rate": 3.360219100976346e-05, "loss": 0.3462, "step": 17688 }, { "epoch": 1.83, "learning_rate": 3.359974738566247e-05, "loss": 1.9172, "step": 17689 }, { "epoch": 1.83, "learning_rate": 3.359730373732881e-05, "loss": 0.6142, "step": 17690 }, { "epoch": 1.83, "learning_rate": 3.359486006477895e-05, "loss": 0.5308, "step": 17691 }, { "epoch": 1.83, "learning_rate": 3.359241636802934e-05, "loss": 0.4285, "step": 17692 }, { "epoch": 1.83, "learning_rate": 3.358997264709641e-05, "loss": 0.6737, "step": 17693 }, { "epoch": 1.83, "learning_rate": 3.3587528901996624e-05, "loss": 0.9432, "step": 17694 }, { "epoch": 1.83, "learning_rate": 3.3585085132746445e-05, "loss": 1.0676, "step": 17695 }, { "epoch": 1.83, "learning_rate": 3.3582641339362305e-05, "loss": 0.8499, "step": 17696 }, { "epoch": 1.83, "learning_rate": 3.358019752186066e-05, "loss": 0.4968, "step": 17697 }, { "epoch": 1.83, "learning_rate": 3.357775368025796e-05, "loss": 0.1521, "step": 17698 }, { "epoch": 1.83, "learning_rate": 3.357530981457066e-05, "loss": 0.1462, "step": 17699 }, { "epoch": 1.83, "learning_rate": 3.357286592481522e-05, "loss": 0.7666, "step": 17700 }, { "epoch": 1.83, "learning_rate": 3.357042201100808e-05, "loss": 0.6288, "step": 17701 }, { "epoch": 1.83, "learning_rate": 3.35679780731657e-05, "loss": 0.4911, "step": 17702 }, { "epoch": 1.83, "learning_rate": 3.356553411130451e-05, "loss": 0.7137, "step": 17703 }, { "epoch": 1.83, "learning_rate": 3.356309012544099e-05, "loss": 0.6666, "step": 17704 }, { "epoch": 1.83, "learning_rate": 3.356064611559157e-05, "loss": 0.4131, "step": 17705 }, { "epoch": 1.83, "learning_rate": 3.355820208177273e-05, "loss": 0.5587, "step": 17706 }, { "epoch": 1.83, "learning_rate": 3.3555758024000895e-05, "loss": 0.2895, "step": 17707 }, { "epoch": 1.83, "learning_rate": 3.355331394229254e-05, "loss": 0.4097, "step": 17708 }, { "epoch": 1.83, "learning_rate": 3.355086983666409e-05, "loss": 0.995, "step": 17709 }, { "epoch": 1.83, "learning_rate": 3.354842570713202e-05, "loss": 0.9391, "step": 17710 }, { "epoch": 1.83, "learning_rate": 3.354598155371279e-05, "loss": 0.8333, "step": 17711 }, { "epoch": 1.83, "learning_rate": 3.354353737642283e-05, "loss": 0.8435, "step": 17712 }, { "epoch": 1.83, "learning_rate": 3.354109317527861e-05, "loss": 1.5361, "step": 17713 }, { "epoch": 1.83, "learning_rate": 3.3538648950296586e-05, "loss": 0.5134, "step": 17714 }, { "epoch": 1.83, "learning_rate": 3.3536204701493195e-05, "loss": 0.424, "step": 17715 }, { "epoch": 1.83, "learning_rate": 3.3533760428884905e-05, "loss": 0.6059, "step": 17716 }, { "epoch": 1.83, "learning_rate": 3.353131613248816e-05, "loss": 0.9112, "step": 17717 }, { "epoch": 1.83, "learning_rate": 3.352887181231943e-05, "loss": 0.5366, "step": 17718 }, { "epoch": 1.84, "learning_rate": 3.352642746839515e-05, "loss": 0.2359, "step": 17719 }, { "epoch": 1.84, "learning_rate": 3.35239831007318e-05, "loss": 0.9082, "step": 17720 }, { "epoch": 1.84, "learning_rate": 3.3521538709345815e-05, "loss": 0.5365, "step": 17721 }, { "epoch": 1.84, "learning_rate": 3.3519094294253655e-05, "loss": 1.3052, "step": 17722 }, { "epoch": 1.84, "learning_rate": 3.351664985547177e-05, "loss": 0.6695, "step": 17723 }, { "epoch": 1.84, "learning_rate": 3.351420539301662e-05, "loss": 1.0814, "step": 17724 }, { "epoch": 1.84, "learning_rate": 3.3511760906904675e-05, "loss": 0.4914, "step": 17725 }, { "epoch": 1.84, "learning_rate": 3.350931639715237e-05, "loss": 0.3073, "step": 17726 }, { "epoch": 1.84, "learning_rate": 3.350687186377617e-05, "loss": 0.692, "step": 17727 }, { "epoch": 1.84, "learning_rate": 3.3504427306792526e-05, "loss": 0.7619, "step": 17728 }, { "epoch": 1.84, "learning_rate": 3.350198272621789e-05, "loss": 0.713, "step": 17729 }, { "epoch": 1.84, "learning_rate": 3.349953812206874e-05, "loss": 0.2904, "step": 17730 }, { "epoch": 1.84, "learning_rate": 3.3497093494361505e-05, "loss": 0.7942, "step": 17731 }, { "epoch": 1.84, "learning_rate": 3.349464884311266e-05, "loss": 0.8709, "step": 17732 }, { "epoch": 1.84, "learning_rate": 3.3492204168338666e-05, "loss": 1.2886, "step": 17733 }, { "epoch": 1.84, "learning_rate": 3.348975947005596e-05, "loss": 0.5618, "step": 17734 }, { "epoch": 1.84, "learning_rate": 3.3487314748281013e-05, "loss": 0.1816, "step": 17735 }, { "epoch": 1.84, "learning_rate": 3.3484870003030284e-05, "loss": 0.5527, "step": 17736 }, { "epoch": 1.84, "learning_rate": 3.348242523432021e-05, "loss": 0.8297, "step": 17737 }, { "epoch": 1.84, "learning_rate": 3.3479980442167275e-05, "loss": 1.1858, "step": 17738 }, { "epoch": 1.84, "learning_rate": 3.3477535626587935e-05, "loss": 0.2578, "step": 17739 }, { "epoch": 1.84, "learning_rate": 3.347509078759863e-05, "loss": 0.6286, "step": 17740 }, { "epoch": 1.84, "learning_rate": 3.347264592521583e-05, "loss": 0.4057, "step": 17741 }, { "epoch": 1.84, "learning_rate": 3.347020103945597e-05, "loss": 0.6819, "step": 17742 }, { "epoch": 1.84, "learning_rate": 3.346775613033555e-05, "loss": 0.2277, "step": 17743 }, { "epoch": 1.84, "learning_rate": 3.3465311197871006e-05, "loss": 0.3989, "step": 17744 }, { "epoch": 1.84, "learning_rate": 3.34628662420788e-05, "loss": 0.6396, "step": 17745 }, { "epoch": 1.84, "learning_rate": 3.346042126297538e-05, "loss": 0.2872, "step": 17746 }, { "epoch": 1.84, "learning_rate": 3.345797626057721e-05, "loss": 0.5109, "step": 17747 }, { "epoch": 1.84, "learning_rate": 3.345553123490076e-05, "loss": 0.4224, "step": 17748 }, { "epoch": 1.84, "learning_rate": 3.345308618596249e-05, "loss": 0.9044, "step": 17749 }, { "epoch": 1.84, "learning_rate": 3.345064111377883e-05, "loss": 0.4191, "step": 17750 }, { "epoch": 1.84, "learning_rate": 3.344819601836628e-05, "loss": 0.6627, "step": 17751 }, { "epoch": 1.84, "learning_rate": 3.3445750899741274e-05, "loss": 1.0708, "step": 17752 }, { "epoch": 1.84, "learning_rate": 3.344330575792028e-05, "loss": 1.2516, "step": 17753 }, { "epoch": 1.84, "learning_rate": 3.344086059291976e-05, "loss": 0.1929, "step": 17754 }, { "epoch": 1.84, "learning_rate": 3.3438415404756174e-05, "loss": 0.1893, "step": 17755 }, { "epoch": 1.84, "learning_rate": 3.343597019344597e-05, "loss": 1.1481, "step": 17756 }, { "epoch": 1.84, "learning_rate": 3.3433524959005624e-05, "loss": 0.2108, "step": 17757 }, { "epoch": 1.84, "learning_rate": 3.3431079701451594e-05, "loss": 1.1561, "step": 17758 }, { "epoch": 1.84, "learning_rate": 3.3428634420800344e-05, "loss": 0.9186, "step": 17759 }, { "epoch": 1.84, "learning_rate": 3.3426189117068314e-05, "loss": 0.4926, "step": 17760 }, { "epoch": 1.84, "learning_rate": 3.342374379027199e-05, "loss": 1.4893, "step": 17761 }, { "epoch": 1.84, "learning_rate": 3.342129844042782e-05, "loss": 0.4862, "step": 17762 }, { "epoch": 1.84, "learning_rate": 3.341885306755228e-05, "loss": 0.4942, "step": 17763 }, { "epoch": 1.84, "learning_rate": 3.3416407671661816e-05, "loss": 0.1917, "step": 17764 }, { "epoch": 1.84, "learning_rate": 3.341396225277288e-05, "loss": 0.2534, "step": 17765 }, { "epoch": 1.84, "learning_rate": 3.341151681090196e-05, "loss": 0.5906, "step": 17766 }, { "epoch": 1.84, "learning_rate": 3.340907134606551e-05, "loss": 0.9629, "step": 17767 }, { "epoch": 1.84, "learning_rate": 3.340662585827999e-05, "loss": 0.6563, "step": 17768 }, { "epoch": 1.84, "learning_rate": 3.340418034756185e-05, "loss": 0.5559, "step": 17769 }, { "epoch": 1.84, "learning_rate": 3.3401734813927575e-05, "loss": 1.0068, "step": 17770 }, { "epoch": 1.84, "learning_rate": 3.3399289257393615e-05, "loss": 1.0099, "step": 17771 }, { "epoch": 1.84, "learning_rate": 3.3396843677976436e-05, "loss": 0.3243, "step": 17772 }, { "epoch": 1.84, "learning_rate": 3.3394398075692505e-05, "loss": 1.2444, "step": 17773 }, { "epoch": 1.84, "learning_rate": 3.339195245055827e-05, "loss": 0.2829, "step": 17774 }, { "epoch": 1.84, "learning_rate": 3.3389506802590205e-05, "loss": 0.8661, "step": 17775 }, { "epoch": 1.84, "learning_rate": 3.338706113180478e-05, "loss": 0.363, "step": 17776 }, { "epoch": 1.84, "learning_rate": 3.338461543821845e-05, "loss": 0.7553, "step": 17777 }, { "epoch": 1.84, "learning_rate": 3.338216972184769e-05, "loss": 0.7948, "step": 17778 }, { "epoch": 1.84, "learning_rate": 3.337972398270894e-05, "loss": 0.1943, "step": 17779 }, { "epoch": 1.84, "learning_rate": 3.337727822081868e-05, "loss": 0.818, "step": 17780 }, { "epoch": 1.84, "learning_rate": 3.337483243619337e-05, "loss": 0.76, "step": 17781 }, { "epoch": 1.84, "learning_rate": 3.337238662884949e-05, "loss": 0.6631, "step": 17782 }, { "epoch": 1.84, "learning_rate": 3.33699407988035e-05, "loss": 0.5679, "step": 17783 }, { "epoch": 1.84, "learning_rate": 3.336749494607184e-05, "loss": 0.6286, "step": 17784 }, { "epoch": 1.84, "learning_rate": 3.336504907067099e-05, "loss": 0.8909, "step": 17785 }, { "epoch": 1.84, "learning_rate": 3.336260317261742e-05, "loss": 0.1878, "step": 17786 }, { "epoch": 1.84, "learning_rate": 3.3360157251927604e-05, "loss": 0.7155, "step": 17787 }, { "epoch": 1.84, "learning_rate": 3.335771130861798e-05, "loss": 0.256, "step": 17788 }, { "epoch": 1.84, "learning_rate": 3.3355265342705044e-05, "loss": 0.4933, "step": 17789 }, { "epoch": 1.84, "learning_rate": 3.3352819354205233e-05, "loss": 0.6939, "step": 17790 }, { "epoch": 1.84, "learning_rate": 3.335037334313503e-05, "loss": 0.7073, "step": 17791 }, { "epoch": 1.84, "learning_rate": 3.334792730951091e-05, "loss": 1.0427, "step": 17792 }, { "epoch": 1.84, "learning_rate": 3.334548125334932e-05, "loss": 1.2695, "step": 17793 }, { "epoch": 1.84, "learning_rate": 3.3343035174666716e-05, "loss": 0.8912, "step": 17794 }, { "epoch": 1.84, "learning_rate": 3.33405890734796e-05, "loss": 0.299, "step": 17795 }, { "epoch": 1.84, "learning_rate": 3.333814294980442e-05, "loss": 0.7473, "step": 17796 }, { "epoch": 1.84, "learning_rate": 3.333569680365764e-05, "loss": 0.4323, "step": 17797 }, { "epoch": 1.84, "learning_rate": 3.333325063505572e-05, "loss": 0.8814, "step": 17798 }, { "epoch": 1.84, "learning_rate": 3.3330804444015146e-05, "loss": 0.8778, "step": 17799 }, { "epoch": 1.84, "learning_rate": 3.3328358230552376e-05, "loss": 0.9892, "step": 17800 }, { "epoch": 1.84, "learning_rate": 3.332591199468388e-05, "loss": 1.214, "step": 17801 }, { "epoch": 1.84, "learning_rate": 3.332346573642612e-05, "loss": 0.3928, "step": 17802 }, { "epoch": 1.84, "learning_rate": 3.3321019455795565e-05, "loss": 0.3321, "step": 17803 }, { "epoch": 1.84, "learning_rate": 3.3318573152808677e-05, "loss": 0.9115, "step": 17804 }, { "epoch": 1.84, "learning_rate": 3.331612682748194e-05, "loss": 0.3593, "step": 17805 }, { "epoch": 1.84, "learning_rate": 3.331368047983181e-05, "loss": 0.1318, "step": 17806 }, { "epoch": 1.84, "learning_rate": 3.3311234109874756e-05, "loss": 1.0895, "step": 17807 }, { "epoch": 1.84, "learning_rate": 3.330878771762726e-05, "loss": 0.2925, "step": 17808 }, { "epoch": 1.84, "learning_rate": 3.330634130310577e-05, "loss": 0.6513, "step": 17809 }, { "epoch": 1.84, "learning_rate": 3.3303894866326754e-05, "loss": 1.1633, "step": 17810 }, { "epoch": 1.84, "learning_rate": 3.3301448407306707e-05, "loss": 0.9262, "step": 17811 }, { "epoch": 1.84, "learning_rate": 3.3299001926062074e-05, "loss": 0.9675, "step": 17812 }, { "epoch": 1.84, "learning_rate": 3.3296555422609334e-05, "loss": 0.2508, "step": 17813 }, { "epoch": 1.85, "learning_rate": 3.3294108896964956e-05, "loss": 0.8672, "step": 17814 }, { "epoch": 1.85, "learning_rate": 3.329166234914541e-05, "loss": 0.603, "step": 17815 }, { "epoch": 1.85, "learning_rate": 3.328921577916716e-05, "loss": 0.2647, "step": 17816 }, { "epoch": 1.85, "learning_rate": 3.328676918704668e-05, "loss": 0.2536, "step": 17817 }, { "epoch": 1.85, "learning_rate": 3.328432257280043e-05, "loss": 0.6195, "step": 17818 }, { "epoch": 1.85, "learning_rate": 3.3281875936444894e-05, "loss": 0.8276, "step": 17819 }, { "epoch": 1.85, "learning_rate": 3.327942927799655e-05, "loss": 0.764, "step": 17820 }, { "epoch": 1.85, "learning_rate": 3.327698259747185e-05, "loss": 0.9223, "step": 17821 }, { "epoch": 1.85, "learning_rate": 3.3274535894887256e-05, "loss": 0.4654, "step": 17822 }, { "epoch": 1.85, "learning_rate": 3.3272089170259266e-05, "loss": 0.5097, "step": 17823 }, { "epoch": 1.85, "learning_rate": 3.326964242360433e-05, "loss": 0.3834, "step": 17824 }, { "epoch": 1.85, "learning_rate": 3.3267195654938924e-05, "loss": 0.5616, "step": 17825 }, { "epoch": 1.85, "learning_rate": 3.3264748864279526e-05, "loss": 0.0629, "step": 17826 }, { "epoch": 1.85, "learning_rate": 3.32623020516426e-05, "loss": 0.3815, "step": 17827 }, { "epoch": 1.85, "learning_rate": 3.325985521704463e-05, "loss": 0.558, "step": 17828 }, { "epoch": 1.85, "learning_rate": 3.325740836050207e-05, "loss": 0.6852, "step": 17829 }, { "epoch": 1.85, "learning_rate": 3.32549614820314e-05, "loss": 0.9517, "step": 17830 }, { "epoch": 1.85, "learning_rate": 3.325251458164909e-05, "loss": 0.1683, "step": 17831 }, { "epoch": 1.85, "learning_rate": 3.325006765937161e-05, "loss": 1.0476, "step": 17832 }, { "epoch": 1.85, "learning_rate": 3.3247620715215435e-05, "loss": 0.6157, "step": 17833 }, { "epoch": 1.85, "learning_rate": 3.324517374919704e-05, "loss": 0.7235, "step": 17834 }, { "epoch": 1.85, "learning_rate": 3.32427267613329e-05, "loss": 1.222, "step": 17835 }, { "epoch": 1.85, "learning_rate": 3.324027975163948e-05, "loss": 0.8212, "step": 17836 }, { "epoch": 1.85, "learning_rate": 3.3237832720133246e-05, "loss": 0.169, "step": 17837 }, { "epoch": 1.85, "learning_rate": 3.323538566683068e-05, "loss": 0.9589, "step": 17838 }, { "epoch": 1.85, "learning_rate": 3.3232938591748264e-05, "loss": 0.5665, "step": 17839 }, { "epoch": 1.85, "learning_rate": 3.323049149490247e-05, "loss": 0.7356, "step": 17840 }, { "epoch": 1.85, "learning_rate": 3.322804437630974e-05, "loss": 0.5598, "step": 17841 }, { "epoch": 1.85, "learning_rate": 3.3225597235986585e-05, "loss": 1.306, "step": 17842 }, { "epoch": 1.85, "learning_rate": 3.322315007394946e-05, "loss": 0.4486, "step": 17843 }, { "epoch": 1.85, "learning_rate": 3.322070289021485e-05, "loss": 0.1456, "step": 17844 }, { "epoch": 1.85, "learning_rate": 3.3218255684799206e-05, "loss": 0.2599, "step": 17845 }, { "epoch": 1.85, "learning_rate": 3.321580845771904e-05, "loss": 1.0501, "step": 17846 }, { "epoch": 1.85, "learning_rate": 3.3213361208990787e-05, "loss": 0.676, "step": 17847 }, { "epoch": 1.85, "learning_rate": 3.3210913938630934e-05, "loss": 0.8334, "step": 17848 }, { "epoch": 1.85, "learning_rate": 3.320846664665597e-05, "loss": 0.9338, "step": 17849 }, { "epoch": 1.85, "learning_rate": 3.3206019333082356e-05, "loss": 1.0967, "step": 17850 }, { "epoch": 1.85, "learning_rate": 3.320357199792657e-05, "loss": 0.5375, "step": 17851 }, { "epoch": 1.85, "learning_rate": 3.3201124641205095e-05, "loss": 0.7685, "step": 17852 }, { "epoch": 1.85, "learning_rate": 3.319867726293439e-05, "loss": 0.4228, "step": 17853 }, { "epoch": 1.85, "learning_rate": 3.3196229863130936e-05, "loss": 1.1658, "step": 17854 }, { "epoch": 1.85, "learning_rate": 3.319378244181122e-05, "loss": 0.6834, "step": 17855 }, { "epoch": 1.85, "learning_rate": 3.3191334998991696e-05, "loss": 0.6786, "step": 17856 }, { "epoch": 1.85, "learning_rate": 3.318888753468885e-05, "loss": 0.3274, "step": 17857 }, { "epoch": 1.85, "learning_rate": 3.3186440048919176e-05, "loss": 0.1842, "step": 17858 }, { "epoch": 1.85, "learning_rate": 3.318399254169913e-05, "loss": 0.3531, "step": 17859 }, { "epoch": 1.85, "learning_rate": 3.318154501304518e-05, "loss": 0.8958, "step": 17860 }, { "epoch": 1.85, "learning_rate": 3.317909746297382e-05, "loss": 0.4303, "step": 17861 }, { "epoch": 1.85, "learning_rate": 3.317664989150151e-05, "loss": 0.5221, "step": 17862 }, { "epoch": 1.85, "learning_rate": 3.3174202298644746e-05, "loss": 0.9524, "step": 17863 }, { "epoch": 1.85, "learning_rate": 3.317175468441999e-05, "loss": 0.6441, "step": 17864 }, { "epoch": 1.85, "learning_rate": 3.316930704884373e-05, "loss": 0.8437, "step": 17865 }, { "epoch": 1.85, "learning_rate": 3.316685939193243e-05, "loss": 0.4051, "step": 17866 }, { "epoch": 1.85, "learning_rate": 3.3164411713702576e-05, "loss": 0.744, "step": 17867 }, { "epoch": 1.85, "learning_rate": 3.316196401417065e-05, "loss": 0.5565, "step": 17868 }, { "epoch": 1.85, "learning_rate": 3.315951629335311e-05, "loss": 0.3101, "step": 17869 }, { "epoch": 1.85, "learning_rate": 3.315706855126645e-05, "loss": 1.3455, "step": 17870 }, { "epoch": 1.85, "learning_rate": 3.315462078792715e-05, "loss": 0.6074, "step": 17871 }, { "epoch": 1.85, "learning_rate": 3.315217300335167e-05, "loss": 0.4584, "step": 17872 }, { "epoch": 1.85, "learning_rate": 3.3149725197556503e-05, "loss": 0.4253, "step": 17873 }, { "epoch": 1.85, "learning_rate": 3.314727737055813e-05, "loss": 0.3581, "step": 17874 }, { "epoch": 1.85, "learning_rate": 3.314482952237301e-05, "loss": 0.6688, "step": 17875 }, { "epoch": 1.85, "learning_rate": 3.3142381653017636e-05, "loss": 0.454, "step": 17876 }, { "epoch": 1.85, "learning_rate": 3.313993376250849e-05, "loss": 0.6869, "step": 17877 }, { "epoch": 1.85, "learning_rate": 3.313748585086205e-05, "loss": 0.3709, "step": 17878 }, { "epoch": 1.85, "learning_rate": 3.313503791809478e-05, "loss": 0.3397, "step": 17879 }, { "epoch": 1.85, "learning_rate": 3.3132589964223166e-05, "loss": 0.958, "step": 17880 }, { "epoch": 1.85, "learning_rate": 3.3130141989263686e-05, "loss": 0.6272, "step": 17881 }, { "epoch": 1.85, "learning_rate": 3.312769399323284e-05, "loss": 0.8339, "step": 17882 }, { "epoch": 1.85, "learning_rate": 3.3125245976147076e-05, "loss": 0.6014, "step": 17883 }, { "epoch": 1.85, "learning_rate": 3.312279793802289e-05, "loss": 0.6586, "step": 17884 }, { "epoch": 1.85, "learning_rate": 3.312034987887676e-05, "loss": 0.4153, "step": 17885 }, { "epoch": 1.85, "learning_rate": 3.311790179872517e-05, "loss": 0.7082, "step": 17886 }, { "epoch": 1.85, "learning_rate": 3.3115453697584585e-05, "loss": 0.9111, "step": 17887 }, { "epoch": 1.85, "learning_rate": 3.3113005575471494e-05, "loss": 0.3333, "step": 17888 }, { "epoch": 1.85, "learning_rate": 3.311055743240238e-05, "loss": 0.4192, "step": 17889 }, { "epoch": 1.85, "learning_rate": 3.310810926839373e-05, "loss": 0.4067, "step": 17890 }, { "epoch": 1.85, "learning_rate": 3.310566108346201e-05, "loss": 1.3342, "step": 17891 }, { "epoch": 1.85, "learning_rate": 3.310321287762371e-05, "loss": 0.2275, "step": 17892 }, { "epoch": 1.85, "learning_rate": 3.31007646508953e-05, "loss": 0.7924, "step": 17893 }, { "epoch": 1.85, "learning_rate": 3.309831640329326e-05, "loss": 0.4597, "step": 17894 }, { "epoch": 1.85, "learning_rate": 3.3095868134834095e-05, "loss": 0.7495, "step": 17895 }, { "epoch": 1.85, "learning_rate": 3.3093419845534264e-05, "loss": 0.3152, "step": 17896 }, { "epoch": 1.85, "learning_rate": 3.3090971535410264e-05, "loss": 1.2418, "step": 17897 }, { "epoch": 1.85, "learning_rate": 3.3088523204478554e-05, "loss": 0.4564, "step": 17898 }, { "epoch": 1.85, "learning_rate": 3.308607485275563e-05, "loss": 0.2112, "step": 17899 }, { "epoch": 1.85, "learning_rate": 3.308362648025797e-05, "loss": 0.344, "step": 17900 }, { "epoch": 1.85, "learning_rate": 3.3081178087002065e-05, "loss": 1.1695, "step": 17901 }, { "epoch": 1.85, "learning_rate": 3.3078729673004395e-05, "loss": 0.3796, "step": 17902 }, { "epoch": 1.85, "learning_rate": 3.307628123828142e-05, "loss": 0.6372, "step": 17903 }, { "epoch": 1.85, "learning_rate": 3.307383278284966e-05, "loss": 0.5649, "step": 17904 }, { "epoch": 1.85, "learning_rate": 3.307138430672556e-05, "loss": 0.7487, "step": 17905 }, { "epoch": 1.85, "learning_rate": 3.306893580992563e-05, "loss": 1.0927, "step": 17906 }, { "epoch": 1.85, "learning_rate": 3.306648729246633e-05, "loss": 1.032, "step": 17907 }, { "epoch": 1.85, "learning_rate": 3.3064038754364166e-05, "loss": 0.7845, "step": 17908 }, { "epoch": 1.85, "learning_rate": 3.306159019563561e-05, "loss": 0.7494, "step": 17909 }, { "epoch": 1.86, "learning_rate": 3.305914161629714e-05, "loss": 0.4986, "step": 17910 }, { "epoch": 1.86, "learning_rate": 3.305669301636525e-05, "loss": 1.0116, "step": 17911 }, { "epoch": 1.86, "learning_rate": 3.305424439585641e-05, "loss": 0.605, "step": 17912 }, { "epoch": 1.86, "learning_rate": 3.3051795754787116e-05, "loss": 0.4331, "step": 17913 }, { "epoch": 1.86, "learning_rate": 3.304934709317384e-05, "loss": 0.7825, "step": 17914 }, { "epoch": 1.86, "learning_rate": 3.304689841103308e-05, "loss": 0.1566, "step": 17915 }, { "epoch": 1.86, "learning_rate": 3.3044449708381323e-05, "loss": 0.334, "step": 17916 }, { "epoch": 1.86, "learning_rate": 3.304200098523503e-05, "loss": 0.4204, "step": 17917 }, { "epoch": 1.86, "learning_rate": 3.30395522416107e-05, "loss": 0.6474, "step": 17918 }, { "epoch": 1.86, "learning_rate": 3.3037103477524816e-05, "loss": 1.2712, "step": 17919 }, { "epoch": 1.86, "learning_rate": 3.3034654692993865e-05, "loss": 1.087, "step": 17920 }, { "epoch": 1.86, "learning_rate": 3.303220588803433e-05, "loss": 0.3439, "step": 17921 }, { "epoch": 1.86, "learning_rate": 3.302975706266269e-05, "loss": 0.389, "step": 17922 }, { "epoch": 1.86, "learning_rate": 3.302730821689543e-05, "loss": 1.055, "step": 17923 }, { "epoch": 1.86, "learning_rate": 3.302485935074904e-05, "loss": 0.2042, "step": 17924 }, { "epoch": 1.86, "learning_rate": 3.302241046424001e-05, "loss": 0.4096, "step": 17925 }, { "epoch": 1.86, "learning_rate": 3.301996155738482e-05, "loss": 1.2171, "step": 17926 }, { "epoch": 1.86, "learning_rate": 3.301751263019995e-05, "loss": 0.8345, "step": 17927 }, { "epoch": 1.86, "learning_rate": 3.30150636827019e-05, "loss": 0.2652, "step": 17928 }, { "epoch": 1.86, "learning_rate": 3.301261471490714e-05, "loss": 0.449, "step": 17929 }, { "epoch": 1.86, "learning_rate": 3.3010165726832164e-05, "loss": 0.5398, "step": 17930 }, { "epoch": 1.86, "learning_rate": 3.300771671849346e-05, "loss": 0.9248, "step": 17931 }, { "epoch": 1.86, "learning_rate": 3.3005267689907506e-05, "loss": 0.6917, "step": 17932 }, { "epoch": 1.86, "learning_rate": 3.300281864109079e-05, "loss": 0.7899, "step": 17933 }, { "epoch": 1.86, "learning_rate": 3.300036957205981e-05, "loss": 1.1069, "step": 17934 }, { "epoch": 1.86, "learning_rate": 3.299792048283104e-05, "loss": 0.3565, "step": 17935 }, { "epoch": 1.86, "learning_rate": 3.299547137342097e-05, "loss": 0.9324, "step": 17936 }, { "epoch": 1.86, "learning_rate": 3.299302224384609e-05, "loss": 0.2913, "step": 17937 }, { "epoch": 1.86, "learning_rate": 3.299057309412288e-05, "loss": 0.8475, "step": 17938 }, { "epoch": 1.86, "learning_rate": 3.298812392426784e-05, "loss": 0.6169, "step": 17939 }, { "epoch": 1.86, "learning_rate": 3.298567473429744e-05, "loss": 0.5338, "step": 17940 }, { "epoch": 1.86, "learning_rate": 3.2983225524228184e-05, "loss": 0.9298, "step": 17941 }, { "epoch": 1.86, "learning_rate": 3.2980776294076546e-05, "loss": 0.934, "step": 17942 }, { "epoch": 1.86, "learning_rate": 3.297832704385902e-05, "loss": 0.3781, "step": 17943 }, { "epoch": 1.86, "learning_rate": 3.2975877773592095e-05, "loss": 1.0419, "step": 17944 }, { "epoch": 1.86, "learning_rate": 3.2973428483292255e-05, "loss": 0.4919, "step": 17945 }, { "epoch": 1.86, "learning_rate": 3.2970979172975996e-05, "loss": 0.4668, "step": 17946 }, { "epoch": 1.86, "learning_rate": 3.296852984265979e-05, "loss": 0.4582, "step": 17947 }, { "epoch": 1.86, "learning_rate": 3.296608049236014e-05, "loss": 0.8475, "step": 17948 }, { "epoch": 1.86, "learning_rate": 3.296363112209354e-05, "loss": 0.5825, "step": 17949 }, { "epoch": 1.86, "learning_rate": 3.296118173187646e-05, "loss": 0.9062, "step": 17950 }, { "epoch": 1.86, "learning_rate": 3.2958732321725395e-05, "loss": 0.5813, "step": 17951 }, { "epoch": 1.86, "learning_rate": 3.295628289165684e-05, "loss": 0.4915, "step": 17952 }, { "epoch": 1.86, "learning_rate": 3.295383344168728e-05, "loss": 0.728, "step": 17953 }, { "epoch": 1.86, "learning_rate": 3.2951383971833214e-05, "loss": 0.3716, "step": 17954 }, { "epoch": 1.86, "learning_rate": 3.29489344821111e-05, "loss": 0.8148, "step": 17955 }, { "epoch": 1.86, "learning_rate": 3.294648497253746e-05, "loss": 0.266, "step": 17956 }, { "epoch": 1.86, "learning_rate": 3.2944035443128774e-05, "loss": 1.4157, "step": 17957 }, { "epoch": 1.86, "learning_rate": 3.294158589390153e-05, "loss": 0.522, "step": 17958 }, { "epoch": 1.86, "learning_rate": 3.2939136324872226e-05, "loss": 1.1304, "step": 17959 }, { "epoch": 1.86, "learning_rate": 3.2936686736057334e-05, "loss": 0.7927, "step": 17960 }, { "epoch": 1.86, "learning_rate": 3.293423712747336e-05, "loss": 0.9834, "step": 17961 }, { "epoch": 1.86, "learning_rate": 3.293178749913677e-05, "loss": 0.4754, "step": 17962 }, { "epoch": 1.86, "learning_rate": 3.29293378510641e-05, "loss": 0.9548, "step": 17963 }, { "epoch": 1.86, "learning_rate": 3.292688818327179e-05, "loss": 0.7596, "step": 17964 }, { "epoch": 1.86, "learning_rate": 3.292443849577637e-05, "loss": 0.624, "step": 17965 }, { "epoch": 1.86, "learning_rate": 3.292198878859431e-05, "loss": 0.3195, "step": 17966 }, { "epoch": 1.86, "learning_rate": 3.29195390617421e-05, "loss": 0.9483, "step": 17967 }, { "epoch": 1.86, "learning_rate": 3.291708931523624e-05, "loss": 0.3549, "step": 17968 }, { "epoch": 1.86, "learning_rate": 3.291463954909321e-05, "loss": 0.8154, "step": 17969 }, { "epoch": 1.86, "learning_rate": 3.2912189763329514e-05, "loss": 1.0333, "step": 17970 }, { "epoch": 1.86, "learning_rate": 3.2909739957961643e-05, "loss": 0.2215, "step": 17971 }, { "epoch": 1.86, "learning_rate": 3.2907290133006075e-05, "loss": 0.4646, "step": 17972 }, { "epoch": 1.86, "learning_rate": 3.2904840288479317e-05, "loss": 0.7294, "step": 17973 }, { "epoch": 1.86, "learning_rate": 3.290239042439785e-05, "loss": 0.7052, "step": 17974 }, { "epoch": 1.86, "learning_rate": 3.2899940540778166e-05, "loss": 1.2286, "step": 17975 }, { "epoch": 1.86, "learning_rate": 3.289749063763677e-05, "loss": 0.9441, "step": 17976 }, { "epoch": 1.86, "learning_rate": 3.289504071499014e-05, "loss": 1.4198, "step": 17977 }, { "epoch": 1.86, "learning_rate": 3.289259077285478e-05, "loss": 0.3889, "step": 17978 }, { "epoch": 1.86, "learning_rate": 3.289014081124716e-05, "loss": 0.301, "step": 17979 }, { "epoch": 1.86, "learning_rate": 3.28876908301838e-05, "loss": 1.1283, "step": 17980 }, { "epoch": 1.86, "learning_rate": 3.2885240829681184e-05, "loss": 0.8619, "step": 17981 }, { "epoch": 1.86, "learning_rate": 3.2882790809755795e-05, "loss": 0.9666, "step": 17982 }, { "epoch": 1.86, "learning_rate": 3.288034077042413e-05, "loss": 0.3859, "step": 17983 }, { "epoch": 1.86, "learning_rate": 3.28778907117027e-05, "loss": 0.3511, "step": 17984 }, { "epoch": 1.86, "learning_rate": 3.287544063360796e-05, "loss": 0.8946, "step": 17985 }, { "epoch": 1.86, "learning_rate": 3.287299053615644e-05, "loss": 0.1536, "step": 17986 }, { "epoch": 1.86, "learning_rate": 3.287054041936462e-05, "loss": 0.1581, "step": 17987 }, { "epoch": 1.86, "learning_rate": 3.2868090283248994e-05, "loss": 1.0052, "step": 17988 }, { "epoch": 1.86, "learning_rate": 3.286564012782605e-05, "loss": 0.4, "step": 17989 }, { "epoch": 1.86, "learning_rate": 3.28631899531123e-05, "loss": 0.9, "step": 17990 }, { "epoch": 1.86, "learning_rate": 3.286073975912421e-05, "loss": 0.4664, "step": 17991 }, { "epoch": 1.86, "learning_rate": 3.28582895458783e-05, "loss": 0.4752, "step": 17992 }, { "epoch": 1.86, "learning_rate": 3.285583931339105e-05, "loss": 0.3566, "step": 17993 }, { "epoch": 1.86, "learning_rate": 3.285338906167895e-05, "loss": 0.5864, "step": 17994 }, { "epoch": 1.86, "learning_rate": 3.28509387907585e-05, "loss": 0.3814, "step": 17995 }, { "epoch": 1.86, "learning_rate": 3.284848850064621e-05, "loss": 0.6799, "step": 17996 }, { "epoch": 1.86, "learning_rate": 3.284603819135857e-05, "loss": 0.7548, "step": 17997 }, { "epoch": 1.86, "learning_rate": 3.284358786291205e-05, "loss": 0.6344, "step": 17998 }, { "epoch": 1.86, "learning_rate": 3.284113751532317e-05, "loss": 0.5549, "step": 17999 }, { "epoch": 1.86, "learning_rate": 3.2838687148608406e-05, "loss": 0.342, "step": 18000 }, { "epoch": 1.86, "learning_rate": 3.283623676278428e-05, "loss": 0.7669, "step": 18001 }, { "epoch": 1.86, "learning_rate": 3.283378635786726e-05, "loss": 1.0238, "step": 18002 }, { "epoch": 1.86, "learning_rate": 3.283133593387386e-05, "loss": 0.6233, "step": 18003 }, { "epoch": 1.86, "learning_rate": 3.2828885490820565e-05, "loss": 0.5871, "step": 18004 }, { "epoch": 1.86, "learning_rate": 3.2826435028723874e-05, "loss": 0.3128, "step": 18005 }, { "epoch": 1.87, "learning_rate": 3.282398454760029e-05, "loss": 1.1364, "step": 18006 }, { "epoch": 1.87, "learning_rate": 3.2821534047466294e-05, "loss": 1.5853, "step": 18007 }, { "epoch": 1.87, "learning_rate": 3.28190835283384e-05, "loss": 0.3566, "step": 18008 }, { "epoch": 1.87, "learning_rate": 3.281663299023309e-05, "loss": 1.0902, "step": 18009 }, { "epoch": 1.87, "learning_rate": 3.281418243316686e-05, "loss": 0.5064, "step": 18010 }, { "epoch": 1.87, "learning_rate": 3.2811731857156225e-05, "loss": 0.2754, "step": 18011 }, { "epoch": 1.87, "learning_rate": 3.280928126221766e-05, "loss": 1.0687, "step": 18012 }, { "epoch": 1.87, "learning_rate": 3.280683064836767e-05, "loss": 0.7257, "step": 18013 }, { "epoch": 1.87, "learning_rate": 3.280438001562276e-05, "loss": 0.1806, "step": 18014 }, { "epoch": 1.87, "learning_rate": 3.280192936399941e-05, "loss": 0.6199, "step": 18015 }, { "epoch": 1.87, "learning_rate": 3.2799478693514144e-05, "loss": 0.5729, "step": 18016 }, { "epoch": 1.87, "learning_rate": 3.279702800418343e-05, "loss": 0.8045, "step": 18017 }, { "epoch": 1.87, "learning_rate": 3.2794577296023775e-05, "loss": 0.56, "step": 18018 }, { "epoch": 1.87, "learning_rate": 3.279212656905169e-05, "loss": 0.4076, "step": 18019 }, { "epoch": 1.87, "learning_rate": 3.2789675823283655e-05, "loss": 0.3704, "step": 18020 }, { "epoch": 1.87, "learning_rate": 3.278722505873618e-05, "loss": 0.7702, "step": 18021 }, { "epoch": 1.87, "learning_rate": 3.278477427542575e-05, "loss": 0.3177, "step": 18022 }, { "epoch": 1.87, "learning_rate": 3.278232347336888e-05, "loss": 0.5166, "step": 18023 }, { "epoch": 1.87, "learning_rate": 3.277987265258206e-05, "loss": 1.0041, "step": 18024 }, { "epoch": 1.87, "learning_rate": 3.277742181308178e-05, "loss": 0.6532, "step": 18025 }, { "epoch": 1.87, "learning_rate": 3.277497095488455e-05, "loss": 1.0651, "step": 18026 }, { "epoch": 1.87, "learning_rate": 3.277252007800686e-05, "loss": 0.8992, "step": 18027 }, { "epoch": 1.87, "learning_rate": 3.2770069182465224e-05, "loss": 0.3942, "step": 18028 }, { "epoch": 1.87, "learning_rate": 3.276761826827612e-05, "loss": 0.5489, "step": 18029 }, { "epoch": 1.87, "learning_rate": 3.276516733545607e-05, "loss": 0.9715, "step": 18030 }, { "epoch": 1.87, "learning_rate": 3.2762716384021544e-05, "loss": 0.3286, "step": 18031 }, { "epoch": 1.87, "learning_rate": 3.2760265413989064e-05, "loss": 0.4939, "step": 18032 }, { "epoch": 1.87, "learning_rate": 3.2757814425375126e-05, "loss": 0.1805, "step": 18033 }, { "epoch": 1.87, "learning_rate": 3.2755363418196226e-05, "loss": 0.5152, "step": 18034 }, { "epoch": 1.87, "learning_rate": 3.275291239246887e-05, "loss": 0.1937, "step": 18035 }, { "epoch": 1.87, "learning_rate": 3.2750461348209537e-05, "loss": 1.2757, "step": 18036 }, { "epoch": 1.87, "learning_rate": 3.2748010285434755e-05, "loss": 0.3142, "step": 18037 }, { "epoch": 1.87, "learning_rate": 3.2745559204161e-05, "loss": 0.5138, "step": 18038 }, { "epoch": 1.87, "learning_rate": 3.274310810440479e-05, "loss": 0.4653, "step": 18039 }, { "epoch": 1.87, "learning_rate": 3.274065698618262e-05, "loss": 0.9999, "step": 18040 }, { "epoch": 1.87, "learning_rate": 3.273820584951099e-05, "loss": 0.5818, "step": 18041 }, { "epoch": 1.87, "learning_rate": 3.2735754694406385e-05, "loss": 0.2161, "step": 18042 }, { "epoch": 1.87, "learning_rate": 3.273330352088533e-05, "loss": 0.9156, "step": 18043 }, { "epoch": 1.87, "learning_rate": 3.273085232896432e-05, "loss": 0.4069, "step": 18044 }, { "epoch": 1.87, "learning_rate": 3.272840111865984e-05, "loss": 0.823, "step": 18045 }, { "epoch": 1.87, "learning_rate": 3.2725949889988406e-05, "loss": 0.8468, "step": 18046 }, { "epoch": 1.87, "learning_rate": 3.272349864296652e-05, "loss": 0.826, "step": 18047 }, { "epoch": 1.87, "learning_rate": 3.272104737761068e-05, "loss": 1.0713, "step": 18048 }, { "epoch": 1.87, "learning_rate": 3.271859609393738e-05, "loss": 0.7965, "step": 18049 }, { "epoch": 1.87, "learning_rate": 3.2716144791963124e-05, "loss": 0.3838, "step": 18050 }, { "epoch": 1.87, "learning_rate": 3.271369347170442e-05, "loss": 1.1653, "step": 18051 }, { "epoch": 1.87, "learning_rate": 3.2711242133177765e-05, "loss": 0.6206, "step": 18052 }, { "epoch": 1.87, "learning_rate": 3.270879077639966e-05, "loss": 0.2201, "step": 18053 }, { "epoch": 1.87, "learning_rate": 3.270633940138662e-05, "loss": 0.2055, "step": 18054 }, { "epoch": 1.87, "learning_rate": 3.270388800815513e-05, "loss": 0.2244, "step": 18055 }, { "epoch": 1.87, "learning_rate": 3.27014365967217e-05, "loss": 0.6023, "step": 18056 }, { "epoch": 1.87, "learning_rate": 3.269898516710283e-05, "loss": 0.8892, "step": 18057 }, { "epoch": 1.87, "learning_rate": 3.269653371931502e-05, "loss": 0.3607, "step": 18058 }, { "epoch": 1.87, "learning_rate": 3.269408225337479e-05, "loss": 0.9534, "step": 18059 }, { "epoch": 1.87, "learning_rate": 3.269163076929861e-05, "loss": 0.5096, "step": 18060 }, { "epoch": 1.87, "learning_rate": 3.2689179267103005e-05, "loss": 0.2535, "step": 18061 }, { "epoch": 1.87, "learning_rate": 3.2686727746804474e-05, "loss": 0.3027, "step": 18062 }, { "epoch": 1.87, "learning_rate": 3.2684276208419534e-05, "loss": 0.7551, "step": 18063 }, { "epoch": 1.87, "learning_rate": 3.268182465196466e-05, "loss": 0.9575, "step": 18064 }, { "epoch": 1.87, "learning_rate": 3.2679373077456376e-05, "loss": 0.7366, "step": 18065 }, { "epoch": 1.87, "learning_rate": 3.267692148491118e-05, "loss": 0.8368, "step": 18066 }, { "epoch": 1.87, "learning_rate": 3.2674469874345574e-05, "loss": 1.35, "step": 18067 }, { "epoch": 1.87, "learning_rate": 3.267201824577606e-05, "loss": 0.3812, "step": 18068 }, { "epoch": 1.87, "learning_rate": 3.266956659921915e-05, "loss": 0.3179, "step": 18069 }, { "epoch": 1.87, "learning_rate": 3.266711493469134e-05, "loss": 0.6458, "step": 18070 }, { "epoch": 1.87, "learning_rate": 3.2664663252209127e-05, "loss": 0.4045, "step": 18071 }, { "epoch": 1.87, "learning_rate": 3.266221155178904e-05, "loss": 0.411, "step": 18072 }, { "epoch": 1.87, "learning_rate": 3.2659759833447565e-05, "loss": 0.2006, "step": 18073 }, { "epoch": 1.87, "learning_rate": 3.2657308097201196e-05, "loss": 0.5655, "step": 18074 }, { "epoch": 1.87, "learning_rate": 3.265485634306646e-05, "loss": 0.7689, "step": 18075 }, { "epoch": 1.87, "learning_rate": 3.2652404571059845e-05, "loss": 0.5563, "step": 18076 }, { "epoch": 1.87, "learning_rate": 3.2649952781197883e-05, "loss": 0.2535, "step": 18077 }, { "epoch": 1.87, "learning_rate": 3.264750097349705e-05, "loss": 1.0617, "step": 18078 }, { "epoch": 1.87, "learning_rate": 3.264504914797385e-05, "loss": 0.1703, "step": 18079 }, { "epoch": 1.87, "learning_rate": 3.26425973046448e-05, "loss": 0.9633, "step": 18080 }, { "epoch": 1.87, "learning_rate": 3.26401454435264e-05, "loss": 0.9604, "step": 18081 }, { "epoch": 1.87, "learning_rate": 3.263769356463517e-05, "loss": 0.2147, "step": 18082 }, { "epoch": 1.87, "learning_rate": 3.26352416679876e-05, "loss": 1.1255, "step": 18083 }, { "epoch": 1.87, "learning_rate": 3.263278975360019e-05, "loss": 0.6927, "step": 18084 }, { "epoch": 1.87, "learning_rate": 3.2630337821489476e-05, "loss": 0.5754, "step": 18085 }, { "epoch": 1.87, "learning_rate": 3.262788587167193e-05, "loss": 1.2553, "step": 18086 }, { "epoch": 1.87, "learning_rate": 3.262543390416408e-05, "loss": 0.7028, "step": 18087 }, { "epoch": 1.87, "learning_rate": 3.262298191898241e-05, "loss": 0.9633, "step": 18088 }, { "epoch": 1.87, "learning_rate": 3.2620529916143446e-05, "loss": 0.496, "step": 18089 }, { "epoch": 1.87, "learning_rate": 3.261807789566368e-05, "loss": 1.1338, "step": 18090 }, { "epoch": 1.87, "learning_rate": 3.261562585755964e-05, "loss": 0.5101, "step": 18091 }, { "epoch": 1.87, "learning_rate": 3.261317380184781e-05, "loss": 0.2877, "step": 18092 }, { "epoch": 1.87, "learning_rate": 3.261072172854471e-05, "loss": 0.5791, "step": 18093 }, { "epoch": 1.87, "learning_rate": 3.260826963766684e-05, "loss": 0.4909, "step": 18094 }, { "epoch": 1.87, "learning_rate": 3.2605817529230704e-05, "loss": 0.7454, "step": 18095 }, { "epoch": 1.87, "learning_rate": 3.260336540325282e-05, "loss": 0.8607, "step": 18096 }, { "epoch": 1.87, "learning_rate": 3.26009132597497e-05, "loss": 0.6883, "step": 18097 }, { "epoch": 1.87, "learning_rate": 3.259846109873782e-05, "loss": 1.0735, "step": 18098 }, { "epoch": 1.87, "learning_rate": 3.259600892023372e-05, "loss": 0.5814, "step": 18099 }, { "epoch": 1.87, "learning_rate": 3.259355672425389e-05, "loss": 0.7216, "step": 18100 }, { "epoch": 1.87, "learning_rate": 3.2591104510814845e-05, "loss": 0.3223, "step": 18101 }, { "epoch": 1.88, "learning_rate": 3.258865227993309e-05, "loss": 0.4032, "step": 18102 }, { "epoch": 1.88, "learning_rate": 3.258620003162514e-05, "loss": 0.5742, "step": 18103 }, { "epoch": 1.88, "learning_rate": 3.2583747765907485e-05, "loss": 0.3063, "step": 18104 }, { "epoch": 1.88, "learning_rate": 3.258129548279665e-05, "loss": 1.1824, "step": 18105 }, { "epoch": 1.88, "learning_rate": 3.257884318230914e-05, "loss": 0.7942, "step": 18106 }, { "epoch": 1.88, "learning_rate": 3.257639086446146e-05, "loss": 0.8002, "step": 18107 }, { "epoch": 1.88, "learning_rate": 3.257393852927011e-05, "loss": 0.431, "step": 18108 }, { "epoch": 1.88, "learning_rate": 3.257148617675162e-05, "loss": 1.0892, "step": 18109 }, { "epoch": 1.88, "learning_rate": 3.2569033806922484e-05, "loss": 0.375, "step": 18110 }, { "epoch": 1.88, "learning_rate": 3.2566581419799214e-05, "loss": 0.279, "step": 18111 }, { "epoch": 1.88, "learning_rate": 3.256412901539832e-05, "loss": 0.1668, "step": 18112 }, { "epoch": 1.88, "learning_rate": 3.25616765937363e-05, "loss": 0.3887, "step": 18113 }, { "epoch": 1.88, "learning_rate": 3.2559224154829685e-05, "loss": 0.5843, "step": 18114 }, { "epoch": 1.88, "learning_rate": 3.255677169869497e-05, "loss": 0.8111, "step": 18115 }, { "epoch": 1.88, "learning_rate": 3.255431922534866e-05, "loss": 1.0165, "step": 18116 }, { "epoch": 1.88, "learning_rate": 3.255186673480727e-05, "loss": 0.2738, "step": 18117 }, { "epoch": 1.88, "learning_rate": 3.254941422708732e-05, "loss": 0.4256, "step": 18118 }, { "epoch": 1.88, "learning_rate": 3.25469617022053e-05, "loss": 0.3361, "step": 18119 }, { "epoch": 1.88, "learning_rate": 3.2544509160177736e-05, "loss": 0.2886, "step": 18120 }, { "epoch": 1.88, "learning_rate": 3.2542056601021126e-05, "loss": 0.8445, "step": 18121 }, { "epoch": 1.88, "learning_rate": 3.253960402475199e-05, "loss": 0.3197, "step": 18122 }, { "epoch": 1.88, "learning_rate": 3.253715143138683e-05, "loss": 0.7657, "step": 18123 }, { "epoch": 1.88, "learning_rate": 3.2534698820942164e-05, "loss": 0.7037, "step": 18124 }, { "epoch": 1.88, "learning_rate": 3.25322461934345e-05, "loss": 0.8353, "step": 18125 }, { "epoch": 1.88, "learning_rate": 3.252979354888034e-05, "loss": 1.1532, "step": 18126 }, { "epoch": 1.88, "learning_rate": 3.2527340887296196e-05, "loss": 0.8513, "step": 18127 }, { "epoch": 1.88, "learning_rate": 3.25248882086986e-05, "loss": 0.7496, "step": 18128 }, { "epoch": 1.88, "learning_rate": 3.2522435513104046e-05, "loss": 1.5444, "step": 18129 }, { "epoch": 1.88, "learning_rate": 3.251998280052905e-05, "loss": 0.8755, "step": 18130 }, { "epoch": 1.88, "learning_rate": 3.2517530070990106e-05, "loss": 0.6149, "step": 18131 }, { "epoch": 1.88, "learning_rate": 3.251507732450374e-05, "loss": 0.3664, "step": 18132 }, { "epoch": 1.88, "learning_rate": 3.2512624561086466e-05, "loss": 0.7029, "step": 18133 }, { "epoch": 1.88, "learning_rate": 3.2510171780754794e-05, "loss": 1.0654, "step": 18134 }, { "epoch": 1.88, "learning_rate": 3.2507718983525235e-05, "loss": 0.3211, "step": 18135 }, { "epoch": 1.88, "learning_rate": 3.2505266169414296e-05, "loss": 0.9724, "step": 18136 }, { "epoch": 1.88, "learning_rate": 3.2502813338438484e-05, "loss": 0.9933, "step": 18137 }, { "epoch": 1.88, "learning_rate": 3.250036049061432e-05, "loss": 0.4726, "step": 18138 }, { "epoch": 1.88, "learning_rate": 3.249790762595832e-05, "loss": 0.9983, "step": 18139 }, { "epoch": 1.88, "learning_rate": 3.2495454744486986e-05, "loss": 0.5202, "step": 18140 }, { "epoch": 1.88, "learning_rate": 3.2493001846216846e-05, "loss": 1.1499, "step": 18141 }, { "epoch": 1.88, "learning_rate": 3.249054893116438e-05, "loss": 0.3614, "step": 18142 }, { "epoch": 1.88, "learning_rate": 3.248809599934614e-05, "loss": 0.5901, "step": 18143 }, { "epoch": 1.88, "learning_rate": 3.248564305077861e-05, "loss": 0.6251, "step": 18144 }, { "epoch": 1.88, "learning_rate": 3.2483190085478317e-05, "loss": 0.4461, "step": 18145 }, { "epoch": 1.88, "learning_rate": 3.2480737103461764e-05, "loss": 0.1344, "step": 18146 }, { "epoch": 1.88, "learning_rate": 3.2478284104745477e-05, "loss": 1.3041, "step": 18147 }, { "epoch": 1.88, "learning_rate": 3.247583108934596e-05, "loss": 0.867, "step": 18148 }, { "epoch": 1.88, "learning_rate": 3.247337805727972e-05, "loss": 0.2644, "step": 18149 }, { "epoch": 1.88, "learning_rate": 3.247092500856328e-05, "loss": 0.7894, "step": 18150 }, { "epoch": 1.88, "learning_rate": 3.2468471943213155e-05, "loss": 0.3052, "step": 18151 }, { "epoch": 1.88, "learning_rate": 3.2466018861245855e-05, "loss": 0.5753, "step": 18152 }, { "epoch": 1.88, "learning_rate": 3.246356576267789e-05, "loss": 1.1691, "step": 18153 }, { "epoch": 1.88, "learning_rate": 3.246111264752578e-05, "loss": 0.3486, "step": 18154 }, { "epoch": 1.88, "learning_rate": 3.245865951580603e-05, "loss": 0.1711, "step": 18155 }, { "epoch": 1.88, "learning_rate": 3.2456206367535164e-05, "loss": 0.6966, "step": 18156 }, { "epoch": 1.88, "learning_rate": 3.245375320272968e-05, "loss": 0.3445, "step": 18157 }, { "epoch": 1.88, "learning_rate": 3.245130002140612e-05, "loss": 1.0758, "step": 18158 }, { "epoch": 1.88, "learning_rate": 3.2448846823580976e-05, "loss": 0.3092, "step": 18159 }, { "epoch": 1.88, "learning_rate": 3.2446393609270774e-05, "loss": 0.9793, "step": 18160 }, { "epoch": 1.88, "learning_rate": 3.2443940378492016e-05, "loss": 0.6849, "step": 18161 }, { "epoch": 1.88, "learning_rate": 3.2441487131261215e-05, "loss": 0.5421, "step": 18162 }, { "epoch": 1.88, "learning_rate": 3.2439033867594905e-05, "loss": 0.2369, "step": 18163 }, { "epoch": 1.88, "learning_rate": 3.2436580587509586e-05, "loss": 0.204, "step": 18164 }, { "epoch": 1.88, "learning_rate": 3.243412729102177e-05, "loss": 1.1286, "step": 18165 }, { "epoch": 1.88, "learning_rate": 3.2431673978148e-05, "loss": 0.427, "step": 18166 }, { "epoch": 1.88, "learning_rate": 3.242922064890475e-05, "loss": 0.3112, "step": 18167 }, { "epoch": 1.88, "learning_rate": 3.242676730330856e-05, "loss": 0.2042, "step": 18168 }, { "epoch": 1.88, "learning_rate": 3.242431394137594e-05, "loss": 1.1032, "step": 18169 }, { "epoch": 1.88, "learning_rate": 3.2421860563123407e-05, "loss": 0.6672, "step": 18170 }, { "epoch": 1.88, "learning_rate": 3.2419407168567474e-05, "loss": 0.2629, "step": 18171 }, { "epoch": 1.88, "learning_rate": 3.241695375772466e-05, "loss": 0.3666, "step": 18172 }, { "epoch": 1.88, "learning_rate": 3.241450033061149e-05, "loss": 0.2803, "step": 18173 }, { "epoch": 1.88, "learning_rate": 3.241204688724445e-05, "loss": 0.1644, "step": 18174 }, { "epoch": 1.88, "learning_rate": 3.2409593427640084e-05, "loss": 0.7268, "step": 18175 }, { "epoch": 1.88, "learning_rate": 3.240713995181489e-05, "loss": 0.502, "step": 18176 }, { "epoch": 1.88, "learning_rate": 3.24046864597854e-05, "loss": 0.8323, "step": 18177 }, { "epoch": 1.88, "learning_rate": 3.240223295156812e-05, "loss": 0.707, "step": 18178 }, { "epoch": 1.88, "learning_rate": 3.239977942717958e-05, "loss": 0.8687, "step": 18179 }, { "epoch": 1.88, "learning_rate": 3.239732588663628e-05, "loss": 0.4506, "step": 18180 }, { "epoch": 1.88, "learning_rate": 3.239487232995474e-05, "loss": 0.3416, "step": 18181 }, { "epoch": 1.88, "learning_rate": 3.2392418757151484e-05, "loss": 0.7149, "step": 18182 }, { "epoch": 1.88, "learning_rate": 3.238996516824302e-05, "loss": 0.3028, "step": 18183 }, { "epoch": 1.88, "learning_rate": 3.2387511563245866e-05, "loss": 1.1353, "step": 18184 }, { "epoch": 1.88, "learning_rate": 3.238505794217656e-05, "loss": 0.2011, "step": 18185 }, { "epoch": 1.88, "learning_rate": 3.238260430505159e-05, "loss": 0.2682, "step": 18186 }, { "epoch": 1.88, "learning_rate": 3.238015065188749e-05, "loss": 0.8949, "step": 18187 }, { "epoch": 1.88, "learning_rate": 3.237769698270077e-05, "loss": 0.3552, "step": 18188 }, { "epoch": 1.88, "learning_rate": 3.237524329750794e-05, "loss": 0.2474, "step": 18189 }, { "epoch": 1.88, "learning_rate": 3.237278959632553e-05, "loss": 0.7022, "step": 18190 }, { "epoch": 1.88, "learning_rate": 3.2370335879170075e-05, "loss": 0.4503, "step": 18191 }, { "epoch": 1.88, "learning_rate": 3.236788214605807e-05, "loss": 0.7525, "step": 18192 }, { "epoch": 1.88, "learning_rate": 3.236542839700602e-05, "loss": 1.1393, "step": 18193 }, { "epoch": 1.88, "learning_rate": 3.236297463203047e-05, "loss": 0.3678, "step": 18194 }, { "epoch": 1.88, "learning_rate": 3.236052085114792e-05, "loss": 0.5787, "step": 18195 }, { "epoch": 1.88, "learning_rate": 3.235806705437491e-05, "loss": 0.0609, "step": 18196 }, { "epoch": 1.89, "learning_rate": 3.2355613241727924e-05, "loss": 0.1506, "step": 18197 }, { "epoch": 1.89, "learning_rate": 3.2353159413223514e-05, "loss": 0.9455, "step": 18198 }, { "epoch": 1.89, "learning_rate": 3.2350705568878185e-05, "loss": 0.6061, "step": 18199 }, { "epoch": 1.89, "learning_rate": 3.234825170870845e-05, "loss": 0.2488, "step": 18200 }, { "epoch": 1.89, "learning_rate": 3.2345797832730846e-05, "loss": 0.395, "step": 18201 }, { "epoch": 1.89, "learning_rate": 3.234334394096187e-05, "loss": 0.3178, "step": 18202 }, { "epoch": 1.89, "learning_rate": 3.234089003341805e-05, "loss": 0.7704, "step": 18203 }, { "epoch": 1.89, "learning_rate": 3.233843611011592e-05, "loss": 0.4381, "step": 18204 }, { "epoch": 1.89, "learning_rate": 3.233598217107197e-05, "loss": 0.7178, "step": 18205 }, { "epoch": 1.89, "learning_rate": 3.233352821630274e-05, "loss": 0.7737, "step": 18206 }, { "epoch": 1.89, "learning_rate": 3.2331074245824745e-05, "loss": 0.4555, "step": 18207 }, { "epoch": 1.89, "learning_rate": 3.2328620259654496e-05, "loss": 0.3274, "step": 18208 }, { "epoch": 1.89, "learning_rate": 3.232616625780853e-05, "loss": 0.7189, "step": 18209 }, { "epoch": 1.89, "learning_rate": 3.232371224030335e-05, "loss": 0.7665, "step": 18210 }, { "epoch": 1.89, "learning_rate": 3.23212582071555e-05, "loss": 0.2343, "step": 18211 }, { "epoch": 1.89, "learning_rate": 3.231880415838146e-05, "loss": 0.2022, "step": 18212 }, { "epoch": 1.89, "learning_rate": 3.231635009399778e-05, "loss": 1.0182, "step": 18213 }, { "epoch": 1.89, "learning_rate": 3.231389601402097e-05, "loss": 0.5921, "step": 18214 }, { "epoch": 1.89, "learning_rate": 3.231144191846757e-05, "loss": 0.9995, "step": 18215 }, { "epoch": 1.89, "learning_rate": 3.230898780735407e-05, "loss": 0.6429, "step": 18216 }, { "epoch": 1.89, "learning_rate": 3.2306533680697005e-05, "loss": 0.2981, "step": 18217 }, { "epoch": 1.89, "learning_rate": 3.2304079538512896e-05, "loss": 0.8234, "step": 18218 }, { "epoch": 1.89, "learning_rate": 3.230162538081826e-05, "loss": 0.3302, "step": 18219 }, { "epoch": 1.89, "learning_rate": 3.229917120762963e-05, "loss": 0.9975, "step": 18220 }, { "epoch": 1.89, "learning_rate": 3.229671701896351e-05, "loss": 1.0472, "step": 18221 }, { "epoch": 1.89, "learning_rate": 3.229426281483643e-05, "loss": 0.5218, "step": 18222 }, { "epoch": 1.89, "learning_rate": 3.229180859526491e-05, "loss": 0.5984, "step": 18223 }, { "epoch": 1.89, "learning_rate": 3.2289354360265464e-05, "loss": 0.4214, "step": 18224 }, { "epoch": 1.89, "learning_rate": 3.2286900109854634e-05, "loss": 0.4141, "step": 18225 }, { "epoch": 1.89, "learning_rate": 3.228444584404891e-05, "loss": 0.413, "step": 18226 }, { "epoch": 1.89, "learning_rate": 3.228199156286485e-05, "loss": 0.3552, "step": 18227 }, { "epoch": 1.89, "learning_rate": 3.227953726631894e-05, "loss": 0.8259, "step": 18228 }, { "epoch": 1.89, "learning_rate": 3.227708295442773e-05, "loss": 0.5504, "step": 18229 }, { "epoch": 1.89, "learning_rate": 3.227462862720773e-05, "loss": 0.6297, "step": 18230 }, { "epoch": 1.89, "learning_rate": 3.2272174284675456e-05, "loss": 0.2858, "step": 18231 }, { "epoch": 1.89, "learning_rate": 3.226971992684743e-05, "loss": 0.5868, "step": 18232 }, { "epoch": 1.89, "learning_rate": 3.226726555374019e-05, "loss": 0.0982, "step": 18233 }, { "epoch": 1.89, "learning_rate": 3.226481116537026e-05, "loss": 0.692, "step": 18234 }, { "epoch": 1.89, "learning_rate": 3.226235676175414e-05, "loss": 0.489, "step": 18235 }, { "epoch": 1.89, "learning_rate": 3.225990234290836e-05, "loss": 0.8691, "step": 18236 }, { "epoch": 1.89, "learning_rate": 3.225744790884945e-05, "loss": 0.8045, "step": 18237 }, { "epoch": 1.89, "learning_rate": 3.225499345959392e-05, "loss": 1.1041, "step": 18238 }, { "epoch": 1.89, "learning_rate": 3.2252538995158316e-05, "loss": 0.3138, "step": 18239 }, { "epoch": 1.89, "learning_rate": 3.225008451555914e-05, "loss": 0.3135, "step": 18240 }, { "epoch": 1.89, "learning_rate": 3.224763002081292e-05, "loss": 0.388, "step": 18241 }, { "epoch": 1.89, "learning_rate": 3.224517551093618e-05, "loss": 0.5095, "step": 18242 }, { "epoch": 1.89, "learning_rate": 3.224272098594545e-05, "loss": 0.8293, "step": 18243 }, { "epoch": 1.89, "learning_rate": 3.224026644585724e-05, "loss": 0.443, "step": 18244 }, { "epoch": 1.89, "learning_rate": 3.2237811890688085e-05, "loss": 1.006, "step": 18245 }, { "epoch": 1.89, "learning_rate": 3.223535732045451e-05, "loss": 1.068, "step": 18246 }, { "epoch": 1.89, "learning_rate": 3.223290273517302e-05, "loss": 0.3901, "step": 18247 }, { "epoch": 1.89, "learning_rate": 3.2230448134860154e-05, "loss": 0.1922, "step": 18248 }, { "epoch": 1.89, "learning_rate": 3.222799351953245e-05, "loss": 0.2887, "step": 18249 }, { "epoch": 1.89, "learning_rate": 3.2225538889206395e-05, "loss": 0.5896, "step": 18250 }, { "epoch": 1.89, "learning_rate": 3.222308424389853e-05, "loss": 0.1913, "step": 18251 }, { "epoch": 1.89, "learning_rate": 3.2220629583625386e-05, "loss": 0.85, "step": 18252 }, { "epoch": 1.89, "learning_rate": 3.22181749084035e-05, "loss": 0.7411, "step": 18253 }, { "epoch": 1.89, "learning_rate": 3.221572021824937e-05, "loss": 0.4986, "step": 18254 }, { "epoch": 1.89, "learning_rate": 3.221326551317953e-05, "loss": 0.8045, "step": 18255 }, { "epoch": 1.89, "learning_rate": 3.2210810793210496e-05, "loss": 0.9463, "step": 18256 }, { "epoch": 1.89, "learning_rate": 3.22083560583588e-05, "loss": 0.7322, "step": 18257 }, { "epoch": 1.89, "learning_rate": 3.2205901308640976e-05, "loss": 0.8514, "step": 18258 }, { "epoch": 1.89, "learning_rate": 3.2203446544073545e-05, "loss": 0.1743, "step": 18259 }, { "epoch": 1.89, "learning_rate": 3.220099176467303e-05, "loss": 0.9379, "step": 18260 }, { "epoch": 1.89, "learning_rate": 3.219853697045593e-05, "loss": 0.2602, "step": 18261 }, { "epoch": 1.89, "learning_rate": 3.2196082161438814e-05, "loss": 1.2819, "step": 18262 }, { "epoch": 1.89, "learning_rate": 3.2193627337638185e-05, "loss": 0.2266, "step": 18263 }, { "epoch": 1.89, "learning_rate": 3.219117249907056e-05, "loss": 0.1856, "step": 18264 }, { "epoch": 1.89, "learning_rate": 3.218871764575248e-05, "loss": 1.1045, "step": 18265 }, { "epoch": 1.89, "learning_rate": 3.218626277770046e-05, "loss": 1.0092, "step": 18266 }, { "epoch": 1.89, "learning_rate": 3.218380789493104e-05, "loss": 0.7878, "step": 18267 }, { "epoch": 1.89, "learning_rate": 3.218135299746074e-05, "loss": 0.3061, "step": 18268 }, { "epoch": 1.89, "learning_rate": 3.217889808530607e-05, "loss": 0.1796, "step": 18269 }, { "epoch": 1.89, "learning_rate": 3.217644315848357e-05, "loss": 0.2644, "step": 18270 }, { "epoch": 1.89, "learning_rate": 3.217398821700977e-05, "loss": 0.3125, "step": 18271 }, { "epoch": 1.89, "learning_rate": 3.217153326090118e-05, "loss": 0.3614, "step": 18272 }, { "epoch": 1.89, "learning_rate": 3.2169078290174355e-05, "loss": 1.1119, "step": 18273 }, { "epoch": 1.89, "learning_rate": 3.2166623304845776e-05, "loss": 0.2117, "step": 18274 }, { "epoch": 1.89, "learning_rate": 3.2164168304932015e-05, "loss": 0.9675, "step": 18275 }, { "epoch": 1.89, "learning_rate": 3.2161713290449566e-05, "loss": 0.4329, "step": 18276 }, { "epoch": 1.89, "learning_rate": 3.215925826141498e-05, "loss": 0.8968, "step": 18277 }, { "epoch": 1.89, "learning_rate": 3.215680321784477e-05, "loss": 0.4608, "step": 18278 }, { "epoch": 1.89, "learning_rate": 3.215434815975547e-05, "loss": 0.3031, "step": 18279 }, { "epoch": 1.89, "learning_rate": 3.2151893087163594e-05, "loss": 0.2192, "step": 18280 }, { "epoch": 1.89, "learning_rate": 3.214943800008568e-05, "loss": 0.0763, "step": 18281 }, { "epoch": 1.89, "learning_rate": 3.2146982898538256e-05, "loss": 1.062, "step": 18282 }, { "epoch": 1.89, "learning_rate": 3.214452778253784e-05, "loss": 0.3579, "step": 18283 }, { "epoch": 1.89, "learning_rate": 3.214207265210097e-05, "loss": 0.2065, "step": 18284 }, { "epoch": 1.89, "learning_rate": 3.2139617507244155e-05, "loss": 0.2863, "step": 18285 }, { "epoch": 1.89, "learning_rate": 3.213716234798395e-05, "loss": 1.2042, "step": 18286 }, { "epoch": 1.89, "learning_rate": 3.2134707174336865e-05, "loss": 0.3587, "step": 18287 }, { "epoch": 1.89, "learning_rate": 3.213225198631943e-05, "loss": 0.7868, "step": 18288 }, { "epoch": 1.89, "learning_rate": 3.212979678394817e-05, "loss": 1.301, "step": 18289 }, { "epoch": 1.89, "learning_rate": 3.212734156723962e-05, "loss": 1.0556, "step": 18290 }, { "epoch": 1.89, "learning_rate": 3.2124886336210304e-05, "loss": 1.544, "step": 18291 }, { "epoch": 1.89, "learning_rate": 3.2122431090876756e-05, "loss": 0.2272, "step": 18292 }, { "epoch": 1.9, "learning_rate": 3.21199758312555e-05, "loss": 1.026, "step": 18293 }, { "epoch": 1.9, "learning_rate": 3.211752055736304e-05, "loss": 0.9303, "step": 18294 }, { "epoch": 1.9, "learning_rate": 3.211506526921595e-05, "loss": 0.8067, "step": 18295 }, { "epoch": 1.9, "learning_rate": 3.211260996683073e-05, "loss": 0.9584, "step": 18296 }, { "epoch": 1.9, "learning_rate": 3.211015465022391e-05, "loss": 0.3313, "step": 18297 }, { "epoch": 1.9, "learning_rate": 3.2107699319412026e-05, "loss": 0.2276, "step": 18298 }, { "epoch": 1.9, "learning_rate": 3.2105243974411595e-05, "loss": 0.3616, "step": 18299 }, { "epoch": 1.9, "learning_rate": 3.210278861523916e-05, "loss": 0.8853, "step": 18300 }, { "epoch": 1.9, "learning_rate": 3.2100333241911246e-05, "loss": 0.3018, "step": 18301 }, { "epoch": 1.9, "learning_rate": 3.209787785444438e-05, "loss": 0.5213, "step": 18302 }, { "epoch": 1.9, "learning_rate": 3.209542245285509e-05, "loss": 0.1956, "step": 18303 }, { "epoch": 1.9, "learning_rate": 3.2092967037159914e-05, "loss": 0.7665, "step": 18304 }, { "epoch": 1.9, "learning_rate": 3.209051160737536e-05, "loss": 0.303, "step": 18305 }, { "epoch": 1.9, "learning_rate": 3.208805616351798e-05, "loss": 0.1863, "step": 18306 }, { "epoch": 1.9, "learning_rate": 3.208560070560429e-05, "loss": 1.3363, "step": 18307 }, { "epoch": 1.9, "learning_rate": 3.208314523365082e-05, "loss": 0.7414, "step": 18308 }, { "epoch": 1.9, "learning_rate": 3.2080689747674115e-05, "loss": 0.3251, "step": 18309 }, { "epoch": 1.9, "learning_rate": 3.207823424769069e-05, "loss": 0.401, "step": 18310 }, { "epoch": 1.9, "learning_rate": 3.2075778733717085e-05, "loss": 1.0747, "step": 18311 }, { "epoch": 1.9, "learning_rate": 3.207332320576981e-05, "loss": 0.1187, "step": 18312 }, { "epoch": 1.9, "learning_rate": 3.207086766386541e-05, "loss": 0.1423, "step": 18313 }, { "epoch": 1.9, "learning_rate": 3.206841210802041e-05, "loss": 0.4616, "step": 18314 }, { "epoch": 1.9, "learning_rate": 3.2065956538251356e-05, "loss": 0.2566, "step": 18315 }, { "epoch": 1.9, "learning_rate": 3.206350095457475e-05, "loss": 0.6839, "step": 18316 }, { "epoch": 1.9, "learning_rate": 3.2061045357007155e-05, "loss": 0.1423, "step": 18317 }, { "epoch": 1.9, "learning_rate": 3.2058589745565076e-05, "loss": 1.0806, "step": 18318 }, { "epoch": 1.9, "learning_rate": 3.205613412026506e-05, "loss": 0.1563, "step": 18319 }, { "epoch": 1.9, "learning_rate": 3.2053678481123626e-05, "loss": 0.271, "step": 18320 }, { "epoch": 1.9, "learning_rate": 3.205122282815731e-05, "loss": 0.9486, "step": 18321 }, { "epoch": 1.9, "learning_rate": 3.204876716138263e-05, "loss": 0.1737, "step": 18322 }, { "epoch": 1.9, "learning_rate": 3.204631148081614e-05, "loss": 0.6463, "step": 18323 }, { "epoch": 1.9, "learning_rate": 3.2043855786474364e-05, "loss": 1.05, "step": 18324 }, { "epoch": 1.9, "learning_rate": 3.2041400078373825e-05, "loss": 0.2962, "step": 18325 }, { "epoch": 1.9, "learning_rate": 3.203894435653106e-05, "loss": 0.6808, "step": 18326 }, { "epoch": 1.9, "learning_rate": 3.203648862096259e-05, "loss": 0.6546, "step": 18327 }, { "epoch": 1.9, "learning_rate": 3.203403287168496e-05, "loss": 0.8371, "step": 18328 }, { "epoch": 1.9, "learning_rate": 3.203157710871469e-05, "loss": 0.9619, "step": 18329 }, { "epoch": 1.9, "learning_rate": 3.2029121332068336e-05, "loss": 0.8412, "step": 18330 }, { "epoch": 1.9, "learning_rate": 3.2026665541762395e-05, "loss": 0.6715, "step": 18331 }, { "epoch": 1.9, "learning_rate": 3.202420973781343e-05, "loss": 1.0638, "step": 18332 }, { "epoch": 1.9, "learning_rate": 3.202175392023794e-05, "loss": 0.298, "step": 18333 }, { "epoch": 1.9, "learning_rate": 3.201929808905249e-05, "loss": 0.4111, "step": 18334 }, { "epoch": 1.9, "learning_rate": 3.20168422442736e-05, "loss": 0.3984, "step": 18335 }, { "epoch": 1.9, "learning_rate": 3.201438638591779e-05, "loss": 0.5201, "step": 18336 }, { "epoch": 1.9, "learning_rate": 3.201193051400161e-05, "loss": 0.7628, "step": 18337 }, { "epoch": 1.9, "learning_rate": 3.200947462854157e-05, "loss": 0.7913, "step": 18338 }, { "epoch": 1.9, "learning_rate": 3.200701872955424e-05, "loss": 0.3001, "step": 18339 }, { "epoch": 1.9, "learning_rate": 3.200456281705611e-05, "loss": 1.1158, "step": 18340 }, { "epoch": 1.9, "learning_rate": 3.200210689106374e-05, "loss": 0.7301, "step": 18341 }, { "epoch": 1.9, "learning_rate": 3.1999650951593663e-05, "loss": 0.2991, "step": 18342 }, { "epoch": 1.9, "learning_rate": 3.199719499866239e-05, "loss": 0.4296, "step": 18343 }, { "epoch": 1.9, "learning_rate": 3.199473903228648e-05, "loss": 0.4474, "step": 18344 }, { "epoch": 1.9, "learning_rate": 3.199228305248244e-05, "loss": 0.3655, "step": 18345 }, { "epoch": 1.9, "learning_rate": 3.198982705926682e-05, "loss": 0.4405, "step": 18346 }, { "epoch": 1.9, "learning_rate": 3.198737105265615e-05, "loss": 0.8834, "step": 18347 }, { "epoch": 1.9, "learning_rate": 3.198491503266697e-05, "loss": 0.8575, "step": 18348 }, { "epoch": 1.9, "learning_rate": 3.198245899931581e-05, "loss": 1.2305, "step": 18349 }, { "epoch": 1.9, "learning_rate": 3.198000295261918e-05, "loss": 1.056, "step": 18350 }, { "epoch": 1.9, "learning_rate": 3.197754689259365e-05, "loss": 0.7398, "step": 18351 }, { "epoch": 1.9, "learning_rate": 3.197509081925573e-05, "loss": 0.7423, "step": 18352 }, { "epoch": 1.9, "learning_rate": 3.197263473262196e-05, "loss": 0.2075, "step": 18353 }, { "epoch": 1.9, "learning_rate": 3.197017863270888e-05, "loss": 1.2306, "step": 18354 }, { "epoch": 1.9, "learning_rate": 3.196772251953303e-05, "loss": 0.9936, "step": 18355 }, { "epoch": 1.9, "learning_rate": 3.1965266393110916e-05, "loss": 0.6068, "step": 18356 }, { "epoch": 1.9, "learning_rate": 3.1962810253459084e-05, "loss": 0.5084, "step": 18357 }, { "epoch": 1.9, "learning_rate": 3.196035410059409e-05, "loss": 0.4982, "step": 18358 }, { "epoch": 1.9, "learning_rate": 3.195789793453243e-05, "loss": 0.7152, "step": 18359 }, { "epoch": 1.9, "learning_rate": 3.1955441755290674e-05, "loss": 0.4332, "step": 18360 }, { "epoch": 1.9, "learning_rate": 3.195298556288533e-05, "loss": 0.6372, "step": 18361 }, { "epoch": 1.9, "learning_rate": 3.195052935733296e-05, "loss": 0.5132, "step": 18362 }, { "epoch": 1.9, "learning_rate": 3.194807313865008e-05, "loss": 0.8925, "step": 18363 }, { "epoch": 1.9, "learning_rate": 3.194561690685322e-05, "loss": 1.0105, "step": 18364 }, { "epoch": 1.9, "learning_rate": 3.194316066195892e-05, "loss": 0.9057, "step": 18365 }, { "epoch": 1.9, "learning_rate": 3.194070440398372e-05, "loss": 0.3482, "step": 18366 }, { "epoch": 1.9, "learning_rate": 3.1938248132944155e-05, "loss": 0.1895, "step": 18367 }, { "epoch": 1.9, "learning_rate": 3.193579184885676e-05, "loss": 0.3599, "step": 18368 }, { "epoch": 1.9, "learning_rate": 3.1933335551738055e-05, "loss": 0.2218, "step": 18369 }, { "epoch": 1.9, "learning_rate": 3.19308792416046e-05, "loss": 2.1546, "step": 18370 }, { "epoch": 1.9, "learning_rate": 3.19284229184729e-05, "loss": 0.6084, "step": 18371 }, { "epoch": 1.9, "learning_rate": 3.192596658235953e-05, "loss": 1.1132, "step": 18372 }, { "epoch": 1.9, "learning_rate": 3.1923510233281e-05, "loss": 0.1168, "step": 18373 }, { "epoch": 1.9, "learning_rate": 3.192105387125384e-05, "loss": 0.3695, "step": 18374 }, { "epoch": 1.9, "learning_rate": 3.1918597496294596e-05, "loss": 0.2943, "step": 18375 }, { "epoch": 1.9, "learning_rate": 3.191614110841981e-05, "loss": 0.5108, "step": 18376 }, { "epoch": 1.9, "learning_rate": 3.191368470764601e-05, "loss": 0.7721, "step": 18377 }, { "epoch": 1.9, "learning_rate": 3.191122829398972e-05, "loss": 0.2245, "step": 18378 }, { "epoch": 1.9, "learning_rate": 3.190877186746749e-05, "loss": 0.9998, "step": 18379 }, { "epoch": 1.9, "learning_rate": 3.190631542809587e-05, "loss": 0.6292, "step": 18380 }, { "epoch": 1.9, "learning_rate": 3.190385897589137e-05, "loss": 0.483, "step": 18381 }, { "epoch": 1.9, "learning_rate": 3.190140251087054e-05, "loss": 0.4902, "step": 18382 }, { "epoch": 1.9, "learning_rate": 3.189894603304991e-05, "loss": 0.494, "step": 18383 }, { "epoch": 1.9, "learning_rate": 3.189648954244601e-05, "loss": 0.3731, "step": 18384 }, { "epoch": 1.9, "learning_rate": 3.18940330390754e-05, "loss": 0.9728, "step": 18385 }, { "epoch": 1.9, "learning_rate": 3.1891576522954603e-05, "loss": 0.6713, "step": 18386 }, { "epoch": 1.9, "learning_rate": 3.188911999410016e-05, "loss": 1.0203, "step": 18387 }, { "epoch": 1.9, "learning_rate": 3.1886663452528585e-05, "loss": 0.8875, "step": 18388 }, { "epoch": 1.91, "learning_rate": 3.188420689825645e-05, "loss": 0.592, "step": 18389 }, { "epoch": 1.91, "learning_rate": 3.188175033130026e-05, "loss": 0.4002, "step": 18390 }, { "epoch": 1.91, "learning_rate": 3.1879293751676586e-05, "loss": 0.7868, "step": 18391 }, { "epoch": 1.91, "learning_rate": 3.187683715940193e-05, "loss": 0.875, "step": 18392 }, { "epoch": 1.91, "learning_rate": 3.187438055449285e-05, "loss": 0.3844, "step": 18393 }, { "epoch": 1.91, "learning_rate": 3.187192393696588e-05, "loss": 0.3842, "step": 18394 }, { "epoch": 1.91, "learning_rate": 3.1869467306837557e-05, "loss": 0.7258, "step": 18395 }, { "epoch": 1.91, "learning_rate": 3.186701066412442e-05, "loss": 0.9477, "step": 18396 }, { "epoch": 1.91, "learning_rate": 3.1864554008843e-05, "loss": 0.2356, "step": 18397 }, { "epoch": 1.91, "learning_rate": 3.186209734100983e-05, "loss": 0.6304, "step": 18398 }, { "epoch": 1.91, "learning_rate": 3.1859640660641475e-05, "loss": 0.6088, "step": 18399 }, { "epoch": 1.91, "learning_rate": 3.185718396775444e-05, "loss": 1.859, "step": 18400 }, { "epoch": 1.91, "learning_rate": 3.185472726236528e-05, "loss": 0.6587, "step": 18401 }, { "epoch": 1.91, "learning_rate": 3.185227054449054e-05, "loss": 0.2376, "step": 18402 }, { "epoch": 1.91, "learning_rate": 3.184981381414673e-05, "loss": 0.0282, "step": 18403 }, { "epoch": 1.91, "learning_rate": 3.184735707135041e-05, "loss": 0.404, "step": 18404 }, { "epoch": 1.91, "learning_rate": 3.184490031611813e-05, "loss": 1.1284, "step": 18405 }, { "epoch": 1.91, "learning_rate": 3.1842443548466406e-05, "loss": 0.9944, "step": 18406 }, { "epoch": 1.91, "learning_rate": 3.183998676841178e-05, "loss": 0.1889, "step": 18407 }, { "epoch": 1.91, "learning_rate": 3.183752997597079e-05, "loss": 0.2747, "step": 18408 }, { "epoch": 1.91, "learning_rate": 3.1835073171159984e-05, "loss": 0.2916, "step": 18409 }, { "epoch": 1.91, "learning_rate": 3.183261635399589e-05, "loss": 1.077, "step": 18410 }, { "epoch": 1.91, "learning_rate": 3.183015952449506e-05, "loss": 0.6356, "step": 18411 }, { "epoch": 1.91, "learning_rate": 3.182770268267402e-05, "loss": 0.9603, "step": 18412 }, { "epoch": 1.91, "learning_rate": 3.182524582854931e-05, "loss": 0.3514, "step": 18413 }, { "epoch": 1.91, "learning_rate": 3.182278896213747e-05, "loss": 0.5239, "step": 18414 }, { "epoch": 1.91, "learning_rate": 3.182033208345505e-05, "loss": 0.7539, "step": 18415 }, { "epoch": 1.91, "learning_rate": 3.181787519251857e-05, "loss": 0.6917, "step": 18416 }, { "epoch": 1.91, "learning_rate": 3.181541828934459e-05, "loss": 0.957, "step": 18417 }, { "epoch": 1.91, "learning_rate": 3.181296137394963e-05, "loss": 0.7448, "step": 18418 }, { "epoch": 1.91, "learning_rate": 3.181050444635024e-05, "loss": 0.5545, "step": 18419 }, { "epoch": 1.91, "learning_rate": 3.1808047506562965e-05, "loss": 0.8726, "step": 18420 }, { "epoch": 1.91, "learning_rate": 3.1805590554604335e-05, "loss": 0.3855, "step": 18421 }, { "epoch": 1.91, "learning_rate": 3.180313359049088e-05, "loss": 0.6237, "step": 18422 }, { "epoch": 1.91, "learning_rate": 3.180067661423916e-05, "loss": 0.9935, "step": 18423 }, { "epoch": 1.91, "learning_rate": 3.179821962586571e-05, "loss": 1.0685, "step": 18424 }, { "epoch": 1.91, "learning_rate": 3.179576262538707e-05, "loss": 1.3466, "step": 18425 }, { "epoch": 1.91, "learning_rate": 3.1793305612819765e-05, "loss": 0.3937, "step": 18426 }, { "epoch": 1.91, "learning_rate": 3.179084858818035e-05, "loss": 0.295, "step": 18427 }, { "epoch": 1.91, "learning_rate": 3.178839155148536e-05, "loss": 0.246, "step": 18428 }, { "epoch": 1.91, "learning_rate": 3.178593450275134e-05, "loss": 0.3706, "step": 18429 }, { "epoch": 1.91, "learning_rate": 3.178347744199483e-05, "loss": 1.167, "step": 18430 }, { "epoch": 1.91, "learning_rate": 3.1781020369232356e-05, "loss": 0.6427, "step": 18431 }, { "epoch": 1.91, "learning_rate": 3.177856328448048e-05, "loss": 1.1672, "step": 18432 }, { "epoch": 1.91, "learning_rate": 3.177610618775573e-05, "loss": 0.4778, "step": 18433 }, { "epoch": 1.91, "learning_rate": 3.177364907907465e-05, "loss": 0.3151, "step": 18434 }, { "epoch": 1.91, "learning_rate": 3.177119195845377e-05, "loss": 0.1312, "step": 18435 }, { "epoch": 1.91, "learning_rate": 3.1768734825909646e-05, "loss": 1.0549, "step": 18436 }, { "epoch": 1.91, "learning_rate": 3.1766277681458814e-05, "loss": 0.7087, "step": 18437 }, { "epoch": 1.91, "learning_rate": 3.176382052511781e-05, "loss": 0.8874, "step": 18438 }, { "epoch": 1.91, "learning_rate": 3.176136335690318e-05, "loss": 0.3786, "step": 18439 }, { "epoch": 1.91, "learning_rate": 3.175890617683147e-05, "loss": 0.778, "step": 18440 }, { "epoch": 1.91, "learning_rate": 3.175644898491921e-05, "loss": 1.0789, "step": 18441 }, { "epoch": 1.91, "learning_rate": 3.175399178118295e-05, "loss": 0.418, "step": 18442 }, { "epoch": 1.91, "learning_rate": 3.175153456563922e-05, "loss": 0.8836, "step": 18443 }, { "epoch": 1.91, "learning_rate": 3.174907733830458e-05, "loss": 0.378, "step": 18444 }, { "epoch": 1.91, "learning_rate": 3.174662009919556e-05, "loss": 0.3114, "step": 18445 }, { "epoch": 1.91, "learning_rate": 3.1744162848328694e-05, "loss": 1.3993, "step": 18446 }, { "epoch": 1.91, "learning_rate": 3.1741705585720534e-05, "loss": 0.5637, "step": 18447 }, { "epoch": 1.91, "learning_rate": 3.173924831138762e-05, "loss": 1.2542, "step": 18448 }, { "epoch": 1.91, "learning_rate": 3.17367910253465e-05, "loss": 0.1148, "step": 18449 }, { "epoch": 1.91, "learning_rate": 3.17343337276137e-05, "loss": 1.0926, "step": 18450 }, { "epoch": 1.91, "learning_rate": 3.1731876418205775e-05, "loss": 0.7874, "step": 18451 }, { "epoch": 1.91, "learning_rate": 3.1729419097139256e-05, "loss": 0.2915, "step": 18452 }, { "epoch": 1.91, "learning_rate": 3.1726961764430704e-05, "loss": 0.8681, "step": 18453 }, { "epoch": 1.91, "learning_rate": 3.1724504420096644e-05, "loss": 0.1993, "step": 18454 }, { "epoch": 1.91, "learning_rate": 3.1722047064153626e-05, "loss": 0.8855, "step": 18455 }, { "epoch": 1.91, "learning_rate": 3.171958969661819e-05, "loss": 1.0772, "step": 18456 }, { "epoch": 1.91, "learning_rate": 3.171713231750687e-05, "loss": 0.2672, "step": 18457 }, { "epoch": 1.91, "learning_rate": 3.171467492683623e-05, "loss": 0.9413, "step": 18458 }, { "epoch": 1.91, "learning_rate": 3.1712217524622796e-05, "loss": 0.928, "step": 18459 }, { "epoch": 1.91, "learning_rate": 3.17097601108831e-05, "loss": 0.8578, "step": 18460 }, { "epoch": 1.91, "learning_rate": 3.170730268563372e-05, "loss": 0.2867, "step": 18461 }, { "epoch": 1.91, "learning_rate": 3.170484524889116e-05, "loss": 0.3074, "step": 18462 }, { "epoch": 1.91, "learning_rate": 3.1702387800671996e-05, "loss": 0.8709, "step": 18463 }, { "epoch": 1.91, "learning_rate": 3.1699930340992744e-05, "loss": 0.4437, "step": 18464 }, { "epoch": 1.91, "learning_rate": 3.169747286986997e-05, "loss": 0.1736, "step": 18465 }, { "epoch": 1.91, "learning_rate": 3.169501538732019e-05, "loss": 0.1816, "step": 18466 }, { "epoch": 1.91, "learning_rate": 3.1692557893359976e-05, "loss": 0.5766, "step": 18467 }, { "epoch": 1.91, "learning_rate": 3.169010038800586e-05, "loss": 0.5943, "step": 18468 }, { "epoch": 1.91, "learning_rate": 3.168764287127438e-05, "loss": 1.1486, "step": 18469 }, { "epoch": 1.91, "learning_rate": 3.168518534318208e-05, "loss": 0.1271, "step": 18470 }, { "epoch": 1.91, "learning_rate": 3.16827278037455e-05, "loss": 0.6491, "step": 18471 }, { "epoch": 1.91, "learning_rate": 3.16802702529812e-05, "loss": 0.4842, "step": 18472 }, { "epoch": 1.91, "learning_rate": 3.1677812690905716e-05, "loss": 0.491, "step": 18473 }, { "epoch": 1.91, "learning_rate": 3.167535511753559e-05, "loss": 0.2844, "step": 18474 }, { "epoch": 1.91, "learning_rate": 3.167289753288736e-05, "loss": 0.3242, "step": 18475 }, { "epoch": 1.91, "learning_rate": 3.167043993697757e-05, "loss": 0.8438, "step": 18476 }, { "epoch": 1.91, "learning_rate": 3.166798232982278e-05, "loss": 0.7388, "step": 18477 }, { "epoch": 1.91, "learning_rate": 3.1665524711439516e-05, "loss": 0.8249, "step": 18478 }, { "epoch": 1.91, "learning_rate": 3.166306708184432e-05, "loss": 0.6939, "step": 18479 }, { "epoch": 1.91, "learning_rate": 3.166060944105377e-05, "loss": 0.1414, "step": 18480 }, { "epoch": 1.91, "learning_rate": 3.165815178908437e-05, "loss": 1.0301, "step": 18481 }, { "epoch": 1.91, "learning_rate": 3.165569412595269e-05, "loss": 1.0028, "step": 18482 }, { "epoch": 1.91, "learning_rate": 3.165323645167525e-05, "loss": 0.2071, "step": 18483 }, { "epoch": 1.91, "learning_rate": 3.165077876626862e-05, "loss": 0.2482, "step": 18484 }, { "epoch": 1.92, "learning_rate": 3.164832106974932e-05, "loss": 0.2461, "step": 18485 }, { "epoch": 1.92, "learning_rate": 3.164586336213393e-05, "loss": 0.3804, "step": 18486 }, { "epoch": 1.92, "learning_rate": 3.164340564343896e-05, "loss": 0.9555, "step": 18487 }, { "epoch": 1.92, "learning_rate": 3.164094791368097e-05, "loss": 0.6801, "step": 18488 }, { "epoch": 1.92, "learning_rate": 3.1638490172876495e-05, "loss": 0.462, "step": 18489 }, { "epoch": 1.92, "learning_rate": 3.1636032421042095e-05, "loss": 0.1195, "step": 18490 }, { "epoch": 1.92, "learning_rate": 3.163357465819431e-05, "loss": 0.768, "step": 18491 }, { "epoch": 1.92, "learning_rate": 3.163111688434968e-05, "loss": 0.892, "step": 18492 }, { "epoch": 1.92, "learning_rate": 3.162865909952475e-05, "loss": 0.7321, "step": 18493 }, { "epoch": 1.92, "learning_rate": 3.162620130373607e-05, "loss": 0.2503, "step": 18494 }, { "epoch": 1.92, "learning_rate": 3.1623743497000184e-05, "loss": 0.2835, "step": 18495 }, { "epoch": 1.92, "learning_rate": 3.162128567933364e-05, "loss": 0.2859, "step": 18496 }, { "epoch": 1.92, "learning_rate": 3.161882785075297e-05, "loss": 0.5673, "step": 18497 }, { "epoch": 1.92, "learning_rate": 3.1616370011274736e-05, "loss": 0.3834, "step": 18498 }, { "epoch": 1.92, "learning_rate": 3.161391216091548e-05, "loss": 1.1403, "step": 18499 }, { "epoch": 1.92, "learning_rate": 3.1611454299691744e-05, "loss": 0.9349, "step": 18500 }, { "epoch": 1.92, "learning_rate": 3.160899642762008e-05, "loss": 0.2098, "step": 18501 }, { "epoch": 1.92, "learning_rate": 3.1606538544717016e-05, "loss": 1.715, "step": 18502 }, { "epoch": 1.92, "learning_rate": 3.160408065099911e-05, "loss": 0.1105, "step": 18503 }, { "epoch": 1.92, "learning_rate": 3.160162274648291e-05, "loss": 0.8056, "step": 18504 }, { "epoch": 1.92, "learning_rate": 3.159916483118497e-05, "loss": 0.6102, "step": 18505 }, { "epoch": 1.92, "learning_rate": 3.159670690512183e-05, "loss": 1.0558, "step": 18506 }, { "epoch": 1.92, "learning_rate": 3.159424896831002e-05, "loss": 0.5336, "step": 18507 }, { "epoch": 1.92, "learning_rate": 3.1591791020766104e-05, "loss": 0.1803, "step": 18508 }, { "epoch": 1.92, "learning_rate": 3.158933306250662e-05, "loss": 0.6506, "step": 18509 }, { "epoch": 1.92, "learning_rate": 3.158687509354813e-05, "loss": 0.1966, "step": 18510 }, { "epoch": 1.92, "learning_rate": 3.158441711390715e-05, "loss": 0.9318, "step": 18511 }, { "epoch": 1.92, "learning_rate": 3.158195912360025e-05, "loss": 0.0477, "step": 18512 }, { "epoch": 1.92, "learning_rate": 3.1579501122643975e-05, "loss": 0.2918, "step": 18513 }, { "epoch": 1.92, "learning_rate": 3.1577043111054866e-05, "loss": 0.4764, "step": 18514 }, { "epoch": 1.92, "learning_rate": 3.157458508884948e-05, "loss": 0.9564, "step": 18515 }, { "epoch": 1.92, "learning_rate": 3.1572127056044344e-05, "loss": 1.4389, "step": 18516 }, { "epoch": 1.92, "learning_rate": 3.156966901265602e-05, "loss": 0.3934, "step": 18517 }, { "epoch": 1.92, "learning_rate": 3.1567210958701054e-05, "loss": 0.136, "step": 18518 }, { "epoch": 1.92, "learning_rate": 3.156475289419599e-05, "loss": 1.088, "step": 18519 }, { "epoch": 1.92, "learning_rate": 3.1562294819157375e-05, "loss": 0.3662, "step": 18520 }, { "epoch": 1.92, "learning_rate": 3.155983673360175e-05, "loss": 0.2351, "step": 18521 }, { "epoch": 1.92, "learning_rate": 3.155737863754568e-05, "loss": 0.2191, "step": 18522 }, { "epoch": 1.92, "learning_rate": 3.15549205310057e-05, "loss": 0.7834, "step": 18523 }, { "epoch": 1.92, "learning_rate": 3.1552462413998353e-05, "loss": 0.7771, "step": 18524 }, { "epoch": 1.92, "learning_rate": 3.15500042865402e-05, "loss": 0.7508, "step": 18525 }, { "epoch": 1.92, "learning_rate": 3.154754614864778e-05, "loss": 0.4529, "step": 18526 }, { "epoch": 1.92, "learning_rate": 3.154508800033763e-05, "loss": 0.148, "step": 18527 }, { "epoch": 1.92, "learning_rate": 3.154262984162632e-05, "loss": 0.2674, "step": 18528 }, { "epoch": 1.92, "learning_rate": 3.1540171672530385e-05, "loss": 0.2741, "step": 18529 }, { "epoch": 1.92, "learning_rate": 3.1537713493066366e-05, "loss": 0.6096, "step": 18530 }, { "epoch": 1.92, "learning_rate": 3.153525530325083e-05, "loss": 0.6937, "step": 18531 }, { "epoch": 1.92, "learning_rate": 3.153279710310031e-05, "loss": 1.2029, "step": 18532 }, { "epoch": 1.92, "learning_rate": 3.1530338892631356e-05, "loss": 0.4045, "step": 18533 }, { "epoch": 1.92, "learning_rate": 3.152788067186053e-05, "loss": 0.197, "step": 18534 }, { "epoch": 1.92, "learning_rate": 3.152542244080436e-05, "loss": 0.9701, "step": 18535 }, { "epoch": 1.92, "learning_rate": 3.15229641994794e-05, "loss": 0.3807, "step": 18536 }, { "epoch": 1.92, "learning_rate": 3.152050594790221e-05, "loss": 0.5437, "step": 18537 }, { "epoch": 1.92, "learning_rate": 3.151804768608932e-05, "loss": 0.4779, "step": 18538 }, { "epoch": 1.92, "learning_rate": 3.15155894140573e-05, "loss": 0.7917, "step": 18539 }, { "epoch": 1.92, "learning_rate": 3.151313113182268e-05, "loss": 0.6806, "step": 18540 }, { "epoch": 1.92, "learning_rate": 3.1510672839402014e-05, "loss": 0.3971, "step": 18541 }, { "epoch": 1.92, "learning_rate": 3.150821453681185e-05, "loss": 0.7065, "step": 18542 }, { "epoch": 1.92, "learning_rate": 3.150575622406875e-05, "loss": 1.0611, "step": 18543 }, { "epoch": 1.92, "learning_rate": 3.1503297901189244e-05, "loss": 0.462, "step": 18544 }, { "epoch": 1.92, "learning_rate": 3.1500839568189886e-05, "loss": 0.6556, "step": 18545 }, { "epoch": 1.92, "learning_rate": 3.1498381225087226e-05, "loss": 0.2747, "step": 18546 }, { "epoch": 1.92, "learning_rate": 3.1495922871897814e-05, "loss": 0.3744, "step": 18547 }, { "epoch": 1.92, "learning_rate": 3.1493464508638203e-05, "loss": 0.5961, "step": 18548 }, { "epoch": 1.92, "learning_rate": 3.149100613532494e-05, "loss": 0.4667, "step": 18549 }, { "epoch": 1.92, "learning_rate": 3.148854775197458e-05, "loss": 0.3757, "step": 18550 }, { "epoch": 1.92, "learning_rate": 3.148608935860365e-05, "loss": 0.5843, "step": 18551 }, { "epoch": 1.92, "learning_rate": 3.148363095522871e-05, "loss": 0.5359, "step": 18552 }, { "epoch": 1.92, "learning_rate": 3.1481172541866325e-05, "loss": 0.2739, "step": 18553 }, { "epoch": 1.92, "learning_rate": 3.147871411853303e-05, "loss": 1.22, "step": 18554 }, { "epoch": 1.92, "learning_rate": 3.147625568524538e-05, "loss": 0.3071, "step": 18555 }, { "epoch": 1.92, "learning_rate": 3.147379724201993e-05, "loss": 0.9808, "step": 18556 }, { "epoch": 1.92, "learning_rate": 3.14713387888732e-05, "loss": 0.8605, "step": 18557 }, { "epoch": 1.92, "learning_rate": 3.146888032582178e-05, "loss": 0.5957, "step": 18558 }, { "epoch": 1.92, "learning_rate": 3.146642185288219e-05, "loss": 0.1185, "step": 18559 }, { "epoch": 1.92, "learning_rate": 3.146396337007099e-05, "loss": 0.7904, "step": 18560 }, { "epoch": 1.92, "learning_rate": 3.1461504877404745e-05, "loss": 0.7016, "step": 18561 }, { "epoch": 1.92, "learning_rate": 3.1459046374899985e-05, "loss": 0.6753, "step": 18562 }, { "epoch": 1.92, "learning_rate": 3.145658786257327e-05, "loss": 1.0744, "step": 18563 }, { "epoch": 1.92, "learning_rate": 3.1454129340441136e-05, "loss": 0.2924, "step": 18564 }, { "epoch": 1.92, "learning_rate": 3.145167080852015e-05, "loss": 0.8134, "step": 18565 }, { "epoch": 1.92, "learning_rate": 3.144921226682685e-05, "loss": 0.6035, "step": 18566 }, { "epoch": 1.92, "learning_rate": 3.14467537153778e-05, "loss": 0.8865, "step": 18567 }, { "epoch": 1.92, "learning_rate": 3.144429515418955e-05, "loss": 0.2721, "step": 18568 }, { "epoch": 1.92, "learning_rate": 3.144183658327863e-05, "loss": 0.6135, "step": 18569 }, { "epoch": 1.92, "learning_rate": 3.14393780026616e-05, "loss": 1.1201, "step": 18570 }, { "epoch": 1.92, "learning_rate": 3.1436919412355015e-05, "loss": 0.87, "step": 18571 }, { "epoch": 1.92, "learning_rate": 3.143446081237543e-05, "loss": 0.7765, "step": 18572 }, { "epoch": 1.92, "learning_rate": 3.143200220273939e-05, "loss": 0.9321, "step": 18573 }, { "epoch": 1.92, "learning_rate": 3.142954358346345e-05, "loss": 0.5544, "step": 18574 }, { "epoch": 1.92, "learning_rate": 3.142708495456415e-05, "loss": 0.588, "step": 18575 }, { "epoch": 1.92, "learning_rate": 3.142462631605805e-05, "loss": 0.7926, "step": 18576 }, { "epoch": 1.92, "learning_rate": 3.142216766796171e-05, "loss": 0.7584, "step": 18577 }, { "epoch": 1.92, "learning_rate": 3.141970901029165e-05, "loss": 0.7161, "step": 18578 }, { "epoch": 1.92, "learning_rate": 3.141725034306445e-05, "loss": 0.8598, "step": 18579 }, { "epoch": 1.93, "learning_rate": 3.1414791666296647e-05, "loss": 0.6419, "step": 18580 }, { "epoch": 1.93, "learning_rate": 3.141233298000481e-05, "loss": 1.1438, "step": 18581 }, { "epoch": 1.93, "learning_rate": 3.140987428420548e-05, "loss": 0.368, "step": 18582 }, { "epoch": 1.93, "learning_rate": 3.140741557891519e-05, "loss": 0.7468, "step": 18583 }, { "epoch": 1.93, "learning_rate": 3.140495686415052e-05, "loss": 0.5497, "step": 18584 }, { "epoch": 1.93, "learning_rate": 3.1402498139928e-05, "loss": 0.3038, "step": 18585 }, { "epoch": 1.93, "learning_rate": 3.14000394062642e-05, "loss": 0.5299, "step": 18586 }, { "epoch": 1.93, "learning_rate": 3.139758066317566e-05, "loss": 0.7433, "step": 18587 }, { "epoch": 1.93, "learning_rate": 3.139512191067893e-05, "loss": 0.4734, "step": 18588 }, { "epoch": 1.93, "learning_rate": 3.139266314879057e-05, "loss": 1.1522, "step": 18589 }, { "epoch": 1.93, "learning_rate": 3.139020437752712e-05, "loss": 0.7238, "step": 18590 }, { "epoch": 1.93, "learning_rate": 3.138774559690514e-05, "loss": 0.5454, "step": 18591 }, { "epoch": 1.93, "learning_rate": 3.138528680694119e-05, "loss": 1.4752, "step": 18592 }, { "epoch": 1.93, "learning_rate": 3.1382828007651815e-05, "loss": 0.2961, "step": 18593 }, { "epoch": 1.93, "learning_rate": 3.138036919905356e-05, "loss": 0.2853, "step": 18594 }, { "epoch": 1.93, "learning_rate": 3.137791038116298e-05, "loss": 1.0474, "step": 18595 }, { "epoch": 1.93, "learning_rate": 3.137545155399663e-05, "loss": 0.19, "step": 18596 }, { "epoch": 1.93, "learning_rate": 3.137299271757106e-05, "loss": 0.4111, "step": 18597 }, { "epoch": 1.93, "learning_rate": 3.137053387190283e-05, "loss": 0.551, "step": 18598 }, { "epoch": 1.93, "learning_rate": 3.136807501700848e-05, "loss": 0.6798, "step": 18599 }, { "epoch": 1.93, "learning_rate": 3.136561615290458e-05, "loss": 0.7559, "step": 18600 }, { "epoch": 1.93, "learning_rate": 3.136315727960767e-05, "loss": 1.0622, "step": 18601 }, { "epoch": 1.93, "learning_rate": 3.13606983971343e-05, "loss": 0.8307, "step": 18602 }, { "epoch": 1.93, "learning_rate": 3.1358239505501024e-05, "loss": 0.3492, "step": 18603 }, { "epoch": 1.93, "learning_rate": 3.135578060472439e-05, "loss": 0.9016, "step": 18604 }, { "epoch": 1.93, "learning_rate": 3.135332169482097e-05, "loss": 0.6064, "step": 18605 }, { "epoch": 1.93, "learning_rate": 3.135086277580731e-05, "loss": 1.0544, "step": 18606 }, { "epoch": 1.93, "learning_rate": 3.1348403847699956e-05, "loss": 0.7099, "step": 18607 }, { "epoch": 1.93, "learning_rate": 3.134594491051545e-05, "loss": 0.2733, "step": 18608 }, { "epoch": 1.93, "learning_rate": 3.1343485964270365e-05, "loss": 0.4815, "step": 18609 }, { "epoch": 1.93, "learning_rate": 3.134102700898125e-05, "loss": 0.7094, "step": 18610 }, { "epoch": 1.93, "learning_rate": 3.1338568044664655e-05, "loss": 1.1113, "step": 18611 }, { "epoch": 1.93, "learning_rate": 3.1336109071337126e-05, "loss": 0.6773, "step": 18612 }, { "epoch": 1.93, "learning_rate": 3.1333650089015225e-05, "loss": 0.6169, "step": 18613 }, { "epoch": 1.93, "learning_rate": 3.133119109771551e-05, "loss": 0.4771, "step": 18614 }, { "epoch": 1.93, "learning_rate": 3.132873209745453e-05, "loss": 0.2323, "step": 18615 }, { "epoch": 1.93, "learning_rate": 3.1326273088248824e-05, "loss": 0.1755, "step": 18616 }, { "epoch": 1.93, "learning_rate": 3.132381407011496e-05, "loss": 0.7175, "step": 18617 }, { "epoch": 1.93, "learning_rate": 3.13213550430695e-05, "loss": 0.3638, "step": 18618 }, { "epoch": 1.93, "learning_rate": 3.131889600712898e-05, "loss": 0.4215, "step": 18619 }, { "epoch": 1.93, "learning_rate": 3.1316436962309966e-05, "loss": 0.0121, "step": 18620 }, { "epoch": 1.93, "learning_rate": 3.1313977908628996e-05, "loss": 0.9107, "step": 18621 }, { "epoch": 1.93, "learning_rate": 3.131151884610264e-05, "loss": 0.1804, "step": 18622 }, { "epoch": 1.93, "learning_rate": 3.130905977474745e-05, "loss": 0.7629, "step": 18623 }, { "epoch": 1.93, "learning_rate": 3.1306600694579976e-05, "loss": 1.0957, "step": 18624 }, { "epoch": 1.93, "learning_rate": 3.130414160561677e-05, "loss": 0.1617, "step": 18625 }, { "epoch": 1.93, "learning_rate": 3.130168250787438e-05, "loss": 0.1891, "step": 18626 }, { "epoch": 1.93, "learning_rate": 3.129922340136938e-05, "loss": 0.4263, "step": 18627 }, { "epoch": 1.93, "learning_rate": 3.12967642861183e-05, "loss": 0.6325, "step": 18628 }, { "epoch": 1.93, "learning_rate": 3.129430516213771e-05, "loss": 0.6368, "step": 18629 }, { "epoch": 1.93, "learning_rate": 3.129184602944416e-05, "loss": 0.8168, "step": 18630 }, { "epoch": 1.93, "learning_rate": 3.128938688805421e-05, "loss": 0.7085, "step": 18631 }, { "epoch": 1.93, "learning_rate": 3.128692773798441e-05, "loss": 0.8377, "step": 18632 }, { "epoch": 1.93, "learning_rate": 3.128446857925131e-05, "loss": 0.4862, "step": 18633 }, { "epoch": 1.93, "learning_rate": 3.1282009411871466e-05, "loss": 0.2714, "step": 18634 }, { "epoch": 1.93, "learning_rate": 3.127955023586144e-05, "loss": 0.4312, "step": 18635 }, { "epoch": 1.93, "learning_rate": 3.1277091051237776e-05, "loss": 1.1047, "step": 18636 }, { "epoch": 1.93, "learning_rate": 3.127463185801704e-05, "loss": 0.8516, "step": 18637 }, { "epoch": 1.93, "learning_rate": 3.127217265621577e-05, "loss": 0.3318, "step": 18638 }, { "epoch": 1.93, "learning_rate": 3.126971344585054e-05, "loss": 0.5949, "step": 18639 }, { "epoch": 1.93, "learning_rate": 3.126725422693789e-05, "loss": 0.7396, "step": 18640 }, { "epoch": 1.93, "learning_rate": 3.126479499949439e-05, "loss": 0.4773, "step": 18641 }, { "epoch": 1.93, "learning_rate": 3.1262335763536576e-05, "loss": 0.6121, "step": 18642 }, { "epoch": 1.93, "learning_rate": 3.1259876519081024e-05, "loss": 0.2699, "step": 18643 }, { "epoch": 1.93, "learning_rate": 3.1257417266144274e-05, "loss": 0.7311, "step": 18644 }, { "epoch": 1.93, "learning_rate": 3.1254958004742886e-05, "loss": 0.3774, "step": 18645 }, { "epoch": 1.93, "learning_rate": 3.125249873489341e-05, "loss": 0.7989, "step": 18646 }, { "epoch": 1.93, "learning_rate": 3.12500394566124e-05, "loss": 0.6529, "step": 18647 }, { "epoch": 1.93, "learning_rate": 3.124758016991642e-05, "loss": 0.5537, "step": 18648 }, { "epoch": 1.93, "learning_rate": 3.1245120874822036e-05, "loss": 0.8811, "step": 18649 }, { "epoch": 1.93, "learning_rate": 3.1242661571345785e-05, "loss": 1.1212, "step": 18650 }, { "epoch": 1.93, "learning_rate": 3.1240202259504215e-05, "loss": 0.3322, "step": 18651 }, { "epoch": 1.93, "learning_rate": 3.12377429393139e-05, "loss": 0.8132, "step": 18652 }, { "epoch": 1.93, "learning_rate": 3.123528361079139e-05, "loss": 0.2954, "step": 18653 }, { "epoch": 1.93, "learning_rate": 3.123282427395324e-05, "loss": 1.0013, "step": 18654 }, { "epoch": 1.93, "learning_rate": 3.1230364928816004e-05, "loss": 0.592, "step": 18655 }, { "epoch": 1.93, "learning_rate": 3.1227905575396244e-05, "loss": 0.8668, "step": 18656 }, { "epoch": 1.93, "learning_rate": 3.12254462137105e-05, "loss": 0.3869, "step": 18657 }, { "epoch": 1.93, "learning_rate": 3.122298684377535e-05, "loss": 0.5051, "step": 18658 }, { "epoch": 1.93, "learning_rate": 3.122052746560734e-05, "loss": 1.2222, "step": 18659 }, { "epoch": 1.93, "learning_rate": 3.121806807922301e-05, "loss": 0.6839, "step": 18660 }, { "epoch": 1.93, "learning_rate": 3.121560868463894e-05, "loss": 0.6467, "step": 18661 }, { "epoch": 1.93, "learning_rate": 3.1213149281871675e-05, "loss": 0.191, "step": 18662 }, { "epoch": 1.93, "learning_rate": 3.121068987093778e-05, "loss": 1.3949, "step": 18663 }, { "epoch": 1.93, "learning_rate": 3.1208230451853804e-05, "loss": 0.6061, "step": 18664 }, { "epoch": 1.93, "learning_rate": 3.120577102463629e-05, "loss": 0.4351, "step": 18665 }, { "epoch": 1.93, "learning_rate": 3.120331158930182e-05, "loss": 0.544, "step": 18666 }, { "epoch": 1.93, "learning_rate": 3.120085214586693e-05, "loss": 0.6753, "step": 18667 }, { "epoch": 1.93, "learning_rate": 3.1198392694348185e-05, "loss": 0.8365, "step": 18668 }, { "epoch": 1.93, "learning_rate": 3.119593323476215e-05, "loss": 0.6255, "step": 18669 }, { "epoch": 1.93, "learning_rate": 3.119347376712537e-05, "loss": 0.459, "step": 18670 }, { "epoch": 1.93, "learning_rate": 3.1191014291454386e-05, "loss": 0.15, "step": 18671 }, { "epoch": 1.93, "learning_rate": 3.1188554807765796e-05, "loss": 0.2873, "step": 18672 }, { "epoch": 1.93, "learning_rate": 3.118609531607612e-05, "loss": 0.6721, "step": 18673 }, { "epoch": 1.93, "learning_rate": 3.118363581640193e-05, "loss": 0.2921, "step": 18674 }, { "epoch": 1.93, "learning_rate": 3.1181176308759786e-05, "loss": 0.3783, "step": 18675 }, { "epoch": 1.94, "learning_rate": 3.117871679316623e-05, "loss": 0.6308, "step": 18676 }, { "epoch": 1.94, "learning_rate": 3.117625726963784e-05, "loss": 0.5895, "step": 18677 }, { "epoch": 1.94, "learning_rate": 3.117379773819116e-05, "loss": 0.6655, "step": 18678 }, { "epoch": 1.94, "learning_rate": 3.1171338198842734e-05, "loss": 0.8099, "step": 18679 }, { "epoch": 1.94, "learning_rate": 3.116887865160914e-05, "loss": 0.7963, "step": 18680 }, { "epoch": 1.94, "learning_rate": 3.1166419096506935e-05, "loss": 0.6891, "step": 18681 }, { "epoch": 1.94, "learning_rate": 3.116395953355267e-05, "loss": 0.5992, "step": 18682 }, { "epoch": 1.94, "learning_rate": 3.1161499962762897e-05, "loss": 0.5431, "step": 18683 }, { "epoch": 1.94, "learning_rate": 3.1159040384154175e-05, "loss": 0.5601, "step": 18684 }, { "epoch": 1.94, "learning_rate": 3.115658079774307e-05, "loss": 0.7128, "step": 18685 }, { "epoch": 1.94, "learning_rate": 3.1154121203546135e-05, "loss": 0.253, "step": 18686 }, { "epoch": 1.94, "learning_rate": 3.115166160157992e-05, "loss": 0.8158, "step": 18687 }, { "epoch": 1.94, "learning_rate": 3.114920199186099e-05, "loss": 0.7111, "step": 18688 }, { "epoch": 1.94, "learning_rate": 3.11467423744059e-05, "loss": 0.2975, "step": 18689 }, { "epoch": 1.94, "learning_rate": 3.114428274923121e-05, "loss": 0.1125, "step": 18690 }, { "epoch": 1.94, "learning_rate": 3.114182311635348e-05, "loss": 0.717, "step": 18691 }, { "epoch": 1.94, "learning_rate": 3.1139363475789256e-05, "loss": 0.7085, "step": 18692 }, { "epoch": 1.94, "learning_rate": 3.1136903827555104e-05, "loss": 0.919, "step": 18693 }, { "epoch": 1.94, "learning_rate": 3.1134444171667596e-05, "loss": 0.4626, "step": 18694 }, { "epoch": 1.94, "learning_rate": 3.113198450814326e-05, "loss": 0.3461, "step": 18695 }, { "epoch": 1.94, "learning_rate": 3.1129524836998676e-05, "loss": 0.4758, "step": 18696 }, { "epoch": 1.94, "learning_rate": 3.11270651582504e-05, "loss": 1.1971, "step": 18697 }, { "epoch": 1.94, "learning_rate": 3.112460547191497e-05, "loss": 0.633, "step": 18698 }, { "epoch": 1.94, "learning_rate": 3.1122145778008965e-05, "loss": 1.3526, "step": 18699 }, { "epoch": 1.94, "learning_rate": 3.1119686076548944e-05, "loss": 0.4499, "step": 18700 }, { "epoch": 1.94, "learning_rate": 3.111722636755146e-05, "loss": 1.0499, "step": 18701 }, { "epoch": 1.94, "learning_rate": 3.111476665103306e-05, "loss": 0.3727, "step": 18702 }, { "epoch": 1.94, "learning_rate": 3.111230692701031e-05, "loss": 0.1722, "step": 18703 }, { "epoch": 1.94, "learning_rate": 3.110984719549978e-05, "loss": 1.3567, "step": 18704 }, { "epoch": 1.94, "learning_rate": 3.110738745651801e-05, "loss": 0.2596, "step": 18705 }, { "epoch": 1.94, "learning_rate": 3.110492771008158e-05, "loss": 0.7281, "step": 18706 }, { "epoch": 1.94, "learning_rate": 3.110246795620702e-05, "loss": 0.6394, "step": 18707 }, { "epoch": 1.94, "learning_rate": 3.1100008194910906e-05, "loss": 0.2935, "step": 18708 }, { "epoch": 1.94, "learning_rate": 3.10975484262098e-05, "loss": 0.3358, "step": 18709 }, { "epoch": 1.94, "learning_rate": 3.109508865012026e-05, "loss": 0.9327, "step": 18710 }, { "epoch": 1.94, "learning_rate": 3.109262886665883e-05, "loss": 0.6019, "step": 18711 }, { "epoch": 1.94, "learning_rate": 3.1090169075842074e-05, "loss": 0.9161, "step": 18712 }, { "epoch": 1.94, "learning_rate": 3.108770927768657e-05, "loss": 0.7546, "step": 18713 }, { "epoch": 1.94, "learning_rate": 3.1085249472208846e-05, "loss": 0.3375, "step": 18714 }, { "epoch": 1.94, "learning_rate": 3.108278965942549e-05, "loss": 0.7424, "step": 18715 }, { "epoch": 1.94, "learning_rate": 3.108032983935303e-05, "loss": 1.0205, "step": 18716 }, { "epoch": 1.94, "learning_rate": 3.107787001200806e-05, "loss": 0.7855, "step": 18717 }, { "epoch": 1.94, "learning_rate": 3.107541017740711e-05, "loss": 0.9854, "step": 18718 }, { "epoch": 1.94, "learning_rate": 3.1072950335566754e-05, "loss": 0.2472, "step": 18719 }, { "epoch": 1.94, "learning_rate": 3.107049048650356e-05, "loss": 0.4611, "step": 18720 }, { "epoch": 1.94, "learning_rate": 3.1068030630234055e-05, "loss": 0.2609, "step": 18721 }, { "epoch": 1.94, "learning_rate": 3.106557076677482e-05, "loss": 0.844, "step": 18722 }, { "epoch": 1.94, "learning_rate": 3.106311089614241e-05, "loss": 0.8496, "step": 18723 }, { "epoch": 1.94, "learning_rate": 3.1060651018353406e-05, "loss": 0.2016, "step": 18724 }, { "epoch": 1.94, "learning_rate": 3.105819113342434e-05, "loss": 0.6925, "step": 18725 }, { "epoch": 1.94, "learning_rate": 3.105573124137177e-05, "loss": 0.8535, "step": 18726 }, { "epoch": 1.94, "learning_rate": 3.1053271342212265e-05, "loss": 0.2349, "step": 18727 }, { "epoch": 1.94, "learning_rate": 3.1050811435962384e-05, "loss": 0.1231, "step": 18728 }, { "epoch": 1.94, "learning_rate": 3.1048351522638687e-05, "loss": 0.7659, "step": 18729 }, { "epoch": 1.94, "learning_rate": 3.1045891602257735e-05, "loss": 0.2869, "step": 18730 }, { "epoch": 1.94, "learning_rate": 3.104343167483608e-05, "loss": 0.5052, "step": 18731 }, { "epoch": 1.94, "learning_rate": 3.1040971740390304e-05, "loss": 1.0736, "step": 18732 }, { "epoch": 1.94, "learning_rate": 3.1038511798936934e-05, "loss": 0.818, "step": 18733 }, { "epoch": 1.94, "learning_rate": 3.1036051850492554e-05, "loss": 0.9671, "step": 18734 }, { "epoch": 1.94, "learning_rate": 3.1033591895073704e-05, "loss": 0.2168, "step": 18735 }, { "epoch": 1.94, "learning_rate": 3.103113193269696e-05, "loss": 0.6719, "step": 18736 }, { "epoch": 1.94, "learning_rate": 3.102867196337888e-05, "loss": 0.7833, "step": 18737 }, { "epoch": 1.94, "learning_rate": 3.102621198713602e-05, "loss": 0.533, "step": 18738 }, { "epoch": 1.94, "learning_rate": 3.102375200398495e-05, "loss": 0.2753, "step": 18739 }, { "epoch": 1.94, "learning_rate": 3.102129201394221e-05, "loss": 0.7217, "step": 18740 }, { "epoch": 1.94, "learning_rate": 3.101883201702437e-05, "loss": 0.7316, "step": 18741 }, { "epoch": 1.94, "learning_rate": 3.101637201324799e-05, "loss": 0.3065, "step": 18742 }, { "epoch": 1.94, "learning_rate": 3.101391200262964e-05, "loss": 1.2559, "step": 18743 }, { "epoch": 1.94, "learning_rate": 3.1011451985185876e-05, "loss": 0.2466, "step": 18744 }, { "epoch": 1.94, "learning_rate": 3.100899196093325e-05, "loss": 0.9041, "step": 18745 }, { "epoch": 1.94, "learning_rate": 3.1006531929888314e-05, "loss": 0.4799, "step": 18746 }, { "epoch": 1.94, "learning_rate": 3.100407189206764e-05, "loss": 0.3998, "step": 18747 }, { "epoch": 1.94, "learning_rate": 3.1001611847487806e-05, "loss": 0.9086, "step": 18748 }, { "epoch": 1.94, "learning_rate": 3.0999151796165345e-05, "loss": 0.655, "step": 18749 }, { "epoch": 1.94, "learning_rate": 3.099669173811683e-05, "loss": 0.795, "step": 18750 }, { "epoch": 1.94, "learning_rate": 3.0994231673358816e-05, "loss": 0.687, "step": 18751 }, { "epoch": 1.94, "learning_rate": 3.099177160190787e-05, "loss": 0.3982, "step": 18752 }, { "epoch": 1.94, "learning_rate": 3.098931152378055e-05, "loss": 0.4264, "step": 18753 }, { "epoch": 1.94, "learning_rate": 3.098685143899342e-05, "loss": 0.2658, "step": 18754 }, { "epoch": 1.94, "learning_rate": 3.0984391347563026e-05, "loss": 1.0215, "step": 18755 }, { "epoch": 1.94, "learning_rate": 3.098193124950595e-05, "loss": 0.17, "step": 18756 }, { "epoch": 1.94, "learning_rate": 3.097947114483874e-05, "loss": 0.2149, "step": 18757 }, { "epoch": 1.94, "learning_rate": 3.097701103357797e-05, "loss": 0.2291, "step": 18758 }, { "epoch": 1.94, "learning_rate": 3.097455091574017e-05, "loss": 0.7218, "step": 18759 }, { "epoch": 1.94, "learning_rate": 3.097209079134193e-05, "loss": 0.2394, "step": 18760 }, { "epoch": 1.94, "learning_rate": 3.09696306603998e-05, "loss": 0.4386, "step": 18761 }, { "epoch": 1.94, "learning_rate": 3.096717052293035e-05, "loss": 1.1752, "step": 18762 }, { "epoch": 1.94, "learning_rate": 3.096471037895013e-05, "loss": 0.4646, "step": 18763 }, { "epoch": 1.94, "learning_rate": 3.096225022847571e-05, "loss": 0.5263, "step": 18764 }, { "epoch": 1.94, "learning_rate": 3.095979007152363e-05, "loss": 0.7583, "step": 18765 }, { "epoch": 1.94, "learning_rate": 3.095732990811048e-05, "loss": 0.9188, "step": 18766 }, { "epoch": 1.94, "learning_rate": 3.095486973825282e-05, "loss": 0.1799, "step": 18767 }, { "epoch": 1.94, "learning_rate": 3.0952409561967183e-05, "loss": 0.4043, "step": 18768 }, { "epoch": 1.94, "learning_rate": 3.094994937927016e-05, "loss": 0.4906, "step": 18769 }, { "epoch": 1.94, "learning_rate": 3.094748919017829e-05, "loss": 0.5727, "step": 18770 }, { "epoch": 1.94, "learning_rate": 3.094502899470815e-05, "loss": 0.9164, "step": 18771 }, { "epoch": 1.95, "learning_rate": 3.094256879287629e-05, "loss": 0.2947, "step": 18772 }, { "epoch": 1.95, "learning_rate": 3.094010858469929e-05, "loss": 0.856, "step": 18773 }, { "epoch": 1.95, "learning_rate": 3.093764837019369e-05, "loss": 0.5899, "step": 18774 }, { "epoch": 1.95, "learning_rate": 3.0935188149376065e-05, "loss": 0.2592, "step": 18775 }, { "epoch": 1.95, "learning_rate": 3.093272792226296e-05, "loss": 0.5451, "step": 18776 }, { "epoch": 1.95, "learning_rate": 3.093026768887096e-05, "loss": 0.2557, "step": 18777 }, { "epoch": 1.95, "learning_rate": 3.092780744921662e-05, "loss": 0.5437, "step": 18778 }, { "epoch": 1.95, "learning_rate": 3.0925347203316485e-05, "loss": 0.2588, "step": 18779 }, { "epoch": 1.95, "learning_rate": 3.092288695118713e-05, "loss": 0.5708, "step": 18780 }, { "epoch": 1.95, "learning_rate": 3.092042669284512e-05, "loss": 0.8193, "step": 18781 }, { "epoch": 1.95, "learning_rate": 3.091796642830702e-05, "loss": 0.1445, "step": 18782 }, { "epoch": 1.95, "learning_rate": 3.0915506157589384e-05, "loss": 0.4129, "step": 18783 }, { "epoch": 1.95, "learning_rate": 3.091304588070876e-05, "loss": 0.1388, "step": 18784 }, { "epoch": 1.95, "learning_rate": 3.0910585597681736e-05, "loss": 1.3547, "step": 18785 }, { "epoch": 1.95, "learning_rate": 3.0908125308524864e-05, "loss": 0.3436, "step": 18786 }, { "epoch": 1.95, "learning_rate": 3.09056650132547e-05, "loss": 0.8475, "step": 18787 }, { "epoch": 1.95, "learning_rate": 3.0903204711887814e-05, "loss": 0.7197, "step": 18788 }, { "epoch": 1.95, "learning_rate": 3.0900744404440764e-05, "loss": 0.2104, "step": 18789 }, { "epoch": 1.95, "learning_rate": 3.089828409093011e-05, "loss": 0.3652, "step": 18790 }, { "epoch": 1.95, "learning_rate": 3.089582377137243e-05, "loss": 0.8783, "step": 18791 }, { "epoch": 1.95, "learning_rate": 3.0893363445784255e-05, "loss": 0.4578, "step": 18792 }, { "epoch": 1.95, "learning_rate": 3.089090311418218e-05, "loss": 1.0931, "step": 18793 }, { "epoch": 1.95, "learning_rate": 3.0888442776582746e-05, "loss": 1.0611, "step": 18794 }, { "epoch": 1.95, "learning_rate": 3.0885982433002524e-05, "loss": 0.3601, "step": 18795 }, { "epoch": 1.95, "learning_rate": 3.088352208345809e-05, "loss": 0.3726, "step": 18796 }, { "epoch": 1.95, "learning_rate": 3.088106172796598e-05, "loss": 0.7842, "step": 18797 }, { "epoch": 1.95, "learning_rate": 3.087860136654276e-05, "loss": 0.776, "step": 18798 }, { "epoch": 1.95, "learning_rate": 3.087614099920501e-05, "loss": 1.0881, "step": 18799 }, { "epoch": 1.95, "learning_rate": 3.087368062596929e-05, "loss": 0.3307, "step": 18800 }, { "epoch": 1.95, "learning_rate": 3.087122024685215e-05, "loss": 0.8254, "step": 18801 }, { "epoch": 1.95, "learning_rate": 3.0868759861870165e-05, "loss": 0.3716, "step": 18802 }, { "epoch": 1.95, "learning_rate": 3.086629947103988e-05, "loss": 0.2538, "step": 18803 }, { "epoch": 1.95, "learning_rate": 3.0863839074377876e-05, "loss": 1.0647, "step": 18804 }, { "epoch": 1.95, "learning_rate": 3.0861378671900715e-05, "loss": 0.2559, "step": 18805 }, { "epoch": 1.95, "learning_rate": 3.0858918263624944e-05, "loss": 1.0288, "step": 18806 }, { "epoch": 1.95, "learning_rate": 3.085645784956715e-05, "loss": 0.424, "step": 18807 }, { "epoch": 1.95, "learning_rate": 3.0853997429743865e-05, "loss": 0.5113, "step": 18808 }, { "epoch": 1.95, "learning_rate": 3.085153700417168e-05, "loss": 0.1645, "step": 18809 }, { "epoch": 1.95, "learning_rate": 3.084907657286715e-05, "loss": 0.4344, "step": 18810 }, { "epoch": 1.95, "learning_rate": 3.0846616135846826e-05, "loss": 0.6262, "step": 18811 }, { "epoch": 1.95, "learning_rate": 3.084415569312728e-05, "loss": 1.8855, "step": 18812 }, { "epoch": 1.95, "learning_rate": 3.0841695244725084e-05, "loss": 0.1296, "step": 18813 }, { "epoch": 1.95, "learning_rate": 3.0839234790656786e-05, "loss": 0.4554, "step": 18814 }, { "epoch": 1.95, "learning_rate": 3.083677433093896e-05, "loss": 0.7002, "step": 18815 }, { "epoch": 1.95, "learning_rate": 3.083431386558816e-05, "loss": 0.1773, "step": 18816 }, { "epoch": 1.95, "learning_rate": 3.083185339462096e-05, "loss": 0.1591, "step": 18817 }, { "epoch": 1.95, "learning_rate": 3.082939291805391e-05, "loss": 1.2445, "step": 18818 }, { "epoch": 1.95, "learning_rate": 3.0826932435903594e-05, "loss": 0.5674, "step": 18819 }, { "epoch": 1.95, "learning_rate": 3.0824471948186554e-05, "loss": 0.314, "step": 18820 }, { "epoch": 1.95, "learning_rate": 3.082201145491936e-05, "loss": 0.2601, "step": 18821 }, { "epoch": 1.95, "learning_rate": 3.0819550956118575e-05, "loss": 0.4801, "step": 18822 }, { "epoch": 1.95, "learning_rate": 3.0817090451800767e-05, "loss": 0.5735, "step": 18823 }, { "epoch": 1.95, "learning_rate": 3.081462994198251e-05, "loss": 0.5695, "step": 18824 }, { "epoch": 1.95, "learning_rate": 3.0812169426680336e-05, "loss": 0.6598, "step": 18825 }, { "epoch": 1.95, "learning_rate": 3.080970890591084e-05, "loss": 0.4357, "step": 18826 }, { "epoch": 1.95, "learning_rate": 3.0807248379690565e-05, "loss": 0.808, "step": 18827 }, { "epoch": 1.95, "learning_rate": 3.080478784803609e-05, "loss": 0.2906, "step": 18828 }, { "epoch": 1.95, "learning_rate": 3.0802327310963964e-05, "loss": 0.7793, "step": 18829 }, { "epoch": 1.95, "learning_rate": 3.0799866768490764e-05, "loss": 1.2375, "step": 18830 }, { "epoch": 1.95, "learning_rate": 3.079740622063304e-05, "loss": 0.4208, "step": 18831 }, { "epoch": 1.95, "learning_rate": 3.0794945667407375e-05, "loss": 0.9971, "step": 18832 }, { "epoch": 1.95, "learning_rate": 3.0792485108830316e-05, "loss": 0.1895, "step": 18833 }, { "epoch": 1.95, "learning_rate": 3.079002454491844e-05, "loss": 0.6386, "step": 18834 }, { "epoch": 1.95, "learning_rate": 3.078756397568829e-05, "loss": 0.274, "step": 18835 }, { "epoch": 1.95, "learning_rate": 3.078510340115645e-05, "loss": 1.2615, "step": 18836 }, { "epoch": 1.95, "learning_rate": 3.078264282133948e-05, "loss": 0.3247, "step": 18837 }, { "epoch": 1.95, "learning_rate": 3.078018223625394e-05, "loss": 0.2474, "step": 18838 }, { "epoch": 1.95, "learning_rate": 3.07777216459164e-05, "loss": 0.4007, "step": 18839 }, { "epoch": 1.95, "learning_rate": 3.077526105034342e-05, "loss": 0.6391, "step": 18840 }, { "epoch": 1.95, "learning_rate": 3.077280044955155e-05, "loss": 1.2672, "step": 18841 }, { "epoch": 1.95, "learning_rate": 3.077033984355738e-05, "loss": 0.3698, "step": 18842 }, { "epoch": 1.95, "learning_rate": 3.076787923237746e-05, "loss": 0.2491, "step": 18843 }, { "epoch": 1.95, "learning_rate": 3.076541861602836e-05, "loss": 1.1456, "step": 18844 }, { "epoch": 1.95, "learning_rate": 3.076295799452664e-05, "loss": 0.6132, "step": 18845 }, { "epoch": 1.95, "learning_rate": 3.0760497367888855e-05, "loss": 0.8967, "step": 18846 }, { "epoch": 1.95, "learning_rate": 3.0758036736131585e-05, "loss": 0.6215, "step": 18847 }, { "epoch": 1.95, "learning_rate": 3.0755576099271396e-05, "loss": 0.2241, "step": 18848 }, { "epoch": 1.95, "learning_rate": 3.075311545732485e-05, "loss": 0.6113, "step": 18849 }, { "epoch": 1.95, "learning_rate": 3.075065481030849e-05, "loss": 0.8904, "step": 18850 }, { "epoch": 1.95, "learning_rate": 3.074819415823891e-05, "loss": 0.7334, "step": 18851 }, { "epoch": 1.95, "learning_rate": 3.0745733501132654e-05, "loss": 0.4258, "step": 18852 }, { "epoch": 1.95, "learning_rate": 3.07432728390063e-05, "loss": 1.0418, "step": 18853 }, { "epoch": 1.95, "learning_rate": 3.0740812171876404e-05, "loss": 0.4284, "step": 18854 }, { "epoch": 1.95, "learning_rate": 3.073835149975953e-05, "loss": 0.3694, "step": 18855 }, { "epoch": 1.95, "learning_rate": 3.073589082267226e-05, "loss": 0.825, "step": 18856 }, { "epoch": 1.95, "learning_rate": 3.0733430140631134e-05, "loss": 1.3328, "step": 18857 }, { "epoch": 1.95, "learning_rate": 3.0730969453652737e-05, "loss": 0.3607, "step": 18858 }, { "epoch": 1.95, "learning_rate": 3.0728508761753613e-05, "loss": 0.8221, "step": 18859 }, { "epoch": 1.95, "learning_rate": 3.072604806495034e-05, "loss": 0.7236, "step": 18860 }, { "epoch": 1.95, "learning_rate": 3.072358736325948e-05, "loss": 1.296, "step": 18861 }, { "epoch": 1.95, "learning_rate": 3.0721126656697614e-05, "loss": 0.5685, "step": 18862 }, { "epoch": 1.95, "learning_rate": 3.071866594528129e-05, "loss": 0.3192, "step": 18863 }, { "epoch": 1.95, "learning_rate": 3.071620522902706e-05, "loss": 0.2769, "step": 18864 }, { "epoch": 1.95, "learning_rate": 3.0713744507951514e-05, "loss": 0.3623, "step": 18865 }, { "epoch": 1.95, "learning_rate": 3.07112837820712e-05, "loss": 0.7845, "step": 18866 }, { "epoch": 1.95, "learning_rate": 3.070882305140269e-05, "loss": 0.4037, "step": 18867 }, { "epoch": 1.96, "learning_rate": 3.0706362315962565e-05, "loss": 0.2538, "step": 18868 }, { "epoch": 1.96, "learning_rate": 3.070390157576735e-05, "loss": 0.9295, "step": 18869 }, { "epoch": 1.96, "learning_rate": 3.070144083083365e-05, "loss": 1.445, "step": 18870 }, { "epoch": 1.96, "learning_rate": 3.069898008117801e-05, "loss": 0.3903, "step": 18871 }, { "epoch": 1.96, "learning_rate": 3.069651932681701e-05, "loss": 0.564, "step": 18872 }, { "epoch": 1.96, "learning_rate": 3.0694058567767194e-05, "loss": 0.5147, "step": 18873 }, { "epoch": 1.96, "learning_rate": 3.069159780404514e-05, "loss": 0.3553, "step": 18874 }, { "epoch": 1.96, "learning_rate": 3.0689137035667404e-05, "loss": 0.6741, "step": 18875 }, { "epoch": 1.96, "learning_rate": 3.068667626265057e-05, "loss": 0.2233, "step": 18876 }, { "epoch": 1.96, "learning_rate": 3.0684215485011195e-05, "loss": 1.107, "step": 18877 }, { "epoch": 1.96, "learning_rate": 3.068175470276584e-05, "loss": 0.9435, "step": 18878 }, { "epoch": 1.96, "learning_rate": 3.067929391593105e-05, "loss": 1.1649, "step": 18879 }, { "epoch": 1.96, "learning_rate": 3.0676833124523426e-05, "loss": 0.3268, "step": 18880 }, { "epoch": 1.96, "learning_rate": 3.067437232855953e-05, "loss": 0.9374, "step": 18881 }, { "epoch": 1.96, "learning_rate": 3.0671911528055915e-05, "loss": 0.618, "step": 18882 }, { "epoch": 1.96, "learning_rate": 3.066945072302914e-05, "loss": 0.655, "step": 18883 }, { "epoch": 1.96, "learning_rate": 3.066698991349578e-05, "loss": 1.1367, "step": 18884 }, { "epoch": 1.96, "learning_rate": 3.06645290994724e-05, "loss": 0.1192, "step": 18885 }, { "epoch": 1.96, "learning_rate": 3.066206828097557e-05, "loss": 1.2678, "step": 18886 }, { "epoch": 1.96, "learning_rate": 3.065960745802185e-05, "loss": 0.2882, "step": 18887 }, { "epoch": 1.96, "learning_rate": 3.065714663062781e-05, "loss": 0.6692, "step": 18888 }, { "epoch": 1.96, "learning_rate": 3.0654685798810015e-05, "loss": 0.8398, "step": 18889 }, { "epoch": 1.96, "learning_rate": 3.0652224962585016e-05, "loss": 1.1209, "step": 18890 }, { "epoch": 1.96, "learning_rate": 3.0649764121969404e-05, "loss": 0.4957, "step": 18891 }, { "epoch": 1.96, "learning_rate": 3.064730327697973e-05, "loss": 0.355, "step": 18892 }, { "epoch": 1.96, "learning_rate": 3.064484242763255e-05, "loss": 0.5446, "step": 18893 }, { "epoch": 1.96, "learning_rate": 3.064238157394445e-05, "loss": 1.1751, "step": 18894 }, { "epoch": 1.96, "learning_rate": 3.0639920715931985e-05, "loss": 0.5506, "step": 18895 }, { "epoch": 1.96, "learning_rate": 3.063745985361173e-05, "loss": 0.2609, "step": 18896 }, { "epoch": 1.96, "learning_rate": 3.063499898700024e-05, "loss": 1.02, "step": 18897 }, { "epoch": 1.96, "learning_rate": 3.063253811611408e-05, "loss": 0.495, "step": 18898 }, { "epoch": 1.96, "learning_rate": 3.063007724096982e-05, "loss": 0.3988, "step": 18899 }, { "epoch": 1.96, "learning_rate": 3.062761636158404e-05, "loss": 0.1772, "step": 18900 }, { "epoch": 1.96, "learning_rate": 3.0625155477973296e-05, "loss": 1.2596, "step": 18901 }, { "epoch": 1.96, "learning_rate": 3.062269459015414e-05, "loss": 0.2433, "step": 18902 }, { "epoch": 1.96, "learning_rate": 3.062023369814315e-05, "loss": 1.1977, "step": 18903 }, { "epoch": 1.96, "learning_rate": 3.0617772801956897e-05, "loss": 0.9719, "step": 18904 }, { "epoch": 1.96, "learning_rate": 3.0615311901611944e-05, "loss": 0.9926, "step": 18905 }, { "epoch": 1.96, "learning_rate": 3.061285099712485e-05, "loss": 0.3742, "step": 18906 }, { "epoch": 1.96, "learning_rate": 3.0610390088512185e-05, "loss": 0.9186, "step": 18907 }, { "epoch": 1.96, "learning_rate": 3.0607929175790516e-05, "loss": 0.7674, "step": 18908 }, { "epoch": 1.96, "learning_rate": 3.0605468258976417e-05, "loss": 0.333, "step": 18909 }, { "epoch": 1.96, "learning_rate": 3.060300733808645e-05, "loss": 0.836, "step": 18910 }, { "epoch": 1.96, "learning_rate": 3.060054641313717e-05, "loss": 0.7597, "step": 18911 }, { "epoch": 1.96, "learning_rate": 3.059808548414515e-05, "loss": 0.8498, "step": 18912 }, { "epoch": 1.96, "learning_rate": 3.0595624551126955e-05, "loss": 0.4512, "step": 18913 }, { "epoch": 1.96, "learning_rate": 3.0593163614099174e-05, "loss": 0.7311, "step": 18914 }, { "epoch": 1.96, "learning_rate": 3.0590702673078344e-05, "loss": 0.5607, "step": 18915 }, { "epoch": 1.96, "learning_rate": 3.058824172808103e-05, "loss": 0.803, "step": 18916 }, { "epoch": 1.96, "learning_rate": 3.058578077912382e-05, "loss": 0.2622, "step": 18917 }, { "epoch": 1.96, "learning_rate": 3.058331982622327e-05, "loss": 0.932, "step": 18918 }, { "epoch": 1.96, "learning_rate": 3.058085886939596e-05, "loss": 0.8928, "step": 18919 }, { "epoch": 1.96, "learning_rate": 3.057839790865844e-05, "loss": 0.5162, "step": 18920 }, { "epoch": 1.96, "learning_rate": 3.0575936944027264e-05, "loss": 0.8785, "step": 18921 }, { "epoch": 1.96, "learning_rate": 3.057347597551903e-05, "loss": 0.1627, "step": 18922 }, { "epoch": 1.96, "learning_rate": 3.057101500315028e-05, "loss": 0.2286, "step": 18923 }, { "epoch": 1.96, "learning_rate": 3.05685540269376e-05, "loss": 0.442, "step": 18924 }, { "epoch": 1.96, "learning_rate": 3.056609304689754e-05, "loss": 1.2155, "step": 18925 }, { "epoch": 1.96, "learning_rate": 3.056363206304668e-05, "loss": 0.2068, "step": 18926 }, { "epoch": 1.96, "learning_rate": 3.0561171075401574e-05, "loss": 0.5506, "step": 18927 }, { "epoch": 1.96, "learning_rate": 3.05587100839788e-05, "loss": 0.8798, "step": 18928 }, { "epoch": 1.96, "learning_rate": 3.055624908879493e-05, "loss": 0.2963, "step": 18929 }, { "epoch": 1.96, "learning_rate": 3.055378808986651e-05, "loss": 1.2222, "step": 18930 }, { "epoch": 1.96, "learning_rate": 3.055132708721011e-05, "loss": 0.8363, "step": 18931 }, { "epoch": 1.96, "learning_rate": 3.054886608084232e-05, "loss": 0.7139, "step": 18932 }, { "epoch": 1.96, "learning_rate": 3.054640507077969e-05, "loss": 0.6447, "step": 18933 }, { "epoch": 1.96, "learning_rate": 3.054394405703878e-05, "loss": 0.6626, "step": 18934 }, { "epoch": 1.96, "learning_rate": 3.054148303963618e-05, "loss": 0.2803, "step": 18935 }, { "epoch": 1.96, "learning_rate": 3.053902201858843e-05, "loss": 0.1648, "step": 18936 }, { "epoch": 1.96, "learning_rate": 3.0536560993912105e-05, "loss": 0.5008, "step": 18937 }, { "epoch": 1.96, "learning_rate": 3.0534099965623794e-05, "loss": 0.5095, "step": 18938 }, { "epoch": 1.96, "learning_rate": 3.0531638933740046e-05, "loss": 0.6384, "step": 18939 }, { "epoch": 1.96, "learning_rate": 3.052917789827742e-05, "loss": 0.9562, "step": 18940 }, { "epoch": 1.96, "learning_rate": 3.0526716859252486e-05, "loss": 1.186, "step": 18941 }, { "epoch": 1.96, "learning_rate": 3.0524255816681824e-05, "loss": 0.2559, "step": 18942 }, { "epoch": 1.96, "learning_rate": 3.0521794770582e-05, "loss": 1.2184, "step": 18943 }, { "epoch": 1.96, "learning_rate": 3.051933372096958e-05, "loss": 0.5036, "step": 18944 }, { "epoch": 1.96, "learning_rate": 3.0516872667861117e-05, "loss": 0.4295, "step": 18945 }, { "epoch": 1.96, "learning_rate": 3.0514411611273182e-05, "loss": 0.2697, "step": 18946 }, { "epoch": 1.96, "learning_rate": 3.051195055122236e-05, "loss": 0.493, "step": 18947 }, { "epoch": 1.96, "learning_rate": 3.0509489487725203e-05, "loss": 0.6144, "step": 18948 }, { "epoch": 1.96, "learning_rate": 3.0507028420798278e-05, "loss": 0.5082, "step": 18949 }, { "epoch": 1.96, "learning_rate": 3.050456735045816e-05, "loss": 0.5987, "step": 18950 }, { "epoch": 1.96, "learning_rate": 3.0502106276721422e-05, "loss": 0.5961, "step": 18951 }, { "epoch": 1.96, "learning_rate": 3.049964519960461e-05, "loss": 0.1475, "step": 18952 }, { "epoch": 1.96, "learning_rate": 3.0497184119124308e-05, "loss": 1.1742, "step": 18953 }, { "epoch": 1.96, "learning_rate": 3.0494723035297073e-05, "loss": 0.3946, "step": 18954 }, { "epoch": 1.96, "learning_rate": 3.0492261948139485e-05, "loss": 0.9568, "step": 18955 }, { "epoch": 1.96, "learning_rate": 3.0489800857668094e-05, "loss": 0.2636, "step": 18956 }, { "epoch": 1.96, "learning_rate": 3.0487339763899495e-05, "loss": 0.8893, "step": 18957 }, { "epoch": 1.96, "learning_rate": 3.0484878666850232e-05, "loss": 0.3125, "step": 18958 }, { "epoch": 1.96, "learning_rate": 3.0482417566536877e-05, "loss": 0.9922, "step": 18959 }, { "epoch": 1.96, "learning_rate": 3.0479956462975994e-05, "loss": 0.645, "step": 18960 }, { "epoch": 1.96, "learning_rate": 3.0477495356184165e-05, "loss": 0.5429, "step": 18961 }, { "epoch": 1.96, "learning_rate": 3.0475034246177948e-05, "loss": 0.6099, "step": 18962 }, { "epoch": 1.97, "learning_rate": 3.047257313297391e-05, "loss": 0.9586, "step": 18963 }, { "epoch": 1.97, "learning_rate": 3.0470112016588626e-05, "loss": 0.4352, "step": 18964 }, { "epoch": 1.97, "learning_rate": 3.046765089703865e-05, "loss": 0.2554, "step": 18965 }, { "epoch": 1.97, "learning_rate": 3.0465189774340557e-05, "loss": 0.293, "step": 18966 }, { "epoch": 1.97, "learning_rate": 3.0462728648510924e-05, "loss": 0.4095, "step": 18967 }, { "epoch": 1.97, "learning_rate": 3.0460267519566303e-05, "loss": 0.5596, "step": 18968 }, { "epoch": 1.97, "learning_rate": 3.0457806387523267e-05, "loss": 0.2351, "step": 18969 }, { "epoch": 1.97, "learning_rate": 3.0455345252398396e-05, "loss": 0.8531, "step": 18970 }, { "epoch": 1.97, "learning_rate": 3.0452884114208235e-05, "loss": 1.0925, "step": 18971 }, { "epoch": 1.97, "learning_rate": 3.0450422972969375e-05, "loss": 0.7431, "step": 18972 }, { "epoch": 1.97, "learning_rate": 3.0447961828698357e-05, "loss": 1.0146, "step": 18973 }, { "epoch": 1.97, "learning_rate": 3.0445500681411777e-05, "loss": 0.3503, "step": 18974 }, { "epoch": 1.97, "learning_rate": 3.0443039531126185e-05, "loss": 0.7026, "step": 18975 }, { "epoch": 1.97, "learning_rate": 3.0440578377858167e-05, "loss": 0.9881, "step": 18976 }, { "epoch": 1.97, "learning_rate": 3.0438117221624266e-05, "loss": 0.3002, "step": 18977 }, { "epoch": 1.97, "learning_rate": 3.0435656062441067e-05, "loss": 0.2238, "step": 18978 }, { "epoch": 1.97, "learning_rate": 3.0433194900325132e-05, "loss": 1.3312, "step": 18979 }, { "epoch": 1.97, "learning_rate": 3.0430733735293025e-05, "loss": 0.5908, "step": 18980 }, { "epoch": 1.97, "learning_rate": 3.0428272567361328e-05, "loss": 0.2746, "step": 18981 }, { "epoch": 1.97, "learning_rate": 3.04258113965466e-05, "loss": 1.2809, "step": 18982 }, { "epoch": 1.97, "learning_rate": 3.0423350222865408e-05, "loss": 1.0619, "step": 18983 }, { "epoch": 1.97, "learning_rate": 3.042088904633432e-05, "loss": 0.9158, "step": 18984 }, { "epoch": 1.97, "learning_rate": 3.0418427866969894e-05, "loss": 0.8878, "step": 18985 }, { "epoch": 1.97, "learning_rate": 3.041596668478873e-05, "loss": 0.555, "step": 18986 }, { "epoch": 1.97, "learning_rate": 3.0413505499807362e-05, "loss": 0.9529, "step": 18987 }, { "epoch": 1.97, "learning_rate": 3.041104431204238e-05, "loss": 0.4643, "step": 18988 }, { "epoch": 1.97, "learning_rate": 3.040858312151034e-05, "loss": 0.8601, "step": 18989 }, { "epoch": 1.97, "learning_rate": 3.040612192822781e-05, "loss": 0.6386, "step": 18990 }, { "epoch": 1.97, "learning_rate": 3.0403660732211372e-05, "loss": 1.1162, "step": 18991 }, { "epoch": 1.97, "learning_rate": 3.0401199533477576e-05, "loss": 0.5043, "step": 18992 }, { "epoch": 1.97, "learning_rate": 3.0398738332042995e-05, "loss": 0.1973, "step": 18993 }, { "epoch": 1.97, "learning_rate": 3.0396277127924207e-05, "loss": 0.4566, "step": 18994 }, { "epoch": 1.97, "learning_rate": 3.0393815921137777e-05, "loss": 1.11, "step": 18995 }, { "epoch": 1.97, "learning_rate": 3.039135471170027e-05, "loss": 0.4822, "step": 18996 }, { "epoch": 1.97, "learning_rate": 3.038889349962825e-05, "loss": 0.6589, "step": 18997 }, { "epoch": 1.97, "learning_rate": 3.038643228493829e-05, "loss": 0.5352, "step": 18998 }, { "epoch": 1.97, "learning_rate": 3.0383971067646958e-05, "loss": 0.9781, "step": 18999 }, { "epoch": 1.97, "learning_rate": 3.0381509847770826e-05, "loss": 0.4218, "step": 19000 }, { "epoch": 1.97, "learning_rate": 3.0379048625326457e-05, "loss": 0.3218, "step": 19001 }, { "epoch": 1.97, "learning_rate": 3.0376587400330428e-05, "loss": 0.8717, "step": 19002 }, { "epoch": 1.97, "learning_rate": 3.0374126172799294e-05, "loss": 0.5357, "step": 19003 }, { "epoch": 1.97, "learning_rate": 3.037166494274963e-05, "loss": 0.2566, "step": 19004 }, { "epoch": 1.97, "learning_rate": 3.036920371019801e-05, "loss": 0.5803, "step": 19005 }, { "epoch": 1.97, "learning_rate": 3.0366742475160987e-05, "loss": 0.2135, "step": 19006 }, { "epoch": 1.97, "learning_rate": 3.0364281237655146e-05, "loss": 0.8062, "step": 19007 }, { "epoch": 1.97, "learning_rate": 3.0361819997697052e-05, "loss": 0.1082, "step": 19008 }, { "epoch": 1.97, "learning_rate": 3.0359358755303265e-05, "loss": 0.692, "step": 19009 }, { "epoch": 1.97, "learning_rate": 3.035689751049036e-05, "loss": 0.4891, "step": 19010 }, { "epoch": 1.97, "learning_rate": 3.0354436263274906e-05, "loss": 0.278, "step": 19011 }, { "epoch": 1.97, "learning_rate": 3.0351975013673468e-05, "loss": 1.2581, "step": 19012 }, { "epoch": 1.97, "learning_rate": 3.034951376170261e-05, "loss": 0.9361, "step": 19013 }, { "epoch": 1.97, "learning_rate": 3.034705250737892e-05, "loss": 0.7444, "step": 19014 }, { "epoch": 1.97, "learning_rate": 3.0344591250718953e-05, "loss": 0.1991, "step": 19015 }, { "epoch": 1.97, "learning_rate": 3.0342129991739268e-05, "loss": 0.2754, "step": 19016 }, { "epoch": 1.97, "learning_rate": 3.0339668730456453e-05, "loss": 0.8261, "step": 19017 }, { "epoch": 1.97, "learning_rate": 3.033720746688706e-05, "loss": 1.1262, "step": 19018 }, { "epoch": 1.97, "learning_rate": 3.0334746201047677e-05, "loss": 0.4337, "step": 19019 }, { "epoch": 1.97, "learning_rate": 3.0332284932954855e-05, "loss": 0.8131, "step": 19020 }, { "epoch": 1.97, "learning_rate": 3.0329823662625162e-05, "loss": 0.5943, "step": 19021 }, { "epoch": 1.97, "learning_rate": 3.032736239007518e-05, "loss": 0.4861, "step": 19022 }, { "epoch": 1.97, "learning_rate": 3.0324901115321467e-05, "loss": 0.2686, "step": 19023 }, { "epoch": 1.97, "learning_rate": 3.0322439838380597e-05, "loss": 0.872, "step": 19024 }, { "epoch": 1.97, "learning_rate": 3.0319978559269134e-05, "loss": 0.2416, "step": 19025 }, { "epoch": 1.97, "learning_rate": 3.0317517278003656e-05, "loss": 0.4606, "step": 19026 }, { "epoch": 1.97, "learning_rate": 3.0315055994600728e-05, "loss": 0.2938, "step": 19027 }, { "epoch": 1.97, "learning_rate": 3.031259470907691e-05, "loss": 1.1277, "step": 19028 }, { "epoch": 1.97, "learning_rate": 3.0310133421448788e-05, "loss": 0.3985, "step": 19029 }, { "epoch": 1.97, "learning_rate": 3.030767213173291e-05, "loss": 0.5705, "step": 19030 }, { "epoch": 1.97, "learning_rate": 3.0305210839945856e-05, "loss": 0.4272, "step": 19031 }, { "epoch": 1.97, "learning_rate": 3.0302749546104192e-05, "loss": 0.2416, "step": 19032 }, { "epoch": 1.97, "learning_rate": 3.0300288250224496e-05, "loss": 0.5246, "step": 19033 }, { "epoch": 1.97, "learning_rate": 3.0297826952323335e-05, "loss": 0.6975, "step": 19034 }, { "epoch": 1.97, "learning_rate": 3.029536565241726e-05, "loss": 1.108, "step": 19035 }, { "epoch": 1.97, "learning_rate": 3.0292904350522857e-05, "loss": 0.5423, "step": 19036 }, { "epoch": 1.97, "learning_rate": 3.029044304665669e-05, "loss": 1.109, "step": 19037 }, { "epoch": 1.97, "learning_rate": 3.0287981740835328e-05, "loss": 0.8093, "step": 19038 }, { "epoch": 1.97, "learning_rate": 3.028552043307535e-05, "loss": 0.7699, "step": 19039 }, { "epoch": 1.97, "learning_rate": 3.0283059123393303e-05, "loss": 0.9542, "step": 19040 }, { "epoch": 1.97, "learning_rate": 3.0280597811805766e-05, "loss": 0.2409, "step": 19041 }, { "epoch": 1.97, "learning_rate": 3.0278136498329316e-05, "loss": 0.6855, "step": 19042 }, { "epoch": 1.97, "learning_rate": 3.0275675182980523e-05, "loss": 0.2701, "step": 19043 }, { "epoch": 1.97, "learning_rate": 3.027321386577594e-05, "loss": 0.6348, "step": 19044 }, { "epoch": 1.97, "learning_rate": 3.027075254673214e-05, "loss": 0.9749, "step": 19045 }, { "epoch": 1.97, "learning_rate": 3.0268291225865714e-05, "loss": 0.6689, "step": 19046 }, { "epoch": 1.97, "learning_rate": 3.0265829903193197e-05, "loss": 0.8727, "step": 19047 }, { "epoch": 1.97, "learning_rate": 3.0263368578731194e-05, "loss": 0.2051, "step": 19048 }, { "epoch": 1.97, "learning_rate": 3.0260907252496242e-05, "loss": 1.065, "step": 19049 }, { "epoch": 1.97, "learning_rate": 3.025844592450492e-05, "loss": 0.5236, "step": 19050 }, { "epoch": 1.97, "learning_rate": 3.0255984594773806e-05, "loss": 0.334, "step": 19051 }, { "epoch": 1.97, "learning_rate": 3.0253523263319467e-05, "loss": 0.0912, "step": 19052 }, { "epoch": 1.97, "learning_rate": 3.025106193015847e-05, "loss": 0.6327, "step": 19053 }, { "epoch": 1.97, "learning_rate": 3.0248600595307373e-05, "loss": 1.2535, "step": 19054 }, { "epoch": 1.97, "learning_rate": 3.024613925878276e-05, "loss": 0.119, "step": 19055 }, { "epoch": 1.97, "learning_rate": 3.0243677920601195e-05, "loss": 0.4793, "step": 19056 }, { "epoch": 1.97, "learning_rate": 3.0241216580779245e-05, "loss": 0.3763, "step": 19057 }, { "epoch": 1.97, "learning_rate": 3.02387552393335e-05, "loss": 1.0962, "step": 19058 }, { "epoch": 1.98, "learning_rate": 3.0236293896280488e-05, "loss": 0.2064, "step": 19059 }, { "epoch": 1.98, "learning_rate": 3.0233832551636806e-05, "loss": 0.6668, "step": 19060 }, { "epoch": 1.98, "learning_rate": 3.023137120541902e-05, "loss": 0.7568, "step": 19061 }, { "epoch": 1.98, "learning_rate": 3.0228909857643698e-05, "loss": 0.3731, "step": 19062 }, { "epoch": 1.98, "learning_rate": 3.0226448508327406e-05, "loss": 0.4489, "step": 19063 }, { "epoch": 1.98, "learning_rate": 3.0223987157486723e-05, "loss": 0.3065, "step": 19064 }, { "epoch": 1.98, "learning_rate": 3.02215258051382e-05, "loss": 0.6651, "step": 19065 }, { "epoch": 1.98, "learning_rate": 3.021906445129842e-05, "loss": 0.6493, "step": 19066 }, { "epoch": 1.98, "learning_rate": 3.0216603095983957e-05, "loss": 0.5962, "step": 19067 }, { "epoch": 1.98, "learning_rate": 3.0214141739211364e-05, "loss": 0.8749, "step": 19068 }, { "epoch": 1.98, "learning_rate": 3.0211680380997218e-05, "loss": 0.4799, "step": 19069 }, { "epoch": 1.98, "learning_rate": 3.0209219021358096e-05, "loss": 0.2424, "step": 19070 }, { "epoch": 1.98, "learning_rate": 3.0206757660310554e-05, "loss": 0.3983, "step": 19071 }, { "epoch": 1.98, "learning_rate": 3.0204296297871178e-05, "loss": 0.4958, "step": 19072 }, { "epoch": 1.98, "learning_rate": 3.020183493405652e-05, "loss": 0.3019, "step": 19073 }, { "epoch": 1.98, "learning_rate": 3.0199373568883154e-05, "loss": 0.4143, "step": 19074 }, { "epoch": 1.98, "learning_rate": 3.0196912202367655e-05, "loss": 1.0814, "step": 19075 }, { "epoch": 1.98, "learning_rate": 3.0194450834526584e-05, "loss": 0.6082, "step": 19076 }, { "epoch": 1.98, "learning_rate": 3.0191989465376532e-05, "loss": 0.8907, "step": 19077 }, { "epoch": 1.98, "learning_rate": 3.0189528094934036e-05, "loss": 1.0263, "step": 19078 }, { "epoch": 1.98, "learning_rate": 3.0187066723215685e-05, "loss": 1.0668, "step": 19079 }, { "epoch": 1.98, "learning_rate": 3.0184605350238043e-05, "loss": 0.5745, "step": 19080 }, { "epoch": 1.98, "learning_rate": 3.018214397601769e-05, "loss": 0.4493, "step": 19081 }, { "epoch": 1.98, "learning_rate": 3.0179682600571173e-05, "loss": 1.1974, "step": 19082 }, { "epoch": 1.98, "learning_rate": 3.0177221223915088e-05, "loss": 1.2204, "step": 19083 }, { "epoch": 1.98, "learning_rate": 3.0174759846065984e-05, "loss": 0.6343, "step": 19084 }, { "epoch": 1.98, "learning_rate": 3.0172298467040432e-05, "loss": 0.2648, "step": 19085 }, { "epoch": 1.98, "learning_rate": 3.0169837086855014e-05, "loss": 0.39, "step": 19086 }, { "epoch": 1.98, "learning_rate": 3.0167375705526292e-05, "loss": 0.6551, "step": 19087 }, { "epoch": 1.98, "learning_rate": 3.0164914323070832e-05, "loss": 0.5381, "step": 19088 }, { "epoch": 1.98, "learning_rate": 3.0162452939505204e-05, "loss": 0.8079, "step": 19089 }, { "epoch": 1.98, "learning_rate": 3.0159991554845993e-05, "loss": 0.8654, "step": 19090 }, { "epoch": 1.98, "learning_rate": 3.0157530169109758e-05, "loss": 0.6865, "step": 19091 }, { "epoch": 1.98, "learning_rate": 3.0155068782313055e-05, "loss": 0.2636, "step": 19092 }, { "epoch": 1.98, "learning_rate": 3.015260739447247e-05, "loss": 0.7796, "step": 19093 }, { "epoch": 1.98, "learning_rate": 3.0150146005604563e-05, "loss": 0.4746, "step": 19094 }, { "epoch": 1.98, "learning_rate": 3.0147684615725914e-05, "loss": 0.4193, "step": 19095 }, { "epoch": 1.98, "learning_rate": 3.0145223224853092e-05, "loss": 1.0307, "step": 19096 }, { "epoch": 1.98, "learning_rate": 3.0142761833002655e-05, "loss": 0.7049, "step": 19097 }, { "epoch": 1.98, "learning_rate": 3.0140300440191178e-05, "loss": 0.3077, "step": 19098 }, { "epoch": 1.98, "learning_rate": 3.013783904643523e-05, "loss": 0.2587, "step": 19099 }, { "epoch": 1.98, "learning_rate": 3.0135377651751378e-05, "loss": 0.6193, "step": 19100 }, { "epoch": 1.98, "learning_rate": 3.0132916256156205e-05, "loss": 0.7591, "step": 19101 }, { "epoch": 1.98, "learning_rate": 3.013045485966627e-05, "loss": 1.0007, "step": 19102 }, { "epoch": 1.98, "learning_rate": 3.012799346229814e-05, "loss": 0.5614, "step": 19103 }, { "epoch": 1.98, "learning_rate": 3.0125532064068385e-05, "loss": 0.5788, "step": 19104 }, { "epoch": 1.98, "learning_rate": 3.0123070664993584e-05, "loss": 0.138, "step": 19105 }, { "epoch": 1.98, "learning_rate": 3.0120609265090293e-05, "loss": 0.1459, "step": 19106 }, { "epoch": 1.98, "learning_rate": 3.011814786437509e-05, "loss": 0.5853, "step": 19107 }, { "epoch": 1.98, "learning_rate": 3.0115686462864554e-05, "loss": 1.3343, "step": 19108 }, { "epoch": 1.98, "learning_rate": 3.0113225060575233e-05, "loss": 0.3372, "step": 19109 }, { "epoch": 1.98, "learning_rate": 3.0110763657523713e-05, "loss": 0.2714, "step": 19110 }, { "epoch": 1.98, "learning_rate": 3.010830225372655e-05, "loss": 0.6006, "step": 19111 }, { "epoch": 1.98, "learning_rate": 3.0105840849200323e-05, "loss": 0.4235, "step": 19112 }, { "epoch": 1.98, "learning_rate": 3.01033794439616e-05, "loss": 0.9483, "step": 19113 }, { "epoch": 1.98, "learning_rate": 3.0100918038026956e-05, "loss": 0.7417, "step": 19114 }, { "epoch": 1.98, "learning_rate": 3.0098456631412965e-05, "loss": 1.4593, "step": 19115 }, { "epoch": 1.98, "learning_rate": 3.0095995224136174e-05, "loss": 0.3114, "step": 19116 }, { "epoch": 1.98, "learning_rate": 3.0093533816213165e-05, "loss": 0.0825, "step": 19117 }, { "epoch": 1.98, "learning_rate": 3.0091072407660505e-05, "loss": 0.4735, "step": 19118 }, { "epoch": 1.98, "learning_rate": 3.008861099849477e-05, "loss": 0.4941, "step": 19119 }, { "epoch": 1.98, "learning_rate": 3.008614958873253e-05, "loss": 0.5778, "step": 19120 }, { "epoch": 1.98, "learning_rate": 3.0083688178390358e-05, "loss": 0.7107, "step": 19121 }, { "epoch": 1.98, "learning_rate": 3.0081226767484807e-05, "loss": 0.1191, "step": 19122 }, { "epoch": 1.98, "learning_rate": 3.0078765356032456e-05, "loss": 0.1533, "step": 19123 }, { "epoch": 1.98, "learning_rate": 3.0076303944049882e-05, "loss": 0.6024, "step": 19124 }, { "epoch": 1.98, "learning_rate": 3.0073842531553637e-05, "loss": 1.0965, "step": 19125 }, { "epoch": 1.98, "learning_rate": 3.0071381118560302e-05, "loss": 1.0176, "step": 19126 }, { "epoch": 1.98, "learning_rate": 3.0068919705086463e-05, "loss": 0.1877, "step": 19127 }, { "epoch": 1.98, "learning_rate": 3.006645829114866e-05, "loss": 0.5292, "step": 19128 }, { "epoch": 1.98, "learning_rate": 3.0063996876763476e-05, "loss": 0.9412, "step": 19129 }, { "epoch": 1.98, "learning_rate": 3.0061535461947476e-05, "loss": 0.5701, "step": 19130 }, { "epoch": 1.98, "learning_rate": 3.005907404671724e-05, "loss": 0.1152, "step": 19131 }, { "epoch": 1.98, "learning_rate": 3.0056612631089327e-05, "loss": 0.6677, "step": 19132 }, { "epoch": 1.98, "learning_rate": 3.0054151215080307e-05, "loss": 0.8863, "step": 19133 }, { "epoch": 1.98, "learning_rate": 3.005168979870677e-05, "loss": 0.5321, "step": 19134 }, { "epoch": 1.98, "learning_rate": 3.004922838198526e-05, "loss": 0.76, "step": 19135 }, { "epoch": 1.98, "learning_rate": 3.0046766964932353e-05, "loss": 0.3499, "step": 19136 }, { "epoch": 1.98, "learning_rate": 3.0044305547564618e-05, "loss": 1.3365, "step": 19137 }, { "epoch": 1.98, "learning_rate": 3.004184412989863e-05, "loss": 0.3225, "step": 19138 }, { "epoch": 1.98, "learning_rate": 3.0039382711950965e-05, "loss": 0.352, "step": 19139 }, { "epoch": 1.98, "learning_rate": 3.003692129373819e-05, "loss": 0.8815, "step": 19140 }, { "epoch": 1.98, "learning_rate": 3.0034459875276857e-05, "loss": 1.1738, "step": 19141 }, { "epoch": 1.98, "learning_rate": 3.003199845658355e-05, "loss": 0.3977, "step": 19142 }, { "epoch": 1.98, "learning_rate": 3.0029537037674835e-05, "loss": 0.2104, "step": 19143 }, { "epoch": 1.98, "learning_rate": 3.002707561856729e-05, "loss": 0.9191, "step": 19144 }, { "epoch": 1.98, "learning_rate": 3.0024614199277476e-05, "loss": 0.5774, "step": 19145 }, { "epoch": 1.98, "learning_rate": 3.002215277982197e-05, "loss": 0.8362, "step": 19146 }, { "epoch": 1.98, "learning_rate": 3.0019691360217327e-05, "loss": 0.9537, "step": 19147 }, { "epoch": 1.98, "learning_rate": 3.0017229940480137e-05, "loss": 1.0503, "step": 19148 }, { "epoch": 1.98, "learning_rate": 3.001476852062695e-05, "loss": 0.565, "step": 19149 }, { "epoch": 1.98, "learning_rate": 3.001230710067435e-05, "loss": 0.5732, "step": 19150 }, { "epoch": 1.98, "learning_rate": 3.0009845680638897e-05, "loss": 1.0196, "step": 19151 }, { "epoch": 1.98, "learning_rate": 3.000738426053717e-05, "loss": 0.5843, "step": 19152 }, { "epoch": 1.98, "learning_rate": 3.0004922840385736e-05, "loss": 0.2697, "step": 19153 }, { "epoch": 1.98, "learning_rate": 3.0002461420201156e-05, "loss": 0.6376, "step": 19154 }, { "epoch": 1.99, "learning_rate": 3e-05, "loss": 0.199, "step": 19155 }, { "epoch": 1.99, "learning_rate": 2.9997538579798853e-05, "loss": 0.6723, "step": 19156 }, { "epoch": 1.99, "learning_rate": 2.9995077159614273e-05, "loss": 0.5048, "step": 19157 }, { "epoch": 1.99, "learning_rate": 2.999261573946284e-05, "loss": 0.6409, "step": 19158 }, { "epoch": 1.99, "learning_rate": 2.9990154319361105e-05, "loss": 0.4381, "step": 19159 }, { "epoch": 1.99, "learning_rate": 2.998769289932565e-05, "loss": 0.9025, "step": 19160 }, { "epoch": 1.99, "learning_rate": 2.998523147937305e-05, "loss": 0.3376, "step": 19161 }, { "epoch": 1.99, "learning_rate": 2.9982770059519872e-05, "loss": 0.9885, "step": 19162 }, { "epoch": 1.99, "learning_rate": 2.998030863978268e-05, "loss": 0.7758, "step": 19163 }, { "epoch": 1.99, "learning_rate": 2.997784722017804e-05, "loss": 0.4216, "step": 19164 }, { "epoch": 1.99, "learning_rate": 2.9975385800722525e-05, "loss": 0.6482, "step": 19165 }, { "epoch": 1.99, "learning_rate": 2.997292438143271e-05, "loss": 0.5558, "step": 19166 }, { "epoch": 1.99, "learning_rate": 2.9970462962325166e-05, "loss": 0.6939, "step": 19167 }, { "epoch": 1.99, "learning_rate": 2.9968001543416455e-05, "loss": 0.5064, "step": 19168 }, { "epoch": 1.99, "learning_rate": 2.9965540124723144e-05, "loss": 0.6237, "step": 19169 }, { "epoch": 1.99, "learning_rate": 2.9963078706261815e-05, "loss": 0.8736, "step": 19170 }, { "epoch": 1.99, "learning_rate": 2.996061728804903e-05, "loss": 1.1146, "step": 19171 }, { "epoch": 1.99, "learning_rate": 2.9958155870101365e-05, "loss": 0.6705, "step": 19172 }, { "epoch": 1.99, "learning_rate": 2.995569445243538e-05, "loss": 0.7047, "step": 19173 }, { "epoch": 1.99, "learning_rate": 2.9953233035067662e-05, "loss": 0.3031, "step": 19174 }, { "epoch": 1.99, "learning_rate": 2.9950771618014753e-05, "loss": 0.9232, "step": 19175 }, { "epoch": 1.99, "learning_rate": 2.9948310201293243e-05, "loss": 0.9414, "step": 19176 }, { "epoch": 1.99, "learning_rate": 2.9945848784919698e-05, "loss": 1.2156, "step": 19177 }, { "epoch": 1.99, "learning_rate": 2.9943387368910682e-05, "loss": 0.4235, "step": 19178 }, { "epoch": 1.99, "learning_rate": 2.994092595328277e-05, "loss": 0.4786, "step": 19179 }, { "epoch": 1.99, "learning_rate": 2.993846453805253e-05, "loss": 0.609, "step": 19180 }, { "epoch": 1.99, "learning_rate": 2.9936003123236532e-05, "loss": 0.9758, "step": 19181 }, { "epoch": 1.99, "learning_rate": 2.993354170885135e-05, "loss": 1.1199, "step": 19182 }, { "epoch": 1.99, "learning_rate": 2.9931080294913546e-05, "loss": 0.3278, "step": 19183 }, { "epoch": 1.99, "learning_rate": 2.9928618881439696e-05, "loss": 0.2043, "step": 19184 }, { "epoch": 1.99, "learning_rate": 2.992615746844636e-05, "loss": 0.4835, "step": 19185 }, { "epoch": 1.99, "learning_rate": 2.9923696055950126e-05, "loss": 1.4117, "step": 19186 }, { "epoch": 1.99, "learning_rate": 2.992123464396756e-05, "loss": 0.1682, "step": 19187 }, { "epoch": 1.99, "learning_rate": 2.9918773232515208e-05, "loss": 1.1445, "step": 19188 }, { "epoch": 1.99, "learning_rate": 2.9916311821609653e-05, "loss": 1.3156, "step": 19189 }, { "epoch": 1.99, "learning_rate": 2.9913850411267474e-05, "loss": 0.3898, "step": 19190 }, { "epoch": 1.99, "learning_rate": 2.9911389001505235e-05, "loss": 0.3165, "step": 19191 }, { "epoch": 1.99, "learning_rate": 2.99089275923395e-05, "loss": 0.8705, "step": 19192 }, { "epoch": 1.99, "learning_rate": 2.9906466183786843e-05, "loss": 0.9606, "step": 19193 }, { "epoch": 1.99, "learning_rate": 2.990400477586383e-05, "loss": 0.6021, "step": 19194 }, { "epoch": 1.99, "learning_rate": 2.9901543368587044e-05, "loss": 0.4804, "step": 19195 }, { "epoch": 1.99, "learning_rate": 2.9899081961973042e-05, "loss": 0.9544, "step": 19196 }, { "epoch": 1.99, "learning_rate": 2.9896620556038396e-05, "loss": 0.8652, "step": 19197 }, { "epoch": 1.99, "learning_rate": 2.9894159150799675e-05, "loss": 0.758, "step": 19198 }, { "epoch": 1.99, "learning_rate": 2.989169774627346e-05, "loss": 1.0083, "step": 19199 }, { "epoch": 1.99, "learning_rate": 2.98892363424763e-05, "loss": 0.162, "step": 19200 }, { "epoch": 1.99, "learning_rate": 2.988677493942478e-05, "loss": 0.3033, "step": 19201 }, { "epoch": 1.99, "learning_rate": 2.988431353713546e-05, "loss": 1.1255, "step": 19202 }, { "epoch": 1.99, "learning_rate": 2.9881852135624914e-05, "loss": 0.355, "step": 19203 }, { "epoch": 1.99, "learning_rate": 2.9879390734909712e-05, "loss": 0.2092, "step": 19204 }, { "epoch": 1.99, "learning_rate": 2.987692933500642e-05, "loss": 0.933, "step": 19205 }, { "epoch": 1.99, "learning_rate": 2.987446793593162e-05, "loss": 0.3254, "step": 19206 }, { "epoch": 1.99, "learning_rate": 2.9872006537701866e-05, "loss": 0.5623, "step": 19207 }, { "epoch": 1.99, "learning_rate": 2.9869545140333734e-05, "loss": 0.5501, "step": 19208 }, { "epoch": 1.99, "learning_rate": 2.9867083743843793e-05, "loss": 0.0766, "step": 19209 }, { "epoch": 1.99, "learning_rate": 2.986462234824862e-05, "loss": 0.2816, "step": 19210 }, { "epoch": 1.99, "learning_rate": 2.9862160953564773e-05, "loss": 0.6425, "step": 19211 }, { "epoch": 1.99, "learning_rate": 2.9859699559808837e-05, "loss": 0.2799, "step": 19212 }, { "epoch": 1.99, "learning_rate": 2.9857238166997357e-05, "loss": 0.8389, "step": 19213 }, { "epoch": 1.99, "learning_rate": 2.985477677514692e-05, "loss": 0.9855, "step": 19214 }, { "epoch": 1.99, "learning_rate": 2.9852315384274094e-05, "loss": 0.8692, "step": 19215 }, { "epoch": 1.99, "learning_rate": 2.984985399439544e-05, "loss": 0.692, "step": 19216 }, { "epoch": 1.99, "learning_rate": 2.9847392605527536e-05, "loss": 0.8737, "step": 19217 }, { "epoch": 1.99, "learning_rate": 2.984493121768695e-05, "loss": 0.9621, "step": 19218 }, { "epoch": 1.99, "learning_rate": 2.9842469830890247e-05, "loss": 0.4035, "step": 19219 }, { "epoch": 1.99, "learning_rate": 2.984000844515401e-05, "loss": 0.4984, "step": 19220 }, { "epoch": 1.99, "learning_rate": 2.983754706049479e-05, "loss": 0.9499, "step": 19221 }, { "epoch": 1.99, "learning_rate": 2.9835085676929166e-05, "loss": 0.5124, "step": 19222 }, { "epoch": 1.99, "learning_rate": 2.983262429447371e-05, "loss": 0.9836, "step": 19223 }, { "epoch": 1.99, "learning_rate": 2.983016291314499e-05, "loss": 0.2471, "step": 19224 }, { "epoch": 1.99, "learning_rate": 2.982770153295958e-05, "loss": 0.7202, "step": 19225 }, { "epoch": 1.99, "learning_rate": 2.982524015393403e-05, "loss": 0.8078, "step": 19226 }, { "epoch": 1.99, "learning_rate": 2.9822778776084927e-05, "loss": 0.5409, "step": 19227 }, { "epoch": 1.99, "learning_rate": 2.9820317399428835e-05, "loss": 0.4843, "step": 19228 }, { "epoch": 1.99, "learning_rate": 2.981785602398232e-05, "loss": 1.0454, "step": 19229 }, { "epoch": 1.99, "learning_rate": 2.9815394649761965e-05, "loss": 0.657, "step": 19230 }, { "epoch": 1.99, "learning_rate": 2.9812933276784323e-05, "loss": 0.9705, "step": 19231 }, { "epoch": 1.99, "learning_rate": 2.981047190506597e-05, "loss": 0.9336, "step": 19232 }, { "epoch": 1.99, "learning_rate": 2.9808010534623477e-05, "loss": 0.8852, "step": 19233 }, { "epoch": 1.99, "learning_rate": 2.9805549165473414e-05, "loss": 0.629, "step": 19234 }, { "epoch": 1.99, "learning_rate": 2.9803087797632346e-05, "loss": 0.2366, "step": 19235 }, { "epoch": 1.99, "learning_rate": 2.9800626431116844e-05, "loss": 1.0354, "step": 19236 }, { "epoch": 1.99, "learning_rate": 2.9798165065943493e-05, "loss": 0.393, "step": 19237 }, { "epoch": 1.99, "learning_rate": 2.979570370212883e-05, "loss": 0.6477, "step": 19238 }, { "epoch": 1.99, "learning_rate": 2.9793242339689454e-05, "loss": 0.5103, "step": 19239 }, { "epoch": 1.99, "learning_rate": 2.9790780978641913e-05, "loss": 0.2811, "step": 19240 }, { "epoch": 1.99, "learning_rate": 2.9788319619002787e-05, "loss": 0.3892, "step": 19241 }, { "epoch": 1.99, "learning_rate": 2.9785858260788645e-05, "loss": 1.0959, "step": 19242 }, { "epoch": 1.99, "learning_rate": 2.9783396904016048e-05, "loss": 0.9925, "step": 19243 }, { "epoch": 1.99, "learning_rate": 2.9780935548701584e-05, "loss": 0.6393, "step": 19244 }, { "epoch": 1.99, "learning_rate": 2.9778474194861805e-05, "loss": 1.0312, "step": 19245 }, { "epoch": 1.99, "learning_rate": 2.9776012842513285e-05, "loss": 0.3415, "step": 19246 }, { "epoch": 1.99, "learning_rate": 2.9773551491672592e-05, "loss": 1.3291, "step": 19247 }, { "epoch": 1.99, "learning_rate": 2.9771090142356304e-05, "loss": 1.2328, "step": 19248 }, { "epoch": 1.99, "learning_rate": 2.976862879458098e-05, "loss": 1.0149, "step": 19249 }, { "epoch": 1.99, "learning_rate": 2.9766167448363205e-05, "loss": 0.7831, "step": 19250 }, { "epoch": 2.0, "learning_rate": 2.976370610371952e-05, "loss": 0.5832, "step": 19251 }, { "epoch": 2.0, "learning_rate": 2.9761244760666516e-05, "loss": 1.0997, "step": 19252 }, { "epoch": 2.0, "learning_rate": 2.975878341922076e-05, "loss": 0.575, "step": 19253 }, { "epoch": 2.0, "learning_rate": 2.975632207939881e-05, "loss": 0.413, "step": 19254 }, { "epoch": 2.0, "learning_rate": 2.975386074121724e-05, "loss": 0.4248, "step": 19255 }, { "epoch": 2.0, "learning_rate": 2.975139940469263e-05, "loss": 0.5097, "step": 19256 }, { "epoch": 2.0, "learning_rate": 2.9748938069841532e-05, "loss": 1.4421, "step": 19257 }, { "epoch": 2.0, "learning_rate": 2.9746476736680538e-05, "loss": 0.5909, "step": 19258 }, { "epoch": 2.0, "learning_rate": 2.9744015405226193e-05, "loss": 0.5408, "step": 19259 }, { "epoch": 2.0, "learning_rate": 2.974155407549508e-05, "loss": 0.8278, "step": 19260 }, { "epoch": 2.0, "learning_rate": 2.973909274750376e-05, "loss": 0.3526, "step": 19261 }, { "epoch": 2.0, "learning_rate": 2.973663142126882e-05, "loss": 0.8859, "step": 19262 }, { "epoch": 2.0, "learning_rate": 2.973417009680681e-05, "loss": 0.1648, "step": 19263 }, { "epoch": 2.0, "learning_rate": 2.97317087741343e-05, "loss": 0.2313, "step": 19264 }, { "epoch": 2.0, "learning_rate": 2.9729247453267864e-05, "loss": 0.7886, "step": 19265 }, { "epoch": 2.0, "learning_rate": 2.972678613422407e-05, "loss": 0.9668, "step": 19266 }, { "epoch": 2.0, "learning_rate": 2.972432481701948e-05, "loss": 0.464, "step": 19267 }, { "epoch": 2.0, "learning_rate": 2.972186350167069e-05, "loss": 0.7881, "step": 19268 }, { "epoch": 2.0, "learning_rate": 2.9719402188194235e-05, "loss": 0.4731, "step": 19269 }, { "epoch": 2.0, "learning_rate": 2.9716940876606706e-05, "loss": 0.4033, "step": 19270 }, { "epoch": 2.0, "learning_rate": 2.9714479566924658e-05, "loss": 1.2863, "step": 19271 }, { "epoch": 2.0, "learning_rate": 2.9712018259164677e-05, "loss": 0.8716, "step": 19272 }, { "epoch": 2.0, "learning_rate": 2.970955695334331e-05, "loss": 0.144, "step": 19273 }, { "epoch": 2.0, "learning_rate": 2.970709564947714e-05, "loss": 0.6736, "step": 19274 }, { "epoch": 2.0, "learning_rate": 2.970463434758275e-05, "loss": 0.8352, "step": 19275 }, { "epoch": 2.0, "learning_rate": 2.9702173047676677e-05, "loss": 0.7007, "step": 19276 }, { "epoch": 2.0, "learning_rate": 2.9699711749775516e-05, "loss": 0.373, "step": 19277 }, { "epoch": 2.0, "learning_rate": 2.9697250453895813e-05, "loss": 0.9988, "step": 19278 }, { "epoch": 2.0, "learning_rate": 2.969478916005415e-05, "loss": 0.66, "step": 19279 }, { "epoch": 2.0, "learning_rate": 2.96923278682671e-05, "loss": 0.5713, "step": 19280 }, { "epoch": 2.0, "learning_rate": 2.968986657855122e-05, "loss": 1.1103, "step": 19281 }, { "epoch": 2.0, "learning_rate": 2.9687405290923095e-05, "loss": 0.5527, "step": 19282 }, { "epoch": 2.0, "learning_rate": 2.9684944005399277e-05, "loss": 0.376, "step": 19283 }, { "epoch": 2.0, "learning_rate": 2.9682482721996342e-05, "loss": 0.4002, "step": 19284 }, { "epoch": 2.0, "learning_rate": 2.968002144073086e-05, "loss": 0.6737, "step": 19285 }, { "epoch": 2.0, "learning_rate": 2.9677560161619405e-05, "loss": 0.8722, "step": 19286 }, { "epoch": 2.0, "learning_rate": 2.967509888467855e-05, "loss": 1.0109, "step": 19287 }, { "epoch": 2.0, "learning_rate": 2.9672637609924836e-05, "loss": 0.3754, "step": 19288 }, { "epoch": 2.0, "learning_rate": 2.967017633737485e-05, "loss": 0.604, "step": 19289 }, { "epoch": 2.0, "learning_rate": 2.9667715067045156e-05, "loss": 1.0559, "step": 19290 }, { "epoch": 2.0, "learning_rate": 2.9665253798952332e-05, "loss": 0.9672, "step": 19291 }, { "epoch": 2.0, "learning_rate": 2.9662792533112943e-05, "loss": 1.0057, "step": 19292 }, { "epoch": 2.0, "learning_rate": 2.9660331269543556e-05, "loss": 0.6265, "step": 19293 }, { "epoch": 2.0, "learning_rate": 2.9657870008260734e-05, "loss": 0.3427, "step": 19294 }, { "epoch": 2.0, "learning_rate": 2.965540874928105e-05, "loss": 0.4368, "step": 19295 }, { "epoch": 2.0, "learning_rate": 2.9652947492621085e-05, "loss": 0.8544, "step": 19296 }, { "epoch": 2.0, "learning_rate": 2.9650486238297383e-05, "loss": 0.1723, "step": 19297 }, { "epoch": 2.0, "learning_rate": 2.9648024986326533e-05, "loss": 0.3575, "step": 19298 }, { "epoch": 2.0, "learning_rate": 2.9645563736725095e-05, "loss": 0.9041, "step": 19299 }, { "epoch": 2.0, "learning_rate": 2.964310248950965e-05, "loss": 0.628, "step": 19300 }, { "epoch": 2.0, "learning_rate": 2.964064124469675e-05, "loss": 1.2003, "step": 19301 }, { "epoch": 2.0, "learning_rate": 2.963818000230296e-05, "loss": 1.3694, "step": 19302 }, { "epoch": 2.0, "learning_rate": 2.963571876234486e-05, "loss": 0.5187, "step": 19303 }, { "epoch": 2.0, "learning_rate": 2.9633257524839018e-05, "loss": 1.1161, "step": 19304 }, { "epoch": 2.0, "learning_rate": 2.9630796289802e-05, "loss": 0.7246, "step": 19305 }, { "epoch": 2.0, "learning_rate": 2.9628335057250375e-05, "loss": 0.5551, "step": 19306 }, { "epoch": 2.0, "learning_rate": 2.962587382720071e-05, "loss": 0.6288, "step": 19307 }, { "epoch": 2.0, "learning_rate": 2.9623412599669577e-05, "loss": 0.1614, "step": 19308 }, { "epoch": 2.0, "learning_rate": 2.9620951374673538e-05, "loss": 0.8873, "step": 19309 }, { "epoch": 2.0, "learning_rate": 2.9618490152229175e-05, "loss": 0.4384, "step": 19310 }, { "epoch": 2.0, "learning_rate": 2.961602893235304e-05, "loss": 0.2786, "step": 19311 }, { "epoch": 2.0, "learning_rate": 2.9613567715061708e-05, "loss": 0.716, "step": 19312 }, { "epoch": 2.0, "learning_rate": 2.961110650037176e-05, "loss": 0.2847, "step": 19313 }, { "epoch": 2.0, "learning_rate": 2.960864528829974e-05, "loss": 0.4022, "step": 19314 }, { "epoch": 2.0, "learning_rate": 2.9606184078862227e-05, "loss": 0.1912, "step": 19315 }, { "epoch": 2.0, "learning_rate": 2.9603722872075798e-05, "loss": 0.299, "step": 19316 }, { "epoch": 2.0, "learning_rate": 2.9601261667957006e-05, "loss": 1.2635, "step": 19317 }, { "epoch": 2.0, "learning_rate": 2.9598800466522432e-05, "loss": 1.41, "step": 19318 }, { "epoch": 2.0, "learning_rate": 2.9596339267788636e-05, "loss": 1.1097, "step": 19319 }, { "epoch": 2.0, "learning_rate": 2.9593878071772196e-05, "loss": 0.4083, "step": 19320 }, { "epoch": 2.0, "learning_rate": 2.9591416878489667e-05, "loss": 0.4475, "step": 19321 }, { "epoch": 2.0, "learning_rate": 2.9588955687957623e-05, "loss": 0.7941, "step": 19322 }, { "epoch": 2.0, "learning_rate": 2.9586494500192633e-05, "loss": 0.733, "step": 19323 }, { "epoch": 2.0, "learning_rate": 2.9584033315211277e-05, "loss": 0.5183, "step": 19324 }, { "epoch": 2.0, "learning_rate": 2.9581572133030114e-05, "loss": 1.0157, "step": 19325 }, { "epoch": 2.0, "learning_rate": 2.9579110953665695e-05, "loss": 0.4049, "step": 19326 }, { "epoch": 2.0, "learning_rate": 2.9576649777134604e-05, "loss": 0.4552, "step": 19327 }, { "epoch": 2.0, "learning_rate": 2.957418860345341e-05, "loss": 0.3913, "step": 19328 }, { "epoch": 2.0, "learning_rate": 2.9571727432638677e-05, "loss": 0.1063, "step": 19329 }, { "epoch": 2.0, "learning_rate": 2.9569266264706976e-05, "loss": 0.5491, "step": 19330 }, { "epoch": 2.0, "learning_rate": 2.9566805099674876e-05, "loss": 0.2659, "step": 19331 }, { "epoch": 2.0, "learning_rate": 2.9564343937558938e-05, "loss": 0.1188, "step": 19332 }, { "epoch": 2.0, "learning_rate": 2.9561882778375736e-05, "loss": 0.2454, "step": 19333 }, { "epoch": 2.0, "learning_rate": 2.9559421622141842e-05, "loss": 0.5855, "step": 19334 }, { "epoch": 2.0, "learning_rate": 2.9556960468873813e-05, "loss": 0.6887, "step": 19335 }, { "epoch": 2.0, "learning_rate": 2.9554499318588225e-05, "loss": 0.7246, "step": 19336 }, { "epoch": 2.0, "learning_rate": 2.9552038171301638e-05, "loss": 0.4391, "step": 19337 }, { "epoch": 2.0, "learning_rate": 2.954957702703064e-05, "loss": 0.7844, "step": 19338 }, { "epoch": 2.0, "learning_rate": 2.954711588579178e-05, "loss": 0.4155, "step": 19339 }, { "epoch": 2.0, "learning_rate": 2.954465474760162e-05, "loss": 0.5383, "step": 19340 }, { "epoch": 2.0, "learning_rate": 2.9542193612476738e-05, "loss": 0.5888, "step": 19341 }, { "epoch": 2.0, "learning_rate": 2.9539732480433705e-05, "loss": 0.1545, "step": 19342 }, { "epoch": 2.0, "learning_rate": 2.9537271351489084e-05, "loss": 0.6456, "step": 19343 }, { "epoch": 2.0, "learning_rate": 2.9534810225659448e-05, "loss": 0.2554, "step": 19344 }, { "epoch": 2.0, "learning_rate": 2.9532349102961353e-05, "loss": 1.2671, "step": 19345 }, { "epoch": 2.01, "learning_rate": 2.9529887983411382e-05, "loss": 0.0771, "step": 19346 }, { "epoch": 2.01, "learning_rate": 2.952742686702609e-05, "loss": 1.1102, "step": 19347 }, { "epoch": 2.01, "learning_rate": 2.9524965753822054e-05, "loss": 0.852, "step": 19348 }, { "epoch": 2.01, "learning_rate": 2.9522504643815837e-05, "loss": 1.3063, "step": 19349 }, { "epoch": 2.01, "learning_rate": 2.9520043537024014e-05, "loss": 0.2469, "step": 19350 }, { "epoch": 2.01, "learning_rate": 2.9517582433463135e-05, "loss": 1.2674, "step": 19351 }, { "epoch": 2.01, "learning_rate": 2.951512133314978e-05, "loss": 0.3944, "step": 19352 }, { "epoch": 2.01, "learning_rate": 2.951266023610051e-05, "loss": 0.6646, "step": 19353 }, { "epoch": 2.01, "learning_rate": 2.9510199142331908e-05, "loss": 0.6468, "step": 19354 }, { "epoch": 2.01, "learning_rate": 2.9507738051860523e-05, "loss": 0.6826, "step": 19355 }, { "epoch": 2.01, "learning_rate": 2.950527696470293e-05, "loss": 1.1993, "step": 19356 }, { "epoch": 2.01, "learning_rate": 2.9502815880875694e-05, "loss": 0.7257, "step": 19357 }, { "epoch": 2.01, "learning_rate": 2.9500354800395395e-05, "loss": 0.7289, "step": 19358 }, { "epoch": 2.01, "learning_rate": 2.9497893723278586e-05, "loss": 0.3029, "step": 19359 }, { "epoch": 2.01, "learning_rate": 2.9495432649541835e-05, "loss": 0.7242, "step": 19360 }, { "epoch": 2.01, "learning_rate": 2.9492971579201717e-05, "loss": 1.1866, "step": 19361 }, { "epoch": 2.01, "learning_rate": 2.9490510512274802e-05, "loss": 0.3919, "step": 19362 }, { "epoch": 2.01, "learning_rate": 2.9488049448777653e-05, "loss": 0.6979, "step": 19363 }, { "epoch": 2.01, "learning_rate": 2.9485588388726826e-05, "loss": 0.9429, "step": 19364 }, { "epoch": 2.01, "learning_rate": 2.9483127332138898e-05, "loss": 0.5368, "step": 19365 }, { "epoch": 2.01, "learning_rate": 2.948066627903043e-05, "loss": 0.5222, "step": 19366 }, { "epoch": 2.01, "learning_rate": 2.9478205229418002e-05, "loss": 0.6565, "step": 19367 }, { "epoch": 2.01, "learning_rate": 2.9475744183318177e-05, "loss": 0.9298, "step": 19368 }, { "epoch": 2.01, "learning_rate": 2.9473283140747516e-05, "loss": 0.7942, "step": 19369 }, { "epoch": 2.01, "learning_rate": 2.947082210172259e-05, "loss": 1.0618, "step": 19370 }, { "epoch": 2.01, "learning_rate": 2.9468361066259963e-05, "loss": 0.7736, "step": 19371 }, { "epoch": 2.01, "learning_rate": 2.946590003437621e-05, "loss": 1.6369, "step": 19372 }, { "epoch": 2.01, "learning_rate": 2.946343900608789e-05, "loss": 0.7864, "step": 19373 }, { "epoch": 2.01, "learning_rate": 2.9460977981411572e-05, "loss": 0.4008, "step": 19374 }, { "epoch": 2.01, "learning_rate": 2.9458516960363838e-05, "loss": 0.6898, "step": 19375 }, { "epoch": 2.01, "learning_rate": 2.945605594296123e-05, "loss": 1.3797, "step": 19376 }, { "epoch": 2.01, "learning_rate": 2.945359492922032e-05, "loss": 0.5613, "step": 19377 }, { "epoch": 2.01, "learning_rate": 2.945113391915769e-05, "loss": 0.5114, "step": 19378 }, { "epoch": 2.01, "learning_rate": 2.9448672912789892e-05, "loss": 1.4389, "step": 19379 }, { "epoch": 2.01, "learning_rate": 2.94462119101335e-05, "loss": 0.1765, "step": 19380 }, { "epoch": 2.01, "learning_rate": 2.944375091120508e-05, "loss": 0.4271, "step": 19381 }, { "epoch": 2.01, "learning_rate": 2.94412899160212e-05, "loss": 0.224, "step": 19382 }, { "epoch": 2.01, "learning_rate": 2.9438828924598427e-05, "loss": 0.3153, "step": 19383 }, { "epoch": 2.01, "learning_rate": 2.9436367936953323e-05, "loss": 0.6266, "step": 19384 }, { "epoch": 2.01, "learning_rate": 2.9433906953102458e-05, "loss": 1.0446, "step": 19385 }, { "epoch": 2.01, "learning_rate": 2.9431445973062404e-05, "loss": 0.4284, "step": 19386 }, { "epoch": 2.01, "learning_rate": 2.942898499684972e-05, "loss": 1.1709, "step": 19387 }, { "epoch": 2.01, "learning_rate": 2.9426524024480986e-05, "loss": 0.2203, "step": 19388 }, { "epoch": 2.01, "learning_rate": 2.9424063055972744e-05, "loss": 0.7746, "step": 19389 }, { "epoch": 2.01, "learning_rate": 2.9421602091341575e-05, "loss": 0.8132, "step": 19390 }, { "epoch": 2.01, "learning_rate": 2.9419141130604044e-05, "loss": 0.3726, "step": 19391 }, { "epoch": 2.01, "learning_rate": 2.9416680173776728e-05, "loss": 0.3574, "step": 19392 }, { "epoch": 2.01, "learning_rate": 2.941421922087618e-05, "loss": 0.5578, "step": 19393 }, { "epoch": 2.01, "learning_rate": 2.9411758271918966e-05, "loss": 0.8395, "step": 19394 }, { "epoch": 2.01, "learning_rate": 2.940929732692166e-05, "loss": 0.536, "step": 19395 }, { "epoch": 2.01, "learning_rate": 2.9406836385900838e-05, "loss": 0.8779, "step": 19396 }, { "epoch": 2.01, "learning_rate": 2.940437544887304e-05, "loss": 0.6738, "step": 19397 }, { "epoch": 2.01, "learning_rate": 2.940191451585485e-05, "loss": 0.5217, "step": 19398 }, { "epoch": 2.01, "learning_rate": 2.9399453586862834e-05, "loss": 0.6414, "step": 19399 }, { "epoch": 2.01, "learning_rate": 2.9396992661913566e-05, "loss": 0.5317, "step": 19400 }, { "epoch": 2.01, "learning_rate": 2.93945317410236e-05, "loss": 0.3186, "step": 19401 }, { "epoch": 2.01, "learning_rate": 2.9392070824209493e-05, "loss": 0.9803, "step": 19402 }, { "epoch": 2.01, "learning_rate": 2.9389609911487826e-05, "loss": 0.4243, "step": 19403 }, { "epoch": 2.01, "learning_rate": 2.938714900287516e-05, "loss": 0.4137, "step": 19404 }, { "epoch": 2.01, "learning_rate": 2.9384688098388065e-05, "loss": 0.2044, "step": 19405 }, { "epoch": 2.0, "learning_rate": 2.9382227198043112e-05, "loss": 0.1468, "step": 19406 }, { "epoch": 2.0, "learning_rate": 2.9379766301856857e-05, "loss": 0.4315, "step": 19407 }, { "epoch": 2.0, "learning_rate": 2.937730540984586e-05, "loss": 0.1321, "step": 19408 }, { "epoch": 2.0, "learning_rate": 2.937484452202671e-05, "loss": 0.8109, "step": 19409 }, { "epoch": 2.0, "learning_rate": 2.9372383638415963e-05, "loss": 0.1269, "step": 19410 }, { "epoch": 2.0, "learning_rate": 2.9369922759030173e-05, "loss": 0.2481, "step": 19411 }, { "epoch": 2.0, "learning_rate": 2.9367461883885914e-05, "loss": 0.1196, "step": 19412 }, { "epoch": 2.0, "learning_rate": 2.936500101299977e-05, "loss": 0.2918, "step": 19413 }, { "epoch": 2.0, "learning_rate": 2.936254014638828e-05, "loss": 0.7408, "step": 19414 }, { "epoch": 2.0, "learning_rate": 2.936007928406802e-05, "loss": 0.6739, "step": 19415 }, { "epoch": 2.0, "learning_rate": 2.935761842605556e-05, "loss": 0.3549, "step": 19416 }, { "epoch": 2.0, "learning_rate": 2.9355157572367456e-05, "loss": 0.6329, "step": 19417 }, { "epoch": 2.0, "learning_rate": 2.9352696723020283e-05, "loss": 0.7106, "step": 19418 }, { "epoch": 2.0, "learning_rate": 2.93502358780306e-05, "loss": 0.3253, "step": 19419 }, { "epoch": 2.0, "learning_rate": 2.9347775037414986e-05, "loss": 0.0827, "step": 19420 }, { "epoch": 2.0, "learning_rate": 2.9345314201189993e-05, "loss": 0.37, "step": 19421 }, { "epoch": 2.0, "learning_rate": 2.934285336937219e-05, "loss": 0.4101, "step": 19422 }, { "epoch": 2.0, "learning_rate": 2.9340392541978144e-05, "loss": 0.0672, "step": 19423 }, { "epoch": 2.0, "learning_rate": 2.9337931719024426e-05, "loss": 0.698, "step": 19424 }, { "epoch": 2.0, "learning_rate": 2.9335470900527596e-05, "loss": 0.1974, "step": 19425 }, { "epoch": 2.0, "learning_rate": 2.9333010086504228e-05, "loss": 0.6148, "step": 19426 }, { "epoch": 2.0, "learning_rate": 2.9330549276970867e-05, "loss": 0.0623, "step": 19427 }, { "epoch": 2.0, "learning_rate": 2.9328088471944097e-05, "loss": 0.1342, "step": 19428 }, { "epoch": 2.0, "learning_rate": 2.9325627671440473e-05, "loss": 0.4284, "step": 19429 }, { "epoch": 2.0, "learning_rate": 2.9323166875476572e-05, "loss": 0.067, "step": 19430 }, { "epoch": 2.0, "learning_rate": 2.932070608406895e-05, "loss": 0.3423, "step": 19431 }, { "epoch": 2.0, "learning_rate": 2.9318245297234174e-05, "loss": 0.542, "step": 19432 }, { "epoch": 2.0, "learning_rate": 2.9315784514988813e-05, "loss": 0.1237, "step": 19433 }, { "epoch": 2.0, "learning_rate": 2.9313323737349434e-05, "loss": 0.3281, "step": 19434 }, { "epoch": 2.0, "learning_rate": 2.9310862964332594e-05, "loss": 0.2307, "step": 19435 }, { "epoch": 2.0, "learning_rate": 2.930840219595486e-05, "loss": 0.1744, "step": 19436 }, { "epoch": 2.0, "learning_rate": 2.9305941432232804e-05, "loss": 0.6068, "step": 19437 }, { "epoch": 2.0, "learning_rate": 2.9303480673183003e-05, "loss": 0.1962, "step": 19438 }, { "epoch": 2.0, "learning_rate": 2.930101991882199e-05, "loss": 0.6216, "step": 19439 }, { "epoch": 2.0, "learning_rate": 2.929855916916636e-05, "loss": 0.1334, "step": 19440 }, { "epoch": 2.0, "learning_rate": 2.9296098424232654e-05, "loss": 0.0464, "step": 19441 }, { "epoch": 2.0, "learning_rate": 2.9293637684037447e-05, "loss": 0.0328, "step": 19442 }, { "epoch": 2.0, "learning_rate": 2.9291176948597306e-05, "loss": 0.0529, "step": 19443 }, { "epoch": 2.0, "learning_rate": 2.9288716217928808e-05, "loss": 0.3296, "step": 19444 }, { "epoch": 2.0, "learning_rate": 2.9286255492048495e-05, "loss": 0.1481, "step": 19445 }, { "epoch": 2.0, "learning_rate": 2.9283794770972942e-05, "loss": 0.2965, "step": 19446 }, { "epoch": 2.0, "learning_rate": 2.9281334054718718e-05, "loss": 0.4692, "step": 19447 }, { "epoch": 2.0, "learning_rate": 2.927887334330239e-05, "loss": 0.162, "step": 19448 }, { "epoch": 2.0, "learning_rate": 2.9276412636740514e-05, "loss": 0.2685, "step": 19449 }, { "epoch": 2.0, "learning_rate": 2.9273951935049656e-05, "loss": 0.8955, "step": 19450 }, { "epoch": 2.0, "learning_rate": 2.9271491238246398e-05, "loss": 0.3266, "step": 19451 }, { "epoch": 2.0, "learning_rate": 2.9269030546347275e-05, "loss": 0.1686, "step": 19452 }, { "epoch": 2.01, "learning_rate": 2.926656985936887e-05, "loss": 0.5375, "step": 19453 }, { "epoch": 2.01, "learning_rate": 2.9264109177327755e-05, "loss": 0.4885, "step": 19454 }, { "epoch": 2.01, "learning_rate": 2.926164850024047e-05, "loss": 0.5275, "step": 19455 }, { "epoch": 2.01, "learning_rate": 2.92591878281236e-05, "loss": 0.6682, "step": 19456 }, { "epoch": 2.01, "learning_rate": 2.92567271609937e-05, "loss": 0.3743, "step": 19457 }, { "epoch": 2.01, "learning_rate": 2.925426649886735e-05, "loss": 0.4385, "step": 19458 }, { "epoch": 2.01, "learning_rate": 2.9251805841761096e-05, "loss": 0.2195, "step": 19459 }, { "epoch": 2.01, "learning_rate": 2.9249345189691512e-05, "loss": 0.19, "step": 19460 }, { "epoch": 2.01, "learning_rate": 2.9246884542675157e-05, "loss": 0.1484, "step": 19461 }, { "epoch": 2.01, "learning_rate": 2.9244423900728606e-05, "loss": 0.671, "step": 19462 }, { "epoch": 2.01, "learning_rate": 2.9241963263868416e-05, "loss": 0.7056, "step": 19463 }, { "epoch": 2.01, "learning_rate": 2.9239502632111153e-05, "loss": 0.0499, "step": 19464 }, { "epoch": 2.01, "learning_rate": 2.9237042005473374e-05, "loss": 0.4231, "step": 19465 }, { "epoch": 2.01, "learning_rate": 2.923458138397165e-05, "loss": 0.29, "step": 19466 }, { "epoch": 2.01, "learning_rate": 2.9232120767622543e-05, "loss": 0.6156, "step": 19467 }, { "epoch": 2.01, "learning_rate": 2.9229660156442628e-05, "loss": 0.1586, "step": 19468 }, { "epoch": 2.01, "learning_rate": 2.9227199550448455e-05, "loss": 0.5868, "step": 19469 }, { "epoch": 2.01, "learning_rate": 2.922473894965659e-05, "loss": 0.8883, "step": 19470 }, { "epoch": 2.01, "learning_rate": 2.9222278354083604e-05, "loss": 0.0723, "step": 19471 }, { "epoch": 2.01, "learning_rate": 2.9219817763746062e-05, "loss": 0.3221, "step": 19472 }, { "epoch": 2.01, "learning_rate": 2.921735717866052e-05, "loss": 0.8368, "step": 19473 }, { "epoch": 2.01, "learning_rate": 2.921489659884355e-05, "loss": 0.7378, "step": 19474 }, { "epoch": 2.01, "learning_rate": 2.9212436024311702e-05, "loss": 0.101, "step": 19475 }, { "epoch": 2.01, "learning_rate": 2.9209975455081575e-05, "loss": 0.2317, "step": 19476 }, { "epoch": 2.01, "learning_rate": 2.920751489116969e-05, "loss": 0.3432, "step": 19477 }, { "epoch": 2.01, "learning_rate": 2.9205054332592633e-05, "loss": 0.5157, "step": 19478 }, { "epoch": 2.01, "learning_rate": 2.920259377936696e-05, "loss": 0.5004, "step": 19479 }, { "epoch": 2.01, "learning_rate": 2.920013323150924e-05, "loss": 0.4787, "step": 19480 }, { "epoch": 2.01, "learning_rate": 2.9197672689036037e-05, "loss": 0.3161, "step": 19481 }, { "epoch": 2.01, "learning_rate": 2.919521215196392e-05, "loss": 0.4998, "step": 19482 }, { "epoch": 2.01, "learning_rate": 2.919275162030944e-05, "loss": 0.421, "step": 19483 }, { "epoch": 2.01, "learning_rate": 2.9190291094089165e-05, "loss": 0.1524, "step": 19484 }, { "epoch": 2.01, "learning_rate": 2.9187830573319662e-05, "loss": 0.2098, "step": 19485 }, { "epoch": 2.01, "learning_rate": 2.91853700580175e-05, "loss": 0.5385, "step": 19486 }, { "epoch": 2.01, "learning_rate": 2.9182909548199228e-05, "loss": 0.243, "step": 19487 }, { "epoch": 2.01, "learning_rate": 2.9180449043881433e-05, "loss": 0.2993, "step": 19488 }, { "epoch": 2.01, "learning_rate": 2.917798854508065e-05, "loss": 0.2872, "step": 19489 }, { "epoch": 2.01, "learning_rate": 2.9175528051813458e-05, "loss": 0.1225, "step": 19490 }, { "epoch": 2.01, "learning_rate": 2.9173067564096414e-05, "loss": 0.2482, "step": 19491 }, { "epoch": 2.01, "learning_rate": 2.917060708194609e-05, "loss": 0.7719, "step": 19492 }, { "epoch": 2.01, "learning_rate": 2.9168146605379047e-05, "loss": 0.0707, "step": 19493 }, { "epoch": 2.01, "learning_rate": 2.916568613441184e-05, "loss": 0.6347, "step": 19494 }, { "epoch": 2.01, "learning_rate": 2.9163225669061043e-05, "loss": 0.2588, "step": 19495 }, { "epoch": 2.01, "learning_rate": 2.9160765209343215e-05, "loss": 0.1203, "step": 19496 }, { "epoch": 2.01, "learning_rate": 2.9158304755274918e-05, "loss": 0.2111, "step": 19497 }, { "epoch": 2.01, "learning_rate": 2.9155844306872715e-05, "loss": 0.1322, "step": 19498 }, { "epoch": 2.01, "learning_rate": 2.915338386415317e-05, "loss": 0.193, "step": 19499 }, { "epoch": 2.01, "learning_rate": 2.9150923427132856e-05, "loss": 0.4393, "step": 19500 }, { "epoch": 2.01, "learning_rate": 2.9148462995828328e-05, "loss": 0.1163, "step": 19501 }, { "epoch": 2.01, "learning_rate": 2.9146002570256144e-05, "loss": 0.4472, "step": 19502 }, { "epoch": 2.01, "learning_rate": 2.9143542150432866e-05, "loss": 0.2804, "step": 19503 }, { "epoch": 2.01, "learning_rate": 2.914108173637506e-05, "loss": 0.2849, "step": 19504 }, { "epoch": 2.01, "learning_rate": 2.9138621328099293e-05, "loss": 0.4163, "step": 19505 }, { "epoch": 2.01, "learning_rate": 2.913616092562213e-05, "loss": 0.5549, "step": 19506 }, { "epoch": 2.01, "learning_rate": 2.9133700528960124e-05, "loss": 0.5271, "step": 19507 }, { "epoch": 2.01, "learning_rate": 2.9131240138129843e-05, "loss": 0.6495, "step": 19508 }, { "epoch": 2.01, "learning_rate": 2.912877975314785e-05, "loss": 0.1537, "step": 19509 }, { "epoch": 2.01, "learning_rate": 2.9126319374030716e-05, "loss": 0.3385, "step": 19510 }, { "epoch": 2.01, "learning_rate": 2.912385900079499e-05, "loss": 0.6202, "step": 19511 }, { "epoch": 2.01, "learning_rate": 2.9121398633457235e-05, "loss": 0.2474, "step": 19512 }, { "epoch": 2.01, "learning_rate": 2.9118938272034024e-05, "loss": 0.0899, "step": 19513 }, { "epoch": 2.01, "learning_rate": 2.9116477916541923e-05, "loss": 0.1991, "step": 19514 }, { "epoch": 2.01, "learning_rate": 2.9114017566997474e-05, "loss": 0.3202, "step": 19515 }, { "epoch": 2.01, "learning_rate": 2.911155722341726e-05, "loss": 0.3347, "step": 19516 }, { "epoch": 2.01, "learning_rate": 2.910909688581783e-05, "loss": 0.2094, "step": 19517 }, { "epoch": 2.01, "learning_rate": 2.9106636554215746e-05, "loss": 0.5776, "step": 19518 }, { "epoch": 2.01, "learning_rate": 2.910417622862758e-05, "loss": 0.3326, "step": 19519 }, { "epoch": 2.01, "learning_rate": 2.9101715909069895e-05, "loss": 0.2148, "step": 19520 }, { "epoch": 2.01, "learning_rate": 2.9099255595559244e-05, "loss": 0.3473, "step": 19521 }, { "epoch": 2.01, "learning_rate": 2.9096795288112187e-05, "loss": 0.2979, "step": 19522 }, { "epoch": 2.01, "learning_rate": 2.90943349867453e-05, "loss": 0.2828, "step": 19523 }, { "epoch": 2.01, "learning_rate": 2.909187469147514e-05, "loss": 0.2933, "step": 19524 }, { "epoch": 2.01, "learning_rate": 2.9089414402318266e-05, "loss": 0.6008, "step": 19525 }, { "epoch": 2.01, "learning_rate": 2.908695411929125e-05, "loss": 0.654, "step": 19526 }, { "epoch": 2.01, "learning_rate": 2.9084493842410634e-05, "loss": 0.3993, "step": 19527 }, { "epoch": 2.01, "learning_rate": 2.9082033571692986e-05, "loss": 0.125, "step": 19528 }, { "epoch": 2.01, "learning_rate": 2.907957330715488e-05, "loss": 0.3653, "step": 19529 }, { "epoch": 2.01, "learning_rate": 2.9077113048812874e-05, "loss": 0.4482, "step": 19530 }, { "epoch": 2.01, "learning_rate": 2.9074652796683523e-05, "loss": 0.4077, "step": 19531 }, { "epoch": 2.01, "learning_rate": 2.907219255078339e-05, "loss": 0.4137, "step": 19532 }, { "epoch": 2.01, "learning_rate": 2.906973231112904e-05, "loss": 0.297, "step": 19533 }, { "epoch": 2.01, "learning_rate": 2.9067272077737044e-05, "loss": 0.393, "step": 19534 }, { "epoch": 2.01, "learning_rate": 2.9064811850623943e-05, "loss": 0.2946, "step": 19535 }, { "epoch": 2.01, "learning_rate": 2.9062351629806314e-05, "loss": 0.605, "step": 19536 }, { "epoch": 2.01, "learning_rate": 2.9059891415300713e-05, "loss": 0.3493, "step": 19537 }, { "epoch": 2.01, "learning_rate": 2.905743120712371e-05, "loss": 0.7779, "step": 19538 }, { "epoch": 2.01, "learning_rate": 2.9054971005291858e-05, "loss": 0.1532, "step": 19539 }, { "epoch": 2.01, "learning_rate": 2.905251080982172e-05, "loss": 0.5799, "step": 19540 }, { "epoch": 2.01, "learning_rate": 2.9050050620729852e-05, "loss": 0.485, "step": 19541 }, { "epoch": 2.01, "learning_rate": 2.904759043803282e-05, "loss": 0.3644, "step": 19542 }, { "epoch": 2.01, "learning_rate": 2.9045130261747187e-05, "loss": 0.5697, "step": 19543 }, { "epoch": 2.01, "learning_rate": 2.9042670091889522e-05, "loss": 0.1847, "step": 19544 }, { "epoch": 2.01, "learning_rate": 2.904020992847637e-05, "loss": 0.3287, "step": 19545 }, { "epoch": 2.01, "learning_rate": 2.90377497715243e-05, "loss": 0.544, "step": 19546 }, { "epoch": 2.01, "learning_rate": 2.9035289621049872e-05, "loss": 0.3111, "step": 19547 }, { "epoch": 2.01, "learning_rate": 2.9032829477069658e-05, "loss": 0.4164, "step": 19548 }, { "epoch": 2.02, "learning_rate": 2.9030369339600202e-05, "loss": 0.2759, "step": 19549 }, { "epoch": 2.02, "learning_rate": 2.902790920865807e-05, "loss": 0.4718, "step": 19550 }, { "epoch": 2.02, "learning_rate": 2.902544908425984e-05, "loss": 0.4948, "step": 19551 }, { "epoch": 2.02, "learning_rate": 2.9022988966422052e-05, "loss": 0.1391, "step": 19552 }, { "epoch": 2.02, "learning_rate": 2.9020528855161266e-05, "loss": 0.1589, "step": 19553 }, { "epoch": 2.02, "learning_rate": 2.901806875049406e-05, "loss": 0.3533, "step": 19554 }, { "epoch": 2.02, "learning_rate": 2.9015608652436976e-05, "loss": 0.4033, "step": 19555 }, { "epoch": 2.02, "learning_rate": 2.901314856100659e-05, "loss": 0.1439, "step": 19556 }, { "epoch": 2.02, "learning_rate": 2.901068847621945e-05, "loss": 0.6189, "step": 19557 }, { "epoch": 2.02, "learning_rate": 2.9008228398092137e-05, "loss": 0.6786, "step": 19558 }, { "epoch": 2.02, "learning_rate": 2.9005768326641185e-05, "loss": 0.4316, "step": 19559 }, { "epoch": 2.02, "learning_rate": 2.9003308261883172e-05, "loss": 0.3768, "step": 19560 }, { "epoch": 2.02, "learning_rate": 2.9000848203834657e-05, "loss": 0.2238, "step": 19561 }, { "epoch": 2.02, "learning_rate": 2.8998388152512203e-05, "loss": 0.1634, "step": 19562 }, { "epoch": 2.02, "learning_rate": 2.8995928107932355e-05, "loss": 0.2086, "step": 19563 }, { "epoch": 2.02, "learning_rate": 2.89934680701117e-05, "loss": 0.1814, "step": 19564 }, { "epoch": 2.02, "learning_rate": 2.899100803906677e-05, "loss": 0.3707, "step": 19565 }, { "epoch": 2.02, "learning_rate": 2.898854801481414e-05, "loss": 0.0925, "step": 19566 }, { "epoch": 2.02, "learning_rate": 2.8986087997370363e-05, "loss": 0.2043, "step": 19567 }, { "epoch": 2.02, "learning_rate": 2.8983627986752013e-05, "loss": 0.2752, "step": 19568 }, { "epoch": 2.02, "learning_rate": 2.8981167982975636e-05, "loss": 0.3119, "step": 19569 }, { "epoch": 2.02, "learning_rate": 2.8978707986057796e-05, "loss": 0.5314, "step": 19570 }, { "epoch": 2.02, "learning_rate": 2.8976247996015056e-05, "loss": 0.3734, "step": 19571 }, { "epoch": 2.02, "learning_rate": 2.897378801286398e-05, "loss": 0.3006, "step": 19572 }, { "epoch": 2.02, "learning_rate": 2.8971328036621122e-05, "loss": 0.1957, "step": 19573 }, { "epoch": 2.02, "learning_rate": 2.8968868067303038e-05, "loss": 0.5027, "step": 19574 }, { "epoch": 2.02, "learning_rate": 2.8966408104926294e-05, "loss": 0.1523, "step": 19575 }, { "epoch": 2.02, "learning_rate": 2.896394814950746e-05, "loss": 0.2097, "step": 19576 }, { "epoch": 2.02, "learning_rate": 2.8961488201063075e-05, "loss": 0.1791, "step": 19577 }, { "epoch": 2.02, "learning_rate": 2.895902825960971e-05, "loss": 1.1552, "step": 19578 }, { "epoch": 2.02, "learning_rate": 2.8956568325163923e-05, "loss": 0.1965, "step": 19579 }, { "epoch": 2.02, "learning_rate": 2.895410839774227e-05, "loss": 0.1039, "step": 19580 }, { "epoch": 2.02, "learning_rate": 2.8951648477361315e-05, "loss": 0.0851, "step": 19581 }, { "epoch": 2.02, "learning_rate": 2.8949188564037625e-05, "loss": 0.24, "step": 19582 }, { "epoch": 2.02, "learning_rate": 2.8946728657787743e-05, "loss": 0.367, "step": 19583 }, { "epoch": 2.02, "learning_rate": 2.894426875862824e-05, "loss": 0.2967, "step": 19584 }, { "epoch": 2.02, "learning_rate": 2.894180886657567e-05, "loss": 0.1563, "step": 19585 }, { "epoch": 2.02, "learning_rate": 2.8939348981646602e-05, "loss": 0.4561, "step": 19586 }, { "epoch": 2.02, "learning_rate": 2.8936889103857583e-05, "loss": 0.256, "step": 19587 }, { "epoch": 2.02, "learning_rate": 2.8934429233225177e-05, "loss": 0.1677, "step": 19588 }, { "epoch": 2.02, "learning_rate": 2.893196936976596e-05, "loss": 0.1603, "step": 19589 }, { "epoch": 2.02, "learning_rate": 2.8929509513496458e-05, "loss": 0.3899, "step": 19590 }, { "epoch": 2.02, "learning_rate": 2.8927049664433247e-05, "loss": 0.26, "step": 19591 }, { "epoch": 2.02, "learning_rate": 2.89245898225929e-05, "loss": 0.2928, "step": 19592 }, { "epoch": 2.02, "learning_rate": 2.892212998799195e-05, "loss": 0.7776, "step": 19593 }, { "epoch": 2.02, "learning_rate": 2.891967016064697e-05, "loss": 0.666, "step": 19594 }, { "epoch": 2.02, "learning_rate": 2.891721034057452e-05, "loss": 0.2692, "step": 19595 }, { "epoch": 2.02, "learning_rate": 2.891475052779116e-05, "loss": 0.2448, "step": 19596 }, { "epoch": 2.02, "learning_rate": 2.891229072231344e-05, "loss": 0.7762, "step": 19597 }, { "epoch": 2.02, "learning_rate": 2.8909830924157924e-05, "loss": 0.76, "step": 19598 }, { "epoch": 2.02, "learning_rate": 2.890737113334117e-05, "loss": 0.4108, "step": 19599 }, { "epoch": 2.02, "learning_rate": 2.8904911349879747e-05, "loss": 0.4985, "step": 19600 }, { "epoch": 2.02, "learning_rate": 2.8902451573790204e-05, "loss": 0.4866, "step": 19601 }, { "epoch": 2.02, "learning_rate": 2.88999918050891e-05, "loss": 0.1251, "step": 19602 }, { "epoch": 2.02, "learning_rate": 2.889753204379299e-05, "loss": 0.1925, "step": 19603 }, { "epoch": 2.02, "learning_rate": 2.889507228991843e-05, "loss": 0.4547, "step": 19604 }, { "epoch": 2.02, "learning_rate": 2.889261254348199e-05, "loss": 0.3971, "step": 19605 }, { "epoch": 2.02, "learning_rate": 2.8890152804500227e-05, "loss": 0.3112, "step": 19606 }, { "epoch": 2.02, "learning_rate": 2.888769307298969e-05, "loss": 0.3852, "step": 19607 }, { "epoch": 2.02, "learning_rate": 2.8885233348966942e-05, "loss": 1.2599, "step": 19608 }, { "epoch": 2.02, "learning_rate": 2.8882773632448545e-05, "loss": 0.0879, "step": 19609 }, { "epoch": 2.02, "learning_rate": 2.8880313923451058e-05, "loss": 0.0875, "step": 19610 }, { "epoch": 2.02, "learning_rate": 2.887785422199103e-05, "loss": 0.9353, "step": 19611 }, { "epoch": 2.02, "learning_rate": 2.8875394528085027e-05, "loss": 0.3966, "step": 19612 }, { "epoch": 2.02, "learning_rate": 2.8872934841749602e-05, "loss": 0.3742, "step": 19613 }, { "epoch": 2.02, "learning_rate": 2.8870475163001332e-05, "loss": 0.5989, "step": 19614 }, { "epoch": 2.02, "learning_rate": 2.886801549185674e-05, "loss": 0.1721, "step": 19615 }, { "epoch": 2.02, "learning_rate": 2.8865555828332415e-05, "loss": 0.068, "step": 19616 }, { "epoch": 2.02, "learning_rate": 2.8863096172444897e-05, "loss": 0.7691, "step": 19617 }, { "epoch": 2.02, "learning_rate": 2.8860636524210745e-05, "loss": 0.3793, "step": 19618 }, { "epoch": 2.02, "learning_rate": 2.8858176883646528e-05, "loss": 0.5558, "step": 19619 }, { "epoch": 2.02, "learning_rate": 2.88557172507688e-05, "loss": 0.221, "step": 19620 }, { "epoch": 2.02, "learning_rate": 2.8853257625594106e-05, "loss": 0.5279, "step": 19621 }, { "epoch": 2.02, "learning_rate": 2.8850798008139014e-05, "loss": 0.5963, "step": 19622 }, { "epoch": 2.02, "learning_rate": 2.8848338398420084e-05, "loss": 0.1711, "step": 19623 }, { "epoch": 2.02, "learning_rate": 2.8845878796453877e-05, "loss": 0.3452, "step": 19624 }, { "epoch": 2.02, "learning_rate": 2.8843419202256934e-05, "loss": 0.2896, "step": 19625 }, { "epoch": 2.02, "learning_rate": 2.8840959615845823e-05, "loss": 0.4306, "step": 19626 }, { "epoch": 2.02, "learning_rate": 2.8838500037237115e-05, "loss": 0.3872, "step": 19627 }, { "epoch": 2.02, "learning_rate": 2.883604046644734e-05, "loss": 0.1899, "step": 19628 }, { "epoch": 2.02, "learning_rate": 2.883358090349307e-05, "loss": 0.215, "step": 19629 }, { "epoch": 2.02, "learning_rate": 2.8831121348390865e-05, "loss": 0.3784, "step": 19630 }, { "epoch": 2.02, "learning_rate": 2.882866180115727e-05, "loss": 0.2795, "step": 19631 }, { "epoch": 2.02, "learning_rate": 2.8826202261808852e-05, "loss": 0.7686, "step": 19632 }, { "epoch": 2.02, "learning_rate": 2.8823742730362163e-05, "loss": 0.275, "step": 19633 }, { "epoch": 2.02, "learning_rate": 2.8821283206833772e-05, "loss": 0.3929, "step": 19634 }, { "epoch": 2.02, "learning_rate": 2.881882369124022e-05, "loss": 0.6775, "step": 19635 }, { "epoch": 2.02, "learning_rate": 2.881636418359807e-05, "loss": 0.2465, "step": 19636 }, { "epoch": 2.02, "learning_rate": 2.8813904683923875e-05, "loss": 0.3793, "step": 19637 }, { "epoch": 2.02, "learning_rate": 2.881144519223421e-05, "loss": 0.1377, "step": 19638 }, { "epoch": 2.02, "learning_rate": 2.8808985708545612e-05, "loss": 0.4023, "step": 19639 }, { "epoch": 2.02, "learning_rate": 2.8806526232874647e-05, "loss": 0.3169, "step": 19640 }, { "epoch": 2.02, "learning_rate": 2.880406676523786e-05, "loss": 0.2987, "step": 19641 }, { "epoch": 2.02, "learning_rate": 2.8801607305651817e-05, "loss": 0.6315, "step": 19642 }, { "epoch": 2.02, "learning_rate": 2.8799147854133074e-05, "loss": 0.1205, "step": 19643 }, { "epoch": 2.02, "learning_rate": 2.879668841069819e-05, "loss": 0.2368, "step": 19644 }, { "epoch": 2.03, "learning_rate": 2.8794228975363714e-05, "loss": 0.1604, "step": 19645 }, { "epoch": 2.03, "learning_rate": 2.8791769548146204e-05, "loss": 0.2418, "step": 19646 }, { "epoch": 2.03, "learning_rate": 2.8789310129062223e-05, "loss": 0.4883, "step": 19647 }, { "epoch": 2.03, "learning_rate": 2.8786850718128323e-05, "loss": 0.8652, "step": 19648 }, { "epoch": 2.03, "learning_rate": 2.878439131536106e-05, "loss": 0.2148, "step": 19649 }, { "epoch": 2.03, "learning_rate": 2.8781931920776985e-05, "loss": 0.341, "step": 19650 }, { "epoch": 2.03, "learning_rate": 2.8779472534392664e-05, "loss": 0.5895, "step": 19651 }, { "epoch": 2.03, "learning_rate": 2.8777013156224658e-05, "loss": 0.6769, "step": 19652 }, { "epoch": 2.03, "learning_rate": 2.8774553786289503e-05, "loss": 0.1365, "step": 19653 }, { "epoch": 2.03, "learning_rate": 2.8772094424603768e-05, "loss": 0.7064, "step": 19654 }, { "epoch": 2.03, "learning_rate": 2.8769635071184004e-05, "loss": 0.196, "step": 19655 }, { "epoch": 2.03, "learning_rate": 2.876717572604677e-05, "loss": 0.304, "step": 19656 }, { "epoch": 2.03, "learning_rate": 2.8764716389208613e-05, "loss": 0.7226, "step": 19657 }, { "epoch": 2.03, "learning_rate": 2.876225706068611e-05, "loss": 0.5189, "step": 19658 }, { "epoch": 2.03, "learning_rate": 2.875979774049579e-05, "loss": 0.5215, "step": 19659 }, { "epoch": 2.03, "learning_rate": 2.8757338428654224e-05, "loss": 0.2134, "step": 19660 }, { "epoch": 2.03, "learning_rate": 2.8754879125177965e-05, "loss": 0.2269, "step": 19661 }, { "epoch": 2.03, "learning_rate": 2.8752419830083575e-05, "loss": 0.6992, "step": 19662 }, { "epoch": 2.03, "learning_rate": 2.87499605433876e-05, "loss": 0.3252, "step": 19663 }, { "epoch": 2.03, "learning_rate": 2.874750126510661e-05, "loss": 0.2939, "step": 19664 }, { "epoch": 2.03, "learning_rate": 2.8745041995257132e-05, "loss": 0.3521, "step": 19665 }, { "epoch": 2.03, "learning_rate": 2.8742582733855734e-05, "loss": 0.1588, "step": 19666 }, { "epoch": 2.03, "learning_rate": 2.8740123480918984e-05, "loss": 0.1815, "step": 19667 }, { "epoch": 2.03, "learning_rate": 2.873766423646343e-05, "loss": 0.1059, "step": 19668 }, { "epoch": 2.03, "learning_rate": 2.8735205000505617e-05, "loss": 0.7176, "step": 19669 }, { "epoch": 2.03, "learning_rate": 2.873274577306211e-05, "loss": 0.713, "step": 19670 }, { "epoch": 2.03, "learning_rate": 2.8730286554149463e-05, "loss": 0.7192, "step": 19671 }, { "epoch": 2.03, "learning_rate": 2.8727827343784232e-05, "loss": 0.6592, "step": 19672 }, { "epoch": 2.03, "learning_rate": 2.8725368141982966e-05, "loss": 0.2868, "step": 19673 }, { "epoch": 2.03, "learning_rate": 2.8722908948762225e-05, "loss": 0.4298, "step": 19674 }, { "epoch": 2.03, "learning_rate": 2.872044976413856e-05, "loss": 0.6285, "step": 19675 }, { "epoch": 2.03, "learning_rate": 2.8717990588128536e-05, "loss": 0.1426, "step": 19676 }, { "epoch": 2.03, "learning_rate": 2.87155314207487e-05, "loss": 0.2343, "step": 19677 }, { "epoch": 2.03, "learning_rate": 2.8713072262015603e-05, "loss": 0.0907, "step": 19678 }, { "epoch": 2.03, "learning_rate": 2.8710613111945797e-05, "loss": 0.1298, "step": 19679 }, { "epoch": 2.03, "learning_rate": 2.8708153970555842e-05, "loss": 0.0868, "step": 19680 }, { "epoch": 2.03, "learning_rate": 2.870569483786229e-05, "loss": 0.3186, "step": 19681 }, { "epoch": 2.03, "learning_rate": 2.870323571388171e-05, "loss": 0.1703, "step": 19682 }, { "epoch": 2.03, "learning_rate": 2.8700776598630634e-05, "loss": 0.3998, "step": 19683 }, { "epoch": 2.03, "learning_rate": 2.8698317492125623e-05, "loss": 0.3007, "step": 19684 }, { "epoch": 2.03, "learning_rate": 2.869585839438324e-05, "loss": 0.0885, "step": 19685 }, { "epoch": 2.03, "learning_rate": 2.8693399305420033e-05, "loss": 0.1469, "step": 19686 }, { "epoch": 2.03, "learning_rate": 2.8690940225252552e-05, "loss": 0.7001, "step": 19687 }, { "epoch": 2.03, "learning_rate": 2.8688481153897358e-05, "loss": 0.4039, "step": 19688 }, { "epoch": 2.03, "learning_rate": 2.8686022091371012e-05, "loss": 0.1486, "step": 19689 }, { "epoch": 2.03, "learning_rate": 2.868356303769005e-05, "loss": 0.5156, "step": 19690 }, { "epoch": 2.03, "learning_rate": 2.868110399287103e-05, "loss": 0.1077, "step": 19691 }, { "epoch": 2.03, "learning_rate": 2.8678644956930512e-05, "loss": 0.1471, "step": 19692 }, { "epoch": 2.03, "learning_rate": 2.8676185929885042e-05, "loss": 0.4855, "step": 19693 }, { "epoch": 2.03, "learning_rate": 2.867372691175118e-05, "loss": 0.3549, "step": 19694 }, { "epoch": 2.03, "learning_rate": 2.8671267902545478e-05, "loss": 0.7059, "step": 19695 }, { "epoch": 2.03, "learning_rate": 2.8668808902284497e-05, "loss": 0.6824, "step": 19696 }, { "epoch": 2.03, "learning_rate": 2.866634991098478e-05, "loss": 0.5079, "step": 19697 }, { "epoch": 2.03, "learning_rate": 2.8663890928662876e-05, "loss": 0.5755, "step": 19698 }, { "epoch": 2.03, "learning_rate": 2.866143195533535e-05, "loss": 0.6449, "step": 19699 }, { "epoch": 2.03, "learning_rate": 2.8658972991018758e-05, "loss": 0.3957, "step": 19700 }, { "epoch": 2.03, "learning_rate": 2.8656514035729633e-05, "loss": 0.1413, "step": 19701 }, { "epoch": 2.03, "learning_rate": 2.8654055089484563e-05, "loss": 0.0097, "step": 19702 }, { "epoch": 2.03, "learning_rate": 2.8651596152300062e-05, "loss": 0.5133, "step": 19703 }, { "epoch": 2.03, "learning_rate": 2.86491372241927e-05, "loss": 0.3694, "step": 19704 }, { "epoch": 2.03, "learning_rate": 2.864667830517903e-05, "loss": 0.3353, "step": 19705 }, { "epoch": 2.03, "learning_rate": 2.864421939527561e-05, "loss": 0.1531, "step": 19706 }, { "epoch": 2.03, "learning_rate": 2.8641760494498985e-05, "loss": 0.0925, "step": 19707 }, { "epoch": 2.03, "learning_rate": 2.863930160286571e-05, "loss": 0.3717, "step": 19708 }, { "epoch": 2.03, "learning_rate": 2.8636842720392338e-05, "loss": 0.2914, "step": 19709 }, { "epoch": 2.03, "learning_rate": 2.8634383847095426e-05, "loss": 0.8442, "step": 19710 }, { "epoch": 2.03, "learning_rate": 2.8631924982991515e-05, "loss": 0.5193, "step": 19711 }, { "epoch": 2.03, "learning_rate": 2.862946612809717e-05, "loss": 0.3654, "step": 19712 }, { "epoch": 2.03, "learning_rate": 2.8627007282428935e-05, "loss": 0.3795, "step": 19713 }, { "epoch": 2.03, "learning_rate": 2.8624548446003374e-05, "loss": 0.144, "step": 19714 }, { "epoch": 2.03, "learning_rate": 2.862208961883703e-05, "loss": 0.5319, "step": 19715 }, { "epoch": 2.03, "learning_rate": 2.8619630800946455e-05, "loss": 0.2658, "step": 19716 }, { "epoch": 2.03, "learning_rate": 2.8617171992348197e-05, "loss": 0.2746, "step": 19717 }, { "epoch": 2.03, "learning_rate": 2.8614713193058813e-05, "loss": 0.1677, "step": 19718 }, { "epoch": 2.03, "learning_rate": 2.8612254403094856e-05, "loss": 0.4747, "step": 19719 }, { "epoch": 2.03, "learning_rate": 2.8609795622472886e-05, "loss": 0.6876, "step": 19720 }, { "epoch": 2.03, "learning_rate": 2.860733685120944e-05, "loss": 0.3639, "step": 19721 }, { "epoch": 2.03, "learning_rate": 2.8604878089321075e-05, "loss": 0.1756, "step": 19722 }, { "epoch": 2.03, "learning_rate": 2.8602419336824343e-05, "loss": 0.3, "step": 19723 }, { "epoch": 2.03, "learning_rate": 2.8599960593735806e-05, "loss": 0.7772, "step": 19724 }, { "epoch": 2.03, "learning_rate": 2.8597501860071998e-05, "loss": 0.1487, "step": 19725 }, { "epoch": 2.03, "learning_rate": 2.8595043135849483e-05, "loss": 0.6055, "step": 19726 }, { "epoch": 2.03, "learning_rate": 2.859258442108482e-05, "loss": 0.4465, "step": 19727 }, { "epoch": 2.03, "learning_rate": 2.8590125715794534e-05, "loss": 0.4236, "step": 19728 }, { "epoch": 2.03, "learning_rate": 2.8587667019995192e-05, "loss": 0.5666, "step": 19729 }, { "epoch": 2.03, "learning_rate": 2.8585208333703355e-05, "loss": 0.2535, "step": 19730 }, { "epoch": 2.03, "learning_rate": 2.8582749656935556e-05, "loss": 0.1348, "step": 19731 }, { "epoch": 2.03, "learning_rate": 2.8580290989708353e-05, "loss": 0.3882, "step": 19732 }, { "epoch": 2.03, "learning_rate": 2.85778323320383e-05, "loss": 0.5355, "step": 19733 }, { "epoch": 2.03, "learning_rate": 2.8575373683941954e-05, "loss": 0.1905, "step": 19734 }, { "epoch": 2.03, "learning_rate": 2.8572915045435853e-05, "loss": 0.438, "step": 19735 }, { "epoch": 2.03, "learning_rate": 2.8570456416536556e-05, "loss": 0.6703, "step": 19736 }, { "epoch": 2.03, "learning_rate": 2.8567997797260607e-05, "loss": 0.7611, "step": 19737 }, { "epoch": 2.03, "learning_rate": 2.8565539187624574e-05, "loss": 0.2326, "step": 19738 }, { "epoch": 2.03, "learning_rate": 2.8563080587644984e-05, "loss": 0.1276, "step": 19739 }, { "epoch": 2.03, "learning_rate": 2.8560621997338415e-05, "loss": 0.2532, "step": 19740 }, { "epoch": 2.04, "learning_rate": 2.8558163416721388e-05, "loss": 0.1783, "step": 19741 }, { "epoch": 2.04, "learning_rate": 2.8555704845810468e-05, "loss": 0.5999, "step": 19742 }, { "epoch": 2.04, "learning_rate": 2.8553246284622202e-05, "loss": 0.2725, "step": 19743 }, { "epoch": 2.04, "learning_rate": 2.8550787733173154e-05, "loss": 0.3644, "step": 19744 }, { "epoch": 2.04, "learning_rate": 2.8548329191479856e-05, "loss": 0.0801, "step": 19745 }, { "epoch": 2.04, "learning_rate": 2.8545870659558866e-05, "loss": 0.0849, "step": 19746 }, { "epoch": 2.04, "learning_rate": 2.8543412137426733e-05, "loss": 0.1378, "step": 19747 }, { "epoch": 2.04, "learning_rate": 2.854095362510002e-05, "loss": 0.6279, "step": 19748 }, { "epoch": 2.04, "learning_rate": 2.8538495122595254e-05, "loss": 0.1875, "step": 19749 }, { "epoch": 2.04, "learning_rate": 2.8536036629929002e-05, "loss": 0.3781, "step": 19750 }, { "epoch": 2.04, "learning_rate": 2.8533578147117806e-05, "loss": 0.2596, "step": 19751 }, { "epoch": 2.04, "learning_rate": 2.853111967417823e-05, "loss": 0.2717, "step": 19752 }, { "epoch": 2.04, "learning_rate": 2.8528661211126804e-05, "loss": 0.286, "step": 19753 }, { "epoch": 2.04, "learning_rate": 2.8526202757980088e-05, "loss": 0.2074, "step": 19754 }, { "epoch": 2.04, "learning_rate": 2.8523744314754627e-05, "loss": 0.8677, "step": 19755 }, { "epoch": 2.04, "learning_rate": 2.8521285881466974e-05, "loss": 0.4345, "step": 19756 }, { "epoch": 2.04, "learning_rate": 2.8518827458133676e-05, "loss": 0.5353, "step": 19757 }, { "epoch": 2.04, "learning_rate": 2.8516369044771294e-05, "loss": 0.2566, "step": 19758 }, { "epoch": 2.04, "learning_rate": 2.8513910641396362e-05, "loss": 0.3306, "step": 19759 }, { "epoch": 2.04, "learning_rate": 2.8511452248025435e-05, "loss": 0.1979, "step": 19760 }, { "epoch": 2.04, "learning_rate": 2.850899386467506e-05, "loss": 0.7973, "step": 19761 }, { "epoch": 2.04, "learning_rate": 2.8506535491361798e-05, "loss": 0.3157, "step": 19762 }, { "epoch": 2.04, "learning_rate": 2.8504077128102184e-05, "loss": 0.1898, "step": 19763 }, { "epoch": 2.04, "learning_rate": 2.8501618774912775e-05, "loss": 0.4158, "step": 19764 }, { "epoch": 2.04, "learning_rate": 2.849916043181013e-05, "loss": 0.1842, "step": 19765 }, { "epoch": 2.04, "learning_rate": 2.849670209881077e-05, "loss": 0.1499, "step": 19766 }, { "epoch": 2.04, "learning_rate": 2.849424377593126e-05, "loss": 0.374, "step": 19767 }, { "epoch": 2.04, "learning_rate": 2.849178546318816e-05, "loss": 0.2419, "step": 19768 }, { "epoch": 2.04, "learning_rate": 2.8489327160597995e-05, "loss": 0.4997, "step": 19769 }, { "epoch": 2.04, "learning_rate": 2.8486868868177328e-05, "loss": 0.2804, "step": 19770 }, { "epoch": 2.04, "learning_rate": 2.8484410585942704e-05, "loss": 0.2441, "step": 19771 }, { "epoch": 2.04, "learning_rate": 2.848195231391068e-05, "loss": 0.3655, "step": 19772 }, { "epoch": 2.04, "learning_rate": 2.8479494052097796e-05, "loss": 0.3861, "step": 19773 }, { "epoch": 2.04, "learning_rate": 2.8477035800520597e-05, "loss": 0.5912, "step": 19774 }, { "epoch": 2.04, "learning_rate": 2.8474577559195638e-05, "loss": 0.1933, "step": 19775 }, { "epoch": 2.04, "learning_rate": 2.8472119328139476e-05, "loss": 0.2601, "step": 19776 }, { "epoch": 2.04, "learning_rate": 2.8469661107368646e-05, "loss": 0.148, "step": 19777 }, { "epoch": 2.04, "learning_rate": 2.84672028968997e-05, "loss": 0.1221, "step": 19778 }, { "epoch": 2.04, "learning_rate": 2.846474469674918e-05, "loss": 0.3114, "step": 19779 }, { "epoch": 2.04, "learning_rate": 2.846228650693364e-05, "loss": 0.1845, "step": 19780 }, { "epoch": 2.04, "learning_rate": 2.8459828327469624e-05, "loss": 0.1603, "step": 19781 }, { "epoch": 2.04, "learning_rate": 2.845737015837369e-05, "loss": 0.2695, "step": 19782 }, { "epoch": 2.04, "learning_rate": 2.8454911999662373e-05, "loss": 0.4539, "step": 19783 }, { "epoch": 2.04, "learning_rate": 2.845245385135223e-05, "loss": 0.2976, "step": 19784 }, { "epoch": 2.04, "learning_rate": 2.8449995713459802e-05, "loss": 0.5952, "step": 19785 }, { "epoch": 2.04, "learning_rate": 2.8447537586001648e-05, "loss": 0.3893, "step": 19786 }, { "epoch": 2.04, "learning_rate": 2.8445079468994305e-05, "loss": 0.447, "step": 19787 }, { "epoch": 2.04, "learning_rate": 2.8442621362454318e-05, "loss": 0.5267, "step": 19788 }, { "epoch": 2.04, "learning_rate": 2.844016326639824e-05, "loss": 0.2137, "step": 19789 }, { "epoch": 2.04, "learning_rate": 2.8437705180842633e-05, "loss": 0.3261, "step": 19790 }, { "epoch": 2.04, "learning_rate": 2.8435247105804014e-05, "loss": 0.3414, "step": 19791 }, { "epoch": 2.04, "learning_rate": 2.8432789041298955e-05, "loss": 0.2274, "step": 19792 }, { "epoch": 2.04, "learning_rate": 2.8430330987343985e-05, "loss": 0.2377, "step": 19793 }, { "epoch": 2.04, "learning_rate": 2.842787294395566e-05, "loss": 0.3428, "step": 19794 }, { "epoch": 2.04, "learning_rate": 2.8425414911150528e-05, "loss": 0.2073, "step": 19795 }, { "epoch": 2.04, "learning_rate": 2.842295688894514e-05, "loss": 0.2661, "step": 19796 }, { "epoch": 2.04, "learning_rate": 2.8420498877356027e-05, "loss": 0.4503, "step": 19797 }, { "epoch": 2.04, "learning_rate": 2.841804087639975e-05, "loss": 0.1312, "step": 19798 }, { "epoch": 2.04, "learning_rate": 2.8415582886092852e-05, "loss": 0.421, "step": 19799 }, { "epoch": 2.04, "learning_rate": 2.8413124906451885e-05, "loss": 0.5561, "step": 19800 }, { "epoch": 2.04, "learning_rate": 2.8410666937493382e-05, "loss": 0.3519, "step": 19801 }, { "epoch": 2.04, "learning_rate": 2.840820897923391e-05, "loss": 0.5388, "step": 19802 }, { "epoch": 2.04, "learning_rate": 2.840575103168999e-05, "loss": 0.3549, "step": 19803 }, { "epoch": 2.04, "learning_rate": 2.840329309487818e-05, "loss": 0.5802, "step": 19804 }, { "epoch": 2.04, "learning_rate": 2.8400835168815032e-05, "loss": 0.5276, "step": 19805 }, { "epoch": 2.04, "learning_rate": 2.839837725351709e-05, "loss": 0.2606, "step": 19806 }, { "epoch": 2.04, "learning_rate": 2.839591934900089e-05, "loss": 0.3579, "step": 19807 }, { "epoch": 2.04, "learning_rate": 2.839346145528299e-05, "loss": 0.7273, "step": 19808 }, { "epoch": 2.04, "learning_rate": 2.839100357237993e-05, "loss": 0.1882, "step": 19809 }, { "epoch": 2.04, "learning_rate": 2.838854570030826e-05, "loss": 0.2581, "step": 19810 }, { "epoch": 2.04, "learning_rate": 2.8386087839084522e-05, "loss": 0.3517, "step": 19811 }, { "epoch": 2.04, "learning_rate": 2.8383629988725262e-05, "loss": 0.9142, "step": 19812 }, { "epoch": 2.04, "learning_rate": 2.8381172149247028e-05, "loss": 0.1634, "step": 19813 }, { "epoch": 2.04, "learning_rate": 2.8378714320666367e-05, "loss": 0.5975, "step": 19814 }, { "epoch": 2.04, "learning_rate": 2.8376256502999824e-05, "loss": 0.4598, "step": 19815 }, { "epoch": 2.04, "learning_rate": 2.837379869626394e-05, "loss": 0.3406, "step": 19816 }, { "epoch": 2.04, "learning_rate": 2.8371340900475258e-05, "loss": 0.2441, "step": 19817 }, { "epoch": 2.04, "learning_rate": 2.8368883115650327e-05, "loss": 0.2089, "step": 19818 }, { "epoch": 2.04, "learning_rate": 2.8366425341805694e-05, "loss": 0.6001, "step": 19819 }, { "epoch": 2.04, "learning_rate": 2.836396757895791e-05, "loss": 0.5095, "step": 19820 }, { "epoch": 2.04, "learning_rate": 2.836150982712351e-05, "loss": 0.2118, "step": 19821 }, { "epoch": 2.04, "learning_rate": 2.8359052086319037e-05, "loss": 0.2478, "step": 19822 }, { "epoch": 2.04, "learning_rate": 2.8356594356561042e-05, "loss": 0.6307, "step": 19823 }, { "epoch": 2.04, "learning_rate": 2.835413663786608e-05, "loss": 0.5804, "step": 19824 }, { "epoch": 2.04, "learning_rate": 2.8351678930250676e-05, "loss": 0.7453, "step": 19825 }, { "epoch": 2.04, "learning_rate": 2.8349221233731383e-05, "loss": 0.0964, "step": 19826 }, { "epoch": 2.04, "learning_rate": 2.834676354832475e-05, "loss": 0.7039, "step": 19827 }, { "epoch": 2.04, "learning_rate": 2.8344305874047328e-05, "loss": 0.4205, "step": 19828 }, { "epoch": 2.04, "learning_rate": 2.8341848210915637e-05, "loss": 0.1931, "step": 19829 }, { "epoch": 2.04, "learning_rate": 2.8339390558946245e-05, "loss": 0.2202, "step": 19830 }, { "epoch": 2.04, "learning_rate": 2.833693291815568e-05, "loss": 0.1449, "step": 19831 }, { "epoch": 2.04, "learning_rate": 2.8334475288560492e-05, "loss": 0.5078, "step": 19832 }, { "epoch": 2.04, "learning_rate": 2.833201767017723e-05, "loss": 0.2831, "step": 19833 }, { "epoch": 2.04, "learning_rate": 2.8329560063022437e-05, "loss": 0.9137, "step": 19834 }, { "epoch": 2.04, "learning_rate": 2.832710246711265e-05, "loss": 0.1003, "step": 19835 }, { "epoch": 2.05, "learning_rate": 2.832464488246442e-05, "loss": 0.3944, "step": 19836 }, { "epoch": 2.05, "learning_rate": 2.832218730909429e-05, "loss": 0.0922, "step": 19837 }, { "epoch": 2.05, "learning_rate": 2.8319729747018802e-05, "loss": 0.259, "step": 19838 }, { "epoch": 2.05, "learning_rate": 2.83172721962545e-05, "loss": 0.7035, "step": 19839 }, { "epoch": 2.05, "learning_rate": 2.8314814656817937e-05, "loss": 0.4465, "step": 19840 }, { "epoch": 2.05, "learning_rate": 2.8312357128725636e-05, "loss": 0.3077, "step": 19841 }, { "epoch": 2.05, "learning_rate": 2.830989961199415e-05, "loss": 0.3338, "step": 19842 }, { "epoch": 2.05, "learning_rate": 2.830744210664003e-05, "loss": 0.491, "step": 19843 }, { "epoch": 2.05, "learning_rate": 2.8304984612679812e-05, "loss": 0.4502, "step": 19844 }, { "epoch": 2.05, "learning_rate": 2.830252713013004e-05, "loss": 0.7643, "step": 19845 }, { "epoch": 2.05, "learning_rate": 2.8300069659007258e-05, "loss": 0.6706, "step": 19846 }, { "epoch": 2.05, "learning_rate": 2.8297612199328006e-05, "loss": 0.7711, "step": 19847 }, { "epoch": 2.05, "learning_rate": 2.829515475110884e-05, "loss": 0.1287, "step": 19848 }, { "epoch": 2.05, "learning_rate": 2.829269731436629e-05, "loss": 0.1374, "step": 19849 }, { "epoch": 2.05, "learning_rate": 2.8290239889116894e-05, "loss": 0.2168, "step": 19850 }, { "epoch": 2.05, "learning_rate": 2.828778247537721e-05, "loss": 0.4139, "step": 19851 }, { "epoch": 2.05, "learning_rate": 2.8285325073163775e-05, "loss": 0.1997, "step": 19852 }, { "epoch": 2.05, "learning_rate": 2.828286768249313e-05, "loss": 0.7364, "step": 19853 }, { "epoch": 2.05, "learning_rate": 2.8280410303381823e-05, "loss": 0.428, "step": 19854 }, { "epoch": 2.05, "learning_rate": 2.8277952935846383e-05, "loss": 0.8162, "step": 19855 }, { "epoch": 2.05, "learning_rate": 2.827549557990336e-05, "loss": 0.1331, "step": 19856 }, { "epoch": 2.05, "learning_rate": 2.82730382355693e-05, "loss": 0.7276, "step": 19857 }, { "epoch": 2.05, "learning_rate": 2.827058090286075e-05, "loss": 0.3249, "step": 19858 }, { "epoch": 2.05, "learning_rate": 2.8268123581794233e-05, "loss": 0.2429, "step": 19859 }, { "epoch": 2.05, "learning_rate": 2.8265666272386304e-05, "loss": 0.1265, "step": 19860 }, { "epoch": 2.05, "learning_rate": 2.8263208974653503e-05, "loss": 0.5743, "step": 19861 }, { "epoch": 2.05, "learning_rate": 2.8260751688612383e-05, "loss": 0.1971, "step": 19862 }, { "epoch": 2.05, "learning_rate": 2.8258294414279468e-05, "loss": 0.1206, "step": 19863 }, { "epoch": 2.05, "learning_rate": 2.8255837151671308e-05, "loss": 0.1563, "step": 19864 }, { "epoch": 2.05, "learning_rate": 2.8253379900804454e-05, "loss": 0.526, "step": 19865 }, { "epoch": 2.05, "learning_rate": 2.8250922661695428e-05, "loss": 0.3622, "step": 19866 }, { "epoch": 2.05, "learning_rate": 2.824846543436078e-05, "loss": 0.1195, "step": 19867 }, { "epoch": 2.05, "learning_rate": 2.8246008218817063e-05, "loss": 0.1576, "step": 19868 }, { "epoch": 2.05, "learning_rate": 2.8243551015080793e-05, "loss": 0.2542, "step": 19869 }, { "epoch": 2.05, "learning_rate": 2.8241093823168537e-05, "loss": 0.4328, "step": 19870 }, { "epoch": 2.05, "learning_rate": 2.823863664309682e-05, "loss": 0.5508, "step": 19871 }, { "epoch": 2.05, "learning_rate": 2.8236179474882193e-05, "loss": 0.6986, "step": 19872 }, { "epoch": 2.05, "learning_rate": 2.823372231854119e-05, "loss": 0.1948, "step": 19873 }, { "epoch": 2.05, "learning_rate": 2.8231265174090356e-05, "loss": 0.2379, "step": 19874 }, { "epoch": 2.05, "learning_rate": 2.822880804154623e-05, "loss": 0.2526, "step": 19875 }, { "epoch": 2.05, "learning_rate": 2.8226350920925362e-05, "loss": 0.265, "step": 19876 }, { "epoch": 2.05, "learning_rate": 2.8223893812244274e-05, "loss": 0.19, "step": 19877 }, { "epoch": 2.05, "learning_rate": 2.8221436715519534e-05, "loss": 0.657, "step": 19878 }, { "epoch": 2.05, "learning_rate": 2.821897963076765e-05, "loss": 0.2051, "step": 19879 }, { "epoch": 2.05, "learning_rate": 2.821652255800518e-05, "loss": 0.4141, "step": 19880 }, { "epoch": 2.05, "learning_rate": 2.821406549724866e-05, "loss": 0.2144, "step": 19881 }, { "epoch": 2.05, "learning_rate": 2.8211608448514647e-05, "loss": 0.2403, "step": 19882 }, { "epoch": 2.05, "learning_rate": 2.8209151411819657e-05, "loss": 0.1721, "step": 19883 }, { "epoch": 2.05, "learning_rate": 2.820669438718024e-05, "loss": 0.5689, "step": 19884 }, { "epoch": 2.05, "learning_rate": 2.8204237374612933e-05, "loss": 0.3442, "step": 19885 }, { "epoch": 2.05, "learning_rate": 2.820178037413429e-05, "loss": 0.1502, "step": 19886 }, { "epoch": 2.05, "learning_rate": 2.819932338576084e-05, "loss": 0.4703, "step": 19887 }, { "epoch": 2.05, "learning_rate": 2.8196866409509118e-05, "loss": 0.619, "step": 19888 }, { "epoch": 2.05, "learning_rate": 2.8194409445395667e-05, "loss": 0.7415, "step": 19889 }, { "epoch": 2.05, "learning_rate": 2.8191952493437047e-05, "loss": 0.5757, "step": 19890 }, { "epoch": 2.05, "learning_rate": 2.8189495553649762e-05, "loss": 0.5512, "step": 19891 }, { "epoch": 2.05, "learning_rate": 2.818703862605038e-05, "loss": 0.3943, "step": 19892 }, { "epoch": 2.05, "learning_rate": 2.818458171065542e-05, "loss": 0.4337, "step": 19893 }, { "epoch": 2.05, "learning_rate": 2.8182124807481433e-05, "loss": 0.3421, "step": 19894 }, { "epoch": 2.05, "learning_rate": 2.8179667916544957e-05, "loss": 0.3994, "step": 19895 }, { "epoch": 2.05, "learning_rate": 2.8177211037862537e-05, "loss": 0.6915, "step": 19896 }, { "epoch": 2.05, "learning_rate": 2.8174754171450698e-05, "loss": 0.3304, "step": 19897 }, { "epoch": 2.05, "learning_rate": 2.817229731732599e-05, "loss": 0.2288, "step": 19898 }, { "epoch": 2.05, "learning_rate": 2.8169840475504942e-05, "loss": 0.3949, "step": 19899 }, { "epoch": 2.05, "learning_rate": 2.816738364600411e-05, "loss": 0.3072, "step": 19900 }, { "epoch": 2.05, "learning_rate": 2.8164926828840018e-05, "loss": 0.6933, "step": 19901 }, { "epoch": 2.05, "learning_rate": 2.8162470024029208e-05, "loss": 0.2481, "step": 19902 }, { "epoch": 2.05, "learning_rate": 2.816001323158823e-05, "loss": 0.2204, "step": 19903 }, { "epoch": 2.05, "learning_rate": 2.8157556451533605e-05, "loss": 0.5889, "step": 19904 }, { "epoch": 2.05, "learning_rate": 2.8155099683881877e-05, "loss": 0.7535, "step": 19905 }, { "epoch": 2.05, "learning_rate": 2.815264292864959e-05, "loss": 0.2315, "step": 19906 }, { "epoch": 2.05, "learning_rate": 2.8150186185853274e-05, "loss": 0.7185, "step": 19907 }, { "epoch": 2.05, "learning_rate": 2.814772945550947e-05, "loss": 0.4959, "step": 19908 }, { "epoch": 2.05, "learning_rate": 2.8145272737634722e-05, "loss": 0.655, "step": 19909 }, { "epoch": 2.05, "learning_rate": 2.8142816032245567e-05, "loss": 0.1425, "step": 19910 }, { "epoch": 2.05, "learning_rate": 2.8140359339358533e-05, "loss": 0.3807, "step": 19911 }, { "epoch": 2.05, "learning_rate": 2.8137902658990165e-05, "loss": 0.4797, "step": 19912 }, { "epoch": 2.05, "learning_rate": 2.8135445991157e-05, "loss": 0.2948, "step": 19913 }, { "epoch": 2.05, "learning_rate": 2.8132989335875587e-05, "loss": 0.0806, "step": 19914 }, { "epoch": 2.05, "learning_rate": 2.8130532693162445e-05, "loss": 0.8387, "step": 19915 }, { "epoch": 2.05, "learning_rate": 2.8128076063034132e-05, "loss": 0.6224, "step": 19916 }, { "epoch": 2.05, "learning_rate": 2.812561944550716e-05, "loss": 0.2151, "step": 19917 }, { "epoch": 2.05, "learning_rate": 2.8123162840598077e-05, "loss": 0.404, "step": 19918 }, { "epoch": 2.05, "learning_rate": 2.8120706248323426e-05, "loss": 0.1343, "step": 19919 }, { "epoch": 2.05, "learning_rate": 2.8118249668699744e-05, "loss": 0.3373, "step": 19920 }, { "epoch": 2.05, "learning_rate": 2.811579310174356e-05, "loss": 0.3113, "step": 19921 }, { "epoch": 2.05, "learning_rate": 2.8113336547471417e-05, "loss": 0.1154, "step": 19922 }, { "epoch": 2.05, "learning_rate": 2.811088000589985e-05, "loss": 0.1521, "step": 19923 }, { "epoch": 2.05, "learning_rate": 2.81084234770454e-05, "loss": 0.1098, "step": 19924 }, { "epoch": 2.05, "learning_rate": 2.81059669609246e-05, "loss": 0.4635, "step": 19925 }, { "epoch": 2.05, "learning_rate": 2.810351045755398e-05, "loss": 0.149, "step": 19926 }, { "epoch": 2.05, "learning_rate": 2.810105396695009e-05, "loss": 0.6588, "step": 19927 }, { "epoch": 2.05, "learning_rate": 2.8098597489129473e-05, "loss": 0.3325, "step": 19928 }, { "epoch": 2.05, "learning_rate": 2.8096141024108637e-05, "loss": 0.3329, "step": 19929 }, { "epoch": 2.05, "learning_rate": 2.8093684571904145e-05, "loss": 0.5695, "step": 19930 }, { "epoch": 2.05, "learning_rate": 2.8091228132532514e-05, "loss": 0.3428, "step": 19931 }, { "epoch": 2.06, "learning_rate": 2.8088771706010285e-05, "loss": 0.2819, "step": 19932 }, { "epoch": 2.06, "learning_rate": 2.8086315292354005e-05, "loss": 0.6696, "step": 19933 }, { "epoch": 2.06, "learning_rate": 2.8083858891580205e-05, "loss": 0.4146, "step": 19934 }, { "epoch": 2.06, "learning_rate": 2.808140250370541e-05, "loss": 0.2676, "step": 19935 }, { "epoch": 2.06, "learning_rate": 2.8078946128746164e-05, "loss": 0.3107, "step": 19936 }, { "epoch": 2.06, "learning_rate": 2.8076489766719006e-05, "loss": 0.9573, "step": 19937 }, { "epoch": 2.06, "learning_rate": 2.8074033417640473e-05, "loss": 0.1682, "step": 19938 }, { "epoch": 2.06, "learning_rate": 2.8071577081527092e-05, "loss": 0.7005, "step": 19939 }, { "epoch": 2.06, "learning_rate": 2.8069120758395405e-05, "loss": 0.3285, "step": 19940 }, { "epoch": 2.06, "learning_rate": 2.8066664448261953e-05, "loss": 0.4201, "step": 19941 }, { "epoch": 2.06, "learning_rate": 2.806420815114325e-05, "loss": 0.4268, "step": 19942 }, { "epoch": 2.06, "learning_rate": 2.8061751867055853e-05, "loss": 0.6829, "step": 19943 }, { "epoch": 2.06, "learning_rate": 2.805929559601629e-05, "loss": 0.2944, "step": 19944 }, { "epoch": 2.06, "learning_rate": 2.8056839338041087e-05, "loss": 0.2091, "step": 19945 }, { "epoch": 2.06, "learning_rate": 2.805438309314679e-05, "loss": 0.1558, "step": 19946 }, { "epoch": 2.06, "learning_rate": 2.8051926861349928e-05, "loss": 0.1014, "step": 19947 }, { "epoch": 2.06, "learning_rate": 2.804947064266705e-05, "loss": 0.2533, "step": 19948 }, { "epoch": 2.06, "learning_rate": 2.8047014437114665e-05, "loss": 0.2224, "step": 19949 }, { "epoch": 2.06, "learning_rate": 2.8044558244709327e-05, "loss": 0.1815, "step": 19950 }, { "epoch": 2.06, "learning_rate": 2.804210206546757e-05, "loss": 0.1676, "step": 19951 }, { "epoch": 2.06, "learning_rate": 2.803964589940592e-05, "loss": 0.651, "step": 19952 }, { "epoch": 2.06, "learning_rate": 2.803718974654092e-05, "loss": 0.0882, "step": 19953 }, { "epoch": 2.06, "learning_rate": 2.80347336068891e-05, "loss": 0.2851, "step": 19954 }, { "epoch": 2.06, "learning_rate": 2.8032277480466988e-05, "loss": 0.2743, "step": 19955 }, { "epoch": 2.06, "learning_rate": 2.802982136729112e-05, "loss": 0.5213, "step": 19956 }, { "epoch": 2.06, "learning_rate": 2.8027365267378035e-05, "loss": 0.2048, "step": 19957 }, { "epoch": 2.06, "learning_rate": 2.8024909180744272e-05, "loss": 0.29, "step": 19958 }, { "epoch": 2.06, "learning_rate": 2.8022453107406352e-05, "loss": 0.411, "step": 19959 }, { "epoch": 2.06, "learning_rate": 2.8019997047380815e-05, "loss": 0.4539, "step": 19960 }, { "epoch": 2.06, "learning_rate": 2.8017541000684195e-05, "loss": 0.5705, "step": 19961 }, { "epoch": 2.06, "learning_rate": 2.8015084967333034e-05, "loss": 0.7873, "step": 19962 }, { "epoch": 2.06, "learning_rate": 2.8012628947343847e-05, "loss": 0.5537, "step": 19963 }, { "epoch": 2.06, "learning_rate": 2.8010172940733177e-05, "loss": 0.1069, "step": 19964 }, { "epoch": 2.06, "learning_rate": 2.8007716947517557e-05, "loss": 0.3631, "step": 19965 }, { "epoch": 2.06, "learning_rate": 2.8005260967713534e-05, "loss": 0.2559, "step": 19966 }, { "epoch": 2.06, "learning_rate": 2.8002805001337613e-05, "loss": 0.5664, "step": 19967 }, { "epoch": 2.06, "learning_rate": 2.800034904840635e-05, "loss": 0.3828, "step": 19968 }, { "epoch": 2.06, "learning_rate": 2.7997893108936263e-05, "loss": 0.5315, "step": 19969 }, { "epoch": 2.06, "learning_rate": 2.799543718294389e-05, "loss": 0.219, "step": 19970 }, { "epoch": 2.06, "learning_rate": 2.7992981270445764e-05, "loss": 0.3762, "step": 19971 }, { "epoch": 2.06, "learning_rate": 2.7990525371458426e-05, "loss": 0.5352, "step": 19972 }, { "epoch": 2.06, "learning_rate": 2.7988069485998397e-05, "loss": 0.4029, "step": 19973 }, { "epoch": 2.06, "learning_rate": 2.798561361408221e-05, "loss": 0.3058, "step": 19974 }, { "epoch": 2.06, "learning_rate": 2.7983157755726403e-05, "loss": 0.2978, "step": 19975 }, { "epoch": 2.06, "learning_rate": 2.7980701910947513e-05, "loss": 0.4155, "step": 19976 }, { "epoch": 2.06, "learning_rate": 2.7978246079762055e-05, "loss": 0.2646, "step": 19977 }, { "epoch": 2.06, "learning_rate": 2.7975790262186587e-05, "loss": 0.4923, "step": 19978 }, { "epoch": 2.06, "learning_rate": 2.7973334458237613e-05, "loss": 0.1889, "step": 19979 }, { "epoch": 2.06, "learning_rate": 2.7970878667931676e-05, "loss": 0.2227, "step": 19980 }, { "epoch": 2.06, "learning_rate": 2.7968422891285308e-05, "loss": 0.3288, "step": 19981 }, { "epoch": 2.06, "learning_rate": 2.7965967128315052e-05, "loss": 0.0857, "step": 19982 }, { "epoch": 2.06, "learning_rate": 2.7963511379037416e-05, "loss": 0.8246, "step": 19983 }, { "epoch": 2.06, "learning_rate": 2.796105564346895e-05, "loss": 0.4587, "step": 19984 }, { "epoch": 2.06, "learning_rate": 2.7958599921626183e-05, "loss": 0.5492, "step": 19985 }, { "epoch": 2.06, "learning_rate": 2.7956144213525644e-05, "loss": 0.2937, "step": 19986 }, { "epoch": 2.06, "learning_rate": 2.795368851918386e-05, "loss": 0.3836, "step": 19987 }, { "epoch": 2.06, "learning_rate": 2.7951232838617365e-05, "loss": 0.2453, "step": 19988 }, { "epoch": 2.06, "learning_rate": 2.7948777171842693e-05, "loss": 1.6943, "step": 19989 }, { "epoch": 2.06, "learning_rate": 2.794632151887638e-05, "loss": 0.3841, "step": 19990 }, { "epoch": 2.06, "learning_rate": 2.794386587973495e-05, "loss": 0.3284, "step": 19991 }, { "epoch": 2.06, "learning_rate": 2.7941410254434932e-05, "loss": 0.4796, "step": 19992 }, { "epoch": 2.06, "learning_rate": 2.793895464299285e-05, "loss": 0.3471, "step": 19993 }, { "epoch": 2.06, "learning_rate": 2.793649904542525e-05, "loss": 0.5342, "step": 19994 }, { "epoch": 2.06, "learning_rate": 2.793404346174865e-05, "loss": 0.2019, "step": 19995 }, { "epoch": 2.06, "learning_rate": 2.793158789197959e-05, "loss": 0.4832, "step": 19996 }, { "epoch": 2.06, "learning_rate": 2.79291323361346e-05, "loss": 0.6791, "step": 19997 }, { "epoch": 2.06, "learning_rate": 2.7926676794230197e-05, "loss": 0.5871, "step": 19998 }, { "epoch": 2.06, "learning_rate": 2.7924221266282927e-05, "loss": 0.5935, "step": 19999 }, { "epoch": 2.06, "learning_rate": 2.7921765752309316e-05, "loss": 0.1799, "step": 20000 }, { "epoch": 2.06, "learning_rate": 2.7919310252325887e-05, "loss": 0.5145, "step": 20001 }, { "epoch": 2.06, "learning_rate": 2.7916854766349174e-05, "loss": 0.1717, "step": 20002 }, { "epoch": 2.06, "learning_rate": 2.7914399294395722e-05, "loss": 0.4172, "step": 20003 }, { "epoch": 2.06, "learning_rate": 2.7911943836482028e-05, "loss": 0.1435, "step": 20004 }, { "epoch": 2.06, "learning_rate": 2.7909488392624645e-05, "loss": 0.3477, "step": 20005 }, { "epoch": 2.06, "learning_rate": 2.79070329628401e-05, "loss": 0.383, "step": 20006 }, { "epoch": 2.06, "learning_rate": 2.7904577547144914e-05, "loss": 0.7469, "step": 20007 }, { "epoch": 2.06, "learning_rate": 2.790212214555562e-05, "loss": 0.428, "step": 20008 }, { "epoch": 2.06, "learning_rate": 2.7899666758088752e-05, "loss": 0.1933, "step": 20009 }, { "epoch": 2.06, "learning_rate": 2.789721138476084e-05, "loss": 0.2715, "step": 20010 }, { "epoch": 2.06, "learning_rate": 2.7894756025588407e-05, "loss": 0.519, "step": 20011 }, { "epoch": 2.06, "learning_rate": 2.7892300680587982e-05, "loss": 0.5353, "step": 20012 }, { "epoch": 2.06, "learning_rate": 2.7889845349776094e-05, "loss": 0.5805, "step": 20013 }, { "epoch": 2.06, "learning_rate": 2.7887390033169278e-05, "loss": 0.294, "step": 20014 }, { "epoch": 2.06, "learning_rate": 2.7884934730784056e-05, "loss": 0.6986, "step": 20015 }, { "epoch": 2.06, "learning_rate": 2.7882479442636967e-05, "loss": 0.6149, "step": 20016 }, { "epoch": 2.06, "learning_rate": 2.788002416874452e-05, "loss": 0.1789, "step": 20017 }, { "epoch": 2.06, "learning_rate": 2.7877568909123255e-05, "loss": 0.3612, "step": 20018 }, { "epoch": 2.06, "learning_rate": 2.7875113663789698e-05, "loss": 0.2208, "step": 20019 }, { "epoch": 2.06, "learning_rate": 2.7872658432760385e-05, "loss": 0.1714, "step": 20020 }, { "epoch": 2.06, "learning_rate": 2.787020321605183e-05, "loss": 0.1361, "step": 20021 }, { "epoch": 2.06, "learning_rate": 2.786774801368057e-05, "loss": 0.5229, "step": 20022 }, { "epoch": 2.06, "learning_rate": 2.7865292825663137e-05, "loss": 0.4999, "step": 20023 }, { "epoch": 2.06, "learning_rate": 2.7862837652016054e-05, "loss": 0.2541, "step": 20024 }, { "epoch": 2.06, "learning_rate": 2.7860382492755843e-05, "loss": 0.3887, "step": 20025 }, { "epoch": 2.06, "learning_rate": 2.7857927347899034e-05, "loss": 0.126, "step": 20026 }, { "epoch": 2.06, "learning_rate": 2.785547221746216e-05, "loss": 1.1474, "step": 20027 }, { "epoch": 2.07, "learning_rate": 2.785301710146175e-05, "loss": 0.5876, "step": 20028 }, { "epoch": 2.07, "learning_rate": 2.7850561999914328e-05, "loss": 0.2512, "step": 20029 }, { "epoch": 2.07, "learning_rate": 2.7848106912836414e-05, "loss": 0.6265, "step": 20030 }, { "epoch": 2.07, "learning_rate": 2.7845651840244543e-05, "loss": 0.3007, "step": 20031 }, { "epoch": 2.07, "learning_rate": 2.784319678215523e-05, "loss": 0.1438, "step": 20032 }, { "epoch": 2.07, "learning_rate": 2.784074173858502e-05, "loss": 0.0983, "step": 20033 }, { "epoch": 2.07, "learning_rate": 2.7838286709550436e-05, "loss": 0.5091, "step": 20034 }, { "epoch": 2.07, "learning_rate": 2.7835831695067993e-05, "loss": 0.5275, "step": 20035 }, { "epoch": 2.07, "learning_rate": 2.7833376695154222e-05, "loss": 0.3716, "step": 20036 }, { "epoch": 2.07, "learning_rate": 2.7830921709825657e-05, "loss": 0.5498, "step": 20037 }, { "epoch": 2.07, "learning_rate": 2.7828466739098822e-05, "loss": 0.4277, "step": 20038 }, { "epoch": 2.07, "learning_rate": 2.7826011782990237e-05, "loss": 0.5645, "step": 20039 }, { "epoch": 2.07, "learning_rate": 2.782355684151643e-05, "loss": 0.1771, "step": 20040 }, { "epoch": 2.07, "learning_rate": 2.7821101914693942e-05, "loss": 0.8683, "step": 20041 }, { "epoch": 2.07, "learning_rate": 2.7818647002539273e-05, "loss": 0.052, "step": 20042 }, { "epoch": 2.07, "learning_rate": 2.7816192105068963e-05, "loss": 0.0483, "step": 20043 }, { "epoch": 2.07, "learning_rate": 2.7813737222299547e-05, "loss": 0.4859, "step": 20044 }, { "epoch": 2.07, "learning_rate": 2.781128235424753e-05, "loss": 0.1359, "step": 20045 }, { "epoch": 2.07, "learning_rate": 2.7808827500929445e-05, "loss": 0.3665, "step": 20046 }, { "epoch": 2.07, "learning_rate": 2.7806372662361827e-05, "loss": 0.5041, "step": 20047 }, { "epoch": 2.07, "learning_rate": 2.7803917838561198e-05, "loss": 0.4643, "step": 20048 }, { "epoch": 2.07, "learning_rate": 2.780146302954407e-05, "loss": 0.2222, "step": 20049 }, { "epoch": 2.07, "learning_rate": 2.7799008235326984e-05, "loss": 0.4145, "step": 20050 }, { "epoch": 2.07, "learning_rate": 2.779655345592646e-05, "loss": 0.6036, "step": 20051 }, { "epoch": 2.07, "learning_rate": 2.7794098691359025e-05, "loss": 0.2156, "step": 20052 }, { "epoch": 2.07, "learning_rate": 2.7791643941641196e-05, "loss": 0.249, "step": 20053 }, { "epoch": 2.07, "learning_rate": 2.778918920678952e-05, "loss": 0.5023, "step": 20054 }, { "epoch": 2.07, "learning_rate": 2.778673448682049e-05, "loss": 0.3351, "step": 20055 }, { "epoch": 2.07, "learning_rate": 2.778427978175064e-05, "loss": 0.6623, "step": 20056 }, { "epoch": 2.07, "learning_rate": 2.7781825091596507e-05, "loss": 0.3832, "step": 20057 }, { "epoch": 2.07, "learning_rate": 2.7779370416374612e-05, "loss": 0.3925, "step": 20058 }, { "epoch": 2.07, "learning_rate": 2.7776915756101473e-05, "loss": 0.1071, "step": 20059 }, { "epoch": 2.07, "learning_rate": 2.7774461110793614e-05, "loss": 0.3158, "step": 20060 }, { "epoch": 2.07, "learning_rate": 2.777200648046756e-05, "loss": 0.2654, "step": 20061 }, { "epoch": 2.07, "learning_rate": 2.7769551865139847e-05, "loss": 0.3875, "step": 20062 }, { "epoch": 2.07, "learning_rate": 2.776709726482698e-05, "loss": 0.5105, "step": 20063 }, { "epoch": 2.07, "learning_rate": 2.7764642679545494e-05, "loss": 0.7113, "step": 20064 }, { "epoch": 2.07, "learning_rate": 2.7762188109311906e-05, "loss": 0.3977, "step": 20065 }, { "epoch": 2.07, "learning_rate": 2.7759733554142762e-05, "loss": 0.0709, "step": 20066 }, { "epoch": 2.07, "learning_rate": 2.7757279014054555e-05, "loss": 0.5617, "step": 20067 }, { "epoch": 2.07, "learning_rate": 2.7754824489063826e-05, "loss": 0.4309, "step": 20068 }, { "epoch": 2.07, "learning_rate": 2.7752369979187086e-05, "loss": 0.385, "step": 20069 }, { "epoch": 2.07, "learning_rate": 2.7749915484440864e-05, "loss": 0.6061, "step": 20070 }, { "epoch": 2.07, "learning_rate": 2.7747461004841686e-05, "loss": 0.2711, "step": 20071 }, { "epoch": 2.07, "learning_rate": 2.774500654040608e-05, "loss": 0.7606, "step": 20072 }, { "epoch": 2.07, "learning_rate": 2.7742552091150557e-05, "loss": 0.2932, "step": 20073 }, { "epoch": 2.07, "learning_rate": 2.7740097657091644e-05, "loss": 0.1674, "step": 20074 }, { "epoch": 2.07, "learning_rate": 2.7737643238245863e-05, "loss": 0.0743, "step": 20075 }, { "epoch": 2.07, "learning_rate": 2.773518883462975e-05, "loss": 0.8421, "step": 20076 }, { "epoch": 2.07, "learning_rate": 2.7732734446259806e-05, "loss": 0.5548, "step": 20077 }, { "epoch": 2.07, "learning_rate": 2.773028007315256e-05, "loss": 0.1986, "step": 20078 }, { "epoch": 2.07, "learning_rate": 2.7727825715324556e-05, "loss": 0.2284, "step": 20079 }, { "epoch": 2.07, "learning_rate": 2.7725371372792283e-05, "loss": 0.7333, "step": 20080 }, { "epoch": 2.07, "learning_rate": 2.7722917045572275e-05, "loss": 0.4513, "step": 20081 }, { "epoch": 2.07, "learning_rate": 2.7720462733681063e-05, "loss": 0.2106, "step": 20082 }, { "epoch": 2.07, "learning_rate": 2.7718008437135164e-05, "loss": 0.5756, "step": 20083 }, { "epoch": 2.07, "learning_rate": 2.771555415595109e-05, "loss": 0.1831, "step": 20084 }, { "epoch": 2.07, "learning_rate": 2.7713099890145374e-05, "loss": 0.2697, "step": 20085 }, { "epoch": 2.07, "learning_rate": 2.7710645639734537e-05, "loss": 0.2779, "step": 20086 }, { "epoch": 2.07, "learning_rate": 2.77081914047351e-05, "loss": 0.3825, "step": 20087 }, { "epoch": 2.07, "learning_rate": 2.7705737185163573e-05, "loss": 0.1259, "step": 20088 }, { "epoch": 2.07, "learning_rate": 2.770328298103649e-05, "loss": 0.5211, "step": 20089 }, { "epoch": 2.07, "learning_rate": 2.7700828792370378e-05, "loss": 0.3425, "step": 20090 }, { "epoch": 2.07, "learning_rate": 2.7698374619181743e-05, "loss": 0.3797, "step": 20091 }, { "epoch": 2.07, "learning_rate": 2.7695920461487116e-05, "loss": 0.2325, "step": 20092 }, { "epoch": 2.07, "learning_rate": 2.7693466319303007e-05, "loss": 0.1025, "step": 20093 }, { "epoch": 2.07, "learning_rate": 2.769101219264594e-05, "loss": 0.1032, "step": 20094 }, { "epoch": 2.07, "learning_rate": 2.768855808153244e-05, "loss": 0.3841, "step": 20095 }, { "epoch": 2.07, "learning_rate": 2.7686103985979033e-05, "loss": 0.2448, "step": 20096 }, { "epoch": 2.07, "learning_rate": 2.7683649906002223e-05, "loss": 0.3041, "step": 20097 }, { "epoch": 2.07, "learning_rate": 2.7681195841618545e-05, "loss": 0.161, "step": 20098 }, { "epoch": 2.07, "learning_rate": 2.7678741792844514e-05, "loss": 0.7316, "step": 20099 }, { "epoch": 2.07, "learning_rate": 2.767628775969665e-05, "loss": 0.078, "step": 20100 }, { "epoch": 2.07, "learning_rate": 2.7673833742191474e-05, "loss": 0.4408, "step": 20101 }, { "epoch": 2.07, "learning_rate": 2.7671379740345502e-05, "loss": 0.6824, "step": 20102 }, { "epoch": 2.07, "learning_rate": 2.7668925754175257e-05, "loss": 0.1364, "step": 20103 }, { "epoch": 2.07, "learning_rate": 2.766647178369727e-05, "loss": 0.2989, "step": 20104 }, { "epoch": 2.07, "learning_rate": 2.766401782892804e-05, "loss": 0.1835, "step": 20105 }, { "epoch": 2.07, "learning_rate": 2.7661563889884095e-05, "loss": 0.3167, "step": 20106 }, { "epoch": 2.07, "learning_rate": 2.7659109966581953e-05, "loss": 0.3308, "step": 20107 }, { "epoch": 2.07, "learning_rate": 2.7656656059038134e-05, "loss": 0.5319, "step": 20108 }, { "epoch": 2.07, "learning_rate": 2.765420216726916e-05, "loss": 0.5721, "step": 20109 }, { "epoch": 2.07, "learning_rate": 2.765174829129155e-05, "loss": 0.2099, "step": 20110 }, { "epoch": 2.07, "learning_rate": 2.764929443112182e-05, "loss": 0.1809, "step": 20111 }, { "epoch": 2.07, "learning_rate": 2.7646840586776484e-05, "loss": 0.1552, "step": 20112 }, { "epoch": 2.07, "learning_rate": 2.764438675827207e-05, "loss": 0.0453, "step": 20113 }, { "epoch": 2.07, "learning_rate": 2.76419329456251e-05, "loss": 0.4393, "step": 20114 }, { "epoch": 2.07, "learning_rate": 2.763947914885208e-05, "loss": 0.1637, "step": 20115 }, { "epoch": 2.07, "learning_rate": 2.7637025367969532e-05, "loss": 0.15, "step": 20116 }, { "epoch": 2.07, "learning_rate": 2.7634571602993992e-05, "loss": 0.6099, "step": 20117 }, { "epoch": 2.07, "learning_rate": 2.7632117853941947e-05, "loss": 0.3219, "step": 20118 }, { "epoch": 2.07, "learning_rate": 2.7629664120829933e-05, "loss": 0.4241, "step": 20119 }, { "epoch": 2.07, "learning_rate": 2.762721040367447e-05, "loss": 1.0558, "step": 20120 }, { "epoch": 2.07, "learning_rate": 2.7624756702492064e-05, "loss": 0.1949, "step": 20121 }, { "epoch": 2.07, "learning_rate": 2.7622303017299238e-05, "loss": 0.2943, "step": 20122 }, { "epoch": 2.07, "learning_rate": 2.7619849348112517e-05, "loss": 0.6372, "step": 20123 }, { "epoch": 2.08, "learning_rate": 2.7617395694948416e-05, "loss": 0.2304, "step": 20124 }, { "epoch": 2.08, "learning_rate": 2.7614942057823447e-05, "loss": 0.2936, "step": 20125 }, { "epoch": 2.08, "learning_rate": 2.7612488436754126e-05, "loss": 0.2747, "step": 20126 }, { "epoch": 2.08, "learning_rate": 2.7610034831756977e-05, "loss": 0.3141, "step": 20127 }, { "epoch": 2.08, "learning_rate": 2.760758124284852e-05, "loss": 0.3255, "step": 20128 }, { "epoch": 2.08, "learning_rate": 2.7605127670045265e-05, "loss": 0.3235, "step": 20129 }, { "epoch": 2.08, "learning_rate": 2.7602674113363733e-05, "loss": 0.1567, "step": 20130 }, { "epoch": 2.08, "learning_rate": 2.7600220572820432e-05, "loss": 0.2275, "step": 20131 }, { "epoch": 2.08, "learning_rate": 2.759776704843188e-05, "loss": 0.1772, "step": 20132 }, { "epoch": 2.08, "learning_rate": 2.75953135402146e-05, "loss": 0.2003, "step": 20133 }, { "epoch": 2.08, "learning_rate": 2.7592860048185114e-05, "loss": 0.5416, "step": 20134 }, { "epoch": 2.08, "learning_rate": 2.7590406572359924e-05, "loss": 0.4674, "step": 20135 }, { "epoch": 2.08, "learning_rate": 2.7587953112755554e-05, "loss": 0.3671, "step": 20136 }, { "epoch": 2.08, "learning_rate": 2.7585499669388518e-05, "loss": 0.2047, "step": 20137 }, { "epoch": 2.08, "learning_rate": 2.758304624227534e-05, "loss": 0.3585, "step": 20138 }, { "epoch": 2.08, "learning_rate": 2.7580592831432524e-05, "loss": 0.3624, "step": 20139 }, { "epoch": 2.08, "learning_rate": 2.7578139436876588e-05, "loss": 0.8957, "step": 20140 }, { "epoch": 2.08, "learning_rate": 2.757568605862405e-05, "loss": 0.3956, "step": 20141 }, { "epoch": 2.08, "learning_rate": 2.7573232696691446e-05, "loss": 0.367, "step": 20142 }, { "epoch": 2.08, "learning_rate": 2.7570779351095254e-05, "loss": 0.543, "step": 20143 }, { "epoch": 2.08, "learning_rate": 2.7568326021852013e-05, "loss": 0.3604, "step": 20144 }, { "epoch": 2.08, "learning_rate": 2.756587270897823e-05, "loss": 0.2125, "step": 20145 }, { "epoch": 2.08, "learning_rate": 2.756341941249042e-05, "loss": 0.6182, "step": 20146 }, { "epoch": 2.08, "learning_rate": 2.7560966132405096e-05, "loss": 0.4752, "step": 20147 }, { "epoch": 2.08, "learning_rate": 2.7558512868738787e-05, "loss": 0.399, "step": 20148 }, { "epoch": 2.08, "learning_rate": 2.7556059621507996e-05, "loss": 0.4609, "step": 20149 }, { "epoch": 2.08, "learning_rate": 2.7553606390729234e-05, "loss": 0.3348, "step": 20150 }, { "epoch": 2.08, "learning_rate": 2.7551153176419022e-05, "loss": 0.036, "step": 20151 }, { "epoch": 2.08, "learning_rate": 2.754869997859388e-05, "loss": 0.3319, "step": 20152 }, { "epoch": 2.08, "learning_rate": 2.754624679727031e-05, "loss": 0.1612, "step": 20153 }, { "epoch": 2.08, "learning_rate": 2.7543793632464848e-05, "loss": 0.4569, "step": 20154 }, { "epoch": 2.08, "learning_rate": 2.7541340484193976e-05, "loss": 0.5128, "step": 20155 }, { "epoch": 2.08, "learning_rate": 2.753888735247423e-05, "loss": 0.9012, "step": 20156 }, { "epoch": 2.08, "learning_rate": 2.7536434237322117e-05, "loss": 0.1783, "step": 20157 }, { "epoch": 2.08, "learning_rate": 2.7533981138754153e-05, "loss": 0.1775, "step": 20158 }, { "epoch": 2.08, "learning_rate": 2.753152805678685e-05, "loss": 0.4083, "step": 20159 }, { "epoch": 2.08, "learning_rate": 2.752907499143672e-05, "loss": 0.6404, "step": 20160 }, { "epoch": 2.08, "learning_rate": 2.752662194272028e-05, "loss": 0.3824, "step": 20161 }, { "epoch": 2.08, "learning_rate": 2.752416891065405e-05, "loss": 0.3402, "step": 20162 }, { "epoch": 2.08, "learning_rate": 2.752171589525453e-05, "loss": 0.1894, "step": 20163 }, { "epoch": 2.08, "learning_rate": 2.7519262896538234e-05, "loss": 0.1685, "step": 20164 }, { "epoch": 2.08, "learning_rate": 2.7516809914521685e-05, "loss": 0.7925, "step": 20165 }, { "epoch": 2.08, "learning_rate": 2.7514356949221392e-05, "loss": 0.0839, "step": 20166 }, { "epoch": 2.08, "learning_rate": 2.751190400065387e-05, "loss": 0.3462, "step": 20167 }, { "epoch": 2.08, "learning_rate": 2.7509451068835624e-05, "loss": 0.2747, "step": 20168 }, { "epoch": 2.08, "learning_rate": 2.750699815378317e-05, "loss": 0.2711, "step": 20169 }, { "epoch": 2.08, "learning_rate": 2.7504545255513015e-05, "loss": 0.2164, "step": 20170 }, { "epoch": 2.08, "learning_rate": 2.7502092374041683e-05, "loss": 0.2957, "step": 20171 }, { "epoch": 2.08, "learning_rate": 2.7499639509385682e-05, "loss": 0.5557, "step": 20172 }, { "epoch": 2.08, "learning_rate": 2.7497186661561524e-05, "loss": 0.3734, "step": 20173 }, { "epoch": 2.08, "learning_rate": 2.7494733830585715e-05, "loss": 0.2023, "step": 20174 }, { "epoch": 2.08, "learning_rate": 2.749228101647477e-05, "loss": 0.7072, "step": 20175 }, { "epoch": 2.08, "learning_rate": 2.748982821924521e-05, "loss": 0.4425, "step": 20176 }, { "epoch": 2.08, "learning_rate": 2.7487375438913536e-05, "loss": 0.4157, "step": 20177 }, { "epoch": 2.08, "learning_rate": 2.7484922675496258e-05, "loss": 0.5308, "step": 20178 }, { "epoch": 2.08, "learning_rate": 2.7482469929009906e-05, "loss": 0.3224, "step": 20179 }, { "epoch": 2.08, "learning_rate": 2.7480017199470964e-05, "loss": 0.1181, "step": 20180 }, { "epoch": 2.08, "learning_rate": 2.747756448689596e-05, "loss": 0.114, "step": 20181 }, { "epoch": 2.08, "learning_rate": 2.7475111791301406e-05, "loss": 0.6812, "step": 20182 }, { "epoch": 2.08, "learning_rate": 2.74726591127038e-05, "loss": 0.6254, "step": 20183 }, { "epoch": 2.08, "learning_rate": 2.7470206451119664e-05, "loss": 0.7207, "step": 20184 }, { "epoch": 2.08, "learning_rate": 2.7467753806565507e-05, "loss": 0.2995, "step": 20185 }, { "epoch": 2.08, "learning_rate": 2.7465301179057848e-05, "loss": 0.8044, "step": 20186 }, { "epoch": 2.08, "learning_rate": 2.7462848568613175e-05, "loss": 0.2, "step": 20187 }, { "epoch": 2.08, "learning_rate": 2.7460395975248015e-05, "loss": 0.2572, "step": 20188 }, { "epoch": 2.08, "learning_rate": 2.7457943398978876e-05, "loss": 0.1994, "step": 20189 }, { "epoch": 2.08, "learning_rate": 2.7455490839822272e-05, "loss": 0.2148, "step": 20190 }, { "epoch": 2.08, "learning_rate": 2.7453038297794704e-05, "loss": 0.056, "step": 20191 }, { "epoch": 2.08, "learning_rate": 2.74505857729127e-05, "loss": 0.205, "step": 20192 }, { "epoch": 2.08, "learning_rate": 2.7448133265192737e-05, "loss": 0.2954, "step": 20193 }, { "epoch": 2.08, "learning_rate": 2.744568077465135e-05, "loss": 0.2689, "step": 20194 }, { "epoch": 2.08, "learning_rate": 2.7443228301305037e-05, "loss": 0.6444, "step": 20195 }, { "epoch": 2.08, "learning_rate": 2.7440775845170323e-05, "loss": 0.2805, "step": 20196 }, { "epoch": 2.08, "learning_rate": 2.7438323406263697e-05, "loss": 0.6122, "step": 20197 }, { "epoch": 2.08, "learning_rate": 2.7435870984601686e-05, "loss": 0.1949, "step": 20198 }, { "epoch": 2.08, "learning_rate": 2.7433418580200784e-05, "loss": 0.3043, "step": 20199 }, { "epoch": 2.08, "learning_rate": 2.743096619307752e-05, "loss": 0.6853, "step": 20200 }, { "epoch": 2.08, "learning_rate": 2.742851382324838e-05, "loss": 0.0995, "step": 20201 }, { "epoch": 2.08, "learning_rate": 2.7426061470729885e-05, "loss": 0.1278, "step": 20202 }, { "epoch": 2.08, "learning_rate": 2.742360913553854e-05, "loss": 0.2254, "step": 20203 }, { "epoch": 2.08, "learning_rate": 2.7421156817690873e-05, "loss": 0.2615, "step": 20204 }, { "epoch": 2.08, "learning_rate": 2.741870451720336e-05, "loss": 0.1693, "step": 20205 }, { "epoch": 2.08, "learning_rate": 2.7416252234092526e-05, "loss": 0.9104, "step": 20206 }, { "epoch": 2.08, "learning_rate": 2.7413799968374875e-05, "loss": 0.1414, "step": 20207 }, { "epoch": 2.08, "learning_rate": 2.7411347720066915e-05, "loss": 0.3467, "step": 20208 }, { "epoch": 2.08, "learning_rate": 2.740889548918516e-05, "loss": 0.1985, "step": 20209 }, { "epoch": 2.08, "learning_rate": 2.7406443275746118e-05, "loss": 0.488, "step": 20210 }, { "epoch": 2.08, "learning_rate": 2.7403991079766288e-05, "loss": 0.4618, "step": 20211 }, { "epoch": 2.08, "learning_rate": 2.740153890126218e-05, "loss": 0.2924, "step": 20212 }, { "epoch": 2.08, "learning_rate": 2.739908674025031e-05, "loss": 0.625, "step": 20213 }, { "epoch": 2.08, "learning_rate": 2.739663459674718e-05, "loss": 0.4476, "step": 20214 }, { "epoch": 2.08, "learning_rate": 2.7394182470769294e-05, "loss": 0.1731, "step": 20215 }, { "epoch": 2.08, "learning_rate": 2.739173036233316e-05, "loss": 0.1486, "step": 20216 }, { "epoch": 2.08, "learning_rate": 2.73892782714553e-05, "loss": 0.5964, "step": 20217 }, { "epoch": 2.08, "learning_rate": 2.7386826198152195e-05, "loss": 0.5754, "step": 20218 }, { "epoch": 2.09, "learning_rate": 2.7384374142440364e-05, "loss": 0.9393, "step": 20219 }, { "epoch": 2.09, "learning_rate": 2.7381922104336322e-05, "loss": 0.2887, "step": 20220 }, { "epoch": 2.09, "learning_rate": 2.7379470083856562e-05, "loss": 0.76, "step": 20221 }, { "epoch": 2.09, "learning_rate": 2.7377018081017595e-05, "loss": 0.3508, "step": 20222 }, { "epoch": 2.09, "learning_rate": 2.737456609583593e-05, "loss": 0.3891, "step": 20223 }, { "epoch": 2.09, "learning_rate": 2.737211412832808e-05, "loss": 0.0339, "step": 20224 }, { "epoch": 2.09, "learning_rate": 2.736966217851053e-05, "loss": 0.2399, "step": 20225 }, { "epoch": 2.09, "learning_rate": 2.7367210246399803e-05, "loss": 0.1865, "step": 20226 }, { "epoch": 2.09, "learning_rate": 2.73647583320124e-05, "loss": 0.1124, "step": 20227 }, { "epoch": 2.09, "learning_rate": 2.7362306435364832e-05, "loss": 0.831, "step": 20228 }, { "epoch": 2.09, "learning_rate": 2.7359854556473595e-05, "loss": 0.5976, "step": 20229 }, { "epoch": 2.09, "learning_rate": 2.7357402695355214e-05, "loss": 0.3413, "step": 20230 }, { "epoch": 2.09, "learning_rate": 2.735495085202616e-05, "loss": 0.5819, "step": 20231 }, { "epoch": 2.09, "learning_rate": 2.7352499026502966e-05, "loss": 0.5744, "step": 20232 }, { "epoch": 2.09, "learning_rate": 2.7350047218802128e-05, "loss": 0.3079, "step": 20233 }, { "epoch": 2.09, "learning_rate": 2.7347595428940153e-05, "loss": 0.6535, "step": 20234 }, { "epoch": 2.09, "learning_rate": 2.7345143656933544e-05, "loss": 0.2345, "step": 20235 }, { "epoch": 2.09, "learning_rate": 2.7342691902798805e-05, "loss": 0.3303, "step": 20236 }, { "epoch": 2.09, "learning_rate": 2.7340240166552443e-05, "loss": 0.3042, "step": 20237 }, { "epoch": 2.09, "learning_rate": 2.733778844821097e-05, "loss": 0.5159, "step": 20238 }, { "epoch": 2.09, "learning_rate": 2.733533674779087e-05, "loss": 0.4257, "step": 20239 }, { "epoch": 2.09, "learning_rate": 2.7332885065308662e-05, "loss": 0.1632, "step": 20240 }, { "epoch": 2.09, "learning_rate": 2.7330433400780848e-05, "loss": 0.4534, "step": 20241 }, { "epoch": 2.09, "learning_rate": 2.7327981754223945e-05, "loss": 0.3041, "step": 20242 }, { "epoch": 2.09, "learning_rate": 2.7325530125654434e-05, "loss": 0.1782, "step": 20243 }, { "epoch": 2.09, "learning_rate": 2.732307851508883e-05, "loss": 0.1612, "step": 20244 }, { "epoch": 2.09, "learning_rate": 2.7320626922543632e-05, "loss": 0.767, "step": 20245 }, { "epoch": 2.09, "learning_rate": 2.7318175348035342e-05, "loss": 0.2696, "step": 20246 }, { "epoch": 2.09, "learning_rate": 2.7315723791580475e-05, "loss": 0.1448, "step": 20247 }, { "epoch": 2.09, "learning_rate": 2.7313272253195527e-05, "loss": 0.3492, "step": 20248 }, { "epoch": 2.09, "learning_rate": 2.7310820732897e-05, "loss": 0.4396, "step": 20249 }, { "epoch": 2.09, "learning_rate": 2.7308369230701393e-05, "loss": 0.6563, "step": 20250 }, { "epoch": 2.09, "learning_rate": 2.730591774662522e-05, "loss": 0.3034, "step": 20251 }, { "epoch": 2.09, "learning_rate": 2.730346628068498e-05, "loss": 0.0796, "step": 20252 }, { "epoch": 2.09, "learning_rate": 2.7301014832897174e-05, "loss": 0.2747, "step": 20253 }, { "epoch": 2.09, "learning_rate": 2.72985634032783e-05, "loss": 0.1737, "step": 20254 }, { "epoch": 2.09, "learning_rate": 2.729611199184488e-05, "loss": 0.0477, "step": 20255 }, { "epoch": 2.09, "learning_rate": 2.7293660598613386e-05, "loss": 0.0957, "step": 20256 }, { "epoch": 2.09, "learning_rate": 2.729120922360034e-05, "loss": 0.8854, "step": 20257 }, { "epoch": 2.09, "learning_rate": 2.728875786682224e-05, "loss": 0.1932, "step": 20258 }, { "epoch": 2.09, "learning_rate": 2.7286306528295587e-05, "loss": 0.4411, "step": 20259 }, { "epoch": 2.09, "learning_rate": 2.728385520803688e-05, "loss": 0.2557, "step": 20260 }, { "epoch": 2.09, "learning_rate": 2.7281403906062627e-05, "loss": 0.7686, "step": 20261 }, { "epoch": 2.09, "learning_rate": 2.727895262238933e-05, "loss": 0.1529, "step": 20262 }, { "epoch": 2.09, "learning_rate": 2.7276501357033483e-05, "loss": 0.6955, "step": 20263 }, { "epoch": 2.09, "learning_rate": 2.7274050110011592e-05, "loss": 0.2052, "step": 20264 }, { "epoch": 2.09, "learning_rate": 2.7271598881340153e-05, "loss": 0.4717, "step": 20265 }, { "epoch": 2.09, "learning_rate": 2.7269147671035684e-05, "loss": 0.338, "step": 20266 }, { "epoch": 2.09, "learning_rate": 2.7266696479114673e-05, "loss": 0.1834, "step": 20267 }, { "epoch": 2.09, "learning_rate": 2.726424530559362e-05, "loss": 0.3462, "step": 20268 }, { "epoch": 2.09, "learning_rate": 2.726179415048902e-05, "loss": 0.4047, "step": 20269 }, { "epoch": 2.09, "learning_rate": 2.7259343013817387e-05, "loss": 0.5371, "step": 20270 }, { "epoch": 2.09, "learning_rate": 2.725689189559521e-05, "loss": 0.3232, "step": 20271 }, { "epoch": 2.09, "learning_rate": 2.7254440795839e-05, "loss": 0.1557, "step": 20272 }, { "epoch": 2.09, "learning_rate": 2.7251989714565254e-05, "loss": 0.207, "step": 20273 }, { "epoch": 2.09, "learning_rate": 2.724953865179046e-05, "loss": 0.1777, "step": 20274 }, { "epoch": 2.09, "learning_rate": 2.7247087607531135e-05, "loss": 0.5957, "step": 20275 }, { "epoch": 2.09, "learning_rate": 2.724463658180378e-05, "loss": 0.671, "step": 20276 }, { "epoch": 2.09, "learning_rate": 2.7242185574624875e-05, "loss": 0.5967, "step": 20277 }, { "epoch": 2.09, "learning_rate": 2.7239734586010935e-05, "loss": 0.2139, "step": 20278 }, { "epoch": 2.09, "learning_rate": 2.7237283615978454e-05, "loss": 0.2246, "step": 20279 }, { "epoch": 2.09, "learning_rate": 2.7234832664543947e-05, "loss": 0.7682, "step": 20280 }, { "epoch": 2.09, "learning_rate": 2.7232381731723887e-05, "loss": 0.2323, "step": 20281 }, { "epoch": 2.09, "learning_rate": 2.7229930817534788e-05, "loss": 0.7448, "step": 20282 }, { "epoch": 2.09, "learning_rate": 2.7227479921993146e-05, "loss": 0.8669, "step": 20283 }, { "epoch": 2.09, "learning_rate": 2.7225029045115455e-05, "loss": 0.1773, "step": 20284 }, { "epoch": 2.09, "learning_rate": 2.7222578186918222e-05, "loss": 0.134, "step": 20285 }, { "epoch": 2.09, "learning_rate": 2.722012734741795e-05, "loss": 0.0913, "step": 20286 }, { "epoch": 2.09, "learning_rate": 2.7217676526631123e-05, "loss": 0.2704, "step": 20287 }, { "epoch": 2.09, "learning_rate": 2.7215225724574248e-05, "loss": 0.0502, "step": 20288 }, { "epoch": 2.09, "learning_rate": 2.721277494126382e-05, "loss": 0.13, "step": 20289 }, { "epoch": 2.09, "learning_rate": 2.7210324176716347e-05, "loss": 0.3099, "step": 20290 }, { "epoch": 2.09, "learning_rate": 2.7207873430948314e-05, "loss": 0.2242, "step": 20291 }, { "epoch": 2.09, "learning_rate": 2.720542270397623e-05, "loss": 0.2399, "step": 20292 }, { "epoch": 2.09, "learning_rate": 2.720297199581658e-05, "loss": 0.5059, "step": 20293 }, { "epoch": 2.09, "learning_rate": 2.7200521306485864e-05, "loss": 0.1284, "step": 20294 }, { "epoch": 2.09, "learning_rate": 2.7198070636000585e-05, "loss": 0.73, "step": 20295 }, { "epoch": 2.09, "learning_rate": 2.7195619984377245e-05, "loss": 0.1613, "step": 20296 }, { "epoch": 2.09, "learning_rate": 2.7193169351632332e-05, "loss": 0.4486, "step": 20297 }, { "epoch": 2.09, "learning_rate": 2.719071873778234e-05, "loss": 0.1759, "step": 20298 }, { "epoch": 2.09, "learning_rate": 2.718826814284378e-05, "loss": 0.0612, "step": 20299 }, { "epoch": 2.09, "learning_rate": 2.718581756683314e-05, "loss": 0.3168, "step": 20300 }, { "epoch": 2.09, "learning_rate": 2.7183367009766914e-05, "loss": 0.2072, "step": 20301 }, { "epoch": 2.09, "learning_rate": 2.7180916471661602e-05, "loss": 0.4285, "step": 20302 }, { "epoch": 2.09, "learning_rate": 2.71784659525337e-05, "loss": 0.2632, "step": 20303 }, { "epoch": 2.09, "learning_rate": 2.7176015452399714e-05, "loss": 0.3713, "step": 20304 }, { "epoch": 2.09, "learning_rate": 2.717356497127613e-05, "loss": 0.6174, "step": 20305 }, { "epoch": 2.09, "learning_rate": 2.7171114509179446e-05, "loss": 0.4997, "step": 20306 }, { "epoch": 2.09, "learning_rate": 2.7168664066126152e-05, "loss": 0.3863, "step": 20307 }, { "epoch": 2.09, "learning_rate": 2.7166213642132743e-05, "loss": 0.3158, "step": 20308 }, { "epoch": 2.09, "learning_rate": 2.716376323721573e-05, "loss": 0.2026, "step": 20309 }, { "epoch": 2.09, "learning_rate": 2.7161312851391596e-05, "loss": 0.5032, "step": 20310 }, { "epoch": 2.09, "learning_rate": 2.7158862484676838e-05, "loss": 0.6019, "step": 20311 }, { "epoch": 2.09, "learning_rate": 2.715641213708795e-05, "loss": 0.3299, "step": 20312 }, { "epoch": 2.09, "learning_rate": 2.7153961808641434e-05, "loss": 0.4727, "step": 20313 }, { "epoch": 2.09, "learning_rate": 2.7151511499353786e-05, "loss": 0.4736, "step": 20314 }, { "epoch": 2.1, "learning_rate": 2.7149061209241493e-05, "loss": 0.3346, "step": 20315 }, { "epoch": 2.1, "learning_rate": 2.7146610938321048e-05, "loss": 0.3243, "step": 20316 }, { "epoch": 2.1, "learning_rate": 2.7144160686608953e-05, "loss": 0.4445, "step": 20317 }, { "epoch": 2.1, "learning_rate": 2.714171045412171e-05, "loss": 0.4516, "step": 20318 }, { "epoch": 2.1, "learning_rate": 2.7139260240875794e-05, "loss": 0.3867, "step": 20319 }, { "epoch": 2.1, "learning_rate": 2.7136810046887717e-05, "loss": 0.6016, "step": 20320 }, { "epoch": 2.1, "learning_rate": 2.7134359872173953e-05, "loss": 0.1614, "step": 20321 }, { "epoch": 2.1, "learning_rate": 2.713190971675101e-05, "loss": 0.2317, "step": 20322 }, { "epoch": 2.1, "learning_rate": 2.7129459580635383e-05, "loss": 0.3433, "step": 20323 }, { "epoch": 2.1, "learning_rate": 2.7127009463843564e-05, "loss": 0.3931, "step": 20324 }, { "epoch": 2.1, "learning_rate": 2.712455936639204e-05, "loss": 0.2806, "step": 20325 }, { "epoch": 2.1, "learning_rate": 2.712210928829731e-05, "loss": 0.7167, "step": 20326 }, { "epoch": 2.1, "learning_rate": 2.7119659229575865e-05, "loss": 0.3071, "step": 20327 }, { "epoch": 2.1, "learning_rate": 2.7117209190244213e-05, "loss": 0.4035, "step": 20328 }, { "epoch": 2.1, "learning_rate": 2.711475917031882e-05, "loss": 0.5392, "step": 20329 }, { "epoch": 2.1, "learning_rate": 2.711230916981621e-05, "loss": 0.3797, "step": 20330 }, { "epoch": 2.1, "learning_rate": 2.7109859188752846e-05, "loss": 0.4262, "step": 20331 }, { "epoch": 2.1, "learning_rate": 2.7107409227145233e-05, "loss": 0.6895, "step": 20332 }, { "epoch": 2.1, "learning_rate": 2.7104959285009863e-05, "loss": 0.2625, "step": 20333 }, { "epoch": 2.1, "learning_rate": 2.7102509362363237e-05, "loss": 0.1516, "step": 20334 }, { "epoch": 2.1, "learning_rate": 2.7100059459221836e-05, "loss": 0.3169, "step": 20335 }, { "epoch": 2.1, "learning_rate": 2.7097609575602155e-05, "loss": 0.1142, "step": 20336 }, { "epoch": 2.1, "learning_rate": 2.7095159711520685e-05, "loss": 0.1111, "step": 20337 }, { "epoch": 2.1, "learning_rate": 2.7092709866993926e-05, "loss": 0.393, "step": 20338 }, { "epoch": 2.1, "learning_rate": 2.7090260042038365e-05, "loss": 0.3939, "step": 20339 }, { "epoch": 2.1, "learning_rate": 2.7087810236670485e-05, "loss": 0.6117, "step": 20340 }, { "epoch": 2.1, "learning_rate": 2.708536045090679e-05, "loss": 0.2649, "step": 20341 }, { "epoch": 2.1, "learning_rate": 2.7082910684763764e-05, "loss": 0.6038, "step": 20342 }, { "epoch": 2.1, "learning_rate": 2.7080460938257908e-05, "loss": 0.3107, "step": 20343 }, { "epoch": 2.1, "learning_rate": 2.7078011211405706e-05, "loss": 0.4927, "step": 20344 }, { "epoch": 2.1, "learning_rate": 2.7075561504223642e-05, "loss": 0.3134, "step": 20345 }, { "epoch": 2.1, "learning_rate": 2.7073111816728213e-05, "loss": 0.3498, "step": 20346 }, { "epoch": 2.1, "learning_rate": 2.7070662148935908e-05, "loss": 0.1828, "step": 20347 }, { "epoch": 2.1, "learning_rate": 2.706821250086323e-05, "loss": 0.5992, "step": 20348 }, { "epoch": 2.1, "learning_rate": 2.706576287252665e-05, "loss": 0.5814, "step": 20349 }, { "epoch": 2.1, "learning_rate": 2.706331326394267e-05, "loss": 0.1571, "step": 20350 }, { "epoch": 2.1, "learning_rate": 2.7060863675127776e-05, "loss": 0.3958, "step": 20351 }, { "epoch": 2.1, "learning_rate": 2.705841410609847e-05, "loss": 0.2365, "step": 20352 }, { "epoch": 2.1, "learning_rate": 2.7055964556871225e-05, "loss": 0.3134, "step": 20353 }, { "epoch": 2.1, "learning_rate": 2.7053515027462535e-05, "loss": 0.3558, "step": 20354 }, { "epoch": 2.1, "learning_rate": 2.7051065517888907e-05, "loss": 0.1314, "step": 20355 }, { "epoch": 2.1, "learning_rate": 2.7048616028166804e-05, "loss": 0.5181, "step": 20356 }, { "epoch": 2.1, "learning_rate": 2.7046166558312723e-05, "loss": 0.4068, "step": 20357 }, { "epoch": 2.1, "learning_rate": 2.7043717108343168e-05, "loss": 0.3645, "step": 20358 }, { "epoch": 2.1, "learning_rate": 2.704126767827461e-05, "loss": 0.741, "step": 20359 }, { "epoch": 2.1, "learning_rate": 2.7038818268123545e-05, "loss": 0.2765, "step": 20360 }, { "epoch": 2.1, "learning_rate": 2.7036368877906465e-05, "loss": 0.3243, "step": 20361 }, { "epoch": 2.1, "learning_rate": 2.703391950763986e-05, "loss": 0.2658, "step": 20362 }, { "epoch": 2.1, "learning_rate": 2.703147015734021e-05, "loss": 0.4393, "step": 20363 }, { "epoch": 2.1, "learning_rate": 2.7029020827024006e-05, "loss": 0.1402, "step": 20364 }, { "epoch": 2.1, "learning_rate": 2.702657151670774e-05, "loss": 0.5505, "step": 20365 }, { "epoch": 2.1, "learning_rate": 2.702412222640791e-05, "loss": 0.3036, "step": 20366 }, { "epoch": 2.1, "learning_rate": 2.702167295614098e-05, "loss": 0.4304, "step": 20367 }, { "epoch": 2.1, "learning_rate": 2.701922370592347e-05, "loss": 0.5921, "step": 20368 }, { "epoch": 2.1, "learning_rate": 2.7016774475771828e-05, "loss": 0.5632, "step": 20369 }, { "epoch": 2.1, "learning_rate": 2.7014325265702565e-05, "loss": 0.6079, "step": 20370 }, { "epoch": 2.1, "learning_rate": 2.7011876075732168e-05, "loss": 0.5819, "step": 20371 }, { "epoch": 2.1, "learning_rate": 2.7009426905877126e-05, "loss": 0.364, "step": 20372 }, { "epoch": 2.1, "learning_rate": 2.700697775615392e-05, "loss": 0.8565, "step": 20373 }, { "epoch": 2.1, "learning_rate": 2.7004528626579035e-05, "loss": 1.2093, "step": 20374 }, { "epoch": 2.1, "learning_rate": 2.7002079517168962e-05, "loss": 0.6656, "step": 20375 }, { "epoch": 2.1, "learning_rate": 2.6999630427940195e-05, "loss": 0.2729, "step": 20376 }, { "epoch": 2.1, "learning_rate": 2.6997181358909213e-05, "loss": 0.3391, "step": 20377 }, { "epoch": 2.1, "learning_rate": 2.69947323100925e-05, "loss": 0.1144, "step": 20378 }, { "epoch": 2.1, "learning_rate": 2.6992283281506544e-05, "loss": 0.3211, "step": 20379 }, { "epoch": 2.1, "learning_rate": 2.6989834273167848e-05, "loss": 0.3289, "step": 20380 }, { "epoch": 2.1, "learning_rate": 2.698738528509287e-05, "loss": 0.1949, "step": 20381 }, { "epoch": 2.1, "learning_rate": 2.6984936317298113e-05, "loss": 0.6069, "step": 20382 }, { "epoch": 2.1, "learning_rate": 2.698248736980006e-05, "loss": 0.2153, "step": 20383 }, { "epoch": 2.1, "learning_rate": 2.698003844261519e-05, "loss": 0.3833, "step": 20384 }, { "epoch": 2.1, "learning_rate": 2.6977589535759995e-05, "loss": 0.1156, "step": 20385 }, { "epoch": 2.1, "learning_rate": 2.6975140649250966e-05, "loss": 0.9224, "step": 20386 }, { "epoch": 2.1, "learning_rate": 2.6972691783104578e-05, "loss": 0.1122, "step": 20387 }, { "epoch": 2.1, "learning_rate": 2.697024293733732e-05, "loss": 0.2368, "step": 20388 }, { "epoch": 2.1, "learning_rate": 2.6967794111965677e-05, "loss": 0.8445, "step": 20389 }, { "epoch": 2.1, "learning_rate": 2.6965345307006144e-05, "loss": 0.2523, "step": 20390 }, { "epoch": 2.1, "learning_rate": 2.6962896522475185e-05, "loss": 0.331, "step": 20391 }, { "epoch": 2.1, "learning_rate": 2.69604477583893e-05, "loss": 0.6052, "step": 20392 }, { "epoch": 2.1, "learning_rate": 2.6957999014764983e-05, "loss": 0.2005, "step": 20393 }, { "epoch": 2.1, "learning_rate": 2.6955550291618688e-05, "loss": 0.4802, "step": 20394 }, { "epoch": 2.1, "learning_rate": 2.695310158896692e-05, "loss": 0.2971, "step": 20395 }, { "epoch": 2.1, "learning_rate": 2.695065290682616e-05, "loss": 0.4027, "step": 20396 }, { "epoch": 2.1, "learning_rate": 2.694820424521289e-05, "loss": 0.4038, "step": 20397 }, { "epoch": 2.1, "learning_rate": 2.6945755604143596e-05, "loss": 0.2168, "step": 20398 }, { "epoch": 2.1, "learning_rate": 2.6943306983634757e-05, "loss": 0.5875, "step": 20399 }, { "epoch": 2.1, "learning_rate": 2.6940858383702866e-05, "loss": 0.0574, "step": 20400 }, { "epoch": 2.1, "learning_rate": 2.6938409804364395e-05, "loss": 0.2081, "step": 20401 }, { "epoch": 2.1, "learning_rate": 2.6935961245635835e-05, "loss": 0.1363, "step": 20402 }, { "epoch": 2.1, "learning_rate": 2.6933512707533663e-05, "loss": 0.0748, "step": 20403 }, { "epoch": 2.1, "learning_rate": 2.6931064190074378e-05, "loss": 0.4441, "step": 20404 }, { "epoch": 2.1, "learning_rate": 2.6928615693274447e-05, "loss": 0.3403, "step": 20405 }, { "epoch": 2.1, "learning_rate": 2.6926167217150357e-05, "loss": 0.2799, "step": 20406 }, { "epoch": 2.1, "learning_rate": 2.6923718761718583e-05, "loss": 0.7625, "step": 20407 }, { "epoch": 2.1, "learning_rate": 2.6921270326995616e-05, "loss": 0.4963, "step": 20408 }, { "epoch": 2.1, "learning_rate": 2.6918821912997936e-05, "loss": 0.3746, "step": 20409 }, { "epoch": 2.1, "learning_rate": 2.6916373519742034e-05, "loss": 0.6395, "step": 20410 }, { "epoch": 2.11, "learning_rate": 2.6913925147244374e-05, "loss": 0.4439, "step": 20411 }, { "epoch": 2.11, "learning_rate": 2.691147679552145e-05, "loss": 0.0516, "step": 20412 }, { "epoch": 2.11, "learning_rate": 2.6909028464589744e-05, "loss": 0.5406, "step": 20413 }, { "epoch": 2.11, "learning_rate": 2.6906580154465737e-05, "loss": 0.2659, "step": 20414 }, { "epoch": 2.11, "learning_rate": 2.6904131865165904e-05, "loss": 0.3163, "step": 20415 }, { "epoch": 2.11, "learning_rate": 2.690168359670673e-05, "loss": 0.2678, "step": 20416 }, { "epoch": 2.11, "learning_rate": 2.68992353491047e-05, "loss": 0.7646, "step": 20417 }, { "epoch": 2.11, "learning_rate": 2.6896787122376306e-05, "loss": 0.1087, "step": 20418 }, { "epoch": 2.11, "learning_rate": 2.6894338916537996e-05, "loss": 0.4082, "step": 20419 }, { "epoch": 2.11, "learning_rate": 2.689189073160628e-05, "loss": 0.3247, "step": 20420 }, { "epoch": 2.11, "learning_rate": 2.688944256759762e-05, "loss": 0.0301, "step": 20421 }, { "epoch": 2.11, "learning_rate": 2.6886994424528507e-05, "loss": 0.0753, "step": 20422 }, { "epoch": 2.11, "learning_rate": 2.688454630241542e-05, "loss": 0.2481, "step": 20423 }, { "epoch": 2.11, "learning_rate": 2.688209820127484e-05, "loss": 0.2268, "step": 20424 }, { "epoch": 2.11, "learning_rate": 2.6879650121123244e-05, "loss": 0.3117, "step": 20425 }, { "epoch": 2.11, "learning_rate": 2.6877202061977112e-05, "loss": 0.6645, "step": 20426 }, { "epoch": 2.11, "learning_rate": 2.6874754023852925e-05, "loss": 0.0641, "step": 20427 }, { "epoch": 2.11, "learning_rate": 2.6872306006767166e-05, "loss": 0.329, "step": 20428 }, { "epoch": 2.11, "learning_rate": 2.686985801073631e-05, "loss": 0.31, "step": 20429 }, { "epoch": 2.11, "learning_rate": 2.6867410035776832e-05, "loss": 0.55, "step": 20430 }, { "epoch": 2.11, "learning_rate": 2.6864962081905237e-05, "loss": 0.2005, "step": 20431 }, { "epoch": 2.11, "learning_rate": 2.6862514149137966e-05, "loss": 0.4536, "step": 20432 }, { "epoch": 2.11, "learning_rate": 2.6860066237491512e-05, "loss": 0.1927, "step": 20433 }, { "epoch": 2.11, "learning_rate": 2.685761834698237e-05, "loss": 0.7683, "step": 20434 }, { "epoch": 2.11, "learning_rate": 2.6855170477626996e-05, "loss": 0.3443, "step": 20435 }, { "epoch": 2.11, "learning_rate": 2.685272262944188e-05, "loss": 0.1812, "step": 20436 }, { "epoch": 2.11, "learning_rate": 2.6850274802443498e-05, "loss": 0.3034, "step": 20437 }, { "epoch": 2.11, "learning_rate": 2.6847826996648336e-05, "loss": 0.2233, "step": 20438 }, { "epoch": 2.11, "learning_rate": 2.6845379212072857e-05, "loss": 0.1464, "step": 20439 }, { "epoch": 2.11, "learning_rate": 2.6842931448733547e-05, "loss": 0.3075, "step": 20440 }, { "epoch": 2.11, "learning_rate": 2.6840483706646887e-05, "loss": 0.5144, "step": 20441 }, { "epoch": 2.11, "learning_rate": 2.683803598582936e-05, "loss": 0.4199, "step": 20442 }, { "epoch": 2.11, "learning_rate": 2.683558828629743e-05, "loss": 0.4222, "step": 20443 }, { "epoch": 2.11, "learning_rate": 2.683314060806758e-05, "loss": 0.3404, "step": 20444 }, { "epoch": 2.11, "learning_rate": 2.683069295115628e-05, "loss": 0.3168, "step": 20445 }, { "epoch": 2.11, "learning_rate": 2.6828245315580012e-05, "loss": 0.2219, "step": 20446 }, { "epoch": 2.11, "learning_rate": 2.682579770135526e-05, "loss": 0.4403, "step": 20447 }, { "epoch": 2.11, "learning_rate": 2.6823350108498498e-05, "loss": 0.5039, "step": 20448 }, { "epoch": 2.11, "learning_rate": 2.682090253702619e-05, "loss": 0.2894, "step": 20449 }, { "epoch": 2.11, "learning_rate": 2.6818454986954826e-05, "loss": 0.2463, "step": 20450 }, { "epoch": 2.11, "learning_rate": 2.681600745830088e-05, "loss": 0.6392, "step": 20451 }, { "epoch": 2.11, "learning_rate": 2.6813559951080833e-05, "loss": 0.7403, "step": 20452 }, { "epoch": 2.11, "learning_rate": 2.6811112465311147e-05, "loss": 0.2778, "step": 20453 }, { "epoch": 2.11, "learning_rate": 2.6808665001008302e-05, "loss": 0.3895, "step": 20454 }, { "epoch": 2.11, "learning_rate": 2.6806217558188784e-05, "loss": 0.6232, "step": 20455 }, { "epoch": 2.11, "learning_rate": 2.680377013686907e-05, "loss": 0.7433, "step": 20456 }, { "epoch": 2.11, "learning_rate": 2.6801322737065618e-05, "loss": 0.2431, "step": 20457 }, { "epoch": 2.11, "learning_rate": 2.6798875358794917e-05, "loss": 0.1583, "step": 20458 }, { "epoch": 2.11, "learning_rate": 2.6796428002073435e-05, "loss": 0.6987, "step": 20459 }, { "epoch": 2.11, "learning_rate": 2.6793980666917645e-05, "loss": 0.1428, "step": 20460 }, { "epoch": 2.11, "learning_rate": 2.6791533353344032e-05, "loss": 0.6257, "step": 20461 }, { "epoch": 2.11, "learning_rate": 2.6789086061369068e-05, "loss": 0.6753, "step": 20462 }, { "epoch": 2.11, "learning_rate": 2.6786638791009225e-05, "loss": 0.1531, "step": 20463 }, { "epoch": 2.11, "learning_rate": 2.678419154228097e-05, "loss": 0.7586, "step": 20464 }, { "epoch": 2.11, "learning_rate": 2.678174431520079e-05, "loss": 0.0983, "step": 20465 }, { "epoch": 2.11, "learning_rate": 2.677929710978516e-05, "loss": 0.2558, "step": 20466 }, { "epoch": 2.11, "learning_rate": 2.677684992605054e-05, "loss": 0.2059, "step": 20467 }, { "epoch": 2.11, "learning_rate": 2.6774402764013426e-05, "loss": 0.2367, "step": 20468 }, { "epoch": 2.11, "learning_rate": 2.6771955623690265e-05, "loss": 0.2072, "step": 20469 }, { "epoch": 2.11, "learning_rate": 2.6769508505097548e-05, "loss": 0.2691, "step": 20470 }, { "epoch": 2.11, "learning_rate": 2.6767061408251737e-05, "loss": 0.7716, "step": 20471 }, { "epoch": 2.11, "learning_rate": 2.6764614333169323e-05, "loss": 0.1346, "step": 20472 }, { "epoch": 2.11, "learning_rate": 2.676216727986676e-05, "loss": 0.4239, "step": 20473 }, { "epoch": 2.11, "learning_rate": 2.675972024836053e-05, "loss": 0.1305, "step": 20474 }, { "epoch": 2.11, "learning_rate": 2.6757273238667103e-05, "loss": 0.1874, "step": 20475 }, { "epoch": 2.11, "learning_rate": 2.675482625080296e-05, "loss": 0.8257, "step": 20476 }, { "epoch": 2.11, "learning_rate": 2.6752379284784566e-05, "loss": 0.211, "step": 20477 }, { "epoch": 2.11, "learning_rate": 2.674993234062839e-05, "loss": 0.4176, "step": 20478 }, { "epoch": 2.11, "learning_rate": 2.674748541835091e-05, "loss": 0.1355, "step": 20479 }, { "epoch": 2.11, "learning_rate": 2.6745038517968605e-05, "loss": 0.2867, "step": 20480 }, { "epoch": 2.11, "learning_rate": 2.674259163949794e-05, "loss": 0.2152, "step": 20481 }, { "epoch": 2.11, "learning_rate": 2.6740144782955385e-05, "loss": 0.2231, "step": 20482 }, { "epoch": 2.11, "learning_rate": 2.6737697948357402e-05, "loss": 0.2183, "step": 20483 }, { "epoch": 2.11, "learning_rate": 2.6735251135720476e-05, "loss": 0.9882, "step": 20484 }, { "epoch": 2.11, "learning_rate": 2.6732804345061077e-05, "loss": 0.2406, "step": 20485 }, { "epoch": 2.11, "learning_rate": 2.673035757639568e-05, "loss": 0.4088, "step": 20486 }, { "epoch": 2.11, "learning_rate": 2.6727910829740745e-05, "loss": 0.7548, "step": 20487 }, { "epoch": 2.11, "learning_rate": 2.672546410511275e-05, "loss": 0.3357, "step": 20488 }, { "epoch": 2.11, "learning_rate": 2.672301740252816e-05, "loss": 0.1811, "step": 20489 }, { "epoch": 2.11, "learning_rate": 2.672057072200346e-05, "loss": 0.644, "step": 20490 }, { "epoch": 2.11, "learning_rate": 2.67181240635551e-05, "loss": 0.2021, "step": 20491 }, { "epoch": 2.11, "learning_rate": 2.6715677427199565e-05, "loss": 0.1596, "step": 20492 }, { "epoch": 2.11, "learning_rate": 2.6713230812953335e-05, "loss": 0.2587, "step": 20493 }, { "epoch": 2.11, "learning_rate": 2.6710784220832848e-05, "loss": 0.5315, "step": 20494 }, { "epoch": 2.11, "learning_rate": 2.67083376508546e-05, "loss": 0.0564, "step": 20495 }, { "epoch": 2.11, "learning_rate": 2.6705891103035053e-05, "loss": 0.1389, "step": 20496 }, { "epoch": 2.11, "learning_rate": 2.670344457739067e-05, "loss": 1.0981, "step": 20497 }, { "epoch": 2.11, "learning_rate": 2.6700998073937928e-05, "loss": 0.4151, "step": 20498 }, { "epoch": 2.11, "learning_rate": 2.6698551592693295e-05, "loss": 0.5728, "step": 20499 }, { "epoch": 2.11, "learning_rate": 2.6696105133673248e-05, "loss": 0.1034, "step": 20500 }, { "epoch": 2.11, "learning_rate": 2.669365869689424e-05, "loss": 0.1471, "step": 20501 }, { "epoch": 2.11, "learning_rate": 2.669121228237275e-05, "loss": 0.0999, "step": 20502 }, { "epoch": 2.11, "learning_rate": 2.6688765890125242e-05, "loss": 0.6919, "step": 20503 }, { "epoch": 2.11, "learning_rate": 2.6686319520168196e-05, "loss": 0.2255, "step": 20504 }, { "epoch": 2.11, "learning_rate": 2.6683873172518063e-05, "loss": 0.2325, "step": 20505 }, { "epoch": 2.11, "learning_rate": 2.6681426847191332e-05, "loss": 0.4572, "step": 20506 }, { "epoch": 2.12, "learning_rate": 2.667898054420445e-05, "loss": 0.2613, "step": 20507 }, { "epoch": 2.12, "learning_rate": 2.6676534263573893e-05, "loss": 0.2413, "step": 20508 }, { "epoch": 2.12, "learning_rate": 2.6674088005316128e-05, "loss": 0.4472, "step": 20509 }, { "epoch": 2.12, "learning_rate": 2.6671641769447632e-05, "loss": 0.6666, "step": 20510 }, { "epoch": 2.12, "learning_rate": 2.6669195555984856e-05, "loss": 0.3371, "step": 20511 }, { "epoch": 2.12, "learning_rate": 2.6666749364944278e-05, "loss": 0.2929, "step": 20512 }, { "epoch": 2.12, "learning_rate": 2.6664303196342366e-05, "loss": 0.4547, "step": 20513 }, { "epoch": 2.12, "learning_rate": 2.6661857050195588e-05, "loss": 0.3376, "step": 20514 }, { "epoch": 2.12, "learning_rate": 2.66594109265204e-05, "loss": 0.4618, "step": 20515 }, { "epoch": 2.12, "learning_rate": 2.6656964825333275e-05, "loss": 0.7004, "step": 20516 }, { "epoch": 2.12, "learning_rate": 2.6654518746650686e-05, "loss": 0.1379, "step": 20517 }, { "epoch": 2.12, "learning_rate": 2.6652072690489097e-05, "loss": 0.1779, "step": 20518 }, { "epoch": 2.12, "learning_rate": 2.6649626656864968e-05, "loss": 0.0908, "step": 20519 }, { "epoch": 2.12, "learning_rate": 2.664718064579477e-05, "loss": 0.8483, "step": 20520 }, { "epoch": 2.12, "learning_rate": 2.664473465729497e-05, "loss": 0.3249, "step": 20521 }, { "epoch": 2.12, "learning_rate": 2.664228869138202e-05, "loss": 0.0392, "step": 20522 }, { "epoch": 2.12, "learning_rate": 2.66398427480724e-05, "loss": 0.5494, "step": 20523 }, { "epoch": 2.12, "learning_rate": 2.663739682738258e-05, "loss": 0.994, "step": 20524 }, { "epoch": 2.12, "learning_rate": 2.6634950929329013e-05, "loss": 0.5324, "step": 20525 }, { "epoch": 2.12, "learning_rate": 2.663250505392817e-05, "loss": 0.4723, "step": 20526 }, { "epoch": 2.12, "learning_rate": 2.663005920119651e-05, "loss": 0.5551, "step": 20527 }, { "epoch": 2.12, "learning_rate": 2.662761337115051e-05, "loss": 0.1917, "step": 20528 }, { "epoch": 2.12, "learning_rate": 2.6625167563806624e-05, "loss": 1.5204, "step": 20529 }, { "epoch": 2.12, "learning_rate": 2.6622721779181318e-05, "loss": 0.2883, "step": 20530 }, { "epoch": 2.12, "learning_rate": 2.6620276017291074e-05, "loss": 0.162, "step": 20531 }, { "epoch": 2.12, "learning_rate": 2.6617830278152328e-05, "loss": 0.1412, "step": 20532 }, { "epoch": 2.12, "learning_rate": 2.6615384561781557e-05, "loss": 0.2513, "step": 20533 }, { "epoch": 2.12, "learning_rate": 2.661293886819523e-05, "loss": 0.2254, "step": 20534 }, { "epoch": 2.12, "learning_rate": 2.66104931974098e-05, "loss": 0.2939, "step": 20535 }, { "epoch": 2.12, "learning_rate": 2.6608047549441737e-05, "loss": 0.4411, "step": 20536 }, { "epoch": 2.12, "learning_rate": 2.6605601924307503e-05, "loss": 0.397, "step": 20537 }, { "epoch": 2.12, "learning_rate": 2.6603156322023572e-05, "loss": 0.5987, "step": 20538 }, { "epoch": 2.12, "learning_rate": 2.6600710742606387e-05, "loss": 0.2614, "step": 20539 }, { "epoch": 2.12, "learning_rate": 2.6598265186072427e-05, "loss": 0.7696, "step": 20540 }, { "epoch": 2.12, "learning_rate": 2.6595819652438145e-05, "loss": 0.5754, "step": 20541 }, { "epoch": 2.12, "learning_rate": 2.6593374141720017e-05, "loss": 0.4063, "step": 20542 }, { "epoch": 2.12, "learning_rate": 2.6590928653934494e-05, "loss": 0.8932, "step": 20543 }, { "epoch": 2.12, "learning_rate": 2.658848318909805e-05, "loss": 0.4774, "step": 20544 }, { "epoch": 2.12, "learning_rate": 2.6586037747227127e-05, "loss": 0.7103, "step": 20545 }, { "epoch": 2.12, "learning_rate": 2.6583592328338202e-05, "loss": 0.2131, "step": 20546 }, { "epoch": 2.12, "learning_rate": 2.6581146932447734e-05, "loss": 0.2258, "step": 20547 }, { "epoch": 2.12, "learning_rate": 2.6578701559572187e-05, "loss": 0.1594, "step": 20548 }, { "epoch": 2.12, "learning_rate": 2.657625620972802e-05, "loss": 0.0571, "step": 20549 }, { "epoch": 2.12, "learning_rate": 2.6573810882931688e-05, "loss": 0.4138, "step": 20550 }, { "epoch": 2.12, "learning_rate": 2.6571365579199664e-05, "loss": 0.4479, "step": 20551 }, { "epoch": 2.12, "learning_rate": 2.656892029854841e-05, "loss": 0.4584, "step": 20552 }, { "epoch": 2.12, "learning_rate": 2.6566475040994374e-05, "loss": 0.7957, "step": 20553 }, { "epoch": 2.12, "learning_rate": 2.6564029806554026e-05, "loss": 0.3663, "step": 20554 }, { "epoch": 2.12, "learning_rate": 2.6561584595243828e-05, "loss": 0.7503, "step": 20555 }, { "epoch": 2.12, "learning_rate": 2.6559139407080248e-05, "loss": 0.1346, "step": 20556 }, { "epoch": 2.12, "learning_rate": 2.6556694242079726e-05, "loss": 0.1025, "step": 20557 }, { "epoch": 2.12, "learning_rate": 2.6554249100258734e-05, "loss": 0.64, "step": 20558 }, { "epoch": 2.12, "learning_rate": 2.655180398163373e-05, "loss": 0.1437, "step": 20559 }, { "epoch": 2.12, "learning_rate": 2.654935888622117e-05, "loss": 0.5396, "step": 20560 }, { "epoch": 2.12, "learning_rate": 2.654691381403752e-05, "loss": 0.2507, "step": 20561 }, { "epoch": 2.12, "learning_rate": 2.6544468765099246e-05, "loss": 0.189, "step": 20562 }, { "epoch": 2.12, "learning_rate": 2.6542023739422792e-05, "loss": 0.6724, "step": 20563 }, { "epoch": 2.12, "learning_rate": 2.6539578737024627e-05, "loss": 0.6064, "step": 20564 }, { "epoch": 2.12, "learning_rate": 2.653713375792121e-05, "loss": 0.3393, "step": 20565 }, { "epoch": 2.12, "learning_rate": 2.6534688802129e-05, "loss": 0.0452, "step": 20566 }, { "epoch": 2.12, "learning_rate": 2.653224386966445e-05, "loss": 0.0715, "step": 20567 }, { "epoch": 2.12, "learning_rate": 2.652979896054402e-05, "loss": 0.615, "step": 20568 }, { "epoch": 2.12, "learning_rate": 2.652735407478419e-05, "loss": 0.0816, "step": 20569 }, { "epoch": 2.12, "learning_rate": 2.652490921240138e-05, "loss": 0.1346, "step": 20570 }, { "epoch": 2.12, "learning_rate": 2.6522464373412073e-05, "loss": 0.8534, "step": 20571 }, { "epoch": 2.12, "learning_rate": 2.6520019557832726e-05, "loss": 0.6626, "step": 20572 }, { "epoch": 2.12, "learning_rate": 2.651757476567979e-05, "loss": 0.1725, "step": 20573 }, { "epoch": 2.12, "learning_rate": 2.6515129996969724e-05, "loss": 0.6566, "step": 20574 }, { "epoch": 2.12, "learning_rate": 2.6512685251718988e-05, "loss": 0.7666, "step": 20575 }, { "epoch": 2.12, "learning_rate": 2.6510240529944044e-05, "loss": 0.3527, "step": 20576 }, { "epoch": 2.12, "learning_rate": 2.650779583166134e-05, "loss": 0.3808, "step": 20577 }, { "epoch": 2.12, "learning_rate": 2.6505351156887335e-05, "loss": 0.1331, "step": 20578 }, { "epoch": 2.12, "learning_rate": 2.650290650563849e-05, "loss": 0.1921, "step": 20579 }, { "epoch": 2.12, "learning_rate": 2.6500461877931264e-05, "loss": 0.506, "step": 20580 }, { "epoch": 2.12, "learning_rate": 2.6498017273782112e-05, "loss": 0.2013, "step": 20581 }, { "epoch": 2.12, "learning_rate": 2.649557269320749e-05, "loss": 0.2306, "step": 20582 }, { "epoch": 2.12, "learning_rate": 2.6493128136223842e-05, "loss": 0.6645, "step": 20583 }, { "epoch": 2.12, "learning_rate": 2.6490683602847637e-05, "loss": 0.1715, "step": 20584 }, { "epoch": 2.12, "learning_rate": 2.648823909309533e-05, "loss": 0.5934, "step": 20585 }, { "epoch": 2.12, "learning_rate": 2.648579460698338e-05, "loss": 0.5795, "step": 20586 }, { "epoch": 2.12, "learning_rate": 2.6483350144528233e-05, "loss": 0.3286, "step": 20587 }, { "epoch": 2.12, "learning_rate": 2.648090570574635e-05, "loss": 0.3086, "step": 20588 }, { "epoch": 2.12, "learning_rate": 2.6478461290654186e-05, "loss": 0.4131, "step": 20589 }, { "epoch": 2.12, "learning_rate": 2.6476016899268204e-05, "loss": 0.3909, "step": 20590 }, { "epoch": 2.12, "learning_rate": 2.6473572531604844e-05, "loss": 0.1392, "step": 20591 }, { "epoch": 2.12, "learning_rate": 2.647112818768057e-05, "loss": 0.2045, "step": 20592 }, { "epoch": 2.12, "learning_rate": 2.6468683867511835e-05, "loss": 0.6574, "step": 20593 }, { "epoch": 2.12, "learning_rate": 2.6466239571115106e-05, "loss": 0.6742, "step": 20594 }, { "epoch": 2.12, "learning_rate": 2.6463795298506813e-05, "loss": 0.499, "step": 20595 }, { "epoch": 2.12, "learning_rate": 2.646135104970343e-05, "loss": 0.1229, "step": 20596 }, { "epoch": 2.12, "learning_rate": 2.64589068247214e-05, "loss": 0.1214, "step": 20597 }, { "epoch": 2.12, "learning_rate": 2.6456462623577175e-05, "loss": 0.6106, "step": 20598 }, { "epoch": 2.12, "learning_rate": 2.6454018446287217e-05, "loss": 0.2956, "step": 20599 }, { "epoch": 2.12, "learning_rate": 2.6451574292867982e-05, "loss": 0.4534, "step": 20600 }, { "epoch": 2.12, "learning_rate": 2.644913016333591e-05, "loss": 0.4189, "step": 20601 }, { "epoch": 2.13, "learning_rate": 2.644668605770747e-05, "loss": 0.1569, "step": 20602 }, { "epoch": 2.13, "learning_rate": 2.6444241975999103e-05, "loss": 0.186, "step": 20603 }, { "epoch": 2.13, "learning_rate": 2.6441797918227273e-05, "loss": 0.3689, "step": 20604 }, { "epoch": 2.13, "learning_rate": 2.6439353884408423e-05, "loss": 0.5041, "step": 20605 }, { "epoch": 2.13, "learning_rate": 2.6436909874559023e-05, "loss": 0.1742, "step": 20606 }, { "epoch": 2.13, "learning_rate": 2.64344658886955e-05, "loss": 0.5442, "step": 20607 }, { "epoch": 2.13, "learning_rate": 2.6432021926834314e-05, "loss": 0.3366, "step": 20608 }, { "epoch": 2.13, "learning_rate": 2.642957798899192e-05, "loss": 0.3506, "step": 20609 }, { "epoch": 2.13, "learning_rate": 2.6427134075184785e-05, "loss": 0.4807, "step": 20610 }, { "epoch": 2.13, "learning_rate": 2.6424690185429336e-05, "loss": 0.5208, "step": 20611 }, { "epoch": 2.13, "learning_rate": 2.6422246319742037e-05, "loss": 0.4176, "step": 20612 }, { "epoch": 2.13, "learning_rate": 2.641980247813934e-05, "loss": 0.0853, "step": 20613 }, { "epoch": 2.13, "learning_rate": 2.6417358660637703e-05, "loss": 0.0854, "step": 20614 }, { "epoch": 2.13, "learning_rate": 2.641491486725356e-05, "loss": 0.3781, "step": 20615 }, { "epoch": 2.13, "learning_rate": 2.6412471098003375e-05, "loss": 0.7177, "step": 20616 }, { "epoch": 2.13, "learning_rate": 2.641002735290359e-05, "loss": 0.7338, "step": 20617 }, { "epoch": 2.13, "learning_rate": 2.640758363197067e-05, "loss": 0.5584, "step": 20618 }, { "epoch": 2.13, "learning_rate": 2.6405139935221053e-05, "loss": 0.1106, "step": 20619 }, { "epoch": 2.13, "learning_rate": 2.6402696262671198e-05, "loss": 0.9406, "step": 20620 }, { "epoch": 2.13, "learning_rate": 2.640025261433754e-05, "loss": 0.2814, "step": 20621 }, { "epoch": 2.13, "learning_rate": 2.639780899023654e-05, "loss": 0.09, "step": 20622 }, { "epoch": 2.13, "learning_rate": 2.6395365390384647e-05, "loss": 0.1957, "step": 20623 }, { "epoch": 2.13, "learning_rate": 2.639292181479832e-05, "loss": 0.2222, "step": 20624 }, { "epoch": 2.13, "learning_rate": 2.639047826349399e-05, "loss": 0.2907, "step": 20625 }, { "epoch": 2.13, "learning_rate": 2.638803473648812e-05, "loss": 0.4763, "step": 20626 }, { "epoch": 2.13, "learning_rate": 2.6385591233797145e-05, "loss": 0.2512, "step": 20627 }, { "epoch": 2.13, "learning_rate": 2.638314775543754e-05, "loss": 0.4341, "step": 20628 }, { "epoch": 2.13, "learning_rate": 2.6380704301425728e-05, "loss": 0.7346, "step": 20629 }, { "epoch": 2.13, "learning_rate": 2.637826087177817e-05, "loss": 0.4663, "step": 20630 }, { "epoch": 2.13, "learning_rate": 2.637581746651131e-05, "loss": 0.4385, "step": 20631 }, { "epoch": 2.13, "learning_rate": 2.637337408564161e-05, "loss": 0.3142, "step": 20632 }, { "epoch": 2.13, "learning_rate": 2.6370930729185498e-05, "loss": 0.2461, "step": 20633 }, { "epoch": 2.13, "learning_rate": 2.6368487397159435e-05, "loss": 0.2766, "step": 20634 }, { "epoch": 2.13, "learning_rate": 2.636604408957986e-05, "loss": 0.3066, "step": 20635 }, { "epoch": 2.13, "learning_rate": 2.636360080646323e-05, "loss": 0.4491, "step": 20636 }, { "epoch": 2.13, "learning_rate": 2.636115754782598e-05, "loss": 0.2222, "step": 20637 }, { "epoch": 2.13, "learning_rate": 2.6358714313684578e-05, "loss": 0.9561, "step": 20638 }, { "epoch": 2.13, "learning_rate": 2.635627110405545e-05, "loss": 0.2001, "step": 20639 }, { "epoch": 2.13, "learning_rate": 2.6353827918955056e-05, "loss": 0.9093, "step": 20640 }, { "epoch": 2.13, "learning_rate": 2.6351384758399836e-05, "loss": 0.5017, "step": 20641 }, { "epoch": 2.13, "learning_rate": 2.6348941622406247e-05, "loss": 0.241, "step": 20642 }, { "epoch": 2.13, "learning_rate": 2.6346498510990725e-05, "loss": 0.3424, "step": 20643 }, { "epoch": 2.13, "learning_rate": 2.634405542416973e-05, "loss": 0.4635, "step": 20644 }, { "epoch": 2.13, "learning_rate": 2.6341612361959687e-05, "loss": 0.1602, "step": 20645 }, { "epoch": 2.13, "learning_rate": 2.6339169324377053e-05, "loss": 0.3563, "step": 20646 }, { "epoch": 2.13, "learning_rate": 2.6336726311438276e-05, "loss": 0.275, "step": 20647 }, { "epoch": 2.13, "learning_rate": 2.6334283323159803e-05, "loss": 0.3405, "step": 20648 }, { "epoch": 2.13, "learning_rate": 2.6331840359558077e-05, "loss": 0.534, "step": 20649 }, { "epoch": 2.13, "learning_rate": 2.632939742064954e-05, "loss": 0.3692, "step": 20650 }, { "epoch": 2.13, "learning_rate": 2.6326954506450636e-05, "loss": 1.1236, "step": 20651 }, { "epoch": 2.13, "learning_rate": 2.6324511616977828e-05, "loss": 0.3696, "step": 20652 }, { "epoch": 2.13, "learning_rate": 2.6322068752247537e-05, "loss": 0.3133, "step": 20653 }, { "epoch": 2.13, "learning_rate": 2.6319625912276222e-05, "loss": 0.4511, "step": 20654 }, { "epoch": 2.13, "learning_rate": 2.6317183097080323e-05, "loss": 0.752, "step": 20655 }, { "epoch": 2.13, "learning_rate": 2.631474030667629e-05, "loss": 0.6351, "step": 20656 }, { "epoch": 2.13, "learning_rate": 2.6312297541080568e-05, "loss": 0.687, "step": 20657 }, { "epoch": 2.13, "learning_rate": 2.6309854800309595e-05, "loss": 0.3309, "step": 20658 }, { "epoch": 2.13, "learning_rate": 2.6307412084379804e-05, "loss": 0.2058, "step": 20659 }, { "epoch": 2.13, "learning_rate": 2.6304969393307655e-05, "loss": 0.5465, "step": 20660 }, { "epoch": 2.13, "learning_rate": 2.6302526727109588e-05, "loss": 0.4018, "step": 20661 }, { "epoch": 2.13, "learning_rate": 2.630008408580205e-05, "loss": 0.1964, "step": 20662 }, { "epoch": 2.13, "learning_rate": 2.6297641469401474e-05, "loss": 0.5499, "step": 20663 }, { "epoch": 2.13, "learning_rate": 2.629519887792431e-05, "loss": 0.3963, "step": 20664 }, { "epoch": 2.13, "learning_rate": 2.6292756311387e-05, "loss": 0.2968, "step": 20665 }, { "epoch": 2.13, "learning_rate": 2.6290313769805996e-05, "loss": 0.321, "step": 20666 }, { "epoch": 2.13, "learning_rate": 2.6287871253197725e-05, "loss": 0.2431, "step": 20667 }, { "epoch": 2.13, "learning_rate": 2.6285428761578633e-05, "loss": 0.2795, "step": 20668 }, { "epoch": 2.13, "learning_rate": 2.628298629496518e-05, "loss": 0.3342, "step": 20669 }, { "epoch": 2.13, "learning_rate": 2.628054385337378e-05, "loss": 0.4799, "step": 20670 }, { "epoch": 2.13, "learning_rate": 2.627810143682089e-05, "loss": 0.2014, "step": 20671 }, { "epoch": 2.13, "learning_rate": 2.6275659045322955e-05, "loss": 0.1602, "step": 20672 }, { "epoch": 2.13, "learning_rate": 2.6273216678896407e-05, "loss": 0.2211, "step": 20673 }, { "epoch": 2.13, "learning_rate": 2.627077433755769e-05, "loss": 0.6423, "step": 20674 }, { "epoch": 2.13, "learning_rate": 2.626833202132325e-05, "loss": 0.2383, "step": 20675 }, { "epoch": 2.13, "learning_rate": 2.626588973020953e-05, "loss": 0.3977, "step": 20676 }, { "epoch": 2.13, "learning_rate": 2.6263447464232963e-05, "loss": 0.6227, "step": 20677 }, { "epoch": 2.13, "learning_rate": 2.626100522340999e-05, "loss": 0.7121, "step": 20678 }, { "epoch": 2.13, "learning_rate": 2.625856300775706e-05, "loss": 0.2268, "step": 20679 }, { "epoch": 2.13, "learning_rate": 2.6256120817290607e-05, "loss": 0.3485, "step": 20680 }, { "epoch": 2.13, "learning_rate": 2.6253678652027076e-05, "loss": 0.0674, "step": 20681 }, { "epoch": 2.13, "learning_rate": 2.6251236511982908e-05, "loss": 0.6897, "step": 20682 }, { "epoch": 2.13, "learning_rate": 2.624879439717453e-05, "loss": 0.4634, "step": 20683 }, { "epoch": 2.13, "learning_rate": 2.624635230761839e-05, "loss": 1.0506, "step": 20684 }, { "epoch": 2.13, "learning_rate": 2.6243910243330928e-05, "loss": 0.3527, "step": 20685 }, { "epoch": 2.13, "learning_rate": 2.624146820432859e-05, "loss": 0.5301, "step": 20686 }, { "epoch": 2.13, "learning_rate": 2.6239026190627802e-05, "loss": 0.044, "step": 20687 }, { "epoch": 2.13, "learning_rate": 2.6236584202245008e-05, "loss": 0.8374, "step": 20688 }, { "epoch": 2.13, "learning_rate": 2.623414223919665e-05, "loss": 0.1524, "step": 20689 }, { "epoch": 2.13, "learning_rate": 2.6231700301499174e-05, "loss": 0.6113, "step": 20690 }, { "epoch": 2.13, "learning_rate": 2.6229258389169e-05, "loss": 0.2446, "step": 20691 }, { "epoch": 2.13, "learning_rate": 2.6226816502222577e-05, "loss": 0.4109, "step": 20692 }, { "epoch": 2.13, "learning_rate": 2.6224374640676343e-05, "loss": 0.2881, "step": 20693 }, { "epoch": 2.13, "learning_rate": 2.622193280454675e-05, "loss": 0.2419, "step": 20694 }, { "epoch": 2.13, "learning_rate": 2.6219490993850206e-05, "loss": 0.3369, "step": 20695 }, { "epoch": 2.13, "learning_rate": 2.621704920860317e-05, "loss": 0.2543, "step": 20696 }, { "epoch": 2.13, "learning_rate": 2.6214607448822068e-05, "loss": 0.2427, "step": 20697 }, { "epoch": 2.14, "learning_rate": 2.621216571452334e-05, "loss": 0.246, "step": 20698 }, { "epoch": 2.14, "learning_rate": 2.6209724005723432e-05, "loss": 0.086, "step": 20699 }, { "epoch": 2.14, "learning_rate": 2.6207282322438778e-05, "loss": 0.2113, "step": 20700 }, { "epoch": 2.14, "learning_rate": 2.6204840664685808e-05, "loss": 0.5005, "step": 20701 }, { "epoch": 2.14, "learning_rate": 2.620239903248096e-05, "loss": 0.1745, "step": 20702 }, { "epoch": 2.14, "learning_rate": 2.6199957425840673e-05, "loss": 0.3046, "step": 20703 }, { "epoch": 2.14, "learning_rate": 2.6197515844781394e-05, "loss": 0.1552, "step": 20704 }, { "epoch": 2.14, "learning_rate": 2.619507428931954e-05, "loss": 0.4322, "step": 20705 }, { "epoch": 2.14, "learning_rate": 2.6192632759471555e-05, "loss": 0.1963, "step": 20706 }, { "epoch": 2.14, "learning_rate": 2.6190191255253888e-05, "loss": 0.3177, "step": 20707 }, { "epoch": 2.14, "learning_rate": 2.618774977668295e-05, "loss": 0.2658, "step": 20708 }, { "epoch": 2.14, "learning_rate": 2.6185308323775186e-05, "loss": 0.5252, "step": 20709 }, { "epoch": 2.14, "learning_rate": 2.618286689654704e-05, "loss": 0.2652, "step": 20710 }, { "epoch": 2.14, "learning_rate": 2.618042549501494e-05, "loss": 0.4819, "step": 20711 }, { "epoch": 2.14, "learning_rate": 2.617798411919532e-05, "loss": 0.0831, "step": 20712 }, { "epoch": 2.14, "learning_rate": 2.6175542769104614e-05, "loss": 1.237, "step": 20713 }, { "epoch": 2.14, "learning_rate": 2.617310144475927e-05, "loss": 0.1404, "step": 20714 }, { "epoch": 2.14, "learning_rate": 2.61706601461757e-05, "loss": 1.3673, "step": 20715 }, { "epoch": 2.14, "learning_rate": 2.616821887337035e-05, "loss": 0.4711, "step": 20716 }, { "epoch": 2.14, "learning_rate": 2.6165777626359655e-05, "loss": 0.6434, "step": 20717 }, { "epoch": 2.14, "learning_rate": 2.616333640516006e-05, "loss": 0.3845, "step": 20718 }, { "epoch": 2.14, "learning_rate": 2.6160895209787972e-05, "loss": 0.7645, "step": 20719 }, { "epoch": 2.14, "learning_rate": 2.6158454040259857e-05, "loss": 0.6557, "step": 20720 }, { "epoch": 2.14, "learning_rate": 2.6156012896592114e-05, "loss": 0.7203, "step": 20721 }, { "epoch": 2.14, "learning_rate": 2.6153571778801198e-05, "loss": 0.2885, "step": 20722 }, { "epoch": 2.14, "learning_rate": 2.615113068690353e-05, "loss": 0.3916, "step": 20723 }, { "epoch": 2.14, "learning_rate": 2.614868962091556e-05, "loss": 0.1322, "step": 20724 }, { "epoch": 2.14, "learning_rate": 2.6146248580853706e-05, "loss": 0.2174, "step": 20725 }, { "epoch": 2.14, "learning_rate": 2.6143807566734405e-05, "loss": 0.5969, "step": 20726 }, { "epoch": 2.14, "learning_rate": 2.614136657857408e-05, "loss": 0.6505, "step": 20727 }, { "epoch": 2.14, "learning_rate": 2.613892561638919e-05, "loss": 0.3525, "step": 20728 }, { "epoch": 2.14, "learning_rate": 2.6136484680196136e-05, "loss": 0.1154, "step": 20729 }, { "epoch": 2.14, "learning_rate": 2.6134043770011366e-05, "loss": 0.2919, "step": 20730 }, { "epoch": 2.14, "learning_rate": 2.613160288585131e-05, "loss": 0.6046, "step": 20731 }, { "epoch": 2.14, "learning_rate": 2.6129162027732407e-05, "loss": 0.412, "step": 20732 }, { "epoch": 2.14, "learning_rate": 2.612672119567107e-05, "loss": 0.35, "step": 20733 }, { "epoch": 2.14, "learning_rate": 2.6124280389683747e-05, "loss": 0.2422, "step": 20734 }, { "epoch": 2.14, "learning_rate": 2.612183960978685e-05, "loss": 0.1827, "step": 20735 }, { "epoch": 2.14, "learning_rate": 2.6119398855996826e-05, "loss": 0.2237, "step": 20736 }, { "epoch": 2.14, "learning_rate": 2.6116958128330102e-05, "loss": 0.6514, "step": 20737 }, { "epoch": 2.14, "learning_rate": 2.611451742680311e-05, "loss": 0.4936, "step": 20738 }, { "epoch": 2.14, "learning_rate": 2.6112076751432273e-05, "loss": 0.4324, "step": 20739 }, { "epoch": 2.14, "learning_rate": 2.6109636102234024e-05, "loss": 0.3745, "step": 20740 }, { "epoch": 2.14, "learning_rate": 2.6107195479224797e-05, "loss": 0.3371, "step": 20741 }, { "epoch": 2.14, "learning_rate": 2.6104754882421024e-05, "loss": 0.1572, "step": 20742 }, { "epoch": 2.14, "learning_rate": 2.6102314311839124e-05, "loss": 0.4149, "step": 20743 }, { "epoch": 2.14, "learning_rate": 2.6099873767495527e-05, "loss": 0.9398, "step": 20744 }, { "epoch": 2.14, "learning_rate": 2.609743324940669e-05, "loss": 0.5835, "step": 20745 }, { "epoch": 2.14, "learning_rate": 2.6094992757589e-05, "loss": 0.3395, "step": 20746 }, { "epoch": 2.14, "learning_rate": 2.6092552292058905e-05, "loss": 0.5358, "step": 20747 }, { "epoch": 2.14, "learning_rate": 2.6090111852832842e-05, "loss": 0.2105, "step": 20748 }, { "epoch": 2.14, "learning_rate": 2.6087671439927224e-05, "loss": 0.6693, "step": 20749 }, { "epoch": 2.14, "learning_rate": 2.6085231053358485e-05, "loss": 0.3965, "step": 20750 }, { "epoch": 2.14, "learning_rate": 2.6082790693143056e-05, "loss": 0.2559, "step": 20751 }, { "epoch": 2.14, "learning_rate": 2.608035035929737e-05, "loss": 0.1247, "step": 20752 }, { "epoch": 2.14, "learning_rate": 2.607791005183784e-05, "loss": 0.1457, "step": 20753 }, { "epoch": 2.14, "learning_rate": 2.6075469770780907e-05, "loss": 0.5755, "step": 20754 }, { "epoch": 2.14, "learning_rate": 2.6073029516142988e-05, "loss": 0.1443, "step": 20755 }, { "epoch": 2.14, "learning_rate": 2.6070589287940523e-05, "loss": 0.6626, "step": 20756 }, { "epoch": 2.14, "learning_rate": 2.6068149086189932e-05, "loss": 0.1457, "step": 20757 }, { "epoch": 2.14, "learning_rate": 2.606570891090764e-05, "loss": 0.1782, "step": 20758 }, { "epoch": 2.14, "learning_rate": 2.6063268762110072e-05, "loss": 0.3173, "step": 20759 }, { "epoch": 2.14, "learning_rate": 2.6060828639813655e-05, "loss": 0.3749, "step": 20760 }, { "epoch": 2.14, "learning_rate": 2.6058388544034817e-05, "loss": 0.0658, "step": 20761 }, { "epoch": 2.14, "learning_rate": 2.6055948474789997e-05, "loss": 0.164, "step": 20762 }, { "epoch": 2.14, "learning_rate": 2.60535084320956e-05, "loss": 0.1237, "step": 20763 }, { "epoch": 2.14, "learning_rate": 2.6051068415968056e-05, "loss": 0.1186, "step": 20764 }, { "epoch": 2.14, "learning_rate": 2.60486284264238e-05, "loss": 0.4242, "step": 20765 }, { "epoch": 2.14, "learning_rate": 2.604618846347926e-05, "loss": 0.0638, "step": 20766 }, { "epoch": 2.14, "learning_rate": 2.6043748527150844e-05, "loss": 0.3502, "step": 20767 }, { "epoch": 2.14, "learning_rate": 2.604130861745499e-05, "loss": 0.4294, "step": 20768 }, { "epoch": 2.14, "learning_rate": 2.6038868734408118e-05, "loss": 0.6435, "step": 20769 }, { "epoch": 2.14, "learning_rate": 2.603642887802667e-05, "loss": 0.1653, "step": 20770 }, { "epoch": 2.14, "learning_rate": 2.6033989048327043e-05, "loss": 0.1296, "step": 20771 }, { "epoch": 2.14, "learning_rate": 2.6031549245325675e-05, "loss": 0.3016, "step": 20772 }, { "epoch": 2.14, "learning_rate": 2.6029109469038988e-05, "loss": 0.3869, "step": 20773 }, { "epoch": 2.14, "learning_rate": 2.6026669719483407e-05, "loss": 0.3862, "step": 20774 }, { "epoch": 2.14, "learning_rate": 2.602422999667535e-05, "loss": 0.0653, "step": 20775 }, { "epoch": 2.14, "learning_rate": 2.602179030063126e-05, "loss": 0.1934, "step": 20776 }, { "epoch": 2.14, "learning_rate": 2.6019350631367537e-05, "loss": 0.1749, "step": 20777 }, { "epoch": 2.14, "learning_rate": 2.6016910988900614e-05, "loss": 0.7311, "step": 20778 }, { "epoch": 2.14, "learning_rate": 2.6014471373246915e-05, "loss": 0.1436, "step": 20779 }, { "epoch": 2.14, "learning_rate": 2.6012031784422866e-05, "loss": 0.4369, "step": 20780 }, { "epoch": 2.14, "learning_rate": 2.6009592222444886e-05, "loss": 0.3539, "step": 20781 }, { "epoch": 2.14, "learning_rate": 2.6007152687329404e-05, "loss": 0.2851, "step": 20782 }, { "epoch": 2.14, "learning_rate": 2.6004713179092822e-05, "loss": 0.2559, "step": 20783 }, { "epoch": 2.14, "learning_rate": 2.6002273697751582e-05, "loss": 0.2206, "step": 20784 }, { "epoch": 2.14, "learning_rate": 2.5999834243322097e-05, "loss": 0.4907, "step": 20785 }, { "epoch": 2.14, "learning_rate": 2.59973948158208e-05, "loss": 0.3158, "step": 20786 }, { "epoch": 2.14, "learning_rate": 2.5994955415264097e-05, "loss": 0.3153, "step": 20787 }, { "epoch": 2.14, "learning_rate": 2.5992516041668416e-05, "loss": 0.3266, "step": 20788 }, { "epoch": 2.14, "learning_rate": 2.599007669505018e-05, "loss": 0.717, "step": 20789 }, { "epoch": 2.14, "learning_rate": 2.5987637375425815e-05, "loss": 0.4947, "step": 20790 }, { "epoch": 2.14, "learning_rate": 2.598519808281173e-05, "loss": 0.3743, "step": 20791 }, { "epoch": 2.14, "learning_rate": 2.5982758817224353e-05, "loss": 0.102, "step": 20792 }, { "epoch": 2.14, "learning_rate": 2.5980319578680104e-05, "loss": 0.8924, "step": 20793 }, { "epoch": 2.15, "learning_rate": 2.597788036719541e-05, "loss": 0.2187, "step": 20794 }, { "epoch": 2.15, "learning_rate": 2.5975441182786682e-05, "loss": 0.257, "step": 20795 }, { "epoch": 2.15, "learning_rate": 2.5973002025470347e-05, "loss": 0.1719, "step": 20796 }, { "epoch": 2.15, "learning_rate": 2.5970562895262808e-05, "loss": 0.5879, "step": 20797 }, { "epoch": 2.15, "learning_rate": 2.59681237921805e-05, "loss": 0.7182, "step": 20798 }, { "epoch": 2.15, "learning_rate": 2.596568471623984e-05, "loss": 0.1377, "step": 20799 }, { "epoch": 2.15, "learning_rate": 2.596324566745725e-05, "loss": 0.4475, "step": 20800 }, { "epoch": 2.15, "learning_rate": 2.5960806645849138e-05, "loss": 0.1962, "step": 20801 }, { "epoch": 2.15, "learning_rate": 2.5958367651431933e-05, "loss": 0.2932, "step": 20802 }, { "epoch": 2.15, "learning_rate": 2.5955928684222054e-05, "loss": 0.3433, "step": 20803 }, { "epoch": 2.15, "learning_rate": 2.595348974423592e-05, "loss": 0.2624, "step": 20804 }, { "epoch": 2.15, "learning_rate": 2.595105083148994e-05, "loss": 0.1135, "step": 20805 }, { "epoch": 2.15, "learning_rate": 2.594861194600054e-05, "loss": 0.3334, "step": 20806 }, { "epoch": 2.15, "learning_rate": 2.594617308778415e-05, "loss": 0.7853, "step": 20807 }, { "epoch": 2.15, "learning_rate": 2.5943734256857158e-05, "loss": 0.4418, "step": 20808 }, { "epoch": 2.15, "learning_rate": 2.5941295453235998e-05, "loss": 0.3271, "step": 20809 }, { "epoch": 2.15, "learning_rate": 2.5938856676937096e-05, "loss": 0.2629, "step": 20810 }, { "epoch": 2.15, "learning_rate": 2.5936417927976853e-05, "loss": 0.3952, "step": 20811 }, { "epoch": 2.15, "learning_rate": 2.5933979206371695e-05, "loss": 0.137, "step": 20812 }, { "epoch": 2.15, "learning_rate": 2.5931540512138038e-05, "loss": 0.2095, "step": 20813 }, { "epoch": 2.15, "learning_rate": 2.5929101845292303e-05, "loss": 0.3849, "step": 20814 }, { "epoch": 2.15, "learning_rate": 2.5926663205850896e-05, "loss": 0.0884, "step": 20815 }, { "epoch": 2.15, "learning_rate": 2.592422459383024e-05, "loss": 0.8264, "step": 20816 }, { "epoch": 2.15, "learning_rate": 2.592178600924675e-05, "loss": 0.1847, "step": 20817 }, { "epoch": 2.15, "learning_rate": 2.591934745211685e-05, "loss": 0.5369, "step": 20818 }, { "epoch": 2.15, "learning_rate": 2.5916908922456936e-05, "loss": 0.4171, "step": 20819 }, { "epoch": 2.15, "learning_rate": 2.5914470420283456e-05, "loss": 0.0599, "step": 20820 }, { "epoch": 2.15, "learning_rate": 2.5912031945612786e-05, "loss": 0.6331, "step": 20821 }, { "epoch": 2.15, "learning_rate": 2.5909593498461364e-05, "loss": 0.2048, "step": 20822 }, { "epoch": 2.15, "learning_rate": 2.5907155078845603e-05, "loss": 0.9343, "step": 20823 }, { "epoch": 2.15, "learning_rate": 2.590471668678192e-05, "loss": 0.1831, "step": 20824 }, { "epoch": 2.15, "learning_rate": 2.590227832228672e-05, "loss": 0.1718, "step": 20825 }, { "epoch": 2.15, "learning_rate": 2.5899839985376425e-05, "loss": 0.2657, "step": 20826 }, { "epoch": 2.15, "learning_rate": 2.5897401676067448e-05, "loss": 0.7245, "step": 20827 }, { "epoch": 2.15, "learning_rate": 2.5894963394376207e-05, "loss": 0.2138, "step": 20828 }, { "epoch": 2.15, "learning_rate": 2.5892525140319107e-05, "loss": 0.6953, "step": 20829 }, { "epoch": 2.15, "learning_rate": 2.5890086913912568e-05, "loss": 0.4728, "step": 20830 }, { "epoch": 2.15, "learning_rate": 2.5887648715173e-05, "loss": 0.6682, "step": 20831 }, { "epoch": 2.15, "learning_rate": 2.5885210544116826e-05, "loss": 0.3738, "step": 20832 }, { "epoch": 2.15, "learning_rate": 2.588277240076045e-05, "loss": 0.8717, "step": 20833 }, { "epoch": 2.15, "learning_rate": 2.588033428512029e-05, "loss": 0.1866, "step": 20834 }, { "epoch": 2.15, "learning_rate": 2.5877896197212747e-05, "loss": 0.3684, "step": 20835 }, { "epoch": 2.15, "learning_rate": 2.5875458137054242e-05, "loss": 0.133, "step": 20836 }, { "epoch": 2.15, "learning_rate": 2.5873020104661193e-05, "loss": 0.2966, "step": 20837 }, { "epoch": 2.15, "learning_rate": 2.5870582100050004e-05, "loss": 0.3354, "step": 20838 }, { "epoch": 2.15, "learning_rate": 2.586814412323709e-05, "loss": 0.3983, "step": 20839 }, { "epoch": 2.15, "learning_rate": 2.586570617423886e-05, "loss": 0.8203, "step": 20840 }, { "epoch": 2.15, "learning_rate": 2.5863268253071733e-05, "loss": 0.1856, "step": 20841 }, { "epoch": 2.15, "learning_rate": 2.586083035975212e-05, "loss": 0.7052, "step": 20842 }, { "epoch": 2.15, "learning_rate": 2.585839249429642e-05, "loss": 0.5749, "step": 20843 }, { "epoch": 2.15, "learning_rate": 2.5855954656721058e-05, "loss": 0.5157, "step": 20844 }, { "epoch": 2.15, "learning_rate": 2.585351684704245e-05, "loss": 0.3395, "step": 20845 }, { "epoch": 2.15, "learning_rate": 2.5851079065276983e-05, "loss": 0.522, "step": 20846 }, { "epoch": 2.15, "learning_rate": 2.584864131144108e-05, "loss": 0.2687, "step": 20847 }, { "epoch": 2.15, "learning_rate": 2.584620358555116e-05, "loss": 0.2169, "step": 20848 }, { "epoch": 2.15, "learning_rate": 2.5843765887623616e-05, "loss": 0.4072, "step": 20849 }, { "epoch": 2.15, "learning_rate": 2.584132821767487e-05, "loss": 0.345, "step": 20850 }, { "epoch": 2.15, "learning_rate": 2.5838890575721332e-05, "loss": 0.4462, "step": 20851 }, { "epoch": 2.15, "learning_rate": 2.5836452961779414e-05, "loss": 1.7612, "step": 20852 }, { "epoch": 2.15, "learning_rate": 2.5834015375865515e-05, "loss": 0.7965, "step": 20853 }, { "epoch": 2.15, "learning_rate": 2.5831577817996047e-05, "loss": 0.6317, "step": 20854 }, { "epoch": 2.15, "learning_rate": 2.582914028818742e-05, "loss": 0.3825, "step": 20855 }, { "epoch": 2.15, "learning_rate": 2.5826702786456055e-05, "loss": 0.2596, "step": 20856 }, { "epoch": 2.15, "learning_rate": 2.582426531281834e-05, "loss": 0.2547, "step": 20857 }, { "epoch": 2.15, "learning_rate": 2.5821827867290714e-05, "loss": 0.6736, "step": 20858 }, { "epoch": 2.15, "learning_rate": 2.5819390449889547e-05, "loss": 0.4135, "step": 20859 }, { "epoch": 2.15, "learning_rate": 2.5816953060631267e-05, "loss": 0.6161, "step": 20860 }, { "epoch": 2.15, "learning_rate": 2.581451569953228e-05, "loss": 0.6763, "step": 20861 }, { "epoch": 2.15, "learning_rate": 2.5812078366608998e-05, "loss": 0.1562, "step": 20862 }, { "epoch": 2.15, "learning_rate": 2.580964106187782e-05, "loss": 0.5596, "step": 20863 }, { "epoch": 2.15, "learning_rate": 2.580720378535516e-05, "loss": 0.4828, "step": 20864 }, { "epoch": 2.15, "learning_rate": 2.580476653705742e-05, "loss": 0.3557, "step": 20865 }, { "epoch": 2.15, "learning_rate": 2.5802329317001015e-05, "loss": 0.0644, "step": 20866 }, { "epoch": 2.15, "learning_rate": 2.5799892125202347e-05, "loss": 0.1807, "step": 20867 }, { "epoch": 2.15, "learning_rate": 2.5797454961677816e-05, "loss": 0.4232, "step": 20868 }, { "epoch": 2.15, "learning_rate": 2.579501782644384e-05, "loss": 0.1695, "step": 20869 }, { "epoch": 2.15, "learning_rate": 2.579258071951683e-05, "loss": 0.1228, "step": 20870 }, { "epoch": 2.15, "learning_rate": 2.5790143640913168e-05, "loss": 0.3993, "step": 20871 }, { "epoch": 2.15, "learning_rate": 2.578770659064928e-05, "loss": 0.4846, "step": 20872 }, { "epoch": 2.15, "learning_rate": 2.578526956874156e-05, "loss": 0.3996, "step": 20873 }, { "epoch": 2.15, "learning_rate": 2.5782832575206424e-05, "loss": 0.6072, "step": 20874 }, { "epoch": 2.15, "learning_rate": 2.5780395610060265e-05, "loss": 0.037, "step": 20875 }, { "epoch": 2.15, "learning_rate": 2.5777958673319508e-05, "loss": 0.5191, "step": 20876 }, { "epoch": 2.15, "learning_rate": 2.5775521765000535e-05, "loss": 0.221, "step": 20877 }, { "epoch": 2.15, "learning_rate": 2.5773084885119762e-05, "loss": 0.154, "step": 20878 }, { "epoch": 2.15, "learning_rate": 2.5770648033693593e-05, "loss": 0.4668, "step": 20879 }, { "epoch": 2.15, "learning_rate": 2.5768211210738438e-05, "loss": 0.5047, "step": 20880 }, { "epoch": 2.15, "learning_rate": 2.576577441627069e-05, "loss": 0.2773, "step": 20881 }, { "epoch": 2.15, "learning_rate": 2.5763337650306756e-05, "loss": 0.1906, "step": 20882 }, { "epoch": 2.15, "learning_rate": 2.5760900912863056e-05, "loss": 0.1639, "step": 20883 }, { "epoch": 2.15, "learning_rate": 2.575846420395596e-05, "loss": 0.1998, "step": 20884 }, { "epoch": 2.15, "learning_rate": 2.57560275236019e-05, "loss": 0.6741, "step": 20885 }, { "epoch": 2.15, "learning_rate": 2.5753590871817273e-05, "loss": 0.4458, "step": 20886 }, { "epoch": 2.15, "learning_rate": 2.575115424861847e-05, "loss": 0.1315, "step": 20887 }, { "epoch": 2.15, "learning_rate": 2.5748717654021904e-05, "loss": 0.4953, "step": 20888 }, { "epoch": 2.15, "learning_rate": 2.5746281088043978e-05, "loss": 0.3899, "step": 20889 }, { "epoch": 2.16, "learning_rate": 2.5743844550701096e-05, "loss": 0.0838, "step": 20890 }, { "epoch": 2.16, "learning_rate": 2.5741408042009652e-05, "loss": 0.2619, "step": 20891 }, { "epoch": 2.16, "learning_rate": 2.573897156198605e-05, "loss": 0.0661, "step": 20892 }, { "epoch": 2.16, "learning_rate": 2.5736535110646697e-05, "loss": 0.3668, "step": 20893 }, { "epoch": 2.16, "learning_rate": 2.5734098688007996e-05, "loss": 0.3054, "step": 20894 }, { "epoch": 2.16, "learning_rate": 2.5731662294086345e-05, "loss": 0.3043, "step": 20895 }, { "epoch": 2.16, "learning_rate": 2.572922592889815e-05, "loss": 0.5785, "step": 20896 }, { "epoch": 2.16, "learning_rate": 2.5726789592459793e-05, "loss": 0.1641, "step": 20897 }, { "epoch": 2.16, "learning_rate": 2.5724353284787688e-05, "loss": 0.2019, "step": 20898 }, { "epoch": 2.16, "learning_rate": 2.572191700589824e-05, "loss": 0.8721, "step": 20899 }, { "epoch": 2.16, "learning_rate": 2.5719480755807857e-05, "loss": 0.4938, "step": 20900 }, { "epoch": 2.16, "learning_rate": 2.5717044534532913e-05, "loss": 0.3382, "step": 20901 }, { "epoch": 2.16, "learning_rate": 2.571460834208983e-05, "loss": 0.5285, "step": 20902 }, { "epoch": 2.16, "learning_rate": 2.5712172178494996e-05, "loss": 0.4198, "step": 20903 }, { "epoch": 2.16, "learning_rate": 2.5709736043764824e-05, "loss": 0.2071, "step": 20904 }, { "epoch": 2.16, "learning_rate": 2.5707299937915697e-05, "loss": 0.6213, "step": 20905 }, { "epoch": 2.16, "learning_rate": 2.5704863860964024e-05, "loss": 0.2207, "step": 20906 }, { "epoch": 2.16, "learning_rate": 2.5702427812926204e-05, "loss": 0.4686, "step": 20907 }, { "epoch": 2.16, "learning_rate": 2.5699991793818644e-05, "loss": 0.1961, "step": 20908 }, { "epoch": 2.16, "learning_rate": 2.5697555803657725e-05, "loss": 0.5572, "step": 20909 }, { "epoch": 2.16, "learning_rate": 2.5695119842459853e-05, "loss": 0.3973, "step": 20910 }, { "epoch": 2.16, "learning_rate": 2.5692683910241425e-05, "loss": 0.9686, "step": 20911 }, { "epoch": 2.16, "learning_rate": 2.5690248007018842e-05, "loss": 0.0645, "step": 20912 }, { "epoch": 2.16, "learning_rate": 2.56878121328085e-05, "loss": 0.3349, "step": 20913 }, { "epoch": 2.16, "learning_rate": 2.56853762876268e-05, "loss": 0.1908, "step": 20914 }, { "epoch": 2.16, "learning_rate": 2.5682940471490137e-05, "loss": 0.1659, "step": 20915 }, { "epoch": 2.16, "learning_rate": 2.5680504684414907e-05, "loss": 0.8467, "step": 20916 }, { "epoch": 2.16, "learning_rate": 2.5678068926417505e-05, "loss": 0.3236, "step": 20917 }, { "epoch": 2.16, "learning_rate": 2.5675633197514343e-05, "loss": 0.2376, "step": 20918 }, { "epoch": 2.16, "learning_rate": 2.5673197497721796e-05, "loss": 0.37, "step": 20919 }, { "epoch": 2.16, "learning_rate": 2.5670761827056273e-05, "loss": 0.2567, "step": 20920 }, { "epoch": 2.16, "learning_rate": 2.5668326185534182e-05, "loss": 0.4087, "step": 20921 }, { "epoch": 2.16, "learning_rate": 2.566589057317189e-05, "loss": 0.5971, "step": 20922 }, { "epoch": 2.16, "learning_rate": 2.5663454989985813e-05, "loss": 0.2578, "step": 20923 }, { "epoch": 2.16, "learning_rate": 2.5661019435992346e-05, "loss": 0.5994, "step": 20924 }, { "epoch": 2.16, "learning_rate": 2.565858391120788e-05, "loss": 0.2218, "step": 20925 }, { "epoch": 2.16, "learning_rate": 2.5656148415648803e-05, "loss": 0.0604, "step": 20926 }, { "epoch": 2.16, "learning_rate": 2.5653712949331523e-05, "loss": 0.1784, "step": 20927 }, { "epoch": 2.16, "learning_rate": 2.5651277512272435e-05, "loss": 0.489, "step": 20928 }, { "epoch": 2.16, "learning_rate": 2.5648842104487923e-05, "loss": 0.5451, "step": 20929 }, { "epoch": 2.16, "learning_rate": 2.5646406725994387e-05, "loss": 0.4826, "step": 20930 }, { "epoch": 2.16, "learning_rate": 2.5643971376808225e-05, "loss": 0.0875, "step": 20931 }, { "epoch": 2.16, "learning_rate": 2.5641536056945832e-05, "loss": 0.5545, "step": 20932 }, { "epoch": 2.16, "learning_rate": 2.56391007664236e-05, "loss": 0.5593, "step": 20933 }, { "epoch": 2.16, "learning_rate": 2.5636665505257923e-05, "loss": 0.1629, "step": 20934 }, { "epoch": 2.16, "learning_rate": 2.5634230273465185e-05, "loss": 0.164, "step": 20935 }, { "epoch": 2.16, "learning_rate": 2.5631795071061786e-05, "loss": 0.4666, "step": 20936 }, { "epoch": 2.16, "learning_rate": 2.5629359898064124e-05, "loss": 0.1573, "step": 20937 }, { "epoch": 2.16, "learning_rate": 2.562692475448859e-05, "loss": 0.1332, "step": 20938 }, { "epoch": 2.16, "learning_rate": 2.5624489640351573e-05, "loss": 0.1433, "step": 20939 }, { "epoch": 2.16, "learning_rate": 2.5622054555669464e-05, "loss": 0.7552, "step": 20940 }, { "epoch": 2.16, "learning_rate": 2.5619619500458663e-05, "loss": 0.1502, "step": 20941 }, { "epoch": 2.16, "learning_rate": 2.5617184474735565e-05, "loss": 0.8036, "step": 20942 }, { "epoch": 2.16, "learning_rate": 2.5614749478516547e-05, "loss": 0.1859, "step": 20943 }, { "epoch": 2.16, "learning_rate": 2.561231451181801e-05, "loss": 0.6541, "step": 20944 }, { "epoch": 2.16, "learning_rate": 2.5609879574656346e-05, "loss": 0.461, "step": 20945 }, { "epoch": 2.16, "learning_rate": 2.560744466704796e-05, "loss": 0.3771, "step": 20946 }, { "epoch": 2.16, "learning_rate": 2.5605009789009213e-05, "loss": 0.278, "step": 20947 }, { "epoch": 2.16, "learning_rate": 2.5602574940556515e-05, "loss": 0.4365, "step": 20948 }, { "epoch": 2.16, "learning_rate": 2.5600140121706253e-05, "loss": 0.4425, "step": 20949 }, { "epoch": 2.16, "learning_rate": 2.5597705332474815e-05, "loss": 0.2981, "step": 20950 }, { "epoch": 2.16, "learning_rate": 2.5595270572878594e-05, "loss": 0.2138, "step": 20951 }, { "epoch": 2.16, "learning_rate": 2.559283584293399e-05, "loss": 0.2669, "step": 20952 }, { "epoch": 2.16, "learning_rate": 2.5590401142657378e-05, "loss": 0.4485, "step": 20953 }, { "epoch": 2.16, "learning_rate": 2.5587966472065153e-05, "loss": 0.5395, "step": 20954 }, { "epoch": 2.16, "learning_rate": 2.5585531831173704e-05, "loss": 0.1993, "step": 20955 }, { "epoch": 2.16, "learning_rate": 2.558309721999943e-05, "loss": 0.298, "step": 20956 }, { "epoch": 2.16, "learning_rate": 2.5580662638558705e-05, "loss": 0.5631, "step": 20957 }, { "epoch": 2.16, "learning_rate": 2.5578228086867937e-05, "loss": 0.3446, "step": 20958 }, { "epoch": 2.16, "learning_rate": 2.5575793564943492e-05, "loss": 0.493, "step": 20959 }, { "epoch": 2.16, "learning_rate": 2.5573359072801766e-05, "loss": 0.4048, "step": 20960 }, { "epoch": 2.16, "learning_rate": 2.5570924610459154e-05, "loss": 0.6095, "step": 20961 }, { "epoch": 2.16, "learning_rate": 2.5568490177932047e-05, "loss": 0.32, "step": 20962 }, { "epoch": 2.16, "learning_rate": 2.556605577523682e-05, "loss": 0.5782, "step": 20963 }, { "epoch": 2.16, "learning_rate": 2.5563621402389867e-05, "loss": 0.1627, "step": 20964 }, { "epoch": 2.16, "learning_rate": 2.556118705940758e-05, "loss": 0.1212, "step": 20965 }, { "epoch": 2.16, "learning_rate": 2.5558752746306346e-05, "loss": 0.5077, "step": 20966 }, { "epoch": 2.16, "learning_rate": 2.5556318463102545e-05, "loss": 0.3894, "step": 20967 }, { "epoch": 2.16, "learning_rate": 2.555388420981257e-05, "loss": 0.6689, "step": 20968 }, { "epoch": 2.16, "learning_rate": 2.5551449986452803e-05, "loss": 0.1542, "step": 20969 }, { "epoch": 2.16, "learning_rate": 2.5549015793039645e-05, "loss": 0.2818, "step": 20970 }, { "epoch": 2.16, "learning_rate": 2.5546581629589468e-05, "loss": 0.6044, "step": 20971 }, { "epoch": 2.16, "learning_rate": 2.554414749611866e-05, "loss": 0.7607, "step": 20972 }, { "epoch": 2.16, "learning_rate": 2.5541713392643605e-05, "loss": 0.5401, "step": 20973 }, { "epoch": 2.16, "learning_rate": 2.5539279319180687e-05, "loss": 0.9476, "step": 20974 }, { "epoch": 2.16, "learning_rate": 2.5536845275746303e-05, "loss": 0.1086, "step": 20975 }, { "epoch": 2.16, "learning_rate": 2.5534411262356835e-05, "loss": 0.3727, "step": 20976 }, { "epoch": 2.16, "learning_rate": 2.553197727902866e-05, "loss": 0.1272, "step": 20977 }, { "epoch": 2.16, "learning_rate": 2.552954332577817e-05, "loss": 0.0702, "step": 20978 }, { "epoch": 2.16, "learning_rate": 2.5527109402621746e-05, "loss": 0.9067, "step": 20979 }, { "epoch": 2.16, "learning_rate": 2.5524675509575785e-05, "loss": 0.5685, "step": 20980 }, { "epoch": 2.16, "learning_rate": 2.552224164665665e-05, "loss": 0.2595, "step": 20981 }, { "epoch": 2.16, "learning_rate": 2.551980781388074e-05, "loss": 0.2034, "step": 20982 }, { "epoch": 2.16, "learning_rate": 2.5517374011264447e-05, "loss": 0.602, "step": 20983 }, { "epoch": 2.16, "learning_rate": 2.551494023882413e-05, "loss": 0.1757, "step": 20984 }, { "epoch": 2.17, "learning_rate": 2.551250649657619e-05, "loss": 0.2953, "step": 20985 }, { "epoch": 2.17, "learning_rate": 2.5510072784537005e-05, "loss": 0.1519, "step": 20986 }, { "epoch": 2.17, "learning_rate": 2.5507639102722958e-05, "loss": 0.1752, "step": 20987 }, { "epoch": 2.17, "learning_rate": 2.5505205451150434e-05, "loss": 0.2529, "step": 20988 }, { "epoch": 2.17, "learning_rate": 2.5502771829835813e-05, "loss": 0.3504, "step": 20989 }, { "epoch": 2.17, "learning_rate": 2.550033823879549e-05, "loss": 0.4178, "step": 20990 }, { "epoch": 2.17, "learning_rate": 2.549790467804583e-05, "loss": 0.2907, "step": 20991 }, { "epoch": 2.17, "learning_rate": 2.5495471147603217e-05, "loss": 0.3129, "step": 20992 }, { "epoch": 2.17, "learning_rate": 2.5493037647484042e-05, "loss": 0.7003, "step": 20993 }, { "epoch": 2.17, "learning_rate": 2.5490604177704687e-05, "loss": 0.1829, "step": 20994 }, { "epoch": 2.17, "learning_rate": 2.5488170738281526e-05, "loss": 0.457, "step": 20995 }, { "epoch": 2.17, "learning_rate": 2.5485737329230956e-05, "loss": 0.4692, "step": 20996 }, { "epoch": 2.17, "learning_rate": 2.5483303950569334e-05, "loss": 0.319, "step": 20997 }, { "epoch": 2.17, "learning_rate": 2.548087060231305e-05, "loss": 0.4228, "step": 20998 }, { "epoch": 2.17, "learning_rate": 2.547843728447849e-05, "loss": 0.5297, "step": 20999 }, { "epoch": 2.17, "learning_rate": 2.5476003997082038e-05, "loss": 0.2126, "step": 21000 }, { "epoch": 2.17, "learning_rate": 2.5473570740140063e-05, "loss": 0.1743, "step": 21001 }, { "epoch": 2.17, "learning_rate": 2.547113751366895e-05, "loss": 0.6406, "step": 21002 }, { "epoch": 2.17, "learning_rate": 2.546870431768508e-05, "loss": 0.3578, "step": 21003 }, { "epoch": 2.17, "learning_rate": 2.546627115220484e-05, "loss": 0.0762, "step": 21004 }, { "epoch": 2.17, "learning_rate": 2.546383801724459e-05, "loss": 1.1182, "step": 21005 }, { "epoch": 2.17, "learning_rate": 2.5461404912820727e-05, "loss": 0.5151, "step": 21006 }, { "epoch": 2.17, "learning_rate": 2.5458971838949622e-05, "loss": 0.1785, "step": 21007 }, { "epoch": 2.17, "learning_rate": 2.545653879564767e-05, "loss": 0.0686, "step": 21008 }, { "epoch": 2.17, "learning_rate": 2.5454105782931223e-05, "loss": 0.4306, "step": 21009 }, { "epoch": 2.17, "learning_rate": 2.5451672800816674e-05, "loss": 0.8639, "step": 21010 }, { "epoch": 2.17, "learning_rate": 2.54492398493204e-05, "loss": 0.378, "step": 21011 }, { "epoch": 2.17, "learning_rate": 2.5446806928458775e-05, "loss": 0.6123, "step": 21012 }, { "epoch": 2.17, "learning_rate": 2.544437403824818e-05, "loss": 0.7559, "step": 21013 }, { "epoch": 2.17, "learning_rate": 2.5441941178705002e-05, "loss": 0.6982, "step": 21014 }, { "epoch": 2.17, "learning_rate": 2.5439508349845598e-05, "loss": 0.4271, "step": 21015 }, { "epoch": 2.17, "learning_rate": 2.543707555168636e-05, "loss": 0.1695, "step": 21016 }, { "epoch": 2.17, "learning_rate": 2.5434642784243662e-05, "loss": 0.1957, "step": 21017 }, { "epoch": 2.17, "learning_rate": 2.5432210047533887e-05, "loss": 0.3193, "step": 21018 }, { "epoch": 2.17, "learning_rate": 2.5429777341573397e-05, "loss": 0.603, "step": 21019 }, { "epoch": 2.17, "learning_rate": 2.542734466637858e-05, "loss": 0.1095, "step": 21020 }, { "epoch": 2.17, "learning_rate": 2.542491202196582e-05, "loss": 0.2951, "step": 21021 }, { "epoch": 2.17, "learning_rate": 2.542247940835147e-05, "loss": 0.7136, "step": 21022 }, { "epoch": 2.17, "learning_rate": 2.5420046825551914e-05, "loss": 0.2974, "step": 21023 }, { "epoch": 2.17, "learning_rate": 2.5417614273583538e-05, "loss": 0.5879, "step": 21024 }, { "epoch": 2.17, "learning_rate": 2.5415181752462706e-05, "loss": 0.3457, "step": 21025 }, { "epoch": 2.17, "learning_rate": 2.54127492622058e-05, "loss": 0.0521, "step": 21026 }, { "epoch": 2.17, "learning_rate": 2.541031680282919e-05, "loss": 0.1604, "step": 21027 }, { "epoch": 2.17, "learning_rate": 2.540788437434926e-05, "loss": 0.5207, "step": 21028 }, { "epoch": 2.17, "learning_rate": 2.5405451976782367e-05, "loss": 0.1004, "step": 21029 }, { "epoch": 2.17, "learning_rate": 2.5403019610144902e-05, "loss": 0.4843, "step": 21030 }, { "epoch": 2.17, "learning_rate": 2.5400587274453234e-05, "loss": 0.4249, "step": 21031 }, { "epoch": 2.17, "learning_rate": 2.5398154969723743e-05, "loss": 0.3022, "step": 21032 }, { "epoch": 2.17, "learning_rate": 2.5395722695972785e-05, "loss": 0.2643, "step": 21033 }, { "epoch": 2.17, "learning_rate": 2.539329045321676e-05, "loss": 0.4126, "step": 21034 }, { "epoch": 2.17, "learning_rate": 2.5390858241472013e-05, "loss": 0.2764, "step": 21035 }, { "epoch": 2.17, "learning_rate": 2.5388426060754926e-05, "loss": 0.4249, "step": 21036 }, { "epoch": 2.17, "learning_rate": 2.538599391108188e-05, "loss": 1.0103, "step": 21037 }, { "epoch": 2.17, "learning_rate": 2.5383561792469252e-05, "loss": 0.1199, "step": 21038 }, { "epoch": 2.17, "learning_rate": 2.5381129704933395e-05, "loss": 0.1831, "step": 21039 }, { "epoch": 2.17, "learning_rate": 2.5378697648490694e-05, "loss": 0.2508, "step": 21040 }, { "epoch": 2.17, "learning_rate": 2.5376265623157515e-05, "loss": 0.1345, "step": 21041 }, { "epoch": 2.17, "learning_rate": 2.537383362895025e-05, "loss": 0.5774, "step": 21042 }, { "epoch": 2.17, "learning_rate": 2.5371401665885236e-05, "loss": 0.1817, "step": 21043 }, { "epoch": 2.17, "learning_rate": 2.5368969733978873e-05, "loss": 0.1294, "step": 21044 }, { "epoch": 2.17, "learning_rate": 2.5366537833247516e-05, "loss": 0.0866, "step": 21045 }, { "epoch": 2.17, "learning_rate": 2.536410596370756e-05, "loss": 0.6764, "step": 21046 }, { "epoch": 2.17, "learning_rate": 2.536167412537534e-05, "loss": 0.5532, "step": 21047 }, { "epoch": 2.17, "learning_rate": 2.5359242318267254e-05, "loss": 0.2816, "step": 21048 }, { "epoch": 2.17, "learning_rate": 2.5356810542399657e-05, "loss": 0.2706, "step": 21049 }, { "epoch": 2.17, "learning_rate": 2.535437879778892e-05, "loss": 0.5105, "step": 21050 }, { "epoch": 2.17, "learning_rate": 2.5351947084451423e-05, "loss": 0.2979, "step": 21051 }, { "epoch": 2.17, "learning_rate": 2.5349515402403537e-05, "loss": 0.9072, "step": 21052 }, { "epoch": 2.17, "learning_rate": 2.5347083751661618e-05, "loss": 0.2358, "step": 21053 }, { "epoch": 2.17, "learning_rate": 2.534465213224204e-05, "loss": 0.2901, "step": 21054 }, { "epoch": 2.17, "learning_rate": 2.5342220544161176e-05, "loss": 0.3422, "step": 21055 }, { "epoch": 2.17, "learning_rate": 2.5339788987435402e-05, "loss": 0.1471, "step": 21056 }, { "epoch": 2.17, "learning_rate": 2.533735746208107e-05, "loss": 0.336, "step": 21057 }, { "epoch": 2.17, "learning_rate": 2.533492596811456e-05, "loss": 0.0752, "step": 21058 }, { "epoch": 2.17, "learning_rate": 2.5332494505552243e-05, "loss": 0.0918, "step": 21059 }, { "epoch": 2.17, "learning_rate": 2.5330063074410474e-05, "loss": 0.1701, "step": 21060 }, { "epoch": 2.17, "learning_rate": 2.5327631674705625e-05, "loss": 0.5047, "step": 21061 }, { "epoch": 2.17, "learning_rate": 2.5325200306454073e-05, "loss": 0.4042, "step": 21062 }, { "epoch": 2.17, "learning_rate": 2.5322768969672174e-05, "loss": 0.3934, "step": 21063 }, { "epoch": 2.17, "learning_rate": 2.5320337664376296e-05, "loss": 0.14, "step": 21064 }, { "epoch": 2.17, "learning_rate": 2.5317906390582813e-05, "loss": 0.2676, "step": 21065 }, { "epoch": 2.17, "learning_rate": 2.5315475148308093e-05, "loss": 0.6621, "step": 21066 }, { "epoch": 2.17, "learning_rate": 2.5313043937568495e-05, "loss": 0.2461, "step": 21067 }, { "epoch": 2.17, "learning_rate": 2.5310612758380386e-05, "loss": 0.6818, "step": 21068 }, { "epoch": 2.17, "learning_rate": 2.5308181610760134e-05, "loss": 0.1688, "step": 21069 }, { "epoch": 2.17, "learning_rate": 2.530575049472411e-05, "loss": 0.6591, "step": 21070 }, { "epoch": 2.17, "learning_rate": 2.5303319410288683e-05, "loss": 0.4444, "step": 21071 }, { "epoch": 2.17, "learning_rate": 2.5300888357470205e-05, "loss": 0.3971, "step": 21072 }, { "epoch": 2.17, "learning_rate": 2.529845733628504e-05, "loss": 0.7936, "step": 21073 }, { "epoch": 2.17, "learning_rate": 2.5296026346749563e-05, "loss": 0.6205, "step": 21074 }, { "epoch": 2.17, "learning_rate": 2.529359538888014e-05, "loss": 0.2707, "step": 21075 }, { "epoch": 2.17, "learning_rate": 2.529116446269313e-05, "loss": 0.5414, "step": 21076 }, { "epoch": 2.17, "learning_rate": 2.5288733568204898e-05, "loss": 0.5531, "step": 21077 }, { "epoch": 2.17, "learning_rate": 2.5286302705431804e-05, "loss": 0.564, "step": 21078 }, { "epoch": 2.17, "learning_rate": 2.5283871874390222e-05, "loss": 0.4461, "step": 21079 }, { "epoch": 2.17, "learning_rate": 2.5281441075096515e-05, "loss": 0.4781, "step": 21080 }, { "epoch": 2.18, "learning_rate": 2.527901030756704e-05, "loss": 0.3534, "step": 21081 }, { "epoch": 2.18, "learning_rate": 2.527657957181816e-05, "loss": 0.1596, "step": 21082 }, { "epoch": 2.18, "learning_rate": 2.5274148867866238e-05, "loss": 0.3518, "step": 21083 }, { "epoch": 2.18, "learning_rate": 2.5271718195727657e-05, "loss": 0.2012, "step": 21084 }, { "epoch": 2.18, "learning_rate": 2.5269287555418748e-05, "loss": 0.282, "step": 21085 }, { "epoch": 2.18, "learning_rate": 2.52668569469559e-05, "loss": 0.7451, "step": 21086 }, { "epoch": 2.18, "learning_rate": 2.5264426370355452e-05, "loss": 0.3841, "step": 21087 }, { "epoch": 2.18, "learning_rate": 2.5261995825633777e-05, "loss": 0.2291, "step": 21088 }, { "epoch": 2.18, "learning_rate": 2.5259565312807237e-05, "loss": 0.4588, "step": 21089 }, { "epoch": 2.18, "learning_rate": 2.52571348318922e-05, "loss": 0.0948, "step": 21090 }, { "epoch": 2.18, "learning_rate": 2.525470438290502e-05, "loss": 0.0593, "step": 21091 }, { "epoch": 2.18, "learning_rate": 2.5252273965862056e-05, "loss": 0.555, "step": 21092 }, { "epoch": 2.18, "learning_rate": 2.5249843580779675e-05, "loss": 0.4092, "step": 21093 }, { "epoch": 2.18, "learning_rate": 2.5247413227674237e-05, "loss": 0.2429, "step": 21094 }, { "epoch": 2.18, "learning_rate": 2.5244982906562105e-05, "loss": 0.101, "step": 21095 }, { "epoch": 2.18, "learning_rate": 2.5242552617459646e-05, "loss": 0.1869, "step": 21096 }, { "epoch": 2.18, "learning_rate": 2.524012236038319e-05, "loss": 0.0872, "step": 21097 }, { "epoch": 2.18, "learning_rate": 2.523769213534912e-05, "loss": 0.5212, "step": 21098 }, { "epoch": 2.18, "learning_rate": 2.5235261942373792e-05, "loss": 0.7981, "step": 21099 }, { "epoch": 2.18, "learning_rate": 2.523283178147357e-05, "loss": 0.663, "step": 21100 }, { "epoch": 2.18, "learning_rate": 2.5230401652664806e-05, "loss": 0.4544, "step": 21101 }, { "epoch": 2.18, "learning_rate": 2.522797155596386e-05, "loss": 1.0819, "step": 21102 }, { "epoch": 2.18, "learning_rate": 2.5225541491387092e-05, "loss": 0.2827, "step": 21103 }, { "epoch": 2.18, "learning_rate": 2.5223111458950873e-05, "loss": 0.3866, "step": 21104 }, { "epoch": 2.18, "learning_rate": 2.5220681458671538e-05, "loss": 0.4474, "step": 21105 }, { "epoch": 2.18, "learning_rate": 2.5218251490565462e-05, "loss": 0.2178, "step": 21106 }, { "epoch": 2.18, "learning_rate": 2.5215821554648993e-05, "loss": 0.2595, "step": 21107 }, { "epoch": 2.18, "learning_rate": 2.5213391650938495e-05, "loss": 0.417, "step": 21108 }, { "epoch": 2.18, "learning_rate": 2.521096177945034e-05, "loss": 0.4294, "step": 21109 }, { "epoch": 2.18, "learning_rate": 2.520853194020086e-05, "loss": 0.4891, "step": 21110 }, { "epoch": 2.18, "learning_rate": 2.5206102133206422e-05, "loss": 0.1666, "step": 21111 }, { "epoch": 2.18, "learning_rate": 2.5203672358483375e-05, "loss": 0.3565, "step": 21112 }, { "epoch": 2.18, "learning_rate": 2.520124261604809e-05, "loss": 0.1127, "step": 21113 }, { "epoch": 2.18, "learning_rate": 2.5198812905916915e-05, "loss": 0.1884, "step": 21114 }, { "epoch": 2.18, "learning_rate": 2.5196383228106208e-05, "loss": 0.3427, "step": 21115 }, { "epoch": 2.18, "learning_rate": 2.5193953582632325e-05, "loss": 0.0798, "step": 21116 }, { "epoch": 2.18, "learning_rate": 2.5191523969511618e-05, "loss": 0.6813, "step": 21117 }, { "epoch": 2.18, "learning_rate": 2.5189094388760458e-05, "loss": 0.2591, "step": 21118 }, { "epoch": 2.18, "learning_rate": 2.5186664840395183e-05, "loss": 0.4608, "step": 21119 }, { "epoch": 2.18, "learning_rate": 2.518423532443215e-05, "loss": 0.2521, "step": 21120 }, { "epoch": 2.18, "learning_rate": 2.5181805840887717e-05, "loss": 0.1699, "step": 21121 }, { "epoch": 2.18, "learning_rate": 2.5179376389778253e-05, "loss": 0.8526, "step": 21122 }, { "epoch": 2.18, "learning_rate": 2.517694697112009e-05, "loss": 0.3212, "step": 21123 }, { "epoch": 2.18, "learning_rate": 2.5174517584929594e-05, "loss": 0.411, "step": 21124 }, { "epoch": 2.18, "learning_rate": 2.5172088231223115e-05, "loss": 0.5783, "step": 21125 }, { "epoch": 2.18, "learning_rate": 2.5169658910017005e-05, "loss": 0.5911, "step": 21126 }, { "epoch": 2.18, "learning_rate": 2.5167229621327623e-05, "loss": 0.446, "step": 21127 }, { "epoch": 2.18, "learning_rate": 2.5164800365171326e-05, "loss": 0.3709, "step": 21128 }, { "epoch": 2.18, "learning_rate": 2.5162371141564453e-05, "loss": 0.3171, "step": 21129 }, { "epoch": 2.18, "learning_rate": 2.5159941950523367e-05, "loss": 0.3824, "step": 21130 }, { "epoch": 2.18, "learning_rate": 2.515751279206442e-05, "loss": 0.2946, "step": 21131 }, { "epoch": 2.18, "learning_rate": 2.515508366620396e-05, "loss": 0.1438, "step": 21132 }, { "epoch": 2.18, "learning_rate": 2.5152654572958356e-05, "loss": 0.2901, "step": 21133 }, { "epoch": 2.18, "learning_rate": 2.515022551234395e-05, "loss": 0.7896, "step": 21134 }, { "epoch": 2.18, "learning_rate": 2.5147796484377077e-05, "loss": 0.3708, "step": 21135 }, { "epoch": 2.18, "learning_rate": 2.5145367489074108e-05, "loss": 0.3714, "step": 21136 }, { "epoch": 2.18, "learning_rate": 2.5142938526451384e-05, "loss": 0.5219, "step": 21137 }, { "epoch": 2.18, "learning_rate": 2.5140509596525266e-05, "loss": 0.5433, "step": 21138 }, { "epoch": 2.18, "learning_rate": 2.51380806993121e-05, "loss": 0.7087, "step": 21139 }, { "epoch": 2.18, "learning_rate": 2.5135651834828232e-05, "loss": 0.4727, "step": 21140 }, { "epoch": 2.18, "learning_rate": 2.5133223003090022e-05, "loss": 0.1899, "step": 21141 }, { "epoch": 2.18, "learning_rate": 2.5130794204113822e-05, "loss": 0.293, "step": 21142 }, { "epoch": 2.18, "learning_rate": 2.5128365437915965e-05, "loss": 0.4251, "step": 21143 }, { "epoch": 2.18, "learning_rate": 2.5125936704512818e-05, "loss": 0.3522, "step": 21144 }, { "epoch": 2.18, "learning_rate": 2.5123508003920722e-05, "loss": 0.2973, "step": 21145 }, { "epoch": 2.18, "learning_rate": 2.512107933615603e-05, "loss": 0.2818, "step": 21146 }, { "epoch": 2.18, "learning_rate": 2.51186507012351e-05, "loss": 0.3078, "step": 21147 }, { "epoch": 2.18, "learning_rate": 2.511622209917427e-05, "loss": 0.1141, "step": 21148 }, { "epoch": 2.18, "learning_rate": 2.5113793529989878e-05, "loss": 0.3994, "step": 21149 }, { "epoch": 2.18, "learning_rate": 2.5111364993698287e-05, "loss": 0.2545, "step": 21150 }, { "epoch": 2.18, "learning_rate": 2.5108936490315844e-05, "loss": 0.3826, "step": 21151 }, { "epoch": 2.18, "learning_rate": 2.5106508019858904e-05, "loss": 0.3866, "step": 21152 }, { "epoch": 2.18, "learning_rate": 2.51040795823438e-05, "loss": 0.8432, "step": 21153 }, { "epoch": 2.18, "learning_rate": 2.5101651177786883e-05, "loss": 0.1619, "step": 21154 }, { "epoch": 2.18, "learning_rate": 2.5099222806204508e-05, "loss": 0.3469, "step": 21155 }, { "epoch": 2.18, "learning_rate": 2.5096794467613016e-05, "loss": 0.0657, "step": 21156 }, { "epoch": 2.18, "learning_rate": 2.509436616202876e-05, "loss": 0.6465, "step": 21157 }, { "epoch": 2.18, "learning_rate": 2.5091937889468086e-05, "loss": 0.3541, "step": 21158 }, { "epoch": 2.18, "learning_rate": 2.5089509649947343e-05, "loss": 0.44, "step": 21159 }, { "epoch": 2.18, "learning_rate": 2.508708144348286e-05, "loss": 0.5664, "step": 21160 }, { "epoch": 2.18, "learning_rate": 2.5084653270091e-05, "loss": 0.2948, "step": 21161 }, { "epoch": 2.18, "learning_rate": 2.5082225129788104e-05, "loss": 0.1891, "step": 21162 }, { "epoch": 2.18, "learning_rate": 2.5079797022590516e-05, "loss": 0.4212, "step": 21163 }, { "epoch": 2.18, "learning_rate": 2.507736894851458e-05, "loss": 0.2883, "step": 21164 }, { "epoch": 2.18, "learning_rate": 2.5074940907576645e-05, "loss": 0.4418, "step": 21165 }, { "epoch": 2.18, "learning_rate": 2.5072512899793065e-05, "loss": 0.0907, "step": 21166 }, { "epoch": 2.18, "learning_rate": 2.5070084925180165e-05, "loss": 0.3947, "step": 21167 }, { "epoch": 2.18, "learning_rate": 2.5067656983754304e-05, "loss": 0.1583, "step": 21168 }, { "epoch": 2.18, "learning_rate": 2.506522907553182e-05, "loss": 0.4285, "step": 21169 }, { "epoch": 2.18, "learning_rate": 2.5062801200529056e-05, "loss": 0.3766, "step": 21170 }, { "epoch": 2.18, "learning_rate": 2.5060373358762368e-05, "loss": 0.0227, "step": 21171 }, { "epoch": 2.18, "learning_rate": 2.50579455502481e-05, "loss": 0.5658, "step": 21172 }, { "epoch": 2.18, "learning_rate": 2.5055517775002567e-05, "loss": 0.571, "step": 21173 }, { "epoch": 2.18, "learning_rate": 2.5053090033042137e-05, "loss": 0.4316, "step": 21174 }, { "epoch": 2.18, "learning_rate": 2.5050662324383146e-05, "loss": 0.493, "step": 21175 }, { "epoch": 2.18, "learning_rate": 2.5048234649041945e-05, "loss": 0.6185, "step": 21176 }, { "epoch": 2.19, "learning_rate": 2.5045807007034862e-05, "loss": 0.2621, "step": 21177 }, { "epoch": 2.19, "learning_rate": 2.5043379398378247e-05, "loss": 0.4831, "step": 21178 }, { "epoch": 2.19, "learning_rate": 2.504095182308844e-05, "loss": 0.3697, "step": 21179 }, { "epoch": 2.19, "learning_rate": 2.503852428118179e-05, "loss": 0.0932, "step": 21180 }, { "epoch": 2.19, "learning_rate": 2.5036096772674637e-05, "loss": 0.3828, "step": 21181 }, { "epoch": 2.19, "learning_rate": 2.5033669297583314e-05, "loss": 0.1133, "step": 21182 }, { "epoch": 2.19, "learning_rate": 2.5031241855924165e-05, "loss": 0.6583, "step": 21183 }, { "epoch": 2.19, "learning_rate": 2.502881444771355e-05, "loss": 0.2361, "step": 21184 }, { "epoch": 2.19, "learning_rate": 2.5026387072967778e-05, "loss": 0.0892, "step": 21185 }, { "epoch": 2.19, "learning_rate": 2.5023959731703212e-05, "loss": 0.2244, "step": 21186 }, { "epoch": 2.19, "learning_rate": 2.5021532423936177e-05, "loss": 0.199, "step": 21187 }, { "epoch": 2.19, "learning_rate": 2.501910514968302e-05, "loss": 0.0684, "step": 21188 }, { "epoch": 2.19, "learning_rate": 2.5016677908960083e-05, "loss": 0.474, "step": 21189 }, { "epoch": 2.19, "learning_rate": 2.501425070178371e-05, "loss": 0.6826, "step": 21190 }, { "epoch": 2.19, "learning_rate": 2.5011823528170226e-05, "loss": 0.3365, "step": 21191 }, { "epoch": 2.19, "learning_rate": 2.5009396388135986e-05, "loss": 0.4519, "step": 21192 }, { "epoch": 2.19, "learning_rate": 2.5006969281697315e-05, "loss": 0.2112, "step": 21193 }, { "epoch": 2.19, "learning_rate": 2.5004542208870558e-05, "loss": 0.2686, "step": 21194 }, { "epoch": 2.19, "learning_rate": 2.5002115169672067e-05, "loss": 0.6421, "step": 21195 }, { "epoch": 2.19, "learning_rate": 2.4999688164118154e-05, "loss": 0.2892, "step": 21196 }, { "epoch": 2.19, "learning_rate": 2.499726119222519e-05, "loss": 0.2787, "step": 21197 }, { "epoch": 2.19, "learning_rate": 2.4994834254009477e-05, "loss": 0.176, "step": 21198 }, { "epoch": 2.19, "learning_rate": 2.4992407349487367e-05, "loss": 0.7634, "step": 21199 }, { "epoch": 2.19, "learning_rate": 2.4989980478675207e-05, "loss": 0.5203, "step": 21200 }, { "epoch": 2.19, "learning_rate": 2.498755364158932e-05, "loss": 0.156, "step": 21201 }, { "epoch": 2.19, "learning_rate": 2.4985126838246048e-05, "loss": 0.5983, "step": 21202 }, { "epoch": 2.19, "learning_rate": 2.4982700068661728e-05, "loss": 0.822, "step": 21203 }, { "epoch": 2.19, "learning_rate": 2.4980273332852704e-05, "loss": 0.2708, "step": 21204 }, { "epoch": 2.19, "learning_rate": 2.49778466308353e-05, "loss": 0.389, "step": 21205 }, { "epoch": 2.19, "learning_rate": 2.497541996262586e-05, "loss": 0.1122, "step": 21206 }, { "epoch": 2.19, "learning_rate": 2.4972993328240714e-05, "loss": 0.4736, "step": 21207 }, { "epoch": 2.19, "learning_rate": 2.49705667276962e-05, "loss": 0.2511, "step": 21208 }, { "epoch": 2.19, "learning_rate": 2.496814016100867e-05, "loss": 0.7107, "step": 21209 }, { "epoch": 2.19, "learning_rate": 2.4965713628194433e-05, "loss": 0.5612, "step": 21210 }, { "epoch": 2.19, "learning_rate": 2.4963287129269828e-05, "loss": 0.3377, "step": 21211 }, { "epoch": 2.19, "learning_rate": 2.4960860664251197e-05, "loss": 0.0878, "step": 21212 }, { "epoch": 2.19, "learning_rate": 2.4958434233154873e-05, "loss": 0.2533, "step": 21213 }, { "epoch": 2.19, "learning_rate": 2.4956007835997198e-05, "loss": 0.3515, "step": 21214 }, { "epoch": 2.19, "learning_rate": 2.495358147279449e-05, "loss": 0.0384, "step": 21215 }, { "epoch": 2.19, "learning_rate": 2.4951155143563092e-05, "loss": 0.3879, "step": 21216 }, { "epoch": 2.19, "learning_rate": 2.4948728848319333e-05, "loss": 0.0911, "step": 21217 }, { "epoch": 2.19, "learning_rate": 2.494630258707955e-05, "loss": 0.7163, "step": 21218 }, { "epoch": 2.19, "learning_rate": 2.4943876359860085e-05, "loss": 0.3327, "step": 21219 }, { "epoch": 2.19, "learning_rate": 2.4941450166677254e-05, "loss": 0.2206, "step": 21220 }, { "epoch": 2.19, "learning_rate": 2.4939024007547394e-05, "loss": 0.3849, "step": 21221 }, { "epoch": 2.19, "learning_rate": 2.4936597882486856e-05, "loss": 0.7006, "step": 21222 }, { "epoch": 2.19, "learning_rate": 2.4934171791511943e-05, "loss": 0.2654, "step": 21223 }, { "epoch": 2.19, "learning_rate": 2.4931745734639007e-05, "loss": 0.0909, "step": 21224 }, { "epoch": 2.19, "learning_rate": 2.4929319711884362e-05, "loss": 0.6691, "step": 21225 }, { "epoch": 2.19, "learning_rate": 2.4926893723264355e-05, "loss": 0.1569, "step": 21226 }, { "epoch": 2.19, "learning_rate": 2.492446776879531e-05, "loss": 0.2748, "step": 21227 }, { "epoch": 2.19, "learning_rate": 2.492204184849357e-05, "loss": 0.219, "step": 21228 }, { "epoch": 2.19, "learning_rate": 2.4919615962375445e-05, "loss": 0.3416, "step": 21229 }, { "epoch": 2.19, "learning_rate": 2.491719011045728e-05, "loss": 0.2256, "step": 21230 }, { "epoch": 2.19, "learning_rate": 2.4914764292755402e-05, "loss": 0.0253, "step": 21231 }, { "epoch": 2.19, "learning_rate": 2.491233850928614e-05, "loss": 0.1686, "step": 21232 }, { "epoch": 2.19, "learning_rate": 2.490991276006583e-05, "loss": 0.2561, "step": 21233 }, { "epoch": 2.19, "learning_rate": 2.4907487045110792e-05, "loss": 0.6624, "step": 21234 }, { "epoch": 2.19, "learning_rate": 2.4905061364437372e-05, "loss": 0.1187, "step": 21235 }, { "epoch": 2.19, "learning_rate": 2.4902635718061874e-05, "loss": 0.4386, "step": 21236 }, { "epoch": 2.19, "learning_rate": 2.490021010600064e-05, "loss": 0.6268, "step": 21237 }, { "epoch": 2.19, "learning_rate": 2.4897784528270004e-05, "loss": 0.1444, "step": 21238 }, { "epoch": 2.19, "learning_rate": 2.489535898488628e-05, "loss": 0.2584, "step": 21239 }, { "epoch": 2.19, "learning_rate": 2.489293347586581e-05, "loss": 1.0196, "step": 21240 }, { "epoch": 2.19, "learning_rate": 2.489050800122491e-05, "loss": 0.4804, "step": 21241 }, { "epoch": 2.19, "learning_rate": 2.4888082560979923e-05, "loss": 0.1566, "step": 21242 }, { "epoch": 2.19, "learning_rate": 2.488565715514717e-05, "loss": 0.4104, "step": 21243 }, { "epoch": 2.19, "learning_rate": 2.488323178374297e-05, "loss": 0.2165, "step": 21244 }, { "epoch": 2.19, "learning_rate": 2.4880806446783656e-05, "loss": 0.604, "step": 21245 }, { "epoch": 2.19, "learning_rate": 2.4878381144285558e-05, "loss": 0.6335, "step": 21246 }, { "epoch": 2.19, "learning_rate": 2.487595587626501e-05, "loss": 0.2024, "step": 21247 }, { "epoch": 2.19, "learning_rate": 2.487353064273832e-05, "loss": 0.541, "step": 21248 }, { "epoch": 2.19, "learning_rate": 2.487110544372182e-05, "loss": 0.5663, "step": 21249 }, { "epoch": 2.19, "learning_rate": 2.4868680279231834e-05, "loss": 0.7413, "step": 21250 }, { "epoch": 2.19, "learning_rate": 2.4866255149284695e-05, "loss": 0.1645, "step": 21251 }, { "epoch": 2.19, "learning_rate": 2.486383005389673e-05, "loss": 0.1131, "step": 21252 }, { "epoch": 2.19, "learning_rate": 2.4861404993084253e-05, "loss": 0.3388, "step": 21253 }, { "epoch": 2.19, "learning_rate": 2.48589799668636e-05, "loss": 0.2687, "step": 21254 }, { "epoch": 2.19, "learning_rate": 2.4856554975251082e-05, "loss": 0.2166, "step": 21255 }, { "epoch": 2.19, "learning_rate": 2.485413001826304e-05, "loss": 0.4652, "step": 21256 }, { "epoch": 2.19, "learning_rate": 2.4851705095915793e-05, "loss": 0.1981, "step": 21257 }, { "epoch": 2.19, "learning_rate": 2.4849280208225653e-05, "loss": 0.5043, "step": 21258 }, { "epoch": 2.19, "learning_rate": 2.484685535520896e-05, "loss": 0.2582, "step": 21259 }, { "epoch": 2.19, "learning_rate": 2.4844430536882043e-05, "loss": 0.1829, "step": 21260 }, { "epoch": 2.19, "learning_rate": 2.4842005753261197e-05, "loss": 0.5117, "step": 21261 }, { "epoch": 2.19, "learning_rate": 2.4839581004362772e-05, "loss": 0.6382, "step": 21262 }, { "epoch": 2.19, "learning_rate": 2.483715629020307e-05, "loss": 0.5641, "step": 21263 }, { "epoch": 2.19, "learning_rate": 2.4834731610798426e-05, "loss": 0.5902, "step": 21264 }, { "epoch": 2.19, "learning_rate": 2.483230696616516e-05, "loss": 0.0691, "step": 21265 }, { "epoch": 2.19, "learning_rate": 2.4829882356319593e-05, "loss": 0.4869, "step": 21266 }, { "epoch": 2.19, "learning_rate": 2.4827457781278055e-05, "loss": 0.3399, "step": 21267 }, { "epoch": 2.19, "learning_rate": 2.482503324105686e-05, "loss": 0.6278, "step": 21268 }, { "epoch": 2.19, "learning_rate": 2.482260873567232e-05, "loss": 0.381, "step": 21269 }, { "epoch": 2.19, "learning_rate": 2.4820184265140773e-05, "loss": 0.9599, "step": 21270 }, { "epoch": 2.19, "learning_rate": 2.481775982947854e-05, "loss": 0.3551, "step": 21271 }, { "epoch": 2.19, "learning_rate": 2.481533542870194e-05, "loss": 0.2695, "step": 21272 }, { "epoch": 2.2, "learning_rate": 2.481291106282728e-05, "loss": 0.3763, "step": 21273 }, { "epoch": 2.2, "learning_rate": 2.4810486731870884e-05, "loss": 0.9303, "step": 21274 }, { "epoch": 2.2, "learning_rate": 2.480806243584908e-05, "loss": 0.1716, "step": 21275 }, { "epoch": 2.2, "learning_rate": 2.480563817477819e-05, "loss": 0.5492, "step": 21276 }, { "epoch": 2.2, "learning_rate": 2.4803213948674525e-05, "loss": 0.1877, "step": 21277 }, { "epoch": 2.2, "learning_rate": 2.48007897575544e-05, "loss": 0.7911, "step": 21278 }, { "epoch": 2.2, "learning_rate": 2.479836560143415e-05, "loss": 0.4362, "step": 21279 }, { "epoch": 2.2, "learning_rate": 2.479594148033008e-05, "loss": 0.1716, "step": 21280 }, { "epoch": 2.2, "learning_rate": 2.4793517394258525e-05, "loss": 0.2062, "step": 21281 }, { "epoch": 2.2, "learning_rate": 2.4791093343235783e-05, "loss": 0.3372, "step": 21282 }, { "epoch": 2.2, "learning_rate": 2.4788669327278183e-05, "loss": 0.1526, "step": 21283 }, { "epoch": 2.2, "learning_rate": 2.4786245346402045e-05, "loss": 0.4312, "step": 21284 }, { "epoch": 2.2, "learning_rate": 2.478382140062369e-05, "loss": 0.1502, "step": 21285 }, { "epoch": 2.2, "learning_rate": 2.4781397489959427e-05, "loss": 0.3906, "step": 21286 }, { "epoch": 2.2, "learning_rate": 2.4778973614425568e-05, "loss": 0.5261, "step": 21287 }, { "epoch": 2.2, "learning_rate": 2.4776549774038436e-05, "loss": 0.6052, "step": 21288 }, { "epoch": 2.2, "learning_rate": 2.477412596881435e-05, "loss": 0.4607, "step": 21289 }, { "epoch": 2.2, "learning_rate": 2.4771702198769628e-05, "loss": 0.6617, "step": 21290 }, { "epoch": 2.2, "learning_rate": 2.4769278463920582e-05, "loss": 0.8913, "step": 21291 }, { "epoch": 2.2, "learning_rate": 2.4766854764283526e-05, "loss": 0.4911, "step": 21292 }, { "epoch": 2.2, "learning_rate": 2.4764431099874784e-05, "loss": 0.1188, "step": 21293 }, { "epoch": 2.2, "learning_rate": 2.4762007470710663e-05, "loss": 0.2864, "step": 21294 }, { "epoch": 2.2, "learning_rate": 2.4759583876807486e-05, "loss": 0.1373, "step": 21295 }, { "epoch": 2.2, "learning_rate": 2.475716031818156e-05, "loss": 0.782, "step": 21296 }, { "epoch": 2.2, "learning_rate": 2.4754736794849216e-05, "loss": 0.2118, "step": 21297 }, { "epoch": 2.2, "learning_rate": 2.4752313306826747e-05, "loss": 0.4609, "step": 21298 }, { "epoch": 2.2, "learning_rate": 2.474988985413048e-05, "loss": 0.6244, "step": 21299 }, { "epoch": 2.2, "learning_rate": 2.4747466436776724e-05, "loss": 0.1908, "step": 21300 }, { "epoch": 2.2, "learning_rate": 2.4745043054781794e-05, "loss": 0.4869, "step": 21301 }, { "epoch": 2.2, "learning_rate": 2.4742619708162e-05, "loss": 0.7643, "step": 21302 }, { "epoch": 2.2, "learning_rate": 2.4740196396933662e-05, "loss": 0.2348, "step": 21303 }, { "epoch": 2.2, "learning_rate": 2.4737773121113092e-05, "loss": 0.0667, "step": 21304 }, { "epoch": 2.2, "learning_rate": 2.473534988071661e-05, "loss": 0.2103, "step": 21305 }, { "epoch": 2.2, "learning_rate": 2.4732926675760513e-05, "loss": 0.596, "step": 21306 }, { "epoch": 2.2, "learning_rate": 2.473050350626112e-05, "loss": 0.5742, "step": 21307 }, { "epoch": 2.2, "learning_rate": 2.4728080372234748e-05, "loss": 0.2751, "step": 21308 }, { "epoch": 2.2, "learning_rate": 2.472565727369771e-05, "loss": 0.3159, "step": 21309 }, { "epoch": 2.2, "learning_rate": 2.4723234210666317e-05, "loss": 0.2711, "step": 21310 }, { "epoch": 2.2, "learning_rate": 2.4720811183156864e-05, "loss": 0.1103, "step": 21311 }, { "epoch": 2.2, "learning_rate": 2.4718388191185678e-05, "loss": 0.0571, "step": 21312 }, { "epoch": 2.2, "learning_rate": 2.4715965234769065e-05, "loss": 0.3769, "step": 21313 }, { "epoch": 2.2, "learning_rate": 2.4713542313923344e-05, "loss": 0.2107, "step": 21314 }, { "epoch": 2.2, "learning_rate": 2.4711119428664814e-05, "loss": 0.0689, "step": 21315 }, { "epoch": 2.2, "learning_rate": 2.470869657900979e-05, "loss": 0.6087, "step": 21316 }, { "epoch": 2.2, "learning_rate": 2.4706273764974584e-05, "loss": 0.2386, "step": 21317 }, { "epoch": 2.2, "learning_rate": 2.4703850986575502e-05, "loss": 0.5143, "step": 21318 }, { "epoch": 2.2, "learning_rate": 2.4701428243828863e-05, "loss": 0.362, "step": 21319 }, { "epoch": 2.2, "learning_rate": 2.4699005536750964e-05, "loss": 0.3382, "step": 21320 }, { "epoch": 2.2, "learning_rate": 2.4696582865358116e-05, "loss": 0.5833, "step": 21321 }, { "epoch": 2.2, "learning_rate": 2.4694160229666633e-05, "loss": 0.6949, "step": 21322 }, { "epoch": 2.2, "learning_rate": 2.469173762969284e-05, "loss": 0.1646, "step": 21323 }, { "epoch": 2.2, "learning_rate": 2.4689315065453013e-05, "loss": 0.5241, "step": 21324 }, { "epoch": 2.2, "learning_rate": 2.4686892536963473e-05, "loss": 0.515, "step": 21325 }, { "epoch": 2.2, "learning_rate": 2.4684470044240526e-05, "loss": 0.1414, "step": 21326 }, { "epoch": 2.2, "learning_rate": 2.4682047587300486e-05, "loss": 0.1432, "step": 21327 }, { "epoch": 2.2, "learning_rate": 2.4679625166159653e-05, "loss": 0.3244, "step": 21328 }, { "epoch": 2.2, "learning_rate": 2.467720278083435e-05, "loss": 0.4304, "step": 21329 }, { "epoch": 2.2, "learning_rate": 2.4674780431340864e-05, "loss": 0.4408, "step": 21330 }, { "epoch": 2.2, "learning_rate": 2.467235811769551e-05, "loss": 0.3094, "step": 21331 }, { "epoch": 2.2, "learning_rate": 2.4669935839914592e-05, "loss": 0.4331, "step": 21332 }, { "epoch": 2.2, "learning_rate": 2.466751359801443e-05, "loss": 0.0964, "step": 21333 }, { "epoch": 2.2, "learning_rate": 2.4665091392011308e-05, "loss": 0.5187, "step": 21334 }, { "epoch": 2.2, "learning_rate": 2.4662669221921557e-05, "loss": 0.3196, "step": 21335 }, { "epoch": 2.2, "learning_rate": 2.4660247087761456e-05, "loss": 0.7549, "step": 21336 }, { "epoch": 2.2, "learning_rate": 2.4657824989547323e-05, "loss": 0.2309, "step": 21337 }, { "epoch": 2.2, "learning_rate": 2.4655402927295465e-05, "loss": 0.3352, "step": 21338 }, { "epoch": 2.2, "learning_rate": 2.465298090102218e-05, "loss": 0.5793, "step": 21339 }, { "epoch": 2.2, "learning_rate": 2.4650558910743776e-05, "loss": 0.3913, "step": 21340 }, { "epoch": 2.2, "learning_rate": 2.464813695647656e-05, "loss": 0.2286, "step": 21341 }, { "epoch": 2.2, "learning_rate": 2.464571503823683e-05, "loss": 0.1613, "step": 21342 }, { "epoch": 2.2, "learning_rate": 2.4643293156040903e-05, "loss": 0.6988, "step": 21343 }, { "epoch": 2.2, "learning_rate": 2.4640871309905063e-05, "loss": 0.4685, "step": 21344 }, { "epoch": 2.2, "learning_rate": 2.463844949984563e-05, "loss": 0.4022, "step": 21345 }, { "epoch": 2.2, "learning_rate": 2.4636027725878894e-05, "loss": 0.1514, "step": 21346 }, { "epoch": 2.2, "learning_rate": 2.4633605988021176e-05, "loss": 0.3801, "step": 21347 }, { "epoch": 2.2, "learning_rate": 2.463118428628877e-05, "loss": 0.5046, "step": 21348 }, { "epoch": 2.2, "learning_rate": 2.462876262069796e-05, "loss": 0.0757, "step": 21349 }, { "epoch": 2.2, "learning_rate": 2.4626340991265068e-05, "loss": 1.0196, "step": 21350 }, { "epoch": 2.2, "learning_rate": 2.4623919398006388e-05, "loss": 0.1078, "step": 21351 }, { "epoch": 2.2, "learning_rate": 2.4621497840938223e-05, "loss": 0.3568, "step": 21352 }, { "epoch": 2.2, "learning_rate": 2.4619076320076885e-05, "loss": 0.2563, "step": 21353 }, { "epoch": 2.2, "learning_rate": 2.461665483543866e-05, "loss": 0.8856, "step": 21354 }, { "epoch": 2.2, "learning_rate": 2.4614233387039856e-05, "loss": 0.2469, "step": 21355 }, { "epoch": 2.2, "learning_rate": 2.4611811974896772e-05, "loss": 0.1976, "step": 21356 }, { "epoch": 2.2, "learning_rate": 2.4609390599025716e-05, "loss": 0.5274, "step": 21357 }, { "epoch": 2.2, "learning_rate": 2.4606969259442975e-05, "loss": 0.1301, "step": 21358 }, { "epoch": 2.2, "learning_rate": 2.460454795616485e-05, "loss": 0.6225, "step": 21359 }, { "epoch": 2.2, "learning_rate": 2.4602126689207667e-05, "loss": 0.5251, "step": 21360 }, { "epoch": 2.2, "learning_rate": 2.4599705458587684e-05, "loss": 0.3205, "step": 21361 }, { "epoch": 2.2, "learning_rate": 2.459728426432123e-05, "loss": 0.1549, "step": 21362 }, { "epoch": 2.2, "learning_rate": 2.4594863106424587e-05, "loss": 0.0417, "step": 21363 }, { "epoch": 2.2, "learning_rate": 2.4592441984914062e-05, "loss": 0.7687, "step": 21364 }, { "epoch": 2.2, "learning_rate": 2.459002089980595e-05, "loss": 0.2834, "step": 21365 }, { "epoch": 2.2, "learning_rate": 2.4587599851116556e-05, "loss": 0.0851, "step": 21366 }, { "epoch": 2.2, "learning_rate": 2.4585178838862176e-05, "loss": 0.6591, "step": 21367 }, { "epoch": 2.21, "learning_rate": 2.45827578630591e-05, "loss": 0.5082, "step": 21368 }, { "epoch": 2.21, "learning_rate": 2.458033692372363e-05, "loss": 0.1585, "step": 21369 }, { "epoch": 2.21, "learning_rate": 2.4577916020872062e-05, "loss": 0.155, "step": 21370 }, { "epoch": 2.21, "learning_rate": 2.4575495154520705e-05, "loss": 0.4681, "step": 21371 }, { "epoch": 2.21, "learning_rate": 2.4573074324685834e-05, "loss": 0.6437, "step": 21372 }, { "epoch": 2.21, "learning_rate": 2.457065353138377e-05, "loss": 0.567, "step": 21373 }, { "epoch": 2.21, "learning_rate": 2.456823277463079e-05, "loss": 0.1569, "step": 21374 }, { "epoch": 2.21, "learning_rate": 2.456581205444319e-05, "loss": 0.0903, "step": 21375 }, { "epoch": 2.21, "learning_rate": 2.456339137083728e-05, "loss": 0.1791, "step": 21376 }, { "epoch": 2.21, "learning_rate": 2.456097072382934e-05, "loss": 0.7329, "step": 21377 }, { "epoch": 2.21, "learning_rate": 2.4558550113435674e-05, "loss": 0.1236, "step": 21378 }, { "epoch": 2.21, "learning_rate": 2.4556129539672575e-05, "loss": 0.1323, "step": 21379 }, { "epoch": 2.21, "learning_rate": 2.455370900255634e-05, "loss": 0.1557, "step": 21380 }, { "epoch": 2.21, "learning_rate": 2.4551288502103263e-05, "loss": 0.5349, "step": 21381 }, { "epoch": 2.21, "learning_rate": 2.4548868038329634e-05, "loss": 0.4266, "step": 21382 }, { "epoch": 2.21, "learning_rate": 2.454644761125175e-05, "loss": 0.6873, "step": 21383 }, { "epoch": 2.21, "learning_rate": 2.45440272208859e-05, "loss": 0.3808, "step": 21384 }, { "epoch": 2.21, "learning_rate": 2.4541606867248404e-05, "loss": 0.4992, "step": 21385 }, { "epoch": 2.21, "learning_rate": 2.453918655035552e-05, "loss": 0.4021, "step": 21386 }, { "epoch": 2.21, "learning_rate": 2.453676627022355e-05, "loss": 0.6044, "step": 21387 }, { "epoch": 2.21, "learning_rate": 2.453434602686879e-05, "loss": 0.2734, "step": 21388 }, { "epoch": 2.21, "learning_rate": 2.4531925820307536e-05, "loss": 0.3018, "step": 21389 }, { "epoch": 2.21, "learning_rate": 2.4529505650556076e-05, "loss": 0.366, "step": 21390 }, { "epoch": 2.21, "learning_rate": 2.452708551763071e-05, "loss": 0.1639, "step": 21391 }, { "epoch": 2.21, "learning_rate": 2.4524665421547722e-05, "loss": 0.2923, "step": 21392 }, { "epoch": 2.21, "learning_rate": 2.4522245362323395e-05, "loss": 0.5705, "step": 21393 }, { "epoch": 2.21, "learning_rate": 2.4519825339974038e-05, "loss": 0.1704, "step": 21394 }, { "epoch": 2.21, "learning_rate": 2.451740535451594e-05, "loss": 0.3857, "step": 21395 }, { "epoch": 2.21, "learning_rate": 2.4514985405965386e-05, "loss": 0.1442, "step": 21396 }, { "epoch": 2.21, "learning_rate": 2.4512565494338657e-05, "loss": 0.2356, "step": 21397 }, { "epoch": 2.21, "learning_rate": 2.451014561965207e-05, "loss": 0.1494, "step": 21398 }, { "epoch": 2.21, "learning_rate": 2.4507725781921885e-05, "loss": 0.2237, "step": 21399 }, { "epoch": 2.21, "learning_rate": 2.4505305981164414e-05, "loss": 0.6318, "step": 21400 }, { "epoch": 2.21, "learning_rate": 2.450288621739593e-05, "loss": 0.4964, "step": 21401 }, { "epoch": 2.21, "learning_rate": 2.4500466490632726e-05, "loss": 0.2192, "step": 21402 }, { "epoch": 2.21, "learning_rate": 2.4498046800891096e-05, "loss": 0.2809, "step": 21403 }, { "epoch": 2.21, "learning_rate": 2.449562714818733e-05, "loss": 0.3542, "step": 21404 }, { "epoch": 2.21, "learning_rate": 2.449320753253772e-05, "loss": 0.3653, "step": 21405 }, { "epoch": 2.21, "learning_rate": 2.4490787953958538e-05, "loss": 0.2556, "step": 21406 }, { "epoch": 2.21, "learning_rate": 2.448836841246609e-05, "loss": 0.2689, "step": 21407 }, { "epoch": 2.21, "learning_rate": 2.448594890807665e-05, "loss": 0.5028, "step": 21408 }, { "epoch": 2.21, "learning_rate": 2.4483529440806516e-05, "loss": 0.5301, "step": 21409 }, { "epoch": 2.21, "learning_rate": 2.4481110010671984e-05, "loss": 0.5478, "step": 21410 }, { "epoch": 2.21, "learning_rate": 2.4478690617689313e-05, "loss": 0.1349, "step": 21411 }, { "epoch": 2.21, "learning_rate": 2.4476271261874807e-05, "loss": 0.1764, "step": 21412 }, { "epoch": 2.21, "learning_rate": 2.447385194324475e-05, "loss": 0.5039, "step": 21413 }, { "epoch": 2.21, "learning_rate": 2.4471432661815426e-05, "loss": 0.1809, "step": 21414 }, { "epoch": 2.21, "learning_rate": 2.4469013417603128e-05, "loss": 0.3061, "step": 21415 }, { "epoch": 2.21, "learning_rate": 2.4466594210624135e-05, "loss": 1.0781, "step": 21416 }, { "epoch": 2.21, "learning_rate": 2.4464175040894734e-05, "loss": 0.0924, "step": 21417 }, { "epoch": 2.21, "learning_rate": 2.446175590843121e-05, "loss": 0.7111, "step": 21418 }, { "epoch": 2.21, "learning_rate": 2.4459336813249858e-05, "loss": 0.2112, "step": 21419 }, { "epoch": 2.21, "learning_rate": 2.4456917755366946e-05, "loss": 0.4144, "step": 21420 }, { "epoch": 2.21, "learning_rate": 2.4454498734798767e-05, "loss": 0.2284, "step": 21421 }, { "epoch": 2.21, "learning_rate": 2.4452079751561606e-05, "loss": 0.2683, "step": 21422 }, { "epoch": 2.21, "learning_rate": 2.444966080567176e-05, "loss": 0.345, "step": 21423 }, { "epoch": 2.21, "learning_rate": 2.444724189714549e-05, "loss": 0.1634, "step": 21424 }, { "epoch": 2.21, "learning_rate": 2.4444823025999087e-05, "loss": 0.0834, "step": 21425 }, { "epoch": 2.21, "learning_rate": 2.4442404192248828e-05, "loss": 0.2565, "step": 21426 }, { "epoch": 2.21, "learning_rate": 2.4439985395911008e-05, "loss": 0.5288, "step": 21427 }, { "epoch": 2.21, "learning_rate": 2.4437566637001904e-05, "loss": 0.3109, "step": 21428 }, { "epoch": 2.21, "learning_rate": 2.4435147915537805e-05, "loss": 0.2004, "step": 21429 }, { "epoch": 2.21, "learning_rate": 2.4432729231534986e-05, "loss": 0.4976, "step": 21430 }, { "epoch": 2.21, "learning_rate": 2.4430310585009725e-05, "loss": 0.3903, "step": 21431 }, { "epoch": 2.21, "learning_rate": 2.4427891975978313e-05, "loss": 0.2839, "step": 21432 }, { "epoch": 2.21, "learning_rate": 2.4425473404457038e-05, "loss": 0.3428, "step": 21433 }, { "epoch": 2.21, "learning_rate": 2.4423054870462164e-05, "loss": 0.5, "step": 21434 }, { "epoch": 2.21, "learning_rate": 2.4420636374009976e-05, "loss": 0.3244, "step": 21435 }, { "epoch": 2.21, "learning_rate": 2.4418217915116774e-05, "loss": 0.5306, "step": 21436 }, { "epoch": 2.21, "learning_rate": 2.441579949379881e-05, "loss": 0.62, "step": 21437 }, { "epoch": 2.21, "learning_rate": 2.4413381110072383e-05, "loss": 0.1985, "step": 21438 }, { "epoch": 2.21, "learning_rate": 2.4410962763953764e-05, "loss": 0.075, "step": 21439 }, { "epoch": 2.21, "learning_rate": 2.440854445545923e-05, "loss": 0.6379, "step": 21440 }, { "epoch": 2.21, "learning_rate": 2.4406126184605072e-05, "loss": 0.569, "step": 21441 }, { "epoch": 2.21, "learning_rate": 2.440370795140756e-05, "loss": 0.4309, "step": 21442 }, { "epoch": 2.21, "learning_rate": 2.4401289755882986e-05, "loss": 0.6624, "step": 21443 }, { "epoch": 2.21, "learning_rate": 2.439887159804761e-05, "loss": 0.4975, "step": 21444 }, { "epoch": 2.21, "learning_rate": 2.439645347791772e-05, "loss": 0.9515, "step": 21445 }, { "epoch": 2.21, "learning_rate": 2.4394035395509597e-05, "loss": 0.3068, "step": 21446 }, { "epoch": 2.21, "learning_rate": 2.439161735083952e-05, "loss": 0.2854, "step": 21447 }, { "epoch": 2.21, "learning_rate": 2.438919934392377e-05, "loss": 0.403, "step": 21448 }, { "epoch": 2.21, "learning_rate": 2.4386781374778604e-05, "loss": 0.1489, "step": 21449 }, { "epoch": 2.21, "learning_rate": 2.4384363443420313e-05, "loss": 0.1816, "step": 21450 }, { "epoch": 2.21, "learning_rate": 2.438194554986517e-05, "loss": 0.6503, "step": 21451 }, { "epoch": 2.21, "learning_rate": 2.437952769412946e-05, "loss": 0.4342, "step": 21452 }, { "epoch": 2.21, "learning_rate": 2.437710987622946e-05, "loss": 0.1128, "step": 21453 }, { "epoch": 2.21, "learning_rate": 2.4374692096181432e-05, "loss": 0.2529, "step": 21454 }, { "epoch": 2.21, "learning_rate": 2.437227435400166e-05, "loss": 0.1721, "step": 21455 }, { "epoch": 2.21, "learning_rate": 2.4369856649706422e-05, "loss": 0.2074, "step": 21456 }, { "epoch": 2.21, "learning_rate": 2.4367438983312e-05, "loss": 0.2276, "step": 21457 }, { "epoch": 2.21, "learning_rate": 2.4365021354834653e-05, "loss": 0.2099, "step": 21458 }, { "epoch": 2.21, "learning_rate": 2.436260376429066e-05, "loss": 0.4124, "step": 21459 }, { "epoch": 2.21, "learning_rate": 2.4360186211696304e-05, "loss": 0.3011, "step": 21460 }, { "epoch": 2.21, "learning_rate": 2.435776869706787e-05, "loss": 0.2833, "step": 21461 }, { "epoch": 2.21, "learning_rate": 2.43553512204216e-05, "loss": 0.594, "step": 21462 }, { "epoch": 2.21, "learning_rate": 2.4352933781773787e-05, "loss": 0.2629, "step": 21463 }, { "epoch": 2.22, "learning_rate": 2.4350516381140703e-05, "loss": 0.6497, "step": 21464 }, { "epoch": 2.22, "learning_rate": 2.434809901853862e-05, "loss": 0.4514, "step": 21465 }, { "epoch": 2.22, "learning_rate": 2.434568169398381e-05, "loss": 0.1792, "step": 21466 }, { "epoch": 2.22, "learning_rate": 2.4343264407492554e-05, "loss": 0.6105, "step": 21467 }, { "epoch": 2.22, "learning_rate": 2.4340847159081113e-05, "loss": 0.2693, "step": 21468 }, { "epoch": 2.22, "learning_rate": 2.4338429948765765e-05, "loss": 0.5581, "step": 21469 }, { "epoch": 2.22, "learning_rate": 2.433601277656278e-05, "loss": 0.4449, "step": 21470 }, { "epoch": 2.22, "learning_rate": 2.433359564248844e-05, "loss": 0.7624, "step": 21471 }, { "epoch": 2.22, "learning_rate": 2.4331178546559002e-05, "loss": 0.3449, "step": 21472 }, { "epoch": 2.22, "learning_rate": 2.4328761488790757e-05, "loss": 0.4629, "step": 21473 }, { "epoch": 2.22, "learning_rate": 2.4326344469199947e-05, "loss": 0.1408, "step": 21474 }, { "epoch": 2.22, "learning_rate": 2.432392748780286e-05, "loss": 0.3989, "step": 21475 }, { "epoch": 2.22, "learning_rate": 2.4321510544615766e-05, "loss": 0.4974, "step": 21476 }, { "epoch": 2.22, "learning_rate": 2.431909363965494e-05, "loss": 0.1698, "step": 21477 }, { "epoch": 2.22, "learning_rate": 2.4316676772936638e-05, "loss": 0.5508, "step": 21478 }, { "epoch": 2.22, "learning_rate": 2.4314259944477145e-05, "loss": 0.9492, "step": 21479 }, { "epoch": 2.22, "learning_rate": 2.4311843154292717e-05, "loss": 0.3056, "step": 21480 }, { "epoch": 2.22, "learning_rate": 2.4309426402399636e-05, "loss": 0.2335, "step": 21481 }, { "epoch": 2.22, "learning_rate": 2.4307009688814163e-05, "loss": 0.4228, "step": 21482 }, { "epoch": 2.22, "learning_rate": 2.4304593013552566e-05, "loss": 0.1226, "step": 21483 }, { "epoch": 2.22, "learning_rate": 2.4302176376631116e-05, "loss": 0.2632, "step": 21484 }, { "epoch": 2.22, "learning_rate": 2.429975977806609e-05, "loss": 0.2809, "step": 21485 }, { "epoch": 2.22, "learning_rate": 2.4297343217873753e-05, "loss": 0.8132, "step": 21486 }, { "epoch": 2.22, "learning_rate": 2.4294926696070352e-05, "loss": 0.1004, "step": 21487 }, { "epoch": 2.22, "learning_rate": 2.4292510212672174e-05, "loss": 0.4698, "step": 21488 }, { "epoch": 2.22, "learning_rate": 2.4290093767695478e-05, "loss": 0.4502, "step": 21489 }, { "epoch": 2.22, "learning_rate": 2.4287677361156534e-05, "loss": 0.6325, "step": 21490 }, { "epoch": 2.22, "learning_rate": 2.428526099307162e-05, "loss": 0.4684, "step": 21491 }, { "epoch": 2.22, "learning_rate": 2.428284466345698e-05, "loss": 0.5545, "step": 21492 }, { "epoch": 2.22, "learning_rate": 2.42804283723289e-05, "loss": 0.5592, "step": 21493 }, { "epoch": 2.22, "learning_rate": 2.427801211970363e-05, "loss": 0.627, "step": 21494 }, { "epoch": 2.22, "learning_rate": 2.4275595905597454e-05, "loss": 0.7208, "step": 21495 }, { "epoch": 2.22, "learning_rate": 2.427317973002662e-05, "loss": 0.3348, "step": 21496 }, { "epoch": 2.22, "learning_rate": 2.42707635930074e-05, "loss": 0.6196, "step": 21497 }, { "epoch": 2.22, "learning_rate": 2.426834749455607e-05, "loss": 0.5503, "step": 21498 }, { "epoch": 2.22, "learning_rate": 2.4265931434688878e-05, "loss": 0.4888, "step": 21499 }, { "epoch": 2.22, "learning_rate": 2.4263515413422084e-05, "loss": 0.1426, "step": 21500 }, { "epoch": 2.22, "learning_rate": 2.4261099430771976e-05, "loss": 0.0613, "step": 21501 }, { "epoch": 2.22, "learning_rate": 2.4258683486754794e-05, "loss": 0.1623, "step": 21502 }, { "epoch": 2.22, "learning_rate": 2.425626758138681e-05, "loss": 0.7751, "step": 21503 }, { "epoch": 2.22, "learning_rate": 2.425385171468429e-05, "loss": 1.1316, "step": 21504 }, { "epoch": 2.22, "learning_rate": 2.4251435886663503e-05, "loss": 0.564, "step": 21505 }, { "epoch": 2.22, "learning_rate": 2.4249020097340697e-05, "loss": 0.1559, "step": 21506 }, { "epoch": 2.22, "learning_rate": 2.4246604346732144e-05, "loss": 0.4123, "step": 21507 }, { "epoch": 2.22, "learning_rate": 2.4244188634854102e-05, "loss": 0.5398, "step": 21508 }, { "epoch": 2.22, "learning_rate": 2.4241772961722844e-05, "loss": 0.3244, "step": 21509 }, { "epoch": 2.22, "learning_rate": 2.4239357327354615e-05, "loss": 0.4731, "step": 21510 }, { "epoch": 2.22, "learning_rate": 2.42369417317657e-05, "loss": 0.1541, "step": 21511 }, { "epoch": 2.22, "learning_rate": 2.423452617497233e-05, "loss": 0.5011, "step": 21512 }, { "epoch": 2.22, "learning_rate": 2.4232110656990777e-05, "loss": 0.5392, "step": 21513 }, { "epoch": 2.22, "learning_rate": 2.4229695177837307e-05, "loss": 0.112, "step": 21514 }, { "epoch": 2.22, "learning_rate": 2.4227279737528193e-05, "loss": 0.4431, "step": 21515 }, { "epoch": 2.22, "learning_rate": 2.4224864336079665e-05, "loss": 0.3375, "step": 21516 }, { "epoch": 2.22, "learning_rate": 2.4222448973508006e-05, "loss": 0.5656, "step": 21517 }, { "epoch": 2.22, "learning_rate": 2.4220033649829463e-05, "loss": 0.2475, "step": 21518 }, { "epoch": 2.22, "learning_rate": 2.421761836506031e-05, "loss": 0.7696, "step": 21519 }, { "epoch": 2.22, "learning_rate": 2.421520311921679e-05, "loss": 0.4897, "step": 21520 }, { "epoch": 2.22, "learning_rate": 2.4212787912315173e-05, "loss": 0.2472, "step": 21521 }, { "epoch": 2.22, "learning_rate": 2.421037274437171e-05, "loss": 0.336, "step": 21522 }, { "epoch": 2.22, "learning_rate": 2.4207957615402672e-05, "loss": 0.2502, "step": 21523 }, { "epoch": 2.22, "learning_rate": 2.420554252542431e-05, "loss": 0.3041, "step": 21524 }, { "epoch": 2.22, "learning_rate": 2.4203127474452874e-05, "loss": 0.1109, "step": 21525 }, { "epoch": 2.22, "learning_rate": 2.420071246250462e-05, "loss": 0.166, "step": 21526 }, { "epoch": 2.22, "learning_rate": 2.4198297489595818e-05, "loss": 0.3794, "step": 21527 }, { "epoch": 2.22, "learning_rate": 2.419588255574272e-05, "loss": 0.1749, "step": 21528 }, { "epoch": 2.22, "learning_rate": 2.4193467660961582e-05, "loss": 0.297, "step": 21529 }, { "epoch": 2.22, "learning_rate": 2.419105280526866e-05, "loss": 0.5337, "step": 21530 }, { "epoch": 2.22, "learning_rate": 2.4188637988680212e-05, "loss": 0.176, "step": 21531 }, { "epoch": 2.22, "learning_rate": 2.4186223211212495e-05, "loss": 0.5188, "step": 21532 }, { "epoch": 2.22, "learning_rate": 2.4183808472881763e-05, "loss": 0.7199, "step": 21533 }, { "epoch": 2.22, "learning_rate": 2.418139377370427e-05, "loss": 0.5645, "step": 21534 }, { "epoch": 2.22, "learning_rate": 2.4178979113696273e-05, "loss": 0.1303, "step": 21535 }, { "epoch": 2.22, "learning_rate": 2.417656449287404e-05, "loss": 1.1819, "step": 21536 }, { "epoch": 2.22, "learning_rate": 2.4174149911253793e-05, "loss": 0.0898, "step": 21537 }, { "epoch": 2.22, "learning_rate": 2.4171735368851813e-05, "loss": 0.0306, "step": 21538 }, { "epoch": 2.22, "learning_rate": 2.416932086568435e-05, "loss": 0.3078, "step": 21539 }, { "epoch": 2.22, "learning_rate": 2.4166906401767646e-05, "loss": 0.7683, "step": 21540 }, { "epoch": 2.22, "learning_rate": 2.4164491977117968e-05, "loss": 0.3451, "step": 21541 }, { "epoch": 2.22, "learning_rate": 2.416207759175156e-05, "loss": 0.1467, "step": 21542 }, { "epoch": 2.22, "learning_rate": 2.4159663245684692e-05, "loss": 0.3444, "step": 21543 }, { "epoch": 2.22, "learning_rate": 2.415724893893359e-05, "loss": 0.3085, "step": 21544 }, { "epoch": 2.22, "learning_rate": 2.4154834671514525e-05, "loss": 0.5065, "step": 21545 }, { "epoch": 2.22, "learning_rate": 2.4152420443443748e-05, "loss": 0.4708, "step": 21546 }, { "epoch": 2.22, "learning_rate": 2.4150006254737512e-05, "loss": 0.4439, "step": 21547 }, { "epoch": 2.22, "learning_rate": 2.4147592105412058e-05, "loss": 0.3329, "step": 21548 }, { "epoch": 2.22, "learning_rate": 2.414517799548366e-05, "loss": 0.174, "step": 21549 }, { "epoch": 2.22, "learning_rate": 2.4142763924968544e-05, "loss": 0.4236, "step": 21550 }, { "epoch": 2.22, "learning_rate": 2.4140349893882964e-05, "loss": 0.262, "step": 21551 }, { "epoch": 2.22, "learning_rate": 2.413793590224318e-05, "loss": 0.2196, "step": 21552 }, { "epoch": 2.22, "learning_rate": 2.4135521950065446e-05, "loss": 0.6936, "step": 21553 }, { "epoch": 2.22, "learning_rate": 2.4133108037366e-05, "loss": 0.5091, "step": 21554 }, { "epoch": 2.22, "learning_rate": 2.4130694164161097e-05, "loss": 0.151, "step": 21555 }, { "epoch": 2.22, "learning_rate": 2.4128280330466987e-05, "loss": 0.2928, "step": 21556 }, { "epoch": 2.22, "learning_rate": 2.412586653629993e-05, "loss": 0.2274, "step": 21557 }, { "epoch": 2.22, "learning_rate": 2.4123452781676155e-05, "loss": 0.7008, "step": 21558 }, { "epoch": 2.22, "learning_rate": 2.4121039066611923e-05, "loss": 0.3493, "step": 21559 }, { "epoch": 2.23, "learning_rate": 2.411862539112348e-05, "loss": 0.2353, "step": 21560 }, { "epoch": 2.23, "learning_rate": 2.411621175522709e-05, "loss": 0.5783, "step": 21561 }, { "epoch": 2.23, "learning_rate": 2.4113798158938967e-05, "loss": 0.2573, "step": 21562 }, { "epoch": 2.23, "learning_rate": 2.4111384602275388e-05, "loss": 0.1549, "step": 21563 }, { "epoch": 2.23, "learning_rate": 2.4108971085252585e-05, "loss": 0.8344, "step": 21564 }, { "epoch": 2.23, "learning_rate": 2.4106557607886804e-05, "loss": 0.4722, "step": 21565 }, { "epoch": 2.23, "learning_rate": 2.4104144170194302e-05, "loss": 0.3196, "step": 21566 }, { "epoch": 2.23, "learning_rate": 2.410173077219133e-05, "loss": 0.0791, "step": 21567 }, { "epoch": 2.23, "learning_rate": 2.409931741389412e-05, "loss": 0.0965, "step": 21568 }, { "epoch": 2.23, "learning_rate": 2.4096904095318922e-05, "loss": 0.4625, "step": 21569 }, { "epoch": 2.23, "learning_rate": 2.4094490816481987e-05, "loss": 0.3129, "step": 21570 }, { "epoch": 2.23, "learning_rate": 2.409207757739956e-05, "loss": 0.3096, "step": 21571 }, { "epoch": 2.23, "learning_rate": 2.4089664378087883e-05, "loss": 0.2455, "step": 21572 }, { "epoch": 2.23, "learning_rate": 2.40872512185632e-05, "loss": 0.0973, "step": 21573 }, { "epoch": 2.23, "learning_rate": 2.4084838098841773e-05, "loss": 0.2136, "step": 21574 }, { "epoch": 2.23, "learning_rate": 2.4082425018939816e-05, "loss": 0.4293, "step": 21575 }, { "epoch": 2.23, "learning_rate": 2.408001197887359e-05, "loss": 0.0951, "step": 21576 }, { "epoch": 2.23, "learning_rate": 2.4077598978659346e-05, "loss": 0.0635, "step": 21577 }, { "epoch": 2.23, "learning_rate": 2.4075186018313314e-05, "loss": 0.0596, "step": 21578 }, { "epoch": 2.23, "learning_rate": 2.4072773097851744e-05, "loss": 0.4565, "step": 21579 }, { "epoch": 2.23, "learning_rate": 2.4070360217290876e-05, "loss": 0.4993, "step": 21580 }, { "epoch": 2.23, "learning_rate": 2.406794737664696e-05, "loss": 0.0359, "step": 21581 }, { "epoch": 2.23, "learning_rate": 2.4065534575936232e-05, "loss": 0.4161, "step": 21582 }, { "epoch": 2.23, "learning_rate": 2.4063121815174934e-05, "loss": 0.1375, "step": 21583 }, { "epoch": 2.23, "learning_rate": 2.4060709094379317e-05, "loss": 0.6051, "step": 21584 }, { "epoch": 2.23, "learning_rate": 2.4058296413565614e-05, "loss": 0.604, "step": 21585 }, { "epoch": 2.23, "learning_rate": 2.4055883772750077e-05, "loss": 0.2323, "step": 21586 }, { "epoch": 2.23, "learning_rate": 2.4053471171948935e-05, "loss": 0.5445, "step": 21587 }, { "epoch": 2.23, "learning_rate": 2.405105861117843e-05, "loss": 0.6853, "step": 21588 }, { "epoch": 2.23, "learning_rate": 2.4048646090454806e-05, "loss": 0.7852, "step": 21589 }, { "epoch": 2.23, "learning_rate": 2.4046233609794303e-05, "loss": 0.4555, "step": 21590 }, { "epoch": 2.23, "learning_rate": 2.4043821169213167e-05, "loss": 0.3221, "step": 21591 }, { "epoch": 2.23, "learning_rate": 2.404140876872763e-05, "loss": 0.3261, "step": 21592 }, { "epoch": 2.23, "learning_rate": 2.4038996408353934e-05, "loss": 0.3317, "step": 21593 }, { "epoch": 2.23, "learning_rate": 2.403658408810832e-05, "loss": 0.3366, "step": 21594 }, { "epoch": 2.23, "learning_rate": 2.403417180800703e-05, "loss": 0.6387, "step": 21595 }, { "epoch": 2.23, "learning_rate": 2.4031759568066295e-05, "loss": 0.2612, "step": 21596 }, { "epoch": 2.23, "learning_rate": 2.402934736830236e-05, "loss": 0.3276, "step": 21597 }, { "epoch": 2.23, "learning_rate": 2.4026935208731457e-05, "loss": 0.388, "step": 21598 }, { "epoch": 2.23, "learning_rate": 2.4024523089369846e-05, "loss": 0.0857, "step": 21599 }, { "epoch": 2.23, "learning_rate": 2.4022111010233732e-05, "loss": 0.2835, "step": 21600 }, { "epoch": 2.23, "learning_rate": 2.4019698971339373e-05, "loss": 0.9217, "step": 21601 }, { "epoch": 2.23, "learning_rate": 2.4017286972703e-05, "loss": 0.5992, "step": 21602 }, { "epoch": 2.23, "learning_rate": 2.4014875014340848e-05, "loss": 0.4388, "step": 21603 }, { "epoch": 2.23, "learning_rate": 2.401246309626916e-05, "loss": 0.4063, "step": 21604 }, { "epoch": 2.23, "learning_rate": 2.401005121850417e-05, "loss": 0.5266, "step": 21605 }, { "epoch": 2.23, "learning_rate": 2.400763938106211e-05, "loss": 0.1776, "step": 21606 }, { "epoch": 2.23, "learning_rate": 2.4005227583959224e-05, "loss": 0.1753, "step": 21607 }, { "epoch": 2.23, "learning_rate": 2.4002815827211734e-05, "loss": 0.4059, "step": 21608 }, { "epoch": 2.23, "learning_rate": 2.40004041108359e-05, "loss": 0.5196, "step": 21609 }, { "epoch": 2.23, "learning_rate": 2.399799243484793e-05, "loss": 0.3031, "step": 21610 }, { "epoch": 2.23, "learning_rate": 2.3995580799264087e-05, "loss": 0.4413, "step": 21611 }, { "epoch": 2.23, "learning_rate": 2.3993169204100572e-05, "loss": 0.3715, "step": 21612 }, { "epoch": 2.23, "learning_rate": 2.3990757649373638e-05, "loss": 0.4554, "step": 21613 }, { "epoch": 2.23, "learning_rate": 2.398834613509952e-05, "loss": 0.6076, "step": 21614 }, { "epoch": 2.23, "learning_rate": 2.3985934661294455e-05, "loss": 0.7917, "step": 21615 }, { "epoch": 2.23, "learning_rate": 2.3983523227974664e-05, "loss": 0.3717, "step": 21616 }, { "epoch": 2.23, "learning_rate": 2.3981111835156383e-05, "loss": 0.6686, "step": 21617 }, { "epoch": 2.23, "learning_rate": 2.3978700482855853e-05, "loss": 0.4958, "step": 21618 }, { "epoch": 2.23, "learning_rate": 2.3976289171089308e-05, "loss": 0.6244, "step": 21619 }, { "epoch": 2.23, "learning_rate": 2.3973877899872968e-05, "loss": 0.3414, "step": 21620 }, { "epoch": 2.23, "learning_rate": 2.397146666922307e-05, "loss": 0.7037, "step": 21621 }, { "epoch": 2.23, "learning_rate": 2.3969055479155853e-05, "loss": 0.217, "step": 21622 }, { "epoch": 2.23, "learning_rate": 2.3966644329687543e-05, "loss": 0.682, "step": 21623 }, { "epoch": 2.23, "learning_rate": 2.3964233220834378e-05, "loss": 0.0793, "step": 21624 }, { "epoch": 2.23, "learning_rate": 2.396182215261258e-05, "loss": 0.6445, "step": 21625 }, { "epoch": 2.23, "learning_rate": 2.3959411125038377e-05, "loss": 0.379, "step": 21626 }, { "epoch": 2.23, "learning_rate": 2.3957000138128002e-05, "loss": 0.4763, "step": 21627 }, { "epoch": 2.23, "learning_rate": 2.3954589191897692e-05, "loss": 0.6186, "step": 21628 }, { "epoch": 2.23, "learning_rate": 2.3952178286363676e-05, "loss": 0.4767, "step": 21629 }, { "epoch": 2.23, "learning_rate": 2.3949767421542177e-05, "loss": 0.5501, "step": 21630 }, { "epoch": 2.23, "learning_rate": 2.3947356597449428e-05, "loss": 0.0659, "step": 21631 }, { "epoch": 2.23, "learning_rate": 2.3944945814101656e-05, "loss": 0.2813, "step": 21632 }, { "epoch": 2.23, "learning_rate": 2.39425350715151e-05, "loss": 0.1613, "step": 21633 }, { "epoch": 2.23, "learning_rate": 2.3940124369705975e-05, "loss": 0.7014, "step": 21634 }, { "epoch": 2.23, "learning_rate": 2.3937713708690514e-05, "loss": 0.3465, "step": 21635 }, { "epoch": 2.23, "learning_rate": 2.3935303088484947e-05, "loss": 0.1835, "step": 21636 }, { "epoch": 2.23, "learning_rate": 2.393289250910551e-05, "loss": 0.2542, "step": 21637 }, { "epoch": 2.23, "learning_rate": 2.393048197056841e-05, "loss": 0.299, "step": 21638 }, { "epoch": 2.23, "learning_rate": 2.3928071472889888e-05, "loss": 0.3171, "step": 21639 }, { "epoch": 2.23, "learning_rate": 2.392566101608617e-05, "loss": 0.148, "step": 21640 }, { "epoch": 2.23, "learning_rate": 2.3923250600173477e-05, "loss": 0.3707, "step": 21641 }, { "epoch": 2.23, "learning_rate": 2.3920840225168035e-05, "loss": 0.2523, "step": 21642 }, { "epoch": 2.23, "learning_rate": 2.3918429891086088e-05, "loss": 0.6189, "step": 21643 }, { "epoch": 2.23, "learning_rate": 2.3916019597943832e-05, "loss": 0.1859, "step": 21644 }, { "epoch": 2.23, "learning_rate": 2.3913609345757516e-05, "loss": 0.5172, "step": 21645 }, { "epoch": 2.23, "learning_rate": 2.3911199134543357e-05, "loss": 0.2607, "step": 21646 }, { "epoch": 2.23, "learning_rate": 2.3908788964317586e-05, "loss": 0.2678, "step": 21647 }, { "epoch": 2.23, "learning_rate": 2.3906378835096415e-05, "loss": 0.2029, "step": 21648 }, { "epoch": 2.23, "learning_rate": 2.390396874689609e-05, "loss": 0.1971, "step": 21649 }, { "epoch": 2.23, "learning_rate": 2.3901558699732807e-05, "loss": 0.1949, "step": 21650 }, { "epoch": 2.23, "learning_rate": 2.3899148693622806e-05, "loss": 0.2242, "step": 21651 }, { "epoch": 2.23, "learning_rate": 2.3896738728582306e-05, "loss": 0.6988, "step": 21652 }, { "epoch": 2.23, "learning_rate": 2.3894328804627543e-05, "loss": 0.1506, "step": 21653 }, { "epoch": 2.23, "learning_rate": 2.389191892177472e-05, "loss": 0.5853, "step": 21654 }, { "epoch": 2.23, "learning_rate": 2.388950908004007e-05, "loss": 0.6047, "step": 21655 }, { "epoch": 2.24, "learning_rate": 2.3887099279439816e-05, "loss": 0.3072, "step": 21656 }, { "epoch": 2.24, "learning_rate": 2.3884689519990187e-05, "loss": 0.1383, "step": 21657 }, { "epoch": 2.24, "learning_rate": 2.3882279801707387e-05, "loss": 0.389, "step": 21658 }, { "epoch": 2.24, "learning_rate": 2.387987012460765e-05, "loss": 0.5436, "step": 21659 }, { "epoch": 2.24, "learning_rate": 2.38774604887072e-05, "loss": 0.4389, "step": 21660 }, { "epoch": 2.24, "learning_rate": 2.3875050894022254e-05, "loss": 0.6816, "step": 21661 }, { "epoch": 2.24, "learning_rate": 2.3872641340569035e-05, "loss": 0.3507, "step": 21662 }, { "epoch": 2.24, "learning_rate": 2.387023182836376e-05, "loss": 0.7145, "step": 21663 }, { "epoch": 2.24, "learning_rate": 2.386782235742264e-05, "loss": 0.5516, "step": 21664 }, { "epoch": 2.24, "learning_rate": 2.386541292776191e-05, "loss": 0.2266, "step": 21665 }, { "epoch": 2.24, "learning_rate": 2.386300353939778e-05, "loss": 0.0666, "step": 21666 }, { "epoch": 2.24, "learning_rate": 2.3860594192346488e-05, "loss": 0.8621, "step": 21667 }, { "epoch": 2.24, "learning_rate": 2.3858184886624227e-05, "loss": 0.3434, "step": 21668 }, { "epoch": 2.24, "learning_rate": 2.385577562224723e-05, "loss": 0.3244, "step": 21669 }, { "epoch": 2.24, "learning_rate": 2.3853366399231712e-05, "loss": 0.1577, "step": 21670 }, { "epoch": 2.24, "learning_rate": 2.3850957217593902e-05, "loss": 0.2893, "step": 21671 }, { "epoch": 2.24, "learning_rate": 2.3848548077350004e-05, "loss": 0.32, "step": 21672 }, { "epoch": 2.24, "learning_rate": 2.384613897851624e-05, "loss": 0.4633, "step": 21673 }, { "epoch": 2.24, "learning_rate": 2.384372992110884e-05, "loss": 0.107, "step": 21674 }, { "epoch": 2.24, "learning_rate": 2.3841320905143998e-05, "loss": 0.235, "step": 21675 }, { "epoch": 2.24, "learning_rate": 2.383891193063794e-05, "loss": 0.7915, "step": 21676 }, { "epoch": 2.24, "learning_rate": 2.3836502997606898e-05, "loss": 0.2833, "step": 21677 }, { "epoch": 2.24, "learning_rate": 2.3834094106067064e-05, "loss": 0.2356, "step": 21678 }, { "epoch": 2.24, "learning_rate": 2.3831685256034668e-05, "loss": 0.2273, "step": 21679 }, { "epoch": 2.24, "learning_rate": 2.382927644752592e-05, "loss": 1.1439, "step": 21680 }, { "epoch": 2.24, "learning_rate": 2.382686768055705e-05, "loss": 0.2319, "step": 21681 }, { "epoch": 2.24, "learning_rate": 2.3824458955144255e-05, "loss": 0.1955, "step": 21682 }, { "epoch": 2.24, "learning_rate": 2.3822050271303755e-05, "loss": 0.0971, "step": 21683 }, { "epoch": 2.24, "learning_rate": 2.381964162905177e-05, "loss": 1.1886, "step": 21684 }, { "epoch": 2.24, "learning_rate": 2.3817233028404515e-05, "loss": 0.3931, "step": 21685 }, { "epoch": 2.24, "learning_rate": 2.38148244693782e-05, "loss": 0.0733, "step": 21686 }, { "epoch": 2.24, "learning_rate": 2.3812415951989044e-05, "loss": 0.1562, "step": 21687 }, { "epoch": 2.24, "learning_rate": 2.3810007476253246e-05, "loss": 0.4114, "step": 21688 }, { "epoch": 2.24, "learning_rate": 2.380759904218703e-05, "loss": 0.4792, "step": 21689 }, { "epoch": 2.24, "learning_rate": 2.380519064980661e-05, "loss": 0.4789, "step": 21690 }, { "epoch": 2.24, "learning_rate": 2.3802782299128204e-05, "loss": 0.3897, "step": 21691 }, { "epoch": 2.24, "learning_rate": 2.380037399016801e-05, "loss": 0.7871, "step": 21692 }, { "epoch": 2.24, "learning_rate": 2.3797965722942247e-05, "loss": 0.103, "step": 21693 }, { "epoch": 2.24, "learning_rate": 2.3795557497467124e-05, "loss": 0.2442, "step": 21694 }, { "epoch": 2.24, "learning_rate": 2.3793149313758868e-05, "loss": 0.5612, "step": 21695 }, { "epoch": 2.24, "learning_rate": 2.3790741171833672e-05, "loss": 0.2122, "step": 21696 }, { "epoch": 2.24, "learning_rate": 2.3788333071707754e-05, "loss": 0.5622, "step": 21697 }, { "epoch": 2.24, "learning_rate": 2.3785925013397317e-05, "loss": 0.2916, "step": 21698 }, { "epoch": 2.24, "learning_rate": 2.37835169969186e-05, "loss": 0.2501, "step": 21699 }, { "epoch": 2.24, "learning_rate": 2.3781109022287778e-05, "loss": 0.3684, "step": 21700 }, { "epoch": 2.24, "learning_rate": 2.377870108952108e-05, "loss": 0.2915, "step": 21701 }, { "epoch": 2.24, "learning_rate": 2.37762931986347e-05, "loss": 0.6592, "step": 21702 }, { "epoch": 2.24, "learning_rate": 2.3773885349644865e-05, "loss": 0.3892, "step": 21703 }, { "epoch": 2.24, "learning_rate": 2.377147754256777e-05, "loss": 0.1836, "step": 21704 }, { "epoch": 2.24, "learning_rate": 2.3769069777419644e-05, "loss": 0.2667, "step": 21705 }, { "epoch": 2.24, "learning_rate": 2.376666205421667e-05, "loss": 0.5796, "step": 21706 }, { "epoch": 2.24, "learning_rate": 2.376425437297507e-05, "loss": 0.1832, "step": 21707 }, { "epoch": 2.24, "learning_rate": 2.3761846733711055e-05, "loss": 0.593, "step": 21708 }, { "epoch": 2.24, "learning_rate": 2.375943913644083e-05, "loss": 0.1643, "step": 21709 }, { "epoch": 2.24, "learning_rate": 2.37570315811806e-05, "loss": 0.393, "step": 21710 }, { "epoch": 2.24, "learning_rate": 2.3754624067946564e-05, "loss": 0.3108, "step": 21711 }, { "epoch": 2.24, "learning_rate": 2.375221659675496e-05, "loss": 0.2929, "step": 21712 }, { "epoch": 2.24, "learning_rate": 2.374980916762195e-05, "loss": 0.263, "step": 21713 }, { "epoch": 2.24, "learning_rate": 2.3747401780563766e-05, "loss": 0.6688, "step": 21714 }, { "epoch": 2.24, "learning_rate": 2.3744994435596618e-05, "loss": 0.2111, "step": 21715 }, { "epoch": 2.24, "learning_rate": 2.3742587132736697e-05, "loss": 0.5289, "step": 21716 }, { "epoch": 2.24, "learning_rate": 2.3740179872000212e-05, "loss": 0.4837, "step": 21717 }, { "epoch": 2.24, "learning_rate": 2.3737772653403375e-05, "loss": 0.2001, "step": 21718 }, { "epoch": 2.24, "learning_rate": 2.3735365476962393e-05, "loss": 0.2642, "step": 21719 }, { "epoch": 2.24, "learning_rate": 2.3732958342693457e-05, "loss": 0.9993, "step": 21720 }, { "epoch": 2.24, "learning_rate": 2.3730551250612782e-05, "loss": 0.6013, "step": 21721 }, { "epoch": 2.24, "learning_rate": 2.3728144200736567e-05, "loss": 0.3872, "step": 21722 }, { "epoch": 2.24, "learning_rate": 2.3725737193081025e-05, "loss": 0.3302, "step": 21723 }, { "epoch": 2.24, "learning_rate": 2.372333022766235e-05, "loss": 0.4924, "step": 21724 }, { "epoch": 2.24, "learning_rate": 2.372092330449675e-05, "loss": 0.1553, "step": 21725 }, { "epoch": 2.24, "learning_rate": 2.3718516423600415e-05, "loss": 0.0936, "step": 21726 }, { "epoch": 2.24, "learning_rate": 2.3716109584989562e-05, "loss": 0.5511, "step": 21727 }, { "epoch": 2.24, "learning_rate": 2.3713702788680385e-05, "loss": 0.7085, "step": 21728 }, { "epoch": 2.24, "learning_rate": 2.37112960346891e-05, "loss": 0.2859, "step": 21729 }, { "epoch": 2.24, "learning_rate": 2.370888932303189e-05, "loss": 0.5455, "step": 21730 }, { "epoch": 2.24, "learning_rate": 2.3706482653724967e-05, "loss": 0.3346, "step": 21731 }, { "epoch": 2.24, "learning_rate": 2.370407602678453e-05, "loss": 0.6103, "step": 21732 }, { "epoch": 2.24, "learning_rate": 2.3701669442226783e-05, "loss": 0.3261, "step": 21733 }, { "epoch": 2.24, "learning_rate": 2.369926290006792e-05, "loss": 0.5893, "step": 21734 }, { "epoch": 2.24, "learning_rate": 2.3696856400324145e-05, "loss": 0.2846, "step": 21735 }, { "epoch": 2.24, "learning_rate": 2.369444994301166e-05, "loss": 0.219, "step": 21736 }, { "epoch": 2.24, "learning_rate": 2.369204352814667e-05, "loss": 0.164, "step": 21737 }, { "epoch": 2.24, "learning_rate": 2.3689637155745356e-05, "loss": 0.8264, "step": 21738 }, { "epoch": 2.24, "learning_rate": 2.3687230825823938e-05, "loss": 0.5497, "step": 21739 }, { "epoch": 2.24, "learning_rate": 2.3684824538398596e-05, "loss": 0.5253, "step": 21740 }, { "epoch": 2.24, "learning_rate": 2.368241829348554e-05, "loss": 0.3567, "step": 21741 }, { "epoch": 2.24, "learning_rate": 2.368001209110096e-05, "loss": 0.2061, "step": 21742 }, { "epoch": 2.24, "learning_rate": 2.367760593126107e-05, "loss": 0.6628, "step": 21743 }, { "epoch": 2.24, "learning_rate": 2.367519981398205e-05, "loss": 0.373, "step": 21744 }, { "epoch": 2.24, "learning_rate": 2.3672793739280106e-05, "loss": 0.1223, "step": 21745 }, { "epoch": 2.24, "learning_rate": 2.3670387707171437e-05, "loss": 0.8927, "step": 21746 }, { "epoch": 2.24, "learning_rate": 2.366798171767224e-05, "loss": 0.2736, "step": 21747 }, { "epoch": 2.24, "learning_rate": 2.3665575770798706e-05, "loss": 0.3123, "step": 21748 }, { "epoch": 2.24, "learning_rate": 2.3663169866567027e-05, "loss": 0.2404, "step": 21749 }, { "epoch": 2.24, "learning_rate": 2.3660764004993422e-05, "loss": 0.6218, "step": 21750 }, { "epoch": 2.25, "learning_rate": 2.365835818609406e-05, "loss": 0.3286, "step": 21751 }, { "epoch": 2.25, "learning_rate": 2.3655952409885142e-05, "loss": 0.8313, "step": 21752 }, { "epoch": 2.25, "learning_rate": 2.3653546676382875e-05, "loss": 1.0044, "step": 21753 }, { "epoch": 2.25, "learning_rate": 2.3651140985603445e-05, "loss": 0.5629, "step": 21754 }, { "epoch": 2.25, "learning_rate": 2.3648735337563045e-05, "loss": 0.3612, "step": 21755 }, { "epoch": 2.25, "learning_rate": 2.3646329732277872e-05, "loss": 0.621, "step": 21756 }, { "epoch": 2.25, "learning_rate": 2.364392416976413e-05, "loss": 0.5172, "step": 21757 }, { "epoch": 2.25, "learning_rate": 2.3641518650037996e-05, "loss": 0.4291, "step": 21758 }, { "epoch": 2.25, "learning_rate": 2.3639113173115666e-05, "loss": 0.2739, "step": 21759 }, { "epoch": 2.25, "learning_rate": 2.3636707739013342e-05, "loss": 0.7275, "step": 21760 }, { "epoch": 2.25, "learning_rate": 2.3634302347747222e-05, "loss": 0.23, "step": 21761 }, { "epoch": 2.25, "learning_rate": 2.3631896999333485e-05, "loss": 0.4672, "step": 21762 }, { "epoch": 2.25, "learning_rate": 2.362949169378833e-05, "loss": 0.3551, "step": 21763 }, { "epoch": 2.25, "learning_rate": 2.3627086431127937e-05, "loss": 0.1931, "step": 21764 }, { "epoch": 2.25, "learning_rate": 2.3624681211368506e-05, "loss": 0.4198, "step": 21765 }, { "epoch": 2.25, "learning_rate": 2.3622276034526232e-05, "loss": 0.2054, "step": 21766 }, { "epoch": 2.25, "learning_rate": 2.361987090061731e-05, "loss": 0.2058, "step": 21767 }, { "epoch": 2.25, "learning_rate": 2.3617465809657916e-05, "loss": 0.4383, "step": 21768 }, { "epoch": 2.25, "learning_rate": 2.361506076166425e-05, "loss": 0.4176, "step": 21769 }, { "epoch": 2.25, "learning_rate": 2.3612655756652504e-05, "loss": 0.1817, "step": 21770 }, { "epoch": 2.25, "learning_rate": 2.3610250794638867e-05, "loss": 0.2177, "step": 21771 }, { "epoch": 2.25, "learning_rate": 2.360784587563952e-05, "loss": 0.2031, "step": 21772 }, { "epoch": 2.25, "learning_rate": 2.360544099967066e-05, "loss": 0.5361, "step": 21773 }, { "epoch": 2.25, "learning_rate": 2.3603036166748474e-05, "loss": 0.5305, "step": 21774 }, { "epoch": 2.25, "learning_rate": 2.360063137688917e-05, "loss": 0.2793, "step": 21775 }, { "epoch": 2.25, "learning_rate": 2.35982266301089e-05, "loss": 0.1425, "step": 21776 }, { "epoch": 2.25, "learning_rate": 2.3595821926423878e-05, "loss": 0.2062, "step": 21777 }, { "epoch": 2.25, "learning_rate": 2.359341726585028e-05, "loss": 0.1902, "step": 21778 }, { "epoch": 2.25, "learning_rate": 2.3591012648404295e-05, "loss": 0.4985, "step": 21779 }, { "epoch": 2.25, "learning_rate": 2.3588608074102113e-05, "loss": 0.6229, "step": 21780 }, { "epoch": 2.25, "learning_rate": 2.358620354295993e-05, "loss": 0.2395, "step": 21781 }, { "epoch": 2.25, "learning_rate": 2.3583799054993916e-05, "loss": 0.5212, "step": 21782 }, { "epoch": 2.25, "learning_rate": 2.358139461022027e-05, "loss": 0.5104, "step": 21783 }, { "epoch": 2.25, "learning_rate": 2.3578990208655167e-05, "loss": 0.0607, "step": 21784 }, { "epoch": 2.25, "learning_rate": 2.357658585031481e-05, "loss": 0.4048, "step": 21785 }, { "epoch": 2.25, "learning_rate": 2.3574181535215366e-05, "loss": 0.1227, "step": 21786 }, { "epoch": 2.25, "learning_rate": 2.357177726337304e-05, "loss": 0.4677, "step": 21787 }, { "epoch": 2.25, "learning_rate": 2.3569373034803995e-05, "loss": 0.2926, "step": 21788 }, { "epoch": 2.25, "learning_rate": 2.3566968849524422e-05, "loss": 0.456, "step": 21789 }, { "epoch": 2.25, "learning_rate": 2.3564564707550517e-05, "loss": 0.0622, "step": 21790 }, { "epoch": 2.25, "learning_rate": 2.3562160608898457e-05, "loss": 0.2197, "step": 21791 }, { "epoch": 2.25, "learning_rate": 2.3559756553584424e-05, "loss": 0.2329, "step": 21792 }, { "epoch": 2.25, "learning_rate": 2.3557352541624598e-05, "loss": 0.246, "step": 21793 }, { "epoch": 2.25, "learning_rate": 2.3554948573035168e-05, "loss": 0.4261, "step": 21794 }, { "epoch": 2.25, "learning_rate": 2.3552544647832325e-05, "loss": 0.2923, "step": 21795 }, { "epoch": 2.25, "learning_rate": 2.3550140766032236e-05, "loss": 0.3293, "step": 21796 }, { "epoch": 2.25, "learning_rate": 2.3547736927651087e-05, "loss": 0.4302, "step": 21797 }, { "epoch": 2.25, "learning_rate": 2.354533313270507e-05, "loss": 0.0929, "step": 21798 }, { "epoch": 2.25, "learning_rate": 2.354292938121036e-05, "loss": 0.5077, "step": 21799 }, { "epoch": 2.25, "learning_rate": 2.354052567318314e-05, "loss": 0.5476, "step": 21800 }, { "epoch": 2.25, "learning_rate": 2.353812200863959e-05, "loss": 0.2383, "step": 21801 }, { "epoch": 2.25, "learning_rate": 2.3535718387595885e-05, "loss": 0.6286, "step": 21802 }, { "epoch": 2.25, "learning_rate": 2.353331481006821e-05, "loss": 0.1071, "step": 21803 }, { "epoch": 2.25, "learning_rate": 2.353091127607275e-05, "loss": 0.4259, "step": 21804 }, { "epoch": 2.25, "learning_rate": 2.3528507785625684e-05, "loss": 0.2979, "step": 21805 }, { "epoch": 2.25, "learning_rate": 2.3526104338743186e-05, "loss": 0.8053, "step": 21806 }, { "epoch": 2.25, "learning_rate": 2.3523700935441438e-05, "loss": 0.1696, "step": 21807 }, { "epoch": 2.25, "learning_rate": 2.3521297575736615e-05, "loss": 0.7478, "step": 21808 }, { "epoch": 2.25, "learning_rate": 2.3518894259644915e-05, "loss": 0.4114, "step": 21809 }, { "epoch": 2.25, "learning_rate": 2.3516490987182494e-05, "loss": 0.3683, "step": 21810 }, { "epoch": 2.25, "learning_rate": 2.3514087758365535e-05, "loss": 0.2062, "step": 21811 }, { "epoch": 2.25, "learning_rate": 2.3511684573210235e-05, "loss": 0.6726, "step": 21812 }, { "epoch": 2.25, "learning_rate": 2.3509281431732744e-05, "loss": 0.3311, "step": 21813 }, { "epoch": 2.25, "learning_rate": 2.3506878333949248e-05, "loss": 0.511, "step": 21814 }, { "epoch": 2.25, "learning_rate": 2.350447527987594e-05, "loss": 0.4248, "step": 21815 }, { "epoch": 2.25, "learning_rate": 2.3502072269528974e-05, "loss": 0.607, "step": 21816 }, { "epoch": 2.25, "learning_rate": 2.3499669302924542e-05, "loss": 0.8286, "step": 21817 }, { "epoch": 2.25, "learning_rate": 2.3497266380078808e-05, "loss": 0.2142, "step": 21818 }, { "epoch": 2.25, "learning_rate": 2.3494863501007967e-05, "loss": 0.1842, "step": 21819 }, { "epoch": 2.25, "learning_rate": 2.3492460665728174e-05, "loss": 0.7078, "step": 21820 }, { "epoch": 2.25, "learning_rate": 2.3490057874255613e-05, "loss": 0.0913, "step": 21821 }, { "epoch": 2.25, "learning_rate": 2.348765512660646e-05, "loss": 1.4262, "step": 21822 }, { "epoch": 2.25, "learning_rate": 2.3485252422796898e-05, "loss": 0.3483, "step": 21823 }, { "epoch": 2.25, "learning_rate": 2.3482849762843082e-05, "loss": 0.6859, "step": 21824 }, { "epoch": 2.25, "learning_rate": 2.348044714676121e-05, "loss": 0.1386, "step": 21825 }, { "epoch": 2.25, "learning_rate": 2.347804457456743e-05, "loss": 0.2041, "step": 21826 }, { "epoch": 2.25, "learning_rate": 2.347564204627793e-05, "loss": 0.1825, "step": 21827 }, { "epoch": 2.25, "learning_rate": 2.3473239561908878e-05, "loss": 0.2957, "step": 21828 }, { "epoch": 2.25, "learning_rate": 2.3470837121476458e-05, "loss": 0.7562, "step": 21829 }, { "epoch": 2.25, "learning_rate": 2.3468434724996828e-05, "loss": 0.3064, "step": 21830 }, { "epoch": 2.25, "learning_rate": 2.3466032372486167e-05, "loss": 0.6214, "step": 21831 }, { "epoch": 2.25, "learning_rate": 2.346363006396065e-05, "loss": 0.5679, "step": 21832 }, { "epoch": 2.25, "learning_rate": 2.346122779943645e-05, "loss": 0.4543, "step": 21833 }, { "epoch": 2.25, "learning_rate": 2.3458825578929728e-05, "loss": 0.1456, "step": 21834 }, { "epoch": 2.25, "learning_rate": 2.345642340245666e-05, "loss": 0.2701, "step": 21835 }, { "epoch": 2.25, "learning_rate": 2.345402127003342e-05, "loss": 0.1798, "step": 21836 }, { "epoch": 2.25, "learning_rate": 2.3451619181676183e-05, "loss": 0.2052, "step": 21837 }, { "epoch": 2.25, "learning_rate": 2.3449217137401116e-05, "loss": 0.6049, "step": 21838 }, { "epoch": 2.25, "learning_rate": 2.3446815137224384e-05, "loss": 0.7075, "step": 21839 }, { "epoch": 2.25, "learning_rate": 2.3444413181162154e-05, "loss": 0.378, "step": 21840 }, { "epoch": 2.25, "learning_rate": 2.34420112692306e-05, "loss": 0.7596, "step": 21841 }, { "epoch": 2.25, "learning_rate": 2.343960940144589e-05, "loss": 0.4225, "step": 21842 }, { "epoch": 2.25, "learning_rate": 2.3437207577824205e-05, "loss": 0.4602, "step": 21843 }, { "epoch": 2.25, "learning_rate": 2.3434805798381697e-05, "loss": 0.2854, "step": 21844 }, { "epoch": 2.25, "learning_rate": 2.3432404063134536e-05, "loss": 0.1414, "step": 21845 }, { "epoch": 2.25, "learning_rate": 2.3430002372098897e-05, "loss": 0.5974, "step": 21846 }, { "epoch": 2.26, "learning_rate": 2.342760072529095e-05, "loss": 0.1238, "step": 21847 }, { "epoch": 2.26, "learning_rate": 2.342519912272685e-05, "loss": 0.5482, "step": 21848 }, { "epoch": 2.26, "learning_rate": 2.342279756442277e-05, "loss": 0.8095, "step": 21849 }, { "epoch": 2.26, "learning_rate": 2.3420396050394895e-05, "loss": 0.5751, "step": 21850 }, { "epoch": 2.26, "learning_rate": 2.341799458065936e-05, "loss": 0.6227, "step": 21851 }, { "epoch": 2.26, "learning_rate": 2.3415593155232344e-05, "loss": 0.6769, "step": 21852 }, { "epoch": 2.26, "learning_rate": 2.341319177413002e-05, "loss": 0.4248, "step": 21853 }, { "epoch": 2.26, "learning_rate": 2.341079043736854e-05, "loss": 0.2287, "step": 21854 }, { "epoch": 2.26, "learning_rate": 2.3408389144964083e-05, "loss": 0.49, "step": 21855 }, { "epoch": 2.26, "learning_rate": 2.3405987896932807e-05, "loss": 0.7102, "step": 21856 }, { "epoch": 2.26, "learning_rate": 2.3403586693290876e-05, "loss": 0.3446, "step": 21857 }, { "epoch": 2.26, "learning_rate": 2.3401185534054457e-05, "loss": 0.3293, "step": 21858 }, { "epoch": 2.26, "learning_rate": 2.3398784419239713e-05, "loss": 0.6759, "step": 21859 }, { "epoch": 2.26, "learning_rate": 2.3396383348862803e-05, "loss": 0.5749, "step": 21860 }, { "epoch": 2.26, "learning_rate": 2.3393982322939908e-05, "loss": 0.3777, "step": 21861 }, { "epoch": 2.26, "learning_rate": 2.3391581341487167e-05, "loss": 1.0782, "step": 21862 }, { "epoch": 2.26, "learning_rate": 2.3389180404520765e-05, "loss": 0.7877, "step": 21863 }, { "epoch": 2.26, "learning_rate": 2.3386779512056845e-05, "loss": 0.1512, "step": 21864 }, { "epoch": 2.26, "learning_rate": 2.3384378664111577e-05, "loss": 0.5713, "step": 21865 }, { "epoch": 2.26, "learning_rate": 2.3381977860701126e-05, "loss": 0.4772, "step": 21866 }, { "epoch": 2.26, "learning_rate": 2.3379577101841654e-05, "loss": 0.4723, "step": 21867 }, { "epoch": 2.26, "learning_rate": 2.3377176387549318e-05, "loss": 0.1084, "step": 21868 }, { "epoch": 2.26, "learning_rate": 2.3374775717840277e-05, "loss": 0.785, "step": 21869 }, { "epoch": 2.26, "learning_rate": 2.3372375092730695e-05, "loss": 1.213, "step": 21870 }, { "epoch": 2.26, "learning_rate": 2.3369974512236745e-05, "loss": 0.5153, "step": 21871 }, { "epoch": 2.26, "learning_rate": 2.3367573976374566e-05, "loss": 0.4445, "step": 21872 }, { "epoch": 2.26, "learning_rate": 2.3365173485160324e-05, "loss": 0.186, "step": 21873 }, { "epoch": 2.26, "learning_rate": 2.3362773038610183e-05, "loss": 0.8062, "step": 21874 }, { "epoch": 2.26, "learning_rate": 2.3360372636740316e-05, "loss": 0.3008, "step": 21875 }, { "epoch": 2.26, "learning_rate": 2.3357972279566855e-05, "loss": 0.5253, "step": 21876 }, { "epoch": 2.26, "learning_rate": 2.3355571967105974e-05, "loss": 0.4397, "step": 21877 }, { "epoch": 2.26, "learning_rate": 2.3353171699373823e-05, "loss": 0.5068, "step": 21878 }, { "epoch": 2.26, "learning_rate": 2.335077147638656e-05, "loss": 0.3112, "step": 21879 }, { "epoch": 2.26, "learning_rate": 2.3348371298160353e-05, "loss": 0.2528, "step": 21880 }, { "epoch": 2.26, "learning_rate": 2.334597116471136e-05, "loss": 0.2267, "step": 21881 }, { "epoch": 2.26, "learning_rate": 2.3343571076055724e-05, "loss": 0.4787, "step": 21882 }, { "epoch": 2.26, "learning_rate": 2.3341171032209614e-05, "loss": 0.2871, "step": 21883 }, { "epoch": 2.26, "learning_rate": 2.3338771033189182e-05, "loss": 0.316, "step": 21884 }, { "epoch": 2.26, "learning_rate": 2.333637107901059e-05, "loss": 0.3221, "step": 21885 }, { "epoch": 2.26, "learning_rate": 2.3333971169689977e-05, "loss": 0.2559, "step": 21886 }, { "epoch": 2.26, "learning_rate": 2.3331571305243514e-05, "loss": 0.6142, "step": 21887 }, { "epoch": 2.26, "learning_rate": 2.3329171485687365e-05, "loss": 0.4131, "step": 21888 }, { "epoch": 2.26, "learning_rate": 2.3326771711037663e-05, "loss": 0.4038, "step": 21889 }, { "epoch": 2.26, "learning_rate": 2.332437198131057e-05, "loss": 0.5606, "step": 21890 }, { "epoch": 2.26, "learning_rate": 2.332197229652225e-05, "loss": 0.579, "step": 21891 }, { "epoch": 2.26, "learning_rate": 2.331957265668884e-05, "loss": 0.2835, "step": 21892 }, { "epoch": 2.26, "learning_rate": 2.331717306182651e-05, "loss": 0.1412, "step": 21893 }, { "epoch": 2.26, "learning_rate": 2.3314773511951404e-05, "loss": 0.5129, "step": 21894 }, { "epoch": 2.26, "learning_rate": 2.3312374007079684e-05, "loss": 0.1963, "step": 21895 }, { "epoch": 2.26, "learning_rate": 2.3309974547227497e-05, "loss": 0.5084, "step": 21896 }, { "epoch": 2.26, "learning_rate": 2.3307575132410987e-05, "loss": 0.2369, "step": 21897 }, { "epoch": 2.26, "learning_rate": 2.330517576264632e-05, "loss": 0.6819, "step": 21898 }, { "epoch": 2.26, "learning_rate": 2.330277643794965e-05, "loss": 0.7118, "step": 21899 }, { "epoch": 2.26, "learning_rate": 2.330037715833712e-05, "loss": 0.6315, "step": 21900 }, { "epoch": 2.26, "learning_rate": 2.3297977923824883e-05, "loss": 0.2578, "step": 21901 }, { "epoch": 2.26, "learning_rate": 2.329557873442909e-05, "loss": 0.4598, "step": 21902 }, { "epoch": 2.26, "learning_rate": 2.3293179590165888e-05, "loss": 0.4425, "step": 21903 }, { "epoch": 2.26, "learning_rate": 2.329078049105143e-05, "loss": 0.2675, "step": 21904 }, { "epoch": 2.26, "learning_rate": 2.328838143710188e-05, "loss": 0.7224, "step": 21905 }, { "epoch": 2.26, "learning_rate": 2.328598242833336e-05, "loss": 0.3858, "step": 21906 }, { "epoch": 2.26, "learning_rate": 2.3283583464762044e-05, "loss": 0.6767, "step": 21907 }, { "epoch": 2.26, "learning_rate": 2.3281184546404067e-05, "loss": 0.0165, "step": 21908 }, { "epoch": 2.26, "learning_rate": 2.3278785673275594e-05, "loss": 0.5013, "step": 21909 }, { "epoch": 2.26, "learning_rate": 2.3276386845392754e-05, "loss": 0.1966, "step": 21910 }, { "epoch": 2.26, "learning_rate": 2.3273988062771706e-05, "loss": 0.1243, "step": 21911 }, { "epoch": 2.26, "learning_rate": 2.3271589325428597e-05, "loss": 0.6347, "step": 21912 }, { "epoch": 2.26, "learning_rate": 2.3269190633379586e-05, "loss": 0.7, "step": 21913 }, { "epoch": 2.26, "learning_rate": 2.3266791986640797e-05, "loss": 0.1731, "step": 21914 }, { "epoch": 2.26, "learning_rate": 2.3264393385228394e-05, "loss": 0.4092, "step": 21915 }, { "epoch": 2.26, "learning_rate": 2.326199482915851e-05, "loss": 0.1656, "step": 21916 }, { "epoch": 2.26, "learning_rate": 2.3259596318447304e-05, "loss": 0.3407, "step": 21917 }, { "epoch": 2.26, "learning_rate": 2.325719785311092e-05, "loss": 0.281, "step": 21918 }, { "epoch": 2.26, "learning_rate": 2.3254799433165507e-05, "loss": 0.157, "step": 21919 }, { "epoch": 2.26, "learning_rate": 2.32524010586272e-05, "loss": 0.1234, "step": 21920 }, { "epoch": 2.26, "learning_rate": 2.325000272951215e-05, "loss": 0.1207, "step": 21921 }, { "epoch": 2.26, "learning_rate": 2.32476044458365e-05, "loss": 0.5519, "step": 21922 }, { "epoch": 2.26, "learning_rate": 2.3245206207616404e-05, "loss": 0.2248, "step": 21923 }, { "epoch": 2.26, "learning_rate": 2.3242808014867992e-05, "loss": 0.1584, "step": 21924 }, { "epoch": 2.26, "learning_rate": 2.324040986760743e-05, "loss": 0.3515, "step": 21925 }, { "epoch": 2.26, "learning_rate": 2.3238011765850836e-05, "loss": 0.4977, "step": 21926 }, { "epoch": 2.26, "learning_rate": 2.323561370961436e-05, "loss": 0.3036, "step": 21927 }, { "epoch": 2.26, "learning_rate": 2.3233215698914157e-05, "loss": 0.6069, "step": 21928 }, { "epoch": 2.26, "learning_rate": 2.3230817733766364e-05, "loss": 0.4563, "step": 21929 }, { "epoch": 2.26, "learning_rate": 2.3228419814187115e-05, "loss": 0.545, "step": 21930 }, { "epoch": 2.26, "learning_rate": 2.3226021940192563e-05, "loss": 1.0384, "step": 21931 }, { "epoch": 2.26, "learning_rate": 2.3223624111798843e-05, "loss": 0.1996, "step": 21932 }, { "epoch": 2.26, "learning_rate": 2.3221226329022107e-05, "loss": 0.5334, "step": 21933 }, { "epoch": 2.26, "learning_rate": 2.3218828591878484e-05, "loss": 0.0978, "step": 21934 }, { "epoch": 2.26, "learning_rate": 2.321643090038412e-05, "loss": 0.4794, "step": 21935 }, { "epoch": 2.26, "learning_rate": 2.3214033254555154e-05, "loss": 0.1568, "step": 21936 }, { "epoch": 2.26, "learning_rate": 2.3211635654407737e-05, "loss": 0.2201, "step": 21937 }, { "epoch": 2.26, "learning_rate": 2.3209238099957998e-05, "loss": 0.3529, "step": 21938 }, { "epoch": 2.26, "learning_rate": 2.320684059122208e-05, "loss": 1.1336, "step": 21939 }, { "epoch": 2.26, "learning_rate": 2.3204443128216118e-05, "loss": 0.0479, "step": 21940 }, { "epoch": 2.26, "learning_rate": 2.3202045710956253e-05, "loss": 0.9072, "step": 21941 }, { "epoch": 2.26, "learning_rate": 2.3199648339458627e-05, "loss": 0.7844, "step": 21942 }, { "epoch": 2.27, "learning_rate": 2.319725101373938e-05, "loss": 0.591, "step": 21943 }, { "epoch": 2.27, "learning_rate": 2.3194853733814643e-05, "loss": 0.2335, "step": 21944 }, { "epoch": 2.27, "learning_rate": 2.319245649970056e-05, "loss": 0.541, "step": 21945 }, { "epoch": 2.27, "learning_rate": 2.3190059311413267e-05, "loss": 0.3357, "step": 21946 }, { "epoch": 2.27, "learning_rate": 2.3187662168968905e-05, "loss": 0.8596, "step": 21947 }, { "epoch": 2.27, "learning_rate": 2.31852650723836e-05, "loss": 0.359, "step": 21948 }, { "epoch": 2.27, "learning_rate": 2.31828680216735e-05, "loss": 0.4273, "step": 21949 }, { "epoch": 2.27, "learning_rate": 2.318047101685474e-05, "loss": 0.6467, "step": 21950 }, { "epoch": 2.27, "learning_rate": 2.317807405794346e-05, "loss": 0.3043, "step": 21951 }, { "epoch": 2.27, "learning_rate": 2.3175677144955777e-05, "loss": 0.3203, "step": 21952 }, { "epoch": 2.27, "learning_rate": 2.317328027790785e-05, "loss": 0.3713, "step": 21953 }, { "epoch": 2.27, "learning_rate": 2.317088345681579e-05, "loss": 0.1972, "step": 21954 }, { "epoch": 2.27, "learning_rate": 2.3168486681695747e-05, "loss": 0.5071, "step": 21955 }, { "epoch": 2.27, "learning_rate": 2.3166089952563855e-05, "loss": 0.7452, "step": 21956 }, { "epoch": 2.27, "learning_rate": 2.316369326943625e-05, "loss": 0.1724, "step": 21957 }, { "epoch": 2.27, "learning_rate": 2.3161296632329062e-05, "loss": 0.1434, "step": 21958 }, { "epoch": 2.27, "learning_rate": 2.315890004125842e-05, "loss": 0.6933, "step": 21959 }, { "epoch": 2.27, "learning_rate": 2.3156503496240462e-05, "loss": 0.6971, "step": 21960 }, { "epoch": 2.27, "learning_rate": 2.315410699729133e-05, "loss": 0.4184, "step": 21961 }, { "epoch": 2.27, "learning_rate": 2.3151710544427138e-05, "loss": 0.5417, "step": 21962 }, { "epoch": 2.27, "learning_rate": 2.3149314137664044e-05, "loss": 0.4109, "step": 21963 }, { "epoch": 2.27, "learning_rate": 2.314691777701815e-05, "loss": 0.0811, "step": 21964 }, { "epoch": 2.27, "learning_rate": 2.3144521462505603e-05, "loss": 0.6679, "step": 21965 }, { "epoch": 2.27, "learning_rate": 2.314212519414253e-05, "loss": 0.3057, "step": 21966 }, { "epoch": 2.27, "learning_rate": 2.3139728971945077e-05, "loss": 0.1327, "step": 21967 }, { "epoch": 2.27, "learning_rate": 2.3137332795929356e-05, "loss": 0.5994, "step": 21968 }, { "epoch": 2.27, "learning_rate": 2.3134936666111503e-05, "loss": 0.3884, "step": 21969 }, { "epoch": 2.27, "learning_rate": 2.313254058250765e-05, "loss": 0.5966, "step": 21970 }, { "epoch": 2.27, "learning_rate": 2.3130144545133933e-05, "loss": 0.3508, "step": 21971 }, { "epoch": 2.27, "learning_rate": 2.312774855400647e-05, "loss": 0.4822, "step": 21972 }, { "epoch": 2.27, "learning_rate": 2.3125352609141395e-05, "loss": 0.0684, "step": 21973 }, { "epoch": 2.27, "learning_rate": 2.3122956710554843e-05, "loss": 0.7079, "step": 21974 }, { "epoch": 2.27, "learning_rate": 2.3120560858262938e-05, "loss": 0.5817, "step": 21975 }, { "epoch": 2.27, "learning_rate": 2.311816505228181e-05, "loss": 0.831, "step": 21976 }, { "epoch": 2.27, "learning_rate": 2.3115769292627584e-05, "loss": 0.3007, "step": 21977 }, { "epoch": 2.27, "learning_rate": 2.311337357931638e-05, "loss": 0.2423, "step": 21978 }, { "epoch": 2.27, "learning_rate": 2.3110977912364337e-05, "loss": 0.4912, "step": 21979 }, { "epoch": 2.27, "learning_rate": 2.310858229178758e-05, "loss": 0.5864, "step": 21980 }, { "epoch": 2.27, "learning_rate": 2.3106186717602238e-05, "loss": 0.4911, "step": 21981 }, { "epoch": 2.27, "learning_rate": 2.3103791189824423e-05, "loss": 0.7236, "step": 21982 }, { "epoch": 2.27, "learning_rate": 2.310139570847028e-05, "loss": 0.2315, "step": 21983 }, { "epoch": 2.27, "learning_rate": 2.309900027355592e-05, "loss": 0.3886, "step": 21984 }, { "epoch": 2.27, "learning_rate": 2.309660488509749e-05, "loss": 0.278, "step": 21985 }, { "epoch": 2.27, "learning_rate": 2.3094209543111086e-05, "loss": 0.3481, "step": 21986 }, { "epoch": 2.27, "learning_rate": 2.3091814247612853e-05, "loss": 0.1316, "step": 21987 }, { "epoch": 2.27, "learning_rate": 2.3089418998618924e-05, "loss": 0.5043, "step": 21988 }, { "epoch": 2.27, "learning_rate": 2.3087023796145394e-05, "loss": 0.944, "step": 21989 }, { "epoch": 2.27, "learning_rate": 2.3084628640208402e-05, "loss": 0.9266, "step": 21990 }, { "epoch": 2.27, "learning_rate": 2.3082233530824077e-05, "loss": 0.4224, "step": 21991 }, { "epoch": 2.27, "learning_rate": 2.3079838468008535e-05, "loss": 0.4431, "step": 21992 }, { "epoch": 2.27, "learning_rate": 2.3077443451777895e-05, "loss": 0.193, "step": 21993 }, { "epoch": 2.27, "learning_rate": 2.3075048482148294e-05, "loss": 0.3672, "step": 21994 }, { "epoch": 2.27, "learning_rate": 2.307265355913585e-05, "loss": 0.1234, "step": 21995 }, { "epoch": 2.27, "learning_rate": 2.307025868275667e-05, "loss": 0.1206, "step": 21996 }, { "epoch": 2.27, "learning_rate": 2.306786385302689e-05, "loss": 0.587, "step": 21997 }, { "epoch": 2.27, "learning_rate": 2.306546906996263e-05, "loss": 0.3878, "step": 21998 }, { "epoch": 2.27, "learning_rate": 2.3063074333580015e-05, "loss": 0.1462, "step": 21999 }, { "epoch": 2.27, "learning_rate": 2.306067964389515e-05, "loss": 0.5166, "step": 22000 }, { "epoch": 2.27, "learning_rate": 2.3058285000924186e-05, "loss": 0.4561, "step": 22001 }, { "epoch": 2.27, "learning_rate": 2.3055890404683208e-05, "loss": 0.6037, "step": 22002 }, { "epoch": 2.27, "learning_rate": 2.3053495855188348e-05, "loss": 0.6982, "step": 22003 }, { "epoch": 2.27, "learning_rate": 2.3051101352455732e-05, "loss": 0.3879, "step": 22004 }, { "epoch": 2.27, "learning_rate": 2.3048706896501482e-05, "loss": 0.2804, "step": 22005 }, { "epoch": 2.27, "learning_rate": 2.3046312487341703e-05, "loss": 0.6791, "step": 22006 }, { "epoch": 2.27, "learning_rate": 2.304391812499252e-05, "loss": 0.1806, "step": 22007 }, { "epoch": 2.27, "learning_rate": 2.3041523809470054e-05, "loss": 0.4213, "step": 22008 }, { "epoch": 2.27, "learning_rate": 2.303912954079043e-05, "loss": 0.4658, "step": 22009 }, { "epoch": 2.27, "learning_rate": 2.303673531896975e-05, "loss": 0.5125, "step": 22010 }, { "epoch": 2.27, "learning_rate": 2.3034341144024135e-05, "loss": 0.4719, "step": 22011 }, { "epoch": 2.27, "learning_rate": 2.303194701596971e-05, "loss": 0.4614, "step": 22012 }, { "epoch": 2.27, "learning_rate": 2.30295529348226e-05, "loss": 0.5858, "step": 22013 }, { "epoch": 2.27, "learning_rate": 2.302715890059889e-05, "loss": 0.2285, "step": 22014 }, { "epoch": 2.27, "learning_rate": 2.3024764913314727e-05, "loss": 0.5152, "step": 22015 }, { "epoch": 2.27, "learning_rate": 2.3022370972986207e-05, "loss": 0.3699, "step": 22016 }, { "epoch": 2.27, "learning_rate": 2.3019977079629452e-05, "loss": 0.6498, "step": 22017 }, { "epoch": 2.27, "learning_rate": 2.3017583233260576e-05, "loss": 0.3925, "step": 22018 }, { "epoch": 2.27, "learning_rate": 2.3015189433895706e-05, "loss": 0.1727, "step": 22019 }, { "epoch": 2.27, "learning_rate": 2.3012795681550937e-05, "loss": 0.4263, "step": 22020 }, { "epoch": 2.27, "learning_rate": 2.3010401976242397e-05, "loss": 0.3082, "step": 22021 }, { "epoch": 2.27, "learning_rate": 2.300800831798619e-05, "loss": 0.5881, "step": 22022 }, { "epoch": 2.27, "learning_rate": 2.3005614706798447e-05, "loss": 0.2708, "step": 22023 }, { "epoch": 2.27, "learning_rate": 2.300322114269526e-05, "loss": 0.6436, "step": 22024 }, { "epoch": 2.27, "learning_rate": 2.300082762569275e-05, "loss": 0.1442, "step": 22025 }, { "epoch": 2.27, "learning_rate": 2.2998434155807045e-05, "loss": 0.3157, "step": 22026 }, { "epoch": 2.27, "learning_rate": 2.299604073305423e-05, "loss": 0.1844, "step": 22027 }, { "epoch": 2.27, "learning_rate": 2.2993647357450433e-05, "loss": 0.2823, "step": 22028 }, { "epoch": 2.27, "learning_rate": 2.2991254029011765e-05, "loss": 0.1028, "step": 22029 }, { "epoch": 2.27, "learning_rate": 2.298886074775433e-05, "loss": 0.0755, "step": 22030 }, { "epoch": 2.27, "learning_rate": 2.298646751369424e-05, "loss": 0.9304, "step": 22031 }, { "epoch": 2.27, "learning_rate": 2.2984074326847613e-05, "loss": 0.2208, "step": 22032 }, { "epoch": 2.27, "learning_rate": 2.2981681187230563e-05, "loss": 0.2382, "step": 22033 }, { "epoch": 2.27, "learning_rate": 2.297928809485919e-05, "loss": 0.5589, "step": 22034 }, { "epoch": 2.27, "learning_rate": 2.2976895049749602e-05, "loss": 0.2013, "step": 22035 }, { "epoch": 2.27, "learning_rate": 2.2974502051917913e-05, "loss": 0.8466, "step": 22036 }, { "epoch": 2.27, "learning_rate": 2.2972109101380243e-05, "loss": 0.598, "step": 22037 }, { "epoch": 2.27, "learning_rate": 2.296971619815268e-05, "loss": 0.0814, "step": 22038 }, { "epoch": 2.28, "learning_rate": 2.2967323342251356e-05, "loss": 0.7076, "step": 22039 }, { "epoch": 2.28, "learning_rate": 2.2964930533692353e-05, "loss": 0.337, "step": 22040 }, { "epoch": 2.28, "learning_rate": 2.2962537772491792e-05, "loss": 0.5444, "step": 22041 }, { "epoch": 2.28, "learning_rate": 2.296014505866578e-05, "loss": 0.7578, "step": 22042 }, { "epoch": 2.28, "learning_rate": 2.295775239223043e-05, "loss": 0.7236, "step": 22043 }, { "epoch": 2.28, "learning_rate": 2.2955359773201837e-05, "loss": 0.4425, "step": 22044 }, { "epoch": 2.28, "learning_rate": 2.2952967201596113e-05, "loss": 0.1763, "step": 22045 }, { "epoch": 2.28, "learning_rate": 2.2950574677429367e-05, "loss": 0.7589, "step": 22046 }, { "epoch": 2.28, "learning_rate": 2.2948182200717707e-05, "loss": 0.2378, "step": 22047 }, { "epoch": 2.28, "learning_rate": 2.2945789771477232e-05, "loss": 0.2041, "step": 22048 }, { "epoch": 2.28, "learning_rate": 2.2943397389724046e-05, "loss": 0.3646, "step": 22049 }, { "epoch": 2.28, "learning_rate": 2.2941005055474258e-05, "loss": 0.154, "step": 22050 }, { "epoch": 2.28, "learning_rate": 2.2938612768743986e-05, "loss": 0.3558, "step": 22051 }, { "epoch": 2.28, "learning_rate": 2.293622052954931e-05, "loss": 0.3346, "step": 22052 }, { "epoch": 2.28, "learning_rate": 2.2933828337906346e-05, "loss": 0.2828, "step": 22053 }, { "epoch": 2.28, "learning_rate": 2.293143619383119e-05, "loss": 0.1197, "step": 22054 }, { "epoch": 2.28, "learning_rate": 2.2929044097339956e-05, "loss": 0.1535, "step": 22055 }, { "epoch": 2.28, "learning_rate": 2.292665204844874e-05, "loss": 0.3753, "step": 22056 }, { "epoch": 2.28, "learning_rate": 2.2924260047173654e-05, "loss": 0.285, "step": 22057 }, { "epoch": 2.28, "learning_rate": 2.2921868093530788e-05, "loss": 1.2076, "step": 22058 }, { "epoch": 2.28, "learning_rate": 2.291947618753625e-05, "loss": 0.6002, "step": 22059 }, { "epoch": 2.28, "learning_rate": 2.2917084329206142e-05, "loss": 0.5198, "step": 22060 }, { "epoch": 2.28, "learning_rate": 2.2914692518556573e-05, "loss": 0.0867, "step": 22061 }, { "epoch": 2.28, "learning_rate": 2.2912300755603622e-05, "loss": 0.1563, "step": 22062 }, { "epoch": 2.28, "learning_rate": 2.2909909040363414e-05, "loss": 0.4413, "step": 22063 }, { "epoch": 2.28, "learning_rate": 2.2907517372852045e-05, "loss": 0.429, "step": 22064 }, { "epoch": 2.28, "learning_rate": 2.29051257530856e-05, "loss": 0.2152, "step": 22065 }, { "epoch": 2.28, "learning_rate": 2.2902734181080187e-05, "loss": 0.2087, "step": 22066 }, { "epoch": 2.28, "learning_rate": 2.2900342656851916e-05, "loss": 0.5345, "step": 22067 }, { "epoch": 2.28, "learning_rate": 2.289795118041687e-05, "loss": 0.5794, "step": 22068 }, { "epoch": 2.28, "learning_rate": 2.289555975179115e-05, "loss": 0.2042, "step": 22069 }, { "epoch": 2.28, "learning_rate": 2.2893168370990865e-05, "loss": 0.3754, "step": 22070 }, { "epoch": 2.28, "learning_rate": 2.289077703803211e-05, "loss": 0.5934, "step": 22071 }, { "epoch": 2.28, "learning_rate": 2.2888385752930977e-05, "loss": 0.5927, "step": 22072 }, { "epoch": 2.28, "learning_rate": 2.2885994515703563e-05, "loss": 0.2024, "step": 22073 }, { "epoch": 2.28, "learning_rate": 2.2883603326365976e-05, "loss": 0.3549, "step": 22074 }, { "epoch": 2.28, "learning_rate": 2.288121218493431e-05, "loss": 0.3641, "step": 22075 }, { "epoch": 2.28, "learning_rate": 2.2878821091424656e-05, "loss": 0.2864, "step": 22076 }, { "epoch": 2.28, "learning_rate": 2.2876430045853114e-05, "loss": 0.2198, "step": 22077 }, { "epoch": 2.28, "learning_rate": 2.287403904823577e-05, "loss": 1.109, "step": 22078 }, { "epoch": 2.28, "learning_rate": 2.2871648098588727e-05, "loss": 0.4807, "step": 22079 }, { "epoch": 2.28, "learning_rate": 2.2869257196928086e-05, "loss": 0.3295, "step": 22080 }, { "epoch": 2.28, "learning_rate": 2.286686634326994e-05, "loss": 0.1113, "step": 22081 }, { "epoch": 2.28, "learning_rate": 2.2864475537630375e-05, "loss": 0.2242, "step": 22082 }, { "epoch": 2.28, "learning_rate": 2.286208478002549e-05, "loss": 0.1405, "step": 22083 }, { "epoch": 2.28, "learning_rate": 2.285969407047138e-05, "loss": 0.2228, "step": 22084 }, { "epoch": 2.28, "learning_rate": 2.2857303408984147e-05, "loss": 0.4699, "step": 22085 }, { "epoch": 2.28, "learning_rate": 2.285491279557987e-05, "loss": 0.1443, "step": 22086 }, { "epoch": 2.28, "learning_rate": 2.2852522230274647e-05, "loss": 0.2087, "step": 22087 }, { "epoch": 2.28, "learning_rate": 2.2850131713084575e-05, "loss": 0.295, "step": 22088 }, { "epoch": 2.28, "learning_rate": 2.284774124402575e-05, "loss": 0.2799, "step": 22089 }, { "epoch": 2.28, "learning_rate": 2.2845350823114248e-05, "loss": 0.3902, "step": 22090 }, { "epoch": 2.28, "learning_rate": 2.2842960450366177e-05, "loss": 0.5867, "step": 22091 }, { "epoch": 2.28, "learning_rate": 2.2840570125797614e-05, "loss": 0.4246, "step": 22092 }, { "epoch": 2.28, "learning_rate": 2.2838179849424664e-05, "loss": 0.2505, "step": 22093 }, { "epoch": 2.28, "learning_rate": 2.2835789621263406e-05, "loss": 0.1093, "step": 22094 }, { "epoch": 2.28, "learning_rate": 2.283339944132994e-05, "loss": 0.5338, "step": 22095 }, { "epoch": 2.28, "learning_rate": 2.2831009309640348e-05, "loss": 0.4265, "step": 22096 }, { "epoch": 2.28, "learning_rate": 2.2828619226210725e-05, "loss": 0.8474, "step": 22097 }, { "epoch": 2.28, "learning_rate": 2.2826229191057156e-05, "loss": 0.6682, "step": 22098 }, { "epoch": 2.28, "learning_rate": 2.2823839204195747e-05, "loss": 0.5977, "step": 22099 }, { "epoch": 2.28, "learning_rate": 2.282144926564256e-05, "loss": 0.6819, "step": 22100 }, { "epoch": 2.28, "learning_rate": 2.281905937541371e-05, "loss": 0.1801, "step": 22101 }, { "epoch": 2.28, "learning_rate": 2.2816669533525264e-05, "loss": 0.5054, "step": 22102 }, { "epoch": 2.28, "learning_rate": 2.2814279739993314e-05, "loss": 0.0768, "step": 22103 }, { "epoch": 2.28, "learning_rate": 2.2811889994833956e-05, "loss": 0.5218, "step": 22104 }, { "epoch": 2.28, "learning_rate": 2.2809500298063273e-05, "loss": 0.5033, "step": 22105 }, { "epoch": 2.28, "learning_rate": 2.2807110649697352e-05, "loss": 0.4777, "step": 22106 }, { "epoch": 2.28, "learning_rate": 2.2804721049752272e-05, "loss": 0.3028, "step": 22107 }, { "epoch": 2.28, "learning_rate": 2.280233149824413e-05, "loss": 0.2111, "step": 22108 }, { "epoch": 2.28, "learning_rate": 2.2799941995189015e-05, "loss": 0.2824, "step": 22109 }, { "epoch": 2.28, "learning_rate": 2.2797552540603e-05, "loss": 0.2973, "step": 22110 }, { "epoch": 2.28, "learning_rate": 2.2795163134502175e-05, "loss": 0.8267, "step": 22111 }, { "epoch": 2.28, "learning_rate": 2.2792773776902623e-05, "loss": 0.7308, "step": 22112 }, { "epoch": 2.28, "learning_rate": 2.279038446782044e-05, "loss": 0.1532, "step": 22113 }, { "epoch": 2.28, "learning_rate": 2.2787995207271706e-05, "loss": 0.426, "step": 22114 }, { "epoch": 2.28, "learning_rate": 2.27856059952725e-05, "loss": 0.5017, "step": 22115 }, { "epoch": 2.28, "learning_rate": 2.2783216831838893e-05, "loss": 0.4415, "step": 22116 }, { "epoch": 2.28, "learning_rate": 2.2780827716986987e-05, "loss": 0.1981, "step": 22117 }, { "epoch": 2.28, "learning_rate": 2.277843865073286e-05, "loss": 0.3103, "step": 22118 }, { "epoch": 2.28, "learning_rate": 2.27760496330926e-05, "loss": 0.563, "step": 22119 }, { "epoch": 2.28, "learning_rate": 2.2773660664082276e-05, "loss": 0.1845, "step": 22120 }, { "epoch": 2.28, "learning_rate": 2.2771271743717978e-05, "loss": 0.675, "step": 22121 }, { "epoch": 2.28, "learning_rate": 2.2768882872015787e-05, "loss": 0.2804, "step": 22122 }, { "epoch": 2.28, "learning_rate": 2.276649404899179e-05, "loss": 0.2136, "step": 22123 }, { "epoch": 2.28, "learning_rate": 2.276410527466206e-05, "loss": 0.1744, "step": 22124 }, { "epoch": 2.28, "learning_rate": 2.2761716549042677e-05, "loss": 0.5271, "step": 22125 }, { "epoch": 2.28, "learning_rate": 2.2759327872149735e-05, "loss": 0.074, "step": 22126 }, { "epoch": 2.28, "learning_rate": 2.2756939243999295e-05, "loss": 0.6599, "step": 22127 }, { "epoch": 2.28, "learning_rate": 2.2754550664607445e-05, "loss": 0.1448, "step": 22128 }, { "epoch": 2.28, "learning_rate": 2.275216213399027e-05, "loss": 0.4944, "step": 22129 }, { "epoch": 2.28, "learning_rate": 2.274977365216384e-05, "loss": 0.5999, "step": 22130 }, { "epoch": 2.28, "learning_rate": 2.2747385219144233e-05, "loss": 0.2742, "step": 22131 }, { "epoch": 2.28, "learning_rate": 2.2744996834947534e-05, "loss": 0.0812, "step": 22132 }, { "epoch": 2.28, "learning_rate": 2.2742608499589824e-05, "loss": 0.1249, "step": 22133 }, { "epoch": 2.29, "learning_rate": 2.2740220213087168e-05, "loss": 0.2404, "step": 22134 }, { "epoch": 2.29, "learning_rate": 2.2737831975455654e-05, "loss": 0.1377, "step": 22135 }, { "epoch": 2.29, "learning_rate": 2.2735443786711354e-05, "loss": 0.3066, "step": 22136 }, { "epoch": 2.29, "learning_rate": 2.2733055646870352e-05, "loss": 0.7446, "step": 22137 }, { "epoch": 2.29, "learning_rate": 2.273066755594872e-05, "loss": 0.6219, "step": 22138 }, { "epoch": 2.29, "learning_rate": 2.2728279513962536e-05, "loss": 0.6375, "step": 22139 }, { "epoch": 2.29, "learning_rate": 2.2725891520927865e-05, "loss": 0.7263, "step": 22140 }, { "epoch": 2.29, "learning_rate": 2.272350357686079e-05, "loss": 0.233, "step": 22141 }, { "epoch": 2.29, "learning_rate": 2.2721115681777388e-05, "loss": 0.5669, "step": 22142 }, { "epoch": 2.29, "learning_rate": 2.2718727835693735e-05, "loss": 0.6687, "step": 22143 }, { "epoch": 2.29, "learning_rate": 2.2716340038625896e-05, "loss": 0.0815, "step": 22144 }, { "epoch": 2.29, "learning_rate": 2.2713952290589955e-05, "loss": 0.1215, "step": 22145 }, { "epoch": 2.29, "learning_rate": 2.2711564591601982e-05, "loss": 0.4032, "step": 22146 }, { "epoch": 2.29, "learning_rate": 2.2709176941678054e-05, "loss": 0.6214, "step": 22147 }, { "epoch": 2.29, "learning_rate": 2.2706789340834238e-05, "loss": 0.3699, "step": 22148 }, { "epoch": 2.29, "learning_rate": 2.2704401789086605e-05, "loss": 0.9857, "step": 22149 }, { "epoch": 2.29, "learning_rate": 2.2702014286451234e-05, "loss": 0.4837, "step": 22150 }, { "epoch": 2.29, "learning_rate": 2.2699626832944204e-05, "loss": 0.2272, "step": 22151 }, { "epoch": 2.29, "learning_rate": 2.2697239428581575e-05, "loss": 0.2011, "step": 22152 }, { "epoch": 2.29, "learning_rate": 2.2694852073379427e-05, "loss": 0.3025, "step": 22153 }, { "epoch": 2.29, "learning_rate": 2.2692464767353816e-05, "loss": 0.4135, "step": 22154 }, { "epoch": 2.29, "learning_rate": 2.269007751052082e-05, "loss": 0.5874, "step": 22155 }, { "epoch": 2.29, "learning_rate": 2.2687690302896514e-05, "loss": 0.1591, "step": 22156 }, { "epoch": 2.29, "learning_rate": 2.2685303144496973e-05, "loss": 0.321, "step": 22157 }, { "epoch": 2.29, "learning_rate": 2.268291603533825e-05, "loss": 0.6273, "step": 22158 }, { "epoch": 2.29, "learning_rate": 2.2680528975436427e-05, "loss": 0.5102, "step": 22159 }, { "epoch": 2.29, "learning_rate": 2.267814196480757e-05, "loss": 0.1876, "step": 22160 }, { "epoch": 2.29, "learning_rate": 2.2675755003467754e-05, "loss": 0.5508, "step": 22161 }, { "epoch": 2.29, "learning_rate": 2.2673368091433037e-05, "loss": 0.2691, "step": 22162 }, { "epoch": 2.29, "learning_rate": 2.2670981228719487e-05, "loss": 0.7539, "step": 22163 }, { "epoch": 2.29, "learning_rate": 2.2668594415343197e-05, "loss": 0.286, "step": 22164 }, { "epoch": 2.29, "learning_rate": 2.2666207651320194e-05, "loss": 0.6183, "step": 22165 }, { "epoch": 2.29, "learning_rate": 2.266382093666657e-05, "loss": 0.7279, "step": 22166 }, { "epoch": 2.29, "learning_rate": 2.2661434271398388e-05, "loss": 0.48, "step": 22167 }, { "epoch": 2.29, "learning_rate": 2.265904765553171e-05, "loss": 0.2843, "step": 22168 }, { "epoch": 2.29, "learning_rate": 2.2656661089082607e-05, "loss": 0.3273, "step": 22169 }, { "epoch": 2.29, "learning_rate": 2.265427457206714e-05, "loss": 0.3221, "step": 22170 }, { "epoch": 2.29, "learning_rate": 2.265188810450139e-05, "loss": 0.6607, "step": 22171 }, { "epoch": 2.29, "learning_rate": 2.2649501686401395e-05, "loss": 0.2559, "step": 22172 }, { "epoch": 2.29, "learning_rate": 2.264711531778324e-05, "loss": 0.3282, "step": 22173 }, { "epoch": 2.29, "learning_rate": 2.264472899866299e-05, "loss": 0.297, "step": 22174 }, { "epoch": 2.29, "learning_rate": 2.26423427290567e-05, "loss": 0.0618, "step": 22175 }, { "epoch": 2.29, "learning_rate": 2.2639956508980434e-05, "loss": 0.7133, "step": 22176 }, { "epoch": 2.29, "learning_rate": 2.263757033845027e-05, "loss": 0.194, "step": 22177 }, { "epoch": 2.29, "learning_rate": 2.2635184217482248e-05, "loss": 0.2962, "step": 22178 }, { "epoch": 2.29, "learning_rate": 2.2632798146092444e-05, "loss": 0.566, "step": 22179 }, { "epoch": 2.29, "learning_rate": 2.2630412124296917e-05, "loss": 0.4542, "step": 22180 }, { "epoch": 2.29, "learning_rate": 2.2628026152111744e-05, "loss": 0.2755, "step": 22181 }, { "epoch": 2.29, "learning_rate": 2.262564022955296e-05, "loss": 0.2035, "step": 22182 }, { "epoch": 2.29, "learning_rate": 2.2623254356636644e-05, "loss": 0.2075, "step": 22183 }, { "epoch": 2.29, "learning_rate": 2.262086853337885e-05, "loss": 0.0618, "step": 22184 }, { "epoch": 2.29, "learning_rate": 2.2618482759795653e-05, "loss": 0.176, "step": 22185 }, { "epoch": 2.29, "learning_rate": 2.2616097035903098e-05, "loss": 0.1693, "step": 22186 }, { "epoch": 2.29, "learning_rate": 2.2613711361717248e-05, "loss": 0.4256, "step": 22187 }, { "epoch": 2.29, "learning_rate": 2.261132573725416e-05, "loss": 0.2095, "step": 22188 }, { "epoch": 2.29, "learning_rate": 2.2608940162529918e-05, "loss": 0.5671, "step": 22189 }, { "epoch": 2.29, "learning_rate": 2.2606554637560546e-05, "loss": 0.2493, "step": 22190 }, { "epoch": 2.29, "learning_rate": 2.2604169162362123e-05, "loss": 0.0981, "step": 22191 }, { "epoch": 2.29, "learning_rate": 2.26017837369507e-05, "loss": 0.1424, "step": 22192 }, { "epoch": 2.29, "learning_rate": 2.2599398361342335e-05, "loss": 0.2398, "step": 22193 }, { "epoch": 2.29, "learning_rate": 2.259701303555309e-05, "loss": 0.4093, "step": 22194 }, { "epoch": 2.29, "learning_rate": 2.2594627759599028e-05, "loss": 0.459, "step": 22195 }, { "epoch": 2.29, "learning_rate": 2.259224253349619e-05, "loss": 0.1955, "step": 22196 }, { "epoch": 2.29, "learning_rate": 2.2589857357260643e-05, "loss": 0.6422, "step": 22197 }, { "epoch": 2.29, "learning_rate": 2.2587472230908446e-05, "loss": 0.0861, "step": 22198 }, { "epoch": 2.29, "learning_rate": 2.2585087154455654e-05, "loss": 0.3622, "step": 22199 }, { "epoch": 2.29, "learning_rate": 2.2582702127918315e-05, "loss": 0.3423, "step": 22200 }, { "epoch": 2.29, "learning_rate": 2.258031715131249e-05, "loss": 0.2524, "step": 22201 }, { "epoch": 2.29, "learning_rate": 2.2577932224654246e-05, "loss": 0.3847, "step": 22202 }, { "epoch": 2.29, "learning_rate": 2.2575547347959613e-05, "loss": 0.258, "step": 22203 }, { "epoch": 2.29, "learning_rate": 2.2573162521244657e-05, "loss": 0.6689, "step": 22204 }, { "epoch": 2.29, "learning_rate": 2.257077774452544e-05, "loss": 0.5144, "step": 22205 }, { "epoch": 2.29, "learning_rate": 2.2568393017818005e-05, "loss": 0.1085, "step": 22206 }, { "epoch": 2.29, "learning_rate": 2.256600834113841e-05, "loss": 0.6053, "step": 22207 }, { "epoch": 2.29, "learning_rate": 2.2563623714502704e-05, "loss": 0.14, "step": 22208 }, { "epoch": 2.29, "learning_rate": 2.2561239137926952e-05, "loss": 0.558, "step": 22209 }, { "epoch": 2.29, "learning_rate": 2.2558854611427188e-05, "loss": 0.289, "step": 22210 }, { "epoch": 2.29, "learning_rate": 2.2556470135019476e-05, "loss": 0.3235, "step": 22211 }, { "epoch": 2.29, "learning_rate": 2.2554085708719867e-05, "loss": 0.3176, "step": 22212 }, { "epoch": 2.29, "learning_rate": 2.2551701332544416e-05, "loss": 0.244, "step": 22213 }, { "epoch": 2.29, "learning_rate": 2.2549317006509165e-05, "loss": 0.4345, "step": 22214 }, { "epoch": 2.29, "learning_rate": 2.2546932730630173e-05, "loss": 0.5624, "step": 22215 }, { "epoch": 2.29, "learning_rate": 2.2544548504923477e-05, "loss": 0.3868, "step": 22216 }, { "epoch": 2.29, "learning_rate": 2.254216432940514e-05, "loss": 0.3399, "step": 22217 }, { "epoch": 2.29, "learning_rate": 2.2539780204091202e-05, "loss": 0.6779, "step": 22218 }, { "epoch": 2.29, "learning_rate": 2.2537396128997732e-05, "loss": 0.6417, "step": 22219 }, { "epoch": 2.29, "learning_rate": 2.2535012104140752e-05, "loss": 0.1767, "step": 22220 }, { "epoch": 2.29, "learning_rate": 2.2532628129536327e-05, "loss": 0.1472, "step": 22221 }, { "epoch": 2.29, "learning_rate": 2.2530244205200498e-05, "loss": 0.2768, "step": 22222 }, { "epoch": 2.29, "learning_rate": 2.252786033114933e-05, "loss": 1.0933, "step": 22223 }, { "epoch": 2.29, "learning_rate": 2.252547650739885e-05, "loss": 0.2947, "step": 22224 }, { "epoch": 2.29, "learning_rate": 2.252309273396511e-05, "loss": 0.8964, "step": 22225 }, { "epoch": 2.29, "learning_rate": 2.252070901086416e-05, "loss": 0.5705, "step": 22226 }, { "epoch": 2.29, "learning_rate": 2.2518325338112062e-05, "loss": 0.7391, "step": 22227 }, { "epoch": 2.29, "learning_rate": 2.2515941715724837e-05, "loss": 0.269, "step": 22228 }, { "epoch": 2.29, "learning_rate": 2.2513558143718542e-05, "loss": 0.6079, "step": 22229 }, { "epoch": 2.3, "learning_rate": 2.2511174622109217e-05, "loss": 0.3273, "step": 22230 }, { "epoch": 2.3, "learning_rate": 2.2508791150912915e-05, "loss": 0.3327, "step": 22231 }, { "epoch": 2.3, "learning_rate": 2.2506407730145678e-05, "loss": 0.2286, "step": 22232 }, { "epoch": 2.3, "learning_rate": 2.250402435982356e-05, "loss": 0.1472, "step": 22233 }, { "epoch": 2.3, "learning_rate": 2.250164103996258e-05, "loss": 0.9728, "step": 22234 }, { "epoch": 2.3, "learning_rate": 2.2499257770578807e-05, "loss": 0.4985, "step": 22235 }, { "epoch": 2.3, "learning_rate": 2.249687455168827e-05, "loss": 0.3828, "step": 22236 }, { "epoch": 2.3, "learning_rate": 2.249449138330703e-05, "loss": 0.086, "step": 22237 }, { "epoch": 2.3, "learning_rate": 2.2492108265451106e-05, "loss": 0.7531, "step": 22238 }, { "epoch": 2.3, "learning_rate": 2.2489725198136555e-05, "loss": 0.1611, "step": 22239 }, { "epoch": 2.3, "learning_rate": 2.2487342181379426e-05, "loss": 0.2305, "step": 22240 }, { "epoch": 2.3, "learning_rate": 2.2484959215195744e-05, "loss": 0.1101, "step": 22241 }, { "epoch": 2.3, "learning_rate": 2.2482576299601555e-05, "loss": 0.447, "step": 22242 }, { "epoch": 2.3, "learning_rate": 2.2480193434612913e-05, "loss": 1.1275, "step": 22243 }, { "epoch": 2.3, "learning_rate": 2.247781062024584e-05, "loss": 0.298, "step": 22244 }, { "epoch": 2.3, "learning_rate": 2.2475427856516385e-05, "loss": 0.1142, "step": 22245 }, { "epoch": 2.3, "learning_rate": 2.2473045143440594e-05, "loss": 0.4996, "step": 22246 }, { "epoch": 2.3, "learning_rate": 2.2470662481034504e-05, "loss": 0.1793, "step": 22247 }, { "epoch": 2.3, "learning_rate": 2.2468279869314148e-05, "loss": 0.2037, "step": 22248 }, { "epoch": 2.3, "learning_rate": 2.2465897308295567e-05, "loss": 0.2065, "step": 22249 }, { "epoch": 2.3, "learning_rate": 2.2463514797994802e-05, "loss": 0.4403, "step": 22250 }, { "epoch": 2.3, "learning_rate": 2.2461132338427903e-05, "loss": 0.3965, "step": 22251 }, { "epoch": 2.3, "learning_rate": 2.2458749929610897e-05, "loss": 0.4789, "step": 22252 }, { "epoch": 2.3, "learning_rate": 2.245636757155982e-05, "loss": 0.3885, "step": 22253 }, { "epoch": 2.3, "learning_rate": 2.2453985264290703e-05, "loss": 0.7278, "step": 22254 }, { "epoch": 2.3, "learning_rate": 2.2451603007819593e-05, "loss": 0.4389, "step": 22255 }, { "epoch": 2.3, "learning_rate": 2.2449220802162525e-05, "loss": 0.3415, "step": 22256 }, { "epoch": 2.3, "learning_rate": 2.2446838647335542e-05, "loss": 0.2322, "step": 22257 }, { "epoch": 2.3, "learning_rate": 2.244445654335467e-05, "loss": 0.5184, "step": 22258 }, { "epoch": 2.3, "learning_rate": 2.2442074490235946e-05, "loss": 0.2931, "step": 22259 }, { "epoch": 2.3, "learning_rate": 2.243969248799541e-05, "loss": 0.3968, "step": 22260 }, { "epoch": 2.3, "learning_rate": 2.24373105366491e-05, "loss": 0.3717, "step": 22261 }, { "epoch": 2.3, "learning_rate": 2.243492863621304e-05, "loss": 0.5189, "step": 22262 }, { "epoch": 2.3, "learning_rate": 2.2432546786703275e-05, "loss": 0.2348, "step": 22263 }, { "epoch": 2.3, "learning_rate": 2.2430164988135828e-05, "loss": 0.6157, "step": 22264 }, { "epoch": 2.3, "learning_rate": 2.242778324052675e-05, "loss": 0.5212, "step": 22265 }, { "epoch": 2.3, "learning_rate": 2.242540154389206e-05, "loss": 0.1425, "step": 22266 }, { "epoch": 2.3, "learning_rate": 2.2423019898247792e-05, "loss": 0.3488, "step": 22267 }, { "epoch": 2.3, "learning_rate": 2.2420638303609982e-05, "loss": 0.089, "step": 22268 }, { "epoch": 2.3, "learning_rate": 2.2418256759994655e-05, "loss": 0.1068, "step": 22269 }, { "epoch": 2.3, "learning_rate": 2.2415875267417854e-05, "loss": 0.1833, "step": 22270 }, { "epoch": 2.3, "learning_rate": 2.2413493825895612e-05, "loss": 0.2042, "step": 22271 }, { "epoch": 2.3, "learning_rate": 2.2411112435443947e-05, "loss": 0.1596, "step": 22272 }, { "epoch": 2.3, "learning_rate": 2.2408731096078892e-05, "loss": 0.2705, "step": 22273 }, { "epoch": 2.3, "learning_rate": 2.240634980781649e-05, "loss": 0.2414, "step": 22274 }, { "epoch": 2.3, "learning_rate": 2.240396857067277e-05, "loss": 0.8347, "step": 22275 }, { "epoch": 2.3, "learning_rate": 2.240158738466375e-05, "loss": 0.5219, "step": 22276 }, { "epoch": 2.3, "learning_rate": 2.239920624980548e-05, "loss": 0.2857, "step": 22277 }, { "epoch": 2.3, "learning_rate": 2.2396825166113956e-05, "loss": 0.4772, "step": 22278 }, { "epoch": 2.3, "learning_rate": 2.239444413360523e-05, "loss": 0.1552, "step": 22279 }, { "epoch": 2.3, "learning_rate": 2.2392063152295325e-05, "loss": 0.8204, "step": 22280 }, { "epoch": 2.3, "learning_rate": 2.238968222220028e-05, "loss": 0.2899, "step": 22281 }, { "epoch": 2.3, "learning_rate": 2.238730134333611e-05, "loss": 0.4433, "step": 22282 }, { "epoch": 2.3, "learning_rate": 2.238492051571884e-05, "loss": 0.2427, "step": 22283 }, { "epoch": 2.3, "learning_rate": 2.2382539739364504e-05, "loss": 0.1026, "step": 22284 }, { "epoch": 2.3, "learning_rate": 2.2380159014289135e-05, "loss": 0.4534, "step": 22285 }, { "epoch": 2.3, "learning_rate": 2.237777834050875e-05, "loss": 0.2525, "step": 22286 }, { "epoch": 2.3, "learning_rate": 2.2375397718039375e-05, "loss": 0.4103, "step": 22287 }, { "epoch": 2.3, "learning_rate": 2.237301714689704e-05, "loss": 1.0407, "step": 22288 }, { "epoch": 2.3, "learning_rate": 2.237063662709777e-05, "loss": 0.2576, "step": 22289 }, { "epoch": 2.3, "learning_rate": 2.2368256158657593e-05, "loss": 0.3075, "step": 22290 }, { "epoch": 2.3, "learning_rate": 2.236587574159253e-05, "loss": 0.4641, "step": 22291 }, { "epoch": 2.3, "learning_rate": 2.23634953759186e-05, "loss": 0.2316, "step": 22292 }, { "epoch": 2.3, "learning_rate": 2.2361115061651828e-05, "loss": 0.4847, "step": 22293 }, { "epoch": 2.3, "learning_rate": 2.2358734798808247e-05, "loss": 0.4616, "step": 22294 }, { "epoch": 2.3, "learning_rate": 2.235635458740388e-05, "loss": 0.3972, "step": 22295 }, { "epoch": 2.3, "learning_rate": 2.2353974427454735e-05, "loss": 1.0388, "step": 22296 }, { "epoch": 2.3, "learning_rate": 2.2351594318976848e-05, "loss": 0.292, "step": 22297 }, { "epoch": 2.3, "learning_rate": 2.2349214261986245e-05, "loss": 0.2339, "step": 22298 }, { "epoch": 2.3, "learning_rate": 2.234683425649894e-05, "loss": 0.2098, "step": 22299 }, { "epoch": 2.3, "learning_rate": 2.2344454302530953e-05, "loss": 0.2466, "step": 22300 }, { "epoch": 2.3, "learning_rate": 2.2342074400098306e-05, "loss": 0.6168, "step": 22301 }, { "epoch": 2.3, "learning_rate": 2.2339694549217036e-05, "loss": 0.3855, "step": 22302 }, { "epoch": 2.3, "learning_rate": 2.233731474990314e-05, "loss": 0.0782, "step": 22303 }, { "epoch": 2.3, "learning_rate": 2.2334935002172644e-05, "loss": 0.3659, "step": 22304 }, { "epoch": 2.3, "learning_rate": 2.233255530604158e-05, "loss": 0.1795, "step": 22305 }, { "epoch": 2.3, "learning_rate": 2.2330175661525952e-05, "loss": 0.1196, "step": 22306 }, { "epoch": 2.3, "learning_rate": 2.232779606864179e-05, "loss": 0.393, "step": 22307 }, { "epoch": 2.3, "learning_rate": 2.2325416527405107e-05, "loss": 0.3596, "step": 22308 }, { "epoch": 2.3, "learning_rate": 2.232303703783193e-05, "loss": 0.2657, "step": 22309 }, { "epoch": 2.3, "learning_rate": 2.232065759993827e-05, "loss": 0.6593, "step": 22310 }, { "epoch": 2.3, "learning_rate": 2.231827821374014e-05, "loss": 0.5946, "step": 22311 }, { "epoch": 2.3, "learning_rate": 2.2315898879253568e-05, "loss": 0.3301, "step": 22312 }, { "epoch": 2.3, "learning_rate": 2.2313519596494572e-05, "loss": 0.3736, "step": 22313 }, { "epoch": 2.3, "learning_rate": 2.2311140365479155e-05, "loss": 0.2565, "step": 22314 }, { "epoch": 2.3, "learning_rate": 2.2308761186223357e-05, "loss": 0.3049, "step": 22315 }, { "epoch": 2.3, "learning_rate": 2.230638205874317e-05, "loss": 0.3335, "step": 22316 }, { "epoch": 2.3, "learning_rate": 2.230400298305461e-05, "loss": 0.6116, "step": 22317 }, { "epoch": 2.3, "learning_rate": 2.230162395917371e-05, "loss": 0.2468, "step": 22318 }, { "epoch": 2.3, "learning_rate": 2.2299244987116478e-05, "loss": 0.3863, "step": 22319 }, { "epoch": 2.3, "learning_rate": 2.229686606689892e-05, "loss": 0.5092, "step": 22320 }, { "epoch": 2.3, "learning_rate": 2.2294487198537058e-05, "loss": 0.1776, "step": 22321 }, { "epoch": 2.3, "learning_rate": 2.2292108382046908e-05, "loss": 0.4505, "step": 22322 }, { "epoch": 2.3, "learning_rate": 2.228972961744448e-05, "loss": 2.0949, "step": 22323 }, { "epoch": 2.3, "learning_rate": 2.228735090474579e-05, "loss": 0.5867, "step": 22324 }, { "epoch": 2.3, "learning_rate": 2.2284972243966845e-05, "loss": 0.1807, "step": 22325 }, { "epoch": 2.31, "learning_rate": 2.2282593635123664e-05, "loss": 0.1374, "step": 22326 }, { "epoch": 2.31, "learning_rate": 2.228021507823227e-05, "loss": 0.8011, "step": 22327 }, { "epoch": 2.31, "learning_rate": 2.2277836573308647e-05, "loss": 0.3431, "step": 22328 }, { "epoch": 2.31, "learning_rate": 2.2275458120368825e-05, "loss": 0.5117, "step": 22329 }, { "epoch": 2.31, "learning_rate": 2.2273079719428807e-05, "loss": 0.9157, "step": 22330 }, { "epoch": 2.31, "learning_rate": 2.227070137050461e-05, "loss": 0.5609, "step": 22331 }, { "epoch": 2.31, "learning_rate": 2.2268323073612243e-05, "loss": 0.1066, "step": 22332 }, { "epoch": 2.31, "learning_rate": 2.2265944828767722e-05, "loss": 0.3159, "step": 22333 }, { "epoch": 2.31, "learning_rate": 2.2263566635987043e-05, "loss": 0.1868, "step": 22334 }, { "epoch": 2.31, "learning_rate": 2.2261188495286226e-05, "loss": 0.2516, "step": 22335 }, { "epoch": 2.31, "learning_rate": 2.2258810406681275e-05, "loss": 0.4356, "step": 22336 }, { "epoch": 2.31, "learning_rate": 2.2256432370188208e-05, "loss": 0.451, "step": 22337 }, { "epoch": 2.31, "learning_rate": 2.2254054385823018e-05, "loss": 0.1693, "step": 22338 }, { "epoch": 2.31, "learning_rate": 2.2251676453601725e-05, "loss": 0.4941, "step": 22339 }, { "epoch": 2.31, "learning_rate": 2.2249298573540346e-05, "loss": 0.3092, "step": 22340 }, { "epoch": 2.31, "learning_rate": 2.2246920745654857e-05, "loss": 0.3391, "step": 22341 }, { "epoch": 2.31, "learning_rate": 2.224454296996129e-05, "loss": 0.5178, "step": 22342 }, { "epoch": 2.31, "learning_rate": 2.224216524647565e-05, "loss": 0.4745, "step": 22343 }, { "epoch": 2.31, "learning_rate": 2.2239787575213926e-05, "loss": 0.3644, "step": 22344 }, { "epoch": 2.31, "learning_rate": 2.2237409956192146e-05, "loss": 0.2084, "step": 22345 }, { "epoch": 2.31, "learning_rate": 2.2235032389426302e-05, "loss": 0.4361, "step": 22346 }, { "epoch": 2.31, "learning_rate": 2.2232654874932406e-05, "loss": 0.6925, "step": 22347 }, { "epoch": 2.31, "learning_rate": 2.2230277412726456e-05, "loss": 0.3734, "step": 22348 }, { "epoch": 2.31, "learning_rate": 2.2227900002824463e-05, "loss": 0.7263, "step": 22349 }, { "epoch": 2.31, "learning_rate": 2.2225522645242428e-05, "loss": 0.2111, "step": 22350 }, { "epoch": 2.31, "learning_rate": 2.222314533999636e-05, "loss": 0.4625, "step": 22351 }, { "epoch": 2.31, "learning_rate": 2.2220768087102253e-05, "loss": 0.431, "step": 22352 }, { "epoch": 2.31, "learning_rate": 2.2218390886576125e-05, "loss": 0.1332, "step": 22353 }, { "epoch": 2.31, "learning_rate": 2.2216013738433956e-05, "loss": 0.4812, "step": 22354 }, { "epoch": 2.31, "learning_rate": 2.2213636642691763e-05, "loss": 1.0168, "step": 22355 }, { "epoch": 2.31, "learning_rate": 2.2211259599365548e-05, "loss": 0.0941, "step": 22356 }, { "epoch": 2.31, "learning_rate": 2.2208882608471316e-05, "loss": 0.63, "step": 22357 }, { "epoch": 2.31, "learning_rate": 2.2206505670025056e-05, "loss": 0.6594, "step": 22358 }, { "epoch": 2.31, "learning_rate": 2.2204128784042776e-05, "loss": 0.4774, "step": 22359 }, { "epoch": 2.31, "learning_rate": 2.220175195054048e-05, "loss": 0.1249, "step": 22360 }, { "epoch": 2.31, "learning_rate": 2.2199375169534167e-05, "loss": 0.189, "step": 22361 }, { "epoch": 2.31, "learning_rate": 2.2196998441039833e-05, "loss": 0.1155, "step": 22362 }, { "epoch": 2.31, "learning_rate": 2.2194621765073473e-05, "loss": 0.7752, "step": 22363 }, { "epoch": 2.31, "learning_rate": 2.21922451416511e-05, "loss": 0.3816, "step": 22364 }, { "epoch": 2.31, "learning_rate": 2.2189868570788718e-05, "loss": 0.4313, "step": 22365 }, { "epoch": 2.31, "learning_rate": 2.21874920525023e-05, "loss": 0.3194, "step": 22366 }, { "epoch": 2.31, "learning_rate": 2.218511558680786e-05, "loss": 0.5768, "step": 22367 }, { "epoch": 2.31, "learning_rate": 2.218273917372139e-05, "loss": 0.3964, "step": 22368 }, { "epoch": 2.31, "learning_rate": 2.2180362813258892e-05, "loss": 0.8178, "step": 22369 }, { "epoch": 2.31, "learning_rate": 2.217798650543636e-05, "loss": 0.1366, "step": 22370 }, { "epoch": 2.31, "learning_rate": 2.2175610250269798e-05, "loss": 0.2168, "step": 22371 }, { "epoch": 2.31, "learning_rate": 2.2173234047775195e-05, "loss": 0.2965, "step": 22372 }, { "epoch": 2.31, "learning_rate": 2.2170857897968546e-05, "loss": 0.2173, "step": 22373 }, { "epoch": 2.31, "learning_rate": 2.2168481800865848e-05, "loss": 0.4167, "step": 22374 }, { "epoch": 2.31, "learning_rate": 2.2166105756483107e-05, "loss": 0.728, "step": 22375 }, { "epoch": 2.31, "learning_rate": 2.2163729764836306e-05, "loss": 0.8004, "step": 22376 }, { "epoch": 2.31, "learning_rate": 2.2161353825941436e-05, "loss": 0.1943, "step": 22377 }, { "epoch": 2.31, "learning_rate": 2.2158977939814513e-05, "loss": 0.4745, "step": 22378 }, { "epoch": 2.31, "learning_rate": 2.2156602106471506e-05, "loss": 0.131, "step": 22379 }, { "epoch": 2.31, "learning_rate": 2.2154226325928415e-05, "loss": 0.6075, "step": 22380 }, { "epoch": 2.31, "learning_rate": 2.2151850598201244e-05, "loss": 0.4066, "step": 22381 }, { "epoch": 2.31, "learning_rate": 2.214947492330597e-05, "loss": 0.3415, "step": 22382 }, { "epoch": 2.31, "learning_rate": 2.2147099301258595e-05, "loss": 0.2173, "step": 22383 }, { "epoch": 2.31, "learning_rate": 2.2144723732075114e-05, "loss": 0.6525, "step": 22384 }, { "epoch": 2.31, "learning_rate": 2.2142348215771517e-05, "loss": 0.5799, "step": 22385 }, { "epoch": 2.31, "learning_rate": 2.213997275236379e-05, "loss": 0.2542, "step": 22386 }, { "epoch": 2.31, "learning_rate": 2.2137597341867923e-05, "loss": 0.4332, "step": 22387 }, { "epoch": 2.31, "learning_rate": 2.2135221984299917e-05, "loss": 0.5174, "step": 22388 }, { "epoch": 2.31, "learning_rate": 2.213284667967576e-05, "loss": 0.1673, "step": 22389 }, { "epoch": 2.31, "learning_rate": 2.2130471428011435e-05, "loss": 0.6914, "step": 22390 }, { "epoch": 2.31, "learning_rate": 2.2128096229322937e-05, "loss": 0.1824, "step": 22391 }, { "epoch": 2.31, "learning_rate": 2.212572108362625e-05, "loss": 0.3987, "step": 22392 }, { "epoch": 2.31, "learning_rate": 2.2123345990937362e-05, "loss": 0.3098, "step": 22393 }, { "epoch": 2.31, "learning_rate": 2.212097095127227e-05, "loss": 0.171, "step": 22394 }, { "epoch": 2.31, "learning_rate": 2.211859596464696e-05, "loss": 0.5193, "step": 22395 }, { "epoch": 2.31, "learning_rate": 2.2116221031077416e-05, "loss": 0.3952, "step": 22396 }, { "epoch": 2.31, "learning_rate": 2.2113846150579623e-05, "loss": 0.235, "step": 22397 }, { "epoch": 2.31, "learning_rate": 2.2111471323169575e-05, "loss": 0.1723, "step": 22398 }, { "epoch": 2.31, "learning_rate": 2.2109096548863262e-05, "loss": 0.5669, "step": 22399 }, { "epoch": 2.31, "learning_rate": 2.210672182767666e-05, "loss": 0.2528, "step": 22400 }, { "epoch": 2.31, "learning_rate": 2.2104347159625757e-05, "loss": 0.425, "step": 22401 }, { "epoch": 2.31, "learning_rate": 2.2101972544726542e-05, "loss": 0.2397, "step": 22402 }, { "epoch": 2.31, "learning_rate": 2.2099597982995013e-05, "loss": 0.1128, "step": 22403 }, { "epoch": 2.31, "learning_rate": 2.2097223474447134e-05, "loss": 1.5866, "step": 22404 }, { "epoch": 2.31, "learning_rate": 2.2094849019098897e-05, "loss": 0.4577, "step": 22405 }, { "epoch": 2.31, "learning_rate": 2.209247461696628e-05, "loss": 0.5888, "step": 22406 }, { "epoch": 2.31, "learning_rate": 2.2090100268065278e-05, "loss": 0.4662, "step": 22407 }, { "epoch": 2.31, "learning_rate": 2.208772597241187e-05, "loss": 0.3661, "step": 22408 }, { "epoch": 2.31, "learning_rate": 2.208535173002204e-05, "loss": 0.3571, "step": 22409 }, { "epoch": 2.31, "learning_rate": 2.208297754091177e-05, "loss": 0.728, "step": 22410 }, { "epoch": 2.31, "learning_rate": 2.2080603405097042e-05, "loss": 0.1011, "step": 22411 }, { "epoch": 2.31, "learning_rate": 2.2078229322593837e-05, "loss": 0.2968, "step": 22412 }, { "epoch": 2.31, "learning_rate": 2.2075855293418146e-05, "loss": 0.473, "step": 22413 }, { "epoch": 2.31, "learning_rate": 2.2073481317585936e-05, "loss": 0.2254, "step": 22414 }, { "epoch": 2.31, "learning_rate": 2.207110739511321e-05, "loss": 0.2298, "step": 22415 }, { "epoch": 2.31, "learning_rate": 2.2068733526015918e-05, "loss": 0.7592, "step": 22416 }, { "epoch": 2.31, "learning_rate": 2.206635971031006e-05, "loss": 0.644, "step": 22417 }, { "epoch": 2.31, "learning_rate": 2.206398594801161e-05, "loss": 0.2041, "step": 22418 }, { "epoch": 2.31, "learning_rate": 2.2061612239136556e-05, "loss": 0.8744, "step": 22419 }, { "epoch": 2.31, "learning_rate": 2.2059238583700864e-05, "loss": 0.3156, "step": 22420 }, { "epoch": 2.31, "learning_rate": 2.205686498172052e-05, "loss": 0.9224, "step": 22421 }, { "epoch": 2.32, "learning_rate": 2.20544914332115e-05, "loss": 0.22, "step": 22422 }, { "epoch": 2.32, "learning_rate": 2.2052117938189798e-05, "loss": 0.2232, "step": 22423 }, { "epoch": 2.32, "learning_rate": 2.2049744496671366e-05, "loss": 0.411, "step": 22424 }, { "epoch": 2.32, "learning_rate": 2.2047371108672195e-05, "loss": 0.1969, "step": 22425 }, { "epoch": 2.32, "learning_rate": 2.204499777420826e-05, "loss": 0.3094, "step": 22426 }, { "epoch": 2.32, "learning_rate": 2.2042624493295546e-05, "loss": 0.31, "step": 22427 }, { "epoch": 2.32, "learning_rate": 2.2040251265950025e-05, "loss": 0.0872, "step": 22428 }, { "epoch": 2.32, "learning_rate": 2.2037878092187665e-05, "loss": 0.2122, "step": 22429 }, { "epoch": 2.32, "learning_rate": 2.2035504972024446e-05, "loss": 0.8935, "step": 22430 }, { "epoch": 2.32, "learning_rate": 2.203313190547634e-05, "loss": 0.5939, "step": 22431 }, { "epoch": 2.32, "learning_rate": 2.2030758892559326e-05, "loss": 0.334, "step": 22432 }, { "epoch": 2.32, "learning_rate": 2.2028385933289385e-05, "loss": 0.2828, "step": 22433 }, { "epoch": 2.32, "learning_rate": 2.2026013027682478e-05, "loss": 0.481, "step": 22434 }, { "epoch": 2.32, "learning_rate": 2.2023640175754586e-05, "loss": 0.4579, "step": 22435 }, { "epoch": 2.32, "learning_rate": 2.2021267377521682e-05, "loss": 0.122, "step": 22436 }, { "epoch": 2.32, "learning_rate": 2.201889463299975e-05, "loss": 0.3745, "step": 22437 }, { "epoch": 2.32, "learning_rate": 2.201652194220474e-05, "loss": 0.4465, "step": 22438 }, { "epoch": 2.32, "learning_rate": 2.201414930515264e-05, "loss": 0.4206, "step": 22439 }, { "epoch": 2.32, "learning_rate": 2.2011776721859414e-05, "loss": 0.4598, "step": 22440 }, { "epoch": 2.32, "learning_rate": 2.200940419234105e-05, "loss": 0.5614, "step": 22441 }, { "epoch": 2.32, "learning_rate": 2.2007031716613505e-05, "loss": 0.3507, "step": 22442 }, { "epoch": 2.32, "learning_rate": 2.2004659294692748e-05, "loss": 0.3783, "step": 22443 }, { "epoch": 2.32, "learning_rate": 2.200228692659475e-05, "loss": 0.2316, "step": 22444 }, { "epoch": 2.32, "learning_rate": 2.1999914612335487e-05, "loss": 0.2469, "step": 22445 }, { "epoch": 2.32, "learning_rate": 2.1997542351930926e-05, "loss": 0.4384, "step": 22446 }, { "epoch": 2.32, "learning_rate": 2.199517014539704e-05, "loss": 0.3979, "step": 22447 }, { "epoch": 2.32, "learning_rate": 2.1992797992749796e-05, "loss": 0.4592, "step": 22448 }, { "epoch": 2.32, "learning_rate": 2.199042589400516e-05, "loss": 0.4025, "step": 22449 }, { "epoch": 2.32, "learning_rate": 2.19880538491791e-05, "loss": 0.2086, "step": 22450 }, { "epoch": 2.32, "learning_rate": 2.1985681858287596e-05, "loss": 0.143, "step": 22451 }, { "epoch": 2.32, "learning_rate": 2.1983309921346598e-05, "loss": 0.5587, "step": 22452 }, { "epoch": 2.32, "learning_rate": 2.1980938038372096e-05, "loss": 0.7487, "step": 22453 }, { "epoch": 2.32, "learning_rate": 2.197856620938003e-05, "loss": 0.2465, "step": 22454 }, { "epoch": 2.32, "learning_rate": 2.1976194434386378e-05, "loss": 0.481, "step": 22455 }, { "epoch": 2.32, "learning_rate": 2.1973822713407113e-05, "loss": 0.2205, "step": 22456 }, { "epoch": 2.32, "learning_rate": 2.1971451046458196e-05, "loss": 0.3134, "step": 22457 }, { "epoch": 2.32, "learning_rate": 2.1969079433555585e-05, "loss": 0.2924, "step": 22458 }, { "epoch": 2.32, "learning_rate": 2.1966707874715258e-05, "loss": 0.7987, "step": 22459 }, { "epoch": 2.32, "learning_rate": 2.196433636995317e-05, "loss": 0.1177, "step": 22460 }, { "epoch": 2.32, "learning_rate": 2.1961964919285297e-05, "loss": 0.4135, "step": 22461 }, { "epoch": 2.32, "learning_rate": 2.1959593522727586e-05, "loss": 0.4051, "step": 22462 }, { "epoch": 2.32, "learning_rate": 2.1957222180296015e-05, "loss": 0.7583, "step": 22463 }, { "epoch": 2.32, "learning_rate": 2.1954850892006543e-05, "loss": 0.3134, "step": 22464 }, { "epoch": 2.32, "learning_rate": 2.1952479657875135e-05, "loss": 0.17, "step": 22465 }, { "epoch": 2.32, "learning_rate": 2.195010847791775e-05, "loss": 0.5057, "step": 22466 }, { "epoch": 2.32, "learning_rate": 2.194773735215035e-05, "loss": 0.3324, "step": 22467 }, { "epoch": 2.32, "learning_rate": 2.1945366280588895e-05, "loss": 0.5906, "step": 22468 }, { "epoch": 2.32, "learning_rate": 2.194299526324935e-05, "loss": 0.4523, "step": 22469 }, { "epoch": 2.32, "learning_rate": 2.1940624300147674e-05, "loss": 0.5624, "step": 22470 }, { "epoch": 2.32, "learning_rate": 2.1938253391299837e-05, "loss": 0.0798, "step": 22471 }, { "epoch": 2.32, "learning_rate": 2.1935882536721785e-05, "loss": 0.7679, "step": 22472 }, { "epoch": 2.32, "learning_rate": 2.193351173642949e-05, "loss": 0.1525, "step": 22473 }, { "epoch": 2.32, "learning_rate": 2.1931140990438902e-05, "loss": 0.3611, "step": 22474 }, { "epoch": 2.32, "learning_rate": 2.192877029876599e-05, "loss": 0.2003, "step": 22475 }, { "epoch": 2.32, "learning_rate": 2.19263996614267e-05, "loss": 0.5089, "step": 22476 }, { "epoch": 2.32, "learning_rate": 2.1924029078437e-05, "loss": 0.2748, "step": 22477 }, { "epoch": 2.32, "learning_rate": 2.192165854981286e-05, "loss": 0.1821, "step": 22478 }, { "epoch": 2.32, "learning_rate": 2.1919288075570212e-05, "loss": 0.3458, "step": 22479 }, { "epoch": 2.32, "learning_rate": 2.1916917655725027e-05, "loss": 0.3676, "step": 22480 }, { "epoch": 2.32, "learning_rate": 2.1914547290293265e-05, "loss": 0.4379, "step": 22481 }, { "epoch": 2.32, "learning_rate": 2.1912176979290875e-05, "loss": 0.6772, "step": 22482 }, { "epoch": 2.32, "learning_rate": 2.1909806722733815e-05, "loss": 0.637, "step": 22483 }, { "epoch": 2.32, "learning_rate": 2.1907436520638045e-05, "loss": 0.1136, "step": 22484 }, { "epoch": 2.32, "learning_rate": 2.1905066373019525e-05, "loss": 0.3633, "step": 22485 }, { "epoch": 2.32, "learning_rate": 2.1902696279894196e-05, "loss": 0.6311, "step": 22486 }, { "epoch": 2.32, "learning_rate": 2.190032624127802e-05, "loss": 0.5885, "step": 22487 }, { "epoch": 2.32, "learning_rate": 2.1897956257186954e-05, "loss": 0.7532, "step": 22488 }, { "epoch": 2.32, "learning_rate": 2.189558632763696e-05, "loss": 1.6816, "step": 22489 }, { "epoch": 2.32, "learning_rate": 2.1893216452643974e-05, "loss": 0.0787, "step": 22490 }, { "epoch": 2.32, "learning_rate": 2.1890846632223972e-05, "loss": 0.5908, "step": 22491 }, { "epoch": 2.32, "learning_rate": 2.1888476866392878e-05, "loss": 0.4782, "step": 22492 }, { "epoch": 2.32, "learning_rate": 2.188610715516666e-05, "loss": 1.1327, "step": 22493 }, { "epoch": 2.32, "learning_rate": 2.188373749856127e-05, "loss": 0.3161, "step": 22494 }, { "epoch": 2.32, "learning_rate": 2.1881367896592672e-05, "loss": 0.9534, "step": 22495 }, { "epoch": 2.32, "learning_rate": 2.187899834927679e-05, "loss": 0.4572, "step": 22496 }, { "epoch": 2.32, "learning_rate": 2.18766288566296e-05, "loss": 0.2577, "step": 22497 }, { "epoch": 2.32, "learning_rate": 2.1874259418667037e-05, "loss": 0.3669, "step": 22498 }, { "epoch": 2.32, "learning_rate": 2.1871890035405068e-05, "loss": 0.7526, "step": 22499 }, { "epoch": 2.32, "learning_rate": 2.1869520706859626e-05, "loss": 0.0755, "step": 22500 }, { "epoch": 2.32, "learning_rate": 2.186715143304667e-05, "loss": 0.2465, "step": 22501 }, { "epoch": 2.32, "learning_rate": 2.1864782213982144e-05, "loss": 0.6425, "step": 22502 }, { "epoch": 2.32, "learning_rate": 2.1862413049682014e-05, "loss": 0.219, "step": 22503 }, { "epoch": 2.32, "learning_rate": 2.1860043940162206e-05, "loss": 0.2892, "step": 22504 }, { "epoch": 2.32, "learning_rate": 2.1857674885438683e-05, "loss": 0.7801, "step": 22505 }, { "epoch": 2.32, "learning_rate": 2.185530588552738e-05, "loss": 0.4418, "step": 22506 }, { "epoch": 2.32, "learning_rate": 2.1852936940444256e-05, "loss": 0.6191, "step": 22507 }, { "epoch": 2.32, "learning_rate": 2.1850568050205245e-05, "loss": 0.5189, "step": 22508 }, { "epoch": 2.32, "learning_rate": 2.1848199214826318e-05, "loss": 0.2654, "step": 22509 }, { "epoch": 2.32, "learning_rate": 2.1845830434323397e-05, "loss": 0.5601, "step": 22510 }, { "epoch": 2.32, "learning_rate": 2.1843461708712435e-05, "loss": 0.3266, "step": 22511 }, { "epoch": 2.32, "learning_rate": 2.1841093038009385e-05, "loss": 0.4793, "step": 22512 }, { "epoch": 2.32, "learning_rate": 2.183872442223019e-05, "loss": 0.5155, "step": 22513 }, { "epoch": 2.32, "learning_rate": 2.183635586139079e-05, "loss": 0.1328, "step": 22514 }, { "epoch": 2.32, "learning_rate": 2.183398735550713e-05, "loss": 0.4615, "step": 22515 }, { "epoch": 2.32, "learning_rate": 2.183161890459517e-05, "loss": 0.7155, "step": 22516 }, { "epoch": 2.33, "learning_rate": 2.1829250508670828e-05, "loss": 0.3364, "step": 22517 }, { "epoch": 2.33, "learning_rate": 2.1826882167750057e-05, "loss": 0.3844, "step": 22518 }, { "epoch": 2.33, "learning_rate": 2.1824513881848813e-05, "loss": 0.247, "step": 22519 }, { "epoch": 2.33, "learning_rate": 2.182214565098302e-05, "loss": 0.3202, "step": 22520 }, { "epoch": 2.33, "learning_rate": 2.181977747516863e-05, "loss": 0.7372, "step": 22521 }, { "epoch": 2.33, "learning_rate": 2.1817409354421584e-05, "loss": 0.7367, "step": 22522 }, { "epoch": 2.33, "learning_rate": 2.181504128875783e-05, "loss": 0.1858, "step": 22523 }, { "epoch": 2.33, "learning_rate": 2.1812673278193292e-05, "loss": 0.6882, "step": 22524 }, { "epoch": 2.33, "learning_rate": 2.181030532274393e-05, "loss": 0.4586, "step": 22525 }, { "epoch": 2.33, "learning_rate": 2.180793742242567e-05, "loss": 0.1934, "step": 22526 }, { "epoch": 2.33, "learning_rate": 2.1805569577254467e-05, "loss": 0.4554, "step": 22527 }, { "epoch": 2.33, "learning_rate": 2.1803201787246253e-05, "loss": 0.4538, "step": 22528 }, { "epoch": 2.33, "learning_rate": 2.1800834052416966e-05, "loss": 0.1799, "step": 22529 }, { "epoch": 2.33, "learning_rate": 2.179846637278254e-05, "loss": 0.5391, "step": 22530 }, { "epoch": 2.33, "learning_rate": 2.179609874835892e-05, "loss": 0.3268, "step": 22531 }, { "epoch": 2.33, "learning_rate": 2.1793731179162043e-05, "loss": 0.6636, "step": 22532 }, { "epoch": 2.33, "learning_rate": 2.1791363665207853e-05, "loss": 0.2596, "step": 22533 }, { "epoch": 2.33, "learning_rate": 2.1788996206512274e-05, "loss": 0.2329, "step": 22534 }, { "epoch": 2.33, "learning_rate": 2.1786628803091255e-05, "loss": 0.2615, "step": 22535 }, { "epoch": 2.33, "learning_rate": 2.178426145496073e-05, "loss": 0.2801, "step": 22536 }, { "epoch": 2.33, "learning_rate": 2.1781894162136634e-05, "loss": 0.6121, "step": 22537 }, { "epoch": 2.33, "learning_rate": 2.1779526924634906e-05, "loss": 0.3246, "step": 22538 }, { "epoch": 2.33, "learning_rate": 2.177715974247147e-05, "loss": 0.3644, "step": 22539 }, { "epoch": 2.33, "learning_rate": 2.1774792615662278e-05, "loss": 0.1963, "step": 22540 }, { "epoch": 2.33, "learning_rate": 2.1772425544223266e-05, "loss": 0.3324, "step": 22541 }, { "epoch": 2.33, "learning_rate": 2.177005852817035e-05, "loss": 0.2121, "step": 22542 }, { "epoch": 2.33, "learning_rate": 2.1767691567519478e-05, "loss": 0.836, "step": 22543 }, { "epoch": 2.33, "learning_rate": 2.1765324662286576e-05, "loss": 0.4335, "step": 22544 }, { "epoch": 2.33, "learning_rate": 2.1762957812487587e-05, "loss": 0.1407, "step": 22545 }, { "epoch": 2.33, "learning_rate": 2.1760591018138433e-05, "loss": 0.0395, "step": 22546 }, { "epoch": 2.33, "learning_rate": 2.1758224279255057e-05, "loss": 0.1259, "step": 22547 }, { "epoch": 2.33, "learning_rate": 2.1755857595853382e-05, "loss": 0.479, "step": 22548 }, { "epoch": 2.33, "learning_rate": 2.1753490967949347e-05, "loss": 0.4521, "step": 22549 }, { "epoch": 2.33, "learning_rate": 2.1751124395558875e-05, "loss": 0.6681, "step": 22550 }, { "epoch": 2.33, "learning_rate": 2.1748757878697915e-05, "loss": 0.4336, "step": 22551 }, { "epoch": 2.33, "learning_rate": 2.174639141738238e-05, "loss": 0.3821, "step": 22552 }, { "epoch": 2.33, "learning_rate": 2.1744025011628206e-05, "loss": 0.1228, "step": 22553 }, { "epoch": 2.33, "learning_rate": 2.174165866145134e-05, "loss": 0.0984, "step": 22554 }, { "epoch": 2.33, "learning_rate": 2.1739292366867674e-05, "loss": 0.6844, "step": 22555 }, { "epoch": 2.33, "learning_rate": 2.1736926127893166e-05, "loss": 0.5821, "step": 22556 }, { "epoch": 2.33, "learning_rate": 2.1734559944543744e-05, "loss": 0.1467, "step": 22557 }, { "epoch": 2.33, "learning_rate": 2.173219381683532e-05, "loss": 0.3168, "step": 22558 }, { "epoch": 2.33, "learning_rate": 2.1729827744783833e-05, "loss": 0.1488, "step": 22559 }, { "epoch": 2.33, "learning_rate": 2.1727461728405215e-05, "loss": 0.208, "step": 22560 }, { "epoch": 2.33, "learning_rate": 2.1725095767715396e-05, "loss": 0.4475, "step": 22561 }, { "epoch": 2.33, "learning_rate": 2.1722729862730288e-05, "loss": 0.187, "step": 22562 }, { "epoch": 2.33, "learning_rate": 2.1720364013465823e-05, "loss": 0.2593, "step": 22563 }, { "epoch": 2.33, "learning_rate": 2.171799821993793e-05, "loss": 0.1732, "step": 22564 }, { "epoch": 2.33, "learning_rate": 2.1715632482162545e-05, "loss": 0.5088, "step": 22565 }, { "epoch": 2.33, "learning_rate": 2.1713266800155584e-05, "loss": 0.2855, "step": 22566 }, { "epoch": 2.33, "learning_rate": 2.171090117393297e-05, "loss": 0.29, "step": 22567 }, { "epoch": 2.33, "learning_rate": 2.1708535603510626e-05, "loss": 0.1307, "step": 22568 }, { "epoch": 2.33, "learning_rate": 2.1706170088904477e-05, "loss": 0.3394, "step": 22569 }, { "epoch": 2.33, "learning_rate": 2.170380463013045e-05, "loss": 0.1351, "step": 22570 }, { "epoch": 2.33, "learning_rate": 2.1701439227204478e-05, "loss": 0.2868, "step": 22571 }, { "epoch": 2.33, "learning_rate": 2.1699073880142465e-05, "loss": 0.4856, "step": 22572 }, { "epoch": 2.33, "learning_rate": 2.169670858896035e-05, "loss": 0.2929, "step": 22573 }, { "epoch": 2.33, "learning_rate": 2.1694343353674044e-05, "loss": 0.354, "step": 22574 }, { "epoch": 2.33, "learning_rate": 2.1691978174299487e-05, "loss": 0.2373, "step": 22575 }, { "epoch": 2.33, "learning_rate": 2.168961305085258e-05, "loss": 0.349, "step": 22576 }, { "epoch": 2.33, "learning_rate": 2.1687247983349253e-05, "loss": 0.5604, "step": 22577 }, { "epoch": 2.33, "learning_rate": 2.1684882971805427e-05, "loss": 0.6854, "step": 22578 }, { "epoch": 2.33, "learning_rate": 2.168251801623704e-05, "loss": 0.2312, "step": 22579 }, { "epoch": 2.33, "learning_rate": 2.1680153116659982e-05, "loss": 0.5185, "step": 22580 }, { "epoch": 2.33, "learning_rate": 2.1677788273090186e-05, "loss": 0.1824, "step": 22581 }, { "epoch": 2.33, "learning_rate": 2.1675423485543574e-05, "loss": 0.5817, "step": 22582 }, { "epoch": 2.33, "learning_rate": 2.167305875403606e-05, "loss": 0.4637, "step": 22583 }, { "epoch": 2.33, "learning_rate": 2.1670694078583566e-05, "loss": 0.4219, "step": 22584 }, { "epoch": 2.33, "learning_rate": 2.166832945920202e-05, "loss": 0.3299, "step": 22585 }, { "epoch": 2.33, "learning_rate": 2.166596489590732e-05, "loss": 0.1295, "step": 22586 }, { "epoch": 2.33, "learning_rate": 2.1663600388715397e-05, "loss": 0.3517, "step": 22587 }, { "epoch": 2.33, "learning_rate": 2.1661235937642163e-05, "loss": 0.5469, "step": 22588 }, { "epoch": 2.33, "learning_rate": 2.1658871542703548e-05, "loss": 0.5047, "step": 22589 }, { "epoch": 2.33, "learning_rate": 2.1656507203915448e-05, "loss": 0.2688, "step": 22590 }, { "epoch": 2.33, "learning_rate": 2.16541429212938e-05, "loss": 0.3965, "step": 22591 }, { "epoch": 2.33, "learning_rate": 2.1651778694854498e-05, "loss": 0.2357, "step": 22592 }, { "epoch": 2.33, "learning_rate": 2.1649414524613472e-05, "loss": 0.0393, "step": 22593 }, { "epoch": 2.33, "learning_rate": 2.1647050410586632e-05, "loss": 0.3204, "step": 22594 }, { "epoch": 2.33, "learning_rate": 2.1644686352789898e-05, "loss": 0.3277, "step": 22595 }, { "epoch": 2.33, "learning_rate": 2.1642322351239177e-05, "loss": 0.6637, "step": 22596 }, { "epoch": 2.33, "learning_rate": 2.1639958405950384e-05, "loss": 0.3431, "step": 22597 }, { "epoch": 2.33, "learning_rate": 2.1637594516939442e-05, "loss": 0.3719, "step": 22598 }, { "epoch": 2.33, "learning_rate": 2.1635230684222252e-05, "loss": 0.5648, "step": 22599 }, { "epoch": 2.33, "learning_rate": 2.1632866907814735e-05, "loss": 0.1267, "step": 22600 }, { "epoch": 2.33, "learning_rate": 2.16305031877328e-05, "loss": 0.4911, "step": 22601 }, { "epoch": 2.33, "learning_rate": 2.162813952399235e-05, "loss": 0.4816, "step": 22602 }, { "epoch": 2.33, "learning_rate": 2.162577591660932e-05, "loss": 0.4046, "step": 22603 }, { "epoch": 2.33, "learning_rate": 2.1623412365599603e-05, "loss": 0.2065, "step": 22604 }, { "epoch": 2.33, "learning_rate": 2.1621048870979117e-05, "loss": 0.2108, "step": 22605 }, { "epoch": 2.33, "learning_rate": 2.1618685432763765e-05, "loss": 0.5751, "step": 22606 }, { "epoch": 2.33, "learning_rate": 2.1616322050969457e-05, "loss": 0.2519, "step": 22607 }, { "epoch": 2.33, "learning_rate": 2.161395872561211e-05, "loss": 0.4765, "step": 22608 }, { "epoch": 2.33, "learning_rate": 2.1611595456707634e-05, "loss": 0.4692, "step": 22609 }, { "epoch": 2.33, "learning_rate": 2.1609232244271933e-05, "loss": 0.621, "step": 22610 }, { "epoch": 2.33, "learning_rate": 2.1606869088320915e-05, "loss": 0.2568, "step": 22611 }, { "epoch": 2.33, "learning_rate": 2.1604505988870488e-05, "loss": 0.4169, "step": 22612 }, { "epoch": 2.34, "learning_rate": 2.1602142945936572e-05, "loss": 0.0591, "step": 22613 }, { "epoch": 2.34, "learning_rate": 2.159977995953506e-05, "loss": 0.7434, "step": 22614 }, { "epoch": 2.34, "learning_rate": 2.159741702968186e-05, "loss": 0.3065, "step": 22615 }, { "epoch": 2.34, "learning_rate": 2.1595054156392894e-05, "loss": 0.4408, "step": 22616 }, { "epoch": 2.34, "learning_rate": 2.1592691339684047e-05, "loss": 0.0681, "step": 22617 }, { "epoch": 2.34, "learning_rate": 2.159032857957123e-05, "loss": 0.5944, "step": 22618 }, { "epoch": 2.34, "learning_rate": 2.1587965876070367e-05, "loss": 0.4467, "step": 22619 }, { "epoch": 2.34, "learning_rate": 2.158560322919734e-05, "loss": 0.2781, "step": 22620 }, { "epoch": 2.34, "learning_rate": 2.1583240638968065e-05, "loss": 0.2103, "step": 22621 }, { "epoch": 2.34, "learning_rate": 2.1580878105398443e-05, "loss": 0.6477, "step": 22622 }, { "epoch": 2.34, "learning_rate": 2.1578515628504384e-05, "loss": 0.5003, "step": 22623 }, { "epoch": 2.34, "learning_rate": 2.1576153208301783e-05, "loss": 0.419, "step": 22624 }, { "epoch": 2.34, "learning_rate": 2.1573790844806548e-05, "loss": 0.7004, "step": 22625 }, { "epoch": 2.34, "learning_rate": 2.157142853803458e-05, "loss": 0.1504, "step": 22626 }, { "epoch": 2.34, "learning_rate": 2.156906628800179e-05, "loss": 1.2926, "step": 22627 }, { "epoch": 2.34, "learning_rate": 2.156670409472407e-05, "loss": 0.6836, "step": 22628 }, { "epoch": 2.34, "learning_rate": 2.1564341958217333e-05, "loss": 0.4167, "step": 22629 }, { "epoch": 2.34, "learning_rate": 2.156197987849746e-05, "loss": 0.4206, "step": 22630 }, { "epoch": 2.34, "learning_rate": 2.1559617855580366e-05, "loss": 0.2324, "step": 22631 }, { "epoch": 2.34, "learning_rate": 2.1557255889481947e-05, "loss": 0.1675, "step": 22632 }, { "epoch": 2.34, "learning_rate": 2.1554893980218114e-05, "loss": 1.8127, "step": 22633 }, { "epoch": 2.34, "learning_rate": 2.155253212780475e-05, "loss": 0.5406, "step": 22634 }, { "epoch": 2.34, "learning_rate": 2.155017033225777e-05, "loss": 0.2111, "step": 22635 }, { "epoch": 2.34, "learning_rate": 2.1547808593593062e-05, "loss": 0.086, "step": 22636 }, { "epoch": 2.34, "learning_rate": 2.154544691182653e-05, "loss": 0.7384, "step": 22637 }, { "epoch": 2.34, "learning_rate": 2.1543085286974074e-05, "loss": 0.2381, "step": 22638 }, { "epoch": 2.34, "learning_rate": 2.1540723719051583e-05, "loss": 0.6659, "step": 22639 }, { "epoch": 2.34, "learning_rate": 2.1538362208074965e-05, "loss": 0.5404, "step": 22640 }, { "epoch": 2.34, "learning_rate": 2.1536000754060116e-05, "loss": 0.2201, "step": 22641 }, { "epoch": 2.34, "learning_rate": 2.153363935702293e-05, "loss": 0.2008, "step": 22642 }, { "epoch": 2.34, "learning_rate": 2.1531278016979302e-05, "loss": 0.5647, "step": 22643 }, { "epoch": 2.34, "learning_rate": 2.1528916733945125e-05, "loss": 0.5524, "step": 22644 }, { "epoch": 2.34, "learning_rate": 2.1526555507936295e-05, "loss": 0.49, "step": 22645 }, { "epoch": 2.34, "learning_rate": 2.152419433896871e-05, "loss": 0.8013, "step": 22646 }, { "epoch": 2.34, "learning_rate": 2.152183322705828e-05, "loss": 0.2492, "step": 22647 }, { "epoch": 2.34, "learning_rate": 2.151947217222087e-05, "loss": 0.2045, "step": 22648 }, { "epoch": 2.34, "learning_rate": 2.1517111174472393e-05, "loss": 0.5831, "step": 22649 }, { "epoch": 2.34, "learning_rate": 2.1514750233828737e-05, "loss": 0.2247, "step": 22650 }, { "epoch": 2.34, "learning_rate": 2.1512389350305804e-05, "loss": 0.0318, "step": 22651 }, { "epoch": 2.34, "learning_rate": 2.1510028523919476e-05, "loss": 0.9311, "step": 22652 }, { "epoch": 2.34, "learning_rate": 2.1507667754685648e-05, "loss": 0.3064, "step": 22653 }, { "epoch": 2.34, "learning_rate": 2.1505307042620226e-05, "loss": 0.1547, "step": 22654 }, { "epoch": 2.34, "learning_rate": 2.1502946387739072e-05, "loss": 0.5668, "step": 22655 }, { "epoch": 2.34, "learning_rate": 2.1500585790058104e-05, "loss": 0.5734, "step": 22656 }, { "epoch": 2.34, "learning_rate": 2.14982252495932e-05, "loss": 0.1487, "step": 22657 }, { "epoch": 2.34, "learning_rate": 2.149586476636026e-05, "loss": 0.11, "step": 22658 }, { "epoch": 2.34, "learning_rate": 2.149350434037516e-05, "loss": 0.3397, "step": 22659 }, { "epoch": 2.34, "learning_rate": 2.1491143971653797e-05, "loss": 0.8493, "step": 22660 }, { "epoch": 2.34, "learning_rate": 2.1488783660212075e-05, "loss": 0.2072, "step": 22661 }, { "epoch": 2.34, "learning_rate": 2.148642340606586e-05, "loss": 0.7169, "step": 22662 }, { "epoch": 2.34, "learning_rate": 2.1484063209231048e-05, "loss": 0.2824, "step": 22663 }, { "epoch": 2.34, "learning_rate": 2.1481703069723534e-05, "loss": 0.228, "step": 22664 }, { "epoch": 2.34, "learning_rate": 2.1479342987559207e-05, "loss": 0.3665, "step": 22665 }, { "epoch": 2.34, "learning_rate": 2.1476982962753943e-05, "loss": 0.44, "step": 22666 }, { "epoch": 2.34, "learning_rate": 2.147462299532365e-05, "loss": 0.4611, "step": 22667 }, { "epoch": 2.34, "learning_rate": 2.1472263085284186e-05, "loss": 0.492, "step": 22668 }, { "epoch": 2.34, "learning_rate": 2.1469903232651447e-05, "loss": 0.324, "step": 22669 }, { "epoch": 2.34, "learning_rate": 2.1467543437441328e-05, "loss": 0.4895, "step": 22670 }, { "epoch": 2.34, "learning_rate": 2.146518369966972e-05, "loss": 0.272, "step": 22671 }, { "epoch": 2.34, "learning_rate": 2.1462824019352487e-05, "loss": 0.092, "step": 22672 }, { "epoch": 2.34, "learning_rate": 2.1460464396505525e-05, "loss": 0.2773, "step": 22673 }, { "epoch": 2.34, "learning_rate": 2.145810483114472e-05, "loss": 0.1787, "step": 22674 }, { "epoch": 2.34, "learning_rate": 2.145574532328596e-05, "loss": 0.3132, "step": 22675 }, { "epoch": 2.34, "learning_rate": 2.145338587294512e-05, "loss": 0.5261, "step": 22676 }, { "epoch": 2.34, "learning_rate": 2.145102648013808e-05, "loss": 0.253, "step": 22677 }, { "epoch": 2.34, "learning_rate": 2.1448667144880737e-05, "loss": 0.3962, "step": 22678 }, { "epoch": 2.34, "learning_rate": 2.144630786718897e-05, "loss": 0.3657, "step": 22679 }, { "epoch": 2.34, "learning_rate": 2.1443948647078654e-05, "loss": 0.184, "step": 22680 }, { "epoch": 2.34, "learning_rate": 2.1441589484565677e-05, "loss": 0.3275, "step": 22681 }, { "epoch": 2.34, "learning_rate": 2.143923037966591e-05, "loss": 0.2168, "step": 22682 }, { "epoch": 2.34, "learning_rate": 2.143687133239524e-05, "loss": 0.5218, "step": 22683 }, { "epoch": 2.34, "learning_rate": 2.1434512342769547e-05, "loss": 0.4252, "step": 22684 }, { "epoch": 2.34, "learning_rate": 2.1432153410804724e-05, "loss": 0.6245, "step": 22685 }, { "epoch": 2.34, "learning_rate": 2.142979453651663e-05, "loss": 0.5503, "step": 22686 }, { "epoch": 2.34, "learning_rate": 2.1427435719921156e-05, "loss": 0.1816, "step": 22687 }, { "epoch": 2.34, "learning_rate": 2.1425076961034173e-05, "loss": 0.561, "step": 22688 }, { "epoch": 2.34, "learning_rate": 2.142271825987158e-05, "loss": 0.2676, "step": 22689 }, { "epoch": 2.34, "learning_rate": 2.1420359616449227e-05, "loss": 0.1816, "step": 22690 }, { "epoch": 2.34, "learning_rate": 2.1418001030783014e-05, "loss": 0.3379, "step": 22691 }, { "epoch": 2.34, "learning_rate": 2.141564250288881e-05, "loss": 0.3509, "step": 22692 }, { "epoch": 2.34, "learning_rate": 2.1413284032782493e-05, "loss": 0.1806, "step": 22693 }, { "epoch": 2.34, "learning_rate": 2.1410925620479932e-05, "loss": 0.4277, "step": 22694 }, { "epoch": 2.34, "learning_rate": 2.1408567265997016e-05, "loss": 0.1589, "step": 22695 }, { "epoch": 2.34, "learning_rate": 2.140620896934961e-05, "loss": 0.1221, "step": 22696 }, { "epoch": 2.34, "learning_rate": 2.1403850730553594e-05, "loss": 0.474, "step": 22697 }, { "epoch": 2.34, "learning_rate": 2.1401492549624846e-05, "loss": 0.5626, "step": 22698 }, { "epoch": 2.34, "learning_rate": 2.1399134426579243e-05, "loss": 0.24, "step": 22699 }, { "epoch": 2.34, "learning_rate": 2.1396776361432647e-05, "loss": 0.134, "step": 22700 }, { "epoch": 2.34, "learning_rate": 2.139441835420094e-05, "loss": 0.5038, "step": 22701 }, { "epoch": 2.34, "learning_rate": 2.13920604049e-05, "loss": 0.6495, "step": 22702 }, { "epoch": 2.34, "learning_rate": 2.1389702513545693e-05, "loss": 0.6452, "step": 22703 }, { "epoch": 2.34, "learning_rate": 2.13873446801539e-05, "loss": 0.3553, "step": 22704 }, { "epoch": 2.34, "learning_rate": 2.1384986904740483e-05, "loss": 0.0952, "step": 22705 }, { "epoch": 2.34, "learning_rate": 2.1382629187321313e-05, "loss": 0.2488, "step": 22706 }, { "epoch": 2.34, "learning_rate": 2.138027152791227e-05, "loss": 0.112, "step": 22707 }, { "epoch": 2.34, "learning_rate": 2.137791392652922e-05, "loss": 0.844, "step": 22708 }, { "epoch": 2.35, "learning_rate": 2.1375556383188048e-05, "loss": 0.3341, "step": 22709 }, { "epoch": 2.35, "learning_rate": 2.13731988979046e-05, "loss": 0.4599, "step": 22710 }, { "epoch": 2.35, "learning_rate": 2.1370841470694756e-05, "loss": 0.1979, "step": 22711 }, { "epoch": 2.35, "learning_rate": 2.1368484101574392e-05, "loss": 0.6749, "step": 22712 }, { "epoch": 2.35, "learning_rate": 2.136612679055938e-05, "loss": 0.3869, "step": 22713 }, { "epoch": 2.35, "learning_rate": 2.1363769537665577e-05, "loss": 0.4467, "step": 22714 }, { "epoch": 2.35, "learning_rate": 2.1361412342908856e-05, "loss": 0.2761, "step": 22715 }, { "epoch": 2.35, "learning_rate": 2.1359055206305083e-05, "loss": 0.2097, "step": 22716 }, { "epoch": 2.35, "learning_rate": 2.1356698127870146e-05, "loss": 0.0844, "step": 22717 }, { "epoch": 2.35, "learning_rate": 2.1354341107619878e-05, "loss": 0.4268, "step": 22718 }, { "epoch": 2.35, "learning_rate": 2.1351984145570172e-05, "loss": 0.1387, "step": 22719 }, { "epoch": 2.35, "learning_rate": 2.1349627241736878e-05, "loss": 0.4208, "step": 22720 }, { "epoch": 2.35, "learning_rate": 2.134727039613587e-05, "loss": 0.6055, "step": 22721 }, { "epoch": 2.35, "learning_rate": 2.1344913608783008e-05, "loss": 0.3943, "step": 22722 }, { "epoch": 2.35, "learning_rate": 2.1342556879694173e-05, "loss": 0.284, "step": 22723 }, { "epoch": 2.35, "learning_rate": 2.1340200208885216e-05, "loss": 0.7492, "step": 22724 }, { "epoch": 2.35, "learning_rate": 2.1337843596372e-05, "loss": 0.9861, "step": 22725 }, { "epoch": 2.35, "learning_rate": 2.1335487042170396e-05, "loss": 0.0751, "step": 22726 }, { "epoch": 2.35, "learning_rate": 2.1333130546296274e-05, "loss": 0.2237, "step": 22727 }, { "epoch": 2.35, "learning_rate": 2.133077410876548e-05, "loss": 0.6145, "step": 22728 }, { "epoch": 2.35, "learning_rate": 2.13284177295939e-05, "loss": 0.2449, "step": 22729 }, { "epoch": 2.35, "learning_rate": 2.132606140879737e-05, "loss": 0.3071, "step": 22730 }, { "epoch": 2.35, "learning_rate": 2.1323705146391767e-05, "loss": 0.5551, "step": 22731 }, { "epoch": 2.35, "learning_rate": 2.132134894239295e-05, "loss": 0.5237, "step": 22732 }, { "epoch": 2.35, "learning_rate": 2.1318992796816787e-05, "loss": 0.2821, "step": 22733 }, { "epoch": 2.35, "learning_rate": 2.1316636709679125e-05, "loss": 0.7658, "step": 22734 }, { "epoch": 2.35, "learning_rate": 2.1314280680995837e-05, "loss": 0.0513, "step": 22735 }, { "epoch": 2.35, "learning_rate": 2.131192471078278e-05, "loss": 0.4413, "step": 22736 }, { "epoch": 2.35, "learning_rate": 2.130956879905581e-05, "loss": 0.1113, "step": 22737 }, { "epoch": 2.35, "learning_rate": 2.130721294583079e-05, "loss": 0.5764, "step": 22738 }, { "epoch": 2.35, "learning_rate": 2.130485715112358e-05, "loss": 0.0397, "step": 22739 }, { "epoch": 2.35, "learning_rate": 2.1302501414950034e-05, "loss": 0.3187, "step": 22740 }, { "epoch": 2.35, "learning_rate": 2.130014573732602e-05, "loss": 0.3172, "step": 22741 }, { "epoch": 2.35, "learning_rate": 2.1297790118267393e-05, "loss": 0.488, "step": 22742 }, { "epoch": 2.35, "learning_rate": 2.1295434557790002e-05, "loss": 0.0926, "step": 22743 }, { "epoch": 2.35, "learning_rate": 2.1293079055909704e-05, "loss": 0.19, "step": 22744 }, { "epoch": 2.35, "learning_rate": 2.1290723612642358e-05, "loss": 0.7076, "step": 22745 }, { "epoch": 2.35, "learning_rate": 2.128836822800383e-05, "loss": 0.2254, "step": 22746 }, { "epoch": 2.35, "learning_rate": 2.128601290200997e-05, "loss": 0.4936, "step": 22747 }, { "epoch": 2.35, "learning_rate": 2.1283657634676627e-05, "loss": 0.1574, "step": 22748 }, { "epoch": 2.35, "learning_rate": 2.128130242601966e-05, "loss": 0.1012, "step": 22749 }, { "epoch": 2.35, "learning_rate": 2.1278947276054925e-05, "loss": 0.1124, "step": 22750 }, { "epoch": 2.35, "learning_rate": 2.1276592184798286e-05, "loss": 0.504, "step": 22751 }, { "epoch": 2.35, "learning_rate": 2.127423715226558e-05, "loss": 0.2053, "step": 22752 }, { "epoch": 2.35, "learning_rate": 2.1271882178472664e-05, "loss": 0.5177, "step": 22753 }, { "epoch": 2.35, "learning_rate": 2.12695272634354e-05, "loss": 0.1644, "step": 22754 }, { "epoch": 2.35, "learning_rate": 2.1267172407169644e-05, "loss": 0.0695, "step": 22755 }, { "epoch": 2.35, "learning_rate": 2.126481760969123e-05, "loss": 0.1877, "step": 22756 }, { "epoch": 2.35, "learning_rate": 2.1262462871016026e-05, "loss": 0.2416, "step": 22757 }, { "epoch": 2.35, "learning_rate": 2.1260108191159868e-05, "loss": 0.4676, "step": 22758 }, { "epoch": 2.35, "learning_rate": 2.1257753570138622e-05, "loss": 0.1618, "step": 22759 }, { "epoch": 2.35, "learning_rate": 2.125539900796813e-05, "loss": 0.5136, "step": 22760 }, { "epoch": 2.35, "learning_rate": 2.125304450466425e-05, "loss": 0.1053, "step": 22761 }, { "epoch": 2.35, "learning_rate": 2.1250690060242825e-05, "loss": 0.2013, "step": 22762 }, { "epoch": 2.35, "learning_rate": 2.1248335674719714e-05, "loss": 0.6576, "step": 22763 }, { "epoch": 2.35, "learning_rate": 2.1245981348110745e-05, "loss": 0.6596, "step": 22764 }, { "epoch": 2.35, "learning_rate": 2.12436270804318e-05, "loss": 0.3554, "step": 22765 }, { "epoch": 2.35, "learning_rate": 2.1241272871698695e-05, "loss": 0.1407, "step": 22766 }, { "epoch": 2.35, "learning_rate": 2.1238918721927306e-05, "loss": 0.4184, "step": 22767 }, { "epoch": 2.35, "learning_rate": 2.1236564631133455e-05, "loss": 0.0747, "step": 22768 }, { "epoch": 2.35, "learning_rate": 2.1234210599333e-05, "loss": 0.1307, "step": 22769 }, { "epoch": 2.35, "learning_rate": 2.123185662654179e-05, "loss": 0.3635, "step": 22770 }, { "epoch": 2.35, "learning_rate": 2.122950271277567e-05, "loss": 0.3743, "step": 22771 }, { "epoch": 2.35, "learning_rate": 2.1227148858050488e-05, "loss": 0.4243, "step": 22772 }, { "epoch": 2.35, "learning_rate": 2.122479506238208e-05, "loss": 0.6761, "step": 22773 }, { "epoch": 2.35, "learning_rate": 2.1222441325786297e-05, "loss": 0.2999, "step": 22774 }, { "epoch": 2.35, "learning_rate": 2.1220087648278995e-05, "loss": 0.2609, "step": 22775 }, { "epoch": 2.35, "learning_rate": 2.1217734029876002e-05, "loss": 0.248, "step": 22776 }, { "epoch": 2.35, "learning_rate": 2.1215380470593175e-05, "loss": 0.1695, "step": 22777 }, { "epoch": 2.35, "learning_rate": 2.1213026970446342e-05, "loss": 0.0225, "step": 22778 }, { "epoch": 2.35, "learning_rate": 2.1210673529451367e-05, "loss": 0.2227, "step": 22779 }, { "epoch": 2.35, "learning_rate": 2.120832014762408e-05, "loss": 0.5196, "step": 22780 }, { "epoch": 2.35, "learning_rate": 2.1205966824980323e-05, "loss": 0.2457, "step": 22781 }, { "epoch": 2.35, "learning_rate": 2.1203613561535936e-05, "loss": 0.0231, "step": 22782 }, { "epoch": 2.35, "learning_rate": 2.120126035730676e-05, "loss": 0.5625, "step": 22783 }, { "epoch": 2.35, "learning_rate": 2.1198907212308643e-05, "loss": 0.7213, "step": 22784 }, { "epoch": 2.35, "learning_rate": 2.119655412655743e-05, "loss": 0.321, "step": 22785 }, { "epoch": 2.35, "learning_rate": 2.1194201100068947e-05, "loss": 0.1368, "step": 22786 }, { "epoch": 2.35, "learning_rate": 2.1191848132859046e-05, "loss": 0.6017, "step": 22787 }, { "epoch": 2.35, "learning_rate": 2.118949522494356e-05, "loss": 0.3367, "step": 22788 }, { "epoch": 2.35, "learning_rate": 2.118714237633834e-05, "loss": 0.5546, "step": 22789 }, { "epoch": 2.35, "learning_rate": 2.11847895870592e-05, "loss": 0.1314, "step": 22790 }, { "epoch": 2.35, "learning_rate": 2.1182436857122e-05, "loss": 0.757, "step": 22791 }, { "epoch": 2.35, "learning_rate": 2.1180084186542585e-05, "loss": 0.3837, "step": 22792 }, { "epoch": 2.35, "learning_rate": 2.1177731575336765e-05, "loss": 0.3885, "step": 22793 }, { "epoch": 2.35, "learning_rate": 2.1175379023520397e-05, "loss": 0.3571, "step": 22794 }, { "epoch": 2.35, "learning_rate": 2.1173026531109312e-05, "loss": 0.4034, "step": 22795 }, { "epoch": 2.35, "learning_rate": 2.1170674098119344e-05, "loss": 0.3262, "step": 22796 }, { "epoch": 2.35, "learning_rate": 2.116832172456633e-05, "loss": 0.2893, "step": 22797 }, { "epoch": 2.35, "learning_rate": 2.116596941046611e-05, "loss": 0.7845, "step": 22798 }, { "epoch": 2.35, "learning_rate": 2.1163617155834525e-05, "loss": 0.3207, "step": 22799 }, { "epoch": 2.35, "learning_rate": 2.1161264960687393e-05, "loss": 0.5972, "step": 22800 }, { "epoch": 2.35, "learning_rate": 2.1158912825040558e-05, "loss": 0.5288, "step": 22801 }, { "epoch": 2.35, "learning_rate": 2.1156560748909853e-05, "loss": 0.3665, "step": 22802 }, { "epoch": 2.35, "learning_rate": 2.1154208732311117e-05, "loss": 0.1485, "step": 22803 }, { "epoch": 2.35, "learning_rate": 2.1151856775260172e-05, "loss": 0.2097, "step": 22804 }, { "epoch": 2.36, "learning_rate": 2.114950487777287e-05, "loss": 0.3205, "step": 22805 }, { "epoch": 2.36, "learning_rate": 2.1147153039865017e-05, "loss": 0.0728, "step": 22806 }, { "epoch": 2.36, "learning_rate": 2.114480126155246e-05, "loss": 0.2461, "step": 22807 }, { "epoch": 2.36, "learning_rate": 2.114244954285103e-05, "loss": 0.2097, "step": 22808 }, { "epoch": 2.36, "learning_rate": 2.114009788377656e-05, "loss": 0.247, "step": 22809 }, { "epoch": 2.36, "learning_rate": 2.1137746284344873e-05, "loss": 0.2317, "step": 22810 }, { "epoch": 2.36, "learning_rate": 2.11353947445718e-05, "loss": 0.6267, "step": 22811 }, { "epoch": 2.36, "learning_rate": 2.1133043264473183e-05, "loss": 0.4403, "step": 22812 }, { "epoch": 2.36, "learning_rate": 2.1130691844064845e-05, "loss": 0.1488, "step": 22813 }, { "epoch": 2.36, "learning_rate": 2.112834048336261e-05, "loss": 0.3258, "step": 22814 }, { "epoch": 2.36, "learning_rate": 2.112598918238231e-05, "loss": 0.894, "step": 22815 }, { "epoch": 2.36, "learning_rate": 2.1123637941139778e-05, "loss": 0.2457, "step": 22816 }, { "epoch": 2.36, "learning_rate": 2.1121286759650852e-05, "loss": 0.0835, "step": 22817 }, { "epoch": 2.36, "learning_rate": 2.111893563793133e-05, "loss": 0.1632, "step": 22818 }, { "epoch": 2.36, "learning_rate": 2.1116584575997058e-05, "loss": 0.4306, "step": 22819 }, { "epoch": 2.36, "learning_rate": 2.111423357386386e-05, "loss": 0.505, "step": 22820 }, { "epoch": 2.36, "learning_rate": 2.111188263154756e-05, "loss": 0.3312, "step": 22821 }, { "epoch": 2.36, "learning_rate": 2.1109531749063988e-05, "loss": 0.2848, "step": 22822 }, { "epoch": 2.36, "learning_rate": 2.1107180926428973e-05, "loss": 0.1819, "step": 22823 }, { "epoch": 2.36, "learning_rate": 2.1104830163658326e-05, "loss": 0.4542, "step": 22824 }, { "epoch": 2.36, "learning_rate": 2.1102479460767884e-05, "loss": 0.1815, "step": 22825 }, { "epoch": 2.36, "learning_rate": 2.110012881777347e-05, "loss": 0.5528, "step": 22826 }, { "epoch": 2.36, "learning_rate": 2.109777823469091e-05, "loss": 0.3348, "step": 22827 }, { "epoch": 2.36, "learning_rate": 2.1095427711536016e-05, "loss": 0.3622, "step": 22828 }, { "epoch": 2.36, "learning_rate": 2.1093077248324626e-05, "loss": 0.3919, "step": 22829 }, { "epoch": 2.36, "learning_rate": 2.1090726845072566e-05, "loss": 0.552, "step": 22830 }, { "epoch": 2.36, "learning_rate": 2.108837650179563e-05, "loss": 0.2448, "step": 22831 }, { "epoch": 2.36, "learning_rate": 2.1086026218509664e-05, "loss": 0.7824, "step": 22832 }, { "epoch": 2.36, "learning_rate": 2.108367599523049e-05, "loss": 0.2399, "step": 22833 }, { "epoch": 2.36, "learning_rate": 2.1081325831973914e-05, "loss": 0.6288, "step": 22834 }, { "epoch": 2.36, "learning_rate": 2.1078975728755764e-05, "loss": 0.1567, "step": 22835 }, { "epoch": 2.36, "learning_rate": 2.107662568559187e-05, "loss": 0.2746, "step": 22836 }, { "epoch": 2.36, "learning_rate": 2.107427570249804e-05, "loss": 0.6746, "step": 22837 }, { "epoch": 2.36, "learning_rate": 2.10719257794901e-05, "loss": 0.5619, "step": 22838 }, { "epoch": 2.36, "learning_rate": 2.106957591658386e-05, "loss": 0.5871, "step": 22839 }, { "epoch": 2.36, "learning_rate": 2.1067226113795148e-05, "loss": 0.3689, "step": 22840 }, { "epoch": 2.36, "learning_rate": 2.1064876371139788e-05, "loss": 0.6727, "step": 22841 }, { "epoch": 2.36, "learning_rate": 2.106252668863358e-05, "loss": 0.3823, "step": 22842 }, { "epoch": 2.36, "learning_rate": 2.106017706629237e-05, "loss": 0.3086, "step": 22843 }, { "epoch": 2.36, "learning_rate": 2.1057827504131937e-05, "loss": 0.2774, "step": 22844 }, { "epoch": 2.36, "learning_rate": 2.1055478002168118e-05, "loss": 0.3984, "step": 22845 }, { "epoch": 2.36, "learning_rate": 2.105312856041673e-05, "loss": 0.3763, "step": 22846 }, { "epoch": 2.36, "learning_rate": 2.1050779178893593e-05, "loss": 0.8556, "step": 22847 }, { "epoch": 2.36, "learning_rate": 2.104842985761451e-05, "loss": 0.5363, "step": 22848 }, { "epoch": 2.36, "learning_rate": 2.10460805965953e-05, "loss": 0.3928, "step": 22849 }, { "epoch": 2.36, "learning_rate": 2.1043731395851783e-05, "loss": 0.1833, "step": 22850 }, { "epoch": 2.36, "learning_rate": 2.104138225539978e-05, "loss": 0.514, "step": 22851 }, { "epoch": 2.36, "learning_rate": 2.1039033175255084e-05, "loss": 0.131, "step": 22852 }, { "epoch": 2.36, "learning_rate": 2.103668415543352e-05, "loss": 0.0703, "step": 22853 }, { "epoch": 2.36, "learning_rate": 2.1034335195950903e-05, "loss": 0.062, "step": 22854 }, { "epoch": 2.36, "learning_rate": 2.103198629682306e-05, "loss": 0.2, "step": 22855 }, { "epoch": 2.36, "learning_rate": 2.102963745806577e-05, "loss": 0.5403, "step": 22856 }, { "epoch": 2.36, "learning_rate": 2.1027288679694868e-05, "loss": 1.0111, "step": 22857 }, { "epoch": 2.36, "learning_rate": 2.1024939961726154e-05, "loss": 0.6078, "step": 22858 }, { "epoch": 2.36, "learning_rate": 2.1022591304175443e-05, "loss": 0.2083, "step": 22859 }, { "epoch": 2.36, "learning_rate": 2.1020242707058548e-05, "loss": 0.1525, "step": 22860 }, { "epoch": 2.36, "learning_rate": 2.1017894170391283e-05, "loss": 0.2028, "step": 22861 }, { "epoch": 2.36, "learning_rate": 2.1015545694189445e-05, "loss": 0.2939, "step": 22862 }, { "epoch": 2.36, "learning_rate": 2.101319727846886e-05, "loss": 0.6325, "step": 22863 }, { "epoch": 2.36, "learning_rate": 2.101084892324532e-05, "loss": 0.5608, "step": 22864 }, { "epoch": 2.36, "learning_rate": 2.1008500628534646e-05, "loss": 0.1775, "step": 22865 }, { "epoch": 2.36, "learning_rate": 2.1006152394352642e-05, "loss": 0.4522, "step": 22866 }, { "epoch": 2.36, "learning_rate": 2.1003804220715114e-05, "loss": 0.283, "step": 22867 }, { "epoch": 2.36, "learning_rate": 2.1001456107637882e-05, "loss": 0.0854, "step": 22868 }, { "epoch": 2.36, "learning_rate": 2.099910805513673e-05, "loss": 1.1324, "step": 22869 }, { "epoch": 2.36, "learning_rate": 2.0996760063227475e-05, "loss": 0.0909, "step": 22870 }, { "epoch": 2.36, "learning_rate": 2.0994412131925933e-05, "loss": 0.6578, "step": 22871 }, { "epoch": 2.36, "learning_rate": 2.0992064261247895e-05, "loss": 0.2538, "step": 22872 }, { "epoch": 2.36, "learning_rate": 2.0989716451209173e-05, "loss": 0.1426, "step": 22873 }, { "epoch": 2.36, "learning_rate": 2.098736870182557e-05, "loss": 0.4578, "step": 22874 }, { "epoch": 2.36, "learning_rate": 2.09850210131129e-05, "loss": 0.1312, "step": 22875 }, { "epoch": 2.36, "learning_rate": 2.0982673385086952e-05, "loss": 0.2329, "step": 22876 }, { "epoch": 2.36, "learning_rate": 2.0980325817763535e-05, "loss": 0.2553, "step": 22877 }, { "epoch": 2.36, "learning_rate": 2.097797831115846e-05, "loss": 0.636, "step": 22878 }, { "epoch": 2.36, "learning_rate": 2.0975630865287525e-05, "loss": 0.2245, "step": 22879 }, { "epoch": 2.36, "learning_rate": 2.0973283480166537e-05, "loss": 0.0645, "step": 22880 }, { "epoch": 2.36, "learning_rate": 2.0970936155811285e-05, "loss": 0.3142, "step": 22881 }, { "epoch": 2.36, "learning_rate": 2.0968588892237576e-05, "loss": 0.8514, "step": 22882 }, { "epoch": 2.36, "learning_rate": 2.0966241689461215e-05, "loss": 0.2528, "step": 22883 }, { "epoch": 2.36, "learning_rate": 2.0963894547498e-05, "loss": 0.6813, "step": 22884 }, { "epoch": 2.36, "learning_rate": 2.0961547466363738e-05, "loss": 0.6127, "step": 22885 }, { "epoch": 2.36, "learning_rate": 2.0959200446074215e-05, "loss": 0.0766, "step": 22886 }, { "epoch": 2.36, "learning_rate": 2.0956853486645244e-05, "loss": 0.2326, "step": 22887 }, { "epoch": 2.36, "learning_rate": 2.0954506588092614e-05, "loss": 0.6334, "step": 22888 }, { "epoch": 2.36, "learning_rate": 2.095215975043214e-05, "loss": 0.6311, "step": 22889 }, { "epoch": 2.36, "learning_rate": 2.09498129736796e-05, "loss": 0.1655, "step": 22890 }, { "epoch": 2.36, "learning_rate": 2.0947466257850808e-05, "loss": 0.5703, "step": 22891 }, { "epoch": 2.36, "learning_rate": 2.0945119602961548e-05, "loss": 0.2526, "step": 22892 }, { "epoch": 2.36, "learning_rate": 2.094277300902764e-05, "loss": 0.3093, "step": 22893 }, { "epoch": 2.36, "learning_rate": 2.0940426476064847e-05, "loss": 0.1196, "step": 22894 }, { "epoch": 2.36, "learning_rate": 2.0938080004088998e-05, "loss": 0.2262, "step": 22895 }, { "epoch": 2.36, "learning_rate": 2.093573359311586e-05, "loss": 0.2075, "step": 22896 }, { "epoch": 2.36, "learning_rate": 2.093338724316125e-05, "loss": 0.2729, "step": 22897 }, { "epoch": 2.36, "learning_rate": 2.093104095424095e-05, "loss": 0.8538, "step": 22898 }, { "epoch": 2.36, "learning_rate": 2.0928694726370768e-05, "loss": 0.004, "step": 22899 }, { "epoch": 2.37, "learning_rate": 2.0926348559566483e-05, "loss": 0.1404, "step": 22900 }, { "epoch": 2.37, "learning_rate": 2.0924002453843903e-05, "loss": 0.5594, "step": 22901 }, { "epoch": 2.37, "learning_rate": 2.0921656409218808e-05, "loss": 0.4275, "step": 22902 }, { "epoch": 2.37, "learning_rate": 2.0919310425707008e-05, "loss": 0.1644, "step": 22903 }, { "epoch": 2.37, "learning_rate": 2.0916964503324275e-05, "loss": 0.3366, "step": 22904 }, { "epoch": 2.37, "learning_rate": 2.0914618642086428e-05, "loss": 0.4256, "step": 22905 }, { "epoch": 2.37, "learning_rate": 2.091227284200923e-05, "loss": 0.4502, "step": 22906 }, { "epoch": 2.37, "learning_rate": 2.0909927103108482e-05, "loss": 0.5621, "step": 22907 }, { "epoch": 2.37, "learning_rate": 2.0907581425399978e-05, "loss": 0.6303, "step": 22908 }, { "epoch": 2.37, "learning_rate": 2.0905235808899517e-05, "loss": 0.2691, "step": 22909 }, { "epoch": 2.37, "learning_rate": 2.0902890253622873e-05, "loss": 0.2693, "step": 22910 }, { "epoch": 2.37, "learning_rate": 2.0900544759585843e-05, "loss": 0.1523, "step": 22911 }, { "epoch": 2.37, "learning_rate": 2.089819932680422e-05, "loss": 0.728, "step": 22912 }, { "epoch": 2.37, "learning_rate": 2.0895853955293793e-05, "loss": 0.6802, "step": 22913 }, { "epoch": 2.37, "learning_rate": 2.089350864507034e-05, "loss": 0.1444, "step": 22914 }, { "epoch": 2.37, "learning_rate": 2.0891163396149655e-05, "loss": 0.5376, "step": 22915 }, { "epoch": 2.37, "learning_rate": 2.0888818208547526e-05, "loss": 0.2941, "step": 22916 }, { "epoch": 2.37, "learning_rate": 2.088647308227975e-05, "loss": 0.3785, "step": 22917 }, { "epoch": 2.37, "learning_rate": 2.0884128017362107e-05, "loss": 0.8559, "step": 22918 }, { "epoch": 2.37, "learning_rate": 2.0881783013810378e-05, "loss": 0.3002, "step": 22919 }, { "epoch": 2.37, "learning_rate": 2.087943807164035e-05, "loss": 0.1386, "step": 22920 }, { "epoch": 2.37, "learning_rate": 2.0877093190867807e-05, "loss": 0.3105, "step": 22921 }, { "epoch": 2.37, "learning_rate": 2.0874748371508535e-05, "loss": 0.3714, "step": 22922 }, { "epoch": 2.37, "learning_rate": 2.0872403613578333e-05, "loss": 0.0457, "step": 22923 }, { "epoch": 2.37, "learning_rate": 2.0870058917092964e-05, "loss": 0.3138, "step": 22924 }, { "epoch": 2.37, "learning_rate": 2.0867714282068228e-05, "loss": 0.2996, "step": 22925 }, { "epoch": 2.37, "learning_rate": 2.08653697085199e-05, "loss": 0.1997, "step": 22926 }, { "epoch": 2.37, "learning_rate": 2.086302519646377e-05, "loss": 0.1752, "step": 22927 }, { "epoch": 2.37, "learning_rate": 2.086068074591561e-05, "loss": 0.2846, "step": 22928 }, { "epoch": 2.37, "learning_rate": 2.0858336356891215e-05, "loss": 0.3978, "step": 22929 }, { "epoch": 2.37, "learning_rate": 2.0855992029406368e-05, "loss": 0.633, "step": 22930 }, { "epoch": 2.37, "learning_rate": 2.0853647763476828e-05, "loss": 0.4258, "step": 22931 }, { "epoch": 2.37, "learning_rate": 2.0851303559118398e-05, "loss": 0.3008, "step": 22932 }, { "epoch": 2.37, "learning_rate": 2.0848959416346852e-05, "loss": 0.6524, "step": 22933 }, { "epoch": 2.37, "learning_rate": 2.0846615335177967e-05, "loss": 0.2016, "step": 22934 }, { "epoch": 2.37, "learning_rate": 2.0844271315627526e-05, "loss": 0.0863, "step": 22935 }, { "epoch": 2.37, "learning_rate": 2.0841927357711306e-05, "loss": 0.1508, "step": 22936 }, { "epoch": 2.37, "learning_rate": 2.0839583461445094e-05, "loss": 0.201, "step": 22937 }, { "epoch": 2.37, "learning_rate": 2.083723962684466e-05, "loss": 0.0886, "step": 22938 }, { "epoch": 2.37, "learning_rate": 2.0834895853925785e-05, "loss": 0.1485, "step": 22939 }, { "epoch": 2.37, "learning_rate": 2.083255214270424e-05, "loss": 0.2996, "step": 22940 }, { "epoch": 2.37, "learning_rate": 2.0830208493195828e-05, "loss": 0.6137, "step": 22941 }, { "epoch": 2.37, "learning_rate": 2.0827864905416288e-05, "loss": 0.405, "step": 22942 }, { "epoch": 2.37, "learning_rate": 2.0825521379381432e-05, "loss": 0.0727, "step": 22943 }, { "epoch": 2.37, "learning_rate": 2.0823177915107008e-05, "loss": 0.2647, "step": 22944 }, { "epoch": 2.37, "learning_rate": 2.08208345126088e-05, "loss": 0.5382, "step": 22945 }, { "epoch": 2.37, "learning_rate": 2.081849117190259e-05, "loss": 0.3268, "step": 22946 }, { "epoch": 2.37, "learning_rate": 2.0816147893004157e-05, "loss": 0.4187, "step": 22947 }, { "epoch": 2.37, "learning_rate": 2.0813804675929258e-05, "loss": 0.256, "step": 22948 }, { "epoch": 2.37, "learning_rate": 2.0811461520693676e-05, "loss": 0.2753, "step": 22949 }, { "epoch": 2.37, "learning_rate": 2.080911842731319e-05, "loss": 0.473, "step": 22950 }, { "epoch": 2.37, "learning_rate": 2.0806775395803565e-05, "loss": 0.3946, "step": 22951 }, { "epoch": 2.37, "learning_rate": 2.0804432426180576e-05, "loss": 0.5784, "step": 22952 }, { "epoch": 2.37, "learning_rate": 2.080208951846e-05, "loss": 0.3266, "step": 22953 }, { "epoch": 2.37, "learning_rate": 2.07997466726576e-05, "loss": 0.2538, "step": 22954 }, { "epoch": 2.37, "learning_rate": 2.079740388878916e-05, "loss": 0.0763, "step": 22955 }, { "epoch": 2.37, "learning_rate": 2.0795061166870448e-05, "loss": 0.3174, "step": 22956 }, { "epoch": 2.37, "learning_rate": 2.0792718506917226e-05, "loss": 0.363, "step": 22957 }, { "epoch": 2.37, "learning_rate": 2.0790375908945264e-05, "loss": 0.1359, "step": 22958 }, { "epoch": 2.37, "learning_rate": 2.0788033372970337e-05, "loss": 0.9343, "step": 22959 }, { "epoch": 2.37, "learning_rate": 2.0785690899008213e-05, "loss": 0.081, "step": 22960 }, { "epoch": 2.37, "learning_rate": 2.0783348487074668e-05, "loss": 0.1529, "step": 22961 }, { "epoch": 2.37, "learning_rate": 2.0781006137185456e-05, "loss": 0.3538, "step": 22962 }, { "epoch": 2.37, "learning_rate": 2.0778663849356355e-05, "loss": 0.1527, "step": 22963 }, { "epoch": 2.37, "learning_rate": 2.0776321623603136e-05, "loss": 0.1911, "step": 22964 }, { "epoch": 2.37, "learning_rate": 2.0773979459941564e-05, "loss": 0.8229, "step": 22965 }, { "epoch": 2.37, "learning_rate": 2.07716373583874e-05, "loss": 0.3212, "step": 22966 }, { "epoch": 2.37, "learning_rate": 2.0769295318956407e-05, "loss": 0.4316, "step": 22967 }, { "epoch": 2.37, "learning_rate": 2.0766953341664378e-05, "loss": 0.2293, "step": 22968 }, { "epoch": 2.37, "learning_rate": 2.076461142652704e-05, "loss": 0.3261, "step": 22969 }, { "epoch": 2.37, "learning_rate": 2.0762269573560183e-05, "loss": 0.1154, "step": 22970 }, { "epoch": 2.37, "learning_rate": 2.0759927782779567e-05, "loss": 0.3903, "step": 22971 }, { "epoch": 2.37, "learning_rate": 2.0757586054200955e-05, "loss": 0.3132, "step": 22972 }, { "epoch": 2.37, "learning_rate": 2.0755244387840104e-05, "loss": 0.5501, "step": 22973 }, { "epoch": 2.37, "learning_rate": 2.075290278371279e-05, "loss": 1.9434, "step": 22974 }, { "epoch": 2.37, "learning_rate": 2.0750561241834775e-05, "loss": 0.0839, "step": 22975 }, { "epoch": 2.37, "learning_rate": 2.0748219762221813e-05, "loss": 0.8013, "step": 22976 }, { "epoch": 2.37, "learning_rate": 2.074587834488967e-05, "loss": 0.2563, "step": 22977 }, { "epoch": 2.37, "learning_rate": 2.0743536989854104e-05, "loss": 0.1603, "step": 22978 }, { "epoch": 2.37, "learning_rate": 2.0741195697130894e-05, "loss": 0.6948, "step": 22979 }, { "epoch": 2.37, "learning_rate": 2.0738854466735777e-05, "loss": 0.3677, "step": 22980 }, { "epoch": 2.37, "learning_rate": 2.0736513298684542e-05, "loss": 0.3451, "step": 22981 }, { "epoch": 2.37, "learning_rate": 2.0734172192992915e-05, "loss": 0.3798, "step": 22982 }, { "epoch": 2.37, "learning_rate": 2.0731831149676677e-05, "loss": 0.4743, "step": 22983 }, { "epoch": 2.37, "learning_rate": 2.072949016875158e-05, "loss": 0.0885, "step": 22984 }, { "epoch": 2.37, "learning_rate": 2.0727149250233395e-05, "loss": 0.4949, "step": 22985 }, { "epoch": 2.37, "learning_rate": 2.072480839413786e-05, "loss": 0.5071, "step": 22986 }, { "epoch": 2.37, "learning_rate": 2.0722467600480752e-05, "loss": 0.2979, "step": 22987 }, { "epoch": 2.37, "learning_rate": 2.0720126869277814e-05, "loss": 0.3632, "step": 22988 }, { "epoch": 2.37, "learning_rate": 2.0717786200544823e-05, "loss": 0.7044, "step": 22989 }, { "epoch": 2.37, "learning_rate": 2.0715445594297514e-05, "loss": 0.8787, "step": 22990 }, { "epoch": 2.37, "learning_rate": 2.0713105050551655e-05, "loss": 0.3912, "step": 22991 }, { "epoch": 2.37, "learning_rate": 2.0710764569322997e-05, "loss": 0.5648, "step": 22992 }, { "epoch": 2.37, "learning_rate": 2.0708424150627312e-05, "loss": 0.6223, "step": 22993 }, { "epoch": 2.37, "learning_rate": 2.0706083794480328e-05, "loss": 0.9652, "step": 22994 }, { "epoch": 2.37, "learning_rate": 2.070374350089782e-05, "loss": 0.8736, "step": 22995 }, { "epoch": 2.38, "learning_rate": 2.0701403269895532e-05, "loss": 0.1767, "step": 22996 }, { "epoch": 2.38, "learning_rate": 2.069906310148922e-05, "loss": 0.3326, "step": 22997 }, { "epoch": 2.38, "learning_rate": 2.069672299569464e-05, "loss": 0.6335, "step": 22998 }, { "epoch": 2.38, "learning_rate": 2.0694382952527547e-05, "loss": 0.5634, "step": 22999 }, { "epoch": 2.38, "learning_rate": 2.069204297200369e-05, "loss": 0.0383, "step": 23000 }, { "epoch": 2.38, "learning_rate": 2.0689703054138813e-05, "loss": 0.4628, "step": 23001 }, { "epoch": 2.38, "learning_rate": 2.0687363198948683e-05, "loss": 0.2087, "step": 23002 }, { "epoch": 2.38, "learning_rate": 2.068502340644905e-05, "loss": 0.7432, "step": 23003 }, { "epoch": 2.38, "learning_rate": 2.0682683676655656e-05, "loss": 0.9371, "step": 23004 }, { "epoch": 2.38, "learning_rate": 2.068034400958425e-05, "loss": 0.5651, "step": 23005 }, { "epoch": 2.38, "learning_rate": 2.0678004405250606e-05, "loss": 0.6218, "step": 23006 }, { "epoch": 2.38, "learning_rate": 2.0675664863670437e-05, "loss": 0.3646, "step": 23007 }, { "epoch": 2.38, "learning_rate": 2.0673325384859512e-05, "loss": 0.3537, "step": 23008 }, { "epoch": 2.38, "learning_rate": 2.0670985968833586e-05, "loss": 0.202, "step": 23009 }, { "epoch": 2.38, "learning_rate": 2.0668646615608394e-05, "loss": 0.4746, "step": 23010 }, { "epoch": 2.38, "learning_rate": 2.0666307325199687e-05, "loss": 0.4252, "step": 23011 }, { "epoch": 2.38, "learning_rate": 2.0663968097623217e-05, "loss": 0.1244, "step": 23012 }, { "epoch": 2.38, "learning_rate": 2.0661628932894738e-05, "loss": 0.0832, "step": 23013 }, { "epoch": 2.38, "learning_rate": 2.0659289831029978e-05, "loss": 0.1936, "step": 23014 }, { "epoch": 2.38, "learning_rate": 2.0656950792044692e-05, "loss": 0.591, "step": 23015 }, { "epoch": 2.38, "learning_rate": 2.0654611815954628e-05, "loss": 0.2861, "step": 23016 }, { "epoch": 2.38, "learning_rate": 2.065227290277554e-05, "loss": 0.3825, "step": 23017 }, { "epoch": 2.38, "learning_rate": 2.0649934052523165e-05, "loss": 0.6046, "step": 23018 }, { "epoch": 2.38, "learning_rate": 2.0647595265213244e-05, "loss": 0.422, "step": 23019 }, { "epoch": 2.38, "learning_rate": 2.064525654086151e-05, "loss": 0.5402, "step": 23020 }, { "epoch": 2.38, "learning_rate": 2.064291787948373e-05, "loss": 0.5588, "step": 23021 }, { "epoch": 2.38, "learning_rate": 2.064057928109564e-05, "loss": 0.069, "step": 23022 }, { "epoch": 2.38, "learning_rate": 2.0638240745712978e-05, "loss": 0.3616, "step": 23023 }, { "epoch": 2.38, "learning_rate": 2.0635902273351483e-05, "loss": 0.5836, "step": 23024 }, { "epoch": 2.38, "learning_rate": 2.0633563864026902e-05, "loss": 0.285, "step": 23025 }, { "epoch": 2.38, "learning_rate": 2.0631225517754983e-05, "loss": 0.5464, "step": 23026 }, { "epoch": 2.38, "learning_rate": 2.062888723455147e-05, "loss": 0.7584, "step": 23027 }, { "epoch": 2.38, "learning_rate": 2.062654901443208e-05, "loss": 0.3278, "step": 23028 }, { "epoch": 2.38, "learning_rate": 2.0624210857412576e-05, "loss": 0.3145, "step": 23029 }, { "epoch": 2.38, "learning_rate": 2.0621872763508685e-05, "loss": 0.3917, "step": 23030 }, { "epoch": 2.38, "learning_rate": 2.0619534732736172e-05, "loss": 0.1854, "step": 23031 }, { "epoch": 2.38, "learning_rate": 2.0617196765110735e-05, "loss": 0.4597, "step": 23032 }, { "epoch": 2.38, "learning_rate": 2.0614858860648148e-05, "loss": 0.2406, "step": 23033 }, { "epoch": 2.38, "learning_rate": 2.0612521019364126e-05, "loss": 0.6176, "step": 23034 }, { "epoch": 2.38, "learning_rate": 2.0610183241274416e-05, "loss": 0.8065, "step": 23035 }, { "epoch": 2.38, "learning_rate": 2.0607845526394757e-05, "loss": 0.4224, "step": 23036 }, { "epoch": 2.38, "learning_rate": 2.0605507874740886e-05, "loss": 0.4771, "step": 23037 }, { "epoch": 2.38, "learning_rate": 2.0603170286328535e-05, "loss": 0.1963, "step": 23038 }, { "epoch": 2.38, "learning_rate": 2.060083276117344e-05, "loss": 0.2599, "step": 23039 }, { "epoch": 2.38, "learning_rate": 2.0598495299291346e-05, "loss": 0.8258, "step": 23040 }, { "epoch": 2.38, "learning_rate": 2.059615790069798e-05, "loss": 0.1505, "step": 23041 }, { "epoch": 2.38, "learning_rate": 2.059382056540908e-05, "loss": 0.5566, "step": 23042 }, { "epoch": 2.38, "learning_rate": 2.0591483293440376e-05, "loss": 0.3616, "step": 23043 }, { "epoch": 2.38, "learning_rate": 2.0589146084807614e-05, "loss": 0.2837, "step": 23044 }, { "epoch": 2.38, "learning_rate": 2.0586808939526512e-05, "loss": 0.4038, "step": 23045 }, { "epoch": 2.38, "learning_rate": 2.058447185761281e-05, "loss": 0.0654, "step": 23046 }, { "epoch": 2.38, "learning_rate": 2.0582134839082242e-05, "loss": 0.1694, "step": 23047 }, { "epoch": 2.38, "learning_rate": 2.0579797883950534e-05, "loss": 0.1162, "step": 23048 }, { "epoch": 2.38, "learning_rate": 2.0577460992233424e-05, "loss": 0.3504, "step": 23049 }, { "epoch": 2.38, "learning_rate": 2.0575124163946637e-05, "loss": 0.8826, "step": 23050 }, { "epoch": 2.38, "learning_rate": 2.057278739910592e-05, "loss": 0.2906, "step": 23051 }, { "epoch": 2.38, "learning_rate": 2.0570450697726987e-05, "loss": 0.1719, "step": 23052 }, { "epoch": 2.38, "learning_rate": 2.056811405982557e-05, "loss": 0.1386, "step": 23053 }, { "epoch": 2.38, "learning_rate": 2.05657774854174e-05, "loss": 0.5117, "step": 23054 }, { "epoch": 2.38, "learning_rate": 2.0563440974518222e-05, "loss": 0.4492, "step": 23055 }, { "epoch": 2.38, "learning_rate": 2.0561104527143748e-05, "loss": 0.6606, "step": 23056 }, { "epoch": 2.38, "learning_rate": 2.0558768143309707e-05, "loss": 0.5231, "step": 23057 }, { "epoch": 2.38, "learning_rate": 2.0556431823031828e-05, "loss": 0.3502, "step": 23058 }, { "epoch": 2.38, "learning_rate": 2.0554095566325842e-05, "loss": 0.1509, "step": 23059 }, { "epoch": 2.38, "learning_rate": 2.055175937320747e-05, "loss": 0.4335, "step": 23060 }, { "epoch": 2.38, "learning_rate": 2.054942324369245e-05, "loss": 0.3947, "step": 23061 }, { "epoch": 2.38, "learning_rate": 2.0547087177796494e-05, "loss": 0.4676, "step": 23062 }, { "epoch": 2.38, "learning_rate": 2.0544751175535336e-05, "loss": 0.3714, "step": 23063 }, { "epoch": 2.38, "learning_rate": 2.05424152369247e-05, "loss": 0.1941, "step": 23064 }, { "epoch": 2.38, "learning_rate": 2.0540079361980326e-05, "loss": 0.6061, "step": 23065 }, { "epoch": 2.38, "learning_rate": 2.0537743550717908e-05, "loss": 0.9802, "step": 23066 }, { "epoch": 2.38, "learning_rate": 2.053540780315319e-05, "loss": 0.3917, "step": 23067 }, { "epoch": 2.38, "learning_rate": 2.0533072119301895e-05, "loss": 0.1306, "step": 23068 }, { "epoch": 2.38, "learning_rate": 2.053073649917975e-05, "loss": 0.7193, "step": 23069 }, { "epoch": 2.38, "learning_rate": 2.0528400942802465e-05, "loss": 0.5379, "step": 23070 }, { "epoch": 2.38, "learning_rate": 2.0526065450185768e-05, "loss": 0.1764, "step": 23071 }, { "epoch": 2.38, "learning_rate": 2.052373002134538e-05, "loss": 0.212, "step": 23072 }, { "epoch": 2.38, "learning_rate": 2.0521394656297025e-05, "loss": 0.5907, "step": 23073 }, { "epoch": 2.38, "learning_rate": 2.051905935505642e-05, "loss": 0.6851, "step": 23074 }, { "epoch": 2.38, "learning_rate": 2.05167241176393e-05, "loss": 0.1573, "step": 23075 }, { "epoch": 2.38, "learning_rate": 2.0514388944061363e-05, "loss": 0.3068, "step": 23076 }, { "epoch": 2.38, "learning_rate": 2.0512053834338344e-05, "loss": 0.2073, "step": 23077 }, { "epoch": 2.38, "learning_rate": 2.050971878848596e-05, "loss": 0.1967, "step": 23078 }, { "epoch": 2.38, "learning_rate": 2.050738380651993e-05, "loss": 0.2787, "step": 23079 }, { "epoch": 2.38, "learning_rate": 2.0505048888455964e-05, "loss": 0.2973, "step": 23080 }, { "epoch": 2.38, "learning_rate": 2.0502714034309806e-05, "loss": 1.0638, "step": 23081 }, { "epoch": 2.38, "learning_rate": 2.050037924409714e-05, "loss": 0.2729, "step": 23082 }, { "epoch": 2.38, "learning_rate": 2.0498044517833694e-05, "loss": 0.4994, "step": 23083 }, { "epoch": 2.38, "learning_rate": 2.0495709855535195e-05, "loss": 0.4497, "step": 23084 }, { "epoch": 2.38, "learning_rate": 2.0493375257217354e-05, "loss": 0.1484, "step": 23085 }, { "epoch": 2.38, "learning_rate": 2.0491040722895885e-05, "loss": 0.2842, "step": 23086 }, { "epoch": 2.38, "learning_rate": 2.04887062525865e-05, "loss": 0.2321, "step": 23087 }, { "epoch": 2.38, "learning_rate": 2.0486371846304924e-05, "loss": 0.1141, "step": 23088 }, { "epoch": 2.38, "learning_rate": 2.0484037504066868e-05, "loss": 0.6358, "step": 23089 }, { "epoch": 2.38, "learning_rate": 2.0481703225888043e-05, "loss": 0.7736, "step": 23090 }, { "epoch": 2.38, "learning_rate": 2.0479369011784167e-05, "loss": 0.4572, "step": 23091 }, { "epoch": 2.39, "learning_rate": 2.0477034861770948e-05, "loss": 0.1499, "step": 23092 }, { "epoch": 2.39, "learning_rate": 2.0474700775864106e-05, "loss": 0.0553, "step": 23093 }, { "epoch": 2.39, "learning_rate": 2.047236675407935e-05, "loss": 0.1741, "step": 23094 }, { "epoch": 2.39, "learning_rate": 2.0470032796432395e-05, "loss": 0.2851, "step": 23095 }, { "epoch": 2.39, "learning_rate": 2.046769890293894e-05, "loss": 0.263, "step": 23096 }, { "epoch": 2.39, "learning_rate": 2.046536507361471e-05, "loss": 0.7371, "step": 23097 }, { "epoch": 2.39, "learning_rate": 2.0463031308475405e-05, "loss": 0.6657, "step": 23098 }, { "epoch": 2.39, "learning_rate": 2.0460697607536754e-05, "loss": 0.6683, "step": 23099 }, { "epoch": 2.39, "learning_rate": 2.0458363970814445e-05, "loss": 0.3367, "step": 23100 }, { "epoch": 2.39, "learning_rate": 2.0456030398324197e-05, "loss": 0.1495, "step": 23101 }, { "epoch": 2.39, "learning_rate": 2.0453696890081722e-05, "loss": 1.0143, "step": 23102 }, { "epoch": 2.39, "learning_rate": 2.0451363446102727e-05, "loss": 0.8613, "step": 23103 }, { "epoch": 2.39, "learning_rate": 2.044903006640292e-05, "loss": 0.2376, "step": 23104 }, { "epoch": 2.39, "learning_rate": 2.0446696750998003e-05, "loss": 0.1614, "step": 23105 }, { "epoch": 2.39, "learning_rate": 2.04443634999037e-05, "loss": 0.2852, "step": 23106 }, { "epoch": 2.39, "learning_rate": 2.0442030313135694e-05, "loss": 0.3986, "step": 23107 }, { "epoch": 2.39, "learning_rate": 2.0439697190709704e-05, "loss": 0.2425, "step": 23108 }, { "epoch": 2.39, "learning_rate": 2.0437364132641446e-05, "loss": 0.4393, "step": 23109 }, { "epoch": 2.39, "learning_rate": 2.0435031138946605e-05, "loss": 0.1104, "step": 23110 }, { "epoch": 2.39, "learning_rate": 2.0432698209640898e-05, "loss": 0.356, "step": 23111 }, { "epoch": 2.39, "learning_rate": 2.0430365344740025e-05, "loss": 0.1582, "step": 23112 }, { "epoch": 2.39, "learning_rate": 2.04280325442597e-05, "loss": 0.5049, "step": 23113 }, { "epoch": 2.39, "learning_rate": 2.042569980821562e-05, "loss": 0.1059, "step": 23114 }, { "epoch": 2.39, "learning_rate": 2.0423367136623485e-05, "loss": 0.2257, "step": 23115 }, { "epoch": 2.39, "learning_rate": 2.0421034529499e-05, "loss": 0.0701, "step": 23116 }, { "epoch": 2.39, "learning_rate": 2.0418701986857878e-05, "loss": 0.1331, "step": 23117 }, { "epoch": 2.39, "learning_rate": 2.0416369508715803e-05, "loss": 0.3007, "step": 23118 }, { "epoch": 2.39, "learning_rate": 2.04140370950885e-05, "loss": 0.466, "step": 23119 }, { "epoch": 2.39, "learning_rate": 2.0411704745991645e-05, "loss": 0.1649, "step": 23120 }, { "epoch": 2.39, "learning_rate": 2.0409372461440946e-05, "loss": 0.457, "step": 23121 }, { "epoch": 2.39, "learning_rate": 2.0407040241452114e-05, "loss": 0.12, "step": 23122 }, { "epoch": 2.39, "learning_rate": 2.0404708086040848e-05, "loss": 0.4151, "step": 23123 }, { "epoch": 2.39, "learning_rate": 2.0402375995222833e-05, "loss": 0.3065, "step": 23124 }, { "epoch": 2.39, "learning_rate": 2.0400043969013775e-05, "loss": 0.419, "step": 23125 }, { "epoch": 2.39, "learning_rate": 2.039771200742938e-05, "loss": 0.3078, "step": 23126 }, { "epoch": 2.39, "learning_rate": 2.0395380110485344e-05, "loss": 0.486, "step": 23127 }, { "epoch": 2.39, "learning_rate": 2.0393048278197354e-05, "loss": 0.6412, "step": 23128 }, { "epoch": 2.39, "learning_rate": 2.039071651058112e-05, "loss": 0.1326, "step": 23129 }, { "epoch": 2.39, "learning_rate": 2.0388384807652328e-05, "loss": 0.8358, "step": 23130 }, { "epoch": 2.39, "learning_rate": 2.0386053169426697e-05, "loss": 0.2077, "step": 23131 }, { "epoch": 2.39, "learning_rate": 2.0383721595919893e-05, "loss": 0.4492, "step": 23132 }, { "epoch": 2.39, "learning_rate": 2.0381390087147636e-05, "loss": 0.314, "step": 23133 }, { "epoch": 2.39, "learning_rate": 2.0379058643125596e-05, "loss": 0.1095, "step": 23134 }, { "epoch": 2.39, "learning_rate": 2.0376727263869487e-05, "loss": 0.3804, "step": 23135 }, { "epoch": 2.39, "learning_rate": 2.0374395949395e-05, "loss": 0.8007, "step": 23136 }, { "epoch": 2.39, "learning_rate": 2.037206469971783e-05, "loss": 0.1917, "step": 23137 }, { "epoch": 2.39, "learning_rate": 2.0369733514853662e-05, "loss": 0.4594, "step": 23138 }, { "epoch": 2.39, "learning_rate": 2.0367402394818194e-05, "loss": 0.082, "step": 23139 }, { "epoch": 2.39, "learning_rate": 2.036507133962712e-05, "loss": 0.2626, "step": 23140 }, { "epoch": 2.39, "learning_rate": 2.0362740349296137e-05, "loss": 0.2257, "step": 23141 }, { "epoch": 2.39, "learning_rate": 2.0360409423840928e-05, "loss": 0.2086, "step": 23142 }, { "epoch": 2.39, "learning_rate": 2.035807856327719e-05, "loss": 0.2799, "step": 23143 }, { "epoch": 2.39, "learning_rate": 2.0355747767620614e-05, "loss": 0.1995, "step": 23144 }, { "epoch": 2.39, "learning_rate": 2.0353417036886883e-05, "loss": 0.7027, "step": 23145 }, { "epoch": 2.39, "learning_rate": 2.0351086371091684e-05, "loss": 0.9326, "step": 23146 }, { "epoch": 2.39, "learning_rate": 2.0348755770250726e-05, "loss": 0.1458, "step": 23147 }, { "epoch": 2.39, "learning_rate": 2.0346425234379675e-05, "loss": 0.3704, "step": 23148 }, { "epoch": 2.39, "learning_rate": 2.0344094763494236e-05, "loss": 0.3939, "step": 23149 }, { "epoch": 2.39, "learning_rate": 2.0341764357610087e-05, "loss": 0.8136, "step": 23150 }, { "epoch": 2.39, "learning_rate": 2.033943401674293e-05, "loss": 0.3751, "step": 23151 }, { "epoch": 2.39, "learning_rate": 2.0337103740908437e-05, "loss": 0.8956, "step": 23152 }, { "epoch": 2.39, "learning_rate": 2.0334773530122295e-05, "loss": 0.6325, "step": 23153 }, { "epoch": 2.39, "learning_rate": 2.03324433844002e-05, "loss": 0.2969, "step": 23154 }, { "epoch": 2.39, "learning_rate": 2.0330113303757842e-05, "loss": 0.2792, "step": 23155 }, { "epoch": 2.39, "learning_rate": 2.0327783288210894e-05, "loss": 0.3898, "step": 23156 }, { "epoch": 2.39, "learning_rate": 2.0325453337775054e-05, "loss": 0.2986, "step": 23157 }, { "epoch": 2.39, "learning_rate": 2.0323123452465988e-05, "loss": 0.1502, "step": 23158 }, { "epoch": 2.39, "learning_rate": 2.0320793632299386e-05, "loss": 0.103, "step": 23159 }, { "epoch": 2.39, "learning_rate": 2.031846387729094e-05, "loss": 0.4717, "step": 23160 }, { "epoch": 2.39, "learning_rate": 2.031613418745634e-05, "loss": 0.3768, "step": 23161 }, { "epoch": 2.39, "learning_rate": 2.0313804562811248e-05, "loss": 0.2867, "step": 23162 }, { "epoch": 2.39, "learning_rate": 2.031147500337136e-05, "loss": 0.2366, "step": 23163 }, { "epoch": 2.39, "learning_rate": 2.0309145509152353e-05, "loss": 0.6847, "step": 23164 }, { "epoch": 2.39, "learning_rate": 2.0306816080169917e-05, "loss": 0.4652, "step": 23165 }, { "epoch": 2.39, "learning_rate": 2.0304486716439722e-05, "loss": 0.6999, "step": 23166 }, { "epoch": 2.39, "learning_rate": 2.0302157417977457e-05, "loss": 0.3519, "step": 23167 }, { "epoch": 2.39, "learning_rate": 2.0299828184798792e-05, "loss": 0.2746, "step": 23168 }, { "epoch": 2.39, "learning_rate": 2.029749901691943e-05, "loss": 0.2058, "step": 23169 }, { "epoch": 2.39, "learning_rate": 2.0295169914355025e-05, "loss": 0.1131, "step": 23170 }, { "epoch": 2.39, "learning_rate": 2.0292840877121268e-05, "loss": 0.5218, "step": 23171 }, { "epoch": 2.39, "learning_rate": 2.029051190523383e-05, "loss": 0.1339, "step": 23172 }, { "epoch": 2.39, "learning_rate": 2.028818299870839e-05, "loss": 0.1842, "step": 23173 }, { "epoch": 2.39, "learning_rate": 2.0285854157560636e-05, "loss": 0.4811, "step": 23174 }, { "epoch": 2.39, "learning_rate": 2.0283525381806242e-05, "loss": 0.3582, "step": 23175 }, { "epoch": 2.39, "learning_rate": 2.0281196671460877e-05, "loss": 0.4988, "step": 23176 }, { "epoch": 2.39, "learning_rate": 2.0278868026540218e-05, "loss": 0.7276, "step": 23177 }, { "epoch": 2.39, "learning_rate": 2.027653944705995e-05, "loss": 0.3891, "step": 23178 }, { "epoch": 2.39, "learning_rate": 2.0274210933035747e-05, "loss": 0.8959, "step": 23179 }, { "epoch": 2.39, "learning_rate": 2.0271882484483275e-05, "loss": 0.2015, "step": 23180 }, { "epoch": 2.39, "learning_rate": 2.0269554101418213e-05, "loss": 0.0598, "step": 23181 }, { "epoch": 2.39, "learning_rate": 2.026722578385625e-05, "loss": 0.5322, "step": 23182 }, { "epoch": 2.39, "learning_rate": 2.0264897531813032e-05, "loss": 0.2972, "step": 23183 }, { "epoch": 2.39, "learning_rate": 2.0262569345304243e-05, "loss": 0.3694, "step": 23184 }, { "epoch": 2.39, "learning_rate": 2.026024122434557e-05, "loss": 0.3047, "step": 23185 }, { "epoch": 2.39, "learning_rate": 2.0257913168952665e-05, "loss": 0.1145, "step": 23186 }, { "epoch": 2.39, "learning_rate": 2.0255585179141212e-05, "loss": 0.3582, "step": 23187 }, { "epoch": 2.4, "learning_rate": 2.0253257254926875e-05, "loss": 0.1428, "step": 23188 }, { "epoch": 2.4, "learning_rate": 2.0250929396325343e-05, "loss": 0.1946, "step": 23189 }, { "epoch": 2.4, "learning_rate": 2.024860160335226e-05, "loss": 0.514, "step": 23190 }, { "epoch": 2.4, "learning_rate": 2.024627387602331e-05, "loss": 0.1949, "step": 23191 }, { "epoch": 2.4, "learning_rate": 2.0243946214354166e-05, "loss": 0.0674, "step": 23192 }, { "epoch": 2.4, "learning_rate": 2.024161861836049e-05, "loss": 0.7708, "step": 23193 }, { "epoch": 2.4, "learning_rate": 2.023929108805797e-05, "loss": 0.701, "step": 23194 }, { "epoch": 2.4, "learning_rate": 2.0236963623462248e-05, "loss": 0.3306, "step": 23195 }, { "epoch": 2.4, "learning_rate": 2.0234636224589e-05, "loss": 1.0248, "step": 23196 }, { "epoch": 2.4, "learning_rate": 2.023230889145389e-05, "loss": 0.1725, "step": 23197 }, { "epoch": 2.4, "learning_rate": 2.0229981624072597e-05, "loss": 0.3584, "step": 23198 }, { "epoch": 2.4, "learning_rate": 2.0227654422460786e-05, "loss": 0.3006, "step": 23199 }, { "epoch": 2.4, "learning_rate": 2.0225327286634115e-05, "loss": 0.2642, "step": 23200 }, { "epoch": 2.4, "learning_rate": 2.0223000216608252e-05, "loss": 0.125, "step": 23201 }, { "epoch": 2.4, "learning_rate": 2.0220673212398865e-05, "loss": 0.2821, "step": 23202 }, { "epoch": 2.4, "learning_rate": 2.021834627402162e-05, "loss": 0.2279, "step": 23203 }, { "epoch": 2.4, "learning_rate": 2.021601940149218e-05, "loss": 0.5238, "step": 23204 }, { "epoch": 2.4, "learning_rate": 2.0213692594826205e-05, "loss": 0.2473, "step": 23205 }, { "epoch": 2.4, "learning_rate": 2.0211365854039355e-05, "loss": 0.1183, "step": 23206 }, { "epoch": 2.4, "learning_rate": 2.020903917914732e-05, "loss": 0.4135, "step": 23207 }, { "epoch": 2.4, "learning_rate": 2.0206712570165728e-05, "loss": 0.0781, "step": 23208 }, { "epoch": 2.4, "learning_rate": 2.0204386027110256e-05, "loss": 0.2294, "step": 23209 }, { "epoch": 2.4, "learning_rate": 2.0202059549996567e-05, "loss": 0.5838, "step": 23210 }, { "epoch": 2.4, "learning_rate": 2.0199733138840317e-05, "loss": 0.5457, "step": 23211 }, { "epoch": 2.4, "learning_rate": 2.0197406793657173e-05, "loss": 0.6936, "step": 23212 }, { "epoch": 2.4, "learning_rate": 2.01950805144628e-05, "loss": 0.4459, "step": 23213 }, { "epoch": 2.4, "learning_rate": 2.0192754301272836e-05, "loss": 0.4763, "step": 23214 }, { "epoch": 2.4, "learning_rate": 2.0190428154102966e-05, "loss": 0.3965, "step": 23215 }, { "epoch": 2.4, "learning_rate": 2.0188102072968832e-05, "loss": 1.0112, "step": 23216 }, { "epoch": 2.4, "learning_rate": 2.0185776057886103e-05, "loss": 0.2994, "step": 23217 }, { "epoch": 2.4, "learning_rate": 2.0183450108870436e-05, "loss": 0.4357, "step": 23218 }, { "epoch": 2.4, "learning_rate": 2.0181124225937492e-05, "loss": 0.2989, "step": 23219 }, { "epoch": 2.4, "learning_rate": 2.0178798409102916e-05, "loss": 0.142, "step": 23220 }, { "epoch": 2.4, "learning_rate": 2.0176472658382365e-05, "loss": 0.4771, "step": 23221 }, { "epoch": 2.4, "learning_rate": 2.0174146973791502e-05, "loss": 0.3416, "step": 23222 }, { "epoch": 2.4, "learning_rate": 2.0171821355345988e-05, "loss": 0.1724, "step": 23223 }, { "epoch": 2.4, "learning_rate": 2.0169495803061468e-05, "loss": 0.258, "step": 23224 }, { "epoch": 2.4, "learning_rate": 2.0167170316953603e-05, "loss": 0.1314, "step": 23225 }, { "epoch": 2.4, "learning_rate": 2.0164844897038043e-05, "loss": 0.1731, "step": 23226 }, { "epoch": 2.4, "learning_rate": 2.0162519543330457e-05, "loss": 0.2513, "step": 23227 }, { "epoch": 2.4, "learning_rate": 2.016019425584648e-05, "loss": 0.2556, "step": 23228 }, { "epoch": 2.4, "learning_rate": 2.015786903460177e-05, "loss": 0.3353, "step": 23229 }, { "epoch": 2.4, "learning_rate": 2.015554387961199e-05, "loss": 0.3073, "step": 23230 }, { "epoch": 2.4, "learning_rate": 2.015321879089278e-05, "loss": 0.618, "step": 23231 }, { "epoch": 2.4, "learning_rate": 2.0150893768459807e-05, "loss": 0.5809, "step": 23232 }, { "epoch": 2.4, "learning_rate": 2.014856881232871e-05, "loss": 0.1706, "step": 23233 }, { "epoch": 2.4, "learning_rate": 2.014624392251514e-05, "loss": 0.0593, "step": 23234 }, { "epoch": 2.4, "learning_rate": 2.0143919099034747e-05, "loss": 0.3942, "step": 23235 }, { "epoch": 2.4, "learning_rate": 2.014159434190318e-05, "loss": 0.1132, "step": 23236 }, { "epoch": 2.4, "learning_rate": 2.0139269651136108e-05, "loss": 0.2543, "step": 23237 }, { "epoch": 2.4, "learning_rate": 2.013694502674915e-05, "loss": 0.3699, "step": 23238 }, { "epoch": 2.4, "learning_rate": 2.0134620468757977e-05, "loss": 0.3422, "step": 23239 }, { "epoch": 2.4, "learning_rate": 2.0132295977178228e-05, "loss": 0.313, "step": 23240 }, { "epoch": 2.4, "learning_rate": 2.012997155202556e-05, "loss": 0.5051, "step": 23241 }, { "epoch": 2.4, "learning_rate": 2.0127647193315608e-05, "loss": 0.4428, "step": 23242 }, { "epoch": 2.4, "learning_rate": 2.012532290106403e-05, "loss": 0.1612, "step": 23243 }, { "epoch": 2.4, "learning_rate": 2.012299867528646e-05, "loss": 0.5993, "step": 23244 }, { "epoch": 2.4, "learning_rate": 2.0120674515998567e-05, "loss": 0.2597, "step": 23245 }, { "epoch": 2.4, "learning_rate": 2.011835042321597e-05, "loss": 0.8479, "step": 23246 }, { "epoch": 2.4, "learning_rate": 2.0116026396954333e-05, "loss": 0.3878, "step": 23247 }, { "epoch": 2.4, "learning_rate": 2.0113702437229286e-05, "loss": 0.5995, "step": 23248 }, { "epoch": 2.4, "learning_rate": 2.011137854405648e-05, "loss": 0.0973, "step": 23249 }, { "epoch": 2.4, "learning_rate": 2.010905471745156e-05, "loss": 0.4767, "step": 23250 }, { "epoch": 2.4, "learning_rate": 2.010673095743018e-05, "loss": 0.0523, "step": 23251 }, { "epoch": 2.4, "learning_rate": 2.010440726400796e-05, "loss": 0.2236, "step": 23252 }, { "epoch": 2.4, "learning_rate": 2.010208363720056e-05, "loss": 0.1978, "step": 23253 }, { "epoch": 2.4, "learning_rate": 2.0099760077023614e-05, "loss": 0.5056, "step": 23254 }, { "epoch": 2.4, "learning_rate": 2.009743658349277e-05, "loss": 0.4183, "step": 23255 }, { "epoch": 2.4, "learning_rate": 2.0095113156623665e-05, "loss": 0.6091, "step": 23256 }, { "epoch": 2.4, "learning_rate": 2.0092789796431953e-05, "loss": 0.7241, "step": 23257 }, { "epoch": 2.4, "learning_rate": 2.009046650293325e-05, "loss": 0.229, "step": 23258 }, { "epoch": 2.4, "learning_rate": 2.0088143276143203e-05, "loss": 0.1194, "step": 23259 }, { "epoch": 2.4, "learning_rate": 2.008582011607746e-05, "loss": 0.2896, "step": 23260 }, { "epoch": 2.4, "learning_rate": 2.008349702275166e-05, "loss": 0.1571, "step": 23261 }, { "epoch": 2.4, "learning_rate": 2.0081173996181436e-05, "loss": 0.936, "step": 23262 }, { "epoch": 2.4, "learning_rate": 2.0078851036382428e-05, "loss": 0.2972, "step": 23263 }, { "epoch": 2.4, "learning_rate": 2.0076528143370264e-05, "loss": 0.1, "step": 23264 }, { "epoch": 2.4, "learning_rate": 2.0074205317160606e-05, "loss": 0.3153, "step": 23265 }, { "epoch": 2.4, "learning_rate": 2.007188255776907e-05, "loss": 0.5322, "step": 23266 }, { "epoch": 2.4, "learning_rate": 2.0069559865211294e-05, "loss": 0.2735, "step": 23267 }, { "epoch": 2.4, "learning_rate": 2.0067237239502914e-05, "loss": 0.2136, "step": 23268 }, { "epoch": 2.4, "learning_rate": 2.0064914680659573e-05, "loss": 0.2091, "step": 23269 }, { "epoch": 2.4, "learning_rate": 2.0062592188696916e-05, "loss": 0.2193, "step": 23270 }, { "epoch": 2.4, "learning_rate": 2.006026976363055e-05, "loss": 0.1207, "step": 23271 }, { "epoch": 2.4, "learning_rate": 2.005794740547612e-05, "loss": 0.5263, "step": 23272 }, { "epoch": 2.4, "learning_rate": 2.0055625114249264e-05, "loss": 0.6249, "step": 23273 }, { "epoch": 2.4, "learning_rate": 2.0053302889965614e-05, "loss": 0.2783, "step": 23274 }, { "epoch": 2.4, "learning_rate": 2.0050980732640802e-05, "loss": 0.3467, "step": 23275 }, { "epoch": 2.4, "learning_rate": 2.004865864229046e-05, "loss": 0.4356, "step": 23276 }, { "epoch": 2.4, "learning_rate": 2.0046336618930214e-05, "loss": 0.4897, "step": 23277 }, { "epoch": 2.4, "learning_rate": 2.0044014662575705e-05, "loss": 0.1217, "step": 23278 }, { "epoch": 2.4, "learning_rate": 2.0041692773242557e-05, "loss": 0.4837, "step": 23279 }, { "epoch": 2.4, "learning_rate": 2.0039370950946412e-05, "loss": 0.1461, "step": 23280 }, { "epoch": 2.4, "learning_rate": 2.003704919570288e-05, "loss": 0.4249, "step": 23281 }, { "epoch": 2.4, "learning_rate": 2.003472750752762e-05, "loss": 0.1111, "step": 23282 }, { "epoch": 2.41, "learning_rate": 2.0032405886436227e-05, "loss": 0.2614, "step": 23283 }, { "epoch": 2.41, "learning_rate": 2.0030084332444348e-05, "loss": 0.2229, "step": 23284 }, { "epoch": 2.41, "learning_rate": 2.002776284556761e-05, "loss": 0.338, "step": 23285 }, { "epoch": 2.41, "learning_rate": 2.0025441425821635e-05, "loss": 0.0637, "step": 23286 }, { "epoch": 2.41, "learning_rate": 2.0023120073222054e-05, "loss": 0.5445, "step": 23287 }, { "epoch": 2.41, "learning_rate": 2.0020798787784498e-05, "loss": 0.4163, "step": 23288 }, { "epoch": 2.41, "learning_rate": 2.0018477569524593e-05, "loss": 0.3884, "step": 23289 }, { "epoch": 2.41, "learning_rate": 2.0016156418457955e-05, "loss": 0.2546, "step": 23290 }, { "epoch": 2.41, "learning_rate": 2.001383533460022e-05, "loss": 0.1716, "step": 23291 }, { "epoch": 2.41, "learning_rate": 2.0011514317967005e-05, "loss": 0.2515, "step": 23292 }, { "epoch": 2.41, "learning_rate": 2.0009193368573935e-05, "loss": 0.3185, "step": 23293 }, { "epoch": 2.41, "learning_rate": 2.0006872486436646e-05, "loss": 0.2579, "step": 23294 }, { "epoch": 2.41, "learning_rate": 2.0004551671570764e-05, "loss": 0.4374, "step": 23295 }, { "epoch": 2.41, "learning_rate": 2.0002230923991886e-05, "loss": 0.497, "step": 23296 }, { "epoch": 2.41, "learning_rate": 1.999991024371565e-05, "loss": 0.4918, "step": 23297 }, { "epoch": 2.41, "learning_rate": 1.999758963075768e-05, "loss": 0.3882, "step": 23298 }, { "epoch": 2.41, "learning_rate": 1.9995269085133598e-05, "loss": 1.0038, "step": 23299 }, { "epoch": 2.41, "learning_rate": 1.9992948606859023e-05, "loss": 0.2784, "step": 23300 }, { "epoch": 2.41, "learning_rate": 1.9990628195949578e-05, "loss": 0.35, "step": 23301 }, { "epoch": 2.41, "learning_rate": 1.9988307852420872e-05, "loss": 0.2733, "step": 23302 }, { "epoch": 2.41, "learning_rate": 1.9985987576288545e-05, "loss": 0.2874, "step": 23303 }, { "epoch": 2.41, "learning_rate": 1.9983667367568205e-05, "loss": 0.2312, "step": 23304 }, { "epoch": 2.41, "learning_rate": 1.998134722627547e-05, "loss": 0.4143, "step": 23305 }, { "epoch": 2.41, "learning_rate": 1.9979027152425968e-05, "loss": 0.2932, "step": 23306 }, { "epoch": 2.41, "learning_rate": 1.9976707146035314e-05, "loss": 0.5705, "step": 23307 }, { "epoch": 2.41, "learning_rate": 1.997438720711911e-05, "loss": 0.4773, "step": 23308 }, { "epoch": 2.41, "learning_rate": 1.9972067335692995e-05, "loss": 0.0788, "step": 23309 }, { "epoch": 2.41, "learning_rate": 1.9969747531772568e-05, "loss": 0.4903, "step": 23310 }, { "epoch": 2.41, "learning_rate": 1.9967427795373457e-05, "loss": 0.4395, "step": 23311 }, { "epoch": 2.41, "learning_rate": 1.996510812651127e-05, "loss": 0.0744, "step": 23312 }, { "epoch": 2.41, "learning_rate": 1.9962788525201636e-05, "loss": 0.1385, "step": 23313 }, { "epoch": 2.41, "learning_rate": 1.9960468991460154e-05, "loss": 0.4181, "step": 23314 }, { "epoch": 2.41, "learning_rate": 1.9958149525302446e-05, "loss": 0.6472, "step": 23315 }, { "epoch": 2.41, "learning_rate": 1.995583012674413e-05, "loss": 0.1239, "step": 23316 }, { "epoch": 2.41, "learning_rate": 1.9953510795800812e-05, "loss": 0.2345, "step": 23317 }, { "epoch": 2.41, "learning_rate": 1.995119153248811e-05, "loss": 0.2131, "step": 23318 }, { "epoch": 2.41, "learning_rate": 1.9948872336821637e-05, "loss": 0.5161, "step": 23319 }, { "epoch": 2.41, "learning_rate": 1.994655320881701e-05, "loss": 0.485, "step": 23320 }, { "epoch": 2.41, "learning_rate": 1.9944234148489827e-05, "loss": 0.1328, "step": 23321 }, { "epoch": 2.41, "learning_rate": 1.9941915155855706e-05, "loss": 0.3383, "step": 23322 }, { "epoch": 2.41, "learning_rate": 1.9939596230930262e-05, "loss": 0.1556, "step": 23323 }, { "epoch": 2.41, "learning_rate": 1.9937277373729097e-05, "loss": 0.3338, "step": 23324 }, { "epoch": 2.41, "learning_rate": 1.9934958584267826e-05, "loss": 0.7878, "step": 23325 }, { "epoch": 2.41, "learning_rate": 1.9932639862562063e-05, "loss": 0.1404, "step": 23326 }, { "epoch": 2.41, "learning_rate": 1.9930321208627412e-05, "loss": 0.3157, "step": 23327 }, { "epoch": 2.41, "learning_rate": 1.9928002622479483e-05, "loss": 0.2567, "step": 23328 }, { "epoch": 2.41, "learning_rate": 1.992568410413388e-05, "loss": 0.1665, "step": 23329 }, { "epoch": 2.41, "learning_rate": 1.9923365653606215e-05, "loss": 0.3497, "step": 23330 }, { "epoch": 2.41, "learning_rate": 1.9921047270912097e-05, "loss": 0.5151, "step": 23331 }, { "epoch": 2.41, "learning_rate": 1.991872895606714e-05, "loss": 0.0746, "step": 23332 }, { "epoch": 2.41, "learning_rate": 1.9916410709086932e-05, "loss": 0.7143, "step": 23333 }, { "epoch": 2.41, "learning_rate": 1.9914092529987083e-05, "loss": 0.5812, "step": 23334 }, { "epoch": 2.41, "learning_rate": 1.991177441878321e-05, "loss": 0.4074, "step": 23335 }, { "epoch": 2.41, "learning_rate": 1.9909456375490904e-05, "loss": 0.3828, "step": 23336 }, { "epoch": 2.41, "learning_rate": 1.9907138400125786e-05, "loss": 0.3371, "step": 23337 }, { "epoch": 2.41, "learning_rate": 1.9904820492703447e-05, "loss": 0.7424, "step": 23338 }, { "epoch": 2.41, "learning_rate": 1.990250265323949e-05, "loss": 0.2866, "step": 23339 }, { "epoch": 2.41, "learning_rate": 1.9900184881749524e-05, "loss": 0.1462, "step": 23340 }, { "epoch": 2.41, "learning_rate": 1.9897867178249147e-05, "loss": 0.6847, "step": 23341 }, { "epoch": 2.41, "learning_rate": 1.9895549542753973e-05, "loss": 0.4762, "step": 23342 }, { "epoch": 2.41, "learning_rate": 1.989323197527959e-05, "loss": 0.0259, "step": 23343 }, { "epoch": 2.41, "learning_rate": 1.9890914475841603e-05, "loss": 0.1713, "step": 23344 }, { "epoch": 2.41, "learning_rate": 1.988859704445563e-05, "loss": 0.1304, "step": 23345 }, { "epoch": 2.41, "learning_rate": 1.9886279681137237e-05, "loss": 0.1743, "step": 23346 }, { "epoch": 2.41, "learning_rate": 1.9883962385902055e-05, "loss": 0.4432, "step": 23347 }, { "epoch": 2.41, "learning_rate": 1.9881645158765663e-05, "loss": 0.8456, "step": 23348 }, { "epoch": 2.41, "learning_rate": 1.987932799974367e-05, "loss": 0.4299, "step": 23349 }, { "epoch": 2.41, "learning_rate": 1.9877010908851672e-05, "loss": 0.3392, "step": 23350 }, { "epoch": 2.41, "learning_rate": 1.987469388610527e-05, "loss": 0.1844, "step": 23351 }, { "epoch": 2.41, "learning_rate": 1.987237693152006e-05, "loss": 0.3323, "step": 23352 }, { "epoch": 2.41, "learning_rate": 1.987006004511163e-05, "loss": 0.0862, "step": 23353 }, { "epoch": 2.41, "learning_rate": 1.9867743226895597e-05, "loss": 0.3785, "step": 23354 }, { "epoch": 2.41, "learning_rate": 1.986542647688754e-05, "loss": 0.3956, "step": 23355 }, { "epoch": 2.41, "learning_rate": 1.9863109795103066e-05, "loss": 0.3169, "step": 23356 }, { "epoch": 2.41, "learning_rate": 1.986079318155776e-05, "loss": 0.4383, "step": 23357 }, { "epoch": 2.41, "learning_rate": 1.9858476636267234e-05, "loss": 0.559, "step": 23358 }, { "epoch": 2.41, "learning_rate": 1.9856160159247058e-05, "loss": 0.3603, "step": 23359 }, { "epoch": 2.41, "learning_rate": 1.9853843750512835e-05, "loss": 0.2866, "step": 23360 }, { "epoch": 2.41, "learning_rate": 1.9851527410080175e-05, "loss": 0.0324, "step": 23361 }, { "epoch": 2.41, "learning_rate": 1.9849211137964648e-05, "loss": 0.6359, "step": 23362 }, { "epoch": 2.41, "learning_rate": 1.9846894934181858e-05, "loss": 0.1375, "step": 23363 }, { "epoch": 2.41, "learning_rate": 1.9844578798747393e-05, "loss": 0.1503, "step": 23364 }, { "epoch": 2.41, "learning_rate": 1.984226273167685e-05, "loss": 0.4224, "step": 23365 }, { "epoch": 2.41, "learning_rate": 1.983994673298582e-05, "loss": 0.2874, "step": 23366 }, { "epoch": 2.41, "learning_rate": 1.983763080268989e-05, "loss": 0.3633, "step": 23367 }, { "epoch": 2.41, "learning_rate": 1.983531494080465e-05, "loss": 0.2224, "step": 23368 }, { "epoch": 2.41, "learning_rate": 1.9832999147345686e-05, "loss": 0.1894, "step": 23369 }, { "epoch": 2.41, "learning_rate": 1.9830683422328615e-05, "loss": 0.3864, "step": 23370 }, { "epoch": 2.41, "learning_rate": 1.9828367765768994e-05, "loss": 0.5795, "step": 23371 }, { "epoch": 2.41, "learning_rate": 1.982605217768241e-05, "loss": 0.8849, "step": 23372 }, { "epoch": 2.41, "learning_rate": 1.9823736658084467e-05, "loss": 0.6957, "step": 23373 }, { "epoch": 2.41, "learning_rate": 1.9821421206990746e-05, "loss": 0.1817, "step": 23374 }, { "epoch": 2.41, "learning_rate": 1.981910582441684e-05, "loss": 0.374, "step": 23375 }, { "epoch": 2.41, "learning_rate": 1.981679051037833e-05, "loss": 0.0452, "step": 23376 }, { "epoch": 2.41, "learning_rate": 1.98144752648908e-05, "loss": 0.1656, "step": 23377 }, { "epoch": 2.41, "learning_rate": 1.9812160087969844e-05, "loss": 0.549, "step": 23378 }, { "epoch": 2.42, "learning_rate": 1.9809844979631034e-05, "loss": 0.3745, "step": 23379 }, { "epoch": 2.42, "learning_rate": 1.980752993988998e-05, "loss": 0.5375, "step": 23380 }, { "epoch": 2.42, "learning_rate": 1.9805214968762234e-05, "loss": 0.5251, "step": 23381 }, { "epoch": 2.42, "learning_rate": 1.98029000662634e-05, "loss": 0.2835, "step": 23382 }, { "epoch": 2.42, "learning_rate": 1.9800585232409073e-05, "loss": 0.9005, "step": 23383 }, { "epoch": 2.42, "learning_rate": 1.9798270467214803e-05, "loss": 0.1563, "step": 23384 }, { "epoch": 2.42, "learning_rate": 1.9795955770696192e-05, "loss": 0.5876, "step": 23385 }, { "epoch": 2.42, "learning_rate": 1.9793641142868822e-05, "loss": 0.1177, "step": 23386 }, { "epoch": 2.42, "learning_rate": 1.9791326583748262e-05, "loss": 0.513, "step": 23387 }, { "epoch": 2.42, "learning_rate": 1.9789012093350108e-05, "loss": 0.8576, "step": 23388 }, { "epoch": 2.42, "learning_rate": 1.9786697671689934e-05, "loss": 0.2759, "step": 23389 }, { "epoch": 2.42, "learning_rate": 1.9784383318783326e-05, "loss": 0.5558, "step": 23390 }, { "epoch": 2.42, "learning_rate": 1.978206903464586e-05, "loss": 0.2895, "step": 23391 }, { "epoch": 2.42, "learning_rate": 1.977975481929311e-05, "loss": 0.1249, "step": 23392 }, { "epoch": 2.42, "learning_rate": 1.977744067274066e-05, "loss": 0.492, "step": 23393 }, { "epoch": 2.42, "learning_rate": 1.9775126595004092e-05, "loss": 0.3105, "step": 23394 }, { "epoch": 2.42, "learning_rate": 1.9772812586098984e-05, "loss": 0.207, "step": 23395 }, { "epoch": 2.42, "learning_rate": 1.97704986460409e-05, "loss": 0.0285, "step": 23396 }, { "epoch": 2.42, "learning_rate": 1.9768184774845423e-05, "loss": 0.3994, "step": 23397 }, { "epoch": 2.42, "learning_rate": 1.9765870972528134e-05, "loss": 0.1454, "step": 23398 }, { "epoch": 2.42, "learning_rate": 1.976355723910461e-05, "loss": 0.6305, "step": 23399 }, { "epoch": 2.42, "learning_rate": 1.9761243574590418e-05, "loss": 0.1262, "step": 23400 }, { "epoch": 2.42, "learning_rate": 1.9758929979001142e-05, "loss": 0.6108, "step": 23401 }, { "epoch": 2.42, "learning_rate": 1.9756616452352355e-05, "loss": 0.6713, "step": 23402 }, { "epoch": 2.42, "learning_rate": 1.9754302994659626e-05, "loss": 0.1896, "step": 23403 }, { "epoch": 2.42, "learning_rate": 1.9751989605938534e-05, "loss": 0.1861, "step": 23404 }, { "epoch": 2.42, "learning_rate": 1.9749676286204646e-05, "loss": 0.2216, "step": 23405 }, { "epoch": 2.42, "learning_rate": 1.9747363035473542e-05, "loss": 0.4842, "step": 23406 }, { "epoch": 2.42, "learning_rate": 1.974504985376079e-05, "loss": 0.25, "step": 23407 }, { "epoch": 2.42, "learning_rate": 1.9742736741081975e-05, "loss": 0.3279, "step": 23408 }, { "epoch": 2.42, "learning_rate": 1.9740423697452653e-05, "loss": 0.3016, "step": 23409 }, { "epoch": 2.42, "learning_rate": 1.973811072288839e-05, "loss": 0.2934, "step": 23410 }, { "epoch": 2.42, "learning_rate": 1.9735797817404758e-05, "loss": 0.6422, "step": 23411 }, { "epoch": 2.42, "learning_rate": 1.9733484981017344e-05, "loss": 0.6365, "step": 23412 }, { "epoch": 2.42, "learning_rate": 1.9731172213741705e-05, "loss": 0.4586, "step": 23413 }, { "epoch": 2.42, "learning_rate": 1.9728859515593412e-05, "loss": 0.4541, "step": 23414 }, { "epoch": 2.42, "learning_rate": 1.9726546886588028e-05, "loss": 0.7088, "step": 23415 }, { "epoch": 2.42, "learning_rate": 1.972423432674113e-05, "loss": 0.3713, "step": 23416 }, { "epoch": 2.42, "learning_rate": 1.972192183606828e-05, "loss": 0.184, "step": 23417 }, { "epoch": 2.42, "learning_rate": 1.9719609414585055e-05, "loss": 0.1307, "step": 23418 }, { "epoch": 2.42, "learning_rate": 1.971729706230701e-05, "loss": 0.3553, "step": 23419 }, { "epoch": 2.42, "learning_rate": 1.9714984779249725e-05, "loss": 0.4218, "step": 23420 }, { "epoch": 2.42, "learning_rate": 1.9712672565428745e-05, "loss": 0.842, "step": 23421 }, { "epoch": 2.42, "learning_rate": 1.9710360420859644e-05, "loss": 0.3566, "step": 23422 }, { "epoch": 2.42, "learning_rate": 1.9708048345557993e-05, "loss": 0.1654, "step": 23423 }, { "epoch": 2.42, "learning_rate": 1.9705736339539352e-05, "loss": 0.3217, "step": 23424 }, { "epoch": 2.42, "learning_rate": 1.9703424402819282e-05, "loss": 0.2916, "step": 23425 }, { "epoch": 2.42, "learning_rate": 1.970111253541335e-05, "loss": 0.0745, "step": 23426 }, { "epoch": 2.42, "learning_rate": 1.969880073733712e-05, "loss": 0.334, "step": 23427 }, { "epoch": 2.42, "learning_rate": 1.9696489008606156e-05, "loss": 0.8836, "step": 23428 }, { "epoch": 2.42, "learning_rate": 1.9694177349236013e-05, "loss": 0.3786, "step": 23429 }, { "epoch": 2.42, "learning_rate": 1.969186575924226e-05, "loss": 0.4679, "step": 23430 }, { "epoch": 2.42, "learning_rate": 1.968955423864045e-05, "loss": 0.6761, "step": 23431 }, { "epoch": 2.42, "learning_rate": 1.9687242787446156e-05, "loss": 0.146, "step": 23432 }, { "epoch": 2.42, "learning_rate": 1.9684931405674938e-05, "loss": 0.0346, "step": 23433 }, { "epoch": 2.42, "learning_rate": 1.9682620093342334e-05, "loss": 0.2201, "step": 23434 }, { "epoch": 2.42, "learning_rate": 1.9680308850463922e-05, "loss": 0.145, "step": 23435 }, { "epoch": 2.42, "learning_rate": 1.967799767705525e-05, "loss": 0.5001, "step": 23436 }, { "epoch": 2.42, "learning_rate": 1.967568657313189e-05, "loss": 0.3289, "step": 23437 }, { "epoch": 2.42, "learning_rate": 1.967337553870939e-05, "loss": 0.1525, "step": 23438 }, { "epoch": 2.42, "learning_rate": 1.9671064573803307e-05, "loss": 0.4473, "step": 23439 }, { "epoch": 2.42, "learning_rate": 1.9668753678429198e-05, "loss": 0.3534, "step": 23440 }, { "epoch": 2.42, "learning_rate": 1.966644285260263e-05, "loss": 0.1186, "step": 23441 }, { "epoch": 2.42, "learning_rate": 1.9664132096339147e-05, "loss": 0.1685, "step": 23442 }, { "epoch": 2.42, "learning_rate": 1.9661821409654308e-05, "loss": 0.5072, "step": 23443 }, { "epoch": 2.42, "learning_rate": 1.9659510792563668e-05, "loss": 0.1489, "step": 23444 }, { "epoch": 2.42, "learning_rate": 1.965720024508278e-05, "loss": 0.001, "step": 23445 }, { "epoch": 2.42, "learning_rate": 1.965488976722721e-05, "loss": 0.8868, "step": 23446 }, { "epoch": 2.42, "learning_rate": 1.9652579359012504e-05, "loss": 0.3499, "step": 23447 }, { "epoch": 2.42, "learning_rate": 1.96502690204542e-05, "loss": 0.4428, "step": 23448 }, { "epoch": 2.42, "learning_rate": 1.9647958751567866e-05, "loss": 0.1934, "step": 23449 }, { "epoch": 2.42, "learning_rate": 1.964564855236905e-05, "loss": 0.3549, "step": 23450 }, { "epoch": 2.42, "learning_rate": 1.96433384228733e-05, "loss": 0.7356, "step": 23451 }, { "epoch": 2.42, "learning_rate": 1.9641028363096188e-05, "loss": 0.2289, "step": 23452 }, { "epoch": 2.42, "learning_rate": 1.963871837305324e-05, "loss": 0.381, "step": 23453 }, { "epoch": 2.42, "learning_rate": 1.963640845276001e-05, "loss": 0.2064, "step": 23454 }, { "epoch": 2.42, "learning_rate": 1.963409860223206e-05, "loss": 0.3676, "step": 23455 }, { "epoch": 2.42, "learning_rate": 1.9631788821484934e-05, "loss": 0.0866, "step": 23456 }, { "epoch": 2.42, "learning_rate": 1.9629479110534176e-05, "loss": 0.1913, "step": 23457 }, { "epoch": 2.42, "learning_rate": 1.9627169469395355e-05, "loss": 0.2209, "step": 23458 }, { "epoch": 2.42, "learning_rate": 1.9624859898083982e-05, "loss": 0.2348, "step": 23459 }, { "epoch": 2.42, "learning_rate": 1.962255039661563e-05, "loss": 0.2963, "step": 23460 }, { "epoch": 2.42, "learning_rate": 1.9620240965005848e-05, "loss": 0.3248, "step": 23461 }, { "epoch": 2.42, "learning_rate": 1.9617931603270162e-05, "loss": 0.2085, "step": 23462 }, { "epoch": 2.42, "learning_rate": 1.9615622311424137e-05, "loss": 0.3982, "step": 23463 }, { "epoch": 2.42, "learning_rate": 1.961331308948331e-05, "loss": 0.666, "step": 23464 }, { "epoch": 2.42, "learning_rate": 1.961100393746323e-05, "loss": 0.8253, "step": 23465 }, { "epoch": 2.42, "learning_rate": 1.960869485537945e-05, "loss": 0.2726, "step": 23466 }, { "epoch": 2.42, "learning_rate": 1.960638584324749e-05, "loss": 0.5357, "step": 23467 }, { "epoch": 2.42, "learning_rate": 1.9604076901082914e-05, "loss": 0.468, "step": 23468 }, { "epoch": 2.42, "learning_rate": 1.9601768028901264e-05, "loss": 0.4479, "step": 23469 }, { "epoch": 2.42, "learning_rate": 1.959945922671808e-05, "loss": 0.3467, "step": 23470 }, { "epoch": 2.42, "learning_rate": 1.959715049454891e-05, "loss": 0.4991, "step": 23471 }, { "epoch": 2.42, "learning_rate": 1.9594841832409275e-05, "loss": 0.2241, "step": 23472 }, { "epoch": 2.42, "learning_rate": 1.9592533240314732e-05, "loss": 0.4418, "step": 23473 }, { "epoch": 2.42, "learning_rate": 1.959022471828082e-05, "loss": 0.6725, "step": 23474 }, { "epoch": 2.43, "learning_rate": 1.9587916266323083e-05, "loss": 0.1872, "step": 23475 }, { "epoch": 2.43, "learning_rate": 1.9585607884457052e-05, "loss": 0.5656, "step": 23476 }, { "epoch": 2.43, "learning_rate": 1.9583299572698278e-05, "loss": 0.7323, "step": 23477 }, { "epoch": 2.43, "learning_rate": 1.958099133106229e-05, "loss": 0.3352, "step": 23478 }, { "epoch": 2.43, "learning_rate": 1.957868315956463e-05, "loss": 0.5097, "step": 23479 }, { "epoch": 2.43, "learning_rate": 1.9576375058220836e-05, "loss": 0.2835, "step": 23480 }, { "epoch": 2.43, "learning_rate": 1.957406702704645e-05, "loss": 0.8196, "step": 23481 }, { "epoch": 2.43, "learning_rate": 1.9571759066057e-05, "loss": 0.3762, "step": 23482 }, { "epoch": 2.43, "learning_rate": 1.9569451175268043e-05, "loss": 0.2225, "step": 23483 }, { "epoch": 2.43, "learning_rate": 1.956714335469509e-05, "loss": 0.4346, "step": 23484 }, { "epoch": 2.43, "learning_rate": 1.9564835604353686e-05, "loss": 0.4015, "step": 23485 }, { "epoch": 2.43, "learning_rate": 1.9562527924259365e-05, "loss": 0.3656, "step": 23486 }, { "epoch": 2.43, "learning_rate": 1.9560220314427667e-05, "loss": 0.2995, "step": 23487 }, { "epoch": 2.43, "learning_rate": 1.955791277487412e-05, "loss": 0.3327, "step": 23488 }, { "epoch": 2.43, "learning_rate": 1.9555605305614258e-05, "loss": 0.1994, "step": 23489 }, { "epoch": 2.43, "learning_rate": 1.955329790666363e-05, "loss": 0.2601, "step": 23490 }, { "epoch": 2.43, "learning_rate": 1.9550990578037742e-05, "loss": 0.1603, "step": 23491 }, { "epoch": 2.43, "learning_rate": 1.9548683319752147e-05, "loss": 0.2876, "step": 23492 }, { "epoch": 2.43, "learning_rate": 1.9546376131822366e-05, "loss": 0.2987, "step": 23493 }, { "epoch": 2.43, "learning_rate": 1.9544069014263943e-05, "loss": 0.6009, "step": 23494 }, { "epoch": 2.43, "learning_rate": 1.95417619670924e-05, "loss": 0.7688, "step": 23495 }, { "epoch": 2.43, "learning_rate": 1.9539454990323274e-05, "loss": 0.2671, "step": 23496 }, { "epoch": 2.43, "learning_rate": 1.953714808397208e-05, "loss": 0.0764, "step": 23497 }, { "epoch": 2.43, "learning_rate": 1.9534841248054354e-05, "loss": 0.2729, "step": 23498 }, { "epoch": 2.43, "learning_rate": 1.9532534482585637e-05, "loss": 0.548, "step": 23499 }, { "epoch": 2.43, "learning_rate": 1.953022778758144e-05, "loss": 0.1364, "step": 23500 }, { "epoch": 2.43, "learning_rate": 1.9527921163057306e-05, "loss": 0.1918, "step": 23501 }, { "epoch": 2.43, "learning_rate": 1.952561460902875e-05, "loss": 0.556, "step": 23502 }, { "epoch": 2.43, "learning_rate": 1.9523308125511308e-05, "loss": 0.4351, "step": 23503 }, { "epoch": 2.43, "learning_rate": 1.9521001712520508e-05, "loss": 0.735, "step": 23504 }, { "epoch": 2.43, "learning_rate": 1.951869537007187e-05, "loss": 0.351, "step": 23505 }, { "epoch": 2.43, "learning_rate": 1.9516389098180922e-05, "loss": 0.2832, "step": 23506 }, { "epoch": 2.43, "learning_rate": 1.9514082896863188e-05, "loss": 0.3625, "step": 23507 }, { "epoch": 2.43, "learning_rate": 1.951177676613421e-05, "loss": 0.3647, "step": 23508 }, { "epoch": 2.43, "learning_rate": 1.9509470706009488e-05, "loss": 0.2696, "step": 23509 }, { "epoch": 2.43, "learning_rate": 1.9507164716504547e-05, "loss": 0.5653, "step": 23510 }, { "epoch": 2.43, "learning_rate": 1.9504858797634916e-05, "loss": 0.0477, "step": 23511 }, { "epoch": 2.43, "learning_rate": 1.9502552949416128e-05, "loss": 0.2043, "step": 23512 }, { "epoch": 2.43, "learning_rate": 1.950024717186369e-05, "loss": 0.1463, "step": 23513 }, { "epoch": 2.43, "learning_rate": 1.949794146499314e-05, "loss": 0.0647, "step": 23514 }, { "epoch": 2.43, "learning_rate": 1.9495635828819983e-05, "loss": 0.2188, "step": 23515 }, { "epoch": 2.43, "learning_rate": 1.9493330263359752e-05, "loss": 0.2764, "step": 23516 }, { "epoch": 2.43, "learning_rate": 1.9491024768627962e-05, "loss": 0.3241, "step": 23517 }, { "epoch": 2.43, "learning_rate": 1.9488719344640136e-05, "loss": 0.9066, "step": 23518 }, { "epoch": 2.43, "learning_rate": 1.9486413991411793e-05, "loss": 0.5265, "step": 23519 }, { "epoch": 2.43, "learning_rate": 1.9484108708958443e-05, "loss": 0.1436, "step": 23520 }, { "epoch": 2.43, "learning_rate": 1.948180349729563e-05, "loss": 0.267, "step": 23521 }, { "epoch": 2.43, "learning_rate": 1.9479498356438844e-05, "loss": 0.171, "step": 23522 }, { "epoch": 2.43, "learning_rate": 1.9477193286403616e-05, "loss": 0.1875, "step": 23523 }, { "epoch": 2.43, "learning_rate": 1.9474888287205453e-05, "loss": 0.1381, "step": 23524 }, { "epoch": 2.43, "learning_rate": 1.947258335885988e-05, "loss": 0.2205, "step": 23525 }, { "epoch": 2.43, "learning_rate": 1.9470278501382415e-05, "loss": 0.2096, "step": 23526 }, { "epoch": 2.43, "learning_rate": 1.9467973714788573e-05, "loss": 0.2284, "step": 23527 }, { "epoch": 2.43, "learning_rate": 1.9465668999093872e-05, "loss": 0.3425, "step": 23528 }, { "epoch": 2.43, "learning_rate": 1.9463364354313816e-05, "loss": 0.5685, "step": 23529 }, { "epoch": 2.43, "learning_rate": 1.9461059780463927e-05, "loss": 0.2764, "step": 23530 }, { "epoch": 2.43, "learning_rate": 1.9458755277559718e-05, "loss": 0.2307, "step": 23531 }, { "epoch": 2.43, "learning_rate": 1.9456450845616704e-05, "loss": 0.3359, "step": 23532 }, { "epoch": 2.43, "learning_rate": 1.9454146484650406e-05, "loss": 0.3309, "step": 23533 }, { "epoch": 2.43, "learning_rate": 1.945184219467631e-05, "loss": 0.2075, "step": 23534 }, { "epoch": 2.43, "learning_rate": 1.944953797570995e-05, "loss": 0.5985, "step": 23535 }, { "epoch": 2.43, "learning_rate": 1.9447233827766826e-05, "loss": 0.0878, "step": 23536 }, { "epoch": 2.43, "learning_rate": 1.9444929750862456e-05, "loss": 0.4075, "step": 23537 }, { "epoch": 2.43, "learning_rate": 1.944262574501236e-05, "loss": 0.3961, "step": 23538 }, { "epoch": 2.43, "learning_rate": 1.9440321810232027e-05, "loss": 0.4628, "step": 23539 }, { "epoch": 2.43, "learning_rate": 1.9438017946536976e-05, "loss": 0.079, "step": 23540 }, { "epoch": 2.43, "learning_rate": 1.943571415394272e-05, "loss": 0.2188, "step": 23541 }, { "epoch": 2.43, "learning_rate": 1.943341043246477e-05, "loss": 0.4592, "step": 23542 }, { "epoch": 2.43, "learning_rate": 1.9431106782118623e-05, "loss": 0.7286, "step": 23543 }, { "epoch": 2.43, "learning_rate": 1.942880320291979e-05, "loss": 0.6229, "step": 23544 }, { "epoch": 2.43, "learning_rate": 1.9426499694883785e-05, "loss": 0.2656, "step": 23545 }, { "epoch": 2.43, "learning_rate": 1.9424196258026116e-05, "loss": 0.2184, "step": 23546 }, { "epoch": 2.43, "learning_rate": 1.942189289236228e-05, "loss": 0.8425, "step": 23547 }, { "epoch": 2.43, "learning_rate": 1.9419589597907777e-05, "loss": 0.5907, "step": 23548 }, { "epoch": 2.43, "learning_rate": 1.9417286374678126e-05, "loss": 0.4611, "step": 23549 }, { "epoch": 2.43, "learning_rate": 1.9414983222688824e-05, "loss": 0.4681, "step": 23550 }, { "epoch": 2.43, "learning_rate": 1.941268014195538e-05, "loss": 0.6154, "step": 23551 }, { "epoch": 2.43, "learning_rate": 1.9410377132493304e-05, "loss": 0.3385, "step": 23552 }, { "epoch": 2.43, "learning_rate": 1.9408074194318076e-05, "loss": 0.1299, "step": 23553 }, { "epoch": 2.43, "learning_rate": 1.9405771327445225e-05, "loss": 0.5103, "step": 23554 }, { "epoch": 2.43, "learning_rate": 1.9403468531890237e-05, "loss": 0.5725, "step": 23555 }, { "epoch": 2.43, "learning_rate": 1.9401165807668627e-05, "loss": 0.6499, "step": 23556 }, { "epoch": 2.43, "learning_rate": 1.939886315479588e-05, "loss": 0.0663, "step": 23557 }, { "epoch": 2.43, "learning_rate": 1.9396560573287512e-05, "loss": 0.4097, "step": 23558 }, { "epoch": 2.43, "learning_rate": 1.9394258063159025e-05, "loss": 0.0376, "step": 23559 }, { "epoch": 2.43, "learning_rate": 1.9391955624425903e-05, "loss": 0.0844, "step": 23560 }, { "epoch": 2.43, "learning_rate": 1.9389653257103654e-05, "loss": 0.5963, "step": 23561 }, { "epoch": 2.43, "learning_rate": 1.9387350961207774e-05, "loss": 0.3816, "step": 23562 }, { "epoch": 2.43, "learning_rate": 1.9385048736753763e-05, "loss": 0.4287, "step": 23563 }, { "epoch": 2.43, "learning_rate": 1.9382746583757122e-05, "loss": 0.6207, "step": 23564 }, { "epoch": 2.43, "learning_rate": 1.9380444502233338e-05, "loss": 0.6599, "step": 23565 }, { "epoch": 2.43, "learning_rate": 1.937814249219793e-05, "loss": 0.2793, "step": 23566 }, { "epoch": 2.43, "learning_rate": 1.9375840553666377e-05, "loss": 0.3458, "step": 23567 }, { "epoch": 2.43, "learning_rate": 1.937353868665418e-05, "loss": 0.6111, "step": 23568 }, { "epoch": 2.43, "learning_rate": 1.9371236891176832e-05, "loss": 0.1905, "step": 23569 }, { "epoch": 2.43, "learning_rate": 1.9368935167249836e-05, "loss": 0.3932, "step": 23570 }, { "epoch": 2.44, "learning_rate": 1.9366633514888688e-05, "loss": 0.5112, "step": 23571 }, { "epoch": 2.44, "learning_rate": 1.936433193410886e-05, "loss": 0.2636, "step": 23572 }, { "epoch": 2.44, "learning_rate": 1.9362030424925866e-05, "loss": 0.6724, "step": 23573 }, { "epoch": 2.44, "learning_rate": 1.935972898735519e-05, "loss": 0.5242, "step": 23574 }, { "epoch": 2.44, "learning_rate": 1.9357427621412334e-05, "loss": 0.4122, "step": 23575 }, { "epoch": 2.44, "learning_rate": 1.9355126327112786e-05, "loss": 0.7555, "step": 23576 }, { "epoch": 2.44, "learning_rate": 1.9352825104472035e-05, "loss": 0.2798, "step": 23577 }, { "epoch": 2.44, "learning_rate": 1.935052395350557e-05, "loss": 0.6096, "step": 23578 }, { "epoch": 2.44, "learning_rate": 1.9348222874228887e-05, "loss": 0.2608, "step": 23579 }, { "epoch": 2.44, "learning_rate": 1.934592186665748e-05, "loss": 0.8419, "step": 23580 }, { "epoch": 2.44, "learning_rate": 1.934362093080683e-05, "loss": 0.388, "step": 23581 }, { "epoch": 2.44, "learning_rate": 1.9341320066692432e-05, "loss": 0.8286, "step": 23582 }, { "epoch": 2.44, "learning_rate": 1.933901927432977e-05, "loss": 0.1528, "step": 23583 }, { "epoch": 2.44, "learning_rate": 1.933671855373435e-05, "loss": 0.1466, "step": 23584 }, { "epoch": 2.44, "learning_rate": 1.933441790492164e-05, "loss": 0.3896, "step": 23585 }, { "epoch": 2.44, "learning_rate": 1.9332117327907126e-05, "loss": 0.4844, "step": 23586 }, { "epoch": 2.44, "learning_rate": 1.9329816822706303e-05, "loss": 0.5267, "step": 23587 }, { "epoch": 2.44, "learning_rate": 1.9327516389334657e-05, "loss": 0.6547, "step": 23588 }, { "epoch": 2.44, "learning_rate": 1.932521602780767e-05, "loss": 0.3742, "step": 23589 }, { "epoch": 2.44, "learning_rate": 1.932291573814084e-05, "loss": 0.2664, "step": 23590 }, { "epoch": 2.44, "learning_rate": 1.9320615520349636e-05, "loss": 0.2678, "step": 23591 }, { "epoch": 2.44, "learning_rate": 1.9318315374449555e-05, "loss": 0.3042, "step": 23592 }, { "epoch": 2.44, "learning_rate": 1.9316015300456068e-05, "loss": 0.3015, "step": 23593 }, { "epoch": 2.44, "learning_rate": 1.9313715298384675e-05, "loss": 0.4911, "step": 23594 }, { "epoch": 2.44, "learning_rate": 1.931141536825084e-05, "loss": 0.075, "step": 23595 }, { "epoch": 2.44, "learning_rate": 1.9309115510070073e-05, "loss": 0.49, "step": 23596 }, { "epoch": 2.44, "learning_rate": 1.9306815723857827e-05, "loss": 0.1873, "step": 23597 }, { "epoch": 2.44, "learning_rate": 1.9304516009629594e-05, "loss": 0.0432, "step": 23598 }, { "epoch": 2.44, "learning_rate": 1.9302216367400856e-05, "loss": 0.5727, "step": 23599 }, { "epoch": 2.44, "learning_rate": 1.92999167971871e-05, "loss": 0.1698, "step": 23600 }, { "epoch": 2.44, "learning_rate": 1.92976172990038e-05, "loss": 0.7222, "step": 23601 }, { "epoch": 2.44, "learning_rate": 1.929531787286643e-05, "loss": 0.5653, "step": 23602 }, { "epoch": 2.44, "learning_rate": 1.9293018518790474e-05, "loss": 0.0979, "step": 23603 }, { "epoch": 2.44, "learning_rate": 1.9290719236791425e-05, "loss": 0.1868, "step": 23604 }, { "epoch": 2.44, "learning_rate": 1.9288420026884742e-05, "loss": 0.3969, "step": 23605 }, { "epoch": 2.44, "learning_rate": 1.9286120889085903e-05, "loss": 0.3484, "step": 23606 }, { "epoch": 2.44, "learning_rate": 1.928382182341039e-05, "loss": 0.5867, "step": 23607 }, { "epoch": 2.44, "learning_rate": 1.9281522829873698e-05, "loss": 0.3853, "step": 23608 }, { "epoch": 2.44, "learning_rate": 1.9279223908491282e-05, "loss": 0.2349, "step": 23609 }, { "epoch": 2.44, "learning_rate": 1.9276925059278615e-05, "loss": 0.0662, "step": 23610 }, { "epoch": 2.44, "learning_rate": 1.927462628225118e-05, "loss": 0.7426, "step": 23611 }, { "epoch": 2.44, "learning_rate": 1.927232757742445e-05, "loss": 0.4543, "step": 23612 }, { "epoch": 2.44, "learning_rate": 1.9270028944813895e-05, "loss": 0.5652, "step": 23613 }, { "epoch": 2.44, "learning_rate": 1.926773038443501e-05, "loss": 0.4423, "step": 23614 }, { "epoch": 2.44, "learning_rate": 1.9265431896303243e-05, "loss": 0.2306, "step": 23615 }, { "epoch": 2.44, "learning_rate": 1.9263133480434078e-05, "loss": 0.2584, "step": 23616 }, { "epoch": 2.44, "learning_rate": 1.926083513684298e-05, "loss": 0.5208, "step": 23617 }, { "epoch": 2.44, "learning_rate": 1.925853686554544e-05, "loss": 0.7006, "step": 23618 }, { "epoch": 2.44, "learning_rate": 1.925623866655691e-05, "loss": 0.4824, "step": 23619 }, { "epoch": 2.44, "learning_rate": 1.9253940539892863e-05, "loss": 0.3212, "step": 23620 }, { "epoch": 2.44, "learning_rate": 1.9251642485568788e-05, "loss": 1.0887, "step": 23621 }, { "epoch": 2.44, "learning_rate": 1.924934450360013e-05, "loss": 0.1733, "step": 23622 }, { "epoch": 2.44, "learning_rate": 1.924704659400237e-05, "loss": 0.1531, "step": 23623 }, { "epoch": 2.44, "learning_rate": 1.9244748756790982e-05, "loss": 0.5521, "step": 23624 }, { "epoch": 2.44, "learning_rate": 1.924245099198142e-05, "loss": 0.3791, "step": 23625 }, { "epoch": 2.44, "learning_rate": 1.9240153299589164e-05, "loss": 0.4696, "step": 23626 }, { "epoch": 2.44, "learning_rate": 1.923785567962968e-05, "loss": 0.5312, "step": 23627 }, { "epoch": 2.44, "learning_rate": 1.9235558132118435e-05, "loss": 0.6468, "step": 23628 }, { "epoch": 2.44, "learning_rate": 1.923326065707089e-05, "loss": 0.6571, "step": 23629 }, { "epoch": 2.44, "learning_rate": 1.9230963254502516e-05, "loss": 0.3597, "step": 23630 }, { "epoch": 2.44, "learning_rate": 1.9228665924428776e-05, "loss": 0.2862, "step": 23631 }, { "epoch": 2.44, "learning_rate": 1.9226368666865145e-05, "loss": 0.7003, "step": 23632 }, { "epoch": 2.44, "learning_rate": 1.9224071481827067e-05, "loss": 0.5017, "step": 23633 }, { "epoch": 2.44, "learning_rate": 1.9221774369330033e-05, "loss": 0.2415, "step": 23634 }, { "epoch": 2.44, "learning_rate": 1.921947732938948e-05, "loss": 0.0577, "step": 23635 }, { "epoch": 2.44, "learning_rate": 1.921718036202089e-05, "loss": 0.5487, "step": 23636 }, { "epoch": 2.44, "learning_rate": 1.9214883467239716e-05, "loss": 0.12, "step": 23637 }, { "epoch": 2.44, "learning_rate": 1.9212586645061425e-05, "loss": 0.1714, "step": 23638 }, { "epoch": 2.44, "learning_rate": 1.9210289895501475e-05, "loss": 0.1195, "step": 23639 }, { "epoch": 2.44, "learning_rate": 1.920799321857532e-05, "loss": 0.2144, "step": 23640 }, { "epoch": 2.44, "learning_rate": 1.9205696614298438e-05, "loss": 0.3842, "step": 23641 }, { "epoch": 2.44, "learning_rate": 1.9203400082686285e-05, "loss": 0.5153, "step": 23642 }, { "epoch": 2.44, "learning_rate": 1.920110362375431e-05, "loss": 0.0929, "step": 23643 }, { "epoch": 2.44, "learning_rate": 1.919880723751798e-05, "loss": 0.5805, "step": 23644 }, { "epoch": 2.44, "learning_rate": 1.919651092399275e-05, "loss": 1.471, "step": 23645 }, { "epoch": 2.44, "learning_rate": 1.9194214683194087e-05, "loss": 0.1643, "step": 23646 }, { "epoch": 2.44, "learning_rate": 1.919191851513745e-05, "loss": 0.3545, "step": 23647 }, { "epoch": 2.44, "learning_rate": 1.9189622419838274e-05, "loss": 0.3871, "step": 23648 }, { "epoch": 2.44, "learning_rate": 1.918732639731204e-05, "loss": 0.6468, "step": 23649 }, { "epoch": 2.44, "learning_rate": 1.918503044757418e-05, "loss": 0.4051, "step": 23650 }, { "epoch": 2.44, "learning_rate": 1.9182734570640176e-05, "loss": 0.2148, "step": 23651 }, { "epoch": 2.44, "learning_rate": 1.9180438766525472e-05, "loss": 0.0905, "step": 23652 }, { "epoch": 2.44, "learning_rate": 1.9178143035245518e-05, "loss": 0.2147, "step": 23653 }, { "epoch": 2.44, "learning_rate": 1.917584737681577e-05, "loss": 0.4563, "step": 23654 }, { "epoch": 2.44, "learning_rate": 1.9173551791251692e-05, "loss": 0.6109, "step": 23655 }, { "epoch": 2.44, "learning_rate": 1.917125627856873e-05, "loss": 0.1284, "step": 23656 }, { "epoch": 2.44, "learning_rate": 1.916896083878233e-05, "loss": 0.5063, "step": 23657 }, { "epoch": 2.44, "learning_rate": 1.9166665471907955e-05, "loss": 0.3913, "step": 23658 }, { "epoch": 2.44, "learning_rate": 1.9164370177961067e-05, "loss": 0.3219, "step": 23659 }, { "epoch": 2.44, "learning_rate": 1.9162074956957088e-05, "loss": 0.5517, "step": 23660 }, { "epoch": 2.44, "learning_rate": 1.9159779808911486e-05, "loss": 0.0504, "step": 23661 }, { "epoch": 2.44, "learning_rate": 1.9157484733839717e-05, "loss": 0.122, "step": 23662 }, { "epoch": 2.44, "learning_rate": 1.9155189731757214e-05, "loss": 0.2402, "step": 23663 }, { "epoch": 2.44, "learning_rate": 1.915289480267944e-05, "loss": 0.6779, "step": 23664 }, { "epoch": 2.44, "learning_rate": 1.915059994662184e-05, "loss": 0.5355, "step": 23665 }, { "epoch": 2.45, "learning_rate": 1.9148305163599865e-05, "loss": 0.1006, "step": 23666 }, { "epoch": 2.45, "learning_rate": 1.914601045362896e-05, "loss": 0.4147, "step": 23667 }, { "epoch": 2.45, "learning_rate": 1.914371581672457e-05, "loss": 0.1154, "step": 23668 }, { "epoch": 2.45, "learning_rate": 1.9141421252902144e-05, "loss": 0.4983, "step": 23669 }, { "epoch": 2.45, "learning_rate": 1.9139126762177135e-05, "loss": 0.1472, "step": 23670 }, { "epoch": 2.45, "learning_rate": 1.913683234456498e-05, "loss": 0.1961, "step": 23671 }, { "epoch": 2.45, "learning_rate": 1.9134538000081137e-05, "loss": 0.257, "step": 23672 }, { "epoch": 2.45, "learning_rate": 1.9132243728741036e-05, "loss": 0.1212, "step": 23673 }, { "epoch": 2.45, "learning_rate": 1.912994953056012e-05, "loss": 0.2637, "step": 23674 }, { "epoch": 2.45, "learning_rate": 1.9127655405553842e-05, "loss": 0.4711, "step": 23675 }, { "epoch": 2.45, "learning_rate": 1.912536135373765e-05, "loss": 0.1, "step": 23676 }, { "epoch": 2.45, "learning_rate": 1.912306737512698e-05, "loss": 0.1583, "step": 23677 }, { "epoch": 2.45, "learning_rate": 1.912077346973727e-05, "loss": 0.4881, "step": 23678 }, { "epoch": 2.45, "learning_rate": 1.911847963758397e-05, "loss": 0.6575, "step": 23679 }, { "epoch": 2.45, "learning_rate": 1.9116185878682524e-05, "loss": 0.1342, "step": 23680 }, { "epoch": 2.45, "learning_rate": 1.911389219304836e-05, "loss": 0.477, "step": 23681 }, { "epoch": 2.45, "learning_rate": 1.9111598580696928e-05, "loss": 0.4928, "step": 23682 }, { "epoch": 2.45, "learning_rate": 1.9109305041643667e-05, "loss": 0.5107, "step": 23683 }, { "epoch": 2.45, "learning_rate": 1.910701157590403e-05, "loss": 0.4855, "step": 23684 }, { "epoch": 2.45, "learning_rate": 1.910471818349343e-05, "loss": 0.1088, "step": 23685 }, { "epoch": 2.45, "learning_rate": 1.9102424864427322e-05, "loss": 0.3892, "step": 23686 }, { "epoch": 2.45, "learning_rate": 1.9100131618721137e-05, "loss": 0.6666, "step": 23687 }, { "epoch": 2.45, "learning_rate": 1.9097838446390313e-05, "loss": 0.1495, "step": 23688 }, { "epoch": 2.45, "learning_rate": 1.9095545347450286e-05, "loss": 0.6226, "step": 23689 }, { "epoch": 2.45, "learning_rate": 1.909325232191651e-05, "loss": 0.3965, "step": 23690 }, { "epoch": 2.45, "learning_rate": 1.9090959369804395e-05, "loss": 0.5653, "step": 23691 }, { "epoch": 2.45, "learning_rate": 1.9088666491129394e-05, "loss": 0.9392, "step": 23692 }, { "epoch": 2.45, "learning_rate": 1.9086373685906935e-05, "loss": 0.2855, "step": 23693 }, { "epoch": 2.45, "learning_rate": 1.908408095415246e-05, "loss": 0.2334, "step": 23694 }, { "epoch": 2.45, "learning_rate": 1.9081788295881393e-05, "loss": 0.3272, "step": 23695 }, { "epoch": 2.45, "learning_rate": 1.9079495711109173e-05, "loss": 0.6142, "step": 23696 }, { "epoch": 2.45, "learning_rate": 1.9077203199851243e-05, "loss": 0.0947, "step": 23697 }, { "epoch": 2.45, "learning_rate": 1.9074910762123014e-05, "loss": 0.1293, "step": 23698 }, { "epoch": 2.45, "learning_rate": 1.9072618397939928e-05, "loss": 0.0313, "step": 23699 }, { "epoch": 2.45, "learning_rate": 1.9070326107317428e-05, "loss": 0.5853, "step": 23700 }, { "epoch": 2.45, "learning_rate": 1.9068033890270924e-05, "loss": 0.342, "step": 23701 }, { "epoch": 2.45, "learning_rate": 1.9065741746815863e-05, "loss": 0.2411, "step": 23702 }, { "epoch": 2.45, "learning_rate": 1.906344967696767e-05, "loss": 0.4781, "step": 23703 }, { "epoch": 2.45, "learning_rate": 1.9061157680741778e-05, "loss": 0.6172, "step": 23704 }, { "epoch": 2.45, "learning_rate": 1.9058865758153606e-05, "loss": 0.1493, "step": 23705 }, { "epoch": 2.45, "learning_rate": 1.905657390921859e-05, "loss": 0.5235, "step": 23706 }, { "epoch": 2.45, "learning_rate": 1.905428213395216e-05, "loss": 0.3015, "step": 23707 }, { "epoch": 2.45, "learning_rate": 1.9051990432369746e-05, "loss": 0.4647, "step": 23708 }, { "epoch": 2.45, "learning_rate": 1.9049698804486773e-05, "loss": 0.6004, "step": 23709 }, { "epoch": 2.45, "learning_rate": 1.9047407250318665e-05, "loss": 0.1724, "step": 23710 }, { "epoch": 2.45, "learning_rate": 1.9045115769880842e-05, "loss": 0.3162, "step": 23711 }, { "epoch": 2.45, "learning_rate": 1.9042824363188736e-05, "loss": 0.1087, "step": 23712 }, { "epoch": 2.45, "learning_rate": 1.9040533030257774e-05, "loss": 0.4181, "step": 23713 }, { "epoch": 2.45, "learning_rate": 1.9038241771103385e-05, "loss": 0.2063, "step": 23714 }, { "epoch": 2.45, "learning_rate": 1.9035950585740978e-05, "loss": 0.566, "step": 23715 }, { "epoch": 2.45, "learning_rate": 1.903365947418599e-05, "loss": 0.138, "step": 23716 }, { "epoch": 2.45, "learning_rate": 1.9031368436453843e-05, "loss": 0.1973, "step": 23717 }, { "epoch": 2.45, "learning_rate": 1.9029077472559957e-05, "loss": 0.602, "step": 23718 }, { "epoch": 2.45, "learning_rate": 1.902678658251975e-05, "loss": 0.3917, "step": 23719 }, { "epoch": 2.45, "learning_rate": 1.902449576634865e-05, "loss": 0.1169, "step": 23720 }, { "epoch": 2.45, "learning_rate": 1.9022205024062078e-05, "loss": 0.4653, "step": 23721 }, { "epoch": 2.45, "learning_rate": 1.9019914355675466e-05, "loss": 0.7486, "step": 23722 }, { "epoch": 2.45, "learning_rate": 1.9017623761204202e-05, "loss": 0.4561, "step": 23723 }, { "epoch": 2.45, "learning_rate": 1.9015333240663735e-05, "loss": 0.3688, "step": 23724 }, { "epoch": 2.45, "learning_rate": 1.9013042794069472e-05, "loss": 0.6022, "step": 23725 }, { "epoch": 2.45, "learning_rate": 1.901075242143683e-05, "loss": 0.3725, "step": 23726 }, { "epoch": 2.45, "learning_rate": 1.9008462122781233e-05, "loss": 0.9061, "step": 23727 }, { "epoch": 2.45, "learning_rate": 1.9006171898118106e-05, "loss": 0.371, "step": 23728 }, { "epoch": 2.45, "learning_rate": 1.9003881747462843e-05, "loss": 0.5805, "step": 23729 }, { "epoch": 2.45, "learning_rate": 1.9001591670830883e-05, "loss": 0.2267, "step": 23730 }, { "epoch": 2.45, "learning_rate": 1.8999301668237633e-05, "loss": 0.2148, "step": 23731 }, { "epoch": 2.45, "learning_rate": 1.8997011739698512e-05, "loss": 0.0609, "step": 23732 }, { "epoch": 2.45, "learning_rate": 1.8994721885228933e-05, "loss": 0.5411, "step": 23733 }, { "epoch": 2.45, "learning_rate": 1.899243210484432e-05, "loss": 0.1749, "step": 23734 }, { "epoch": 2.45, "learning_rate": 1.8990142398560072e-05, "loss": 0.5557, "step": 23735 }, { "epoch": 2.45, "learning_rate": 1.8987852766391603e-05, "loss": 0.2246, "step": 23736 }, { "epoch": 2.45, "learning_rate": 1.8985563208354335e-05, "loss": 0.372, "step": 23737 }, { "epoch": 2.45, "learning_rate": 1.8983273724463683e-05, "loss": 0.3663, "step": 23738 }, { "epoch": 2.45, "learning_rate": 1.8980984314735047e-05, "loss": 0.5871, "step": 23739 }, { "epoch": 2.45, "learning_rate": 1.897869497918385e-05, "loss": 0.2127, "step": 23740 }, { "epoch": 2.45, "learning_rate": 1.89764057178255e-05, "loss": 0.2256, "step": 23741 }, { "epoch": 2.45, "learning_rate": 1.8974116530675407e-05, "loss": 0.1674, "step": 23742 }, { "epoch": 2.45, "learning_rate": 1.8971827417748983e-05, "loss": 0.3622, "step": 23743 }, { "epoch": 2.45, "learning_rate": 1.8969538379061636e-05, "loss": 0.8939, "step": 23744 }, { "epoch": 2.45, "learning_rate": 1.896724941462877e-05, "loss": 0.2679, "step": 23745 }, { "epoch": 2.45, "learning_rate": 1.896496052446581e-05, "loss": 0.3704, "step": 23746 }, { "epoch": 2.45, "learning_rate": 1.896267170858815e-05, "loss": 0.4051, "step": 23747 }, { "epoch": 2.45, "learning_rate": 1.8960382967011207e-05, "loss": 0.529, "step": 23748 }, { "epoch": 2.45, "learning_rate": 1.895809429975037e-05, "loss": 0.3778, "step": 23749 }, { "epoch": 2.45, "learning_rate": 1.8955805706821065e-05, "loss": 0.2799, "step": 23750 }, { "epoch": 2.45, "learning_rate": 1.8953517188238687e-05, "loss": 0.565, "step": 23751 }, { "epoch": 2.45, "learning_rate": 1.895122874401866e-05, "loss": 0.3899, "step": 23752 }, { "epoch": 2.45, "learning_rate": 1.894894037417636e-05, "loss": 0.2271, "step": 23753 }, { "epoch": 2.45, "learning_rate": 1.8946652078727214e-05, "loss": 0.2305, "step": 23754 }, { "epoch": 2.45, "learning_rate": 1.894436385768662e-05, "loss": 0.0095, "step": 23755 }, { "epoch": 2.45, "learning_rate": 1.8942075711069984e-05, "loss": 0.3154, "step": 23756 }, { "epoch": 2.45, "learning_rate": 1.8939787638892702e-05, "loss": 0.6443, "step": 23757 }, { "epoch": 2.45, "learning_rate": 1.8937499641170185e-05, "loss": 0.2804, "step": 23758 }, { "epoch": 2.45, "learning_rate": 1.893521171791783e-05, "loss": 0.6424, "step": 23759 }, { "epoch": 2.45, "learning_rate": 1.8932923869151054e-05, "loss": 0.2496, "step": 23760 }, { "epoch": 2.45, "learning_rate": 1.8930636094885234e-05, "loss": 0.2416, "step": 23761 }, { "epoch": 2.46, "learning_rate": 1.8928348395135782e-05, "loss": 0.2592, "step": 23762 }, { "epoch": 2.46, "learning_rate": 1.8926060769918094e-05, "loss": 0.3737, "step": 23763 }, { "epoch": 2.46, "learning_rate": 1.8923773219247576e-05, "loss": 0.3588, "step": 23764 }, { "epoch": 2.46, "learning_rate": 1.8921485743139627e-05, "loss": 0.548, "step": 23765 }, { "epoch": 2.46, "learning_rate": 1.8919198341609646e-05, "loss": 0.1291, "step": 23766 }, { "epoch": 2.46, "learning_rate": 1.891691101467302e-05, "loss": 0.401, "step": 23767 }, { "epoch": 2.46, "learning_rate": 1.8914623762345163e-05, "loss": 0.4136, "step": 23768 }, { "epoch": 2.46, "learning_rate": 1.891233658464146e-05, "loss": 0.4123, "step": 23769 }, { "epoch": 2.46, "learning_rate": 1.8910049481577324e-05, "loss": 0.0665, "step": 23770 }, { "epoch": 2.46, "learning_rate": 1.890776245316813e-05, "loss": 0.357, "step": 23771 }, { "epoch": 2.46, "learning_rate": 1.8905475499429293e-05, "loss": 0.3118, "step": 23772 }, { "epoch": 2.46, "learning_rate": 1.8903188620376196e-05, "loss": 0.38, "step": 23773 }, { "epoch": 2.46, "learning_rate": 1.8900901816024226e-05, "loss": 0.7209, "step": 23774 }, { "epoch": 2.46, "learning_rate": 1.8898615086388795e-05, "loss": 0.3075, "step": 23775 }, { "epoch": 2.46, "learning_rate": 1.8896328431485296e-05, "loss": 0.2472, "step": 23776 }, { "epoch": 2.46, "learning_rate": 1.8894041851329106e-05, "loss": 0.179, "step": 23777 }, { "epoch": 2.46, "learning_rate": 1.889175534593563e-05, "loss": 0.5881, "step": 23778 }, { "epoch": 2.46, "learning_rate": 1.8889468915320257e-05, "loss": 0.2669, "step": 23779 }, { "epoch": 2.46, "learning_rate": 1.888718255949839e-05, "loss": 0.4147, "step": 23780 }, { "epoch": 2.46, "learning_rate": 1.88848962784854e-05, "loss": 1.197, "step": 23781 }, { "epoch": 2.46, "learning_rate": 1.8882610072296686e-05, "loss": 0.2306, "step": 23782 }, { "epoch": 2.46, "learning_rate": 1.8880323940947643e-05, "loss": 0.1019, "step": 23783 }, { "epoch": 2.46, "learning_rate": 1.887803788445366e-05, "loss": 0.553, "step": 23784 }, { "epoch": 2.46, "learning_rate": 1.8875751902830124e-05, "loss": 0.111, "step": 23785 }, { "epoch": 2.46, "learning_rate": 1.887346599609243e-05, "loss": 0.36, "step": 23786 }, { "epoch": 2.46, "learning_rate": 1.8871180164255947e-05, "loss": 0.2464, "step": 23787 }, { "epoch": 2.46, "learning_rate": 1.8868894407336076e-05, "loss": 0.4789, "step": 23788 }, { "epoch": 2.46, "learning_rate": 1.8866608725348203e-05, "loss": 0.1517, "step": 23789 }, { "epoch": 2.46, "learning_rate": 1.8864323118307722e-05, "loss": 0.5407, "step": 23790 }, { "epoch": 2.46, "learning_rate": 1.8862037586230006e-05, "loss": 0.2625, "step": 23791 }, { "epoch": 2.46, "learning_rate": 1.8859752129130444e-05, "loss": 0.2462, "step": 23792 }, { "epoch": 2.46, "learning_rate": 1.8857466747024432e-05, "loss": 0.0937, "step": 23793 }, { "epoch": 2.46, "learning_rate": 1.8855181439927344e-05, "loss": 0.5825, "step": 23794 }, { "epoch": 2.46, "learning_rate": 1.8852896207854566e-05, "loss": 0.329, "step": 23795 }, { "epoch": 2.46, "learning_rate": 1.885061105082148e-05, "loss": 0.5679, "step": 23796 }, { "epoch": 2.46, "learning_rate": 1.8848325968843487e-05, "loss": 0.1301, "step": 23797 }, { "epoch": 2.46, "learning_rate": 1.884604096193594e-05, "loss": 0.4756, "step": 23798 }, { "epoch": 2.46, "learning_rate": 1.8843756030114237e-05, "loss": 0.1446, "step": 23799 }, { "epoch": 2.46, "learning_rate": 1.8841471173393763e-05, "loss": 0.6147, "step": 23800 }, { "epoch": 2.46, "learning_rate": 1.8839186391789888e-05, "loss": 0.358, "step": 23801 }, { "epoch": 2.46, "learning_rate": 1.8836901685317995e-05, "loss": 0.5071, "step": 23802 }, { "epoch": 2.46, "learning_rate": 1.8834617053993472e-05, "loss": 0.2504, "step": 23803 }, { "epoch": 2.46, "learning_rate": 1.88323324978317e-05, "loss": 0.538, "step": 23804 }, { "epoch": 2.46, "learning_rate": 1.883004801684805e-05, "loss": 0.2578, "step": 23805 }, { "epoch": 2.46, "learning_rate": 1.8827763611057898e-05, "loss": 0.6528, "step": 23806 }, { "epoch": 2.46, "learning_rate": 1.8825479280476628e-05, "loss": 0.4263, "step": 23807 }, { "epoch": 2.46, "learning_rate": 1.8823195025119625e-05, "loss": 0.1399, "step": 23808 }, { "epoch": 2.46, "learning_rate": 1.882091084500225e-05, "loss": 0.2023, "step": 23809 }, { "epoch": 2.46, "learning_rate": 1.88186267401399e-05, "loss": 0.3836, "step": 23810 }, { "epoch": 2.46, "learning_rate": 1.8816342710547926e-05, "loss": 0.763, "step": 23811 }, { "epoch": 2.46, "learning_rate": 1.881405875624172e-05, "loss": 0.5056, "step": 23812 }, { "epoch": 2.46, "learning_rate": 1.881177487723665e-05, "loss": 0.5743, "step": 23813 }, { "epoch": 2.46, "learning_rate": 1.88094910735481e-05, "loss": 0.2799, "step": 23814 }, { "epoch": 2.46, "learning_rate": 1.880720734519143e-05, "loss": 0.8704, "step": 23815 }, { "epoch": 2.46, "learning_rate": 1.8804923692182022e-05, "loss": 0.1943, "step": 23816 }, { "epoch": 2.46, "learning_rate": 1.880264011453525e-05, "loss": 0.7375, "step": 23817 }, { "epoch": 2.46, "learning_rate": 1.880035661226649e-05, "loss": 0.3487, "step": 23818 }, { "epoch": 2.46, "learning_rate": 1.8798073185391102e-05, "loss": 0.2535, "step": 23819 }, { "epoch": 2.46, "learning_rate": 1.8795789833924466e-05, "loss": 0.5149, "step": 23820 }, { "epoch": 2.46, "learning_rate": 1.879350655788195e-05, "loss": 0.4878, "step": 23821 }, { "epoch": 2.46, "learning_rate": 1.8791223357278934e-05, "loss": 0.4618, "step": 23822 }, { "epoch": 2.46, "learning_rate": 1.8788940232130777e-05, "loss": 0.4704, "step": 23823 }, { "epoch": 2.46, "learning_rate": 1.878665718245285e-05, "loss": 0.7884, "step": 23824 }, { "epoch": 2.46, "learning_rate": 1.8784374208260517e-05, "loss": 0.587, "step": 23825 }, { "epoch": 2.46, "learning_rate": 1.8782091309569156e-05, "loss": 0.1366, "step": 23826 }, { "epoch": 2.46, "learning_rate": 1.877980848639413e-05, "loss": 0.2486, "step": 23827 }, { "epoch": 2.46, "learning_rate": 1.8777525738750816e-05, "loss": 0.4866, "step": 23828 }, { "epoch": 2.46, "learning_rate": 1.877524306665456e-05, "loss": 0.2812, "step": 23829 }, { "epoch": 2.46, "learning_rate": 1.8772960470120747e-05, "loss": 0.1971, "step": 23830 }, { "epoch": 2.46, "learning_rate": 1.8770677949164738e-05, "loss": 0.7785, "step": 23831 }, { "epoch": 2.46, "learning_rate": 1.87683955038019e-05, "loss": 0.6369, "step": 23832 }, { "epoch": 2.46, "learning_rate": 1.8766113134047594e-05, "loss": 0.4068, "step": 23833 }, { "epoch": 2.46, "learning_rate": 1.8763830839917183e-05, "loss": 0.353, "step": 23834 }, { "epoch": 2.46, "learning_rate": 1.8761548621426048e-05, "loss": 0.2235, "step": 23835 }, { "epoch": 2.46, "learning_rate": 1.875926647858952e-05, "loss": 0.3977, "step": 23836 }, { "epoch": 2.46, "learning_rate": 1.8756984411422988e-05, "loss": 0.5121, "step": 23837 }, { "epoch": 2.46, "learning_rate": 1.8754702419941806e-05, "loss": 0.1166, "step": 23838 }, { "epoch": 2.46, "learning_rate": 1.875242050416133e-05, "loss": 0.5748, "step": 23839 }, { "epoch": 2.46, "learning_rate": 1.8750138664096934e-05, "loss": 0.4791, "step": 23840 }, { "epoch": 2.46, "learning_rate": 1.8747856899763967e-05, "loss": 0.2693, "step": 23841 }, { "epoch": 2.46, "learning_rate": 1.8745575211177803e-05, "loss": 0.5187, "step": 23842 }, { "epoch": 2.46, "learning_rate": 1.8743293598353782e-05, "loss": 0.085, "step": 23843 }, { "epoch": 2.46, "learning_rate": 1.874101206130728e-05, "loss": 0.1975, "step": 23844 }, { "epoch": 2.46, "learning_rate": 1.873873060005365e-05, "loss": 0.6403, "step": 23845 }, { "epoch": 2.46, "learning_rate": 1.8736449214608255e-05, "loss": 0.4488, "step": 23846 }, { "epoch": 2.46, "learning_rate": 1.8734167904986444e-05, "loss": 0.4032, "step": 23847 }, { "epoch": 2.46, "learning_rate": 1.873188667120359e-05, "loss": 1.2463, "step": 23848 }, { "epoch": 2.46, "learning_rate": 1.8729605513275025e-05, "loss": 0.903, "step": 23849 }, { "epoch": 2.46, "learning_rate": 1.8727324431216123e-05, "loss": 0.4836, "step": 23850 }, { "epoch": 2.46, "learning_rate": 1.8725043425042233e-05, "loss": 1.3736, "step": 23851 }, { "epoch": 2.46, "learning_rate": 1.8722762494768715e-05, "loss": 0.3967, "step": 23852 }, { "epoch": 2.46, "learning_rate": 1.8720481640410918e-05, "loss": 0.1681, "step": 23853 }, { "epoch": 2.46, "learning_rate": 1.8718200861984203e-05, "loss": 0.6722, "step": 23854 }, { "epoch": 2.46, "learning_rate": 1.8715920159503913e-05, "loss": 0.5027, "step": 23855 }, { "epoch": 2.46, "learning_rate": 1.871363953298542e-05, "loss": 0.0967, "step": 23856 }, { "epoch": 2.46, "learning_rate": 1.8711358982444056e-05, "loss": 0.2932, "step": 23857 }, { "epoch": 2.47, "learning_rate": 1.8709078507895182e-05, "loss": 0.5011, "step": 23858 }, { "epoch": 2.47, "learning_rate": 1.8706798109354152e-05, "loss": 0.8463, "step": 23859 }, { "epoch": 2.47, "learning_rate": 1.870451778683632e-05, "loss": 0.1455, "step": 23860 }, { "epoch": 2.47, "learning_rate": 1.8702237540357028e-05, "loss": 0.5826, "step": 23861 }, { "epoch": 2.47, "learning_rate": 1.869995736993163e-05, "loss": 0.176, "step": 23862 }, { "epoch": 2.47, "learning_rate": 1.8697677275575474e-05, "loss": 0.1289, "step": 23863 }, { "epoch": 2.47, "learning_rate": 1.8695397257303906e-05, "loss": 0.6988, "step": 23864 }, { "epoch": 2.47, "learning_rate": 1.8693117315132274e-05, "loss": 0.2385, "step": 23865 }, { "epoch": 2.47, "learning_rate": 1.8690837449075945e-05, "loss": 0.1231, "step": 23866 }, { "epoch": 2.47, "learning_rate": 1.8688557659150247e-05, "loss": 0.4613, "step": 23867 }, { "epoch": 2.47, "learning_rate": 1.8686277945370526e-05, "loss": 0.1972, "step": 23868 }, { "epoch": 2.47, "learning_rate": 1.8683998307752136e-05, "loss": 0.3516, "step": 23869 }, { "epoch": 2.47, "learning_rate": 1.868171874631043e-05, "loss": 0.1149, "step": 23870 }, { "epoch": 2.47, "learning_rate": 1.8679439261060738e-05, "loss": 0.1994, "step": 23871 }, { "epoch": 2.47, "learning_rate": 1.867715985201841e-05, "loss": 0.2206, "step": 23872 }, { "epoch": 2.47, "learning_rate": 1.8674880519198804e-05, "loss": 0.8962, "step": 23873 }, { "epoch": 2.47, "learning_rate": 1.8672601262617243e-05, "loss": 0.6054, "step": 23874 }, { "epoch": 2.47, "learning_rate": 1.867032208228908e-05, "loss": 1.3738, "step": 23875 }, { "epoch": 2.47, "learning_rate": 1.8668042978229663e-05, "loss": 0.7286, "step": 23876 }, { "epoch": 2.47, "learning_rate": 1.8665763950454327e-05, "loss": 0.1106, "step": 23877 }, { "epoch": 2.47, "learning_rate": 1.866348499897841e-05, "loss": 0.4887, "step": 23878 }, { "epoch": 2.47, "learning_rate": 1.8661206123817264e-05, "loss": 0.5867, "step": 23879 }, { "epoch": 2.47, "learning_rate": 1.865892732498623e-05, "loss": 0.3998, "step": 23880 }, { "epoch": 2.47, "learning_rate": 1.865664860250064e-05, "loss": 0.3573, "step": 23881 }, { "epoch": 2.47, "learning_rate": 1.865436995637583e-05, "loss": 0.3978, "step": 23882 }, { "epoch": 2.47, "learning_rate": 1.8652091386627155e-05, "loss": 0.9392, "step": 23883 }, { "epoch": 2.47, "learning_rate": 1.864981289326995e-05, "loss": 0.7756, "step": 23884 }, { "epoch": 2.47, "learning_rate": 1.8647534476319542e-05, "loss": 0.0366, "step": 23885 }, { "epoch": 2.47, "learning_rate": 1.8645256135791283e-05, "loss": 0.2976, "step": 23886 }, { "epoch": 2.47, "learning_rate": 1.8642977871700493e-05, "loss": 0.4177, "step": 23887 }, { "epoch": 2.47, "learning_rate": 1.864069968406252e-05, "loss": 0.8783, "step": 23888 }, { "epoch": 2.47, "learning_rate": 1.86384215728927e-05, "loss": 0.1091, "step": 23889 }, { "epoch": 2.47, "learning_rate": 1.8636143538206375e-05, "loss": 0.422, "step": 23890 }, { "epoch": 2.47, "learning_rate": 1.863386558001886e-05, "loss": 0.1825, "step": 23891 }, { "epoch": 2.47, "learning_rate": 1.863158769834551e-05, "loss": 0.4662, "step": 23892 }, { "epoch": 2.47, "learning_rate": 1.8629309893201646e-05, "loss": 0.2929, "step": 23893 }, { "epoch": 2.47, "learning_rate": 1.8627032164602616e-05, "loss": 0.2796, "step": 23894 }, { "epoch": 2.47, "learning_rate": 1.862475451256374e-05, "loss": 0.148, "step": 23895 }, { "epoch": 2.47, "learning_rate": 1.862247693710035e-05, "loss": 0.7047, "step": 23896 }, { "epoch": 2.47, "learning_rate": 1.8620199438227788e-05, "loss": 0.2519, "step": 23897 }, { "epoch": 2.47, "learning_rate": 1.861792201596139e-05, "loss": 0.4522, "step": 23898 }, { "epoch": 2.47, "learning_rate": 1.861564467031647e-05, "loss": 0.3216, "step": 23899 }, { "epoch": 2.47, "learning_rate": 1.8613367401308364e-05, "loss": 0.5565, "step": 23900 }, { "epoch": 2.47, "learning_rate": 1.8611090208952408e-05, "loss": 0.2431, "step": 23901 }, { "epoch": 2.47, "learning_rate": 1.860881309326392e-05, "loss": 0.3571, "step": 23902 }, { "epoch": 2.47, "learning_rate": 1.860653605425824e-05, "loss": 0.3067, "step": 23903 }, { "epoch": 2.47, "learning_rate": 1.8604259091950704e-05, "loss": 0.204, "step": 23904 }, { "epoch": 2.47, "learning_rate": 1.860198220635662e-05, "loss": 0.2818, "step": 23905 }, { "epoch": 2.47, "learning_rate": 1.8599705397491323e-05, "loss": 0.4097, "step": 23906 }, { "epoch": 2.47, "learning_rate": 1.8597428665370143e-05, "loss": 0.1, "step": 23907 }, { "epoch": 2.47, "learning_rate": 1.8595152010008412e-05, "loss": 0.6076, "step": 23908 }, { "epoch": 2.47, "learning_rate": 1.859287543142144e-05, "loss": 0.2189, "step": 23909 }, { "epoch": 2.47, "learning_rate": 1.8590598929624574e-05, "loss": 0.5284, "step": 23910 }, { "epoch": 2.47, "learning_rate": 1.858832250463312e-05, "loss": 0.3901, "step": 23911 }, { "epoch": 2.47, "learning_rate": 1.8586046156462405e-05, "loss": 0.5884, "step": 23912 }, { "epoch": 2.47, "learning_rate": 1.8583769885127754e-05, "loss": 0.0636, "step": 23913 }, { "epoch": 2.47, "learning_rate": 1.85814936906445e-05, "loss": 0.5089, "step": 23914 }, { "epoch": 2.47, "learning_rate": 1.8579217573027956e-05, "loss": 0.2325, "step": 23915 }, { "epoch": 2.47, "learning_rate": 1.8576941532293445e-05, "loss": 0.2993, "step": 23916 }, { "epoch": 2.47, "learning_rate": 1.857466556845629e-05, "loss": 0.2843, "step": 23917 }, { "epoch": 2.47, "learning_rate": 1.8572389681531822e-05, "loss": 0.2136, "step": 23918 }, { "epoch": 2.47, "learning_rate": 1.8570113871535346e-05, "loss": 0.276, "step": 23919 }, { "epoch": 2.47, "learning_rate": 1.8567838138482185e-05, "loss": 0.5057, "step": 23920 }, { "epoch": 2.47, "learning_rate": 1.856556248238767e-05, "loss": 0.542, "step": 23921 }, { "epoch": 2.47, "learning_rate": 1.8563286903267113e-05, "loss": 0.0715, "step": 23922 }, { "epoch": 2.47, "learning_rate": 1.8561011401135834e-05, "loss": 0.4254, "step": 23923 }, { "epoch": 2.47, "learning_rate": 1.855873597600915e-05, "loss": 0.1913, "step": 23924 }, { "epoch": 2.47, "learning_rate": 1.8556460627902374e-05, "loss": 0.567, "step": 23925 }, { "epoch": 2.47, "learning_rate": 1.8554185356830825e-05, "loss": 0.2234, "step": 23926 }, { "epoch": 2.47, "learning_rate": 1.855191016280982e-05, "loss": 0.2055, "step": 23927 }, { "epoch": 2.47, "learning_rate": 1.8549635045854693e-05, "loss": 0.0478, "step": 23928 }, { "epoch": 2.47, "learning_rate": 1.854736000598073e-05, "loss": 1.037, "step": 23929 }, { "epoch": 2.47, "learning_rate": 1.8545085043203256e-05, "loss": 0.0996, "step": 23930 }, { "epoch": 2.47, "learning_rate": 1.8542810157537593e-05, "loss": 0.1902, "step": 23931 }, { "epoch": 2.47, "learning_rate": 1.854053534899906e-05, "loss": 0.3855, "step": 23932 }, { "epoch": 2.47, "learning_rate": 1.8538260617602956e-05, "loss": 0.055, "step": 23933 }, { "epoch": 2.47, "learning_rate": 1.8535985963364597e-05, "loss": 0.2056, "step": 23934 }, { "epoch": 2.47, "learning_rate": 1.8533711386299308e-05, "loss": 0.4439, "step": 23935 }, { "epoch": 2.47, "learning_rate": 1.8531436886422383e-05, "loss": 0.2112, "step": 23936 }, { "epoch": 2.47, "learning_rate": 1.8529162463749137e-05, "loss": 0.3322, "step": 23937 }, { "epoch": 2.47, "learning_rate": 1.8526888118294898e-05, "loss": 0.1238, "step": 23938 }, { "epoch": 2.47, "learning_rate": 1.8524613850074955e-05, "loss": 0.3483, "step": 23939 }, { "epoch": 2.47, "learning_rate": 1.8522339659104624e-05, "loss": 0.4835, "step": 23940 }, { "epoch": 2.47, "learning_rate": 1.852006554539922e-05, "loss": 0.7069, "step": 23941 }, { "epoch": 2.47, "learning_rate": 1.851779150897405e-05, "loss": 0.787, "step": 23942 }, { "epoch": 2.47, "learning_rate": 1.851551754984442e-05, "loss": 0.3628, "step": 23943 }, { "epoch": 2.47, "learning_rate": 1.8513243668025638e-05, "loss": 0.1815, "step": 23944 }, { "epoch": 2.47, "learning_rate": 1.851096986353301e-05, "loss": 0.2095, "step": 23945 }, { "epoch": 2.47, "learning_rate": 1.8508696136381852e-05, "loss": 0.3225, "step": 23946 }, { "epoch": 2.47, "learning_rate": 1.8506422486587455e-05, "loss": 0.1104, "step": 23947 }, { "epoch": 2.47, "learning_rate": 1.8504148914165146e-05, "loss": 1.2905, "step": 23948 }, { "epoch": 2.47, "learning_rate": 1.8501875419130206e-05, "loss": 0.2071, "step": 23949 }, { "epoch": 2.47, "learning_rate": 1.849960200149795e-05, "loss": 0.3539, "step": 23950 }, { "epoch": 2.47, "learning_rate": 1.849732866128368e-05, "loss": 0.1536, "step": 23951 }, { "epoch": 2.47, "learning_rate": 1.8495055398502712e-05, "loss": 0.2349, "step": 23952 }, { "epoch": 2.47, "learning_rate": 1.849278221317033e-05, "loss": 0.6455, "step": 23953 }, { "epoch": 2.48, "learning_rate": 1.8490509105301845e-05, "loss": 0.5257, "step": 23954 }, { "epoch": 2.48, "learning_rate": 1.848823607491256e-05, "loss": 0.2028, "step": 23955 }, { "epoch": 2.48, "learning_rate": 1.848596312201778e-05, "loss": 0.2994, "step": 23956 }, { "epoch": 2.48, "learning_rate": 1.84836902466328e-05, "loss": 0.3864, "step": 23957 }, { "epoch": 2.48, "learning_rate": 1.8481417448772925e-05, "loss": 0.1551, "step": 23958 }, { "epoch": 2.48, "learning_rate": 1.847914472845345e-05, "loss": 0.6318, "step": 23959 }, { "epoch": 2.48, "learning_rate": 1.847687208568968e-05, "loss": 0.1619, "step": 23960 }, { "epoch": 2.48, "learning_rate": 1.8474599520496916e-05, "loss": 0.3749, "step": 23961 }, { "epoch": 2.48, "learning_rate": 1.8472327032890445e-05, "loss": 0.3627, "step": 23962 }, { "epoch": 2.48, "learning_rate": 1.847005462288557e-05, "loss": 0.3388, "step": 23963 }, { "epoch": 2.48, "learning_rate": 1.8467782290497593e-05, "loss": 0.426, "step": 23964 }, { "epoch": 2.48, "learning_rate": 1.84655100357418e-05, "loss": 0.2761, "step": 23965 }, { "epoch": 2.48, "learning_rate": 1.84632378586335e-05, "loss": 0.2717, "step": 23966 }, { "epoch": 2.48, "learning_rate": 1.8460965759187983e-05, "loss": 0.1455, "step": 23967 }, { "epoch": 2.48, "learning_rate": 1.8458693737420542e-05, "loss": 1.0773, "step": 23968 }, { "epoch": 2.48, "learning_rate": 1.8456421793346477e-05, "loss": 0.1872, "step": 23969 }, { "epoch": 2.48, "learning_rate": 1.8454149926981084e-05, "loss": 0.1699, "step": 23970 }, { "epoch": 2.48, "learning_rate": 1.8451878138339645e-05, "loss": 0.0904, "step": 23971 }, { "epoch": 2.48, "learning_rate": 1.844960642743746e-05, "loss": 0.4862, "step": 23972 }, { "epoch": 2.48, "learning_rate": 1.844733479428983e-05, "loss": 0.3438, "step": 23973 }, { "epoch": 2.48, "learning_rate": 1.8445063238912036e-05, "loss": 0.204, "step": 23974 }, { "epoch": 2.48, "learning_rate": 1.844279176131937e-05, "loss": 0.2295, "step": 23975 }, { "epoch": 2.48, "learning_rate": 1.844052036152713e-05, "loss": 0.2819, "step": 23976 }, { "epoch": 2.48, "learning_rate": 1.8438249039550594e-05, "loss": 0.1779, "step": 23977 }, { "epoch": 2.48, "learning_rate": 1.843597779540506e-05, "loss": 0.1532, "step": 23978 }, { "epoch": 2.48, "learning_rate": 1.843370662910582e-05, "loss": 0.4665, "step": 23979 }, { "epoch": 2.48, "learning_rate": 1.8431435540668168e-05, "loss": 0.3418, "step": 23980 }, { "epoch": 2.48, "learning_rate": 1.842916453010738e-05, "loss": 0.2947, "step": 23981 }, { "epoch": 2.48, "learning_rate": 1.8426893597438744e-05, "loss": 0.1291, "step": 23982 }, { "epoch": 2.48, "learning_rate": 1.8424622742677556e-05, "loss": 0.9288, "step": 23983 }, { "epoch": 2.48, "learning_rate": 1.8422351965839104e-05, "loss": 0.3911, "step": 23984 }, { "epoch": 2.48, "learning_rate": 1.8420081266938664e-05, "loss": 0.1331, "step": 23985 }, { "epoch": 2.48, "learning_rate": 1.841781064599154e-05, "loss": 0.2655, "step": 23986 }, { "epoch": 2.48, "learning_rate": 1.8415540103012992e-05, "loss": 0.3976, "step": 23987 }, { "epoch": 2.48, "learning_rate": 1.841326963801832e-05, "loss": 0.1237, "step": 23988 }, { "epoch": 2.48, "learning_rate": 1.8410999251022806e-05, "loss": 0.6794, "step": 23989 }, { "epoch": 2.48, "learning_rate": 1.840872894204173e-05, "loss": 0.4548, "step": 23990 }, { "epoch": 2.48, "learning_rate": 1.8406458711090382e-05, "loss": 0.1123, "step": 23991 }, { "epoch": 2.48, "learning_rate": 1.840418855818404e-05, "loss": 0.5821, "step": 23992 }, { "epoch": 2.48, "learning_rate": 1.8401918483337986e-05, "loss": 0.1845, "step": 23993 }, { "epoch": 2.48, "learning_rate": 1.839964848656751e-05, "loss": 0.4839, "step": 23994 }, { "epoch": 2.48, "learning_rate": 1.8397378567887877e-05, "loss": 0.4567, "step": 23995 }, { "epoch": 2.48, "learning_rate": 1.8395108727314382e-05, "loss": 0.0287, "step": 23996 }, { "epoch": 2.48, "learning_rate": 1.8392838964862296e-05, "loss": 0.5268, "step": 23997 }, { "epoch": 2.48, "learning_rate": 1.8390569280546916e-05, "loss": 0.4684, "step": 23998 }, { "epoch": 2.48, "learning_rate": 1.8388299674383494e-05, "loss": 0.6375, "step": 23999 }, { "epoch": 2.48, "learning_rate": 1.8386030146387326e-05, "loss": 0.4028, "step": 24000 }, { "epoch": 2.48, "learning_rate": 1.8383760696573684e-05, "loss": 0.1362, "step": 24001 }, { "epoch": 2.48, "learning_rate": 1.8381491324957843e-05, "loss": 0.1671, "step": 24002 }, { "epoch": 2.48, "learning_rate": 1.8379222031555086e-05, "loss": 0.6734, "step": 24003 }, { "epoch": 2.48, "learning_rate": 1.8376952816380694e-05, "loss": 0.2857, "step": 24004 }, { "epoch": 2.48, "learning_rate": 1.8374683679449927e-05, "loss": 0.087, "step": 24005 }, { "epoch": 2.48, "learning_rate": 1.8372414620778072e-05, "loss": 0.6314, "step": 24006 }, { "epoch": 2.48, "learning_rate": 1.8370145640380397e-05, "loss": 0.4055, "step": 24007 }, { "epoch": 2.48, "learning_rate": 1.836787673827219e-05, "loss": 0.0554, "step": 24008 }, { "epoch": 2.48, "learning_rate": 1.836560791446871e-05, "loss": 0.4287, "step": 24009 }, { "epoch": 2.48, "learning_rate": 1.8363339168985232e-05, "loss": 0.1763, "step": 24010 }, { "epoch": 2.48, "learning_rate": 1.836107050183705e-05, "loss": 0.8755, "step": 24011 }, { "epoch": 2.48, "learning_rate": 1.8358801913039403e-05, "loss": 0.1578, "step": 24012 }, { "epoch": 2.48, "learning_rate": 1.8356533402607574e-05, "loss": 0.1513, "step": 24013 }, { "epoch": 2.48, "learning_rate": 1.8354264970556846e-05, "loss": 0.1959, "step": 24014 }, { "epoch": 2.48, "learning_rate": 1.8351996616902477e-05, "loss": 0.5622, "step": 24015 }, { "epoch": 2.48, "learning_rate": 1.8349728341659743e-05, "loss": 0.1638, "step": 24016 }, { "epoch": 2.48, "learning_rate": 1.8347460144843905e-05, "loss": 0.3006, "step": 24017 }, { "epoch": 2.48, "learning_rate": 1.834519202647025e-05, "loss": 0.023, "step": 24018 }, { "epoch": 2.48, "learning_rate": 1.834292398655403e-05, "loss": 0.6399, "step": 24019 }, { "epoch": 2.48, "learning_rate": 1.8340656025110514e-05, "loss": 0.7114, "step": 24020 }, { "epoch": 2.48, "learning_rate": 1.8338388142154976e-05, "loss": 0.1096, "step": 24021 }, { "epoch": 2.48, "learning_rate": 1.833612033770269e-05, "loss": 0.3924, "step": 24022 }, { "epoch": 2.48, "learning_rate": 1.8333852611768905e-05, "loss": 0.3688, "step": 24023 }, { "epoch": 2.48, "learning_rate": 1.83315849643689e-05, "loss": 0.4891, "step": 24024 }, { "epoch": 2.48, "learning_rate": 1.832931739551793e-05, "loss": 0.3441, "step": 24025 }, { "epoch": 2.48, "learning_rate": 1.832704990523126e-05, "loss": 0.3138, "step": 24026 }, { "epoch": 2.48, "learning_rate": 1.8324782493524162e-05, "loss": 0.3144, "step": 24027 }, { "epoch": 2.48, "learning_rate": 1.83225151604119e-05, "loss": 0.2575, "step": 24028 }, { "epoch": 2.48, "learning_rate": 1.832024790590973e-05, "loss": 0.2695, "step": 24029 }, { "epoch": 2.48, "learning_rate": 1.8317980730032915e-05, "loss": 0.6153, "step": 24030 }, { "epoch": 2.48, "learning_rate": 1.8315713632796726e-05, "loss": 0.27, "step": 24031 }, { "epoch": 2.48, "learning_rate": 1.8313446614216417e-05, "loss": 0.3719, "step": 24032 }, { "epoch": 2.48, "learning_rate": 1.8311179674307253e-05, "loss": 0.3952, "step": 24033 }, { "epoch": 2.48, "learning_rate": 1.830891281308449e-05, "loss": 0.2266, "step": 24034 }, { "epoch": 2.48, "learning_rate": 1.830664603056339e-05, "loss": 0.2265, "step": 24035 }, { "epoch": 2.48, "learning_rate": 1.830437932675922e-05, "loss": 0.1976, "step": 24036 }, { "epoch": 2.48, "learning_rate": 1.8302112701687225e-05, "loss": 0.1098, "step": 24037 }, { "epoch": 2.48, "learning_rate": 1.8299846155362678e-05, "loss": 0.188, "step": 24038 }, { "epoch": 2.48, "learning_rate": 1.8297579687800818e-05, "loss": 0.3551, "step": 24039 }, { "epoch": 2.48, "learning_rate": 1.8295313299016918e-05, "loss": 0.088, "step": 24040 }, { "epoch": 2.48, "learning_rate": 1.829304698902623e-05, "loss": 0.5309, "step": 24041 }, { "epoch": 2.48, "learning_rate": 1.829078075784401e-05, "loss": 0.2605, "step": 24042 }, { "epoch": 2.48, "learning_rate": 1.8288514605485515e-05, "loss": 0.2124, "step": 24043 }, { "epoch": 2.48, "learning_rate": 1.8286248531965994e-05, "loss": 0.4921, "step": 24044 }, { "epoch": 2.48, "learning_rate": 1.828398253730071e-05, "loss": 0.4107, "step": 24045 }, { "epoch": 2.48, "learning_rate": 1.828171662150492e-05, "loss": 0.3303, "step": 24046 }, { "epoch": 2.48, "learning_rate": 1.827945078459387e-05, "loss": 0.835, "step": 24047 }, { "epoch": 2.48, "learning_rate": 1.827718502658281e-05, "loss": 0.3689, "step": 24048 }, { "epoch": 2.49, "learning_rate": 1.8274919347487012e-05, "loss": 0.1687, "step": 24049 }, { "epoch": 2.49, "learning_rate": 1.8272653747321696e-05, "loss": 0.5824, "step": 24050 }, { "epoch": 2.49, "learning_rate": 1.8270388226102138e-05, "loss": 0.2526, "step": 24051 }, { "epoch": 2.49, "learning_rate": 1.8268122783843587e-05, "loss": 0.2816, "step": 24052 }, { "epoch": 2.49, "learning_rate": 1.8265857420561277e-05, "loss": 0.1624, "step": 24053 }, { "epoch": 2.49, "learning_rate": 1.8263592136270475e-05, "loss": 0.1692, "step": 24054 }, { "epoch": 2.49, "learning_rate": 1.826132693098642e-05, "loss": 0.4885, "step": 24055 }, { "epoch": 2.49, "learning_rate": 1.8259061804724373e-05, "loss": 0.3618, "step": 24056 }, { "epoch": 2.49, "learning_rate": 1.8256796757499573e-05, "loss": 0.7095, "step": 24057 }, { "epoch": 2.49, "learning_rate": 1.8254531789327263e-05, "loss": 0.1651, "step": 24058 }, { "epoch": 2.49, "learning_rate": 1.8252266900222698e-05, "loss": 0.4008, "step": 24059 }, { "epoch": 2.49, "learning_rate": 1.8250002090201133e-05, "loss": 0.2739, "step": 24060 }, { "epoch": 2.49, "learning_rate": 1.8247737359277806e-05, "loss": 0.1907, "step": 24061 }, { "epoch": 2.49, "learning_rate": 1.824547270746796e-05, "loss": 0.329, "step": 24062 }, { "epoch": 2.49, "learning_rate": 1.824320813478683e-05, "loss": 0.1295, "step": 24063 }, { "epoch": 2.49, "learning_rate": 1.8240943641249675e-05, "loss": 0.773, "step": 24064 }, { "epoch": 2.49, "learning_rate": 1.823867922687174e-05, "loss": 0.2839, "step": 24065 }, { "epoch": 2.49, "learning_rate": 1.8236414891668265e-05, "loss": 0.9096, "step": 24066 }, { "epoch": 2.49, "learning_rate": 1.823415063565449e-05, "loss": 0.7741, "step": 24067 }, { "epoch": 2.49, "learning_rate": 1.8231886458845658e-05, "loss": 0.1641, "step": 24068 }, { "epoch": 2.49, "learning_rate": 1.8229622361257014e-05, "loss": 0.4607, "step": 24069 }, { "epoch": 2.49, "learning_rate": 1.8227358342903802e-05, "loss": 0.1321, "step": 24070 }, { "epoch": 2.49, "learning_rate": 1.8225094403801256e-05, "loss": 0.5856, "step": 24071 }, { "epoch": 2.49, "learning_rate": 1.8222830543964616e-05, "loss": 0.4846, "step": 24072 }, { "epoch": 2.49, "learning_rate": 1.822056676340913e-05, "loss": 0.377, "step": 24073 }, { "epoch": 2.49, "learning_rate": 1.8218303062150043e-05, "loss": 0.7194, "step": 24074 }, { "epoch": 2.49, "learning_rate": 1.8216039440202565e-05, "loss": 0.6463, "step": 24075 }, { "epoch": 2.49, "learning_rate": 1.8213775897581963e-05, "loss": 0.1196, "step": 24076 }, { "epoch": 2.49, "learning_rate": 1.821151243430346e-05, "loss": 0.3312, "step": 24077 }, { "epoch": 2.49, "learning_rate": 1.8209249050382295e-05, "loss": 0.267, "step": 24078 }, { "epoch": 2.49, "learning_rate": 1.8206985745833705e-05, "loss": 0.5387, "step": 24079 }, { "epoch": 2.49, "learning_rate": 1.8204722520672934e-05, "loss": 0.5151, "step": 24080 }, { "epoch": 2.49, "learning_rate": 1.820245937491521e-05, "loss": 0.3566, "step": 24081 }, { "epoch": 2.49, "learning_rate": 1.8200196308575764e-05, "loss": 0.4094, "step": 24082 }, { "epoch": 2.49, "learning_rate": 1.8197933321669836e-05, "loss": 0.7769, "step": 24083 }, { "epoch": 2.49, "learning_rate": 1.8195670414212668e-05, "loss": 0.2072, "step": 24084 }, { "epoch": 2.49, "learning_rate": 1.819340758621948e-05, "loss": 0.9763, "step": 24085 }, { "epoch": 2.49, "learning_rate": 1.819114483770551e-05, "loss": 0.2485, "step": 24086 }, { "epoch": 2.49, "learning_rate": 1.818888216868599e-05, "loss": 0.1384, "step": 24087 }, { "epoch": 2.49, "learning_rate": 1.8186619579176147e-05, "loss": 0.2895, "step": 24088 }, { "epoch": 2.49, "learning_rate": 1.818435706919122e-05, "loss": 0.5253, "step": 24089 }, { "epoch": 2.49, "learning_rate": 1.8182094638746438e-05, "loss": 0.2816, "step": 24090 }, { "epoch": 2.49, "learning_rate": 1.817983228785703e-05, "loss": 0.3897, "step": 24091 }, { "epoch": 2.49, "learning_rate": 1.817757001653822e-05, "loss": 0.5616, "step": 24092 }, { "epoch": 2.49, "learning_rate": 1.8175307824805242e-05, "loss": 0.3558, "step": 24093 }, { "epoch": 2.49, "learning_rate": 1.8173045712673335e-05, "loss": 0.2212, "step": 24094 }, { "epoch": 2.49, "learning_rate": 1.8170783680157708e-05, "loss": 0.3086, "step": 24095 }, { "epoch": 2.49, "learning_rate": 1.8168521727273603e-05, "loss": 0.3131, "step": 24096 }, { "epoch": 2.49, "learning_rate": 1.8166259854036233e-05, "loss": 0.282, "step": 24097 }, { "epoch": 2.49, "learning_rate": 1.816399806046084e-05, "loss": 0.4849, "step": 24098 }, { "epoch": 2.49, "learning_rate": 1.816173634656265e-05, "loss": 0.0788, "step": 24099 }, { "epoch": 2.49, "learning_rate": 1.8159474712356877e-05, "loss": 0.4768, "step": 24100 }, { "epoch": 2.49, "learning_rate": 1.815721315785874e-05, "loss": 0.3048, "step": 24101 }, { "epoch": 2.49, "learning_rate": 1.8154951683083477e-05, "loss": 0.242, "step": 24102 }, { "epoch": 2.49, "learning_rate": 1.8152690288046306e-05, "loss": 0.3748, "step": 24103 }, { "epoch": 2.49, "learning_rate": 1.8150428972762457e-05, "loss": 0.3953, "step": 24104 }, { "epoch": 2.49, "learning_rate": 1.8148167737247147e-05, "loss": 0.4864, "step": 24105 }, { "epoch": 2.49, "learning_rate": 1.814590658151559e-05, "loss": 0.4532, "step": 24106 }, { "epoch": 2.49, "learning_rate": 1.8143645505583026e-05, "loss": 0.4387, "step": 24107 }, { "epoch": 2.49, "learning_rate": 1.8141384509464665e-05, "loss": 0.2608, "step": 24108 }, { "epoch": 2.49, "learning_rate": 1.813912359317573e-05, "loss": 0.2941, "step": 24109 }, { "epoch": 2.49, "learning_rate": 1.813686275673143e-05, "loss": 0.588, "step": 24110 }, { "epoch": 2.49, "learning_rate": 1.8134602000147014e-05, "loss": 0.4394, "step": 24111 }, { "epoch": 2.49, "learning_rate": 1.8132341323437665e-05, "loss": 0.1426, "step": 24112 }, { "epoch": 2.49, "learning_rate": 1.8130080726618618e-05, "loss": 0.4882, "step": 24113 }, { "epoch": 2.49, "learning_rate": 1.8127820209705098e-05, "loss": 0.3823, "step": 24114 }, { "epoch": 2.49, "learning_rate": 1.8125559772712307e-05, "loss": 0.5946, "step": 24115 }, { "epoch": 2.49, "learning_rate": 1.8123299415655473e-05, "loss": 0.365, "step": 24116 }, { "epoch": 2.49, "learning_rate": 1.8121039138549804e-05, "loss": 0.6869, "step": 24117 }, { "epoch": 2.49, "learning_rate": 1.8118778941410524e-05, "loss": 0.7308, "step": 24118 }, { "epoch": 2.49, "learning_rate": 1.8116518824252844e-05, "loss": 0.1993, "step": 24119 }, { "epoch": 2.49, "learning_rate": 1.8114258787091972e-05, "loss": 0.1125, "step": 24120 }, { "epoch": 2.49, "learning_rate": 1.8111998829943133e-05, "loss": 1.8168, "step": 24121 }, { "epoch": 2.49, "learning_rate": 1.8109738952821543e-05, "loss": 0.1156, "step": 24122 }, { "epoch": 2.49, "learning_rate": 1.81074791557424e-05, "loss": 0.5046, "step": 24123 }, { "epoch": 2.49, "learning_rate": 1.8105219438720938e-05, "loss": 0.4261, "step": 24124 }, { "epoch": 2.49, "learning_rate": 1.8102959801772345e-05, "loss": 0.4543, "step": 24125 }, { "epoch": 2.49, "learning_rate": 1.8100700244911843e-05, "loss": 0.1499, "step": 24126 }, { "epoch": 2.49, "learning_rate": 1.809844076815464e-05, "loss": 0.1771, "step": 24127 }, { "epoch": 2.49, "learning_rate": 1.8096181371515958e-05, "loss": 0.5255, "step": 24128 }, { "epoch": 2.49, "learning_rate": 1.8093922055010988e-05, "loss": 0.2064, "step": 24129 }, { "epoch": 2.49, "learning_rate": 1.8091662818654956e-05, "loss": 0.3892, "step": 24130 }, { "epoch": 2.49, "learning_rate": 1.8089403662463057e-05, "loss": 0.3982, "step": 24131 }, { "epoch": 2.49, "learning_rate": 1.8087144586450514e-05, "loss": 0.088, "step": 24132 }, { "epoch": 2.49, "learning_rate": 1.8084885590632516e-05, "loss": 0.8761, "step": 24133 }, { "epoch": 2.49, "learning_rate": 1.8082626675024286e-05, "loss": 0.1765, "step": 24134 }, { "epoch": 2.49, "learning_rate": 1.808036783964102e-05, "loss": 0.2492, "step": 24135 }, { "epoch": 2.49, "learning_rate": 1.8078109084497947e-05, "loss": 0.2103, "step": 24136 }, { "epoch": 2.49, "learning_rate": 1.807585040961024e-05, "loss": 0.5044, "step": 24137 }, { "epoch": 2.49, "learning_rate": 1.807359181499312e-05, "loss": 0.1105, "step": 24138 }, { "epoch": 2.49, "learning_rate": 1.8071333300661786e-05, "loss": 0.2437, "step": 24139 }, { "epoch": 2.49, "learning_rate": 1.806907486663144e-05, "loss": 0.2416, "step": 24140 }, { "epoch": 2.49, "learning_rate": 1.8066816512917303e-05, "loss": 0.3129, "step": 24141 }, { "epoch": 2.49, "learning_rate": 1.806455823953456e-05, "loss": 0.0681, "step": 24142 }, { "epoch": 2.49, "learning_rate": 1.8062300046498417e-05, "loss": 0.1992, "step": 24143 }, { "epoch": 2.49, "learning_rate": 1.8060041933824075e-05, "loss": 0.3944, "step": 24144 }, { "epoch": 2.5, "learning_rate": 1.8057783901526736e-05, "loss": 0.2829, "step": 24145 }, { "epoch": 2.5, "learning_rate": 1.8055525949621612e-05, "loss": 1.3594, "step": 24146 }, { "epoch": 2.5, "learning_rate": 1.8053268078123883e-05, "loss": 0.6028, "step": 24147 }, { "epoch": 2.5, "learning_rate": 1.8051010287048757e-05, "loss": 0.119, "step": 24148 }, { "epoch": 2.5, "learning_rate": 1.8048752576411448e-05, "loss": 0.3634, "step": 24149 }, { "epoch": 2.5, "learning_rate": 1.8046494946227127e-05, "loss": 0.3433, "step": 24150 }, { "epoch": 2.5, "learning_rate": 1.8044237396511007e-05, "loss": 0.2383, "step": 24151 }, { "epoch": 2.5, "learning_rate": 1.8041979927278288e-05, "loss": 0.6624, "step": 24152 }, { "epoch": 2.5, "learning_rate": 1.8039722538544157e-05, "loss": 0.6936, "step": 24153 }, { "epoch": 2.5, "learning_rate": 1.803746523032382e-05, "loss": 0.1952, "step": 24154 }, { "epoch": 2.5, "learning_rate": 1.8035208002632458e-05, "loss": 0.0793, "step": 24155 }, { "epoch": 2.5, "learning_rate": 1.8032950855485287e-05, "loss": 0.5395, "step": 24156 }, { "epoch": 2.5, "learning_rate": 1.8030693788897486e-05, "loss": 0.2151, "step": 24157 }, { "epoch": 2.5, "learning_rate": 1.802843680288425e-05, "loss": 0.4747, "step": 24158 }, { "epoch": 2.5, "learning_rate": 1.8026179897460782e-05, "loss": 0.6827, "step": 24159 }, { "epoch": 2.5, "learning_rate": 1.802392307264227e-05, "loss": 0.5564, "step": 24160 }, { "epoch": 2.5, "learning_rate": 1.80216663284439e-05, "loss": 0.3379, "step": 24161 }, { "epoch": 2.5, "learning_rate": 1.8019409664880887e-05, "loss": 0.1992, "step": 24162 }, { "epoch": 2.5, "learning_rate": 1.801715308196839e-05, "loss": 0.3148, "step": 24163 }, { "epoch": 2.5, "learning_rate": 1.8014896579721612e-05, "loss": 0.4057, "step": 24164 }, { "epoch": 2.5, "learning_rate": 1.801264015815575e-05, "loss": 0.5167, "step": 24165 }, { "epoch": 2.5, "learning_rate": 1.8010383817285997e-05, "loss": 0.5236, "step": 24166 }, { "epoch": 2.5, "learning_rate": 1.8008127557127524e-05, "loss": 1.164, "step": 24167 }, { "epoch": 2.5, "learning_rate": 1.800587137769553e-05, "loss": 0.4219, "step": 24168 }, { "epoch": 2.5, "learning_rate": 1.800361527900521e-05, "loss": 0.3002, "step": 24169 }, { "epoch": 2.5, "learning_rate": 1.800135926107174e-05, "loss": 0.3123, "step": 24170 }, { "epoch": 2.5, "learning_rate": 1.799910332391032e-05, "loss": 0.2029, "step": 24171 }, { "epoch": 2.5, "learning_rate": 1.7996847467536117e-05, "loss": 0.5158, "step": 24172 }, { "epoch": 2.5, "learning_rate": 1.7994591691964333e-05, "loss": 0.7836, "step": 24173 }, { "epoch": 2.5, "learning_rate": 1.7992335997210157e-05, "loss": 0.0682, "step": 24174 }, { "epoch": 2.5, "learning_rate": 1.7990080383288756e-05, "loss": 0.1071, "step": 24175 }, { "epoch": 2.5, "learning_rate": 1.798782485021533e-05, "loss": 0.6096, "step": 24176 }, { "epoch": 2.5, "learning_rate": 1.7985569398005052e-05, "loss": 0.1603, "step": 24177 }, { "epoch": 2.5, "learning_rate": 1.798331402667311e-05, "loss": 0.381, "step": 24178 }, { "epoch": 2.5, "learning_rate": 1.7981058736234683e-05, "loss": 0.587, "step": 24179 }, { "epoch": 2.5, "learning_rate": 1.7978803526704964e-05, "loss": 0.6549, "step": 24180 }, { "epoch": 2.5, "learning_rate": 1.7976548398099123e-05, "loss": 0.4826, "step": 24181 }, { "epoch": 2.5, "learning_rate": 1.7974293350432342e-05, "loss": 0.1489, "step": 24182 }, { "epoch": 2.5, "learning_rate": 1.7972038383719807e-05, "loss": 0.2403, "step": 24183 }, { "epoch": 2.5, "learning_rate": 1.7969783497976696e-05, "loss": 0.3995, "step": 24184 }, { "epoch": 2.5, "learning_rate": 1.7967528693218185e-05, "loss": 0.1173, "step": 24185 }, { "epoch": 2.5, "learning_rate": 1.796527396945946e-05, "loss": 0.2774, "step": 24186 }, { "epoch": 2.5, "learning_rate": 1.79630193267157e-05, "loss": 0.1047, "step": 24187 }, { "epoch": 2.5, "learning_rate": 1.7960764765002066e-05, "loss": 0.3058, "step": 24188 }, { "epoch": 2.5, "learning_rate": 1.795851028433375e-05, "loss": 0.34, "step": 24189 }, { "epoch": 2.5, "learning_rate": 1.795625588472593e-05, "loss": 0.5352, "step": 24190 }, { "epoch": 2.5, "learning_rate": 1.795400156619377e-05, "loss": 0.9533, "step": 24191 }, { "epoch": 2.5, "learning_rate": 1.795174732875246e-05, "loss": 0.3402, "step": 24192 }, { "epoch": 2.5, "learning_rate": 1.794949317241716e-05, "loss": 0.5853, "step": 24193 }, { "epoch": 2.5, "learning_rate": 1.7947239097203058e-05, "loss": 0.4887, "step": 24194 }, { "epoch": 2.5, "learning_rate": 1.794498510312532e-05, "loss": 0.0701, "step": 24195 }, { "epoch": 2.5, "learning_rate": 1.794273119019912e-05, "loss": 0.2582, "step": 24196 }, { "epoch": 2.5, "learning_rate": 1.7940477358439633e-05, "loss": 0.4249, "step": 24197 }, { "epoch": 2.5, "learning_rate": 1.7938223607862038e-05, "loss": 0.0675, "step": 24198 }, { "epoch": 2.5, "learning_rate": 1.7935969938481495e-05, "loss": 0.6083, "step": 24199 }, { "epoch": 2.5, "learning_rate": 1.793371635031318e-05, "loss": 0.2504, "step": 24200 }, { "epoch": 2.5, "learning_rate": 1.7931462843372263e-05, "loss": 0.7536, "step": 24201 }, { "epoch": 2.5, "learning_rate": 1.792920941767391e-05, "loss": 0.5884, "step": 24202 }, { "epoch": 2.5, "learning_rate": 1.7926956073233292e-05, "loss": 0.6356, "step": 24203 }, { "epoch": 2.5, "learning_rate": 1.7924702810065597e-05, "loss": 0.2921, "step": 24204 }, { "epoch": 2.5, "learning_rate": 1.7922449628185963e-05, "loss": 0.1297, "step": 24205 }, { "epoch": 2.5, "learning_rate": 1.7920196527609576e-05, "loss": 0.139, "step": 24206 }, { "epoch": 2.5, "learning_rate": 1.7917943508351595e-05, "loss": 0.7135, "step": 24207 }, { "epoch": 2.5, "learning_rate": 1.79156905704272e-05, "loss": 0.3717, "step": 24208 }, { "epoch": 2.5, "learning_rate": 1.7913437713851543e-05, "loss": 0.4189, "step": 24209 }, { "epoch": 2.5, "learning_rate": 1.7911184938639796e-05, "loss": 0.3796, "step": 24210 }, { "epoch": 2.5, "learning_rate": 1.7908932244807123e-05, "loss": 0.5529, "step": 24211 }, { "epoch": 2.5, "learning_rate": 1.79066796323687e-05, "loss": 0.3281, "step": 24212 }, { "epoch": 2.5, "learning_rate": 1.790442710133967e-05, "loss": 0.3176, "step": 24213 }, { "epoch": 2.5, "learning_rate": 1.790217465173521e-05, "loss": 0.08, "step": 24214 }, { "epoch": 2.5, "learning_rate": 1.7899922283570474e-05, "loss": 0.1576, "step": 24215 }, { "epoch": 2.5, "learning_rate": 1.7897669996860633e-05, "loss": 0.3825, "step": 24216 }, { "epoch": 2.5, "learning_rate": 1.789541779162084e-05, "loss": 0.2796, "step": 24217 }, { "epoch": 2.5, "learning_rate": 1.7893165667866273e-05, "loss": 0.5041, "step": 24218 }, { "epoch": 2.5, "learning_rate": 1.7890913625612074e-05, "loss": 0.6371, "step": 24219 }, { "epoch": 2.5, "learning_rate": 1.7888661664873413e-05, "loss": 0.5109, "step": 24220 }, { "epoch": 2.5, "learning_rate": 1.7886409785665448e-05, "loss": 0.3724, "step": 24221 }, { "epoch": 2.5, "learning_rate": 1.7884157988003343e-05, "loss": 0.7979, "step": 24222 }, { "epoch": 2.5, "learning_rate": 1.7881906271902244e-05, "loss": 0.5666, "step": 24223 }, { "epoch": 2.5, "learning_rate": 1.787965463737733e-05, "loss": 0.3843, "step": 24224 }, { "epoch": 2.5, "learning_rate": 1.7877403084443734e-05, "loss": 0.2241, "step": 24225 }, { "epoch": 2.5, "learning_rate": 1.787515161311662e-05, "loss": 0.5116, "step": 24226 }, { "epoch": 2.5, "learning_rate": 1.7872900223411153e-05, "loss": 0.2239, "step": 24227 }, { "epoch": 2.5, "learning_rate": 1.7870648915342488e-05, "loss": 0.4229, "step": 24228 }, { "epoch": 2.5, "learning_rate": 1.786839768892577e-05, "loss": 0.6927, "step": 24229 }, { "epoch": 2.5, "learning_rate": 1.7866146544176162e-05, "loss": 0.4249, "step": 24230 }, { "epoch": 2.5, "learning_rate": 1.7863895481108815e-05, "loss": 0.4849, "step": 24231 }, { "epoch": 2.5, "learning_rate": 1.786164449973889e-05, "loss": 0.4658, "step": 24232 }, { "epoch": 2.5, "learning_rate": 1.785939360008153e-05, "loss": 0.1297, "step": 24233 }, { "epoch": 2.5, "learning_rate": 1.7857142782151887e-05, "loss": 0.2132, "step": 24234 }, { "epoch": 2.5, "learning_rate": 1.7854892045965123e-05, "loss": 0.4249, "step": 24235 }, { "epoch": 2.5, "learning_rate": 1.7852641391536387e-05, "loss": 1.0606, "step": 24236 }, { "epoch": 2.5, "learning_rate": 1.7850390818880832e-05, "loss": 0.2356, "step": 24237 }, { "epoch": 2.5, "learning_rate": 1.7848140328013596e-05, "loss": 0.3107, "step": 24238 }, { "epoch": 2.5, "learning_rate": 1.784588991894983e-05, "loss": 0.3383, "step": 24239 }, { "epoch": 2.5, "learning_rate": 1.7843639591704694e-05, "loss": 0.2931, "step": 24240 }, { "epoch": 2.51, "learning_rate": 1.784138934629333e-05, "loss": 0.2808, "step": 24241 }, { "epoch": 2.51, "learning_rate": 1.7839139182730895e-05, "loss": 0.5046, "step": 24242 }, { "epoch": 2.51, "learning_rate": 1.7836889101032523e-05, "loss": 0.2218, "step": 24243 }, { "epoch": 2.51, "learning_rate": 1.7834639101213366e-05, "loss": 0.8169, "step": 24244 }, { "epoch": 2.51, "learning_rate": 1.783238918328857e-05, "loss": 0.3734, "step": 24245 }, { "epoch": 2.51, "learning_rate": 1.7830139347273294e-05, "loss": 0.2948, "step": 24246 }, { "epoch": 2.51, "learning_rate": 1.7827889593182668e-05, "loss": 0.1294, "step": 24247 }, { "epoch": 2.51, "learning_rate": 1.7825639921031837e-05, "loss": 0.2189, "step": 24248 }, { "epoch": 2.51, "learning_rate": 1.782339033083595e-05, "loss": 0.1864, "step": 24249 }, { "epoch": 2.51, "learning_rate": 1.7821140822610162e-05, "loss": 0.8504, "step": 24250 }, { "epoch": 2.51, "learning_rate": 1.7818891396369595e-05, "loss": 0.3236, "step": 24251 }, { "epoch": 2.51, "learning_rate": 1.7816642052129402e-05, "loss": 0.2502, "step": 24252 }, { "epoch": 2.51, "learning_rate": 1.7814392789904724e-05, "loss": 0.0881, "step": 24253 }, { "epoch": 2.51, "learning_rate": 1.78121436097107e-05, "loss": 0.2406, "step": 24254 }, { "epoch": 2.51, "learning_rate": 1.7809894511562473e-05, "loss": 0.152, "step": 24255 }, { "epoch": 2.51, "learning_rate": 1.780764549547519e-05, "loss": 0.3932, "step": 24256 }, { "epoch": 2.51, "learning_rate": 1.7805396561463978e-05, "loss": 0.473, "step": 24257 }, { "epoch": 2.51, "learning_rate": 1.780314770954398e-05, "loss": 0.2122, "step": 24258 }, { "epoch": 2.51, "learning_rate": 1.7800898939730345e-05, "loss": 0.6849, "step": 24259 }, { "epoch": 2.51, "learning_rate": 1.7798650252038207e-05, "loss": 0.2924, "step": 24260 }, { "epoch": 2.51, "learning_rate": 1.7796401646482692e-05, "loss": 0.2902, "step": 24261 }, { "epoch": 2.51, "learning_rate": 1.7794153123078964e-05, "loss": 0.3554, "step": 24262 }, { "epoch": 2.51, "learning_rate": 1.7791904681842118e-05, "loss": 0.3595, "step": 24263 }, { "epoch": 2.51, "learning_rate": 1.7789656322787325e-05, "loss": 0.2815, "step": 24264 }, { "epoch": 2.51, "learning_rate": 1.77874080459297e-05, "loss": 0.227, "step": 24265 }, { "epoch": 2.51, "learning_rate": 1.77851598512844e-05, "loss": 0.506, "step": 24266 }, { "epoch": 2.51, "learning_rate": 1.7782911738866534e-05, "loss": 0.8065, "step": 24267 }, { "epoch": 2.51, "learning_rate": 1.778066370869125e-05, "loss": 0.1731, "step": 24268 }, { "epoch": 2.51, "learning_rate": 1.777841576077368e-05, "loss": 0.3946, "step": 24269 }, { "epoch": 2.51, "learning_rate": 1.777616789512896e-05, "loss": 0.2798, "step": 24270 }, { "epoch": 2.51, "learning_rate": 1.7773920111772212e-05, "loss": 0.4006, "step": 24271 }, { "epoch": 2.51, "learning_rate": 1.7771672410718573e-05, "loss": 0.4761, "step": 24272 }, { "epoch": 2.51, "learning_rate": 1.776942479198317e-05, "loss": 0.1989, "step": 24273 }, { "epoch": 2.51, "learning_rate": 1.776717725558115e-05, "loss": 0.0673, "step": 24274 }, { "epoch": 2.51, "learning_rate": 1.7764929801527626e-05, "loss": 0.8618, "step": 24275 }, { "epoch": 2.51, "learning_rate": 1.7762682429837734e-05, "loss": 0.6485, "step": 24276 }, { "epoch": 2.51, "learning_rate": 1.7760435140526597e-05, "loss": 0.7824, "step": 24277 }, { "epoch": 2.51, "learning_rate": 1.7758187933609343e-05, "loss": 0.2626, "step": 24278 }, { "epoch": 2.51, "learning_rate": 1.7755940809101108e-05, "loss": 0.4208, "step": 24279 }, { "epoch": 2.51, "learning_rate": 1.7753693767017014e-05, "loss": 0.3119, "step": 24280 }, { "epoch": 2.51, "learning_rate": 1.7751446807372185e-05, "loss": 0.5607, "step": 24281 }, { "epoch": 2.51, "learning_rate": 1.7749199930181754e-05, "loss": 0.6119, "step": 24282 }, { "epoch": 2.51, "learning_rate": 1.7746953135460835e-05, "loss": 0.5665, "step": 24283 }, { "epoch": 2.51, "learning_rate": 1.7744706423224575e-05, "loss": 0.4399, "step": 24284 }, { "epoch": 2.51, "learning_rate": 1.774245979348807e-05, "loss": 0.7941, "step": 24285 }, { "epoch": 2.51, "learning_rate": 1.7740213246266463e-05, "loss": 0.1887, "step": 24286 }, { "epoch": 2.51, "learning_rate": 1.773796678157488e-05, "loss": 0.5665, "step": 24287 }, { "epoch": 2.51, "learning_rate": 1.7735720399428427e-05, "loss": 0.4084, "step": 24288 }, { "epoch": 2.51, "learning_rate": 1.7733474099842234e-05, "loss": 0.5201, "step": 24289 }, { "epoch": 2.51, "learning_rate": 1.7731227882831427e-05, "loss": 0.3314, "step": 24290 }, { "epoch": 2.51, "learning_rate": 1.772898174841112e-05, "loss": 0.3044, "step": 24291 }, { "epoch": 2.51, "learning_rate": 1.7726735696596436e-05, "loss": 0.797, "step": 24292 }, { "epoch": 2.51, "learning_rate": 1.7724489727402496e-05, "loss": 0.3452, "step": 24293 }, { "epoch": 2.51, "learning_rate": 1.772224384084442e-05, "loss": 0.7234, "step": 24294 }, { "epoch": 2.51, "learning_rate": 1.7719998036937326e-05, "loss": 0.3693, "step": 24295 }, { "epoch": 2.51, "learning_rate": 1.771775231569633e-05, "loss": 0.444, "step": 24296 }, { "epoch": 2.51, "learning_rate": 1.771550667713655e-05, "loss": 0.1585, "step": 24297 }, { "epoch": 2.51, "learning_rate": 1.771326112127311e-05, "loss": 0.22, "step": 24298 }, { "epoch": 2.51, "learning_rate": 1.7711015648121122e-05, "loss": 0.5462, "step": 24299 }, { "epoch": 2.51, "learning_rate": 1.7708770257695703e-05, "loss": 0.348, "step": 24300 }, { "epoch": 2.51, "learning_rate": 1.7706524950011963e-05, "loss": 0.4115, "step": 24301 }, { "epoch": 2.51, "learning_rate": 1.7704279725085016e-05, "loss": 0.5368, "step": 24302 }, { "epoch": 2.51, "learning_rate": 1.7702034582929982e-05, "loss": 0.4035, "step": 24303 }, { "epoch": 2.51, "learning_rate": 1.769978952356198e-05, "loss": 0.2521, "step": 24304 }, { "epoch": 2.51, "learning_rate": 1.769754454699611e-05, "loss": 0.2718, "step": 24305 }, { "epoch": 2.51, "learning_rate": 1.7695299653247496e-05, "loss": 0.2066, "step": 24306 }, { "epoch": 2.51, "learning_rate": 1.769305484233124e-05, "loss": 0.5073, "step": 24307 }, { "epoch": 2.51, "learning_rate": 1.769081011426247e-05, "loss": 0.2804, "step": 24308 }, { "epoch": 2.51, "learning_rate": 1.7688565469056278e-05, "loss": 0.3239, "step": 24309 }, { "epoch": 2.51, "learning_rate": 1.7686320906727782e-05, "loss": 0.2996, "step": 24310 }, { "epoch": 2.51, "learning_rate": 1.7684076427292092e-05, "loss": 0.4044, "step": 24311 }, { "epoch": 2.51, "learning_rate": 1.768183203076433e-05, "loss": 0.157, "step": 24312 }, { "epoch": 2.51, "learning_rate": 1.7679587717159585e-05, "loss": 0.5954, "step": 24313 }, { "epoch": 2.51, "learning_rate": 1.7677343486492975e-05, "loss": 0.0369, "step": 24314 }, { "epoch": 2.51, "learning_rate": 1.7675099338779596e-05, "loss": 0.5381, "step": 24315 }, { "epoch": 2.51, "learning_rate": 1.7672855274034567e-05, "loss": 0.2529, "step": 24316 }, { "epoch": 2.51, "learning_rate": 1.7670611292272994e-05, "loss": 0.1041, "step": 24317 }, { "epoch": 2.51, "learning_rate": 1.7668367393509987e-05, "loss": 0.6646, "step": 24318 }, { "epoch": 2.51, "learning_rate": 1.7666123577760638e-05, "loss": 0.7109, "step": 24319 }, { "epoch": 2.51, "learning_rate": 1.766387984504006e-05, "loss": 0.4835, "step": 24320 }, { "epoch": 2.51, "learning_rate": 1.766163619536335e-05, "loss": 0.3183, "step": 24321 }, { "epoch": 2.51, "learning_rate": 1.7659392628745632e-05, "loss": 0.4239, "step": 24322 }, { "epoch": 2.51, "learning_rate": 1.765714914520199e-05, "loss": 0.3283, "step": 24323 }, { "epoch": 2.51, "learning_rate": 1.765490574474753e-05, "loss": 0.0851, "step": 24324 }, { "epoch": 2.51, "learning_rate": 1.7652662427397362e-05, "loss": 0.1413, "step": 24325 }, { "epoch": 2.51, "learning_rate": 1.7650419193166577e-05, "loss": 0.1858, "step": 24326 }, { "epoch": 2.51, "learning_rate": 1.7648176042070276e-05, "loss": 0.181, "step": 24327 }, { "epoch": 2.51, "learning_rate": 1.7645932974123574e-05, "loss": 0.7444, "step": 24328 }, { "epoch": 2.51, "learning_rate": 1.7643689989341552e-05, "loss": 0.3888, "step": 24329 }, { "epoch": 2.51, "learning_rate": 1.764144708773932e-05, "loss": 0.3817, "step": 24330 }, { "epoch": 2.51, "learning_rate": 1.763920426933197e-05, "loss": 0.2021, "step": 24331 }, { "epoch": 2.51, "learning_rate": 1.7636961534134617e-05, "loss": 0.342, "step": 24332 }, { "epoch": 2.51, "learning_rate": 1.7634718882162336e-05, "loss": 0.0864, "step": 24333 }, { "epoch": 2.51, "learning_rate": 1.7632476313430236e-05, "loss": 0.2114, "step": 24334 }, { "epoch": 2.51, "learning_rate": 1.763023382795341e-05, "loss": 1.086, "step": 24335 }, { "epoch": 2.51, "learning_rate": 1.7627991425746962e-05, "loss": 0.463, "step": 24336 }, { "epoch": 2.52, "learning_rate": 1.762574910682598e-05, "loss": 0.3778, "step": 24337 }, { "epoch": 2.52, "learning_rate": 1.7623506871205558e-05, "loss": 0.1928, "step": 24338 }, { "epoch": 2.52, "learning_rate": 1.762126471890079e-05, "loss": 0.9352, "step": 24339 }, { "epoch": 2.52, "learning_rate": 1.7619022649926773e-05, "loss": 0.6667, "step": 24340 }, { "epoch": 2.52, "learning_rate": 1.7616780664298595e-05, "loss": 0.2348, "step": 24341 }, { "epoch": 2.52, "learning_rate": 1.761453876203136e-05, "loss": 0.7194, "step": 24342 }, { "epoch": 2.52, "learning_rate": 1.761229694314014e-05, "loss": 0.4131, "step": 24343 }, { "epoch": 2.52, "learning_rate": 1.7610055207640047e-05, "loss": 0.2823, "step": 24344 }, { "epoch": 2.52, "learning_rate": 1.7607813555546162e-05, "loss": 0.3235, "step": 24345 }, { "epoch": 2.52, "learning_rate": 1.7605571986873578e-05, "loss": 0.231, "step": 24346 }, { "epoch": 2.52, "learning_rate": 1.7603330501637377e-05, "loss": 0.8958, "step": 24347 }, { "epoch": 2.52, "learning_rate": 1.760108909985266e-05, "loss": 1.0587, "step": 24348 }, { "epoch": 2.52, "learning_rate": 1.759884778153451e-05, "loss": 0.8297, "step": 24349 }, { "epoch": 2.52, "learning_rate": 1.7596606546698022e-05, "loss": 0.176, "step": 24350 }, { "epoch": 2.52, "learning_rate": 1.7594365395358266e-05, "loss": 0.5505, "step": 24351 }, { "epoch": 2.52, "learning_rate": 1.7592124327530347e-05, "loss": 0.5647, "step": 24352 }, { "epoch": 2.52, "learning_rate": 1.7589883343229337e-05, "loss": 0.3398, "step": 24353 }, { "epoch": 2.52, "learning_rate": 1.7587642442470332e-05, "loss": 0.2146, "step": 24354 }, { "epoch": 2.52, "learning_rate": 1.7585401625268407e-05, "loss": 0.2902, "step": 24355 }, { "epoch": 2.52, "learning_rate": 1.7583160891638663e-05, "loss": 0.0616, "step": 24356 }, { "epoch": 2.52, "learning_rate": 1.758092024159617e-05, "loss": 0.2681, "step": 24357 }, { "epoch": 2.52, "learning_rate": 1.7578679675156014e-05, "loss": 0.3252, "step": 24358 }, { "epoch": 2.52, "learning_rate": 1.757643919233328e-05, "loss": 0.0765, "step": 24359 }, { "epoch": 2.52, "learning_rate": 1.7574198793143057e-05, "loss": 0.3547, "step": 24360 }, { "epoch": 2.52, "learning_rate": 1.7571958477600417e-05, "loss": 0.4667, "step": 24361 }, { "epoch": 2.52, "learning_rate": 1.7569718245720444e-05, "loss": 0.1989, "step": 24362 }, { "epoch": 2.52, "learning_rate": 1.7567478097518226e-05, "loss": 0.5622, "step": 24363 }, { "epoch": 2.52, "learning_rate": 1.756523803300883e-05, "loss": 0.3243, "step": 24364 }, { "epoch": 2.52, "learning_rate": 1.7562998052207342e-05, "loss": 0.3294, "step": 24365 }, { "epoch": 2.52, "learning_rate": 1.756075815512884e-05, "loss": 0.152, "step": 24366 }, { "epoch": 2.52, "learning_rate": 1.7558518341788406e-05, "loss": 0.3616, "step": 24367 }, { "epoch": 2.52, "learning_rate": 1.7556278612201113e-05, "loss": 0.2461, "step": 24368 }, { "epoch": 2.52, "learning_rate": 1.755403896638204e-05, "loss": 0.206, "step": 24369 }, { "epoch": 2.52, "learning_rate": 1.7551799404346272e-05, "loss": 0.3871, "step": 24370 }, { "epoch": 2.52, "learning_rate": 1.7549559926108867e-05, "loss": 0.2975, "step": 24371 }, { "epoch": 2.52, "learning_rate": 1.7547320531684914e-05, "loss": 0.6679, "step": 24372 }, { "epoch": 2.52, "learning_rate": 1.7545081221089488e-05, "loss": 0.2593, "step": 24373 }, { "epoch": 2.52, "learning_rate": 1.7542841994337665e-05, "loss": 0.6239, "step": 24374 }, { "epoch": 2.52, "learning_rate": 1.7540602851444513e-05, "loss": 0.2191, "step": 24375 }, { "epoch": 2.52, "learning_rate": 1.7538363792425113e-05, "loss": 0.676, "step": 24376 }, { "epoch": 2.52, "learning_rate": 1.753612481729452e-05, "loss": 0.4052, "step": 24377 }, { "epoch": 2.52, "learning_rate": 1.753388592606782e-05, "loss": 0.6057, "step": 24378 }, { "epoch": 2.52, "learning_rate": 1.7531647118760086e-05, "loss": 0.5192, "step": 24379 }, { "epoch": 2.52, "learning_rate": 1.752940839538639e-05, "loss": 0.0901, "step": 24380 }, { "epoch": 2.52, "learning_rate": 1.7527169755961797e-05, "loss": 0.311, "step": 24381 }, { "epoch": 2.52, "learning_rate": 1.752493120050137e-05, "loss": 0.3389, "step": 24382 }, { "epoch": 2.52, "learning_rate": 1.7522692729020196e-05, "loss": 0.1131, "step": 24383 }, { "epoch": 2.52, "learning_rate": 1.7520454341533337e-05, "loss": 0.238, "step": 24384 }, { "epoch": 2.52, "learning_rate": 1.7518216038055854e-05, "loss": 0.2424, "step": 24385 }, { "epoch": 2.52, "learning_rate": 1.751597781860282e-05, "loss": 0.1525, "step": 24386 }, { "epoch": 2.52, "learning_rate": 1.7513739683189306e-05, "loss": 0.4029, "step": 24387 }, { "epoch": 2.52, "learning_rate": 1.7511501631830386e-05, "loss": 0.793, "step": 24388 }, { "epoch": 2.52, "learning_rate": 1.75092636645411e-05, "loss": 0.1029, "step": 24389 }, { "epoch": 2.52, "learning_rate": 1.7507025781336525e-05, "loss": 0.5599, "step": 24390 }, { "epoch": 2.52, "learning_rate": 1.7504787982231735e-05, "loss": 0.5614, "step": 24391 }, { "epoch": 2.52, "learning_rate": 1.7502550267241782e-05, "loss": 0.6137, "step": 24392 }, { "epoch": 2.52, "learning_rate": 1.7500312636381744e-05, "loss": 0.8259, "step": 24393 }, { "epoch": 2.52, "learning_rate": 1.7498075089666673e-05, "loss": 0.5317, "step": 24394 }, { "epoch": 2.52, "learning_rate": 1.7495837627111638e-05, "loss": 0.5002, "step": 24395 }, { "epoch": 2.52, "learning_rate": 1.7493600248731704e-05, "loss": 0.2481, "step": 24396 }, { "epoch": 2.52, "learning_rate": 1.7491362954541922e-05, "loss": 0.458, "step": 24397 }, { "epoch": 2.52, "learning_rate": 1.7489125744557353e-05, "loss": 0.7635, "step": 24398 }, { "epoch": 2.52, "learning_rate": 1.7486888618793066e-05, "loss": 0.4726, "step": 24399 }, { "epoch": 2.52, "learning_rate": 1.7484651577264122e-05, "loss": 0.1375, "step": 24400 }, { "epoch": 2.52, "learning_rate": 1.748241461998557e-05, "loss": 0.3497, "step": 24401 }, { "epoch": 2.52, "learning_rate": 1.7480177746972474e-05, "loss": 0.8942, "step": 24402 }, { "epoch": 2.52, "learning_rate": 1.747794095823989e-05, "loss": 0.3623, "step": 24403 }, { "epoch": 2.52, "learning_rate": 1.7475704253802876e-05, "loss": 0.8719, "step": 24404 }, { "epoch": 2.52, "learning_rate": 1.7473467633676492e-05, "loss": 0.7322, "step": 24405 }, { "epoch": 2.52, "learning_rate": 1.7471231097875807e-05, "loss": 0.1241, "step": 24406 }, { "epoch": 2.52, "learning_rate": 1.746899464641585e-05, "loss": 0.0967, "step": 24407 }, { "epoch": 2.52, "learning_rate": 1.7466758279311688e-05, "loss": 0.4264, "step": 24408 }, { "epoch": 2.52, "learning_rate": 1.7464521996578375e-05, "loss": 0.6259, "step": 24409 }, { "epoch": 2.52, "learning_rate": 1.746228579823097e-05, "loss": 0.2981, "step": 24410 }, { "epoch": 2.52, "learning_rate": 1.746004968428452e-05, "loss": 0.4257, "step": 24411 }, { "epoch": 2.52, "learning_rate": 1.7457813654754083e-05, "loss": 0.229, "step": 24412 }, { "epoch": 2.52, "learning_rate": 1.7455577709654723e-05, "loss": 0.8059, "step": 24413 }, { "epoch": 2.52, "learning_rate": 1.7453341849001462e-05, "loss": 0.2125, "step": 24414 }, { "epoch": 2.52, "learning_rate": 1.7451106072809373e-05, "loss": 0.2494, "step": 24415 }, { "epoch": 2.52, "learning_rate": 1.7448870381093516e-05, "loss": 0.4726, "step": 24416 }, { "epoch": 2.52, "learning_rate": 1.744663477386891e-05, "loss": 0.2914, "step": 24417 }, { "epoch": 2.52, "learning_rate": 1.744439925115062e-05, "loss": 0.2001, "step": 24418 }, { "epoch": 2.52, "learning_rate": 1.74421638129537e-05, "loss": 0.3177, "step": 24419 }, { "epoch": 2.52, "learning_rate": 1.7439928459293193e-05, "loss": 0.5468, "step": 24420 }, { "epoch": 2.52, "learning_rate": 1.7437693190184153e-05, "loss": 0.1888, "step": 24421 }, { "epoch": 2.52, "learning_rate": 1.7435458005641617e-05, "loss": 0.6241, "step": 24422 }, { "epoch": 2.52, "learning_rate": 1.7433222905680642e-05, "loss": 0.1321, "step": 24423 }, { "epoch": 2.52, "learning_rate": 1.7430987890316284e-05, "loss": 0.5054, "step": 24424 }, { "epoch": 2.52, "learning_rate": 1.7428752959563558e-05, "loss": 0.1822, "step": 24425 }, { "epoch": 2.52, "learning_rate": 1.7426518113437538e-05, "loss": 0.4255, "step": 24426 }, { "epoch": 2.52, "learning_rate": 1.7424283351953247e-05, "loss": 0.217, "step": 24427 }, { "epoch": 2.52, "learning_rate": 1.7422048675125735e-05, "loss": 0.1771, "step": 24428 }, { "epoch": 2.52, "learning_rate": 1.741981408297005e-05, "loss": 0.6849, "step": 24429 }, { "epoch": 2.52, "learning_rate": 1.7417579575501232e-05, "loss": 0.0424, "step": 24430 }, { "epoch": 2.52, "learning_rate": 1.7415345152734324e-05, "loss": 0.2583, "step": 24431 }, { "epoch": 2.53, "learning_rate": 1.741311081468437e-05, "loss": 0.3792, "step": 24432 }, { "epoch": 2.53, "learning_rate": 1.7410876561366407e-05, "loss": 0.5171, "step": 24433 }, { "epoch": 2.53, "learning_rate": 1.740864239279549e-05, "loss": 0.3155, "step": 24434 }, { "epoch": 2.53, "learning_rate": 1.7406408308986633e-05, "loss": 0.3376, "step": 24435 }, { "epoch": 2.53, "learning_rate": 1.7404174309954892e-05, "loss": 0.0885, "step": 24436 }, { "epoch": 2.53, "learning_rate": 1.7401940395715297e-05, "loss": 0.1946, "step": 24437 }, { "epoch": 2.53, "learning_rate": 1.7399706566282906e-05, "loss": 0.6304, "step": 24438 }, { "epoch": 2.53, "learning_rate": 1.739747282167273e-05, "loss": 0.5033, "step": 24439 }, { "epoch": 2.53, "learning_rate": 1.739523916189982e-05, "loss": 0.1874, "step": 24440 }, { "epoch": 2.53, "learning_rate": 1.739300558697921e-05, "loss": 0.6411, "step": 24441 }, { "epoch": 2.53, "learning_rate": 1.739077209692594e-05, "loss": 0.0509, "step": 24442 }, { "epoch": 2.53, "learning_rate": 1.7388538691755035e-05, "loss": 0.3274, "step": 24443 }, { "epoch": 2.53, "learning_rate": 1.7386305371481553e-05, "loss": 0.42, "step": 24444 }, { "epoch": 2.53, "learning_rate": 1.7384072136120497e-05, "loss": 0.3044, "step": 24445 }, { "epoch": 2.53, "learning_rate": 1.738183898568692e-05, "loss": 0.5792, "step": 24446 }, { "epoch": 2.53, "learning_rate": 1.7379605920195847e-05, "loss": 0.2485, "step": 24447 }, { "epoch": 2.53, "learning_rate": 1.7377372939662313e-05, "loss": 0.4927, "step": 24448 }, { "epoch": 2.53, "learning_rate": 1.7375140044101355e-05, "loss": 0.1329, "step": 24449 }, { "epoch": 2.53, "learning_rate": 1.7372907233527993e-05, "loss": 0.3942, "step": 24450 }, { "epoch": 2.53, "learning_rate": 1.7370674507957288e-05, "loss": 0.6532, "step": 24451 }, { "epoch": 2.53, "learning_rate": 1.7368441867404225e-05, "loss": 0.1503, "step": 24452 }, { "epoch": 2.53, "learning_rate": 1.736620931188386e-05, "loss": 0.6313, "step": 24453 }, { "epoch": 2.53, "learning_rate": 1.7363976841411233e-05, "loss": 0.6595, "step": 24454 }, { "epoch": 2.53, "learning_rate": 1.7361744456001343e-05, "loss": 0.2925, "step": 24455 }, { "epoch": 2.53, "learning_rate": 1.7359512155669228e-05, "loss": 0.4484, "step": 24456 }, { "epoch": 2.53, "learning_rate": 1.7357279940429924e-05, "loss": 0.2675, "step": 24457 }, { "epoch": 2.53, "learning_rate": 1.7355047810298454e-05, "loss": 0.6314, "step": 24458 }, { "epoch": 2.53, "learning_rate": 1.735281576528984e-05, "loss": 0.7506, "step": 24459 }, { "epoch": 2.53, "learning_rate": 1.7350583805419113e-05, "loss": 0.1581, "step": 24460 }, { "epoch": 2.53, "learning_rate": 1.7348351930701297e-05, "loss": 0.1067, "step": 24461 }, { "epoch": 2.53, "learning_rate": 1.734612014115142e-05, "loss": 1.0164, "step": 24462 }, { "epoch": 2.53, "learning_rate": 1.73438884367845e-05, "loss": 0.4526, "step": 24463 }, { "epoch": 2.53, "learning_rate": 1.734165681761557e-05, "loss": 0.4163, "step": 24464 }, { "epoch": 2.53, "learning_rate": 1.7339425283659632e-05, "loss": 0.157, "step": 24465 }, { "epoch": 2.53, "learning_rate": 1.733719383493172e-05, "loss": 0.2345, "step": 24466 }, { "epoch": 2.53, "learning_rate": 1.7334962471446858e-05, "loss": 0.1973, "step": 24467 }, { "epoch": 2.53, "learning_rate": 1.7332731193220063e-05, "loss": 0.306, "step": 24468 }, { "epoch": 2.53, "learning_rate": 1.733050000026636e-05, "loss": 0.4897, "step": 24469 }, { "epoch": 2.53, "learning_rate": 1.7328268892600762e-05, "loss": 0.7495, "step": 24470 }, { "epoch": 2.53, "learning_rate": 1.73260378702383e-05, "loss": 0.3665, "step": 24471 }, { "epoch": 2.53, "learning_rate": 1.7323806933193994e-05, "loss": 0.5945, "step": 24472 }, { "epoch": 2.53, "learning_rate": 1.7321576081482843e-05, "loss": 0.5311, "step": 24473 }, { "epoch": 2.53, "learning_rate": 1.731934531511987e-05, "loss": 0.0775, "step": 24474 }, { "epoch": 2.53, "learning_rate": 1.7317114634120102e-05, "loss": 0.8675, "step": 24475 }, { "epoch": 2.53, "learning_rate": 1.7314884038498564e-05, "loss": 0.2382, "step": 24476 }, { "epoch": 2.53, "learning_rate": 1.7312653528270247e-05, "loss": 0.1085, "step": 24477 }, { "epoch": 2.53, "learning_rate": 1.7310423103450173e-05, "loss": 0.2493, "step": 24478 }, { "epoch": 2.53, "learning_rate": 1.730819276405336e-05, "loss": 0.3458, "step": 24479 }, { "epoch": 2.53, "learning_rate": 1.730596251009483e-05, "loss": 0.2057, "step": 24480 }, { "epoch": 2.53, "learning_rate": 1.7303732341589586e-05, "loss": 0.1974, "step": 24481 }, { "epoch": 2.53, "learning_rate": 1.7301502258552663e-05, "loss": 0.0256, "step": 24482 }, { "epoch": 2.53, "learning_rate": 1.729927226099904e-05, "loss": 0.3166, "step": 24483 }, { "epoch": 2.53, "learning_rate": 1.729704234894374e-05, "loss": 0.1519, "step": 24484 }, { "epoch": 2.53, "learning_rate": 1.7294812522401782e-05, "loss": 0.3298, "step": 24485 }, { "epoch": 2.53, "learning_rate": 1.7292582781388177e-05, "loss": 1.126, "step": 24486 }, { "epoch": 2.53, "learning_rate": 1.7290353125917927e-05, "loss": 0.1528, "step": 24487 }, { "epoch": 2.53, "learning_rate": 1.728812355600606e-05, "loss": 0.4921, "step": 24488 }, { "epoch": 2.53, "learning_rate": 1.7285894071667556e-05, "loss": 0.4526, "step": 24489 }, { "epoch": 2.53, "learning_rate": 1.728366467291744e-05, "loss": 0.5885, "step": 24490 }, { "epoch": 2.53, "learning_rate": 1.728143535977072e-05, "loss": 0.251, "step": 24491 }, { "epoch": 2.53, "learning_rate": 1.727920613224241e-05, "loss": 0.489, "step": 24492 }, { "epoch": 2.53, "learning_rate": 1.72769769903475e-05, "loss": 0.5484, "step": 24493 }, { "epoch": 2.53, "learning_rate": 1.7274747934101003e-05, "loss": 0.717, "step": 24494 }, { "epoch": 2.53, "learning_rate": 1.727251896351792e-05, "loss": 0.23, "step": 24495 }, { "epoch": 2.53, "learning_rate": 1.7270290078613266e-05, "loss": 0.2104, "step": 24496 }, { "epoch": 2.53, "learning_rate": 1.7268061279402042e-05, "loss": 0.1543, "step": 24497 }, { "epoch": 2.53, "learning_rate": 1.7265832565899247e-05, "loss": 0.2433, "step": 24498 }, { "epoch": 2.53, "learning_rate": 1.7263603938119892e-05, "loss": 0.4323, "step": 24499 }, { "epoch": 2.53, "learning_rate": 1.726137539607898e-05, "loss": 0.358, "step": 24500 }, { "epoch": 2.53, "learning_rate": 1.7259146939791503e-05, "loss": 0.4416, "step": 24501 }, { "epoch": 2.53, "learning_rate": 1.7256918569272477e-05, "loss": 0.1169, "step": 24502 }, { "epoch": 2.53, "learning_rate": 1.7254690284536882e-05, "loss": 0.3684, "step": 24503 }, { "epoch": 2.53, "learning_rate": 1.725246208559973e-05, "loss": 0.5301, "step": 24504 }, { "epoch": 2.53, "learning_rate": 1.725023397247602e-05, "loss": 0.5514, "step": 24505 }, { "epoch": 2.53, "learning_rate": 1.7248005945180754e-05, "loss": 0.2026, "step": 24506 }, { "epoch": 2.53, "learning_rate": 1.7245778003728924e-05, "loss": 0.431, "step": 24507 }, { "epoch": 2.53, "learning_rate": 1.7243550148135534e-05, "loss": 0.2578, "step": 24508 }, { "epoch": 2.53, "learning_rate": 1.7241322378415578e-05, "loss": 0.1804, "step": 24509 }, { "epoch": 2.53, "learning_rate": 1.723909469458407e-05, "loss": 0.4714, "step": 24510 }, { "epoch": 2.53, "learning_rate": 1.7236867096655977e-05, "loss": 0.2396, "step": 24511 }, { "epoch": 2.53, "learning_rate": 1.723463958464631e-05, "loss": 0.2139, "step": 24512 }, { "epoch": 2.53, "learning_rate": 1.723241215857007e-05, "loss": 0.6026, "step": 24513 }, { "epoch": 2.53, "learning_rate": 1.7230184818442237e-05, "loss": 0.2074, "step": 24514 }, { "epoch": 2.53, "learning_rate": 1.722795756427781e-05, "loss": 0.296, "step": 24515 }, { "epoch": 2.53, "learning_rate": 1.7225730396091782e-05, "loss": 0.4467, "step": 24516 }, { "epoch": 2.53, "learning_rate": 1.7223503313899152e-05, "loss": 0.4821, "step": 24517 }, { "epoch": 2.53, "learning_rate": 1.7221276317714904e-05, "loss": 0.816, "step": 24518 }, { "epoch": 2.53, "learning_rate": 1.7219049407554036e-05, "loss": 0.0262, "step": 24519 }, { "epoch": 2.53, "learning_rate": 1.721682258343155e-05, "loss": 0.1286, "step": 24520 }, { "epoch": 2.53, "learning_rate": 1.7214595845362414e-05, "loss": 0.7071, "step": 24521 }, { "epoch": 2.53, "learning_rate": 1.721236919336162e-05, "loss": 0.416, "step": 24522 }, { "epoch": 2.53, "learning_rate": 1.721014262744417e-05, "loss": 0.1808, "step": 24523 }, { "epoch": 2.53, "learning_rate": 1.7207916147625045e-05, "loss": 0.6032, "step": 24524 }, { "epoch": 2.53, "learning_rate": 1.7205689753919236e-05, "loss": 0.2198, "step": 24525 }, { "epoch": 2.53, "learning_rate": 1.720346344634174e-05, "loss": 0.2773, "step": 24526 }, { "epoch": 2.53, "learning_rate": 1.7201237224907524e-05, "loss": 0.2961, "step": 24527 }, { "epoch": 2.54, "learning_rate": 1.7199011089631586e-05, "loss": 0.103, "step": 24528 }, { "epoch": 2.54, "learning_rate": 1.7196785040528904e-05, "loss": 0.4548, "step": 24529 }, { "epoch": 2.54, "learning_rate": 1.719455907761449e-05, "loss": 0.2109, "step": 24530 }, { "epoch": 2.54, "learning_rate": 1.7192333200903288e-05, "loss": 0.6461, "step": 24531 }, { "epoch": 2.54, "learning_rate": 1.7190107410410305e-05, "loss": 0.5518, "step": 24532 }, { "epoch": 2.54, "learning_rate": 1.7187881706150524e-05, "loss": 0.7138, "step": 24533 }, { "epoch": 2.54, "learning_rate": 1.7185656088138925e-05, "loss": 0.2686, "step": 24534 }, { "epoch": 2.54, "learning_rate": 1.7183430556390486e-05, "loss": 0.2563, "step": 24535 }, { "epoch": 2.54, "learning_rate": 1.7181205110920202e-05, "loss": 0.4282, "step": 24536 }, { "epoch": 2.54, "learning_rate": 1.717897975174304e-05, "loss": 0.2814, "step": 24537 }, { "epoch": 2.54, "learning_rate": 1.7176754478873996e-05, "loss": 0.2878, "step": 24538 }, { "epoch": 2.54, "learning_rate": 1.7174529292328033e-05, "loss": 0.4241, "step": 24539 }, { "epoch": 2.54, "learning_rate": 1.717230419212015e-05, "loss": 0.1773, "step": 24540 }, { "epoch": 2.54, "learning_rate": 1.7170079178265303e-05, "loss": 0.2638, "step": 24541 }, { "epoch": 2.54, "learning_rate": 1.7167854250778476e-05, "loss": 0.4848, "step": 24542 }, { "epoch": 2.54, "learning_rate": 1.7165629409674654e-05, "loss": 0.4122, "step": 24543 }, { "epoch": 2.54, "learning_rate": 1.7163404654968816e-05, "loss": 0.449, "step": 24544 }, { "epoch": 2.54, "learning_rate": 1.7161179986675927e-05, "loss": 0.2581, "step": 24545 }, { "epoch": 2.54, "learning_rate": 1.7158955404810974e-05, "loss": 0.2466, "step": 24546 }, { "epoch": 2.54, "learning_rate": 1.7156730909388928e-05, "loss": 0.6673, "step": 24547 }, { "epoch": 2.54, "learning_rate": 1.7154506500424778e-05, "loss": 0.4655, "step": 24548 }, { "epoch": 2.54, "learning_rate": 1.7152282177933473e-05, "loss": 0.7427, "step": 24549 }, { "epoch": 2.54, "learning_rate": 1.7150057941929998e-05, "loss": 0.7976, "step": 24550 }, { "epoch": 2.54, "learning_rate": 1.7147833792429337e-05, "loss": 0.6623, "step": 24551 }, { "epoch": 2.54, "learning_rate": 1.714560972944644e-05, "loss": 0.3724, "step": 24552 }, { "epoch": 2.54, "learning_rate": 1.714338575299629e-05, "loss": 0.2635, "step": 24553 }, { "epoch": 2.54, "learning_rate": 1.714116186309386e-05, "loss": 0.3433, "step": 24554 }, { "epoch": 2.54, "learning_rate": 1.713893805975412e-05, "loss": 0.2034, "step": 24555 }, { "epoch": 2.54, "learning_rate": 1.7136714342992036e-05, "loss": 0.0845, "step": 24556 }, { "epoch": 2.54, "learning_rate": 1.7134490712822585e-05, "loss": 0.1416, "step": 24557 }, { "epoch": 2.54, "learning_rate": 1.7132267169260742e-05, "loss": 0.2117, "step": 24558 }, { "epoch": 2.54, "learning_rate": 1.7130043712321455e-05, "loss": 0.608, "step": 24559 }, { "epoch": 2.54, "learning_rate": 1.7127820342019704e-05, "loss": 0.2831, "step": 24560 }, { "epoch": 2.54, "learning_rate": 1.712559705837045e-05, "loss": 0.1438, "step": 24561 }, { "epoch": 2.54, "learning_rate": 1.7123373861388667e-05, "loss": 0.4364, "step": 24562 }, { "epoch": 2.54, "learning_rate": 1.712115075108932e-05, "loss": 0.803, "step": 24563 }, { "epoch": 2.54, "learning_rate": 1.7118927727487386e-05, "loss": 0.3726, "step": 24564 }, { "epoch": 2.54, "learning_rate": 1.71167047905978e-05, "loss": 0.3107, "step": 24565 }, { "epoch": 2.54, "learning_rate": 1.7114481940435546e-05, "loss": 0.6657, "step": 24566 }, { "epoch": 2.54, "learning_rate": 1.7112259177015585e-05, "loss": 0.2221, "step": 24567 }, { "epoch": 2.54, "learning_rate": 1.7110036500352888e-05, "loss": 0.6433, "step": 24568 }, { "epoch": 2.54, "learning_rate": 1.7107813910462398e-05, "loss": 0.5198, "step": 24569 }, { "epoch": 2.54, "learning_rate": 1.7105591407359086e-05, "loss": 0.2598, "step": 24570 }, { "epoch": 2.54, "learning_rate": 1.7103368991057925e-05, "loss": 0.3491, "step": 24571 }, { "epoch": 2.54, "learning_rate": 1.7101146661573857e-05, "loss": 0.2808, "step": 24572 }, { "epoch": 2.54, "learning_rate": 1.7098924418921857e-05, "loss": 0.1744, "step": 24573 }, { "epoch": 2.54, "learning_rate": 1.7096702263116874e-05, "loss": 0.5696, "step": 24574 }, { "epoch": 2.54, "learning_rate": 1.7094480194173876e-05, "loss": 0.3171, "step": 24575 }, { "epoch": 2.54, "learning_rate": 1.7092258212107832e-05, "loss": 0.6715, "step": 24576 }, { "epoch": 2.54, "learning_rate": 1.709003631693367e-05, "loss": 0.1615, "step": 24577 }, { "epoch": 2.54, "learning_rate": 1.7087814508666376e-05, "loss": 0.2114, "step": 24578 }, { "epoch": 2.54, "learning_rate": 1.708559278732088e-05, "loss": 0.1187, "step": 24579 }, { "epoch": 2.54, "learning_rate": 1.7083371152912155e-05, "loss": 0.1401, "step": 24580 }, { "epoch": 2.54, "learning_rate": 1.7081149605455148e-05, "loss": 0.7443, "step": 24581 }, { "epoch": 2.54, "learning_rate": 1.7078928144964824e-05, "loss": 0.1248, "step": 24582 }, { "epoch": 2.54, "learning_rate": 1.707670677145613e-05, "loss": 0.0829, "step": 24583 }, { "epoch": 2.54, "learning_rate": 1.707448548494402e-05, "loss": 0.3887, "step": 24584 }, { "epoch": 2.54, "learning_rate": 1.7072264285443456e-05, "loss": 0.257, "step": 24585 }, { "epoch": 2.54, "learning_rate": 1.7070043172969388e-05, "loss": 0.2237, "step": 24586 }, { "epoch": 2.54, "learning_rate": 1.7067822147536757e-05, "loss": 0.3624, "step": 24587 }, { "epoch": 2.54, "learning_rate": 1.7065601209160523e-05, "loss": 0.2311, "step": 24588 }, { "epoch": 2.54, "learning_rate": 1.706338035785564e-05, "loss": 0.1418, "step": 24589 }, { "epoch": 2.54, "learning_rate": 1.7061159593637047e-05, "loss": 0.748, "step": 24590 }, { "epoch": 2.54, "learning_rate": 1.70589389165197e-05, "loss": 0.4884, "step": 24591 }, { "epoch": 2.54, "learning_rate": 1.7056718326518548e-05, "loss": 0.9527, "step": 24592 }, { "epoch": 2.54, "learning_rate": 1.705449782364854e-05, "loss": 0.7035, "step": 24593 }, { "epoch": 2.54, "learning_rate": 1.7052277407924623e-05, "loss": 0.7749, "step": 24594 }, { "epoch": 2.54, "learning_rate": 1.7050057079361746e-05, "loss": 0.4125, "step": 24595 }, { "epoch": 2.54, "learning_rate": 1.7047836837974863e-05, "loss": 0.8339, "step": 24596 }, { "epoch": 2.54, "learning_rate": 1.7045616683778906e-05, "loss": 0.1385, "step": 24597 }, { "epoch": 2.54, "learning_rate": 1.704339661678882e-05, "loss": 0.5444, "step": 24598 }, { "epoch": 2.54, "learning_rate": 1.704117663701956e-05, "loss": 0.1997, "step": 24599 }, { "epoch": 2.54, "learning_rate": 1.7038956744486064e-05, "loss": 0.2374, "step": 24600 }, { "epoch": 2.54, "learning_rate": 1.7036736939203292e-05, "loss": 0.16, "step": 24601 }, { "epoch": 2.54, "learning_rate": 1.7034517221186165e-05, "loss": 0.5382, "step": 24602 }, { "epoch": 2.54, "learning_rate": 1.703229759044963e-05, "loss": 0.4632, "step": 24603 }, { "epoch": 2.54, "learning_rate": 1.7030078047008632e-05, "loss": 0.3906, "step": 24604 }, { "epoch": 2.54, "learning_rate": 1.7027858590878117e-05, "loss": 0.0772, "step": 24605 }, { "epoch": 2.54, "learning_rate": 1.7025639222073037e-05, "loss": 0.1413, "step": 24606 }, { "epoch": 2.54, "learning_rate": 1.7023419940608305e-05, "loss": 0.3471, "step": 24607 }, { "epoch": 2.54, "learning_rate": 1.7021200746498873e-05, "loss": 0.2481, "step": 24608 }, { "epoch": 2.54, "learning_rate": 1.7018981639759678e-05, "loss": 0.5805, "step": 24609 }, { "epoch": 2.54, "learning_rate": 1.7016762620405663e-05, "loss": 0.4693, "step": 24610 }, { "epoch": 2.54, "learning_rate": 1.7014543688451766e-05, "loss": 0.3904, "step": 24611 }, { "epoch": 2.54, "learning_rate": 1.701232484391292e-05, "loss": 0.2219, "step": 24612 }, { "epoch": 2.54, "learning_rate": 1.7010106086804067e-05, "loss": 0.2028, "step": 24613 }, { "epoch": 2.54, "learning_rate": 1.700788741714015e-05, "loss": 0.3253, "step": 24614 }, { "epoch": 2.54, "learning_rate": 1.7005668834936085e-05, "loss": 0.393, "step": 24615 }, { "epoch": 2.54, "learning_rate": 1.7003450340206827e-05, "loss": 0.3175, "step": 24616 }, { "epoch": 2.54, "learning_rate": 1.700123193296729e-05, "loss": 0.1273, "step": 24617 }, { "epoch": 2.54, "learning_rate": 1.6999013613232416e-05, "loss": 0.5305, "step": 24618 }, { "epoch": 2.54, "learning_rate": 1.6996795381017144e-05, "loss": 0.145, "step": 24619 }, { "epoch": 2.54, "learning_rate": 1.6994577236336397e-05, "loss": 1.0616, "step": 24620 }, { "epoch": 2.54, "learning_rate": 1.699235917920512e-05, "loss": 0.0699, "step": 24621 }, { "epoch": 2.54, "learning_rate": 1.699014120963823e-05, "loss": 0.2446, "step": 24622 }, { "epoch": 2.54, "learning_rate": 1.698792332765067e-05, "loss": 0.7486, "step": 24623 }, { "epoch": 2.55, "learning_rate": 1.6985705533257377e-05, "loss": 0.1152, "step": 24624 }, { "epoch": 2.55, "learning_rate": 1.6983487826473257e-05, "loss": 0.3945, "step": 24625 }, { "epoch": 2.55, "learning_rate": 1.6981270207313266e-05, "loss": 0.3306, "step": 24626 }, { "epoch": 2.55, "learning_rate": 1.6979052675792305e-05, "loss": 0.2491, "step": 24627 }, { "epoch": 2.55, "learning_rate": 1.697683523192531e-05, "loss": 0.5655, "step": 24628 }, { "epoch": 2.55, "learning_rate": 1.697461787572722e-05, "loss": 0.2494, "step": 24629 }, { "epoch": 2.55, "learning_rate": 1.697240060721295e-05, "loss": 0.2096, "step": 24630 }, { "epoch": 2.55, "learning_rate": 1.697018342639743e-05, "loss": 0.0389, "step": 24631 }, { "epoch": 2.55, "learning_rate": 1.696796633329559e-05, "loss": 0.4947, "step": 24632 }, { "epoch": 2.55, "learning_rate": 1.6965749327922348e-05, "loss": 0.4117, "step": 24633 }, { "epoch": 2.55, "learning_rate": 1.6963532410292645e-05, "loss": 0.5526, "step": 24634 }, { "epoch": 2.55, "learning_rate": 1.696131558042138e-05, "loss": 0.2794, "step": 24635 }, { "epoch": 2.55, "learning_rate": 1.695909883832349e-05, "loss": 0.6722, "step": 24636 }, { "epoch": 2.55, "learning_rate": 1.695688218401389e-05, "loss": 0.3105, "step": 24637 }, { "epoch": 2.55, "learning_rate": 1.695466561750751e-05, "loss": 0.1815, "step": 24638 }, { "epoch": 2.55, "learning_rate": 1.6952449138819282e-05, "loss": 0.3367, "step": 24639 }, { "epoch": 2.55, "learning_rate": 1.69502327479641e-05, "loss": 0.3157, "step": 24640 }, { "epoch": 2.55, "learning_rate": 1.69480164449569e-05, "loss": 0.1033, "step": 24641 }, { "epoch": 2.55, "learning_rate": 1.6945800229812598e-05, "loss": 0.3065, "step": 24642 }, { "epoch": 2.55, "learning_rate": 1.6943584102546113e-05, "loss": 0.237, "step": 24643 }, { "epoch": 2.55, "learning_rate": 1.694136806317238e-05, "loss": 0.4183, "step": 24644 }, { "epoch": 2.55, "learning_rate": 1.6939152111706286e-05, "loss": 0.2178, "step": 24645 }, { "epoch": 2.55, "learning_rate": 1.6936936248162763e-05, "loss": 1.0713, "step": 24646 }, { "epoch": 2.55, "learning_rate": 1.6934720472556734e-05, "loss": 0.5054, "step": 24647 }, { "epoch": 2.55, "learning_rate": 1.6932504784903106e-05, "loss": 0.1243, "step": 24648 }, { "epoch": 2.55, "learning_rate": 1.69302891852168e-05, "loss": 0.414, "step": 24649 }, { "epoch": 2.55, "learning_rate": 1.692807367351273e-05, "loss": 0.7187, "step": 24650 }, { "epoch": 2.55, "learning_rate": 1.6925858249805807e-05, "loss": 0.7851, "step": 24651 }, { "epoch": 2.55, "learning_rate": 1.6923642914110957e-05, "loss": 0.7135, "step": 24652 }, { "epoch": 2.55, "learning_rate": 1.6921427666443074e-05, "loss": 0.4927, "step": 24653 }, { "epoch": 2.55, "learning_rate": 1.6919212506817094e-05, "loss": 0.1928, "step": 24654 }, { "epoch": 2.55, "learning_rate": 1.69169974352479e-05, "loss": 0.1403, "step": 24655 }, { "epoch": 2.55, "learning_rate": 1.691478245175042e-05, "loss": 0.0743, "step": 24656 }, { "epoch": 2.55, "learning_rate": 1.6912567556339562e-05, "loss": 0.4745, "step": 24657 }, { "epoch": 2.55, "learning_rate": 1.6910352749030234e-05, "loss": 0.5409, "step": 24658 }, { "epoch": 2.55, "learning_rate": 1.6908138029837354e-05, "loss": 0.0747, "step": 24659 }, { "epoch": 2.55, "learning_rate": 1.690592339877582e-05, "loss": 0.5389, "step": 24660 }, { "epoch": 2.55, "learning_rate": 1.6903708855860553e-05, "loss": 0.4282, "step": 24661 }, { "epoch": 2.55, "learning_rate": 1.690149440110646e-05, "loss": 1.0256, "step": 24662 }, { "epoch": 2.55, "learning_rate": 1.6899280034528432e-05, "loss": 0.4406, "step": 24663 }, { "epoch": 2.55, "learning_rate": 1.6897065756141398e-05, "loss": 0.2742, "step": 24664 }, { "epoch": 2.55, "learning_rate": 1.6894851565960235e-05, "loss": 0.4931, "step": 24665 }, { "epoch": 2.55, "learning_rate": 1.689263746399987e-05, "loss": 0.1704, "step": 24666 }, { "epoch": 2.55, "learning_rate": 1.68904234502752e-05, "loss": 0.3112, "step": 24667 }, { "epoch": 2.55, "learning_rate": 1.6888209524801136e-05, "loss": 0.1448, "step": 24668 }, { "epoch": 2.55, "learning_rate": 1.688599568759257e-05, "loss": 0.2235, "step": 24669 }, { "epoch": 2.55, "learning_rate": 1.6883781938664416e-05, "loss": 1.0231, "step": 24670 }, { "epoch": 2.55, "learning_rate": 1.688156827803158e-05, "loss": 0.5331, "step": 24671 }, { "epoch": 2.55, "learning_rate": 1.6879354705708956e-05, "loss": 0.1878, "step": 24672 }, { "epoch": 2.55, "learning_rate": 1.6877141221711447e-05, "loss": 0.3453, "step": 24673 }, { "epoch": 2.55, "learning_rate": 1.6874927826053944e-05, "loss": 0.0459, "step": 24674 }, { "epoch": 2.55, "learning_rate": 1.687271451875136e-05, "loss": 0.5666, "step": 24675 }, { "epoch": 2.55, "learning_rate": 1.6870501299818587e-05, "loss": 0.3221, "step": 24676 }, { "epoch": 2.55, "learning_rate": 1.6868288169270542e-05, "loss": 0.2231, "step": 24677 }, { "epoch": 2.55, "learning_rate": 1.6866075127122098e-05, "loss": 0.6972, "step": 24678 }, { "epoch": 2.55, "learning_rate": 1.6863862173388158e-05, "loss": 0.6016, "step": 24679 }, { "epoch": 2.55, "learning_rate": 1.686164930808363e-05, "loss": 0.4391, "step": 24680 }, { "epoch": 2.55, "learning_rate": 1.68594365312234e-05, "loss": 0.6411, "step": 24681 }, { "epoch": 2.55, "learning_rate": 1.6857223842822382e-05, "loss": 0.3061, "step": 24682 }, { "epoch": 2.55, "learning_rate": 1.6855011242895443e-05, "loss": 0.3118, "step": 24683 }, { "epoch": 2.55, "learning_rate": 1.6852798731457497e-05, "loss": 0.1932, "step": 24684 }, { "epoch": 2.55, "learning_rate": 1.6850586308523436e-05, "loss": 0.5861, "step": 24685 }, { "epoch": 2.55, "learning_rate": 1.6848373974108146e-05, "loss": 0.5244, "step": 24686 }, { "epoch": 2.55, "learning_rate": 1.684616172822653e-05, "loss": 0.1614, "step": 24687 }, { "epoch": 2.55, "learning_rate": 1.684394957089347e-05, "loss": 0.6112, "step": 24688 }, { "epoch": 2.55, "learning_rate": 1.684173750212388e-05, "loss": 0.5275, "step": 24689 }, { "epoch": 2.55, "learning_rate": 1.6839525521932617e-05, "loss": 0.2853, "step": 24690 }, { "epoch": 2.55, "learning_rate": 1.6837313630334595e-05, "loss": 0.074, "step": 24691 }, { "epoch": 2.55, "learning_rate": 1.6835101827344707e-05, "loss": 0.2663, "step": 24692 }, { "epoch": 2.55, "learning_rate": 1.683289011297782e-05, "loss": 0.2773, "step": 24693 }, { "epoch": 2.55, "learning_rate": 1.6830678487248836e-05, "loss": 0.3261, "step": 24694 }, { "epoch": 2.55, "learning_rate": 1.6828466950172643e-05, "loss": 0.4034, "step": 24695 }, { "epoch": 2.55, "learning_rate": 1.6826255501764133e-05, "loss": 0.3158, "step": 24696 }, { "epoch": 2.55, "learning_rate": 1.6824044142038183e-05, "loss": 0.5451, "step": 24697 }, { "epoch": 2.55, "learning_rate": 1.6821832871009685e-05, "loss": 0.5344, "step": 24698 }, { "epoch": 2.55, "learning_rate": 1.6819621688693533e-05, "loss": 0.3269, "step": 24699 }, { "epoch": 2.55, "learning_rate": 1.6817410595104607e-05, "loss": 0.2131, "step": 24700 }, { "epoch": 2.55, "learning_rate": 1.6815199590257777e-05, "loss": 0.4387, "step": 24701 }, { "epoch": 2.55, "learning_rate": 1.6812988674167956e-05, "loss": 0.3864, "step": 24702 }, { "epoch": 2.55, "learning_rate": 1.681077784684999e-05, "loss": 0.4363, "step": 24703 }, { "epoch": 2.55, "learning_rate": 1.6808567108318787e-05, "loss": 0.4948, "step": 24704 }, { "epoch": 2.55, "learning_rate": 1.680635645858922e-05, "loss": 0.4988, "step": 24705 }, { "epoch": 2.55, "learning_rate": 1.680414589767618e-05, "loss": 0.0974, "step": 24706 }, { "epoch": 2.55, "learning_rate": 1.680193542559454e-05, "loss": 0.393, "step": 24707 }, { "epoch": 2.55, "learning_rate": 1.679972504235918e-05, "loss": 0.1787, "step": 24708 }, { "epoch": 2.55, "learning_rate": 1.679751474798498e-05, "loss": 0.5224, "step": 24709 }, { "epoch": 2.55, "learning_rate": 1.6795304542486838e-05, "loss": 0.0804, "step": 24710 }, { "epoch": 2.55, "learning_rate": 1.6793094425879605e-05, "loss": 0.3712, "step": 24711 }, { "epoch": 2.55, "learning_rate": 1.679088439817817e-05, "loss": 0.5312, "step": 24712 }, { "epoch": 2.55, "learning_rate": 1.678867445939741e-05, "loss": 0.2777, "step": 24713 }, { "epoch": 2.55, "learning_rate": 1.678646460955221e-05, "loss": 0.5577, "step": 24714 }, { "epoch": 2.55, "learning_rate": 1.678425484865743e-05, "loss": 0.5691, "step": 24715 }, { "epoch": 2.55, "learning_rate": 1.6782045176727955e-05, "loss": 0.3207, "step": 24716 }, { "epoch": 2.55, "learning_rate": 1.6779835593778662e-05, "loss": 0.2903, "step": 24717 }, { "epoch": 2.55, "learning_rate": 1.6777626099824416e-05, "loss": 0.1784, "step": 24718 }, { "epoch": 2.55, "learning_rate": 1.67754166948801e-05, "loss": 0.3613, "step": 24719 }, { "epoch": 2.56, "learning_rate": 1.6773207378960596e-05, "loss": 0.4259, "step": 24720 }, { "epoch": 2.56, "learning_rate": 1.6770998152080755e-05, "loss": 0.1337, "step": 24721 }, { "epoch": 2.56, "learning_rate": 1.6768789014255457e-05, "loss": 0.2758, "step": 24722 }, { "epoch": 2.56, "learning_rate": 1.6766579965499577e-05, "loss": 0.3795, "step": 24723 }, { "epoch": 2.56, "learning_rate": 1.6764371005827985e-05, "loss": 0.0641, "step": 24724 }, { "epoch": 2.56, "learning_rate": 1.676216213525555e-05, "loss": 0.4642, "step": 24725 }, { "epoch": 2.56, "learning_rate": 1.6759953353797138e-05, "loss": 0.5248, "step": 24726 }, { "epoch": 2.56, "learning_rate": 1.6757744661467637e-05, "loss": 0.3413, "step": 24727 }, { "epoch": 2.56, "learning_rate": 1.6755536058281892e-05, "loss": 0.5027, "step": 24728 }, { "epoch": 2.56, "learning_rate": 1.6753327544254774e-05, "loss": 0.4979, "step": 24729 }, { "epoch": 2.56, "learning_rate": 1.675111911940117e-05, "loss": 0.2409, "step": 24730 }, { "epoch": 2.56, "learning_rate": 1.674891078373592e-05, "loss": 0.4877, "step": 24731 }, { "epoch": 2.56, "learning_rate": 1.67467025372739e-05, "loss": 0.2762, "step": 24732 }, { "epoch": 2.56, "learning_rate": 1.674449438002998e-05, "loss": 0.5274, "step": 24733 }, { "epoch": 2.56, "learning_rate": 1.6742286312019022e-05, "loss": 0.2344, "step": 24734 }, { "epoch": 2.56, "learning_rate": 1.674007833325589e-05, "loss": 0.406, "step": 24735 }, { "epoch": 2.56, "learning_rate": 1.673787044375545e-05, "loss": 0.4416, "step": 24736 }, { "epoch": 2.56, "learning_rate": 1.6735662643532562e-05, "loss": 0.443, "step": 24737 }, { "epoch": 2.56, "learning_rate": 1.67334549326021e-05, "loss": 0.9776, "step": 24738 }, { "epoch": 2.56, "learning_rate": 1.6731247310978902e-05, "loss": 0.2639, "step": 24739 }, { "epoch": 2.56, "learning_rate": 1.6729039778677855e-05, "loss": 0.4687, "step": 24740 }, { "epoch": 2.56, "learning_rate": 1.67268323357138e-05, "loss": 0.2698, "step": 24741 }, { "epoch": 2.56, "learning_rate": 1.67246249821016e-05, "loss": 0.5688, "step": 24742 }, { "epoch": 2.56, "learning_rate": 1.6722417717856114e-05, "loss": 0.3721, "step": 24743 }, { "epoch": 2.56, "learning_rate": 1.6720210542992213e-05, "loss": 0.1051, "step": 24744 }, { "epoch": 2.56, "learning_rate": 1.6718003457524745e-05, "loss": 0.1477, "step": 24745 }, { "epoch": 2.56, "learning_rate": 1.671579646146857e-05, "loss": 0.2757, "step": 24746 }, { "epoch": 2.56, "learning_rate": 1.6713589554838546e-05, "loss": 0.3082, "step": 24747 }, { "epoch": 2.56, "learning_rate": 1.671138273764954e-05, "loss": 0.0732, "step": 24748 }, { "epoch": 2.56, "learning_rate": 1.670917600991638e-05, "loss": 0.0403, "step": 24749 }, { "epoch": 2.56, "learning_rate": 1.670696937165394e-05, "loss": 0.5184, "step": 24750 }, { "epoch": 2.56, "learning_rate": 1.670476282287707e-05, "loss": 0.1702, "step": 24751 }, { "epoch": 2.56, "learning_rate": 1.6702556363600644e-05, "loss": 0.4773, "step": 24752 }, { "epoch": 2.56, "learning_rate": 1.670034999383948e-05, "loss": 0.145, "step": 24753 }, { "epoch": 2.56, "learning_rate": 1.669814371360845e-05, "loss": 0.3063, "step": 24754 }, { "epoch": 2.56, "learning_rate": 1.6695937522922402e-05, "loss": 0.1577, "step": 24755 }, { "epoch": 2.56, "learning_rate": 1.669373142179619e-05, "loss": 0.163, "step": 24756 }, { "epoch": 2.56, "learning_rate": 1.6691525410244666e-05, "loss": 0.4955, "step": 24757 }, { "epoch": 2.56, "learning_rate": 1.668931948828269e-05, "loss": 0.55, "step": 24758 }, { "epoch": 2.56, "learning_rate": 1.6687113655925093e-05, "loss": 0.6949, "step": 24759 }, { "epoch": 2.56, "learning_rate": 1.6684907913186728e-05, "loss": 0.2514, "step": 24760 }, { "epoch": 2.56, "learning_rate": 1.6682702260082446e-05, "loss": 1.014, "step": 24761 }, { "epoch": 2.56, "learning_rate": 1.6680496696627095e-05, "loss": 0.3887, "step": 24762 }, { "epoch": 2.56, "learning_rate": 1.6678291222835528e-05, "loss": 0.4681, "step": 24763 }, { "epoch": 2.56, "learning_rate": 1.667608583872259e-05, "loss": 0.2602, "step": 24764 }, { "epoch": 2.56, "learning_rate": 1.6673880544303133e-05, "loss": 0.0248, "step": 24765 }, { "epoch": 2.56, "learning_rate": 1.667167533959198e-05, "loss": 0.3813, "step": 24766 }, { "epoch": 2.56, "learning_rate": 1.666947022460399e-05, "loss": 0.4423, "step": 24767 }, { "epoch": 2.56, "learning_rate": 1.666726519935403e-05, "loss": 0.9722, "step": 24768 }, { "epoch": 2.56, "learning_rate": 1.6665060263856894e-05, "loss": 0.4577, "step": 24769 }, { "epoch": 2.56, "learning_rate": 1.666285541812746e-05, "loss": 0.2013, "step": 24770 }, { "epoch": 2.56, "learning_rate": 1.666065066218056e-05, "loss": 0.051, "step": 24771 }, { "epoch": 2.56, "learning_rate": 1.665844599603104e-05, "loss": 0.6489, "step": 24772 }, { "epoch": 2.56, "learning_rate": 1.665624141969374e-05, "loss": 0.3454, "step": 24773 }, { "epoch": 2.56, "learning_rate": 1.66540369331835e-05, "loss": 0.3548, "step": 24774 }, { "epoch": 2.56, "learning_rate": 1.665183253651516e-05, "loss": 0.4229, "step": 24775 }, { "epoch": 2.56, "learning_rate": 1.6649628229703576e-05, "loss": 0.3482, "step": 24776 }, { "epoch": 2.56, "learning_rate": 1.6647424012763555e-05, "loss": 0.8935, "step": 24777 }, { "epoch": 2.56, "learning_rate": 1.6645219885709966e-05, "loss": 0.2262, "step": 24778 }, { "epoch": 2.56, "learning_rate": 1.6643015848557617e-05, "loss": 0.4212, "step": 24779 }, { "epoch": 2.56, "learning_rate": 1.6640811901321364e-05, "loss": 0.6662, "step": 24780 }, { "epoch": 2.56, "learning_rate": 1.663860804401604e-05, "loss": 0.7217, "step": 24781 }, { "epoch": 2.56, "learning_rate": 1.663640427665647e-05, "loss": 0.0398, "step": 24782 }, { "epoch": 2.56, "learning_rate": 1.663420059925751e-05, "loss": 0.5428, "step": 24783 }, { "epoch": 2.56, "learning_rate": 1.663199701183398e-05, "loss": 0.3926, "step": 24784 }, { "epoch": 2.56, "learning_rate": 1.6629793514400722e-05, "loss": 0.522, "step": 24785 }, { "epoch": 2.56, "learning_rate": 1.6627590106972576e-05, "loss": 0.2805, "step": 24786 }, { "epoch": 2.56, "learning_rate": 1.6625386789564347e-05, "loss": 0.1649, "step": 24787 }, { "epoch": 2.56, "learning_rate": 1.662318356219089e-05, "loss": 0.0931, "step": 24788 }, { "epoch": 2.56, "learning_rate": 1.6620980424867035e-05, "loss": 0.4872, "step": 24789 }, { "epoch": 2.56, "learning_rate": 1.6618777377607614e-05, "loss": 0.1315, "step": 24790 }, { "epoch": 2.56, "learning_rate": 1.6616574420427444e-05, "loss": 0.4522, "step": 24791 }, { "epoch": 2.56, "learning_rate": 1.6614371553341365e-05, "loss": 0.264, "step": 24792 }, { "epoch": 2.56, "learning_rate": 1.66121687763642e-05, "loss": 0.2263, "step": 24793 }, { "epoch": 2.56, "learning_rate": 1.6609966089510785e-05, "loss": 0.5828, "step": 24794 }, { "epoch": 2.56, "learning_rate": 1.6607763492795948e-05, "loss": 0.5764, "step": 24795 }, { "epoch": 2.56, "learning_rate": 1.660556098623452e-05, "loss": 0.4429, "step": 24796 }, { "epoch": 2.56, "learning_rate": 1.6603358569841316e-05, "loss": 0.6392, "step": 24797 }, { "epoch": 2.56, "learning_rate": 1.6601156243631164e-05, "loss": 0.3676, "step": 24798 }, { "epoch": 2.56, "learning_rate": 1.659895400761889e-05, "loss": 0.1202, "step": 24799 }, { "epoch": 2.56, "learning_rate": 1.659675186181933e-05, "loss": 0.5473, "step": 24800 }, { "epoch": 2.56, "learning_rate": 1.6594549806247294e-05, "loss": 0.2734, "step": 24801 }, { "epoch": 2.56, "learning_rate": 1.6592347840917626e-05, "loss": 0.556, "step": 24802 }, { "epoch": 2.56, "learning_rate": 1.6590145965845122e-05, "loss": 0.4955, "step": 24803 }, { "epoch": 2.56, "learning_rate": 1.6587944181044617e-05, "loss": 0.3977, "step": 24804 }, { "epoch": 2.56, "learning_rate": 1.658574248653094e-05, "loss": 0.8118, "step": 24805 }, { "epoch": 2.56, "learning_rate": 1.6583540882318907e-05, "loss": 0.3167, "step": 24806 }, { "epoch": 2.56, "learning_rate": 1.6581339368423333e-05, "loss": 0.606, "step": 24807 }, { "epoch": 2.56, "learning_rate": 1.6579137944859042e-05, "loss": 0.1119, "step": 24808 }, { "epoch": 2.56, "learning_rate": 1.657693661164085e-05, "loss": 0.4764, "step": 24809 }, { "epoch": 2.56, "learning_rate": 1.6574735368783588e-05, "loss": 0.3731, "step": 24810 }, { "epoch": 2.56, "learning_rate": 1.6572534216302058e-05, "loss": 0.375, "step": 24811 }, { "epoch": 2.56, "learning_rate": 1.6570333154211093e-05, "loss": 0.4605, "step": 24812 }, { "epoch": 2.56, "learning_rate": 1.65681321825255e-05, "loss": 0.8947, "step": 24813 }, { "epoch": 2.56, "learning_rate": 1.6565931301260107e-05, "loss": 0.6179, "step": 24814 }, { "epoch": 2.57, "learning_rate": 1.6563730510429712e-05, "loss": 0.1907, "step": 24815 }, { "epoch": 2.57, "learning_rate": 1.656152981004915e-05, "loss": 0.485, "step": 24816 }, { "epoch": 2.57, "learning_rate": 1.6559329200133216e-05, "loss": 0.4794, "step": 24817 }, { "epoch": 2.57, "learning_rate": 1.655712868069673e-05, "loss": 0.511, "step": 24818 }, { "epoch": 2.57, "learning_rate": 1.6554928251754512e-05, "loss": 0.4614, "step": 24819 }, { "epoch": 2.57, "learning_rate": 1.655272791332137e-05, "loss": 0.279, "step": 24820 }, { "epoch": 2.57, "learning_rate": 1.6550527665412113e-05, "loss": 0.4477, "step": 24821 }, { "epoch": 2.57, "learning_rate": 1.654832750804156e-05, "loss": 0.2388, "step": 24822 }, { "epoch": 2.57, "learning_rate": 1.6546127441224515e-05, "loss": 0.2349, "step": 24823 }, { "epoch": 2.57, "learning_rate": 1.654392746497581e-05, "loss": 0.4164, "step": 24824 }, { "epoch": 2.57, "learning_rate": 1.6541727579310226e-05, "loss": 0.4513, "step": 24825 }, { "epoch": 2.57, "learning_rate": 1.6539527784242575e-05, "loss": 0.6906, "step": 24826 }, { "epoch": 2.57, "learning_rate": 1.653732807978769e-05, "loss": 0.4143, "step": 24827 }, { "epoch": 2.57, "learning_rate": 1.653512846596035e-05, "loss": 0.7468, "step": 24828 }, { "epoch": 2.57, "learning_rate": 1.6532928942775374e-05, "loss": 0.2242, "step": 24829 }, { "epoch": 2.57, "learning_rate": 1.6530729510247572e-05, "loss": 0.3322, "step": 24830 }, { "epoch": 2.57, "learning_rate": 1.6528530168391747e-05, "loss": 0.1485, "step": 24831 }, { "epoch": 2.57, "learning_rate": 1.6526330917222702e-05, "loss": 0.3788, "step": 24832 }, { "epoch": 2.57, "learning_rate": 1.6524131756755245e-05, "loss": 0.4418, "step": 24833 }, { "epoch": 2.57, "learning_rate": 1.6521932687004194e-05, "loss": 0.2742, "step": 24834 }, { "epoch": 2.57, "learning_rate": 1.6519733707984328e-05, "loss": 0.1957, "step": 24835 }, { "epoch": 2.57, "learning_rate": 1.6517534819710458e-05, "loss": 0.2535, "step": 24836 }, { "epoch": 2.57, "learning_rate": 1.6515336022197385e-05, "loss": 0.2359, "step": 24837 }, { "epoch": 2.57, "learning_rate": 1.6513137315459922e-05, "loss": 0.0394, "step": 24838 }, { "epoch": 2.57, "learning_rate": 1.651093869951286e-05, "loss": 0.4729, "step": 24839 }, { "epoch": 2.57, "learning_rate": 1.6508740174371012e-05, "loss": 0.5132, "step": 24840 }, { "epoch": 2.57, "learning_rate": 1.650654174004916e-05, "loss": 0.5835, "step": 24841 }, { "epoch": 2.57, "learning_rate": 1.650434339656211e-05, "loss": 0.2718, "step": 24842 }, { "epoch": 2.57, "learning_rate": 1.650214514392466e-05, "loss": 0.1347, "step": 24843 }, { "epoch": 2.57, "learning_rate": 1.649994698215163e-05, "loss": 0.2327, "step": 24844 }, { "epoch": 2.57, "learning_rate": 1.649774891125778e-05, "loss": 0.0983, "step": 24845 }, { "epoch": 2.57, "learning_rate": 1.6495550931257928e-05, "loss": 0.486, "step": 24846 }, { "epoch": 2.57, "learning_rate": 1.6493353042166864e-05, "loss": 0.0323, "step": 24847 }, { "epoch": 2.57, "learning_rate": 1.6491155243999393e-05, "loss": 0.1465, "step": 24848 }, { "epoch": 2.57, "learning_rate": 1.6488957536770298e-05, "loss": 0.1649, "step": 24849 }, { "epoch": 2.57, "learning_rate": 1.648675992049438e-05, "loss": 0.1609, "step": 24850 }, { "epoch": 2.57, "learning_rate": 1.6484562395186434e-05, "loss": 0.2035, "step": 24851 }, { "epoch": 2.57, "learning_rate": 1.648236496086126e-05, "loss": 0.2593, "step": 24852 }, { "epoch": 2.57, "learning_rate": 1.6480167617533637e-05, "loss": 0.2433, "step": 24853 }, { "epoch": 2.57, "learning_rate": 1.6477970365218372e-05, "loss": 0.2847, "step": 24854 }, { "epoch": 2.57, "learning_rate": 1.6475773203930235e-05, "loss": 0.0575, "step": 24855 }, { "epoch": 2.57, "learning_rate": 1.6473576133684024e-05, "loss": 0.7138, "step": 24856 }, { "epoch": 2.57, "learning_rate": 1.647137915449454e-05, "loss": 0.3091, "step": 24857 }, { "epoch": 2.57, "learning_rate": 1.6469182266376563e-05, "loss": 0.192, "step": 24858 }, { "epoch": 2.57, "learning_rate": 1.6466985469344885e-05, "loss": 0.207, "step": 24859 }, { "epoch": 2.57, "learning_rate": 1.646478876341429e-05, "loss": 0.0789, "step": 24860 }, { "epoch": 2.57, "learning_rate": 1.6462592148599576e-05, "loss": 0.6007, "step": 24861 }, { "epoch": 2.57, "learning_rate": 1.6460395624915534e-05, "loss": 0.1946, "step": 24862 }, { "epoch": 2.57, "learning_rate": 1.6458199192376932e-05, "loss": 0.2924, "step": 24863 }, { "epoch": 2.57, "learning_rate": 1.645600285099856e-05, "loss": 0.3125, "step": 24864 }, { "epoch": 2.57, "learning_rate": 1.6453806600795223e-05, "loss": 0.3688, "step": 24865 }, { "epoch": 2.57, "learning_rate": 1.6451610441781676e-05, "loss": 0.4554, "step": 24866 }, { "epoch": 2.57, "learning_rate": 1.6449414373972714e-05, "loss": 0.5817, "step": 24867 }, { "epoch": 2.57, "learning_rate": 1.6447218397383132e-05, "loss": 0.312, "step": 24868 }, { "epoch": 2.57, "learning_rate": 1.64450225120277e-05, "loss": 0.432, "step": 24869 }, { "epoch": 2.57, "learning_rate": 1.6442826717921205e-05, "loss": 0.1905, "step": 24870 }, { "epoch": 2.57, "learning_rate": 1.6440631015078428e-05, "loss": 0.526, "step": 24871 }, { "epoch": 2.57, "learning_rate": 1.6438435403514164e-05, "loss": 0.4408, "step": 24872 }, { "epoch": 2.57, "learning_rate": 1.6436239883243166e-05, "loss": 0.3219, "step": 24873 }, { "epoch": 2.57, "learning_rate": 1.643404445428023e-05, "loss": 0.2499, "step": 24874 }, { "epoch": 2.57, "learning_rate": 1.6431849116640128e-05, "loss": 0.1798, "step": 24875 }, { "epoch": 2.57, "learning_rate": 1.6429653870337647e-05, "loss": 0.2159, "step": 24876 }, { "epoch": 2.57, "learning_rate": 1.642745871538756e-05, "loss": 0.3254, "step": 24877 }, { "epoch": 2.57, "learning_rate": 1.642526365180466e-05, "loss": 0.9382, "step": 24878 }, { "epoch": 2.57, "learning_rate": 1.642306867960369e-05, "loss": 0.364, "step": 24879 }, { "epoch": 2.57, "learning_rate": 1.642087379879945e-05, "loss": 0.0894, "step": 24880 }, { "epoch": 2.57, "learning_rate": 1.6418679009406713e-05, "loss": 0.6845, "step": 24881 }, { "epoch": 2.57, "learning_rate": 1.6416484311440256e-05, "loss": 0.7961, "step": 24882 }, { "epoch": 2.57, "learning_rate": 1.6414289704914842e-05, "loss": 0.2029, "step": 24883 }, { "epoch": 2.57, "learning_rate": 1.6412095189845245e-05, "loss": 0.1802, "step": 24884 }, { "epoch": 2.57, "learning_rate": 1.6409900766246252e-05, "loss": 0.1948, "step": 24885 }, { "epoch": 2.57, "learning_rate": 1.640770643413262e-05, "loss": 0.2834, "step": 24886 }, { "epoch": 2.57, "learning_rate": 1.640551219351913e-05, "loss": 0.3367, "step": 24887 }, { "epoch": 2.57, "learning_rate": 1.640331804442055e-05, "loss": 0.3713, "step": 24888 }, { "epoch": 2.57, "learning_rate": 1.6401123986851653e-05, "loss": 0.4458, "step": 24889 }, { "epoch": 2.57, "learning_rate": 1.6398930020827224e-05, "loss": 0.5017, "step": 24890 }, { "epoch": 2.57, "learning_rate": 1.6396736146361994e-05, "loss": 0.1585, "step": 24891 }, { "epoch": 2.57, "learning_rate": 1.6394542363470774e-05, "loss": 0.1719, "step": 24892 }, { "epoch": 2.57, "learning_rate": 1.6392348672168295e-05, "loss": 0.4046, "step": 24893 }, { "epoch": 2.57, "learning_rate": 1.639015507246934e-05, "loss": 0.1359, "step": 24894 }, { "epoch": 2.57, "learning_rate": 1.638796156438868e-05, "loss": 0.1491, "step": 24895 }, { "epoch": 2.57, "learning_rate": 1.6385768147941074e-05, "loss": 0.3286, "step": 24896 }, { "epoch": 2.57, "learning_rate": 1.6383574823141294e-05, "loss": 0.61, "step": 24897 }, { "epoch": 2.57, "learning_rate": 1.6381381590004103e-05, "loss": 0.4544, "step": 24898 }, { "epoch": 2.57, "learning_rate": 1.6379188448544266e-05, "loss": 0.692, "step": 24899 }, { "epoch": 2.57, "learning_rate": 1.6376995398776554e-05, "loss": 0.763, "step": 24900 }, { "epoch": 2.57, "learning_rate": 1.637480244071571e-05, "loss": 0.3172, "step": 24901 }, { "epoch": 2.57, "learning_rate": 1.637260957437651e-05, "loss": 0.2151, "step": 24902 }, { "epoch": 2.57, "learning_rate": 1.6370416799773725e-05, "loss": 0.099, "step": 24903 }, { "epoch": 2.57, "learning_rate": 1.636822411692209e-05, "loss": 0.4896, "step": 24904 }, { "epoch": 2.57, "learning_rate": 1.6366031525836384e-05, "loss": 0.3667, "step": 24905 }, { "epoch": 2.57, "learning_rate": 1.6363839026531363e-05, "loss": 0.6225, "step": 24906 }, { "epoch": 2.57, "learning_rate": 1.6361646619021783e-05, "loss": 0.2387, "step": 24907 }, { "epoch": 2.57, "learning_rate": 1.6359454303322413e-05, "loss": 0.3073, "step": 24908 }, { "epoch": 2.57, "learning_rate": 1.6357262079448e-05, "loss": 0.3068, "step": 24909 }, { "epoch": 2.57, "learning_rate": 1.6355069947413318e-05, "loss": 0.1238, "step": 24910 }, { "epoch": 2.58, "learning_rate": 1.6352877907233108e-05, "loss": 0.3927, "step": 24911 }, { "epoch": 2.58, "learning_rate": 1.6350685958922125e-05, "loss": 0.2587, "step": 24912 }, { "epoch": 2.58, "learning_rate": 1.634849410249513e-05, "loss": 0.1867, "step": 24913 }, { "epoch": 2.58, "learning_rate": 1.634630233796688e-05, "loss": 0.2426, "step": 24914 }, { "epoch": 2.58, "learning_rate": 1.634411066535214e-05, "loss": 0.3294, "step": 24915 }, { "epoch": 2.58, "learning_rate": 1.634191908466564e-05, "loss": 0.6179, "step": 24916 }, { "epoch": 2.58, "learning_rate": 1.6339727595922145e-05, "loss": 0.1319, "step": 24917 }, { "epoch": 2.58, "learning_rate": 1.6337536199136408e-05, "loss": 0.6817, "step": 24918 }, { "epoch": 2.58, "learning_rate": 1.6335344894323183e-05, "loss": 0.1864, "step": 24919 }, { "epoch": 2.58, "learning_rate": 1.6333153681497223e-05, "loss": 0.3261, "step": 24920 }, { "epoch": 2.58, "learning_rate": 1.633096256067327e-05, "loss": 0.6011, "step": 24921 }, { "epoch": 2.58, "learning_rate": 1.6328771531866075e-05, "loss": 1.0906, "step": 24922 }, { "epoch": 2.58, "learning_rate": 1.6326580595090393e-05, "loss": 0.3701, "step": 24923 }, { "epoch": 2.58, "learning_rate": 1.632438975036097e-05, "loss": 0.4857, "step": 24924 }, { "epoch": 2.58, "learning_rate": 1.6322198997692555e-05, "loss": 0.52, "step": 24925 }, { "epoch": 2.58, "learning_rate": 1.6320008337099897e-05, "loss": 0.1864, "step": 24926 }, { "epoch": 2.58, "learning_rate": 1.6317817768597745e-05, "loss": 0.6323, "step": 24927 }, { "epoch": 2.58, "learning_rate": 1.6315627292200848e-05, "loss": 0.7317, "step": 24928 }, { "epoch": 2.58, "learning_rate": 1.6313436907923937e-05, "loss": 0.0627, "step": 24929 }, { "epoch": 2.58, "learning_rate": 1.631124661578178e-05, "loss": 0.22, "step": 24930 }, { "epoch": 2.58, "learning_rate": 1.6309056415789096e-05, "loss": 0.3232, "step": 24931 }, { "epoch": 2.58, "learning_rate": 1.6306866307960636e-05, "loss": 0.3871, "step": 24932 }, { "epoch": 2.58, "learning_rate": 1.6304676292311156e-05, "loss": 0.397, "step": 24933 }, { "epoch": 2.58, "learning_rate": 1.6302486368855385e-05, "loss": 0.5237, "step": 24934 }, { "epoch": 2.58, "learning_rate": 1.6300296537608074e-05, "loss": 0.7059, "step": 24935 }, { "epoch": 2.58, "learning_rate": 1.6298106798583956e-05, "loss": 0.3015, "step": 24936 }, { "epoch": 2.58, "learning_rate": 1.6295917151797778e-05, "loss": 0.3895, "step": 24937 }, { "epoch": 2.58, "learning_rate": 1.6293727597264297e-05, "loss": 0.2988, "step": 24938 }, { "epoch": 2.58, "learning_rate": 1.629153813499822e-05, "loss": 0.2424, "step": 24939 }, { "epoch": 2.58, "learning_rate": 1.628934876501431e-05, "loss": 0.3258, "step": 24940 }, { "epoch": 2.58, "learning_rate": 1.6287159487327287e-05, "loss": 0.384, "step": 24941 }, { "epoch": 2.58, "learning_rate": 1.62849703019519e-05, "loss": 0.3289, "step": 24942 }, { "epoch": 2.58, "learning_rate": 1.628278120890288e-05, "loss": 0.8075, "step": 24943 }, { "epoch": 2.58, "learning_rate": 1.6280592208194965e-05, "loss": 0.2498, "step": 24944 }, { "epoch": 2.58, "learning_rate": 1.62784032998429e-05, "loss": 0.0351, "step": 24945 }, { "epoch": 2.58, "learning_rate": 1.6276214483861405e-05, "loss": 0.0981, "step": 24946 }, { "epoch": 2.58, "learning_rate": 1.627402576026523e-05, "loss": 0.3874, "step": 24947 }, { "epoch": 2.58, "learning_rate": 1.6271837129069112e-05, "loss": 1.0291, "step": 24948 }, { "epoch": 2.58, "learning_rate": 1.6269648590287763e-05, "loss": 0.3791, "step": 24949 }, { "epoch": 2.58, "learning_rate": 1.6267460143935926e-05, "loss": 0.1482, "step": 24950 }, { "epoch": 2.58, "learning_rate": 1.626527179002833e-05, "loss": 0.0886, "step": 24951 }, { "epoch": 2.58, "learning_rate": 1.6263083528579717e-05, "loss": 0.1181, "step": 24952 }, { "epoch": 2.58, "learning_rate": 1.626089535960482e-05, "loss": 0.2637, "step": 24953 }, { "epoch": 2.58, "learning_rate": 1.625870728311835e-05, "loss": 0.7423, "step": 24954 }, { "epoch": 2.58, "learning_rate": 1.625651929913505e-05, "loss": 0.552, "step": 24955 }, { "epoch": 2.58, "learning_rate": 1.6254331407669642e-05, "loss": 0.2129, "step": 24956 }, { "epoch": 2.58, "learning_rate": 1.625214360873686e-05, "loss": 0.0853, "step": 24957 }, { "epoch": 2.58, "learning_rate": 1.6249955902351445e-05, "loss": 0.3604, "step": 24958 }, { "epoch": 2.58, "learning_rate": 1.6247768288528092e-05, "loss": 0.7943, "step": 24959 }, { "epoch": 2.58, "learning_rate": 1.6245580767281554e-05, "loss": 0.0176, "step": 24960 }, { "epoch": 2.58, "learning_rate": 1.624339333862654e-05, "loss": 0.209, "step": 24961 }, { "epoch": 2.58, "learning_rate": 1.6241206002577785e-05, "loss": 0.2514, "step": 24962 }, { "epoch": 2.58, "learning_rate": 1.6239018759150017e-05, "loss": 0.1909, "step": 24963 }, { "epoch": 2.58, "learning_rate": 1.6236831608357946e-05, "loss": 0.1413, "step": 24964 }, { "epoch": 2.58, "learning_rate": 1.623464455021631e-05, "loss": 0.246, "step": 24965 }, { "epoch": 2.58, "learning_rate": 1.6232457584739837e-05, "loss": 0.193, "step": 24966 }, { "epoch": 2.58, "learning_rate": 1.6230270711943228e-05, "loss": 0.111, "step": 24967 }, { "epoch": 2.58, "learning_rate": 1.6228083931841225e-05, "loss": 0.4744, "step": 24968 }, { "epoch": 2.58, "learning_rate": 1.6225897244448523e-05, "loss": 0.519, "step": 24969 }, { "epoch": 2.58, "learning_rate": 1.6223710649779864e-05, "loss": 0.4797, "step": 24970 }, { "epoch": 2.58, "learning_rate": 1.6221524147849963e-05, "loss": 0.4117, "step": 24971 }, { "epoch": 2.58, "learning_rate": 1.621933773867353e-05, "loss": 0.9273, "step": 24972 }, { "epoch": 2.58, "learning_rate": 1.6217151422265295e-05, "loss": 0.4097, "step": 24973 }, { "epoch": 2.58, "learning_rate": 1.6214965198639974e-05, "loss": 0.3643, "step": 24974 }, { "epoch": 2.58, "learning_rate": 1.6212779067812276e-05, "loss": 0.0602, "step": 24975 }, { "epoch": 2.58, "learning_rate": 1.621059302979694e-05, "loss": 0.5295, "step": 24976 }, { "epoch": 2.58, "learning_rate": 1.620840708460865e-05, "loss": 0.0887, "step": 24977 }, { "epoch": 2.58, "learning_rate": 1.620622123226215e-05, "loss": 0.4703, "step": 24978 }, { "epoch": 2.58, "learning_rate": 1.6204035472772126e-05, "loss": 0.0931, "step": 24979 }, { "epoch": 2.58, "learning_rate": 1.6201849806153316e-05, "loss": 0.1164, "step": 24980 }, { "epoch": 2.58, "learning_rate": 1.6199664232420418e-05, "loss": 0.3729, "step": 24981 }, { "epoch": 2.58, "learning_rate": 1.6197478751588152e-05, "loss": 0.7353, "step": 24982 }, { "epoch": 2.58, "learning_rate": 1.6195293363671232e-05, "loss": 0.2649, "step": 24983 }, { "epoch": 2.58, "learning_rate": 1.6193108068684365e-05, "loss": 0.7869, "step": 24984 }, { "epoch": 2.58, "learning_rate": 1.6190922866642263e-05, "loss": 0.3285, "step": 24985 }, { "epoch": 2.58, "learning_rate": 1.618873775755965e-05, "loss": 0.542, "step": 24986 }, { "epoch": 2.58, "learning_rate": 1.6186552741451212e-05, "loss": 0.4247, "step": 24987 }, { "epoch": 2.58, "learning_rate": 1.6184367818331666e-05, "loss": 0.0252, "step": 24988 }, { "epoch": 2.58, "learning_rate": 1.6182182988215724e-05, "loss": 0.1485, "step": 24989 }, { "epoch": 2.58, "learning_rate": 1.6179998251118098e-05, "loss": 0.136, "step": 24990 }, { "epoch": 2.58, "learning_rate": 1.6177813607053492e-05, "loss": 0.6764, "step": 24991 }, { "epoch": 2.58, "learning_rate": 1.6175629056036607e-05, "loss": 0.1609, "step": 24992 }, { "epoch": 2.58, "learning_rate": 1.6173444598082155e-05, "loss": 0.8808, "step": 24993 }, { "epoch": 2.58, "learning_rate": 1.617126023320483e-05, "loss": 0.233, "step": 24994 }, { "epoch": 2.58, "learning_rate": 1.6169075961419352e-05, "loss": 0.248, "step": 24995 }, { "epoch": 2.58, "learning_rate": 1.6166891782740424e-05, "loss": 0.4484, "step": 24996 }, { "epoch": 2.58, "learning_rate": 1.616470769718274e-05, "loss": 0.2444, "step": 24997 }, { "epoch": 2.58, "learning_rate": 1.6162523704761003e-05, "loss": 0.3101, "step": 24998 }, { "epoch": 2.58, "learning_rate": 1.616033980548992e-05, "loss": 0.1776, "step": 24999 }, { "epoch": 2.58, "learning_rate": 1.6158155999384187e-05, "loss": 0.5974, "step": 25000 }, { "epoch": 2.58, "learning_rate": 1.6155972286458512e-05, "loss": 0.2325, "step": 25001 }, { "epoch": 2.58, "learning_rate": 1.615378866672759e-05, "loss": 0.0813, "step": 25002 }, { "epoch": 2.58, "learning_rate": 1.6151605140206138e-05, "loss": 0.5976, "step": 25003 }, { "epoch": 2.58, "learning_rate": 1.6149421706908828e-05, "loss": 0.1629, "step": 25004 }, { "epoch": 2.58, "learning_rate": 1.614723836685037e-05, "loss": 0.5459, "step": 25005 }, { "epoch": 2.58, "learning_rate": 1.614505512004547e-05, "loss": 0.529, "step": 25006 }, { "epoch": 2.59, "learning_rate": 1.614287196650881e-05, "loss": 0.3188, "step": 25007 }, { "epoch": 2.59, "learning_rate": 1.614068890625509e-05, "loss": 0.1435, "step": 25008 }, { "epoch": 2.59, "learning_rate": 1.613850593929901e-05, "loss": 0.4911, "step": 25009 }, { "epoch": 2.59, "learning_rate": 1.6136323065655268e-05, "loss": 0.1729, "step": 25010 }, { "epoch": 2.59, "learning_rate": 1.6134140285338554e-05, "loss": 0.2818, "step": 25011 }, { "epoch": 2.59, "learning_rate": 1.613195759836356e-05, "loss": 0.0491, "step": 25012 }, { "epoch": 2.59, "learning_rate": 1.6129775004744988e-05, "loss": 0.2098, "step": 25013 }, { "epoch": 2.59, "learning_rate": 1.6127592504497536e-05, "loss": 0.2596, "step": 25014 }, { "epoch": 2.59, "learning_rate": 1.612541009763587e-05, "loss": 0.5734, "step": 25015 }, { "epoch": 2.59, "learning_rate": 1.6123227784174712e-05, "loss": 0.4488, "step": 25016 }, { "epoch": 2.59, "learning_rate": 1.612104556412872e-05, "loss": 0.1555, "step": 25017 }, { "epoch": 2.59, "learning_rate": 1.611886343751261e-05, "loss": 0.2358, "step": 25018 }, { "epoch": 2.59, "learning_rate": 1.6116681404341056e-05, "loss": 0.3619, "step": 25019 }, { "epoch": 2.59, "learning_rate": 1.611449946462876e-05, "loss": 0.5027, "step": 25020 }, { "epoch": 2.59, "learning_rate": 1.6112317618390404e-05, "loss": 0.2665, "step": 25021 }, { "epoch": 2.59, "learning_rate": 1.6110135865640675e-05, "loss": 0.4431, "step": 25022 }, { "epoch": 2.59, "learning_rate": 1.6107954206394258e-05, "loss": 0.4106, "step": 25023 }, { "epoch": 2.59, "learning_rate": 1.610577264066586e-05, "loss": 0.4161, "step": 25024 }, { "epoch": 2.59, "learning_rate": 1.6103591168470135e-05, "loss": 0.0824, "step": 25025 }, { "epoch": 2.59, "learning_rate": 1.6101409789821783e-05, "loss": 0.1317, "step": 25026 }, { "epoch": 2.59, "learning_rate": 1.6099228504735488e-05, "loss": 0.114, "step": 25027 }, { "epoch": 2.59, "learning_rate": 1.609704731322595e-05, "loss": 0.6278, "step": 25028 }, { "epoch": 2.59, "learning_rate": 1.609486621530782e-05, "loss": 0.4346, "step": 25029 }, { "epoch": 2.59, "learning_rate": 1.6092685210995797e-05, "loss": 0.1947, "step": 25030 }, { "epoch": 2.59, "learning_rate": 1.6090504300304566e-05, "loss": 0.2019, "step": 25031 }, { "epoch": 2.59, "learning_rate": 1.6088323483248806e-05, "loss": 0.1637, "step": 25032 }, { "epoch": 2.59, "learning_rate": 1.6086142759843194e-05, "loss": 0.2134, "step": 25033 }, { "epoch": 2.59, "learning_rate": 1.6083962130102428e-05, "loss": 0.0337, "step": 25034 }, { "epoch": 2.59, "learning_rate": 1.6081781594041157e-05, "loss": 0.2175, "step": 25035 }, { "epoch": 2.59, "learning_rate": 1.6079601151674077e-05, "loss": 0.4357, "step": 25036 }, { "epoch": 2.59, "learning_rate": 1.6077420803015868e-05, "loss": 0.4958, "step": 25037 }, { "epoch": 2.59, "learning_rate": 1.6075240548081204e-05, "loss": 0.3382, "step": 25038 }, { "epoch": 2.59, "learning_rate": 1.6073060386884763e-05, "loss": 0.7439, "step": 25039 }, { "epoch": 2.59, "learning_rate": 1.6070880319441214e-05, "loss": 0.168, "step": 25040 }, { "epoch": 2.59, "learning_rate": 1.6068700345765257e-05, "loss": 0.5473, "step": 25041 }, { "epoch": 2.59, "learning_rate": 1.6066520465871537e-05, "loss": 0.3973, "step": 25042 }, { "epoch": 2.59, "learning_rate": 1.606434067977474e-05, "loss": 0.6694, "step": 25043 }, { "epoch": 2.59, "learning_rate": 1.6062160987489557e-05, "loss": 0.6955, "step": 25044 }, { "epoch": 2.59, "learning_rate": 1.6059981389030632e-05, "loss": 0.233, "step": 25045 }, { "epoch": 2.59, "learning_rate": 1.6057801884412643e-05, "loss": 0.2081, "step": 25046 }, { "epoch": 2.59, "learning_rate": 1.6055622473650276e-05, "loss": 0.5128, "step": 25047 }, { "epoch": 2.59, "learning_rate": 1.6053443156758198e-05, "loss": 0.119, "step": 25048 }, { "epoch": 2.59, "learning_rate": 1.605126393375107e-05, "loss": 0.0814, "step": 25049 }, { "epoch": 2.59, "learning_rate": 1.6049084804643578e-05, "loss": 0.5426, "step": 25050 }, { "epoch": 2.59, "learning_rate": 1.6046905769450378e-05, "loss": 0.5018, "step": 25051 }, { "epoch": 2.59, "learning_rate": 1.6044726828186156e-05, "loss": 0.4968, "step": 25052 }, { "epoch": 2.59, "learning_rate": 1.6042547980865555e-05, "loss": 0.2517, "step": 25053 }, { "epoch": 2.59, "learning_rate": 1.604036922750327e-05, "loss": 0.3021, "step": 25054 }, { "epoch": 2.59, "learning_rate": 1.603819056811394e-05, "loss": 0.4469, "step": 25055 }, { "epoch": 2.59, "learning_rate": 1.6036012002712243e-05, "loss": 0.1433, "step": 25056 }, { "epoch": 2.59, "learning_rate": 1.6033833531312843e-05, "loss": 0.6215, "step": 25057 }, { "epoch": 2.59, "learning_rate": 1.6031655153930414e-05, "loss": 0.1911, "step": 25058 }, { "epoch": 2.59, "learning_rate": 1.602947687057961e-05, "loss": 0.3969, "step": 25059 }, { "epoch": 2.59, "learning_rate": 1.60272986812751e-05, "loss": 0.4076, "step": 25060 }, { "epoch": 2.59, "learning_rate": 1.6025120586031547e-05, "loss": 0.1167, "step": 25061 }, { "epoch": 2.59, "learning_rate": 1.6022942584863623e-05, "loss": 0.2892, "step": 25062 }, { "epoch": 2.59, "learning_rate": 1.6020764677785968e-05, "loss": 0.2847, "step": 25063 }, { "epoch": 2.59, "learning_rate": 1.6018586864813254e-05, "loss": 0.3098, "step": 25064 }, { "epoch": 2.59, "learning_rate": 1.6016409145960144e-05, "loss": 0.2667, "step": 25065 }, { "epoch": 2.59, "learning_rate": 1.6014231521241302e-05, "loss": 0.2691, "step": 25066 }, { "epoch": 2.59, "learning_rate": 1.6012053990671375e-05, "loss": 0.6324, "step": 25067 }, { "epoch": 2.59, "learning_rate": 1.6009876554265027e-05, "loss": 0.7467, "step": 25068 }, { "epoch": 2.59, "learning_rate": 1.6007699212036914e-05, "loss": 0.777, "step": 25069 }, { "epoch": 2.59, "learning_rate": 1.6005521964001697e-05, "loss": 0.4875, "step": 25070 }, { "epoch": 2.59, "learning_rate": 1.600334481017403e-05, "loss": 0.5192, "step": 25071 }, { "epoch": 2.59, "learning_rate": 1.6001167750568585e-05, "loss": 0.0829, "step": 25072 }, { "epoch": 2.59, "learning_rate": 1.599899078519999e-05, "loss": 0.1138, "step": 25073 }, { "epoch": 2.59, "learning_rate": 1.5996813914082915e-05, "loss": 0.2282, "step": 25074 }, { "epoch": 2.59, "learning_rate": 1.599463713723201e-05, "loss": 0.3582, "step": 25075 }, { "epoch": 2.59, "learning_rate": 1.599246045466193e-05, "loss": 0.2781, "step": 25076 }, { "epoch": 2.59, "learning_rate": 1.5990283866387334e-05, "loss": 0.577, "step": 25077 }, { "epoch": 2.59, "learning_rate": 1.5988107372422863e-05, "loss": 0.1465, "step": 25078 }, { "epoch": 2.59, "learning_rate": 1.598593097278319e-05, "loss": 0.5589, "step": 25079 }, { "epoch": 2.59, "learning_rate": 1.5983754667482937e-05, "loss": 0.1819, "step": 25080 }, { "epoch": 2.59, "learning_rate": 1.598157845653677e-05, "loss": 0.0996, "step": 25081 }, { "epoch": 2.59, "learning_rate": 1.5979402339959344e-05, "loss": 0.5012, "step": 25082 }, { "epoch": 2.59, "learning_rate": 1.5977226317765296e-05, "loss": 0.1007, "step": 25083 }, { "epoch": 2.59, "learning_rate": 1.5975050389969275e-05, "loss": 0.8664, "step": 25084 }, { "epoch": 2.59, "learning_rate": 1.597287455658593e-05, "loss": 0.4375, "step": 25085 }, { "epoch": 2.59, "learning_rate": 1.5970698817629917e-05, "loss": 0.2248, "step": 25086 }, { "epoch": 2.59, "learning_rate": 1.5968523173115876e-05, "loss": 0.2542, "step": 25087 }, { "epoch": 2.59, "learning_rate": 1.5966347623058454e-05, "loss": 0.2985, "step": 25088 }, { "epoch": 2.59, "learning_rate": 1.5964172167472294e-05, "loss": 0.1733, "step": 25089 }, { "epoch": 2.59, "learning_rate": 1.596199680637206e-05, "loss": 0.1514, "step": 25090 }, { "epoch": 2.59, "learning_rate": 1.595982153977236e-05, "loss": 0.3717, "step": 25091 }, { "epoch": 2.59, "learning_rate": 1.5957646367687872e-05, "loss": 0.2602, "step": 25092 }, { "epoch": 2.59, "learning_rate": 1.5955471290133207e-05, "loss": 0.2956, "step": 25093 }, { "epoch": 2.59, "learning_rate": 1.595329630712303e-05, "loss": 0.2038, "step": 25094 }, { "epoch": 2.59, "learning_rate": 1.595112141867197e-05, "loss": 0.6171, "step": 25095 }, { "epoch": 2.59, "learning_rate": 1.5948946624794678e-05, "loss": 0.2143, "step": 25096 }, { "epoch": 2.59, "learning_rate": 1.5946771925505785e-05, "loss": 0.716, "step": 25097 }, { "epoch": 2.59, "learning_rate": 1.5944597320819937e-05, "loss": 0.3449, "step": 25098 }, { "epoch": 2.59, "learning_rate": 1.5942422810751773e-05, "loss": 0.1182, "step": 25099 }, { "epoch": 2.59, "learning_rate": 1.594024839531594e-05, "loss": 0.6399, "step": 25100 }, { "epoch": 2.59, "learning_rate": 1.5938074074527052e-05, "loss": 0.5659, "step": 25101 }, { "epoch": 2.59, "learning_rate": 1.5935899848399762e-05, "loss": 0.5717, "step": 25102 }, { "epoch": 2.6, "learning_rate": 1.5933725716948698e-05, "loss": 0.6936, "step": 25103 }, { "epoch": 2.6, "learning_rate": 1.5931551680188515e-05, "loss": 0.3583, "step": 25104 }, { "epoch": 2.6, "learning_rate": 1.592937773813383e-05, "loss": 0.9009, "step": 25105 }, { "epoch": 2.6, "learning_rate": 1.5927203890799276e-05, "loss": 0.7457, "step": 25106 }, { "epoch": 2.6, "learning_rate": 1.592503013819949e-05, "loss": 0.377, "step": 25107 }, { "epoch": 2.6, "learning_rate": 1.592285648034911e-05, "loss": 0.277, "step": 25108 }, { "epoch": 2.6, "learning_rate": 1.5920682917262765e-05, "loss": 0.0761, "step": 25109 }, { "epoch": 2.6, "learning_rate": 1.5918509448955104e-05, "loss": 0.2028, "step": 25110 }, { "epoch": 2.6, "learning_rate": 1.591633607544073e-05, "loss": 0.1536, "step": 25111 }, { "epoch": 2.6, "learning_rate": 1.5914162796734285e-05, "loss": 0.2586, "step": 25112 }, { "epoch": 2.6, "learning_rate": 1.59119896128504e-05, "loss": 0.429, "step": 25113 }, { "epoch": 2.6, "learning_rate": 1.5909816523803703e-05, "loss": 0.3165, "step": 25114 }, { "epoch": 2.6, "learning_rate": 1.590764352960883e-05, "loss": 0.3534, "step": 25115 }, { "epoch": 2.6, "learning_rate": 1.5905470630280408e-05, "loss": 0.7114, "step": 25116 }, { "epoch": 2.6, "learning_rate": 1.590329782583305e-05, "loss": 0.5264, "step": 25117 }, { "epoch": 2.6, "learning_rate": 1.5901125116281396e-05, "loss": 0.5336, "step": 25118 }, { "epoch": 2.6, "learning_rate": 1.5898952501640067e-05, "loss": 0.4567, "step": 25119 }, { "epoch": 2.6, "learning_rate": 1.5896779981923702e-05, "loss": 0.2065, "step": 25120 }, { "epoch": 2.6, "learning_rate": 1.58946075571469e-05, "loss": 0.7626, "step": 25121 }, { "epoch": 2.6, "learning_rate": 1.5892435227324303e-05, "loss": 0.358, "step": 25122 }, { "epoch": 2.6, "learning_rate": 1.589026299247053e-05, "loss": 0.2539, "step": 25123 }, { "epoch": 2.6, "learning_rate": 1.5888090852600204e-05, "loss": 0.2471, "step": 25124 }, { "epoch": 2.6, "learning_rate": 1.588591880772795e-05, "loss": 0.3917, "step": 25125 }, { "epoch": 2.6, "learning_rate": 1.588374685786839e-05, "loss": 0.794, "step": 25126 }, { "epoch": 2.6, "learning_rate": 1.588157500303614e-05, "loss": 0.2052, "step": 25127 }, { "epoch": 2.6, "learning_rate": 1.5879403243245837e-05, "loss": 0.8272, "step": 25128 }, { "epoch": 2.6, "learning_rate": 1.5877231578512077e-05, "loss": 0.2352, "step": 25129 }, { "epoch": 2.6, "learning_rate": 1.58750600088495e-05, "loss": 0.6592, "step": 25130 }, { "epoch": 2.6, "learning_rate": 1.58728885342727e-05, "loss": 0.0754, "step": 25131 }, { "epoch": 2.6, "learning_rate": 1.587071715479631e-05, "loss": 0.3698, "step": 25132 }, { "epoch": 2.6, "learning_rate": 1.586854587043495e-05, "loss": 0.3147, "step": 25133 }, { "epoch": 2.6, "learning_rate": 1.586637468120323e-05, "loss": 0.1247, "step": 25134 }, { "epoch": 2.6, "learning_rate": 1.5864203587115766e-05, "loss": 0.4618, "step": 25135 }, { "epoch": 2.6, "learning_rate": 1.586203258818718e-05, "loss": 0.3153, "step": 25136 }, { "epoch": 2.6, "learning_rate": 1.5859861684432078e-05, "loss": 0.3135, "step": 25137 }, { "epoch": 2.6, "learning_rate": 1.5857690875865093e-05, "loss": 0.4034, "step": 25138 }, { "epoch": 2.6, "learning_rate": 1.585552016250081e-05, "loss": 0.7726, "step": 25139 }, { "epoch": 2.6, "learning_rate": 1.5853349544353862e-05, "loss": 0.2934, "step": 25140 }, { "epoch": 2.6, "learning_rate": 1.585117902143886e-05, "loss": 0.2673, "step": 25141 }, { "epoch": 2.6, "learning_rate": 1.5849008593770403e-05, "loss": 0.6708, "step": 25142 }, { "epoch": 2.6, "learning_rate": 1.5846838261363108e-05, "loss": 0.4433, "step": 25143 }, { "epoch": 2.6, "learning_rate": 1.5844668024231587e-05, "loss": 0.2019, "step": 25144 }, { "epoch": 2.6, "learning_rate": 1.584249788239044e-05, "loss": 0.8441, "step": 25145 }, { "epoch": 2.6, "learning_rate": 1.58403278358543e-05, "loss": 0.3887, "step": 25146 }, { "epoch": 2.6, "learning_rate": 1.5838157884637747e-05, "loss": 0.4404, "step": 25147 }, { "epoch": 2.6, "learning_rate": 1.583598802875542e-05, "loss": 0.2696, "step": 25148 }, { "epoch": 2.6, "learning_rate": 1.5833818268221894e-05, "loss": 0.0856, "step": 25149 }, { "epoch": 2.6, "learning_rate": 1.583164860305179e-05, "loss": 0.2906, "step": 25150 }, { "epoch": 2.6, "learning_rate": 1.5829479033259708e-05, "loss": 0.4477, "step": 25151 }, { "epoch": 2.6, "learning_rate": 1.582730955886026e-05, "loss": 0.4961, "step": 25152 }, { "epoch": 2.6, "learning_rate": 1.582514017986805e-05, "loss": 0.3111, "step": 25153 }, { "epoch": 2.6, "learning_rate": 1.5822970896297687e-05, "loss": 0.4311, "step": 25154 }, { "epoch": 2.6, "learning_rate": 1.5820801708163758e-05, "loss": 0.2542, "step": 25155 }, { "epoch": 2.6, "learning_rate": 1.5818632615480877e-05, "loss": 0.5932, "step": 25156 }, { "epoch": 2.6, "learning_rate": 1.581646361826364e-05, "loss": 0.7764, "step": 25157 }, { "epoch": 2.6, "learning_rate": 1.581429471652666e-05, "loss": 0.1452, "step": 25158 }, { "epoch": 2.6, "learning_rate": 1.5812125910284522e-05, "loss": 0.7206, "step": 25159 }, { "epoch": 2.6, "learning_rate": 1.5809957199551833e-05, "loss": 0.6481, "step": 25160 }, { "epoch": 2.6, "learning_rate": 1.580778858434319e-05, "loss": 0.2438, "step": 25161 }, { "epoch": 2.6, "learning_rate": 1.5805620064673193e-05, "loss": 0.571, "step": 25162 }, { "epoch": 2.6, "learning_rate": 1.580345164055644e-05, "loss": 0.7297, "step": 25163 }, { "epoch": 2.6, "learning_rate": 1.5801283312007528e-05, "loss": 0.3598, "step": 25164 }, { "epoch": 2.6, "learning_rate": 1.5799115079041056e-05, "loss": 0.1838, "step": 25165 }, { "epoch": 2.6, "learning_rate": 1.579694694167163e-05, "loss": 0.501, "step": 25166 }, { "epoch": 2.6, "learning_rate": 1.5794778899913827e-05, "loss": 0.4434, "step": 25167 }, { "epoch": 2.6, "learning_rate": 1.5792610953782252e-05, "loss": 0.154, "step": 25168 }, { "epoch": 2.6, "learning_rate": 1.5790443103291492e-05, "loss": 0.5025, "step": 25169 }, { "epoch": 2.6, "learning_rate": 1.5788275348456135e-05, "loss": 0.149, "step": 25170 }, { "epoch": 2.6, "learning_rate": 1.578610768929079e-05, "loss": 0.3952, "step": 25171 }, { "epoch": 2.6, "learning_rate": 1.578394012581004e-05, "loss": 0.9554, "step": 25172 }, { "epoch": 2.6, "learning_rate": 1.5781772658028484e-05, "loss": 0.9146, "step": 25173 }, { "epoch": 2.6, "learning_rate": 1.5779605285960702e-05, "loss": 0.1482, "step": 25174 }, { "epoch": 2.6, "learning_rate": 1.5777438009621288e-05, "loss": 0.1531, "step": 25175 }, { "epoch": 2.6, "learning_rate": 1.577527082902485e-05, "loss": 0.2752, "step": 25176 }, { "epoch": 2.6, "learning_rate": 1.5773103744185947e-05, "loss": 0.3175, "step": 25177 }, { "epoch": 2.6, "learning_rate": 1.5770936755119183e-05, "loss": 0.3507, "step": 25178 }, { "epoch": 2.6, "learning_rate": 1.576876986183915e-05, "loss": 0.2351, "step": 25179 }, { "epoch": 2.6, "learning_rate": 1.576660306436042e-05, "loss": 0.0996, "step": 25180 }, { "epoch": 2.6, "learning_rate": 1.576443636269759e-05, "loss": 0.1187, "step": 25181 }, { "epoch": 2.6, "learning_rate": 1.576226975686524e-05, "loss": 0.3784, "step": 25182 }, { "epoch": 2.6, "learning_rate": 1.5760103246877958e-05, "loss": 0.1806, "step": 25183 }, { "epoch": 2.6, "learning_rate": 1.5757936832750334e-05, "loss": 1.3848, "step": 25184 }, { "epoch": 2.6, "learning_rate": 1.5755770514496944e-05, "loss": 0.612, "step": 25185 }, { "epoch": 2.6, "learning_rate": 1.5753604292132385e-05, "loss": 0.122, "step": 25186 }, { "epoch": 2.6, "learning_rate": 1.575143816567122e-05, "loss": 0.1649, "step": 25187 }, { "epoch": 2.6, "learning_rate": 1.574927213512803e-05, "loss": 0.5344, "step": 25188 }, { "epoch": 2.6, "learning_rate": 1.5747106200517412e-05, "loss": 0.2115, "step": 25189 }, { "epoch": 2.6, "learning_rate": 1.5744940361853944e-05, "loss": 0.5333, "step": 25190 }, { "epoch": 2.6, "learning_rate": 1.5742774619152197e-05, "loss": 0.4615, "step": 25191 }, { "epoch": 2.6, "learning_rate": 1.5740608972426768e-05, "loss": 0.2819, "step": 25192 }, { "epoch": 2.6, "learning_rate": 1.573844342169221e-05, "loss": 0.1994, "step": 25193 }, { "epoch": 2.6, "learning_rate": 1.5736277966963117e-05, "loss": 0.4009, "step": 25194 }, { "epoch": 2.6, "learning_rate": 1.573411260825406e-05, "loss": 0.6421, "step": 25195 }, { "epoch": 2.6, "learning_rate": 1.5731947345579637e-05, "loss": 0.1687, "step": 25196 }, { "epoch": 2.6, "learning_rate": 1.5729782178954386e-05, "loss": 0.9265, "step": 25197 }, { "epoch": 2.61, "learning_rate": 1.5727617108392907e-05, "loss": 0.5365, "step": 25198 }, { "epoch": 2.61, "learning_rate": 1.5725452133909772e-05, "loss": 0.3884, "step": 25199 }, { "epoch": 2.61, "learning_rate": 1.5723287255519547e-05, "loss": 0.0958, "step": 25200 }, { "epoch": 2.61, "learning_rate": 1.5721122473236818e-05, "loss": 0.2147, "step": 25201 }, { "epoch": 2.61, "learning_rate": 1.571895778707615e-05, "loss": 0.2897, "step": 25202 }, { "epoch": 2.61, "learning_rate": 1.5716793197052114e-05, "loss": 0.2611, "step": 25203 }, { "epoch": 2.61, "learning_rate": 1.57146287031793e-05, "loss": 0.35, "step": 25204 }, { "epoch": 2.61, "learning_rate": 1.5712464305472245e-05, "loss": 0.4548, "step": 25205 }, { "epoch": 2.61, "learning_rate": 1.5710300003945555e-05, "loss": 0.5788, "step": 25206 }, { "epoch": 2.61, "learning_rate": 1.5708135798613768e-05, "loss": 0.1663, "step": 25207 }, { "epoch": 2.61, "learning_rate": 1.5705971689491466e-05, "loss": 0.3327, "step": 25208 }, { "epoch": 2.61, "learning_rate": 1.5703807676593218e-05, "loss": 0.4937, "step": 25209 }, { "epoch": 2.61, "learning_rate": 1.570164375993359e-05, "loss": 0.7094, "step": 25210 }, { "epoch": 2.61, "learning_rate": 1.5699479939527156e-05, "loss": 0.9314, "step": 25211 }, { "epoch": 2.61, "learning_rate": 1.5697316215388474e-05, "loss": 0.5385, "step": 25212 }, { "epoch": 2.61, "learning_rate": 1.5695152587532112e-05, "loss": 0.2951, "step": 25213 }, { "epoch": 2.61, "learning_rate": 1.569298905597265e-05, "loss": 0.6803, "step": 25214 }, { "epoch": 2.61, "learning_rate": 1.5690825620724626e-05, "loss": 0.7562, "step": 25215 }, { "epoch": 2.61, "learning_rate": 1.568866228180261e-05, "loss": 0.4854, "step": 25216 }, { "epoch": 2.61, "learning_rate": 1.568649903922119e-05, "loss": 0.2982, "step": 25217 }, { "epoch": 2.61, "learning_rate": 1.5684335892994894e-05, "loss": 1.0818, "step": 25218 }, { "epoch": 2.61, "learning_rate": 1.56821728431383e-05, "loss": 0.1097, "step": 25219 }, { "epoch": 2.61, "learning_rate": 1.568000988966597e-05, "loss": 0.4883, "step": 25220 }, { "epoch": 2.61, "learning_rate": 1.567784703259246e-05, "loss": 0.0911, "step": 25221 }, { "epoch": 2.61, "learning_rate": 1.567568427193233e-05, "loss": 0.5677, "step": 25222 }, { "epoch": 2.61, "learning_rate": 1.5673521607700148e-05, "loss": 0.3566, "step": 25223 }, { "epoch": 2.61, "learning_rate": 1.5671359039910476e-05, "loss": 0.4146, "step": 25224 }, { "epoch": 2.61, "learning_rate": 1.5669196568577847e-05, "loss": 0.2951, "step": 25225 }, { "epoch": 2.61, "learning_rate": 1.566703419371684e-05, "loss": 0.3211, "step": 25226 }, { "epoch": 2.61, "learning_rate": 1.5664871915342e-05, "loss": 0.3083, "step": 25227 }, { "epoch": 2.61, "learning_rate": 1.566270973346789e-05, "loss": 0.1176, "step": 25228 }, { "epoch": 2.61, "learning_rate": 1.5660547648109074e-05, "loss": 0.4198, "step": 25229 }, { "epoch": 2.61, "learning_rate": 1.565838565928009e-05, "loss": 0.1692, "step": 25230 }, { "epoch": 2.61, "learning_rate": 1.5656223766995493e-05, "loss": 0.0566, "step": 25231 }, { "epoch": 2.61, "learning_rate": 1.5654061971269844e-05, "loss": 0.2159, "step": 25232 }, { "epoch": 2.61, "learning_rate": 1.565190027211769e-05, "loss": 0.0671, "step": 25233 }, { "epoch": 2.61, "learning_rate": 1.56497386695536e-05, "loss": 0.8852, "step": 25234 }, { "epoch": 2.61, "learning_rate": 1.5647577163592097e-05, "loss": 0.1566, "step": 25235 }, { "epoch": 2.61, "learning_rate": 1.5645415754247747e-05, "loss": 0.6911, "step": 25236 }, { "epoch": 2.61, "learning_rate": 1.5643254441535105e-05, "loss": 0.7943, "step": 25237 }, { "epoch": 2.61, "learning_rate": 1.5641093225468708e-05, "loss": 0.1375, "step": 25238 }, { "epoch": 2.61, "learning_rate": 1.5638932106063115e-05, "loss": 0.1811, "step": 25239 }, { "epoch": 2.61, "learning_rate": 1.563677108333287e-05, "loss": 0.1519, "step": 25240 }, { "epoch": 2.61, "learning_rate": 1.5634610157292522e-05, "loss": 0.5493, "step": 25241 }, { "epoch": 2.61, "learning_rate": 1.5632449327956626e-05, "loss": 0.3044, "step": 25242 }, { "epoch": 2.61, "learning_rate": 1.563028859533971e-05, "loss": 0.1095, "step": 25243 }, { "epoch": 2.61, "learning_rate": 1.562812795945634e-05, "loss": 0.5269, "step": 25244 }, { "epoch": 2.61, "learning_rate": 1.5625967420321044e-05, "loss": 0.4286, "step": 25245 }, { "epoch": 2.61, "learning_rate": 1.562380697794836e-05, "loss": 0.0758, "step": 25246 }, { "epoch": 2.61, "learning_rate": 1.5621646632352856e-05, "loss": 0.0917, "step": 25247 }, { "epoch": 2.61, "learning_rate": 1.5619486383549055e-05, "loss": 0.5656, "step": 25248 }, { "epoch": 2.61, "learning_rate": 1.561732623155151e-05, "loss": 0.1358, "step": 25249 }, { "epoch": 2.61, "learning_rate": 1.561516617637476e-05, "loss": 0.3236, "step": 25250 }, { "epoch": 2.61, "learning_rate": 1.5613006218033342e-05, "loss": 0.3424, "step": 25251 }, { "epoch": 2.61, "learning_rate": 1.5610846356541818e-05, "loss": 0.462, "step": 25252 }, { "epoch": 2.61, "learning_rate": 1.560868659191469e-05, "loss": 0.341, "step": 25253 }, { "epoch": 2.61, "learning_rate": 1.560652692416652e-05, "loss": 0.5384, "step": 25254 }, { "epoch": 2.61, "learning_rate": 1.5604367353311858e-05, "loss": 0.2529, "step": 25255 }, { "epoch": 2.61, "learning_rate": 1.5602207879365208e-05, "loss": 0.5763, "step": 25256 }, { "epoch": 2.61, "learning_rate": 1.560004850234113e-05, "loss": 0.3493, "step": 25257 }, { "epoch": 2.61, "learning_rate": 1.559788922225416e-05, "loss": 0.6918, "step": 25258 }, { "epoch": 2.61, "learning_rate": 1.5595730039118823e-05, "loss": 0.2979, "step": 25259 }, { "epoch": 2.61, "learning_rate": 1.559357095294966e-05, "loss": 0.2362, "step": 25260 }, { "epoch": 2.61, "learning_rate": 1.5591411963761214e-05, "loss": 0.1256, "step": 25261 }, { "epoch": 2.61, "learning_rate": 1.5589253071568017e-05, "loss": 0.2273, "step": 25262 }, { "epoch": 2.61, "learning_rate": 1.558709427638459e-05, "loss": 0.1352, "step": 25263 }, { "epoch": 2.61, "learning_rate": 1.558493557822547e-05, "loss": 0.1697, "step": 25264 }, { "epoch": 2.61, "learning_rate": 1.5582776977105184e-05, "loss": 0.1836, "step": 25265 }, { "epoch": 2.61, "learning_rate": 1.558061847303828e-05, "loss": 0.3209, "step": 25266 }, { "epoch": 2.61, "learning_rate": 1.557846006603928e-05, "loss": 0.4872, "step": 25267 }, { "epoch": 2.61, "learning_rate": 1.557630175612271e-05, "loss": 0.3091, "step": 25268 }, { "epoch": 2.61, "learning_rate": 1.5574143543303102e-05, "loss": 0.4938, "step": 25269 }, { "epoch": 2.61, "learning_rate": 1.557198542759498e-05, "loss": 0.2882, "step": 25270 }, { "epoch": 2.61, "learning_rate": 1.5569827409012876e-05, "loss": 0.4921, "step": 25271 }, { "epoch": 2.61, "learning_rate": 1.556766948757133e-05, "loss": 0.6742, "step": 25272 }, { "epoch": 2.61, "learning_rate": 1.5565511663284845e-05, "loss": 0.3294, "step": 25273 }, { "epoch": 2.61, "learning_rate": 1.5563353936167958e-05, "loss": 0.2824, "step": 25274 }, { "epoch": 2.61, "learning_rate": 1.5561196306235192e-05, "loss": 0.5867, "step": 25275 }, { "epoch": 2.61, "learning_rate": 1.5559038773501074e-05, "loss": 0.749, "step": 25276 }, { "epoch": 2.61, "learning_rate": 1.5556881337980137e-05, "loss": 0.1332, "step": 25277 }, { "epoch": 2.61, "learning_rate": 1.5554723999686883e-05, "loss": 0.2135, "step": 25278 }, { "epoch": 2.61, "learning_rate": 1.5552566758635855e-05, "loss": 0.305, "step": 25279 }, { "epoch": 2.61, "learning_rate": 1.5550409614841576e-05, "loss": 0.6022, "step": 25280 }, { "epoch": 2.61, "learning_rate": 1.5548252568318545e-05, "loss": 0.1108, "step": 25281 }, { "epoch": 2.61, "learning_rate": 1.5546095619081314e-05, "loss": 0.4324, "step": 25282 }, { "epoch": 2.61, "learning_rate": 1.5543938767144367e-05, "loss": 0.4784, "step": 25283 }, { "epoch": 2.61, "learning_rate": 1.5541782012522245e-05, "loss": 0.6965, "step": 25284 }, { "epoch": 2.61, "learning_rate": 1.5539625355229465e-05, "loss": 0.4656, "step": 25285 }, { "epoch": 2.61, "learning_rate": 1.5537468795280543e-05, "loss": 0.1304, "step": 25286 }, { "epoch": 2.61, "learning_rate": 1.5535312332689994e-05, "loss": 0.2866, "step": 25287 }, { "epoch": 2.61, "learning_rate": 1.5533155967472338e-05, "loss": 0.1985, "step": 25288 }, { "epoch": 2.61, "learning_rate": 1.5530999699642097e-05, "loss": 0.7447, "step": 25289 }, { "epoch": 2.61, "learning_rate": 1.552884352921379e-05, "loss": 0.1293, "step": 25290 }, { "epoch": 2.61, "learning_rate": 1.5526687456201906e-05, "loss": 0.5455, "step": 25291 }, { "epoch": 2.61, "learning_rate": 1.552453148062099e-05, "loss": 0.2252, "step": 25292 }, { "epoch": 2.61, "learning_rate": 1.552237560248553e-05, "loss": 0.3893, "step": 25293 }, { "epoch": 2.62, "learning_rate": 1.552021982181005e-05, "loss": 0.2016, "step": 25294 }, { "epoch": 2.62, "learning_rate": 1.551806413860906e-05, "loss": 0.8374, "step": 25295 }, { "epoch": 2.62, "learning_rate": 1.5515908552897077e-05, "loss": 0.2845, "step": 25296 }, { "epoch": 2.62, "learning_rate": 1.5513753064688606e-05, "loss": 0.3497, "step": 25297 }, { "epoch": 2.62, "learning_rate": 1.5511597673998164e-05, "loss": 0.9595, "step": 25298 }, { "epoch": 2.62, "learning_rate": 1.550944238084025e-05, "loss": 0.2949, "step": 25299 }, { "epoch": 2.62, "learning_rate": 1.5507287185229396e-05, "loss": 0.3938, "step": 25300 }, { "epoch": 2.62, "learning_rate": 1.550513208718008e-05, "loss": 0.1962, "step": 25301 }, { "epoch": 2.62, "learning_rate": 1.5502977086706823e-05, "loss": 0.3152, "step": 25302 }, { "epoch": 2.62, "learning_rate": 1.5500822183824136e-05, "loss": 0.1239, "step": 25303 }, { "epoch": 2.62, "learning_rate": 1.549866737854652e-05, "loss": 0.8886, "step": 25304 }, { "epoch": 2.62, "learning_rate": 1.5496512670888488e-05, "loss": 0.5951, "step": 25305 }, { "epoch": 2.62, "learning_rate": 1.5494358060864532e-05, "loss": 0.0959, "step": 25306 }, { "epoch": 2.62, "learning_rate": 1.5492203548489163e-05, "loss": 0.3945, "step": 25307 }, { "epoch": 2.62, "learning_rate": 1.5490049133776887e-05, "loss": 0.3382, "step": 25308 }, { "epoch": 2.62, "learning_rate": 1.5487894816742204e-05, "loss": 0.4655, "step": 25309 }, { "epoch": 2.62, "learning_rate": 1.5485740597399628e-05, "loss": 0.4688, "step": 25310 }, { "epoch": 2.62, "learning_rate": 1.548358647576364e-05, "loss": 0.6677, "step": 25311 }, { "epoch": 2.62, "learning_rate": 1.548143245184875e-05, "loss": 0.1947, "step": 25312 }, { "epoch": 2.62, "learning_rate": 1.547927852566946e-05, "loss": 0.1564, "step": 25313 }, { "epoch": 2.62, "learning_rate": 1.5477124697240267e-05, "loss": 0.4879, "step": 25314 }, { "epoch": 2.62, "learning_rate": 1.5474970966575677e-05, "loss": 0.6941, "step": 25315 }, { "epoch": 2.62, "learning_rate": 1.5472817333690183e-05, "loss": 0.4204, "step": 25316 }, { "epoch": 2.62, "learning_rate": 1.5470663798598295e-05, "loss": 0.4813, "step": 25317 }, { "epoch": 2.62, "learning_rate": 1.5468510361314484e-05, "loss": 0.3009, "step": 25318 }, { "epoch": 2.62, "learning_rate": 1.5466357021853264e-05, "loss": 0.4741, "step": 25319 }, { "epoch": 2.62, "learning_rate": 1.546420378022914e-05, "loss": 0.1333, "step": 25320 }, { "epoch": 2.62, "learning_rate": 1.5462050636456584e-05, "loss": 0.3807, "step": 25321 }, { "epoch": 2.62, "learning_rate": 1.5459897590550096e-05, "loss": 0.0953, "step": 25322 }, { "epoch": 2.62, "learning_rate": 1.5457744642524177e-05, "loss": 0.3427, "step": 25323 }, { "epoch": 2.62, "learning_rate": 1.5455591792393325e-05, "loss": 0.3317, "step": 25324 }, { "epoch": 2.62, "learning_rate": 1.545343904017202e-05, "loss": 0.6794, "step": 25325 }, { "epoch": 2.62, "learning_rate": 1.5451286385874764e-05, "loss": 0.2883, "step": 25326 }, { "epoch": 2.62, "learning_rate": 1.544913382951604e-05, "loss": 0.1098, "step": 25327 }, { "epoch": 2.62, "learning_rate": 1.5446981371110343e-05, "loss": 0.5853, "step": 25328 }, { "epoch": 2.62, "learning_rate": 1.5444829010672176e-05, "loss": 0.2129, "step": 25329 }, { "epoch": 2.62, "learning_rate": 1.544267674821601e-05, "loss": 0.5297, "step": 25330 }, { "epoch": 2.62, "learning_rate": 1.544052458375633e-05, "loss": 0.1692, "step": 25331 }, { "epoch": 2.62, "learning_rate": 1.5438372517307633e-05, "loss": 0.5333, "step": 25332 }, { "epoch": 2.62, "learning_rate": 1.54362205488844e-05, "loss": 0.4505, "step": 25333 }, { "epoch": 2.62, "learning_rate": 1.543406867850113e-05, "loss": 0.1855, "step": 25334 }, { "epoch": 2.62, "learning_rate": 1.5431916906172297e-05, "loss": 0.4352, "step": 25335 }, { "epoch": 2.62, "learning_rate": 1.542976523191239e-05, "loss": 0.0993, "step": 25336 }, { "epoch": 2.62, "learning_rate": 1.5427613655735896e-05, "loss": 0.3175, "step": 25337 }, { "epoch": 2.62, "learning_rate": 1.542546217765731e-05, "loss": 0.2568, "step": 25338 }, { "epoch": 2.62, "learning_rate": 1.5423310797691092e-05, "loss": 0.3572, "step": 25339 }, { "epoch": 2.62, "learning_rate": 1.5421159515851727e-05, "loss": 0.3975, "step": 25340 }, { "epoch": 2.62, "learning_rate": 1.5419008332153712e-05, "loss": 1.2139, "step": 25341 }, { "epoch": 2.62, "learning_rate": 1.5416857246611532e-05, "loss": 0.6236, "step": 25342 }, { "epoch": 2.62, "learning_rate": 1.5414706259239644e-05, "loss": 0.269, "step": 25343 }, { "epoch": 2.62, "learning_rate": 1.5412555370052542e-05, "loss": 0.3947, "step": 25344 }, { "epoch": 2.62, "learning_rate": 1.5410404579064698e-05, "loss": 0.5015, "step": 25345 }, { "epoch": 2.62, "learning_rate": 1.5408253886290608e-05, "loss": 0.782, "step": 25346 }, { "epoch": 2.62, "learning_rate": 1.540610329174473e-05, "loss": 0.1476, "step": 25347 }, { "epoch": 2.62, "learning_rate": 1.5403952795441565e-05, "loss": 0.3799, "step": 25348 }, { "epoch": 2.62, "learning_rate": 1.540180239739556e-05, "loss": 0.2021, "step": 25349 }, { "epoch": 2.62, "learning_rate": 1.5399652097621208e-05, "loss": 0.59, "step": 25350 }, { "epoch": 2.62, "learning_rate": 1.539750189613298e-05, "loss": 0.2112, "step": 25351 }, { "epoch": 2.62, "learning_rate": 1.5395351792945354e-05, "loss": 0.4052, "step": 25352 }, { "epoch": 2.62, "learning_rate": 1.5393201788072803e-05, "loss": 0.3269, "step": 25353 }, { "epoch": 2.62, "learning_rate": 1.53910518815298e-05, "loss": 0.2688, "step": 25354 }, { "epoch": 2.62, "learning_rate": 1.5388902073330827e-05, "loss": 0.3138, "step": 25355 }, { "epoch": 2.62, "learning_rate": 1.5386752363490337e-05, "loss": 0.1285, "step": 25356 }, { "epoch": 2.62, "learning_rate": 1.538460275202281e-05, "loss": 0.6236, "step": 25357 }, { "epoch": 2.62, "learning_rate": 1.5382453238942728e-05, "loss": 0.3425, "step": 25358 }, { "epoch": 2.62, "learning_rate": 1.5380303824264543e-05, "loss": 0.7894, "step": 25359 }, { "epoch": 2.62, "learning_rate": 1.5378154508002727e-05, "loss": 0.8382, "step": 25360 }, { "epoch": 2.62, "learning_rate": 1.5376005290171756e-05, "loss": 0.1245, "step": 25361 }, { "epoch": 2.62, "learning_rate": 1.5373856170786093e-05, "loss": 0.4567, "step": 25362 }, { "epoch": 2.62, "learning_rate": 1.5371707149860216e-05, "loss": 0.2847, "step": 25363 }, { "epoch": 2.62, "learning_rate": 1.5369558227408578e-05, "loss": 0.2209, "step": 25364 }, { "epoch": 2.62, "learning_rate": 1.5367409403445648e-05, "loss": 0.2749, "step": 25365 }, { "epoch": 2.62, "learning_rate": 1.5365260677985897e-05, "loss": 0.7475, "step": 25366 }, { "epoch": 2.62, "learning_rate": 1.5363112051043796e-05, "loss": 0.2604, "step": 25367 }, { "epoch": 2.62, "learning_rate": 1.5360963522633806e-05, "loss": 0.2841, "step": 25368 }, { "epoch": 2.62, "learning_rate": 1.5358815092770366e-05, "loss": 0.5266, "step": 25369 }, { "epoch": 2.62, "learning_rate": 1.535666676146796e-05, "loss": 0.5485, "step": 25370 }, { "epoch": 2.62, "learning_rate": 1.535451852874105e-05, "loss": 0.2802, "step": 25371 }, { "epoch": 2.62, "learning_rate": 1.535237039460409e-05, "loss": 0.1034, "step": 25372 }, { "epoch": 2.62, "learning_rate": 1.5350222359071552e-05, "loss": 0.3433, "step": 25373 }, { "epoch": 2.62, "learning_rate": 1.5348074422157882e-05, "loss": 0.4825, "step": 25374 }, { "epoch": 2.62, "learning_rate": 1.5345926583877547e-05, "loss": 0.1013, "step": 25375 }, { "epoch": 2.62, "learning_rate": 1.534377884424502e-05, "loss": 0.7063, "step": 25376 }, { "epoch": 2.62, "learning_rate": 1.534163120327473e-05, "loss": 0.319, "step": 25377 }, { "epoch": 2.62, "learning_rate": 1.5339483660981155e-05, "loss": 0.1705, "step": 25378 }, { "epoch": 2.62, "learning_rate": 1.5337336217378742e-05, "loss": 0.3216, "step": 25379 }, { "epoch": 2.62, "learning_rate": 1.5335188872481956e-05, "loss": 0.3668, "step": 25380 }, { "epoch": 2.62, "learning_rate": 1.5333041626305247e-05, "loss": 0.22, "step": 25381 }, { "epoch": 2.62, "learning_rate": 1.5330894478863063e-05, "loss": 0.1337, "step": 25382 }, { "epoch": 2.62, "learning_rate": 1.5328747430169866e-05, "loss": 0.7062, "step": 25383 }, { "epoch": 2.62, "learning_rate": 1.532660048024011e-05, "loss": 0.2431, "step": 25384 }, { "epoch": 2.62, "learning_rate": 1.5324453629088243e-05, "loss": 0.4449, "step": 25385 }, { "epoch": 2.62, "learning_rate": 1.5322306876728737e-05, "loss": 0.3265, "step": 25386 }, { "epoch": 2.62, "learning_rate": 1.532016022317601e-05, "loss": 0.4839, "step": 25387 }, { "epoch": 2.62, "learning_rate": 1.5318013668444534e-05, "loss": 0.2385, "step": 25388 }, { "epoch": 2.62, "learning_rate": 1.5315867212548756e-05, "loss": 0.1364, "step": 25389 }, { "epoch": 2.63, "learning_rate": 1.5313720855503118e-05, "loss": 0.3909, "step": 25390 }, { "epoch": 2.63, "learning_rate": 1.5311574597322078e-05, "loss": 0.3315, "step": 25391 }, { "epoch": 2.63, "learning_rate": 1.5309428438020083e-05, "loss": 0.2526, "step": 25392 }, { "epoch": 2.63, "learning_rate": 1.5307282377611586e-05, "loss": 0.8211, "step": 25393 }, { "epoch": 2.63, "learning_rate": 1.530513641611102e-05, "loss": 0.3998, "step": 25394 }, { "epoch": 2.63, "learning_rate": 1.530299055353283e-05, "loss": 0.1809, "step": 25395 }, { "epoch": 2.63, "learning_rate": 1.5300844789891486e-05, "loss": 0.1021, "step": 25396 }, { "epoch": 2.63, "learning_rate": 1.52986991252014e-05, "loss": 0.3925, "step": 25397 }, { "epoch": 2.63, "learning_rate": 1.5296553559477035e-05, "loss": 0.3955, "step": 25398 }, { "epoch": 2.63, "learning_rate": 1.5294408092732827e-05, "loss": 0.257, "step": 25399 }, { "epoch": 2.63, "learning_rate": 1.529226272498323e-05, "loss": 0.3806, "step": 25400 }, { "epoch": 2.63, "learning_rate": 1.5290117456242673e-05, "loss": 0.3762, "step": 25401 }, { "epoch": 2.63, "learning_rate": 1.5287972286525608e-05, "loss": 0.0981, "step": 25402 }, { "epoch": 2.63, "learning_rate": 1.5285827215846464e-05, "loss": 0.5378, "step": 25403 }, { "epoch": 2.63, "learning_rate": 1.5283682244219695e-05, "loss": 0.1359, "step": 25404 }, { "epoch": 2.63, "learning_rate": 1.528153737165974e-05, "loss": 0.2451, "step": 25405 }, { "epoch": 2.63, "learning_rate": 1.5279392598181037e-05, "loss": 0.4754, "step": 25406 }, { "epoch": 2.63, "learning_rate": 1.5277247923798007e-05, "loss": 0.1303, "step": 25407 }, { "epoch": 2.63, "learning_rate": 1.5275103348525093e-05, "loss": 0.152, "step": 25408 }, { "epoch": 2.63, "learning_rate": 1.527295887237674e-05, "loss": 0.0715, "step": 25409 }, { "epoch": 2.63, "learning_rate": 1.527081449536739e-05, "loss": 0.07, "step": 25410 }, { "epoch": 2.63, "learning_rate": 1.5268670217511464e-05, "loss": 0.113, "step": 25411 }, { "epoch": 2.63, "learning_rate": 1.5266526038823404e-05, "loss": 0.3025, "step": 25412 }, { "epoch": 2.63, "learning_rate": 1.526438195931764e-05, "loss": 0.3307, "step": 25413 }, { "epoch": 2.63, "learning_rate": 1.5262237979008614e-05, "loss": 0.1291, "step": 25414 }, { "epoch": 2.63, "learning_rate": 1.526009409791076e-05, "loss": 0.1909, "step": 25415 }, { "epoch": 2.63, "learning_rate": 1.5257950316038498e-05, "loss": 0.3014, "step": 25416 }, { "epoch": 2.63, "learning_rate": 1.5255806633406261e-05, "loss": 0.4187, "step": 25417 }, { "epoch": 2.63, "learning_rate": 1.5253663050028498e-05, "loss": 0.2722, "step": 25418 }, { "epoch": 2.63, "learning_rate": 1.525151956591961e-05, "loss": 0.626, "step": 25419 }, { "epoch": 2.63, "learning_rate": 1.5249376181094044e-05, "loss": 0.3394, "step": 25420 }, { "epoch": 2.63, "learning_rate": 1.5247232895566225e-05, "loss": 0.4543, "step": 25421 }, { "epoch": 2.63, "learning_rate": 1.5245089709350585e-05, "loss": 0.3781, "step": 25422 }, { "epoch": 2.63, "learning_rate": 1.5242946622461545e-05, "loss": 0.3136, "step": 25423 }, { "epoch": 2.63, "learning_rate": 1.5240803634913549e-05, "loss": 0.2844, "step": 25424 }, { "epoch": 2.63, "learning_rate": 1.5238660746720998e-05, "loss": 0.6762, "step": 25425 }, { "epoch": 2.63, "learning_rate": 1.5236517957898325e-05, "loss": 0.7177, "step": 25426 }, { "epoch": 2.63, "learning_rate": 1.5234375268459968e-05, "loss": 0.3315, "step": 25427 }, { "epoch": 2.63, "learning_rate": 1.5232232678420333e-05, "loss": 0.1456, "step": 25428 }, { "epoch": 2.63, "learning_rate": 1.5230090187793855e-05, "loss": 0.1466, "step": 25429 }, { "epoch": 2.63, "learning_rate": 1.5227947796594967e-05, "loss": 0.6418, "step": 25430 }, { "epoch": 2.63, "learning_rate": 1.5225805504838067e-05, "loss": 0.3075, "step": 25431 }, { "epoch": 2.63, "learning_rate": 1.5223663312537585e-05, "loss": 0.3495, "step": 25432 }, { "epoch": 2.63, "learning_rate": 1.5221521219707943e-05, "loss": 0.3651, "step": 25433 }, { "epoch": 2.63, "learning_rate": 1.5219379226363575e-05, "loss": 0.5115, "step": 25434 }, { "epoch": 2.63, "learning_rate": 1.521723733251888e-05, "loss": 0.408, "step": 25435 }, { "epoch": 2.63, "learning_rate": 1.5215095538188281e-05, "loss": 0.1567, "step": 25436 }, { "epoch": 2.63, "learning_rate": 1.5212953843386199e-05, "loss": 0.1474, "step": 25437 }, { "epoch": 2.63, "learning_rate": 1.5210812248127051e-05, "loss": 0.1577, "step": 25438 }, { "epoch": 2.63, "learning_rate": 1.5208670752425256e-05, "loss": 0.381, "step": 25439 }, { "epoch": 2.63, "learning_rate": 1.520652935629523e-05, "loss": 0.4913, "step": 25440 }, { "epoch": 2.63, "learning_rate": 1.5204388059751385e-05, "loss": 0.8391, "step": 25441 }, { "epoch": 2.63, "learning_rate": 1.520224686280814e-05, "loss": 0.3738, "step": 25442 }, { "epoch": 2.63, "learning_rate": 1.5200105765479916e-05, "loss": 0.637, "step": 25443 }, { "epoch": 2.63, "learning_rate": 1.5197964767781117e-05, "loss": 0.3174, "step": 25444 }, { "epoch": 2.63, "learning_rate": 1.5195823869726146e-05, "loss": 0.1798, "step": 25445 }, { "epoch": 2.63, "learning_rate": 1.5193683071329426e-05, "loss": 0.3183, "step": 25446 }, { "epoch": 2.63, "learning_rate": 1.519154237260536e-05, "loss": 0.1869, "step": 25447 }, { "epoch": 2.63, "learning_rate": 1.5189401773568374e-05, "loss": 0.5417, "step": 25448 }, { "epoch": 2.63, "learning_rate": 1.5187261274232866e-05, "loss": 0.5783, "step": 25449 }, { "epoch": 2.63, "learning_rate": 1.5185120874613252e-05, "loss": 0.1733, "step": 25450 }, { "epoch": 2.63, "learning_rate": 1.5182980574723932e-05, "loss": 0.0917, "step": 25451 }, { "epoch": 2.63, "learning_rate": 1.5180840374579326e-05, "loss": 1.4372, "step": 25452 }, { "epoch": 2.63, "learning_rate": 1.5178700274193843e-05, "loss": 0.5269, "step": 25453 }, { "epoch": 2.63, "learning_rate": 1.517656027358187e-05, "loss": 0.3374, "step": 25454 }, { "epoch": 2.63, "learning_rate": 1.5174420372757827e-05, "loss": 0.2657, "step": 25455 }, { "epoch": 2.63, "learning_rate": 1.5172280571736129e-05, "loss": 0.1365, "step": 25456 }, { "epoch": 2.63, "learning_rate": 1.5170140870531156e-05, "loss": 0.7951, "step": 25457 }, { "epoch": 2.63, "learning_rate": 1.5168001269157328e-05, "loss": 0.5215, "step": 25458 }, { "epoch": 2.63, "learning_rate": 1.516586176762904e-05, "loss": 0.237, "step": 25459 }, { "epoch": 2.63, "learning_rate": 1.5163722365960704e-05, "loss": 0.1705, "step": 25460 }, { "epoch": 2.63, "learning_rate": 1.5161583064166714e-05, "loss": 0.624, "step": 25461 }, { "epoch": 2.63, "learning_rate": 1.5159443862261489e-05, "loss": 1.0452, "step": 25462 }, { "epoch": 2.63, "learning_rate": 1.5157304760259401e-05, "loss": 0.0655, "step": 25463 }, { "epoch": 2.63, "learning_rate": 1.5155165758174866e-05, "loss": 0.0817, "step": 25464 }, { "epoch": 2.63, "learning_rate": 1.5153026856022282e-05, "loss": 0.5457, "step": 25465 }, { "epoch": 2.63, "learning_rate": 1.5150888053816048e-05, "loss": 0.3602, "step": 25466 }, { "epoch": 2.63, "learning_rate": 1.514874935157056e-05, "loss": 0.5532, "step": 25467 }, { "epoch": 2.63, "learning_rate": 1.5146610749300232e-05, "loss": 0.4033, "step": 25468 }, { "epoch": 2.63, "learning_rate": 1.5144472247019428e-05, "loss": 0.3741, "step": 25469 }, { "epoch": 2.63, "learning_rate": 1.5142333844742563e-05, "loss": 0.4675, "step": 25470 }, { "epoch": 2.63, "learning_rate": 1.514019554248403e-05, "loss": 1.1691, "step": 25471 }, { "epoch": 2.63, "learning_rate": 1.5138057340258233e-05, "loss": 0.0805, "step": 25472 }, { "epoch": 2.63, "learning_rate": 1.5135919238079547e-05, "loss": 0.1733, "step": 25473 }, { "epoch": 2.63, "learning_rate": 1.5133781235962376e-05, "loss": 0.9005, "step": 25474 }, { "epoch": 2.63, "learning_rate": 1.5131643333921112e-05, "loss": 0.3317, "step": 25475 }, { "epoch": 2.63, "learning_rate": 1.5129505531970145e-05, "loss": 0.1188, "step": 25476 }, { "epoch": 2.63, "learning_rate": 1.5127367830123866e-05, "loss": 0.1323, "step": 25477 }, { "epoch": 2.63, "learning_rate": 1.5125230228396667e-05, "loss": 0.6899, "step": 25478 }, { "epoch": 2.63, "learning_rate": 1.5123092726802937e-05, "loss": 0.7533, "step": 25479 }, { "epoch": 2.63, "learning_rate": 1.512095532535707e-05, "loss": 0.0841, "step": 25480 }, { "epoch": 2.63, "learning_rate": 1.511881802407346e-05, "loss": 0.1338, "step": 25481 }, { "epoch": 2.63, "learning_rate": 1.5116680822966484e-05, "loss": 0.3221, "step": 25482 }, { "epoch": 2.63, "learning_rate": 1.5114543722050519e-05, "loss": 0.4025, "step": 25483 }, { "epoch": 2.63, "learning_rate": 1.5112406721339962e-05, "loss": 0.3663, "step": 25484 }, { "epoch": 2.63, "learning_rate": 1.5110269820849199e-05, "loss": 0.1685, "step": 25485 }, { "epoch": 2.64, "learning_rate": 1.5108133020592617e-05, "loss": 0.4411, "step": 25486 }, { "epoch": 2.64, "learning_rate": 1.51059963205846e-05, "loss": 0.487, "step": 25487 }, { "epoch": 2.64, "learning_rate": 1.5103859720839528e-05, "loss": 1.0595, "step": 25488 }, { "epoch": 2.64, "learning_rate": 1.5101723221371789e-05, "loss": 0.3887, "step": 25489 }, { "epoch": 2.64, "learning_rate": 1.5099586822195766e-05, "loss": 0.4614, "step": 25490 }, { "epoch": 2.64, "learning_rate": 1.5097450523325846e-05, "loss": 0.7783, "step": 25491 }, { "epoch": 2.64, "learning_rate": 1.509531432477639e-05, "loss": 0.1773, "step": 25492 }, { "epoch": 2.64, "learning_rate": 1.5093178226561806e-05, "loss": 0.3058, "step": 25493 }, { "epoch": 2.64, "learning_rate": 1.5091042228696443e-05, "loss": 0.2352, "step": 25494 }, { "epoch": 2.64, "learning_rate": 1.5088906331194695e-05, "loss": 0.3832, "step": 25495 }, { "epoch": 2.64, "learning_rate": 1.5086770534070943e-05, "loss": 0.6335, "step": 25496 }, { "epoch": 2.64, "learning_rate": 1.508463483733956e-05, "loss": 0.5514, "step": 25497 }, { "epoch": 2.64, "learning_rate": 1.5082499241014927e-05, "loss": 0.2069, "step": 25498 }, { "epoch": 2.64, "learning_rate": 1.508036374511142e-05, "loss": 0.5939, "step": 25499 }, { "epoch": 2.64, "learning_rate": 1.5078228349643412e-05, "loss": 0.2882, "step": 25500 }, { "epoch": 2.64, "learning_rate": 1.507609305462529e-05, "loss": 0.1053, "step": 25501 }, { "epoch": 2.64, "learning_rate": 1.5073957860071406e-05, "loss": 0.3832, "step": 25502 }, { "epoch": 2.64, "learning_rate": 1.5071822765996147e-05, "loss": 0.5649, "step": 25503 }, { "epoch": 2.64, "learning_rate": 1.5069687772413883e-05, "loss": 0.5919, "step": 25504 }, { "epoch": 2.64, "learning_rate": 1.506755287933899e-05, "loss": 0.2718, "step": 25505 }, { "epoch": 2.64, "learning_rate": 1.5065418086785845e-05, "loss": 0.5325, "step": 25506 }, { "epoch": 2.64, "learning_rate": 1.5063283394768803e-05, "loss": 0.3721, "step": 25507 }, { "epoch": 2.64, "learning_rate": 1.506114880330224e-05, "loss": 0.1348, "step": 25508 }, { "epoch": 2.64, "learning_rate": 1.5059014312400526e-05, "loss": 0.3785, "step": 25509 }, { "epoch": 2.64, "learning_rate": 1.5056879922078048e-05, "loss": 0.1278, "step": 25510 }, { "epoch": 2.64, "learning_rate": 1.5054745632349141e-05, "loss": 0.431, "step": 25511 }, { "epoch": 2.64, "learning_rate": 1.5052611443228193e-05, "loss": 0.3716, "step": 25512 }, { "epoch": 2.64, "learning_rate": 1.5050477354729566e-05, "loss": 0.4167, "step": 25513 }, { "epoch": 2.64, "learning_rate": 1.5048343366867627e-05, "loss": 0.1636, "step": 25514 }, { "epoch": 2.64, "learning_rate": 1.5046209479656746e-05, "loss": 0.6655, "step": 25515 }, { "epoch": 2.64, "learning_rate": 1.5044075693111281e-05, "loss": 0.326, "step": 25516 }, { "epoch": 2.64, "learning_rate": 1.5041942007245597e-05, "loss": 0.2248, "step": 25517 }, { "epoch": 2.64, "learning_rate": 1.5039808422074073e-05, "loss": 0.2744, "step": 25518 }, { "epoch": 2.64, "learning_rate": 1.503767493761105e-05, "loss": 0.2658, "step": 25519 }, { "epoch": 2.64, "learning_rate": 1.5035541553870905e-05, "loss": 0.1235, "step": 25520 }, { "epoch": 2.64, "learning_rate": 1.5033408270867983e-05, "loss": 0.0945, "step": 25521 }, { "epoch": 2.64, "learning_rate": 1.5031275088616656e-05, "loss": 0.2789, "step": 25522 }, { "epoch": 2.64, "learning_rate": 1.5029142007131282e-05, "loss": 0.2978, "step": 25523 }, { "epoch": 2.64, "learning_rate": 1.502700902642622e-05, "loss": 0.749, "step": 25524 }, { "epoch": 2.64, "learning_rate": 1.5024876146515829e-05, "loss": 0.3366, "step": 25525 }, { "epoch": 2.64, "learning_rate": 1.5022743367414471e-05, "loss": 0.4556, "step": 25526 }, { "epoch": 2.64, "learning_rate": 1.5020610689136498e-05, "loss": 0.2475, "step": 25527 }, { "epoch": 2.64, "learning_rate": 1.5018478111696271e-05, "loss": 0.4658, "step": 25528 }, { "epoch": 2.64, "learning_rate": 1.5016345635108154e-05, "loss": 0.1127, "step": 25529 }, { "epoch": 2.64, "learning_rate": 1.5014213259386483e-05, "loss": 0.2275, "step": 25530 }, { "epoch": 2.64, "learning_rate": 1.501208098454563e-05, "loss": 0.4619, "step": 25531 }, { "epoch": 2.64, "learning_rate": 1.500994881059993e-05, "loss": 0.2454, "step": 25532 }, { "epoch": 2.64, "learning_rate": 1.5007816737563747e-05, "loss": 0.3247, "step": 25533 }, { "epoch": 2.64, "learning_rate": 1.5005684765451435e-05, "loss": 0.3876, "step": 25534 }, { "epoch": 2.64, "learning_rate": 1.5003552894277345e-05, "loss": 0.3743, "step": 25535 }, { "epoch": 2.64, "learning_rate": 1.5001421124055828e-05, "loss": 0.4263, "step": 25536 }, { "epoch": 2.64, "learning_rate": 1.4999289454801234e-05, "loss": 0.1373, "step": 25537 }, { "epoch": 2.64, "learning_rate": 1.4997157886527912e-05, "loss": 0.3416, "step": 25538 }, { "epoch": 2.64, "learning_rate": 1.4995026419250229e-05, "loss": 0.1289, "step": 25539 }, { "epoch": 2.64, "learning_rate": 1.4992895052982502e-05, "loss": 0.1847, "step": 25540 }, { "epoch": 2.64, "learning_rate": 1.4990763787739094e-05, "loss": 0.515, "step": 25541 }, { "epoch": 2.64, "learning_rate": 1.4988632623534353e-05, "loss": 0.4374, "step": 25542 }, { "epoch": 2.64, "learning_rate": 1.4986501560382635e-05, "loss": 0.6251, "step": 25543 }, { "epoch": 2.64, "learning_rate": 1.4984370598298264e-05, "loss": 0.4141, "step": 25544 }, { "epoch": 2.64, "learning_rate": 1.4982239737295601e-05, "loss": 0.321, "step": 25545 }, { "epoch": 2.64, "learning_rate": 1.4980108977388983e-05, "loss": 0.7708, "step": 25546 }, { "epoch": 2.64, "learning_rate": 1.4977978318592757e-05, "loss": 0.2011, "step": 25547 }, { "epoch": 2.64, "learning_rate": 1.4975847760921279e-05, "loss": 1.3115, "step": 25548 }, { "epoch": 2.64, "learning_rate": 1.4973717304388866e-05, "loss": 0.795, "step": 25549 }, { "epoch": 2.64, "learning_rate": 1.4971586949009872e-05, "loss": 0.2313, "step": 25550 }, { "epoch": 2.64, "learning_rate": 1.4969456694798639e-05, "loss": 0.547, "step": 25551 }, { "epoch": 2.64, "learning_rate": 1.4967326541769504e-05, "loss": 0.28, "step": 25552 }, { "epoch": 2.64, "learning_rate": 1.4965196489936812e-05, "loss": 0.0541, "step": 25553 }, { "epoch": 2.64, "learning_rate": 1.49630665393149e-05, "loss": 0.1073, "step": 25554 }, { "epoch": 2.64, "learning_rate": 1.4960936689918105e-05, "loss": 0.1993, "step": 25555 }, { "epoch": 2.64, "learning_rate": 1.4958806941760777e-05, "loss": 0.2511, "step": 25556 }, { "epoch": 2.64, "learning_rate": 1.4956677294857226e-05, "loss": 0.0921, "step": 25557 }, { "epoch": 2.64, "learning_rate": 1.4954547749221822e-05, "loss": 0.157, "step": 25558 }, { "epoch": 2.64, "learning_rate": 1.4952418304868866e-05, "loss": 0.0827, "step": 25559 }, { "epoch": 2.64, "learning_rate": 1.495028896181271e-05, "loss": 0.3505, "step": 25560 }, { "epoch": 2.64, "learning_rate": 1.4948159720067687e-05, "loss": 0.6425, "step": 25561 }, { "epoch": 2.64, "learning_rate": 1.4946030579648133e-05, "loss": 0.6093, "step": 25562 }, { "epoch": 2.64, "learning_rate": 1.4943901540568375e-05, "loss": 0.1305, "step": 25563 }, { "epoch": 2.64, "learning_rate": 1.4941772602842752e-05, "loss": 0.1675, "step": 25564 }, { "epoch": 2.64, "learning_rate": 1.4939643766485594e-05, "loss": 0.2691, "step": 25565 }, { "epoch": 2.64, "learning_rate": 1.4937515031511226e-05, "loss": 0.1791, "step": 25566 }, { "epoch": 2.64, "learning_rate": 1.4935386397933996e-05, "loss": 0.5352, "step": 25567 }, { "epoch": 2.64, "learning_rate": 1.4933257865768209e-05, "loss": 0.3209, "step": 25568 }, { "epoch": 2.64, "learning_rate": 1.4931129435028217e-05, "loss": 0.4296, "step": 25569 }, { "epoch": 2.64, "learning_rate": 1.4929001105728321e-05, "loss": 0.1585, "step": 25570 }, { "epoch": 2.64, "learning_rate": 1.4926872877882866e-05, "loss": 0.4403, "step": 25571 }, { "epoch": 2.64, "learning_rate": 1.4924744751506174e-05, "loss": 0.5213, "step": 25572 }, { "epoch": 2.64, "learning_rate": 1.4922616726612574e-05, "loss": 0.2038, "step": 25573 }, { "epoch": 2.64, "learning_rate": 1.4920488803216391e-05, "loss": 0.9083, "step": 25574 }, { "epoch": 2.64, "learning_rate": 1.4918360981331949e-05, "loss": 0.5882, "step": 25575 }, { "epoch": 2.64, "learning_rate": 1.4916233260973573e-05, "loss": 0.3999, "step": 25576 }, { "epoch": 2.64, "learning_rate": 1.4914105642155593e-05, "loss": 0.3804, "step": 25577 }, { "epoch": 2.64, "learning_rate": 1.4911978124892318e-05, "loss": 0.1297, "step": 25578 }, { "epoch": 2.64, "learning_rate": 1.4909850709198074e-05, "loss": 0.0259, "step": 25579 }, { "epoch": 2.64, "learning_rate": 1.4907723395087179e-05, "loss": 0.1887, "step": 25580 }, { "epoch": 2.65, "learning_rate": 1.4905596182573977e-05, "loss": 0.6317, "step": 25581 }, { "epoch": 2.65, "learning_rate": 1.4903469071672756e-05, "loss": 0.1474, "step": 25582 }, { "epoch": 2.65, "learning_rate": 1.4901342062397848e-05, "loss": 0.0185, "step": 25583 }, { "epoch": 2.65, "learning_rate": 1.4899215154763574e-05, "loss": 0.3722, "step": 25584 }, { "epoch": 2.65, "learning_rate": 1.4897088348784246e-05, "loss": 0.5783, "step": 25585 }, { "epoch": 2.65, "learning_rate": 1.489496164447419e-05, "loss": 0.3635, "step": 25586 }, { "epoch": 2.65, "learning_rate": 1.4892835041847727e-05, "loss": 0.0603, "step": 25587 }, { "epoch": 2.65, "learning_rate": 1.4890708540919155e-05, "loss": 0.2299, "step": 25588 }, { "epoch": 2.65, "learning_rate": 1.4888582141702795e-05, "loss": 0.4929, "step": 25589 }, { "epoch": 2.65, "learning_rate": 1.4886455844212965e-05, "loss": 0.2645, "step": 25590 }, { "epoch": 2.65, "learning_rate": 1.488432964846398e-05, "loss": 0.2668, "step": 25591 }, { "epoch": 2.65, "learning_rate": 1.488220355447015e-05, "loss": 0.3127, "step": 25592 }, { "epoch": 2.65, "learning_rate": 1.4880077562245787e-05, "loss": 0.4325, "step": 25593 }, { "epoch": 2.65, "learning_rate": 1.4877951671805219e-05, "loss": 0.233, "step": 25594 }, { "epoch": 2.65, "learning_rate": 1.4875825883162725e-05, "loss": 0.5874, "step": 25595 }, { "epoch": 2.65, "learning_rate": 1.4873700196332645e-05, "loss": 0.3961, "step": 25596 }, { "epoch": 2.65, "learning_rate": 1.487157461132927e-05, "loss": 0.2098, "step": 25597 }, { "epoch": 2.65, "learning_rate": 1.486944912816691e-05, "loss": 0.067, "step": 25598 }, { "epoch": 2.65, "learning_rate": 1.4867323746859879e-05, "loss": 0.7413, "step": 25599 }, { "epoch": 2.65, "learning_rate": 1.4865198467422485e-05, "loss": 0.441, "step": 25600 }, { "epoch": 2.65, "learning_rate": 1.4863073289869033e-05, "loss": 0.267, "step": 25601 }, { "epoch": 2.65, "learning_rate": 1.486094821421383e-05, "loss": 0.9774, "step": 25602 }, { "epoch": 2.65, "learning_rate": 1.4858823240471179e-05, "loss": 0.4028, "step": 25603 }, { "epoch": 2.65, "learning_rate": 1.4856698368655392e-05, "loss": 0.5714, "step": 25604 }, { "epoch": 2.65, "learning_rate": 1.4854573598780777e-05, "loss": 0.4837, "step": 25605 }, { "epoch": 2.65, "learning_rate": 1.4852448930861628e-05, "loss": 0.1723, "step": 25606 }, { "epoch": 2.65, "learning_rate": 1.4850324364912237e-05, "loss": 0.3347, "step": 25607 }, { "epoch": 2.65, "learning_rate": 1.484819990094692e-05, "loss": 0.0997, "step": 25608 }, { "epoch": 2.65, "learning_rate": 1.4846075538979971e-05, "loss": 0.5045, "step": 25609 }, { "epoch": 2.65, "learning_rate": 1.4843951279025702e-05, "loss": 0.4441, "step": 25610 }, { "epoch": 2.65, "learning_rate": 1.4841827121098403e-05, "loss": 0.2689, "step": 25611 }, { "epoch": 2.65, "learning_rate": 1.483970306521238e-05, "loss": 0.0142, "step": 25612 }, { "epoch": 2.65, "learning_rate": 1.4837579111381926e-05, "loss": 0.2949, "step": 25613 }, { "epoch": 2.65, "learning_rate": 1.4835455259621344e-05, "loss": 0.888, "step": 25614 }, { "epoch": 2.65, "learning_rate": 1.4833331509944938e-05, "loss": 0.3735, "step": 25615 }, { "epoch": 2.65, "learning_rate": 1.4831207862366985e-05, "loss": 0.3534, "step": 25616 }, { "epoch": 2.65, "learning_rate": 1.482908431690179e-05, "loss": 0.4442, "step": 25617 }, { "epoch": 2.65, "learning_rate": 1.4826960873563652e-05, "loss": 0.7068, "step": 25618 }, { "epoch": 2.65, "learning_rate": 1.4824837532366876e-05, "loss": 0.3268, "step": 25619 }, { "epoch": 2.65, "learning_rate": 1.4822714293325727e-05, "loss": 0.9216, "step": 25620 }, { "epoch": 2.65, "learning_rate": 1.4820591156454518e-05, "loss": 0.5319, "step": 25621 }, { "epoch": 2.65, "learning_rate": 1.4818468121767537e-05, "loss": 0.3985, "step": 25622 }, { "epoch": 2.65, "learning_rate": 1.4816345189279074e-05, "loss": 0.4481, "step": 25623 }, { "epoch": 2.65, "learning_rate": 1.4814222359003424e-05, "loss": 0.1088, "step": 25624 }, { "epoch": 2.65, "learning_rate": 1.4812099630954887e-05, "loss": 0.6816, "step": 25625 }, { "epoch": 2.65, "learning_rate": 1.4809977005147732e-05, "loss": 0.3044, "step": 25626 }, { "epoch": 2.65, "learning_rate": 1.4807854481596257e-05, "loss": 0.4813, "step": 25627 }, { "epoch": 2.65, "learning_rate": 1.480573206031475e-05, "loss": 0.1582, "step": 25628 }, { "epoch": 2.65, "learning_rate": 1.4803609741317499e-05, "loss": 0.201, "step": 25629 }, { "epoch": 2.65, "learning_rate": 1.4801487524618794e-05, "loss": 0.729, "step": 25630 }, { "epoch": 2.65, "learning_rate": 1.4799365410232927e-05, "loss": 0.1828, "step": 25631 }, { "epoch": 2.65, "learning_rate": 1.4797243398174168e-05, "loss": 0.284, "step": 25632 }, { "epoch": 2.65, "learning_rate": 1.4795121488456805e-05, "loss": 0.1423, "step": 25633 }, { "epoch": 2.65, "learning_rate": 1.479299968109514e-05, "loss": 0.1249, "step": 25634 }, { "epoch": 2.65, "learning_rate": 1.4790877976103433e-05, "loss": 0.126, "step": 25635 }, { "epoch": 2.65, "learning_rate": 1.4788756373495972e-05, "loss": 0.4307, "step": 25636 }, { "epoch": 2.65, "learning_rate": 1.478663487328705e-05, "loss": 0.1929, "step": 25637 }, { "epoch": 2.65, "learning_rate": 1.4784513475490936e-05, "loss": 0.3152, "step": 25638 }, { "epoch": 2.65, "learning_rate": 1.4782392180121921e-05, "loss": 0.1674, "step": 25639 }, { "epoch": 2.65, "learning_rate": 1.4780270987194282e-05, "loss": 0.2796, "step": 25640 }, { "epoch": 2.65, "learning_rate": 1.47781498967223e-05, "loss": 0.4477, "step": 25641 }, { "epoch": 2.65, "learning_rate": 1.4776028908720246e-05, "loss": 0.212, "step": 25642 }, { "epoch": 2.65, "learning_rate": 1.4773908023202417e-05, "loss": 0.3175, "step": 25643 }, { "epoch": 2.65, "learning_rate": 1.4771787240183074e-05, "loss": 0.058, "step": 25644 }, { "epoch": 2.65, "learning_rate": 1.4769666559676487e-05, "loss": 0.634, "step": 25645 }, { "epoch": 2.65, "learning_rate": 1.4767545981696942e-05, "loss": 0.0892, "step": 25646 }, { "epoch": 2.65, "learning_rate": 1.476542550625871e-05, "loss": 0.5379, "step": 25647 }, { "epoch": 2.65, "learning_rate": 1.4763305133376073e-05, "loss": 0.2295, "step": 25648 }, { "epoch": 2.65, "learning_rate": 1.4761184863063301e-05, "loss": 0.334, "step": 25649 }, { "epoch": 2.65, "learning_rate": 1.4759064695334663e-05, "loss": 0.2998, "step": 25650 }, { "epoch": 2.65, "learning_rate": 1.4756944630204435e-05, "loss": 0.3607, "step": 25651 }, { "epoch": 2.65, "learning_rate": 1.4754824667686892e-05, "loss": 0.0964, "step": 25652 }, { "epoch": 2.65, "learning_rate": 1.4752704807796314e-05, "loss": 0.2236, "step": 25653 }, { "epoch": 2.65, "learning_rate": 1.4750585050546947e-05, "loss": 0.1755, "step": 25654 }, { "epoch": 2.65, "learning_rate": 1.4748465395953073e-05, "loss": 0.3176, "step": 25655 }, { "epoch": 2.65, "learning_rate": 1.4746345844028963e-05, "loss": 0.8067, "step": 25656 }, { "epoch": 2.65, "learning_rate": 1.4744226394788895e-05, "loss": 0.2765, "step": 25657 }, { "epoch": 2.65, "learning_rate": 1.4742107048247112e-05, "loss": 0.2436, "step": 25658 }, { "epoch": 2.65, "learning_rate": 1.4739987804417895e-05, "loss": 0.2271, "step": 25659 }, { "epoch": 2.65, "learning_rate": 1.4737868663315512e-05, "loss": 0.1113, "step": 25660 }, { "epoch": 2.65, "learning_rate": 1.4735749624954223e-05, "loss": 0.3091, "step": 25661 }, { "epoch": 2.65, "learning_rate": 1.4733630689348297e-05, "loss": 0.229, "step": 25662 }, { "epoch": 2.65, "learning_rate": 1.473151185651201e-05, "loss": 0.264, "step": 25663 }, { "epoch": 2.65, "learning_rate": 1.4729393126459596e-05, "loss": 0.3119, "step": 25664 }, { "epoch": 2.65, "learning_rate": 1.472727449920534e-05, "loss": 0.8029, "step": 25665 }, { "epoch": 2.65, "learning_rate": 1.4725155974763496e-05, "loss": 0.4889, "step": 25666 }, { "epoch": 2.65, "learning_rate": 1.4723037553148328e-05, "loss": 0.1353, "step": 25667 }, { "epoch": 2.65, "learning_rate": 1.4720919234374098e-05, "loss": 0.5144, "step": 25668 }, { "epoch": 2.65, "learning_rate": 1.4718801018455074e-05, "loss": 1.3748, "step": 25669 }, { "epoch": 2.65, "learning_rate": 1.4716682905405496e-05, "loss": 0.4152, "step": 25670 }, { "epoch": 2.65, "learning_rate": 1.4714564895239633e-05, "loss": 0.3805, "step": 25671 }, { "epoch": 2.65, "learning_rate": 1.4712446987971755e-05, "loss": 0.583, "step": 25672 }, { "epoch": 2.65, "learning_rate": 1.4710329183616095e-05, "loss": 0.4274, "step": 25673 }, { "epoch": 2.65, "learning_rate": 1.4708211482186918e-05, "loss": 0.1829, "step": 25674 }, { "epoch": 2.65, "learning_rate": 1.4706093883698488e-05, "loss": 0.2402, "step": 25675 }, { "epoch": 2.65, "learning_rate": 1.4703976388165052e-05, "loss": 0.6595, "step": 25676 }, { "epoch": 2.66, "learning_rate": 1.470185899560087e-05, "loss": 0.2969, "step": 25677 }, { "epoch": 2.66, "learning_rate": 1.4699741706020193e-05, "loss": 0.3175, "step": 25678 }, { "epoch": 2.66, "learning_rate": 1.4697624519437275e-05, "loss": 0.4359, "step": 25679 }, { "epoch": 2.66, "learning_rate": 1.4695507435866369e-05, "loss": 0.2274, "step": 25680 }, { "epoch": 2.66, "learning_rate": 1.4693390455321736e-05, "loss": 0.0872, "step": 25681 }, { "epoch": 2.66, "learning_rate": 1.4691273577817618e-05, "loss": 0.224, "step": 25682 }, { "epoch": 2.66, "learning_rate": 1.4689156803368251e-05, "loss": 0.2543, "step": 25683 }, { "epoch": 2.66, "learning_rate": 1.46870401319879e-05, "loss": 0.6636, "step": 25684 }, { "epoch": 2.66, "learning_rate": 1.4684923563690813e-05, "loss": 0.5091, "step": 25685 }, { "epoch": 2.66, "learning_rate": 1.4682807098491236e-05, "loss": 0.2303, "step": 25686 }, { "epoch": 2.66, "learning_rate": 1.4680690736403417e-05, "loss": 0.402, "step": 25687 }, { "epoch": 2.66, "learning_rate": 1.4678574477441602e-05, "loss": 0.6397, "step": 25688 }, { "epoch": 2.66, "learning_rate": 1.4676458321620042e-05, "loss": 0.1798, "step": 25689 }, { "epoch": 2.66, "learning_rate": 1.4674342268952975e-05, "loss": 0.4652, "step": 25690 }, { "epoch": 2.66, "learning_rate": 1.4672226319454667e-05, "loss": 0.0756, "step": 25691 }, { "epoch": 2.66, "learning_rate": 1.4670110473139329e-05, "loss": 0.1833, "step": 25692 }, { "epoch": 2.66, "learning_rate": 1.466799473002122e-05, "loss": 0.0834, "step": 25693 }, { "epoch": 2.66, "learning_rate": 1.4665879090114597e-05, "loss": 0.5706, "step": 25694 }, { "epoch": 2.66, "learning_rate": 1.4663763553433678e-05, "loss": 0.1031, "step": 25695 }, { "epoch": 2.66, "learning_rate": 1.4661648119992712e-05, "loss": 0.6475, "step": 25696 }, { "epoch": 2.66, "learning_rate": 1.4659532789805941e-05, "loss": 0.3915, "step": 25697 }, { "epoch": 2.66, "learning_rate": 1.4657417562887608e-05, "loss": 0.3835, "step": 25698 }, { "epoch": 2.66, "learning_rate": 1.4655302439251951e-05, "loss": 0.2165, "step": 25699 }, { "epoch": 2.66, "learning_rate": 1.4653187418913204e-05, "loss": 0.5567, "step": 25700 }, { "epoch": 2.66, "learning_rate": 1.465107250188562e-05, "loss": 0.1599, "step": 25701 }, { "epoch": 2.66, "learning_rate": 1.4648957688183418e-05, "loss": 0.4095, "step": 25702 }, { "epoch": 2.66, "learning_rate": 1.4646842977820836e-05, "loss": 0.1968, "step": 25703 }, { "epoch": 2.66, "learning_rate": 1.4644728370812117e-05, "loss": 0.4372, "step": 25704 }, { "epoch": 2.66, "learning_rate": 1.4642613867171494e-05, "loss": 0.2868, "step": 25705 }, { "epoch": 2.66, "learning_rate": 1.4640499466913203e-05, "loss": 0.405, "step": 25706 }, { "epoch": 2.66, "learning_rate": 1.4638385170051484e-05, "loss": 0.5254, "step": 25707 }, { "epoch": 2.66, "learning_rate": 1.4636270976600549e-05, "loss": 0.7254, "step": 25708 }, { "epoch": 2.66, "learning_rate": 1.4634156886574645e-05, "loss": 1.0567, "step": 25709 }, { "epoch": 2.66, "learning_rate": 1.4632042899988003e-05, "loss": 0.4829, "step": 25710 }, { "epoch": 2.66, "learning_rate": 1.462992901685486e-05, "loss": 0.4258, "step": 25711 }, { "epoch": 2.66, "learning_rate": 1.462781523718943e-05, "loss": 0.3525, "step": 25712 }, { "epoch": 2.66, "learning_rate": 1.4625701561005951e-05, "loss": 0.199, "step": 25713 }, { "epoch": 2.66, "learning_rate": 1.4623587988318649e-05, "loss": 0.4916, "step": 25714 }, { "epoch": 2.66, "learning_rate": 1.462147451914176e-05, "loss": 0.1911, "step": 25715 }, { "epoch": 2.66, "learning_rate": 1.4619361153489503e-05, "loss": 0.3046, "step": 25716 }, { "epoch": 2.66, "learning_rate": 1.461724789137611e-05, "loss": 0.2242, "step": 25717 }, { "epoch": 2.66, "learning_rate": 1.46151347328158e-05, "loss": 0.509, "step": 25718 }, { "epoch": 2.66, "learning_rate": 1.461302167782282e-05, "loss": 0.4099, "step": 25719 }, { "epoch": 2.66, "learning_rate": 1.4610908726411372e-05, "loss": 0.1321, "step": 25720 }, { "epoch": 2.66, "learning_rate": 1.4608795878595681e-05, "loss": 0.056, "step": 25721 }, { "epoch": 2.66, "learning_rate": 1.460668313438997e-05, "loss": 0.4003, "step": 25722 }, { "epoch": 2.66, "learning_rate": 1.4604570493808469e-05, "loss": 0.2941, "step": 25723 }, { "epoch": 2.66, "learning_rate": 1.4602457956865397e-05, "loss": 0.1128, "step": 25724 }, { "epoch": 2.66, "learning_rate": 1.4600345523574973e-05, "loss": 0.7773, "step": 25725 }, { "epoch": 2.66, "learning_rate": 1.459823319395142e-05, "loss": 0.0858, "step": 25726 }, { "epoch": 2.66, "learning_rate": 1.4596120968008961e-05, "loss": 0.0902, "step": 25727 }, { "epoch": 2.66, "learning_rate": 1.4594008845761808e-05, "loss": 0.3544, "step": 25728 }, { "epoch": 2.66, "learning_rate": 1.4591896827224197e-05, "loss": 0.1819, "step": 25729 }, { "epoch": 2.66, "learning_rate": 1.4589784912410316e-05, "loss": 0.2441, "step": 25730 }, { "epoch": 2.66, "learning_rate": 1.4587673101334403e-05, "loss": 0.3331, "step": 25731 }, { "epoch": 2.66, "learning_rate": 1.4585561394010675e-05, "loss": 0.1317, "step": 25732 }, { "epoch": 2.66, "learning_rate": 1.4583449790453329e-05, "loss": 0.3165, "step": 25733 }, { "epoch": 2.66, "learning_rate": 1.4581338290676594e-05, "loss": 0.6976, "step": 25734 }, { "epoch": 2.66, "learning_rate": 1.4579226894694684e-05, "loss": 0.1364, "step": 25735 }, { "epoch": 2.66, "learning_rate": 1.4577115602521806e-05, "loss": 0.054, "step": 25736 }, { "epoch": 2.66, "learning_rate": 1.4575004414172182e-05, "loss": 0.236, "step": 25737 }, { "epoch": 2.66, "learning_rate": 1.4572893329660014e-05, "loss": 0.2487, "step": 25738 }, { "epoch": 2.66, "learning_rate": 1.4570782348999531e-05, "loss": 0.477, "step": 25739 }, { "epoch": 2.66, "learning_rate": 1.4568671472204922e-05, "loss": 0.167, "step": 25740 }, { "epoch": 2.66, "learning_rate": 1.4566560699290406e-05, "loss": 0.2632, "step": 25741 }, { "epoch": 2.66, "learning_rate": 1.456445003027019e-05, "loss": 0.5211, "step": 25742 }, { "epoch": 2.66, "learning_rate": 1.4562339465158485e-05, "loss": 0.2366, "step": 25743 }, { "epoch": 2.66, "learning_rate": 1.4560229003969512e-05, "loss": 0.6071, "step": 25744 }, { "epoch": 2.66, "learning_rate": 1.4558118646717453e-05, "loss": 0.4244, "step": 25745 }, { "epoch": 2.66, "learning_rate": 1.4556008393416524e-05, "loss": 0.1003, "step": 25746 }, { "epoch": 2.66, "learning_rate": 1.4553898244080935e-05, "loss": 0.692, "step": 25747 }, { "epoch": 2.66, "learning_rate": 1.4551788198724887e-05, "loss": 0.2401, "step": 25748 }, { "epoch": 2.66, "learning_rate": 1.4549678257362601e-05, "loss": 0.427, "step": 25749 }, { "epoch": 2.66, "learning_rate": 1.4547568420008253e-05, "loss": 0.4033, "step": 25750 }, { "epoch": 2.66, "learning_rate": 1.4545458686676054e-05, "loss": 0.3737, "step": 25751 }, { "epoch": 2.66, "learning_rate": 1.4543349057380217e-05, "loss": 0.4147, "step": 25752 }, { "epoch": 2.66, "learning_rate": 1.4541239532134938e-05, "loss": 0.0658, "step": 25753 }, { "epoch": 2.66, "learning_rate": 1.4539130110954414e-05, "loss": 0.44, "step": 25754 }, { "epoch": 2.66, "learning_rate": 1.453702079385285e-05, "loss": 0.0377, "step": 25755 }, { "epoch": 2.66, "learning_rate": 1.4534911580844445e-05, "loss": 0.4754, "step": 25756 }, { "epoch": 2.66, "learning_rate": 1.4532802471943403e-05, "loss": 0.1757, "step": 25757 }, { "epoch": 2.66, "learning_rate": 1.4530693467163921e-05, "loss": 0.3312, "step": 25758 }, { "epoch": 2.66, "learning_rate": 1.4528584566520178e-05, "loss": 0.036, "step": 25759 }, { "epoch": 2.66, "learning_rate": 1.4526475770026386e-05, "loss": 0.2195, "step": 25760 }, { "epoch": 2.66, "learning_rate": 1.4524367077696736e-05, "loss": 0.089, "step": 25761 }, { "epoch": 2.66, "learning_rate": 1.4522258489545424e-05, "loss": 0.7081, "step": 25762 }, { "epoch": 2.66, "learning_rate": 1.4520150005586648e-05, "loss": 0.1278, "step": 25763 }, { "epoch": 2.66, "learning_rate": 1.4518041625834602e-05, "loss": 0.4489, "step": 25764 }, { "epoch": 2.66, "learning_rate": 1.4515933350303479e-05, "loss": 0.1443, "step": 25765 }, { "epoch": 2.66, "learning_rate": 1.4513825179007466e-05, "loss": 0.7605, "step": 25766 }, { "epoch": 2.66, "learning_rate": 1.4511717111960771e-05, "loss": 0.3979, "step": 25767 }, { "epoch": 2.66, "learning_rate": 1.4509609149177558e-05, "loss": 0.2057, "step": 25768 }, { "epoch": 2.66, "learning_rate": 1.4507501290672033e-05, "loss": 0.1366, "step": 25769 }, { "epoch": 2.66, "learning_rate": 1.45053935364584e-05, "loss": 0.0219, "step": 25770 }, { "epoch": 2.66, "learning_rate": 1.4503285886550818e-05, "loss": 0.2154, "step": 25771 }, { "epoch": 2.66, "learning_rate": 1.450117834096349e-05, "loss": 0.4568, "step": 25772 }, { "epoch": 2.67, "learning_rate": 1.4499070899710602e-05, "loss": 0.1858, "step": 25773 }, { "epoch": 2.67, "learning_rate": 1.449696356280634e-05, "loss": 0.4699, "step": 25774 }, { "epoch": 2.67, "learning_rate": 1.4494856330264894e-05, "loss": 0.6035, "step": 25775 }, { "epoch": 2.67, "learning_rate": 1.4492749202100449e-05, "loss": 0.401, "step": 25776 }, { "epoch": 2.67, "learning_rate": 1.4490642178327198e-05, "loss": 0.5618, "step": 25777 }, { "epoch": 2.67, "learning_rate": 1.4488535258959303e-05, "loss": 0.3433, "step": 25778 }, { "epoch": 2.67, "learning_rate": 1.448642844401096e-05, "loss": 0.3908, "step": 25779 }, { "epoch": 2.67, "learning_rate": 1.4484321733496348e-05, "loss": 0.432, "step": 25780 }, { "epoch": 2.67, "learning_rate": 1.4482215127429654e-05, "loss": 0.8113, "step": 25781 }, { "epoch": 2.67, "learning_rate": 1.448010862582507e-05, "loss": 0.3495, "step": 25782 }, { "epoch": 2.67, "learning_rate": 1.4478002228696753e-05, "loss": 0.4305, "step": 25783 }, { "epoch": 2.67, "learning_rate": 1.4475895936058888e-05, "loss": 0.2128, "step": 25784 }, { "epoch": 2.67, "learning_rate": 1.4473789747925661e-05, "loss": 0.1231, "step": 25785 }, { "epoch": 2.67, "learning_rate": 1.4471683664311248e-05, "loss": 0.4879, "step": 25786 }, { "epoch": 2.67, "learning_rate": 1.446957768522984e-05, "loss": 0.2646, "step": 25787 }, { "epoch": 2.67, "learning_rate": 1.4467471810695592e-05, "loss": 0.6002, "step": 25788 }, { "epoch": 2.67, "learning_rate": 1.4465366040722685e-05, "loss": 0.0871, "step": 25789 }, { "epoch": 2.67, "learning_rate": 1.4463260375325303e-05, "loss": 0.0757, "step": 25790 }, { "epoch": 2.67, "learning_rate": 1.4461154814517615e-05, "loss": 0.2782, "step": 25791 }, { "epoch": 2.67, "learning_rate": 1.4459049358313796e-05, "loss": 0.3355, "step": 25792 }, { "epoch": 2.67, "learning_rate": 1.445694400672802e-05, "loss": 0.652, "step": 25793 }, { "epoch": 2.67, "learning_rate": 1.4454838759774462e-05, "loss": 0.3765, "step": 25794 }, { "epoch": 2.67, "learning_rate": 1.4452733617467305e-05, "loss": 0.1249, "step": 25795 }, { "epoch": 2.67, "learning_rate": 1.4450628579820695e-05, "loss": 0.1318, "step": 25796 }, { "epoch": 2.67, "learning_rate": 1.4448523646848823e-05, "loss": 0.4823, "step": 25797 }, { "epoch": 2.67, "learning_rate": 1.4446418818565842e-05, "loss": 0.5442, "step": 25798 }, { "epoch": 2.67, "learning_rate": 1.4444314094985932e-05, "loss": 0.3481, "step": 25799 }, { "epoch": 2.67, "learning_rate": 1.4442209476123259e-05, "loss": 0.3047, "step": 25800 }, { "epoch": 2.67, "learning_rate": 1.444010496199199e-05, "loss": 0.5203, "step": 25801 }, { "epoch": 2.67, "learning_rate": 1.4438000552606297e-05, "loss": 0.6742, "step": 25802 }, { "epoch": 2.67, "learning_rate": 1.443589624798034e-05, "loss": 0.3296, "step": 25803 }, { "epoch": 2.67, "learning_rate": 1.4433792048128287e-05, "loss": 0.2302, "step": 25804 }, { "epoch": 2.67, "learning_rate": 1.4431687953064319e-05, "loss": 0.8346, "step": 25805 }, { "epoch": 2.67, "learning_rate": 1.4429583962802571e-05, "loss": 0.1797, "step": 25806 }, { "epoch": 2.67, "learning_rate": 1.4427480077357231e-05, "loss": 0.1117, "step": 25807 }, { "epoch": 2.67, "learning_rate": 1.4425376296742443e-05, "loss": 0.6061, "step": 25808 }, { "epoch": 2.67, "learning_rate": 1.4423272620972374e-05, "loss": 0.2921, "step": 25809 }, { "epoch": 2.67, "learning_rate": 1.442116905006119e-05, "loss": 0.2787, "step": 25810 }, { "epoch": 2.67, "learning_rate": 1.4419065584023054e-05, "loss": 0.5112, "step": 25811 }, { "epoch": 2.67, "learning_rate": 1.441696222287212e-05, "loss": 0.3561, "step": 25812 }, { "epoch": 2.67, "learning_rate": 1.4414858966622551e-05, "loss": 0.6086, "step": 25813 }, { "epoch": 2.67, "learning_rate": 1.4412755815288504e-05, "loss": 0.1543, "step": 25814 }, { "epoch": 2.67, "learning_rate": 1.4410652768884149e-05, "loss": 0.3688, "step": 25815 }, { "epoch": 2.67, "learning_rate": 1.4408549827423623e-05, "loss": 0.1358, "step": 25816 }, { "epoch": 2.67, "learning_rate": 1.440644699092109e-05, "loss": 0.5212, "step": 25817 }, { "epoch": 2.67, "learning_rate": 1.4404344259390707e-05, "loss": 0.4481, "step": 25818 }, { "epoch": 2.67, "learning_rate": 1.4402241632846627e-05, "loss": 0.3578, "step": 25819 }, { "epoch": 2.67, "learning_rate": 1.440013911130302e-05, "loss": 0.341, "step": 25820 }, { "epoch": 2.67, "learning_rate": 1.4398036694774017e-05, "loss": 0.454, "step": 25821 }, { "epoch": 2.67, "learning_rate": 1.4395934383273778e-05, "loss": 0.4954, "step": 25822 }, { "epoch": 2.67, "learning_rate": 1.4393832176816461e-05, "loss": 0.5473, "step": 25823 }, { "epoch": 2.67, "learning_rate": 1.4391730075416209e-05, "loss": 0.1977, "step": 25824 }, { "epoch": 2.67, "learning_rate": 1.4389628079087194e-05, "loss": 0.1431, "step": 25825 }, { "epoch": 2.67, "learning_rate": 1.4387526187843536e-05, "loss": 0.7963, "step": 25826 }, { "epoch": 2.67, "learning_rate": 1.4385424401699402e-05, "loss": 0.3419, "step": 25827 }, { "epoch": 2.67, "learning_rate": 1.4383322720668934e-05, "loss": 0.7218, "step": 25828 }, { "epoch": 2.67, "learning_rate": 1.4381221144766284e-05, "loss": 0.3996, "step": 25829 }, { "epoch": 2.67, "learning_rate": 1.4379119674005602e-05, "loss": 0.5879, "step": 25830 }, { "epoch": 2.67, "learning_rate": 1.4377018308401027e-05, "loss": 0.4684, "step": 25831 }, { "epoch": 2.67, "learning_rate": 1.437491704796672e-05, "loss": 0.5555, "step": 25832 }, { "epoch": 2.67, "learning_rate": 1.4372815892716807e-05, "loss": 0.2229, "step": 25833 }, { "epoch": 2.67, "learning_rate": 1.4370714842665442e-05, "loss": 0.3554, "step": 25834 }, { "epoch": 2.67, "learning_rate": 1.4368613897826776e-05, "loss": 0.1313, "step": 25835 }, { "epoch": 2.67, "learning_rate": 1.4366513058214936e-05, "loss": 0.2814, "step": 25836 }, { "epoch": 2.67, "learning_rate": 1.436441232384407e-05, "loss": 0.2028, "step": 25837 }, { "epoch": 2.67, "learning_rate": 1.4362311694728322e-05, "loss": 0.4821, "step": 25838 }, { "epoch": 2.67, "learning_rate": 1.4360211170881831e-05, "loss": 0.1326, "step": 25839 }, { "epoch": 2.67, "learning_rate": 1.4358110752318741e-05, "loss": 0.2344, "step": 25840 }, { "epoch": 2.67, "learning_rate": 1.4356010439053189e-05, "loss": 0.3653, "step": 25841 }, { "epoch": 2.67, "learning_rate": 1.4353910231099315e-05, "loss": 0.3824, "step": 25842 }, { "epoch": 2.67, "learning_rate": 1.4351810128471265e-05, "loss": 0.2374, "step": 25843 }, { "epoch": 2.67, "learning_rate": 1.4349710131183159e-05, "loss": 0.0624, "step": 25844 }, { "epoch": 2.67, "learning_rate": 1.4347610239249156e-05, "loss": 0.5638, "step": 25845 }, { "epoch": 2.67, "learning_rate": 1.4345510452683365e-05, "loss": 0.1927, "step": 25846 }, { "epoch": 2.67, "learning_rate": 1.4343410771499934e-05, "loss": 0.0834, "step": 25847 }, { "epoch": 2.67, "learning_rate": 1.4341311195712998e-05, "loss": 0.2962, "step": 25848 }, { "epoch": 2.67, "learning_rate": 1.4339211725336693e-05, "loss": 0.5636, "step": 25849 }, { "epoch": 2.67, "learning_rate": 1.433711236038515e-05, "loss": 0.1193, "step": 25850 }, { "epoch": 2.67, "learning_rate": 1.4335013100872501e-05, "loss": 0.546, "step": 25851 }, { "epoch": 2.67, "learning_rate": 1.4332913946812878e-05, "loss": 0.4662, "step": 25852 }, { "epoch": 2.67, "learning_rate": 1.4330814898220424e-05, "loss": 0.2871, "step": 25853 }, { "epoch": 2.67, "learning_rate": 1.4328715955109245e-05, "loss": 0.4515, "step": 25854 }, { "epoch": 2.67, "learning_rate": 1.4326617117493488e-05, "loss": 0.603, "step": 25855 }, { "epoch": 2.67, "learning_rate": 1.4324518385387274e-05, "loss": 0.3326, "step": 25856 }, { "epoch": 2.67, "learning_rate": 1.4322419758804734e-05, "loss": 0.7526, "step": 25857 }, { "epoch": 2.67, "learning_rate": 1.4320321237760009e-05, "loss": 0.2205, "step": 25858 }, { "epoch": 2.67, "learning_rate": 1.43182228222672e-05, "loss": 0.3623, "step": 25859 }, { "epoch": 2.67, "learning_rate": 1.4316124512340447e-05, "loss": 0.2345, "step": 25860 }, { "epoch": 2.67, "learning_rate": 1.4314026307993873e-05, "loss": 0.5486, "step": 25861 }, { "epoch": 2.67, "learning_rate": 1.4311928209241605e-05, "loss": 0.4994, "step": 25862 }, { "epoch": 2.67, "learning_rate": 1.4309830216097775e-05, "loss": 0.2492, "step": 25863 }, { "epoch": 2.67, "learning_rate": 1.4307732328576487e-05, "loss": 0.3835, "step": 25864 }, { "epoch": 2.67, "learning_rate": 1.4305634546691871e-05, "loss": 0.2159, "step": 25865 }, { "epoch": 2.67, "learning_rate": 1.4303536870458053e-05, "loss": 0.803, "step": 25866 }, { "epoch": 2.67, "learning_rate": 1.4301439299889154e-05, "loss": 0.7654, "step": 25867 }, { "epoch": 2.67, "learning_rate": 1.4299341834999288e-05, "loss": 0.3831, "step": 25868 }, { "epoch": 2.68, "learning_rate": 1.4297244475802584e-05, "loss": 0.2353, "step": 25869 }, { "epoch": 2.68, "learning_rate": 1.4295147222313162e-05, "loss": 0.2986, "step": 25870 }, { "epoch": 2.68, "learning_rate": 1.4293050074545125e-05, "loss": 0.399, "step": 25871 }, { "epoch": 2.68, "learning_rate": 1.4290953032512603e-05, "loss": 0.1338, "step": 25872 }, { "epoch": 2.68, "learning_rate": 1.4288856096229718e-05, "loss": 0.1378, "step": 25873 }, { "epoch": 2.68, "learning_rate": 1.4286759265710566e-05, "loss": 0.3387, "step": 25874 }, { "epoch": 2.68, "learning_rate": 1.4284662540969276e-05, "loss": 0.8935, "step": 25875 }, { "epoch": 2.68, "learning_rate": 1.428256592201996e-05, "loss": 0.2607, "step": 25876 }, { "epoch": 2.68, "learning_rate": 1.4280469408876732e-05, "loss": 0.4567, "step": 25877 }, { "epoch": 2.68, "learning_rate": 1.4278373001553704e-05, "loss": 0.1609, "step": 25878 }, { "epoch": 2.68, "learning_rate": 1.4276276700064994e-05, "loss": 0.3472, "step": 25879 }, { "epoch": 2.68, "learning_rate": 1.4274180504424708e-05, "loss": 0.2932, "step": 25880 }, { "epoch": 2.68, "learning_rate": 1.427208441464697e-05, "loss": 0.1527, "step": 25881 }, { "epoch": 2.68, "learning_rate": 1.4269988430745869e-05, "loss": 0.7734, "step": 25882 }, { "epoch": 2.68, "learning_rate": 1.4267892552735538e-05, "loss": 0.5409, "step": 25883 }, { "epoch": 2.68, "learning_rate": 1.4265796780630063e-05, "loss": 0.5493, "step": 25884 }, { "epoch": 2.68, "learning_rate": 1.426370111444356e-05, "loss": 0.1543, "step": 25885 }, { "epoch": 2.68, "learning_rate": 1.4261605554190136e-05, "loss": 0.187, "step": 25886 }, { "epoch": 2.68, "learning_rate": 1.4259510099883908e-05, "loss": 0.3036, "step": 25887 }, { "epoch": 2.68, "learning_rate": 1.4257414751538971e-05, "loss": 0.177, "step": 25888 }, { "epoch": 2.68, "learning_rate": 1.4255319509169434e-05, "loss": 0.1666, "step": 25889 }, { "epoch": 2.68, "learning_rate": 1.4253224372789403e-05, "loss": 0.3316, "step": 25890 }, { "epoch": 2.68, "learning_rate": 1.4251129342412994e-05, "loss": 0.0986, "step": 25891 }, { "epoch": 2.68, "learning_rate": 1.4249034418054283e-05, "loss": 0.8247, "step": 25892 }, { "epoch": 2.68, "learning_rate": 1.424693959972739e-05, "loss": 0.2802, "step": 25893 }, { "epoch": 2.68, "learning_rate": 1.4244844887446412e-05, "loss": 0.6014, "step": 25894 }, { "epoch": 2.68, "learning_rate": 1.4242750281225466e-05, "loss": 0.4381, "step": 25895 }, { "epoch": 2.68, "learning_rate": 1.4240655781078621e-05, "loss": 0.2563, "step": 25896 }, { "epoch": 2.68, "learning_rate": 1.4238561387019999e-05, "loss": 0.2383, "step": 25897 }, { "epoch": 2.68, "learning_rate": 1.4236467099063693e-05, "loss": 0.4665, "step": 25898 }, { "epoch": 2.68, "learning_rate": 1.4234372917223798e-05, "loss": 0.1446, "step": 25899 }, { "epoch": 2.68, "learning_rate": 1.423227884151442e-05, "loss": 0.3332, "step": 25900 }, { "epoch": 2.68, "learning_rate": 1.423018487194966e-05, "loss": 0.3803, "step": 25901 }, { "epoch": 2.68, "learning_rate": 1.4228091008543597e-05, "loss": 0.6561, "step": 25902 }, { "epoch": 2.68, "learning_rate": 1.4225997251310333e-05, "loss": 0.2605, "step": 25903 }, { "epoch": 2.68, "learning_rate": 1.4223903600263965e-05, "loss": 0.6355, "step": 25904 }, { "epoch": 2.68, "learning_rate": 1.422181005541859e-05, "loss": 0.3418, "step": 25905 }, { "epoch": 2.68, "learning_rate": 1.4219716616788293e-05, "loss": 0.1428, "step": 25906 }, { "epoch": 2.68, "learning_rate": 1.4217623284387176e-05, "loss": 0.5844, "step": 25907 }, { "epoch": 2.68, "learning_rate": 1.4215530058229333e-05, "loss": 0.152, "step": 25908 }, { "epoch": 2.68, "learning_rate": 1.4213436938328839e-05, "loss": 0.0635, "step": 25909 }, { "epoch": 2.68, "learning_rate": 1.4211343924699794e-05, "loss": 0.2655, "step": 25910 }, { "epoch": 2.68, "learning_rate": 1.4209251017356299e-05, "loss": 0.1778, "step": 25911 }, { "epoch": 2.68, "learning_rate": 1.420715821631242e-05, "loss": 0.6093, "step": 25912 }, { "epoch": 2.68, "learning_rate": 1.4205065521582258e-05, "loss": 0.272, "step": 25913 }, { "epoch": 2.68, "learning_rate": 1.4202972933179896e-05, "loss": 0.5283, "step": 25914 }, { "epoch": 2.68, "learning_rate": 1.4200880451119428e-05, "loss": 0.8109, "step": 25915 }, { "epoch": 2.68, "learning_rate": 1.419878807541493e-05, "loss": 0.2458, "step": 25916 }, { "epoch": 2.68, "learning_rate": 1.4196695806080499e-05, "loss": 0.3315, "step": 25917 }, { "epoch": 2.68, "learning_rate": 1.419460364313021e-05, "loss": 0.5507, "step": 25918 }, { "epoch": 2.68, "learning_rate": 1.4192511586578166e-05, "loss": 0.2409, "step": 25919 }, { "epoch": 2.68, "learning_rate": 1.4190419636438424e-05, "loss": 0.3411, "step": 25920 }, { "epoch": 2.68, "learning_rate": 1.4188327792725086e-05, "loss": 0.1358, "step": 25921 }, { "epoch": 2.68, "learning_rate": 1.4186236055452218e-05, "loss": 0.4134, "step": 25922 }, { "epoch": 2.68, "learning_rate": 1.4184144424633906e-05, "loss": 0.3617, "step": 25923 }, { "epoch": 2.68, "learning_rate": 1.4182052900284236e-05, "loss": 0.2468, "step": 25924 }, { "epoch": 2.68, "learning_rate": 1.4179961482417284e-05, "loss": 0.4188, "step": 25925 }, { "epoch": 2.68, "learning_rate": 1.4177870171047126e-05, "loss": 0.7719, "step": 25926 }, { "epoch": 2.68, "learning_rate": 1.4175778966187847e-05, "loss": 0.4797, "step": 25927 }, { "epoch": 2.68, "learning_rate": 1.4173687867853523e-05, "loss": 0.4176, "step": 25928 }, { "epoch": 2.68, "learning_rate": 1.4171596876058238e-05, "loss": 0.3686, "step": 25929 }, { "epoch": 2.68, "learning_rate": 1.4169505990816049e-05, "loss": 0.2735, "step": 25930 }, { "epoch": 2.68, "learning_rate": 1.4167415212141042e-05, "loss": 0.6621, "step": 25931 }, { "epoch": 2.68, "learning_rate": 1.416532454004729e-05, "loss": 0.2582, "step": 25932 }, { "epoch": 2.68, "learning_rate": 1.4163233974548877e-05, "loss": 0.4037, "step": 25933 }, { "epoch": 2.68, "learning_rate": 1.4161143515659858e-05, "loss": 0.1415, "step": 25934 }, { "epoch": 2.68, "learning_rate": 1.415905316339432e-05, "loss": 0.164, "step": 25935 }, { "epoch": 2.68, "learning_rate": 1.4156962917766324e-05, "loss": 0.1198, "step": 25936 }, { "epoch": 2.68, "learning_rate": 1.4154872778789946e-05, "loss": 0.2062, "step": 25937 }, { "epoch": 2.68, "learning_rate": 1.415278274647926e-05, "loss": 0.1302, "step": 25938 }, { "epoch": 2.68, "learning_rate": 1.4150692820848345e-05, "loss": 0.7401, "step": 25939 }, { "epoch": 2.68, "learning_rate": 1.4148603001911244e-05, "loss": 0.6743, "step": 25940 }, { "epoch": 2.68, "learning_rate": 1.4146513289682039e-05, "loss": 0.778, "step": 25941 }, { "epoch": 2.68, "learning_rate": 1.4144423684174796e-05, "loss": 0.236, "step": 25942 }, { "epoch": 2.68, "learning_rate": 1.4142334185403584e-05, "loss": 0.0615, "step": 25943 }, { "epoch": 2.68, "learning_rate": 1.4140244793382466e-05, "loss": 0.1944, "step": 25944 }, { "epoch": 2.68, "learning_rate": 1.4138155508125526e-05, "loss": 0.5952, "step": 25945 }, { "epoch": 2.68, "learning_rate": 1.4136066329646794e-05, "loss": 0.4304, "step": 25946 }, { "epoch": 2.68, "learning_rate": 1.4133977257960355e-05, "loss": 0.2573, "step": 25947 }, { "epoch": 2.68, "learning_rate": 1.4131888293080269e-05, "loss": 0.4747, "step": 25948 }, { "epoch": 2.68, "learning_rate": 1.4129799435020608e-05, "loss": 0.145, "step": 25949 }, { "epoch": 2.68, "learning_rate": 1.4127710683795413e-05, "loss": 0.2144, "step": 25950 }, { "epoch": 2.68, "learning_rate": 1.4125622039418754e-05, "loss": 0.4667, "step": 25951 }, { "epoch": 2.68, "learning_rate": 1.4123533501904697e-05, "loss": 0.3963, "step": 25952 }, { "epoch": 2.68, "learning_rate": 1.4121445071267295e-05, "loss": 0.4001, "step": 25953 }, { "epoch": 2.68, "learning_rate": 1.4119356747520608e-05, "loss": 0.2973, "step": 25954 }, { "epoch": 2.68, "learning_rate": 1.4117268530678702e-05, "loss": 0.7446, "step": 25955 }, { "epoch": 2.68, "learning_rate": 1.411518042075562e-05, "loss": 0.5735, "step": 25956 }, { "epoch": 2.68, "learning_rate": 1.4113092417765442e-05, "loss": 0.224, "step": 25957 }, { "epoch": 2.68, "learning_rate": 1.4111004521722199e-05, "loss": 0.188, "step": 25958 }, { "epoch": 2.68, "learning_rate": 1.4108916732639965e-05, "loss": 0.5259, "step": 25959 }, { "epoch": 2.68, "learning_rate": 1.4106829050532774e-05, "loss": 0.2862, "step": 25960 }, { "epoch": 2.68, "learning_rate": 1.410474147541469e-05, "loss": 0.1267, "step": 25961 }, { "epoch": 2.68, "learning_rate": 1.410265400729977e-05, "loss": 0.0722, "step": 25962 }, { "epoch": 2.68, "learning_rate": 1.410056664620206e-05, "loss": 0.2038, "step": 25963 }, { "epoch": 2.69, "learning_rate": 1.4098479392135619e-05, "loss": 0.1658, "step": 25964 }, { "epoch": 2.69, "learning_rate": 1.4096392245114492e-05, "loss": 0.2654, "step": 25965 }, { "epoch": 2.69, "learning_rate": 1.409430520515273e-05, "loss": 0.2745, "step": 25966 }, { "epoch": 2.69, "learning_rate": 1.4092218272264397e-05, "loss": 0.0163, "step": 25967 }, { "epoch": 2.69, "learning_rate": 1.4090131446463518e-05, "loss": 0.2648, "step": 25968 }, { "epoch": 2.69, "learning_rate": 1.408804472776415e-05, "loss": 0.666, "step": 25969 }, { "epoch": 2.69, "learning_rate": 1.4085958116180342e-05, "loss": 0.1254, "step": 25970 }, { "epoch": 2.69, "learning_rate": 1.4083871611726152e-05, "loss": 0.7923, "step": 25971 }, { "epoch": 2.69, "learning_rate": 1.4081785214415603e-05, "loss": 0.3111, "step": 25972 }, { "epoch": 2.69, "learning_rate": 1.4079698924262751e-05, "loss": 0.4275, "step": 25973 }, { "epoch": 2.69, "learning_rate": 1.407761274128164e-05, "loss": 0.5742, "step": 25974 }, { "epoch": 2.69, "learning_rate": 1.4075526665486314e-05, "loss": 0.9155, "step": 25975 }, { "epoch": 2.69, "learning_rate": 1.4073440696890819e-05, "loss": 0.8666, "step": 25976 }, { "epoch": 2.69, "learning_rate": 1.4071354835509204e-05, "loss": 0.7647, "step": 25977 }, { "epoch": 2.69, "learning_rate": 1.406926908135549e-05, "loss": 0.384, "step": 25978 }, { "epoch": 2.69, "learning_rate": 1.406718343444373e-05, "loss": 0.7921, "step": 25979 }, { "epoch": 2.69, "learning_rate": 1.4065097894787962e-05, "loss": 0.3287, "step": 25980 }, { "epoch": 2.69, "learning_rate": 1.406301246240223e-05, "loss": 0.4038, "step": 25981 }, { "epoch": 2.69, "learning_rate": 1.4060927137300566e-05, "loss": 0.251, "step": 25982 }, { "epoch": 2.69, "learning_rate": 1.4058841919497023e-05, "loss": 0.1581, "step": 25983 }, { "epoch": 2.69, "learning_rate": 1.4056756809005615e-05, "loss": 0.6876, "step": 25984 }, { "epoch": 2.69, "learning_rate": 1.4054671805840388e-05, "loss": 0.2424, "step": 25985 }, { "epoch": 2.69, "learning_rate": 1.4052586910015383e-05, "loss": 0.7434, "step": 25986 }, { "epoch": 2.69, "learning_rate": 1.4050502121544641e-05, "loss": 0.0999, "step": 25987 }, { "epoch": 2.69, "learning_rate": 1.4048417440442175e-05, "loss": 0.0436, "step": 25988 }, { "epoch": 2.69, "learning_rate": 1.4046332866722032e-05, "loss": 0.3912, "step": 25989 }, { "epoch": 2.69, "learning_rate": 1.4044248400398239e-05, "loss": 0.7001, "step": 25990 }, { "epoch": 2.69, "learning_rate": 1.4042164041484839e-05, "loss": 0.1194, "step": 25991 }, { "epoch": 2.69, "learning_rate": 1.404007978999585e-05, "loss": 0.1356, "step": 25992 }, { "epoch": 2.69, "learning_rate": 1.4037995645945313e-05, "loss": 0.4601, "step": 25993 }, { "epoch": 2.69, "learning_rate": 1.4035911609347256e-05, "loss": 0.4766, "step": 25994 }, { "epoch": 2.69, "learning_rate": 1.4033827680215715e-05, "loss": 0.101, "step": 25995 }, { "epoch": 2.69, "learning_rate": 1.4031743858564702e-05, "loss": 0.1206, "step": 25996 }, { "epoch": 2.69, "learning_rate": 1.4029660144408261e-05, "loss": 0.4305, "step": 25997 }, { "epoch": 2.69, "learning_rate": 1.4027576537760404e-05, "loss": 0.285, "step": 25998 }, { "epoch": 2.69, "learning_rate": 1.402549303863516e-05, "loss": 0.105, "step": 25999 }, { "epoch": 2.69, "learning_rate": 1.4023409647046562e-05, "loss": 0.5832, "step": 26000 }, { "epoch": 2.69, "learning_rate": 1.4021326363008633e-05, "loss": 0.3729, "step": 26001 }, { "epoch": 2.69, "learning_rate": 1.4019243186535394e-05, "loss": 0.635, "step": 26002 }, { "epoch": 2.69, "learning_rate": 1.4017160117640871e-05, "loss": 0.5634, "step": 26003 }, { "epoch": 2.69, "learning_rate": 1.4015077156339086e-05, "loss": 0.5705, "step": 26004 }, { "epoch": 2.69, "learning_rate": 1.4012994302644072e-05, "loss": 0.1056, "step": 26005 }, { "epoch": 2.69, "learning_rate": 1.401091155656983e-05, "loss": 0.3209, "step": 26006 }, { "epoch": 2.69, "learning_rate": 1.4008828918130391e-05, "loss": 0.2488, "step": 26007 }, { "epoch": 2.69, "learning_rate": 1.4006746387339785e-05, "loss": 0.4044, "step": 26008 }, { "epoch": 2.69, "learning_rate": 1.4004663964212007e-05, "loss": 0.1536, "step": 26009 }, { "epoch": 2.69, "learning_rate": 1.400258164876109e-05, "loss": 0.3277, "step": 26010 }, { "epoch": 2.69, "learning_rate": 1.400049944100105e-05, "loss": 0.4858, "step": 26011 }, { "epoch": 2.69, "learning_rate": 1.3998417340945906e-05, "loss": 0.1517, "step": 26012 }, { "epoch": 2.69, "learning_rate": 1.399633534860967e-05, "loss": 0.4513, "step": 26013 }, { "epoch": 2.69, "learning_rate": 1.3994253464006362e-05, "loss": 0.816, "step": 26014 }, { "epoch": 2.69, "learning_rate": 1.3992171687150002e-05, "loss": 0.1809, "step": 26015 }, { "epoch": 2.69, "learning_rate": 1.3990090018054588e-05, "loss": 0.5562, "step": 26016 }, { "epoch": 2.69, "learning_rate": 1.3988008456734142e-05, "loss": 0.3549, "step": 26017 }, { "epoch": 2.69, "learning_rate": 1.3985927003202676e-05, "loss": 0.6709, "step": 26018 }, { "epoch": 2.69, "learning_rate": 1.3983845657474202e-05, "loss": 0.1988, "step": 26019 }, { "epoch": 2.69, "learning_rate": 1.3981764419562734e-05, "loss": 0.5525, "step": 26020 }, { "epoch": 2.69, "learning_rate": 1.3979683289482288e-05, "loss": 0.4252, "step": 26021 }, { "epoch": 2.69, "learning_rate": 1.3977602267246857e-05, "loss": 0.5437, "step": 26022 }, { "epoch": 2.69, "learning_rate": 1.3975521352870459e-05, "loss": 0.2236, "step": 26023 }, { "epoch": 2.69, "learning_rate": 1.39734405463671e-05, "loss": 0.6987, "step": 26024 }, { "epoch": 2.69, "learning_rate": 1.39713598477508e-05, "loss": 0.2958, "step": 26025 }, { "epoch": 2.69, "learning_rate": 1.3969279257035547e-05, "loss": 0.1581, "step": 26026 }, { "epoch": 2.69, "learning_rate": 1.3967198774235353e-05, "loss": 0.7266, "step": 26027 }, { "epoch": 2.69, "learning_rate": 1.3965118399364224e-05, "loss": 0.7, "step": 26028 }, { "epoch": 2.69, "learning_rate": 1.3963038132436167e-05, "loss": 0.2158, "step": 26029 }, { "epoch": 2.69, "learning_rate": 1.3960957973465186e-05, "loss": 0.3111, "step": 26030 }, { "epoch": 2.69, "learning_rate": 1.395887792246528e-05, "loss": 0.1358, "step": 26031 }, { "epoch": 2.69, "learning_rate": 1.395679797945046e-05, "loss": 0.277, "step": 26032 }, { "epoch": 2.69, "learning_rate": 1.3954718144434728e-05, "loss": 0.202, "step": 26033 }, { "epoch": 2.69, "learning_rate": 1.3952638417432071e-05, "loss": 0.2289, "step": 26034 }, { "epoch": 2.69, "learning_rate": 1.3950558798456509e-05, "loss": 0.2348, "step": 26035 }, { "epoch": 2.69, "learning_rate": 1.3948479287522017e-05, "loss": 0.1284, "step": 26036 }, { "epoch": 2.69, "learning_rate": 1.394639988464261e-05, "loss": 0.0514, "step": 26037 }, { "epoch": 2.69, "learning_rate": 1.3944320589832278e-05, "loss": 0.5356, "step": 26038 }, { "epoch": 2.69, "learning_rate": 1.3942241403105027e-05, "loss": 0.0658, "step": 26039 }, { "epoch": 2.69, "learning_rate": 1.3940162324474846e-05, "loss": 0.1702, "step": 26040 }, { "epoch": 2.69, "learning_rate": 1.3938083353955736e-05, "loss": 0.5513, "step": 26041 }, { "epoch": 2.69, "learning_rate": 1.393600449156169e-05, "loss": 0.2407, "step": 26042 }, { "epoch": 2.69, "learning_rate": 1.3933925737306716e-05, "loss": 0.3337, "step": 26043 }, { "epoch": 2.69, "learning_rate": 1.3931847091204782e-05, "loss": 0.1296, "step": 26044 }, { "epoch": 2.69, "learning_rate": 1.3929768553269897e-05, "loss": 0.5401, "step": 26045 }, { "epoch": 2.69, "learning_rate": 1.3927690123516059e-05, "loss": 0.689, "step": 26046 }, { "epoch": 2.69, "learning_rate": 1.3925611801957238e-05, "loss": 0.2665, "step": 26047 }, { "epoch": 2.69, "learning_rate": 1.3923533588607437e-05, "loss": 0.2288, "step": 26048 }, { "epoch": 2.69, "learning_rate": 1.3921455483480649e-05, "loss": 0.7396, "step": 26049 }, { "epoch": 2.69, "learning_rate": 1.3919377486590858e-05, "loss": 0.5177, "step": 26050 }, { "epoch": 2.69, "learning_rate": 1.3917299597952057e-05, "loss": 0.1171, "step": 26051 }, { "epoch": 2.69, "learning_rate": 1.3915221817578231e-05, "loss": 0.2339, "step": 26052 }, { "epoch": 2.69, "learning_rate": 1.391314414548338e-05, "loss": 0.4378, "step": 26053 }, { "epoch": 2.69, "learning_rate": 1.3911066581681469e-05, "loss": 0.3403, "step": 26054 }, { "epoch": 2.69, "learning_rate": 1.3908989126186491e-05, "loss": 0.2149, "step": 26055 }, { "epoch": 2.69, "learning_rate": 1.3906911779012433e-05, "loss": 0.4879, "step": 26056 }, { "epoch": 2.69, "learning_rate": 1.390483454017328e-05, "loss": 0.2142, "step": 26057 }, { "epoch": 2.69, "learning_rate": 1.3902757409683014e-05, "loss": 0.5109, "step": 26058 }, { "epoch": 2.69, "learning_rate": 1.3900680387555631e-05, "loss": 0.9504, "step": 26059 }, { "epoch": 2.7, "learning_rate": 1.389860347380509e-05, "loss": 0.2134, "step": 26060 }, { "epoch": 2.7, "learning_rate": 1.3896526668445384e-05, "loss": 0.5166, "step": 26061 }, { "epoch": 2.7, "learning_rate": 1.3894449971490491e-05, "loss": 0.2012, "step": 26062 }, { "epoch": 2.7, "learning_rate": 1.38923733829544e-05, "loss": 0.4352, "step": 26063 }, { "epoch": 2.7, "learning_rate": 1.3890296902851077e-05, "loss": 0.2031, "step": 26064 }, { "epoch": 2.7, "learning_rate": 1.3888220531194502e-05, "loss": 0.1459, "step": 26065 }, { "epoch": 2.7, "learning_rate": 1.388614426799866e-05, "loss": 0.3039, "step": 26066 }, { "epoch": 2.7, "learning_rate": 1.3884068113277521e-05, "loss": 0.9652, "step": 26067 }, { "epoch": 2.7, "learning_rate": 1.3881992067045067e-05, "loss": 0.3758, "step": 26068 }, { "epoch": 2.7, "learning_rate": 1.387991612931527e-05, "loss": 0.3959, "step": 26069 }, { "epoch": 2.7, "learning_rate": 1.3877840300102109e-05, "loss": 0.2196, "step": 26070 }, { "epoch": 2.7, "learning_rate": 1.387576457941956e-05, "loss": 0.7006, "step": 26071 }, { "epoch": 2.7, "learning_rate": 1.3873688967281584e-05, "loss": 0.5881, "step": 26072 }, { "epoch": 2.7, "learning_rate": 1.3871613463702171e-05, "loss": 0.2708, "step": 26073 }, { "epoch": 2.7, "learning_rate": 1.386953806869527e-05, "loss": 1.0273, "step": 26074 }, { "epoch": 2.7, "learning_rate": 1.3867462782274866e-05, "loss": 0.4335, "step": 26075 }, { "epoch": 2.7, "learning_rate": 1.3865387604454928e-05, "loss": 0.4563, "step": 26076 }, { "epoch": 2.7, "learning_rate": 1.3863312535249426e-05, "loss": 0.411, "step": 26077 }, { "epoch": 2.7, "learning_rate": 1.3861237574672327e-05, "loss": 0.2616, "step": 26078 }, { "epoch": 2.7, "learning_rate": 1.38591627227376e-05, "loss": 0.3423, "step": 26079 }, { "epoch": 2.7, "learning_rate": 1.3857087979459214e-05, "loss": 0.1871, "step": 26080 }, { "epoch": 2.7, "learning_rate": 1.3855013344851144e-05, "loss": 0.4634, "step": 26081 }, { "epoch": 2.7, "learning_rate": 1.3852938818927338e-05, "loss": 0.2043, "step": 26082 }, { "epoch": 2.7, "learning_rate": 1.3850864401701767e-05, "loss": 0.3585, "step": 26083 }, { "epoch": 2.7, "learning_rate": 1.3848790093188411e-05, "loss": 0.2793, "step": 26084 }, { "epoch": 2.7, "learning_rate": 1.3846715893401211e-05, "loss": 0.5029, "step": 26085 }, { "epoch": 2.7, "learning_rate": 1.384464180235414e-05, "loss": 0.213, "step": 26086 }, { "epoch": 2.7, "learning_rate": 1.3842567820061158e-05, "loss": 0.2476, "step": 26087 }, { "epoch": 2.7, "learning_rate": 1.3840493946536232e-05, "loss": 0.0702, "step": 26088 }, { "epoch": 2.7, "learning_rate": 1.3838420181793317e-05, "loss": 0.751, "step": 26089 }, { "epoch": 2.7, "learning_rate": 1.3836346525846376e-05, "loss": 0.1049, "step": 26090 }, { "epoch": 2.7, "learning_rate": 1.383427297870938e-05, "loss": 0.4467, "step": 26091 }, { "epoch": 2.7, "learning_rate": 1.3832199540396265e-05, "loss": 0.4553, "step": 26092 }, { "epoch": 2.7, "learning_rate": 1.3830126210921e-05, "loss": 0.17, "step": 26093 }, { "epoch": 2.7, "learning_rate": 1.382805299029754e-05, "loss": 0.706, "step": 26094 }, { "epoch": 2.7, "learning_rate": 1.3825979878539846e-05, "loss": 0.4621, "step": 26095 }, { "epoch": 2.7, "learning_rate": 1.382390687566188e-05, "loss": 0.2686, "step": 26096 }, { "epoch": 2.7, "learning_rate": 1.3821833981677578e-05, "loss": 0.3538, "step": 26097 }, { "epoch": 2.7, "learning_rate": 1.3819761196600902e-05, "loss": 0.7691, "step": 26098 }, { "epoch": 2.7, "learning_rate": 1.3817688520445812e-05, "loss": 0.27, "step": 26099 }, { "epoch": 2.7, "learning_rate": 1.3815615953226252e-05, "loss": 0.0557, "step": 26100 }, { "epoch": 2.7, "learning_rate": 1.3813543494956191e-05, "loss": 0.7188, "step": 26101 }, { "epoch": 2.7, "learning_rate": 1.3811471145649559e-05, "loss": 0.2154, "step": 26102 }, { "epoch": 2.7, "learning_rate": 1.3809398905320315e-05, "loss": 0.1184, "step": 26103 }, { "epoch": 2.7, "learning_rate": 1.3807326773982409e-05, "loss": 0.4676, "step": 26104 }, { "epoch": 2.7, "learning_rate": 1.3805254751649787e-05, "loss": 0.1452, "step": 26105 }, { "epoch": 2.7, "learning_rate": 1.3803182838336403e-05, "loss": 0.3847, "step": 26106 }, { "epoch": 2.7, "learning_rate": 1.3801111034056206e-05, "loss": 0.4926, "step": 26107 }, { "epoch": 2.7, "learning_rate": 1.3799039338823134e-05, "loss": 0.2497, "step": 26108 }, { "epoch": 2.7, "learning_rate": 1.3796967752651152e-05, "loss": 0.5891, "step": 26109 }, { "epoch": 2.7, "learning_rate": 1.3794896275554181e-05, "loss": 0.3774, "step": 26110 }, { "epoch": 2.7, "learning_rate": 1.3792824907546187e-05, "loss": 0.2404, "step": 26111 }, { "epoch": 2.7, "learning_rate": 1.3790753648641092e-05, "loss": 0.237, "step": 26112 }, { "epoch": 2.7, "learning_rate": 1.378868249885285e-05, "loss": 0.0543, "step": 26113 }, { "epoch": 2.7, "learning_rate": 1.3786611458195408e-05, "loss": 0.2812, "step": 26114 }, { "epoch": 2.7, "learning_rate": 1.37845405266827e-05, "loss": 0.5931, "step": 26115 }, { "epoch": 2.7, "learning_rate": 1.3782469704328672e-05, "loss": 0.175, "step": 26116 }, { "epoch": 2.7, "learning_rate": 1.3780398991147266e-05, "loss": 0.2282, "step": 26117 }, { "epoch": 2.7, "learning_rate": 1.3778328387152416e-05, "loss": 0.1466, "step": 26118 }, { "epoch": 2.7, "learning_rate": 1.3776257892358078e-05, "loss": 0.2368, "step": 26119 }, { "epoch": 2.7, "learning_rate": 1.377418750677816e-05, "loss": 0.8339, "step": 26120 }, { "epoch": 2.7, "learning_rate": 1.377211723042663e-05, "loss": 0.255, "step": 26121 }, { "epoch": 2.7, "learning_rate": 1.3770047063317396e-05, "loss": 0.4877, "step": 26122 }, { "epoch": 2.7, "learning_rate": 1.3767977005464411e-05, "loss": 0.1316, "step": 26123 }, { "epoch": 2.7, "learning_rate": 1.3765907056881605e-05, "loss": 0.256, "step": 26124 }, { "epoch": 2.7, "learning_rate": 1.3763837217582915e-05, "loss": 0.1747, "step": 26125 }, { "epoch": 2.7, "learning_rate": 1.376176748758227e-05, "loss": 0.2135, "step": 26126 }, { "epoch": 2.7, "learning_rate": 1.375969786689361e-05, "loss": 0.4196, "step": 26127 }, { "epoch": 2.7, "learning_rate": 1.3757628355530865e-05, "loss": 0.5097, "step": 26128 }, { "epoch": 2.7, "learning_rate": 1.3755558953507976e-05, "loss": 0.1901, "step": 26129 }, { "epoch": 2.7, "learning_rate": 1.375348966083885e-05, "loss": 0.3238, "step": 26130 }, { "epoch": 2.7, "learning_rate": 1.3751420477537433e-05, "loss": 0.1463, "step": 26131 }, { "epoch": 2.7, "learning_rate": 1.3749351403617649e-05, "loss": 0.2864, "step": 26132 }, { "epoch": 2.7, "learning_rate": 1.3747282439093432e-05, "loss": 0.3297, "step": 26133 }, { "epoch": 2.7, "learning_rate": 1.3745213583978716e-05, "loss": 0.1477, "step": 26134 }, { "epoch": 2.7, "learning_rate": 1.3743144838287406e-05, "loss": 0.2833, "step": 26135 }, { "epoch": 2.7, "learning_rate": 1.3741076202033445e-05, "loss": 0.4739, "step": 26136 }, { "epoch": 2.7, "learning_rate": 1.3739007675230753e-05, "loss": 0.2088, "step": 26137 }, { "epoch": 2.7, "learning_rate": 1.3736939257893256e-05, "loss": 0.2456, "step": 26138 }, { "epoch": 2.7, "learning_rate": 1.3734870950034889e-05, "loss": 0.5217, "step": 26139 }, { "epoch": 2.7, "learning_rate": 1.3732802751669554e-05, "loss": 0.4239, "step": 26140 }, { "epoch": 2.7, "learning_rate": 1.3730734662811186e-05, "loss": 0.4281, "step": 26141 }, { "epoch": 2.7, "learning_rate": 1.3728666683473707e-05, "loss": 0.2379, "step": 26142 }, { "epoch": 2.7, "learning_rate": 1.3726598813671036e-05, "loss": 0.1111, "step": 26143 }, { "epoch": 2.7, "learning_rate": 1.3724531053417091e-05, "loss": 0.0762, "step": 26144 }, { "epoch": 2.7, "learning_rate": 1.3722463402725797e-05, "loss": 0.3035, "step": 26145 }, { "epoch": 2.7, "learning_rate": 1.3720395861611082e-05, "loss": 0.3173, "step": 26146 }, { "epoch": 2.7, "learning_rate": 1.3718328430086842e-05, "loss": 0.1198, "step": 26147 }, { "epoch": 2.7, "learning_rate": 1.3716261108167006e-05, "loss": 0.3464, "step": 26148 }, { "epoch": 2.7, "learning_rate": 1.3714193895865499e-05, "loss": 0.788, "step": 26149 }, { "epoch": 2.7, "learning_rate": 1.3712126793196217e-05, "loss": 0.6425, "step": 26150 }, { "epoch": 2.7, "learning_rate": 1.371005980017309e-05, "loss": 0.1581, "step": 26151 }, { "epoch": 2.7, "learning_rate": 1.3707992916810025e-05, "loss": 0.2092, "step": 26152 }, { "epoch": 2.7, "learning_rate": 1.3705926143120941e-05, "loss": 0.2921, "step": 26153 }, { "epoch": 2.7, "learning_rate": 1.3703859479119752e-05, "loss": 0.4121, "step": 26154 }, { "epoch": 2.7, "learning_rate": 1.3701792924820363e-05, "loss": 0.2349, "step": 26155 }, { "epoch": 2.71, "learning_rate": 1.3699726480236695e-05, "loss": 0.7201, "step": 26156 }, { "epoch": 2.71, "learning_rate": 1.3697660145382661e-05, "loss": 0.375, "step": 26157 }, { "epoch": 2.71, "learning_rate": 1.3695593920272157e-05, "loss": 0.3932, "step": 26158 }, { "epoch": 2.71, "learning_rate": 1.3693527804919112e-05, "loss": 0.4085, "step": 26159 }, { "epoch": 2.71, "learning_rate": 1.3691461799337407e-05, "loss": 0.0849, "step": 26160 }, { "epoch": 2.71, "learning_rate": 1.368939590354097e-05, "loss": 0.8488, "step": 26161 }, { "epoch": 2.71, "learning_rate": 1.36873301175437e-05, "loss": 0.2024, "step": 26162 }, { "epoch": 2.71, "learning_rate": 1.3685264441359508e-05, "loss": 0.0552, "step": 26163 }, { "epoch": 2.71, "learning_rate": 1.3683198875002298e-05, "loss": 0.2156, "step": 26164 }, { "epoch": 2.71, "learning_rate": 1.3681133418485974e-05, "loss": 0.3148, "step": 26165 }, { "epoch": 2.71, "learning_rate": 1.3679068071824447e-05, "loss": 0.1513, "step": 26166 }, { "epoch": 2.71, "learning_rate": 1.3677002835031622e-05, "loss": 0.5672, "step": 26167 }, { "epoch": 2.71, "learning_rate": 1.3674937708121382e-05, "loss": 0.2987, "step": 26168 }, { "epoch": 2.71, "learning_rate": 1.3672872691107645e-05, "loss": 0.4105, "step": 26169 }, { "epoch": 2.71, "learning_rate": 1.3670807784004306e-05, "loss": 0.3123, "step": 26170 }, { "epoch": 2.71, "learning_rate": 1.3668742986825268e-05, "loss": 0.3927, "step": 26171 }, { "epoch": 2.71, "learning_rate": 1.3666678299584444e-05, "loss": 0.3828, "step": 26172 }, { "epoch": 2.71, "learning_rate": 1.3664613722295708e-05, "loss": 0.69, "step": 26173 }, { "epoch": 2.71, "learning_rate": 1.3662549254972971e-05, "loss": 0.1211, "step": 26174 }, { "epoch": 2.71, "learning_rate": 1.366048489763013e-05, "loss": 0.1061, "step": 26175 }, { "epoch": 2.71, "learning_rate": 1.3658420650281082e-05, "loss": 0.1797, "step": 26176 }, { "epoch": 2.71, "learning_rate": 1.3656356512939732e-05, "loss": 0.1727, "step": 26177 }, { "epoch": 2.71, "learning_rate": 1.3654292485619956e-05, "loss": 0.3594, "step": 26178 }, { "epoch": 2.71, "learning_rate": 1.3652228568335658e-05, "loss": 0.6433, "step": 26179 }, { "epoch": 2.71, "learning_rate": 1.3650164761100731e-05, "loss": 0.3322, "step": 26180 }, { "epoch": 2.71, "learning_rate": 1.3648101063929071e-05, "loss": 0.177, "step": 26181 }, { "epoch": 2.71, "learning_rate": 1.3646037476834569e-05, "loss": 0.5212, "step": 26182 }, { "epoch": 2.71, "learning_rate": 1.3643973999831112e-05, "loss": 0.13, "step": 26183 }, { "epoch": 2.71, "learning_rate": 1.364191063293261e-05, "loss": 0.0863, "step": 26184 }, { "epoch": 2.71, "learning_rate": 1.3639847376152927e-05, "loss": 0.454, "step": 26185 }, { "epoch": 2.71, "learning_rate": 1.3637784229505962e-05, "loss": 0.5736, "step": 26186 }, { "epoch": 2.71, "learning_rate": 1.3635721193005617e-05, "loss": 0.3627, "step": 26187 }, { "epoch": 2.71, "learning_rate": 1.3633658266665756e-05, "loss": 0.0733, "step": 26188 }, { "epoch": 2.71, "learning_rate": 1.3631595450500278e-05, "loss": 0.2948, "step": 26189 }, { "epoch": 2.71, "learning_rate": 1.3629532744523072e-05, "loss": 0.272, "step": 26190 }, { "epoch": 2.71, "learning_rate": 1.3627470148748017e-05, "loss": 0.2037, "step": 26191 }, { "epoch": 2.71, "learning_rate": 1.3625407663189007e-05, "loss": 0.5699, "step": 26192 }, { "epoch": 2.71, "learning_rate": 1.3623345287859919e-05, "loss": 0.4062, "step": 26193 }, { "epoch": 2.71, "learning_rate": 1.3621283022774639e-05, "loss": 0.7229, "step": 26194 }, { "epoch": 2.71, "learning_rate": 1.3619220867947061e-05, "loss": 0.2684, "step": 26195 }, { "epoch": 2.71, "learning_rate": 1.3617158823391042e-05, "loss": 0.4259, "step": 26196 }, { "epoch": 2.71, "learning_rate": 1.361509688912049e-05, "loss": 0.3951, "step": 26197 }, { "epoch": 2.71, "learning_rate": 1.361303506514926e-05, "loss": 0.2624, "step": 26198 }, { "epoch": 2.71, "learning_rate": 1.3610973351491245e-05, "loss": 0.1456, "step": 26199 }, { "epoch": 2.71, "learning_rate": 1.3608911748160317e-05, "loss": 0.4374, "step": 26200 }, { "epoch": 2.71, "learning_rate": 1.3606850255170364e-05, "loss": 0.3233, "step": 26201 }, { "epoch": 2.71, "learning_rate": 1.360478887253526e-05, "loss": 0.194, "step": 26202 }, { "epoch": 2.71, "learning_rate": 1.3602727600268883e-05, "loss": 0.635, "step": 26203 }, { "epoch": 2.71, "learning_rate": 1.3600666438385106e-05, "loss": 0.7804, "step": 26204 }, { "epoch": 2.71, "learning_rate": 1.3598605386897814e-05, "loss": 0.2387, "step": 26205 }, { "epoch": 2.71, "learning_rate": 1.3596544445820862e-05, "loss": 0.2137, "step": 26206 }, { "epoch": 2.71, "learning_rate": 1.3594483615168137e-05, "loss": 0.1175, "step": 26207 }, { "epoch": 2.71, "learning_rate": 1.3592422894953508e-05, "loss": 0.9602, "step": 26208 }, { "epoch": 2.71, "learning_rate": 1.359036228519086e-05, "loss": 0.4715, "step": 26209 }, { "epoch": 2.71, "learning_rate": 1.3588301785894044e-05, "loss": 0.3165, "step": 26210 }, { "epoch": 2.71, "learning_rate": 1.3586241397076937e-05, "loss": 0.3356, "step": 26211 }, { "epoch": 2.71, "learning_rate": 1.3584181118753415e-05, "loss": 0.3984, "step": 26212 }, { "epoch": 2.71, "learning_rate": 1.3582120950937345e-05, "loss": 0.1386, "step": 26213 }, { "epoch": 2.71, "learning_rate": 1.3580060893642595e-05, "loss": 0.6789, "step": 26214 }, { "epoch": 2.71, "learning_rate": 1.3578000946883046e-05, "loss": 0.8332, "step": 26215 }, { "epoch": 2.71, "learning_rate": 1.357594111067254e-05, "loss": 0.3062, "step": 26216 }, { "epoch": 2.71, "learning_rate": 1.3573881385024952e-05, "loss": 0.1845, "step": 26217 }, { "epoch": 2.71, "learning_rate": 1.3571821769954155e-05, "loss": 0.1518, "step": 26218 }, { "epoch": 2.71, "learning_rate": 1.356976226547401e-05, "loss": 0.4349, "step": 26219 }, { "epoch": 2.71, "learning_rate": 1.356770287159838e-05, "loss": 0.2811, "step": 26220 }, { "epoch": 2.71, "learning_rate": 1.3565643588341125e-05, "loss": 0.4788, "step": 26221 }, { "epoch": 2.71, "learning_rate": 1.356358441571613e-05, "loss": 0.4687, "step": 26222 }, { "epoch": 2.71, "learning_rate": 1.3561525353737226e-05, "loss": 0.5395, "step": 26223 }, { "epoch": 2.71, "learning_rate": 1.3559466402418285e-05, "loss": 0.1434, "step": 26224 }, { "epoch": 2.71, "learning_rate": 1.355740756177318e-05, "loss": 0.5584, "step": 26225 }, { "epoch": 2.71, "learning_rate": 1.3555348831815753e-05, "loss": 0.2135, "step": 26226 }, { "epoch": 2.71, "learning_rate": 1.3553290212559868e-05, "loss": 0.7411, "step": 26227 }, { "epoch": 2.71, "learning_rate": 1.355123170401938e-05, "loss": 0.4992, "step": 26228 }, { "epoch": 2.71, "learning_rate": 1.3549173306208159e-05, "loss": 0.599, "step": 26229 }, { "epoch": 2.71, "learning_rate": 1.354711501914005e-05, "loss": 0.1794, "step": 26230 }, { "epoch": 2.71, "learning_rate": 1.3545056842828914e-05, "loss": 1.7115, "step": 26231 }, { "epoch": 2.71, "learning_rate": 1.3542998777288605e-05, "loss": 0.9778, "step": 26232 }, { "epoch": 2.71, "learning_rate": 1.354094082253299e-05, "loss": 0.2801, "step": 26233 }, { "epoch": 2.71, "learning_rate": 1.3538882978575898e-05, "loss": 0.0427, "step": 26234 }, { "epoch": 2.71, "learning_rate": 1.3536825245431206e-05, "loss": 0.2091, "step": 26235 }, { "epoch": 2.71, "learning_rate": 1.3534767623112743e-05, "loss": 0.3178, "step": 26236 }, { "epoch": 2.71, "learning_rate": 1.353271011163437e-05, "loss": 0.128, "step": 26237 }, { "epoch": 2.71, "learning_rate": 1.3530652711009943e-05, "loss": 0.1795, "step": 26238 }, { "epoch": 2.71, "learning_rate": 1.3528595421253304e-05, "loss": 0.2523, "step": 26239 }, { "epoch": 2.71, "learning_rate": 1.352653824237831e-05, "loss": 0.7478, "step": 26240 }, { "epoch": 2.71, "learning_rate": 1.3524481174398806e-05, "loss": 0.6736, "step": 26241 }, { "epoch": 2.71, "learning_rate": 1.3522424217328639e-05, "loss": 0.2566, "step": 26242 }, { "epoch": 2.71, "learning_rate": 1.3520367371181669e-05, "loss": 0.6036, "step": 26243 }, { "epoch": 2.71, "learning_rate": 1.3518310635971715e-05, "loss": 0.5558, "step": 26244 }, { "epoch": 2.71, "learning_rate": 1.3516254011712642e-05, "loss": 0.3826, "step": 26245 }, { "epoch": 2.71, "learning_rate": 1.3514197498418286e-05, "loss": 0.2709, "step": 26246 }, { "epoch": 2.71, "learning_rate": 1.351214109610251e-05, "loss": 0.5871, "step": 26247 }, { "epoch": 2.71, "learning_rate": 1.351008480477913e-05, "loss": 0.6936, "step": 26248 }, { "epoch": 2.71, "learning_rate": 1.3508028624462e-05, "loss": 0.7915, "step": 26249 }, { "epoch": 2.71, "learning_rate": 1.3505972555164964e-05, "loss": 0.1172, "step": 26250 }, { "epoch": 2.71, "learning_rate": 1.3503916596901858e-05, "loss": 0.0811, "step": 26251 }, { "epoch": 2.72, "learning_rate": 1.3501860749686528e-05, "loss": 0.5418, "step": 26252 }, { "epoch": 2.72, "learning_rate": 1.3499805013532823e-05, "loss": 0.4149, "step": 26253 }, { "epoch": 2.72, "learning_rate": 1.3497749388454556e-05, "loss": 0.2018, "step": 26254 }, { "epoch": 2.72, "learning_rate": 1.349569387446558e-05, "loss": 0.3587, "step": 26255 }, { "epoch": 2.72, "learning_rate": 1.3493638471579735e-05, "loss": 0.1391, "step": 26256 }, { "epoch": 2.72, "learning_rate": 1.349158317981085e-05, "loss": 0.3445, "step": 26257 }, { "epoch": 2.72, "learning_rate": 1.3489527999172764e-05, "loss": 0.1858, "step": 26258 }, { "epoch": 2.72, "learning_rate": 1.3487472929679315e-05, "loss": 0.3002, "step": 26259 }, { "epoch": 2.72, "learning_rate": 1.3485417971344342e-05, "loss": 0.2988, "step": 26260 }, { "epoch": 2.72, "learning_rate": 1.3483363124181665e-05, "loss": 0.4323, "step": 26261 }, { "epoch": 2.72, "learning_rate": 1.3481308388205116e-05, "loss": 0.3166, "step": 26262 }, { "epoch": 2.72, "learning_rate": 1.3479253763428552e-05, "loss": 0.9589, "step": 26263 }, { "epoch": 2.72, "learning_rate": 1.347719924986577e-05, "loss": 0.5353, "step": 26264 }, { "epoch": 2.72, "learning_rate": 1.3475144847530618e-05, "loss": 0.077, "step": 26265 }, { "epoch": 2.72, "learning_rate": 1.3473090556436928e-05, "loss": 0.3616, "step": 26266 }, { "epoch": 2.72, "learning_rate": 1.3471036376598523e-05, "loss": 0.3441, "step": 26267 }, { "epoch": 2.72, "learning_rate": 1.346898230802923e-05, "loss": 0.6182, "step": 26268 }, { "epoch": 2.72, "learning_rate": 1.3466928350742885e-05, "loss": 0.5132, "step": 26269 }, { "epoch": 2.72, "learning_rate": 1.3464874504753309e-05, "loss": 0.0989, "step": 26270 }, { "epoch": 2.72, "learning_rate": 1.346282077007434e-05, "loss": 0.6338, "step": 26271 }, { "epoch": 2.72, "learning_rate": 1.3460767146719779e-05, "loss": 0.2179, "step": 26272 }, { "epoch": 2.72, "learning_rate": 1.3458713634703474e-05, "loss": 0.2323, "step": 26273 }, { "epoch": 2.72, "learning_rate": 1.345666023403923e-05, "loss": 0.4711, "step": 26274 }, { "epoch": 2.72, "learning_rate": 1.3454606944740877e-05, "loss": 0.2976, "step": 26275 }, { "epoch": 2.72, "learning_rate": 1.345255376682224e-05, "loss": 0.4108, "step": 26276 }, { "epoch": 2.72, "learning_rate": 1.3450500700297138e-05, "loss": 0.3751, "step": 26277 }, { "epoch": 2.72, "learning_rate": 1.3448447745179394e-05, "loss": 0.3666, "step": 26278 }, { "epoch": 2.72, "learning_rate": 1.3446394901482826e-05, "loss": 0.2612, "step": 26279 }, { "epoch": 2.72, "learning_rate": 1.3444342169221256e-05, "loss": 0.1083, "step": 26280 }, { "epoch": 2.72, "learning_rate": 1.3442289548408512e-05, "loss": 0.4644, "step": 26281 }, { "epoch": 2.72, "learning_rate": 1.3440237039058387e-05, "loss": 2.2685, "step": 26282 }, { "epoch": 2.72, "learning_rate": 1.3438184641184714e-05, "loss": 0.2599, "step": 26283 }, { "epoch": 2.72, "learning_rate": 1.3436132354801304e-05, "loss": 0.2662, "step": 26284 }, { "epoch": 2.72, "learning_rate": 1.3434080179921991e-05, "loss": 0.3798, "step": 26285 }, { "epoch": 2.72, "learning_rate": 1.3432028116560559e-05, "loss": 0.345, "step": 26286 }, { "epoch": 2.72, "learning_rate": 1.3429976164730838e-05, "loss": 0.4821, "step": 26287 }, { "epoch": 2.72, "learning_rate": 1.3427924324446643e-05, "loss": 0.3995, "step": 26288 }, { "epoch": 2.72, "learning_rate": 1.342587259572178e-05, "loss": 0.1563, "step": 26289 }, { "epoch": 2.72, "learning_rate": 1.3423820978570067e-05, "loss": 0.4949, "step": 26290 }, { "epoch": 2.72, "learning_rate": 1.3421769473005325e-05, "loss": 0.6508, "step": 26291 }, { "epoch": 2.72, "learning_rate": 1.3419718079041336e-05, "loss": 0.2755, "step": 26292 }, { "epoch": 2.72, "learning_rate": 1.341766679669193e-05, "loss": 0.2057, "step": 26293 }, { "epoch": 2.72, "learning_rate": 1.3415615625970909e-05, "loss": 0.3346, "step": 26294 }, { "epoch": 2.72, "learning_rate": 1.3413564566892085e-05, "loss": 0.307, "step": 26295 }, { "epoch": 2.72, "learning_rate": 1.3411513619469261e-05, "loss": 0.504, "step": 26296 }, { "epoch": 2.72, "learning_rate": 1.3409462783716257e-05, "loss": 0.391, "step": 26297 }, { "epoch": 2.72, "learning_rate": 1.340741205964686e-05, "loss": 0.0955, "step": 26298 }, { "epoch": 2.72, "learning_rate": 1.340536144727488e-05, "loss": 0.0959, "step": 26299 }, { "epoch": 2.72, "learning_rate": 1.3403310946614123e-05, "loss": 0.0976, "step": 26300 }, { "epoch": 2.72, "learning_rate": 1.3401260557678407e-05, "loss": 0.099, "step": 26301 }, { "epoch": 2.72, "learning_rate": 1.3399210280481507e-05, "loss": 0.7377, "step": 26302 }, { "epoch": 2.72, "learning_rate": 1.3397160115037242e-05, "loss": 0.7867, "step": 26303 }, { "epoch": 2.72, "learning_rate": 1.339511006135941e-05, "loss": 0.118, "step": 26304 }, { "epoch": 2.72, "learning_rate": 1.339306011946181e-05, "loss": 0.1748, "step": 26305 }, { "epoch": 2.72, "learning_rate": 1.3391010289358244e-05, "loss": 0.2836, "step": 26306 }, { "epoch": 2.72, "learning_rate": 1.3388960571062512e-05, "loss": 0.3724, "step": 26307 }, { "epoch": 2.72, "learning_rate": 1.3386910964588411e-05, "loss": 0.8845, "step": 26308 }, { "epoch": 2.72, "learning_rate": 1.3384861469949747e-05, "loss": 0.422, "step": 26309 }, { "epoch": 2.72, "learning_rate": 1.33828120871603e-05, "loss": 0.417, "step": 26310 }, { "epoch": 2.72, "learning_rate": 1.338076281623388e-05, "loss": 0.2113, "step": 26311 }, { "epoch": 2.72, "learning_rate": 1.337871365718427e-05, "loss": 0.2215, "step": 26312 }, { "epoch": 2.72, "learning_rate": 1.337666461002527e-05, "loss": 0.1058, "step": 26313 }, { "epoch": 2.72, "learning_rate": 1.3374615674770671e-05, "loss": 0.4043, "step": 26314 }, { "epoch": 2.72, "learning_rate": 1.3372566851434271e-05, "loss": 0.1043, "step": 26315 }, { "epoch": 2.72, "learning_rate": 1.3370518140029862e-05, "loss": 0.1518, "step": 26316 }, { "epoch": 2.72, "learning_rate": 1.336846954057123e-05, "loss": 0.3354, "step": 26317 }, { "epoch": 2.72, "learning_rate": 1.3366421053072174e-05, "loss": 0.3201, "step": 26318 }, { "epoch": 2.72, "learning_rate": 1.336437267754649e-05, "loss": 0.6925, "step": 26319 }, { "epoch": 2.72, "learning_rate": 1.3362324414007944e-05, "loss": 0.0263, "step": 26320 }, { "epoch": 2.72, "learning_rate": 1.3360276262470337e-05, "loss": 0.5977, "step": 26321 }, { "epoch": 2.72, "learning_rate": 1.3358228222947468e-05, "loss": 0.5926, "step": 26322 }, { "epoch": 2.72, "learning_rate": 1.3356180295453103e-05, "loss": 0.4086, "step": 26323 }, { "epoch": 2.72, "learning_rate": 1.3354132480001037e-05, "loss": 0.2953, "step": 26324 }, { "epoch": 2.72, "learning_rate": 1.3352084776605058e-05, "loss": 0.3301, "step": 26325 }, { "epoch": 2.72, "learning_rate": 1.3350037185278946e-05, "loss": 0.408, "step": 26326 }, { "epoch": 2.72, "learning_rate": 1.3347989706036492e-05, "loss": 0.3852, "step": 26327 }, { "epoch": 2.72, "learning_rate": 1.3345942338891472e-05, "loss": 0.1781, "step": 26328 }, { "epoch": 2.72, "learning_rate": 1.3343895083857679e-05, "loss": 0.1104, "step": 26329 }, { "epoch": 2.72, "learning_rate": 1.3341847940948882e-05, "loss": 0.1795, "step": 26330 }, { "epoch": 2.72, "learning_rate": 1.3339800910178864e-05, "loss": 0.1809, "step": 26331 }, { "epoch": 2.72, "learning_rate": 1.333775399156141e-05, "loss": 0.3151, "step": 26332 }, { "epoch": 2.72, "learning_rate": 1.3335707185110295e-05, "loss": 0.1807, "step": 26333 }, { "epoch": 2.72, "learning_rate": 1.3333660490839302e-05, "loss": 0.223, "step": 26334 }, { "epoch": 2.72, "learning_rate": 1.3331613908762217e-05, "loss": 0.8649, "step": 26335 }, { "epoch": 2.72, "learning_rate": 1.3329567438892798e-05, "loss": 0.2746, "step": 26336 }, { "epoch": 2.72, "learning_rate": 1.3327521081244829e-05, "loss": 0.2735, "step": 26337 }, { "epoch": 2.72, "learning_rate": 1.3325474835832085e-05, "loss": 0.0166, "step": 26338 }, { "epoch": 2.72, "learning_rate": 1.3323428702668357e-05, "loss": 0.2063, "step": 26339 }, { "epoch": 2.72, "learning_rate": 1.3321382681767393e-05, "loss": 0.3756, "step": 26340 }, { "epoch": 2.72, "learning_rate": 1.3319336773142978e-05, "loss": 0.229, "step": 26341 }, { "epoch": 2.72, "learning_rate": 1.3317290976808884e-05, "loss": 0.8907, "step": 26342 }, { "epoch": 2.72, "learning_rate": 1.3315245292778885e-05, "loss": 0.4739, "step": 26343 }, { "epoch": 2.72, "learning_rate": 1.3313199721066751e-05, "loss": 0.8031, "step": 26344 }, { "epoch": 2.72, "learning_rate": 1.3311154261686251e-05, "loss": 0.3185, "step": 26345 }, { "epoch": 2.72, "learning_rate": 1.3309108914651158e-05, "loss": 0.3244, "step": 26346 }, { "epoch": 2.73, "learning_rate": 1.3307063679975247e-05, "loss": 0.3203, "step": 26347 }, { "epoch": 2.73, "learning_rate": 1.3305018557672265e-05, "loss": 0.1356, "step": 26348 }, { "epoch": 2.73, "learning_rate": 1.3302973547756006e-05, "loss": 0.0854, "step": 26349 }, { "epoch": 2.73, "learning_rate": 1.330092865024021e-05, "loss": 0.3829, "step": 26350 }, { "epoch": 2.73, "learning_rate": 1.3298883865138657e-05, "loss": 0.3104, "step": 26351 }, { "epoch": 2.73, "learning_rate": 1.3296839192465106e-05, "loss": 0.0799, "step": 26352 }, { "epoch": 2.73, "learning_rate": 1.3294794632233329e-05, "loss": 0.2488, "step": 26353 }, { "epoch": 2.73, "learning_rate": 1.3292750184457083e-05, "loss": 0.0707, "step": 26354 }, { "epoch": 2.73, "learning_rate": 1.3290705849150138e-05, "loss": 0.4391, "step": 26355 }, { "epoch": 2.73, "learning_rate": 1.3288661626326245e-05, "loss": 0.2834, "step": 26356 }, { "epoch": 2.73, "learning_rate": 1.3286617515999187e-05, "loss": 0.0775, "step": 26357 }, { "epoch": 2.73, "learning_rate": 1.3284573518182698e-05, "loss": 0.3658, "step": 26358 }, { "epoch": 2.73, "learning_rate": 1.3282529632890548e-05, "loss": 0.7493, "step": 26359 }, { "epoch": 2.73, "learning_rate": 1.328048586013651e-05, "loss": 0.1796, "step": 26360 }, { "epoch": 2.73, "learning_rate": 1.3278442199934317e-05, "loss": 0.2448, "step": 26361 }, { "epoch": 2.73, "learning_rate": 1.3276398652297737e-05, "loss": 0.2735, "step": 26362 }, { "epoch": 2.73, "learning_rate": 1.3274355217240529e-05, "loss": 0.2025, "step": 26363 }, { "epoch": 2.73, "learning_rate": 1.327231189477645e-05, "loss": 0.5079, "step": 26364 }, { "epoch": 2.73, "learning_rate": 1.3270268684919252e-05, "loss": 0.4607, "step": 26365 }, { "epoch": 2.73, "learning_rate": 1.3268225587682692e-05, "loss": 0.1601, "step": 26366 }, { "epoch": 2.73, "learning_rate": 1.326618260308053e-05, "loss": 0.2451, "step": 26367 }, { "epoch": 2.73, "learning_rate": 1.3264139731126504e-05, "loss": 0.6067, "step": 26368 }, { "epoch": 2.73, "learning_rate": 1.326209697183437e-05, "loss": 0.7556, "step": 26369 }, { "epoch": 2.73, "learning_rate": 1.3260054325217886e-05, "loss": 0.2073, "step": 26370 }, { "epoch": 2.73, "learning_rate": 1.32580117912908e-05, "loss": 0.6325, "step": 26371 }, { "epoch": 2.73, "learning_rate": 1.325596937006686e-05, "loss": 0.3156, "step": 26372 }, { "epoch": 2.73, "learning_rate": 1.3253927061559824e-05, "loss": 0.2291, "step": 26373 }, { "epoch": 2.73, "learning_rate": 1.3251884865783426e-05, "loss": 0.158, "step": 26374 }, { "epoch": 2.73, "learning_rate": 1.324984278275142e-05, "loss": 0.1771, "step": 26375 }, { "epoch": 2.73, "learning_rate": 1.324780081247755e-05, "loss": 0.5919, "step": 26376 }, { "epoch": 2.73, "learning_rate": 1.324575895497558e-05, "loss": 0.3627, "step": 26377 }, { "epoch": 2.73, "learning_rate": 1.3243717210259227e-05, "loss": 0.6279, "step": 26378 }, { "epoch": 2.73, "learning_rate": 1.3241675578342249e-05, "loss": 0.4482, "step": 26379 }, { "epoch": 2.73, "learning_rate": 1.3239634059238392e-05, "loss": 0.2659, "step": 26380 }, { "epoch": 2.73, "learning_rate": 1.3237592652961397e-05, "loss": 0.578, "step": 26381 }, { "epoch": 2.73, "learning_rate": 1.3235551359525005e-05, "loss": 0.4533, "step": 26382 }, { "epoch": 2.73, "learning_rate": 1.3233510178942958e-05, "loss": 0.0857, "step": 26383 }, { "epoch": 2.73, "learning_rate": 1.3231469111228998e-05, "loss": 0.227, "step": 26384 }, { "epoch": 2.73, "learning_rate": 1.3229428156396876e-05, "loss": 0.7155, "step": 26385 }, { "epoch": 2.73, "learning_rate": 1.322738731446031e-05, "loss": 0.5345, "step": 26386 }, { "epoch": 2.73, "learning_rate": 1.3225346585433055e-05, "loss": 0.2152, "step": 26387 }, { "epoch": 2.73, "learning_rate": 1.3223305969328833e-05, "loss": 0.5218, "step": 26388 }, { "epoch": 2.73, "learning_rate": 1.322126546616139e-05, "loss": 0.2891, "step": 26389 }, { "epoch": 2.73, "learning_rate": 1.3219225075944463e-05, "loss": 0.1127, "step": 26390 }, { "epoch": 2.73, "learning_rate": 1.3217184798691787e-05, "loss": 0.0255, "step": 26391 }, { "epoch": 2.73, "learning_rate": 1.3215144634417093e-05, "loss": 0.3087, "step": 26392 }, { "epoch": 2.73, "learning_rate": 1.321310458313412e-05, "loss": 0.3507, "step": 26393 }, { "epoch": 2.73, "learning_rate": 1.3211064644856599e-05, "loss": 0.299, "step": 26394 }, { "epoch": 2.73, "learning_rate": 1.320902481959827e-05, "loss": 0.1436, "step": 26395 }, { "epoch": 2.73, "learning_rate": 1.320698510737285e-05, "loss": 0.8569, "step": 26396 }, { "epoch": 2.73, "learning_rate": 1.3204945508194077e-05, "loss": 0.1409, "step": 26397 }, { "epoch": 2.73, "learning_rate": 1.3202906022075693e-05, "loss": 0.655, "step": 26398 }, { "epoch": 2.73, "learning_rate": 1.3200866649031401e-05, "loss": 0.1858, "step": 26399 }, { "epoch": 2.73, "learning_rate": 1.3198827389074945e-05, "loss": 0.4414, "step": 26400 }, { "epoch": 2.73, "learning_rate": 1.3196788242220055e-05, "loss": 0.647, "step": 26401 }, { "epoch": 2.73, "learning_rate": 1.3194749208480451e-05, "loss": 0.1762, "step": 26402 }, { "epoch": 2.73, "learning_rate": 1.3192710287869864e-05, "loss": 0.2609, "step": 26403 }, { "epoch": 2.73, "learning_rate": 1.3190671480402022e-05, "loss": 0.4175, "step": 26404 }, { "epoch": 2.73, "learning_rate": 1.3188632786090653e-05, "loss": 0.9058, "step": 26405 }, { "epoch": 2.73, "learning_rate": 1.3186594204949464e-05, "loss": 0.1933, "step": 26406 }, { "epoch": 2.73, "learning_rate": 1.3184555736992193e-05, "loss": 0.8682, "step": 26407 }, { "epoch": 2.73, "learning_rate": 1.3182517382232555e-05, "loss": 0.5663, "step": 26408 }, { "epoch": 2.73, "learning_rate": 1.3180479140684276e-05, "loss": 0.5175, "step": 26409 }, { "epoch": 2.73, "learning_rate": 1.3178441012361085e-05, "loss": 0.4531, "step": 26410 }, { "epoch": 2.73, "learning_rate": 1.3176402997276684e-05, "loss": 0.0466, "step": 26411 }, { "epoch": 2.73, "learning_rate": 1.3174365095444802e-05, "loss": 0.5933, "step": 26412 }, { "epoch": 2.73, "learning_rate": 1.3172327306879159e-05, "loss": 0.3348, "step": 26413 }, { "epoch": 2.73, "learning_rate": 1.3170289631593468e-05, "loss": 0.4802, "step": 26414 }, { "epoch": 2.73, "learning_rate": 1.316825206960146e-05, "loss": 0.6419, "step": 26415 }, { "epoch": 2.73, "learning_rate": 1.316621462091683e-05, "loss": 0.4994, "step": 26416 }, { "epoch": 2.73, "learning_rate": 1.3164177285553308e-05, "loss": 0.5379, "step": 26417 }, { "epoch": 2.73, "learning_rate": 1.3162140063524601e-05, "loss": 0.6038, "step": 26418 }, { "epoch": 2.73, "learning_rate": 1.3160102954844429e-05, "loss": 0.5148, "step": 26419 }, { "epoch": 2.73, "learning_rate": 1.3158065959526505e-05, "loss": 0.224, "step": 26420 }, { "epoch": 2.73, "learning_rate": 1.3156029077584537e-05, "loss": 0.8113, "step": 26421 }, { "epoch": 2.73, "learning_rate": 1.315399230903224e-05, "loss": 0.2035, "step": 26422 }, { "epoch": 2.73, "learning_rate": 1.3151955653883342e-05, "loss": 0.4259, "step": 26423 }, { "epoch": 2.73, "learning_rate": 1.3149919112151522e-05, "loss": 0.5795, "step": 26424 }, { "epoch": 2.73, "learning_rate": 1.3147882683850512e-05, "loss": 0.2967, "step": 26425 }, { "epoch": 2.73, "learning_rate": 1.3145846368994004e-05, "loss": 0.8333, "step": 26426 }, { "epoch": 2.73, "learning_rate": 1.3143810167595716e-05, "loss": 0.1408, "step": 26427 }, { "epoch": 2.73, "learning_rate": 1.3141774079669351e-05, "loss": 0.5007, "step": 26428 }, { "epoch": 2.73, "learning_rate": 1.3139738105228622e-05, "loss": 0.2596, "step": 26429 }, { "epoch": 2.73, "learning_rate": 1.313770224428723e-05, "loss": 0.1707, "step": 26430 }, { "epoch": 2.73, "learning_rate": 1.3135666496858879e-05, "loss": 0.1382, "step": 26431 }, { "epoch": 2.73, "learning_rate": 1.3133630862957276e-05, "loss": 0.4539, "step": 26432 }, { "epoch": 2.73, "learning_rate": 1.3131595342596138e-05, "loss": 0.206, "step": 26433 }, { "epoch": 2.73, "learning_rate": 1.3129559935789139e-05, "loss": 0.1749, "step": 26434 }, { "epoch": 2.73, "learning_rate": 1.3127524642550003e-05, "loss": 0.4522, "step": 26435 }, { "epoch": 2.73, "learning_rate": 1.3125489462892418e-05, "loss": 0.4133, "step": 26436 }, { "epoch": 2.73, "learning_rate": 1.3123454396830087e-05, "loss": 0.0404, "step": 26437 }, { "epoch": 2.73, "learning_rate": 1.3121419444376713e-05, "loss": 0.2598, "step": 26438 }, { "epoch": 2.73, "learning_rate": 1.3119384605545992e-05, "loss": 0.4737, "step": 26439 }, { "epoch": 2.73, "learning_rate": 1.3117349880351625e-05, "loss": 0.8021, "step": 26440 }, { "epoch": 2.73, "learning_rate": 1.3115315268807313e-05, "loss": 0.2384, "step": 26441 }, { "epoch": 2.73, "learning_rate": 1.3113280770926743e-05, "loss": 0.6236, "step": 26442 }, { "epoch": 2.74, "learning_rate": 1.3111246386723626e-05, "loss": 0.6241, "step": 26443 }, { "epoch": 2.74, "learning_rate": 1.3109212116211637e-05, "loss": 0.0943, "step": 26444 }, { "epoch": 2.74, "learning_rate": 1.3107177959404481e-05, "loss": 0.4215, "step": 26445 }, { "epoch": 2.74, "learning_rate": 1.310514391631585e-05, "loss": 0.7596, "step": 26446 }, { "epoch": 2.74, "learning_rate": 1.3103109986959435e-05, "loss": 0.6039, "step": 26447 }, { "epoch": 2.74, "learning_rate": 1.3101076171348944e-05, "loss": 0.3485, "step": 26448 }, { "epoch": 2.74, "learning_rate": 1.3099042469498041e-05, "loss": 0.5988, "step": 26449 }, { "epoch": 2.74, "learning_rate": 1.3097008881420432e-05, "loss": 0.1288, "step": 26450 }, { "epoch": 2.74, "learning_rate": 1.3094975407129804e-05, "loss": 0.0712, "step": 26451 }, { "epoch": 2.74, "learning_rate": 1.3092942046639843e-05, "loss": 0.2401, "step": 26452 }, { "epoch": 2.74, "learning_rate": 1.3090908799964254e-05, "loss": 0.4498, "step": 26453 }, { "epoch": 2.74, "learning_rate": 1.30888756671167e-05, "loss": 0.1656, "step": 26454 }, { "epoch": 2.74, "learning_rate": 1.3086842648110876e-05, "loss": 0.275, "step": 26455 }, { "epoch": 2.74, "learning_rate": 1.3084809742960469e-05, "loss": 0.052, "step": 26456 }, { "epoch": 2.74, "learning_rate": 1.3082776951679166e-05, "loss": 0.5356, "step": 26457 }, { "epoch": 2.74, "learning_rate": 1.3080744274280651e-05, "loss": 0.0607, "step": 26458 }, { "epoch": 2.74, "learning_rate": 1.3078711710778606e-05, "loss": 0.7322, "step": 26459 }, { "epoch": 2.74, "learning_rate": 1.3076679261186726e-05, "loss": 0.1743, "step": 26460 }, { "epoch": 2.74, "learning_rate": 1.307464692551867e-05, "loss": 0.2782, "step": 26461 }, { "epoch": 2.74, "learning_rate": 1.3072614703788133e-05, "loss": 0.7536, "step": 26462 }, { "epoch": 2.74, "learning_rate": 1.3070582596008801e-05, "loss": 0.1383, "step": 26463 }, { "epoch": 2.74, "learning_rate": 1.3068550602194337e-05, "loss": 0.4851, "step": 26464 }, { "epoch": 2.74, "learning_rate": 1.3066518722358427e-05, "loss": 0.3379, "step": 26465 }, { "epoch": 2.74, "learning_rate": 1.3064486956514751e-05, "loss": 0.6291, "step": 26466 }, { "epoch": 2.74, "learning_rate": 1.3062455304676988e-05, "loss": 0.4353, "step": 26467 }, { "epoch": 2.74, "learning_rate": 1.3060423766858811e-05, "loss": 0.5616, "step": 26468 }, { "epoch": 2.74, "learning_rate": 1.30583923430739e-05, "loss": 0.2609, "step": 26469 }, { "epoch": 2.74, "learning_rate": 1.3056361033335924e-05, "loss": 0.0894, "step": 26470 }, { "epoch": 2.74, "learning_rate": 1.3054329837658577e-05, "loss": 0.5153, "step": 26471 }, { "epoch": 2.74, "learning_rate": 1.3052298756055503e-05, "loss": 0.604, "step": 26472 }, { "epoch": 2.74, "learning_rate": 1.3050267788540398e-05, "loss": 0.5214, "step": 26473 }, { "epoch": 2.74, "learning_rate": 1.3048236935126913e-05, "loss": 0.1702, "step": 26474 }, { "epoch": 2.74, "learning_rate": 1.3046206195828734e-05, "loss": 0.895, "step": 26475 }, { "epoch": 2.74, "learning_rate": 1.3044175570659527e-05, "loss": 0.4012, "step": 26476 }, { "epoch": 2.74, "learning_rate": 1.3042145059632959e-05, "loss": 0.6945, "step": 26477 }, { "epoch": 2.74, "learning_rate": 1.3040114662762707e-05, "loss": 0.1957, "step": 26478 }, { "epoch": 2.74, "learning_rate": 1.3038084380062434e-05, "loss": 0.4049, "step": 26479 }, { "epoch": 2.74, "learning_rate": 1.3036054211545806e-05, "loss": 0.0971, "step": 26480 }, { "epoch": 2.74, "learning_rate": 1.3034024157226503e-05, "loss": 0.1497, "step": 26481 }, { "epoch": 2.74, "learning_rate": 1.303199421711817e-05, "loss": 0.1429, "step": 26482 }, { "epoch": 2.74, "learning_rate": 1.3029964391234481e-05, "loss": 0.9024, "step": 26483 }, { "epoch": 2.74, "learning_rate": 1.3027934679589102e-05, "loss": 0.2051, "step": 26484 }, { "epoch": 2.74, "learning_rate": 1.3025905082195695e-05, "loss": 0.0985, "step": 26485 }, { "epoch": 2.74, "learning_rate": 1.3023875599067935e-05, "loss": 0.1484, "step": 26486 }, { "epoch": 2.74, "learning_rate": 1.302184623021946e-05, "loss": 0.2739, "step": 26487 }, { "epoch": 2.74, "learning_rate": 1.3019816975663946e-05, "loss": 0.1214, "step": 26488 }, { "epoch": 2.74, "learning_rate": 1.3017787835415049e-05, "loss": 0.1354, "step": 26489 }, { "epoch": 2.74, "learning_rate": 1.3015758809486431e-05, "loss": 0.2267, "step": 26490 }, { "epoch": 2.74, "learning_rate": 1.3013729897891763e-05, "loss": 0.7204, "step": 26491 }, { "epoch": 2.74, "learning_rate": 1.3011701100644676e-05, "loss": 0.4219, "step": 26492 }, { "epoch": 2.74, "learning_rate": 1.3009672417758847e-05, "loss": 0.4053, "step": 26493 }, { "epoch": 2.74, "learning_rate": 1.3007643849247926e-05, "loss": 0.2972, "step": 26494 }, { "epoch": 2.74, "learning_rate": 1.3005615395125571e-05, "loss": 0.1601, "step": 26495 }, { "epoch": 2.74, "learning_rate": 1.3003587055405434e-05, "loss": 0.6414, "step": 26496 }, { "epoch": 2.74, "learning_rate": 1.3001558830101178e-05, "loss": 0.195, "step": 26497 }, { "epoch": 2.74, "learning_rate": 1.2999530719226457e-05, "loss": 0.3242, "step": 26498 }, { "epoch": 2.74, "learning_rate": 1.2997502722794907e-05, "loss": 0.6198, "step": 26499 }, { "epoch": 2.74, "learning_rate": 1.2995474840820191e-05, "loss": 0.4028, "step": 26500 }, { "epoch": 2.74, "learning_rate": 1.2993447073315969e-05, "loss": 0.5356, "step": 26501 }, { "epoch": 2.74, "learning_rate": 1.2991419420295876e-05, "loss": 0.2107, "step": 26502 }, { "epoch": 2.74, "learning_rate": 1.2989391881773563e-05, "loss": 0.1693, "step": 26503 }, { "epoch": 2.74, "learning_rate": 1.2987364457762685e-05, "loss": 0.1941, "step": 26504 }, { "epoch": 2.74, "learning_rate": 1.2985337148276889e-05, "loss": 0.2961, "step": 26505 }, { "epoch": 2.74, "learning_rate": 1.2983309953329822e-05, "loss": 0.3126, "step": 26506 }, { "epoch": 2.74, "learning_rate": 1.2981282872935131e-05, "loss": 0.4093, "step": 26507 }, { "epoch": 2.74, "learning_rate": 1.2979255907106462e-05, "loss": 0.2154, "step": 26508 }, { "epoch": 2.74, "learning_rate": 1.2977229055857472e-05, "loss": 0.2683, "step": 26509 }, { "epoch": 2.74, "learning_rate": 1.2975202319201779e-05, "loss": 0.2555, "step": 26510 }, { "epoch": 2.74, "learning_rate": 1.2973175697153054e-05, "loss": 0.3098, "step": 26511 }, { "epoch": 2.74, "learning_rate": 1.2971149189724912e-05, "loss": 0.1397, "step": 26512 }, { "epoch": 2.74, "learning_rate": 1.296912279693101e-05, "loss": 0.2211, "step": 26513 }, { "epoch": 2.74, "learning_rate": 1.2967096518784992e-05, "loss": 0.0617, "step": 26514 }, { "epoch": 2.74, "learning_rate": 1.2965070355300494e-05, "loss": 0.8558, "step": 26515 }, { "epoch": 2.74, "learning_rate": 1.2963044306491154e-05, "loss": 0.2483, "step": 26516 }, { "epoch": 2.74, "learning_rate": 1.2961018372370613e-05, "loss": 0.5256, "step": 26517 }, { "epoch": 2.74, "learning_rate": 1.2958992552952514e-05, "loss": 0.3782, "step": 26518 }, { "epoch": 2.74, "learning_rate": 1.2956966848250496e-05, "loss": 0.5159, "step": 26519 }, { "epoch": 2.74, "learning_rate": 1.295494125827818e-05, "loss": 0.7192, "step": 26520 }, { "epoch": 2.74, "learning_rate": 1.2952915783049212e-05, "loss": 0.2133, "step": 26521 }, { "epoch": 2.74, "learning_rate": 1.2950890422577223e-05, "loss": 0.4122, "step": 26522 }, { "epoch": 2.74, "learning_rate": 1.2948865176875868e-05, "loss": 0.0598, "step": 26523 }, { "epoch": 2.74, "learning_rate": 1.2946840045958748e-05, "loss": 0.0922, "step": 26524 }, { "epoch": 2.74, "learning_rate": 1.2944815029839508e-05, "loss": 0.2108, "step": 26525 }, { "epoch": 2.74, "learning_rate": 1.2942790128531787e-05, "loss": 0.1435, "step": 26526 }, { "epoch": 2.74, "learning_rate": 1.294076534204921e-05, "loss": 0.1488, "step": 26527 }, { "epoch": 2.74, "learning_rate": 1.2938740670405408e-05, "loss": 0.3373, "step": 26528 }, { "epoch": 2.74, "learning_rate": 1.2936716113614025e-05, "loss": 0.6021, "step": 26529 }, { "epoch": 2.74, "learning_rate": 1.2934691671688666e-05, "loss": 0.0835, "step": 26530 }, { "epoch": 2.74, "learning_rate": 1.2932667344642972e-05, "loss": 0.1346, "step": 26531 }, { "epoch": 2.74, "learning_rate": 1.2930643132490561e-05, "loss": 0.5915, "step": 26532 }, { "epoch": 2.74, "learning_rate": 1.2928619035245074e-05, "loss": 0.2147, "step": 26533 }, { "epoch": 2.74, "learning_rate": 1.2926595052920124e-05, "loss": 0.1355, "step": 26534 }, { "epoch": 2.74, "learning_rate": 1.292457118552934e-05, "loss": 0.124, "step": 26535 }, { "epoch": 2.74, "learning_rate": 1.2922547433086361e-05, "loss": 0.1167, "step": 26536 }, { "epoch": 2.74, "learning_rate": 1.2920523795604787e-05, "loss": 0.5173, "step": 26537 }, { "epoch": 2.74, "learning_rate": 1.291850027309825e-05, "loss": 0.0289, "step": 26538 }, { "epoch": 2.75, "learning_rate": 1.2916476865580384e-05, "loss": 0.3668, "step": 26539 }, { "epoch": 2.75, "learning_rate": 1.2914453573064787e-05, "loss": 0.1631, "step": 26540 }, { "epoch": 2.75, "learning_rate": 1.2912430395565088e-05, "loss": 0.2595, "step": 26541 }, { "epoch": 2.75, "learning_rate": 1.2910407333094912e-05, "loss": 0.4367, "step": 26542 }, { "epoch": 2.75, "learning_rate": 1.2908384385667877e-05, "loss": 0.2468, "step": 26543 }, { "epoch": 2.75, "learning_rate": 1.2906361553297595e-05, "loss": 0.668, "step": 26544 }, { "epoch": 2.75, "learning_rate": 1.290433883599769e-05, "loss": 0.3974, "step": 26545 }, { "epoch": 2.75, "learning_rate": 1.2902316233781775e-05, "loss": 0.2386, "step": 26546 }, { "epoch": 2.75, "learning_rate": 1.2900293746663478e-05, "loss": 0.7421, "step": 26547 }, { "epoch": 2.75, "learning_rate": 1.2898271374656392e-05, "loss": 0.5143, "step": 26548 }, { "epoch": 2.75, "learning_rate": 1.2896249117774152e-05, "loss": 0.3188, "step": 26549 }, { "epoch": 2.75, "learning_rate": 1.289422697603035e-05, "loss": 0.4765, "step": 26550 }, { "epoch": 2.75, "learning_rate": 1.289220494943861e-05, "loss": 0.2735, "step": 26551 }, { "epoch": 2.75, "learning_rate": 1.2890183038012542e-05, "loss": 0.7735, "step": 26552 }, { "epoch": 2.75, "learning_rate": 1.288816124176576e-05, "loss": 0.2345, "step": 26553 }, { "epoch": 2.75, "learning_rate": 1.2886139560711874e-05, "loss": 0.3625, "step": 26554 }, { "epoch": 2.75, "learning_rate": 1.288411799486449e-05, "loss": 0.1126, "step": 26555 }, { "epoch": 2.75, "learning_rate": 1.288209654423722e-05, "loss": 0.2465, "step": 26556 }, { "epoch": 2.75, "learning_rate": 1.288007520884368e-05, "loss": 0.1527, "step": 26557 }, { "epoch": 2.75, "learning_rate": 1.2878053988697457e-05, "loss": 0.3511, "step": 26558 }, { "epoch": 2.75, "learning_rate": 1.2876032883812167e-05, "loss": 0.1782, "step": 26559 }, { "epoch": 2.75, "learning_rate": 1.2874011894201421e-05, "loss": 0.1812, "step": 26560 }, { "epoch": 2.75, "learning_rate": 1.2871991019878828e-05, "loss": 0.5858, "step": 26561 }, { "epoch": 2.75, "learning_rate": 1.2869970260857975e-05, "loss": 0.0967, "step": 26562 }, { "epoch": 2.75, "learning_rate": 1.2867949617152471e-05, "loss": 0.4736, "step": 26563 }, { "epoch": 2.75, "learning_rate": 1.2865929088775927e-05, "loss": 0.461, "step": 26564 }, { "epoch": 2.75, "learning_rate": 1.2863908675741934e-05, "loss": 0.4684, "step": 26565 }, { "epoch": 2.75, "learning_rate": 1.2861888378064101e-05, "loss": 0.1241, "step": 26566 }, { "epoch": 2.75, "learning_rate": 1.2859868195756034e-05, "loss": 0.2468, "step": 26567 }, { "epoch": 2.75, "learning_rate": 1.2857848128831315e-05, "loss": 0.2324, "step": 26568 }, { "epoch": 2.75, "learning_rate": 1.2855828177303548e-05, "loss": 0.4269, "step": 26569 }, { "epoch": 2.75, "learning_rate": 1.285380834118634e-05, "loss": 0.4965, "step": 26570 }, { "epoch": 2.75, "learning_rate": 1.2851788620493279e-05, "loss": 0.428, "step": 26571 }, { "epoch": 2.75, "learning_rate": 1.2849769015237967e-05, "loss": 0.4419, "step": 26572 }, { "epoch": 2.75, "learning_rate": 1.2847749525433993e-05, "loss": 0.0455, "step": 26573 }, { "epoch": 2.75, "learning_rate": 1.284573015109497e-05, "loss": 0.6888, "step": 26574 }, { "epoch": 2.75, "learning_rate": 1.2843710892234463e-05, "loss": 0.3208, "step": 26575 }, { "epoch": 2.75, "learning_rate": 1.2841691748866087e-05, "loss": 0.5156, "step": 26576 }, { "epoch": 2.75, "learning_rate": 1.2839672721003432e-05, "loss": 0.2265, "step": 26577 }, { "epoch": 2.75, "learning_rate": 1.2837653808660077e-05, "loss": 0.2967, "step": 26578 }, { "epoch": 2.75, "learning_rate": 1.2835635011849619e-05, "loss": 0.5577, "step": 26579 }, { "epoch": 2.75, "learning_rate": 1.2833616330585654e-05, "loss": 0.2697, "step": 26580 }, { "epoch": 2.75, "learning_rate": 1.2831597764881762e-05, "loss": 0.2181, "step": 26581 }, { "epoch": 2.75, "learning_rate": 1.282957931475154e-05, "loss": 0.6347, "step": 26582 }, { "epoch": 2.75, "learning_rate": 1.2827560980208573e-05, "loss": 0.4204, "step": 26583 }, { "epoch": 2.75, "learning_rate": 1.2825542761266444e-05, "loss": 0.6145, "step": 26584 }, { "epoch": 2.75, "learning_rate": 1.2823524657938758e-05, "loss": 0.6676, "step": 26585 }, { "epoch": 2.75, "learning_rate": 1.2821506670239072e-05, "loss": 0.0491, "step": 26586 }, { "epoch": 2.75, "learning_rate": 1.2819488798180995e-05, "loss": 0.2546, "step": 26587 }, { "epoch": 2.75, "learning_rate": 1.2817471041778087e-05, "loss": 0.3675, "step": 26588 }, { "epoch": 2.75, "learning_rate": 1.2815453401043943e-05, "loss": 0.2128, "step": 26589 }, { "epoch": 2.75, "learning_rate": 1.2813435875992147e-05, "loss": 0.5011, "step": 26590 }, { "epoch": 2.75, "learning_rate": 1.2811418466636282e-05, "loss": 0.2326, "step": 26591 }, { "epoch": 2.75, "learning_rate": 1.2809401172989922e-05, "loss": 0.6737, "step": 26592 }, { "epoch": 2.75, "learning_rate": 1.2807383995066656e-05, "loss": 0.1743, "step": 26593 }, { "epoch": 2.75, "learning_rate": 1.2805366932880053e-05, "loss": 0.0911, "step": 26594 }, { "epoch": 2.75, "learning_rate": 1.2803349986443709e-05, "loss": 0.1242, "step": 26595 }, { "epoch": 2.75, "learning_rate": 1.280133315577118e-05, "loss": 0.6986, "step": 26596 }, { "epoch": 2.75, "learning_rate": 1.2799316440876051e-05, "loss": 0.3422, "step": 26597 }, { "epoch": 2.75, "learning_rate": 1.2797299841771898e-05, "loss": 0.2254, "step": 26598 }, { "epoch": 2.75, "learning_rate": 1.279528335847231e-05, "loss": 0.782, "step": 26599 }, { "epoch": 2.75, "learning_rate": 1.2793266990990837e-05, "loss": 0.1681, "step": 26600 }, { "epoch": 2.75, "learning_rate": 1.2791250739341064e-05, "loss": 0.6367, "step": 26601 }, { "epoch": 2.75, "learning_rate": 1.2789234603536568e-05, "loss": 0.4535, "step": 26602 }, { "epoch": 2.75, "learning_rate": 1.2787218583590914e-05, "loss": 0.4589, "step": 26603 }, { "epoch": 2.75, "learning_rate": 1.2785202679517677e-05, "loss": 0.2722, "step": 26604 }, { "epoch": 2.75, "learning_rate": 1.278318689133044e-05, "loss": 0.4388, "step": 26605 }, { "epoch": 2.75, "learning_rate": 1.2781171219042748e-05, "loss": 0.5135, "step": 26606 }, { "epoch": 2.75, "learning_rate": 1.2779155662668184e-05, "loss": 0.2145, "step": 26607 }, { "epoch": 2.75, "learning_rate": 1.2777140222220315e-05, "loss": 0.3738, "step": 26608 }, { "epoch": 2.75, "learning_rate": 1.2775124897712707e-05, "loss": 0.8953, "step": 26609 }, { "epoch": 2.75, "learning_rate": 1.2773109689158928e-05, "loss": 0.2385, "step": 26610 }, { "epoch": 2.75, "learning_rate": 1.2771094596572554e-05, "loss": 0.5673, "step": 26611 }, { "epoch": 2.75, "learning_rate": 1.2769079619967133e-05, "loss": 0.3893, "step": 26612 }, { "epoch": 2.75, "learning_rate": 1.2767064759356231e-05, "loss": 0.2262, "step": 26613 }, { "epoch": 2.75, "learning_rate": 1.2765050014753419e-05, "loss": 0.332, "step": 26614 }, { "epoch": 2.75, "learning_rate": 1.2763035386172269e-05, "loss": 0.4791, "step": 26615 }, { "epoch": 2.75, "learning_rate": 1.276102087362632e-05, "loss": 0.168, "step": 26616 }, { "epoch": 2.75, "learning_rate": 1.2759006477129147e-05, "loss": 0.076, "step": 26617 }, { "epoch": 2.75, "learning_rate": 1.2756992196694306e-05, "loss": 0.6406, "step": 26618 }, { "epoch": 2.75, "learning_rate": 1.275497803233536e-05, "loss": 0.5582, "step": 26619 }, { "epoch": 2.75, "learning_rate": 1.2752963984065866e-05, "loss": 0.9374, "step": 26620 }, { "epoch": 2.75, "learning_rate": 1.2750950051899383e-05, "loss": 0.309, "step": 26621 }, { "epoch": 2.75, "learning_rate": 1.274893623584947e-05, "loss": 0.6221, "step": 26622 }, { "epoch": 2.75, "learning_rate": 1.274692253592969e-05, "loss": 0.3235, "step": 26623 }, { "epoch": 2.75, "learning_rate": 1.2744908952153584e-05, "loss": 0.1676, "step": 26624 }, { "epoch": 2.75, "learning_rate": 1.274289548453472e-05, "loss": 0.1999, "step": 26625 }, { "epoch": 2.75, "learning_rate": 1.2740882133086636e-05, "loss": 0.2624, "step": 26626 }, { "epoch": 2.75, "learning_rate": 1.2738868897822894e-05, "loss": 0.2126, "step": 26627 }, { "epoch": 2.75, "learning_rate": 1.2736855778757052e-05, "loss": 0.3566, "step": 26628 }, { "epoch": 2.75, "learning_rate": 1.2734842775902655e-05, "loss": 0.3293, "step": 26629 }, { "epoch": 2.75, "learning_rate": 1.2732829889273256e-05, "loss": 0.3699, "step": 26630 }, { "epoch": 2.75, "learning_rate": 1.2730817118882407e-05, "loss": 0.5203, "step": 26631 }, { "epoch": 2.75, "learning_rate": 1.2728804464743658e-05, "loss": 0.2649, "step": 26632 }, { "epoch": 2.75, "learning_rate": 1.2726791926870565e-05, "loss": 0.2857, "step": 26633 }, { "epoch": 2.75, "learning_rate": 1.2724779505276654e-05, "loss": 0.2435, "step": 26634 }, { "epoch": 2.76, "learning_rate": 1.272276719997549e-05, "loss": 0.382, "step": 26635 }, { "epoch": 2.76, "learning_rate": 1.2720755010980624e-05, "loss": 0.2153, "step": 26636 }, { "epoch": 2.76, "learning_rate": 1.2718742938305581e-05, "loss": 0.0826, "step": 26637 }, { "epoch": 2.76, "learning_rate": 1.2716730981963914e-05, "loss": 0.4665, "step": 26638 }, { "epoch": 2.76, "learning_rate": 1.2714719141969174e-05, "loss": 0.547, "step": 26639 }, { "epoch": 2.76, "learning_rate": 1.27127074183349e-05, "loss": 0.2785, "step": 26640 }, { "epoch": 2.76, "learning_rate": 1.2710695811074635e-05, "loss": 0.3968, "step": 26641 }, { "epoch": 2.76, "learning_rate": 1.2708684320201917e-05, "loss": 0.4768, "step": 26642 }, { "epoch": 2.76, "learning_rate": 1.2706672945730307e-05, "loss": 0.7394, "step": 26643 }, { "epoch": 2.76, "learning_rate": 1.2704661687673314e-05, "loss": 0.2729, "step": 26644 }, { "epoch": 2.76, "learning_rate": 1.2702650546044492e-05, "loss": 0.4063, "step": 26645 }, { "epoch": 2.76, "learning_rate": 1.2700639520857378e-05, "loss": 0.4185, "step": 26646 }, { "epoch": 2.76, "learning_rate": 1.269862861212551e-05, "loss": 0.1226, "step": 26647 }, { "epoch": 2.76, "learning_rate": 1.269661781986243e-05, "loss": 0.2628, "step": 26648 }, { "epoch": 2.76, "learning_rate": 1.2694607144081678e-05, "loss": 0.4719, "step": 26649 }, { "epoch": 2.76, "learning_rate": 1.2692596584796771e-05, "loss": 0.1776, "step": 26650 }, { "epoch": 2.76, "learning_rate": 1.2690586142021257e-05, "loss": 0.4328, "step": 26651 }, { "epoch": 2.76, "learning_rate": 1.2688575815768665e-05, "loss": 0.4572, "step": 26652 }, { "epoch": 2.76, "learning_rate": 1.2686565606052542e-05, "loss": 0.3817, "step": 26653 }, { "epoch": 2.76, "learning_rate": 1.2684555512886397e-05, "loss": 0.2188, "step": 26654 }, { "epoch": 2.76, "learning_rate": 1.2682545536283773e-05, "loss": 0.5586, "step": 26655 }, { "epoch": 2.76, "learning_rate": 1.2680535676258197e-05, "loss": 0.1997, "step": 26656 }, { "epoch": 2.76, "learning_rate": 1.2678525932823205e-05, "loss": 0.5178, "step": 26657 }, { "epoch": 2.76, "learning_rate": 1.2676516305992326e-05, "loss": 0.608, "step": 26658 }, { "epoch": 2.76, "learning_rate": 1.2674506795779082e-05, "loss": 0.5291, "step": 26659 }, { "epoch": 2.76, "learning_rate": 1.2672497402197008e-05, "loss": 0.2892, "step": 26660 }, { "epoch": 2.76, "learning_rate": 1.2670488125259636e-05, "loss": 0.3679, "step": 26661 }, { "epoch": 2.76, "learning_rate": 1.2668478964980471e-05, "loss": 0.5958, "step": 26662 }, { "epoch": 2.76, "learning_rate": 1.2666469921373062e-05, "loss": 0.8538, "step": 26663 }, { "epoch": 2.76, "learning_rate": 1.266446099445091e-05, "loss": 0.2528, "step": 26664 }, { "epoch": 2.76, "learning_rate": 1.2662452184227553e-05, "loss": 0.8016, "step": 26665 }, { "epoch": 2.76, "learning_rate": 1.2660443490716513e-05, "loss": 0.2091, "step": 26666 }, { "epoch": 2.76, "learning_rate": 1.2658434913931303e-05, "loss": 0.1974, "step": 26667 }, { "epoch": 2.76, "learning_rate": 1.2656426453885457e-05, "loss": 0.319, "step": 26668 }, { "epoch": 2.76, "learning_rate": 1.265441811059249e-05, "loss": 0.2642, "step": 26669 }, { "epoch": 2.76, "learning_rate": 1.2652409884065921e-05, "loss": 0.3211, "step": 26670 }, { "epoch": 2.76, "learning_rate": 1.265040177431928e-05, "loss": 0.615, "step": 26671 }, { "epoch": 2.76, "learning_rate": 1.2648393781366064e-05, "loss": 0.5546, "step": 26672 }, { "epoch": 2.76, "learning_rate": 1.2646385905219803e-05, "loss": 0.528, "step": 26673 }, { "epoch": 2.76, "learning_rate": 1.2644378145894017e-05, "loss": 0.0829, "step": 26674 }, { "epoch": 2.76, "learning_rate": 1.2642370503402212e-05, "loss": 0.2157, "step": 26675 }, { "epoch": 2.76, "learning_rate": 1.2640362977757905e-05, "loss": 0.3197, "step": 26676 }, { "epoch": 2.76, "learning_rate": 1.263835556897461e-05, "loss": 0.7968, "step": 26677 }, { "epoch": 2.76, "learning_rate": 1.2636348277065848e-05, "loss": 0.4018, "step": 26678 }, { "epoch": 2.76, "learning_rate": 1.2634341102045125e-05, "loss": 0.4815, "step": 26679 }, { "epoch": 2.76, "learning_rate": 1.2632334043925952e-05, "loss": 0.2767, "step": 26680 }, { "epoch": 2.76, "learning_rate": 1.2630327102721854e-05, "loss": 0.0757, "step": 26681 }, { "epoch": 2.76, "learning_rate": 1.2628320278446319e-05, "loss": 0.5914, "step": 26682 }, { "epoch": 2.76, "learning_rate": 1.2626313571112868e-05, "loss": 0.6403, "step": 26683 }, { "epoch": 2.76, "learning_rate": 1.2624306980735011e-05, "loss": 0.1351, "step": 26684 }, { "epoch": 2.76, "learning_rate": 1.2622300507326248e-05, "loss": 0.5844, "step": 26685 }, { "epoch": 2.76, "learning_rate": 1.2620294150900097e-05, "loss": 0.4479, "step": 26686 }, { "epoch": 2.76, "learning_rate": 1.2618287911470071e-05, "loss": 0.1944, "step": 26687 }, { "epoch": 2.76, "learning_rate": 1.261628178904965e-05, "loss": 0.1896, "step": 26688 }, { "epoch": 2.76, "learning_rate": 1.2614275783652352e-05, "loss": 0.3367, "step": 26689 }, { "epoch": 2.76, "learning_rate": 1.2612269895291682e-05, "loss": 0.2032, "step": 26690 }, { "epoch": 2.76, "learning_rate": 1.2610264123981156e-05, "loss": 0.5163, "step": 26691 }, { "epoch": 2.76, "learning_rate": 1.260825846973425e-05, "loss": 0.1932, "step": 26692 }, { "epoch": 2.76, "learning_rate": 1.2606252932564478e-05, "loss": 0.4214, "step": 26693 }, { "epoch": 2.76, "learning_rate": 1.2604247512485342e-05, "loss": 0.1946, "step": 26694 }, { "epoch": 2.76, "learning_rate": 1.2602242209510343e-05, "loss": 0.199, "step": 26695 }, { "epoch": 2.76, "learning_rate": 1.2600237023652976e-05, "loss": 0.2981, "step": 26696 }, { "epoch": 2.76, "learning_rate": 1.2598231954926744e-05, "loss": 0.4426, "step": 26697 }, { "epoch": 2.76, "learning_rate": 1.2596227003345143e-05, "loss": 0.2179, "step": 26698 }, { "epoch": 2.76, "learning_rate": 1.2594222168921682e-05, "loss": 0.1668, "step": 26699 }, { "epoch": 2.76, "learning_rate": 1.259221745166983e-05, "loss": 0.4692, "step": 26700 }, { "epoch": 2.76, "learning_rate": 1.259021285160311e-05, "loss": 0.5912, "step": 26701 }, { "epoch": 2.76, "learning_rate": 1.2588208368734993e-05, "loss": 0.9387, "step": 26702 }, { "epoch": 2.76, "learning_rate": 1.2586204003078983e-05, "loss": 0.7056, "step": 26703 }, { "epoch": 2.76, "learning_rate": 1.2584199754648573e-05, "loss": 0.2369, "step": 26704 }, { "epoch": 2.76, "learning_rate": 1.2582195623457253e-05, "loss": 0.5048, "step": 26705 }, { "epoch": 2.76, "learning_rate": 1.2580191609518523e-05, "loss": 0.3228, "step": 26706 }, { "epoch": 2.76, "learning_rate": 1.2578187712845861e-05, "loss": 0.0579, "step": 26707 }, { "epoch": 2.76, "learning_rate": 1.2576183933452763e-05, "loss": 0.168, "step": 26708 }, { "epoch": 2.76, "learning_rate": 1.2574180271352733e-05, "loss": 0.0758, "step": 26709 }, { "epoch": 2.76, "learning_rate": 1.2572176726559232e-05, "loss": 0.0833, "step": 26710 }, { "epoch": 2.76, "learning_rate": 1.2570173299085756e-05, "loss": 0.3595, "step": 26711 }, { "epoch": 2.76, "learning_rate": 1.2568169988945807e-05, "loss": 0.4076, "step": 26712 }, { "epoch": 2.76, "learning_rate": 1.256616679615285e-05, "loss": 0.1572, "step": 26713 }, { "epoch": 2.76, "learning_rate": 1.256416372072038e-05, "loss": 0.349, "step": 26714 }, { "epoch": 2.76, "learning_rate": 1.2562160762661878e-05, "loss": 0.2966, "step": 26715 }, { "epoch": 2.76, "learning_rate": 1.2560157921990831e-05, "loss": 0.4041, "step": 26716 }, { "epoch": 2.76, "learning_rate": 1.2558155198720717e-05, "loss": 0.3649, "step": 26717 }, { "epoch": 2.76, "learning_rate": 1.2556152592865023e-05, "loss": 0.2456, "step": 26718 }, { "epoch": 2.76, "learning_rate": 1.255415010443724e-05, "loss": 0.2124, "step": 26719 }, { "epoch": 2.76, "learning_rate": 1.2552147733450826e-05, "loss": 0.4145, "step": 26720 }, { "epoch": 2.76, "learning_rate": 1.2550145479919271e-05, "loss": 0.6691, "step": 26721 }, { "epoch": 2.76, "learning_rate": 1.2548143343856053e-05, "loss": 0.3383, "step": 26722 }, { "epoch": 2.76, "learning_rate": 1.254614132527465e-05, "loss": 0.1568, "step": 26723 }, { "epoch": 2.76, "learning_rate": 1.2544139424188553e-05, "loss": 0.1138, "step": 26724 }, { "epoch": 2.76, "learning_rate": 1.254213764061121e-05, "loss": 0.3382, "step": 26725 }, { "epoch": 2.76, "learning_rate": 1.2540135974556113e-05, "loss": 0.8417, "step": 26726 }, { "epoch": 2.76, "learning_rate": 1.2538134426036736e-05, "loss": 0.2919, "step": 26727 }, { "epoch": 2.76, "learning_rate": 1.2536132995066555e-05, "loss": 0.1049, "step": 26728 }, { "epoch": 2.76, "learning_rate": 1.2534131681659048e-05, "loss": 0.458, "step": 26729 }, { "epoch": 2.77, "learning_rate": 1.2532130485827668e-05, "loss": 0.4918, "step": 26730 }, { "epoch": 2.77, "learning_rate": 1.2530129407585903e-05, "loss": 0.7473, "step": 26731 }, { "epoch": 2.77, "learning_rate": 1.2528128446947215e-05, "loss": 0.6716, "step": 26732 }, { "epoch": 2.77, "learning_rate": 1.252612760392508e-05, "loss": 0.6537, "step": 26733 }, { "epoch": 2.77, "learning_rate": 1.2524126878532963e-05, "loss": 0.114, "step": 26734 }, { "epoch": 2.77, "learning_rate": 1.2522126270784339e-05, "loss": 0.2089, "step": 26735 }, { "epoch": 2.77, "learning_rate": 1.2520125780692669e-05, "loss": 0.411, "step": 26736 }, { "epoch": 2.77, "learning_rate": 1.2518125408271432e-05, "loss": 0.6408, "step": 26737 }, { "epoch": 2.77, "learning_rate": 1.2516125153534077e-05, "loss": 0.1875, "step": 26738 }, { "epoch": 2.77, "learning_rate": 1.2514125016494084e-05, "loss": 0.2974, "step": 26739 }, { "epoch": 2.77, "learning_rate": 1.25121249971649e-05, "loss": 0.0777, "step": 26740 }, { "epoch": 2.77, "learning_rate": 1.2510125095559998e-05, "loss": 0.366, "step": 26741 }, { "epoch": 2.77, "learning_rate": 1.2508125311692844e-05, "loss": 0.7157, "step": 26742 }, { "epoch": 2.77, "learning_rate": 1.2506125645576898e-05, "loss": 0.2363, "step": 26743 }, { "epoch": 2.77, "learning_rate": 1.2504126097225617e-05, "loss": 0.0943, "step": 26744 }, { "epoch": 2.77, "learning_rate": 1.250212666665247e-05, "loss": 0.1464, "step": 26745 }, { "epoch": 2.77, "learning_rate": 1.2500127353870909e-05, "loss": 0.4091, "step": 26746 }, { "epoch": 2.77, "learning_rate": 1.2498128158894405e-05, "loss": 0.2695, "step": 26747 }, { "epoch": 2.77, "learning_rate": 1.2496129081736399e-05, "loss": 0.2087, "step": 26748 }, { "epoch": 2.77, "learning_rate": 1.249413012241037e-05, "loss": 0.9362, "step": 26749 }, { "epoch": 2.77, "learning_rate": 1.2492131280929748e-05, "loss": 0.467, "step": 26750 }, { "epoch": 2.77, "learning_rate": 1.2490132557308001e-05, "loss": 0.828, "step": 26751 }, { "epoch": 2.77, "learning_rate": 1.2488133951558582e-05, "loss": 0.203, "step": 26752 }, { "epoch": 2.77, "learning_rate": 1.2486135463694953e-05, "loss": 0.0502, "step": 26753 }, { "epoch": 2.77, "learning_rate": 1.248413709373056e-05, "loss": 0.6222, "step": 26754 }, { "epoch": 2.77, "learning_rate": 1.2482138841678855e-05, "loss": 0.3562, "step": 26755 }, { "epoch": 2.77, "learning_rate": 1.2480140707553294e-05, "loss": 0.5257, "step": 26756 }, { "epoch": 2.77, "learning_rate": 1.247814269136734e-05, "loss": 0.5609, "step": 26757 }, { "epoch": 2.77, "learning_rate": 1.2476144793134417e-05, "loss": 0.1763, "step": 26758 }, { "epoch": 2.77, "learning_rate": 1.2474147012867984e-05, "loss": 0.2447, "step": 26759 }, { "epoch": 2.77, "learning_rate": 1.2472149350581498e-05, "loss": 0.0912, "step": 26760 }, { "epoch": 2.77, "learning_rate": 1.24701518062884e-05, "loss": 0.2446, "step": 26761 }, { "epoch": 2.77, "learning_rate": 1.2468154380002144e-05, "loss": 0.1665, "step": 26762 }, { "epoch": 2.77, "learning_rate": 1.2466157071736167e-05, "loss": 0.7832, "step": 26763 }, { "epoch": 2.77, "learning_rate": 1.2464159881503913e-05, "loss": 0.1747, "step": 26764 }, { "epoch": 2.77, "learning_rate": 1.2462162809318834e-05, "loss": 0.3457, "step": 26765 }, { "epoch": 2.77, "learning_rate": 1.2460165855194371e-05, "loss": 0.1369, "step": 26766 }, { "epoch": 2.77, "learning_rate": 1.2458169019143979e-05, "loss": 0.2173, "step": 26767 }, { "epoch": 2.77, "learning_rate": 1.2456172301181077e-05, "loss": 0.5481, "step": 26768 }, { "epoch": 2.77, "learning_rate": 1.2454175701319118e-05, "loss": 0.4332, "step": 26769 }, { "epoch": 2.77, "learning_rate": 1.2452179219571541e-05, "loss": 0.1531, "step": 26770 }, { "epoch": 2.77, "learning_rate": 1.2450182855951791e-05, "loss": 0.5797, "step": 26771 }, { "epoch": 2.77, "learning_rate": 1.2448186610473302e-05, "loss": 0.49, "step": 26772 }, { "epoch": 2.77, "learning_rate": 1.2446190483149511e-05, "loss": 0.172, "step": 26773 }, { "epoch": 2.77, "learning_rate": 1.2444194473993864e-05, "loss": 0.4838, "step": 26774 }, { "epoch": 2.77, "learning_rate": 1.24421985830198e-05, "loss": 0.2026, "step": 26775 }, { "epoch": 2.77, "learning_rate": 1.2440202810240737e-05, "loss": 0.4392, "step": 26776 }, { "epoch": 2.77, "learning_rate": 1.243820715567013e-05, "loss": 0.3445, "step": 26777 }, { "epoch": 2.77, "learning_rate": 1.2436211619321392e-05, "loss": 0.345, "step": 26778 }, { "epoch": 2.77, "learning_rate": 1.243421620120797e-05, "loss": 0.3476, "step": 26779 }, { "epoch": 2.77, "learning_rate": 1.2432220901343294e-05, "loss": 0.174, "step": 26780 }, { "epoch": 2.77, "learning_rate": 1.2430225719740795e-05, "loss": 0.1605, "step": 26781 }, { "epoch": 2.77, "learning_rate": 1.2428230656413906e-05, "loss": 0.2521, "step": 26782 }, { "epoch": 2.77, "learning_rate": 1.2426235711376056e-05, "loss": 0.9499, "step": 26783 }, { "epoch": 2.77, "learning_rate": 1.2424240884640675e-05, "loss": 0.0517, "step": 26784 }, { "epoch": 2.77, "learning_rate": 1.2422246176221205e-05, "loss": 0.4961, "step": 26785 }, { "epoch": 2.77, "learning_rate": 1.242025158613105e-05, "loss": 0.1522, "step": 26786 }, { "epoch": 2.77, "learning_rate": 1.2418257114383656e-05, "loss": 0.1777, "step": 26787 }, { "epoch": 2.77, "learning_rate": 1.2416262760992433e-05, "loss": 0.4456, "step": 26788 }, { "epoch": 2.77, "learning_rate": 1.2414268525970814e-05, "loss": 0.265, "step": 26789 }, { "epoch": 2.77, "learning_rate": 1.2412274409332223e-05, "loss": 0.3729, "step": 26790 }, { "epoch": 2.77, "learning_rate": 1.2410280411090088e-05, "loss": 0.2026, "step": 26791 }, { "epoch": 2.77, "learning_rate": 1.2408286531257829e-05, "loss": 0.6197, "step": 26792 }, { "epoch": 2.77, "learning_rate": 1.2406292769848867e-05, "loss": 0.1529, "step": 26793 }, { "epoch": 2.77, "learning_rate": 1.2404299126876628e-05, "loss": 0.355, "step": 26794 }, { "epoch": 2.77, "learning_rate": 1.240230560235454e-05, "loss": 0.3901, "step": 26795 }, { "epoch": 2.77, "learning_rate": 1.2400312196296003e-05, "loss": 0.8083, "step": 26796 }, { "epoch": 2.77, "learning_rate": 1.2398318908714446e-05, "loss": 0.4735, "step": 26797 }, { "epoch": 2.77, "learning_rate": 1.2396325739623285e-05, "loss": 0.5236, "step": 26798 }, { "epoch": 2.77, "learning_rate": 1.2394332689035944e-05, "loss": 0.5075, "step": 26799 }, { "epoch": 2.77, "learning_rate": 1.2392339756965845e-05, "loss": 0.1601, "step": 26800 }, { "epoch": 2.77, "learning_rate": 1.2390346943426382e-05, "loss": 0.2473, "step": 26801 }, { "epoch": 2.77, "learning_rate": 1.2388354248430988e-05, "loss": 0.3588, "step": 26802 }, { "epoch": 2.77, "learning_rate": 1.2386361671993069e-05, "loss": 0.7581, "step": 26803 }, { "epoch": 2.77, "learning_rate": 1.2384369214126044e-05, "loss": 0.1701, "step": 26804 }, { "epoch": 2.77, "learning_rate": 1.2382376874843332e-05, "loss": 0.3079, "step": 26805 }, { "epoch": 2.77, "learning_rate": 1.2380384654158328e-05, "loss": 0.4239, "step": 26806 }, { "epoch": 2.77, "learning_rate": 1.2378392552084452e-05, "loss": 0.5546, "step": 26807 }, { "epoch": 2.77, "learning_rate": 1.237640056863511e-05, "loss": 0.1068, "step": 26808 }, { "epoch": 2.77, "learning_rate": 1.2374408703823721e-05, "loss": 0.2808, "step": 26809 }, { "epoch": 2.77, "learning_rate": 1.2372416957663686e-05, "loss": 0.638, "step": 26810 }, { "epoch": 2.77, "learning_rate": 1.2370425330168416e-05, "loss": 0.6605, "step": 26811 }, { "epoch": 2.77, "learning_rate": 1.2368433821351329e-05, "loss": 0.2447, "step": 26812 }, { "epoch": 2.77, "learning_rate": 1.2366442431225809e-05, "loss": 0.3008, "step": 26813 }, { "epoch": 2.77, "learning_rate": 1.2364451159805272e-05, "loss": 0.2827, "step": 26814 }, { "epoch": 2.77, "learning_rate": 1.2362460007103133e-05, "loss": 0.2095, "step": 26815 }, { "epoch": 2.77, "learning_rate": 1.2360468973132779e-05, "loss": 0.2916, "step": 26816 }, { "epoch": 2.77, "learning_rate": 1.2358478057907618e-05, "loss": 0.5048, "step": 26817 }, { "epoch": 2.77, "learning_rate": 1.2356487261441058e-05, "loss": 0.0293, "step": 26818 }, { "epoch": 2.77, "learning_rate": 1.2354496583746492e-05, "loss": 0.2511, "step": 26819 }, { "epoch": 2.77, "learning_rate": 1.235250602483733e-05, "loss": 0.633, "step": 26820 }, { "epoch": 2.77, "learning_rate": 1.2350515584726968e-05, "loss": 0.6635, "step": 26821 }, { "epoch": 2.77, "learning_rate": 1.2348525263428806e-05, "loss": 0.3434, "step": 26822 }, { "epoch": 2.77, "learning_rate": 1.2346535060956252e-05, "loss": 0.7302, "step": 26823 }, { "epoch": 2.77, "learning_rate": 1.2344544977322683e-05, "loss": 0.5455, "step": 26824 }, { "epoch": 2.77, "learning_rate": 1.2342555012541515e-05, "loss": 0.0816, "step": 26825 }, { "epoch": 2.78, "learning_rate": 1.2340565166626125e-05, "loss": 0.1049, "step": 26826 }, { "epoch": 2.78, "learning_rate": 1.2338575439589919e-05, "loss": 0.3215, "step": 26827 }, { "epoch": 2.78, "learning_rate": 1.2336585831446291e-05, "loss": 0.1962, "step": 26828 }, { "epoch": 2.78, "learning_rate": 1.2334596342208635e-05, "loss": 0.7545, "step": 26829 }, { "epoch": 2.78, "learning_rate": 1.233260697189034e-05, "loss": 0.4569, "step": 26830 }, { "epoch": 2.78, "learning_rate": 1.2330617720504805e-05, "loss": 0.4931, "step": 26831 }, { "epoch": 2.78, "learning_rate": 1.2328628588065413e-05, "loss": 0.2235, "step": 26832 }, { "epoch": 2.78, "learning_rate": 1.2326639574585574e-05, "loss": 0.1647, "step": 26833 }, { "epoch": 2.78, "learning_rate": 1.2324650680078646e-05, "loss": 0.3287, "step": 26834 }, { "epoch": 2.78, "learning_rate": 1.232266190455804e-05, "loss": 0.3394, "step": 26835 }, { "epoch": 2.78, "learning_rate": 1.2320673248037132e-05, "loss": 0.2246, "step": 26836 }, { "epoch": 2.78, "learning_rate": 1.2318684710529331e-05, "loss": 0.0841, "step": 26837 }, { "epoch": 2.78, "learning_rate": 1.2316696292047993e-05, "loss": 0.3461, "step": 26838 }, { "epoch": 2.78, "learning_rate": 1.2314707992606522e-05, "loss": 0.3076, "step": 26839 }, { "epoch": 2.78, "learning_rate": 1.2312719812218294e-05, "loss": 0.4954, "step": 26840 }, { "epoch": 2.78, "learning_rate": 1.23107317508967e-05, "loss": 0.2864, "step": 26841 }, { "epoch": 2.78, "learning_rate": 1.2308743808655118e-05, "loss": 0.1385, "step": 26842 }, { "epoch": 2.78, "learning_rate": 1.230675598550695e-05, "loss": 0.2344, "step": 26843 }, { "epoch": 2.78, "learning_rate": 1.2304768281465544e-05, "loss": 0.4469, "step": 26844 }, { "epoch": 2.78, "learning_rate": 1.2302780696544302e-05, "loss": 0.1779, "step": 26845 }, { "epoch": 2.78, "learning_rate": 1.2300793230756598e-05, "loss": 0.3187, "step": 26846 }, { "epoch": 2.78, "learning_rate": 1.2298805884115811e-05, "loss": 0.2837, "step": 26847 }, { "epoch": 2.78, "learning_rate": 1.2296818656635324e-05, "loss": 0.3117, "step": 26848 }, { "epoch": 2.78, "learning_rate": 1.229483154832851e-05, "loss": 0.1955, "step": 26849 }, { "epoch": 2.78, "learning_rate": 1.2292844559208757e-05, "loss": 0.3562, "step": 26850 }, { "epoch": 2.78, "learning_rate": 1.2290857689289422e-05, "loss": 0.1323, "step": 26851 }, { "epoch": 2.78, "learning_rate": 1.2288870938583888e-05, "loss": 0.0764, "step": 26852 }, { "epoch": 2.78, "learning_rate": 1.2286884307105545e-05, "loss": 0.1801, "step": 26853 }, { "epoch": 2.78, "learning_rate": 1.2284897794867739e-05, "loss": 0.283, "step": 26854 }, { "epoch": 2.78, "learning_rate": 1.2282911401883856e-05, "loss": 0.1527, "step": 26855 }, { "epoch": 2.78, "learning_rate": 1.228092512816727e-05, "loss": 0.2169, "step": 26856 }, { "epoch": 2.78, "learning_rate": 1.2278938973731349e-05, "loss": 0.4207, "step": 26857 }, { "epoch": 2.78, "learning_rate": 1.2276952938589462e-05, "loss": 0.6918, "step": 26858 }, { "epoch": 2.78, "learning_rate": 1.2274967022754983e-05, "loss": 0.5309, "step": 26859 }, { "epoch": 2.78, "learning_rate": 1.2272981226241276e-05, "loss": 0.642, "step": 26860 }, { "epoch": 2.78, "learning_rate": 1.2270995549061726e-05, "loss": 0.503, "step": 26861 }, { "epoch": 2.78, "learning_rate": 1.2269009991229675e-05, "loss": 0.0959, "step": 26862 }, { "epoch": 2.78, "learning_rate": 1.2267024552758508e-05, "loss": 0.1729, "step": 26863 }, { "epoch": 2.78, "learning_rate": 1.2265039233661574e-05, "loss": 0.1474, "step": 26864 }, { "epoch": 2.78, "learning_rate": 1.2263054033952244e-05, "loss": 0.1007, "step": 26865 }, { "epoch": 2.78, "learning_rate": 1.2261068953643886e-05, "loss": 0.3343, "step": 26866 }, { "epoch": 2.78, "learning_rate": 1.2259083992749861e-05, "loss": 0.1204, "step": 26867 }, { "epoch": 2.78, "learning_rate": 1.2257099151283532e-05, "loss": 0.2793, "step": 26868 }, { "epoch": 2.78, "learning_rate": 1.2255114429258258e-05, "loss": 0.1691, "step": 26869 }, { "epoch": 2.78, "learning_rate": 1.2253129826687403e-05, "loss": 0.1652, "step": 26870 }, { "epoch": 2.78, "learning_rate": 1.2251145343584338e-05, "loss": 0.5681, "step": 26871 }, { "epoch": 2.78, "learning_rate": 1.2249160979962397e-05, "loss": 0.368, "step": 26872 }, { "epoch": 2.78, "learning_rate": 1.2247176735834951e-05, "loss": 0.1112, "step": 26873 }, { "epoch": 2.78, "learning_rate": 1.2245192611215359e-05, "loss": 0.3321, "step": 26874 }, { "epoch": 2.78, "learning_rate": 1.2243208606116984e-05, "loss": 0.1837, "step": 26875 }, { "epoch": 2.78, "learning_rate": 1.2241224720553167e-05, "loss": 0.6074, "step": 26876 }, { "epoch": 2.78, "learning_rate": 1.2239240954537267e-05, "loss": 0.6001, "step": 26877 }, { "epoch": 2.78, "learning_rate": 1.2237257308082642e-05, "loss": 0.4304, "step": 26878 }, { "epoch": 2.78, "learning_rate": 1.2235273781202643e-05, "loss": 0.2717, "step": 26879 }, { "epoch": 2.78, "learning_rate": 1.2233290373910626e-05, "loss": 0.2363, "step": 26880 }, { "epoch": 2.78, "learning_rate": 1.2231307086219951e-05, "loss": 0.2385, "step": 26881 }, { "epoch": 2.78, "learning_rate": 1.222932391814395e-05, "loss": 0.4468, "step": 26882 }, { "epoch": 2.78, "learning_rate": 1.2227340869695981e-05, "loss": 0.2812, "step": 26883 }, { "epoch": 2.78, "learning_rate": 1.2225357940889394e-05, "loss": 0.0898, "step": 26884 }, { "epoch": 2.78, "learning_rate": 1.2223375131737539e-05, "loss": 0.095, "step": 26885 }, { "epoch": 2.78, "learning_rate": 1.2221392442253763e-05, "loss": 0.3672, "step": 26886 }, { "epoch": 2.78, "learning_rate": 1.2219409872451417e-05, "loss": 0.1697, "step": 26887 }, { "epoch": 2.78, "learning_rate": 1.2217427422343851e-05, "loss": 0.3179, "step": 26888 }, { "epoch": 2.78, "learning_rate": 1.2215445091944391e-05, "loss": 0.4277, "step": 26889 }, { "epoch": 2.78, "learning_rate": 1.2213462881266394e-05, "loss": 0.6913, "step": 26890 }, { "epoch": 2.78, "learning_rate": 1.2211480790323216e-05, "loss": 1.634, "step": 26891 }, { "epoch": 2.78, "learning_rate": 1.2209498819128174e-05, "loss": 0.2953, "step": 26892 }, { "epoch": 2.78, "learning_rate": 1.2207516967694624e-05, "loss": 0.4729, "step": 26893 }, { "epoch": 2.78, "learning_rate": 1.2205535236035908e-05, "loss": 0.8277, "step": 26894 }, { "epoch": 2.78, "learning_rate": 1.2203553624165363e-05, "loss": 0.1647, "step": 26895 }, { "epoch": 2.78, "learning_rate": 1.2201572132096332e-05, "loss": 0.2972, "step": 26896 }, { "epoch": 2.78, "learning_rate": 1.2199590759842154e-05, "loss": 0.9592, "step": 26897 }, { "epoch": 2.78, "learning_rate": 1.2197609507416164e-05, "loss": 0.9054, "step": 26898 }, { "epoch": 2.78, "learning_rate": 1.2195628374831714e-05, "loss": 0.2992, "step": 26899 }, { "epoch": 2.78, "learning_rate": 1.2193647362102118e-05, "loss": 0.3493, "step": 26900 }, { "epoch": 2.78, "learning_rate": 1.219166646924073e-05, "loss": 0.4848, "step": 26901 }, { "epoch": 2.78, "learning_rate": 1.2189685696260864e-05, "loss": 0.3317, "step": 26902 }, { "epoch": 2.78, "learning_rate": 1.2187705043175869e-05, "loss": 0.2995, "step": 26903 }, { "epoch": 2.78, "learning_rate": 1.2185724509999076e-05, "loss": 0.968, "step": 26904 }, { "epoch": 2.78, "learning_rate": 1.218374409674382e-05, "loss": 0.1441, "step": 26905 }, { "epoch": 2.78, "learning_rate": 1.2181763803423427e-05, "loss": 0.5219, "step": 26906 }, { "epoch": 2.78, "learning_rate": 1.217978363005123e-05, "loss": 0.0236, "step": 26907 }, { "epoch": 2.78, "learning_rate": 1.2177803576640561e-05, "loss": 0.718, "step": 26908 }, { "epoch": 2.78, "learning_rate": 1.2175823643204757e-05, "loss": 0.2676, "step": 26909 }, { "epoch": 2.78, "learning_rate": 1.217384382975713e-05, "loss": 0.8796, "step": 26910 }, { "epoch": 2.78, "learning_rate": 1.2171864136311016e-05, "loss": 0.3676, "step": 26911 }, { "epoch": 2.78, "learning_rate": 1.216988456287974e-05, "loss": 0.5449, "step": 26912 }, { "epoch": 2.78, "learning_rate": 1.216790510947664e-05, "loss": 0.2008, "step": 26913 }, { "epoch": 2.78, "learning_rate": 1.2165925776115017e-05, "loss": 0.7946, "step": 26914 }, { "epoch": 2.78, "learning_rate": 1.2163946562808214e-05, "loss": 0.4246, "step": 26915 }, { "epoch": 2.78, "learning_rate": 1.2161967469569548e-05, "loss": 0.3468, "step": 26916 }, { "epoch": 2.78, "learning_rate": 1.215998849641234e-05, "loss": 0.6994, "step": 26917 }, { "epoch": 2.78, "learning_rate": 1.2158009643349915e-05, "loss": 0.2826, "step": 26918 }, { "epoch": 2.78, "learning_rate": 1.215603091039561e-05, "loss": 0.2216, "step": 26919 }, { "epoch": 2.78, "learning_rate": 1.2154052297562715e-05, "loss": 0.2243, "step": 26920 }, { "epoch": 2.78, "learning_rate": 1.2152073804864567e-05, "loss": 0.565, "step": 26921 }, { "epoch": 2.79, "learning_rate": 1.2150095432314478e-05, "loss": 0.2097, "step": 26922 }, { "epoch": 2.79, "learning_rate": 1.2148117179925774e-05, "loss": 0.2029, "step": 26923 }, { "epoch": 2.79, "learning_rate": 1.2146139047711766e-05, "loss": 0.5282, "step": 26924 }, { "epoch": 2.79, "learning_rate": 1.2144161035685784e-05, "loss": 0.2563, "step": 26925 }, { "epoch": 2.79, "learning_rate": 1.2142183143861122e-05, "loss": 0.0853, "step": 26926 }, { "epoch": 2.79, "learning_rate": 1.2140205372251104e-05, "loss": 0.5324, "step": 26927 }, { "epoch": 2.79, "learning_rate": 1.2138227720869044e-05, "loss": 0.4824, "step": 26928 }, { "epoch": 2.79, "learning_rate": 1.2136250189728268e-05, "loss": 0.3687, "step": 26929 }, { "epoch": 2.79, "learning_rate": 1.2134272778842064e-05, "loss": 0.4516, "step": 26930 }, { "epoch": 2.79, "learning_rate": 1.2132295488223755e-05, "loss": 0.1728, "step": 26931 }, { "epoch": 2.79, "learning_rate": 1.2130318317886653e-05, "loss": 0.327, "step": 26932 }, { "epoch": 2.79, "learning_rate": 1.2128341267844067e-05, "loss": 0.1137, "step": 26933 }, { "epoch": 2.79, "learning_rate": 1.2126364338109306e-05, "loss": 0.1248, "step": 26934 }, { "epoch": 2.79, "learning_rate": 1.2124387528695677e-05, "loss": 0.3268, "step": 26935 }, { "epoch": 2.79, "learning_rate": 1.2122410839616488e-05, "loss": 0.7282, "step": 26936 }, { "epoch": 2.79, "learning_rate": 1.212043427088506e-05, "loss": 0.4996, "step": 26937 }, { "epoch": 2.79, "learning_rate": 1.211845782251467e-05, "loss": 0.5756, "step": 26938 }, { "epoch": 2.79, "learning_rate": 1.2116481494518651e-05, "loss": 0.255, "step": 26939 }, { "epoch": 2.79, "learning_rate": 1.2114505286910284e-05, "loss": 0.4043, "step": 26940 }, { "epoch": 2.79, "learning_rate": 1.2112529199702883e-05, "loss": 0.1965, "step": 26941 }, { "epoch": 2.79, "learning_rate": 1.211055323290975e-05, "loss": 0.3304, "step": 26942 }, { "epoch": 2.79, "learning_rate": 1.2108577386544185e-05, "loss": 0.4644, "step": 26943 }, { "epoch": 2.79, "learning_rate": 1.2106601660619491e-05, "loss": 0.3619, "step": 26944 }, { "epoch": 2.79, "learning_rate": 1.2104626055148969e-05, "loss": 0.5473, "step": 26945 }, { "epoch": 2.79, "learning_rate": 1.2102650570145919e-05, "loss": 0.2238, "step": 26946 }, { "epoch": 2.79, "learning_rate": 1.2100675205623646e-05, "loss": 0.0667, "step": 26947 }, { "epoch": 2.79, "learning_rate": 1.2098699961595428e-05, "loss": 0.4668, "step": 26948 }, { "epoch": 2.79, "learning_rate": 1.2096724838074576e-05, "loss": 0.1299, "step": 26949 }, { "epoch": 2.79, "learning_rate": 1.2094749835074392e-05, "loss": 0.3701, "step": 26950 }, { "epoch": 2.79, "learning_rate": 1.2092774952608151e-05, "loss": 0.3759, "step": 26951 }, { "epoch": 2.79, "learning_rate": 1.209080019068916e-05, "loss": 0.234, "step": 26952 }, { "epoch": 2.79, "learning_rate": 1.2088825549330714e-05, "loss": 0.2228, "step": 26953 }, { "epoch": 2.79, "learning_rate": 1.2086851028546098e-05, "loss": 0.2247, "step": 26954 }, { "epoch": 2.79, "learning_rate": 1.2084876628348616e-05, "loss": 0.1195, "step": 26955 }, { "epoch": 2.79, "learning_rate": 1.2082902348751548e-05, "loss": 0.3105, "step": 26956 }, { "epoch": 2.79, "learning_rate": 1.2080928189768204e-05, "loss": 0.1289, "step": 26957 }, { "epoch": 2.79, "learning_rate": 1.2078954151411847e-05, "loss": 0.3764, "step": 26958 }, { "epoch": 2.79, "learning_rate": 1.2076980233695772e-05, "loss": 0.3553, "step": 26959 }, { "epoch": 2.79, "learning_rate": 1.2075006436633279e-05, "loss": 0.3412, "step": 26960 }, { "epoch": 2.79, "learning_rate": 1.2073032760237646e-05, "loss": 0.3859, "step": 26961 }, { "epoch": 2.79, "learning_rate": 1.2071059204522161e-05, "loss": 0.5703, "step": 26962 }, { "epoch": 2.79, "learning_rate": 1.2069085769500122e-05, "loss": 0.3234, "step": 26963 }, { "epoch": 2.79, "learning_rate": 1.206711245518479e-05, "loss": 0.1447, "step": 26964 }, { "epoch": 2.79, "learning_rate": 1.2065139261589464e-05, "loss": 0.239, "step": 26965 }, { "epoch": 2.79, "learning_rate": 1.2063166188727423e-05, "loss": 0.441, "step": 26966 }, { "epoch": 2.79, "learning_rate": 1.2061193236611959e-05, "loss": 0.268, "step": 26967 }, { "epoch": 2.79, "learning_rate": 1.2059220405256337e-05, "loss": 0.6477, "step": 26968 }, { "epoch": 2.79, "learning_rate": 1.2057247694673847e-05, "loss": 0.2845, "step": 26969 }, { "epoch": 2.79, "learning_rate": 1.2055275104877762e-05, "loss": 0.2576, "step": 26970 }, { "epoch": 2.79, "learning_rate": 1.205330263588137e-05, "loss": 0.3793, "step": 26971 }, { "epoch": 2.79, "learning_rate": 1.2051330287697944e-05, "loss": 0.7847, "step": 26972 }, { "epoch": 2.79, "learning_rate": 1.2049358060340762e-05, "loss": 0.3998, "step": 26973 }, { "epoch": 2.79, "learning_rate": 1.2047385953823104e-05, "loss": 0.7341, "step": 26974 }, { "epoch": 2.79, "learning_rate": 1.2045413968158253e-05, "loss": 0.3241, "step": 26975 }, { "epoch": 2.79, "learning_rate": 1.2043442103359462e-05, "loss": 0.2866, "step": 26976 }, { "epoch": 2.79, "learning_rate": 1.2041470359440026e-05, "loss": 0.673, "step": 26977 }, { "epoch": 2.79, "learning_rate": 1.20394987364132e-05, "loss": 0.399, "step": 26978 }, { "epoch": 2.79, "learning_rate": 1.2037527234292268e-05, "loss": 0.1912, "step": 26979 }, { "epoch": 2.79, "learning_rate": 1.2035555853090499e-05, "loss": 0.3182, "step": 26980 }, { "epoch": 2.79, "learning_rate": 1.2033584592821169e-05, "loss": 0.3651, "step": 26981 }, { "epoch": 2.79, "learning_rate": 1.2031613453497538e-05, "loss": 0.1526, "step": 26982 }, { "epoch": 2.79, "learning_rate": 1.2029642435132882e-05, "loss": 0.0914, "step": 26983 }, { "epoch": 2.79, "learning_rate": 1.202767153774047e-05, "loss": 0.6087, "step": 26984 }, { "epoch": 2.79, "learning_rate": 1.202570076133358e-05, "loss": 0.1659, "step": 26985 }, { "epoch": 2.79, "learning_rate": 1.2023730105925453e-05, "loss": 0.246, "step": 26986 }, { "epoch": 2.79, "learning_rate": 1.2021759571529369e-05, "loss": 0.602, "step": 26987 }, { "epoch": 2.79, "learning_rate": 1.2019789158158603e-05, "loss": 0.2782, "step": 26988 }, { "epoch": 2.79, "learning_rate": 1.2017818865826402e-05, "loss": 0.3464, "step": 26989 }, { "epoch": 2.79, "learning_rate": 1.2015848694546033e-05, "loss": 0.275, "step": 26990 }, { "epoch": 2.79, "learning_rate": 1.2013878644330765e-05, "loss": 0.3257, "step": 26991 }, { "epoch": 2.79, "learning_rate": 1.201190871519386e-05, "loss": 0.2997, "step": 26992 }, { "epoch": 2.79, "learning_rate": 1.2009938907148572e-05, "loss": 0.1433, "step": 26993 }, { "epoch": 2.79, "learning_rate": 1.2007969220208167e-05, "loss": 0.2886, "step": 26994 }, { "epoch": 2.79, "learning_rate": 1.2005999654385916e-05, "loss": 0.1909, "step": 26995 }, { "epoch": 2.79, "learning_rate": 1.2004030209695052e-05, "loss": 0.4321, "step": 26996 }, { "epoch": 2.79, "learning_rate": 1.2002060886148846e-05, "loss": 0.3095, "step": 26997 }, { "epoch": 2.79, "learning_rate": 1.2000091683760558e-05, "loss": 0.0494, "step": 26998 }, { "epoch": 2.79, "learning_rate": 1.199812260254344e-05, "loss": 0.2525, "step": 26999 }, { "epoch": 2.79, "learning_rate": 1.1996153642510745e-05, "loss": 0.5017, "step": 27000 }, { "epoch": 2.79, "learning_rate": 1.1994184803675747e-05, "loss": 0.4081, "step": 27001 }, { "epoch": 2.79, "learning_rate": 1.199221608605167e-05, "loss": 0.3499, "step": 27002 }, { "epoch": 2.79, "learning_rate": 1.199024748965178e-05, "loss": 0.0987, "step": 27003 }, { "epoch": 2.79, "learning_rate": 1.1988279014489335e-05, "loss": 0.1248, "step": 27004 }, { "epoch": 2.79, "learning_rate": 1.1986310660577588e-05, "loss": 0.3735, "step": 27005 }, { "epoch": 2.79, "learning_rate": 1.198434242792977e-05, "loss": 0.4761, "step": 27006 }, { "epoch": 2.79, "learning_rate": 1.1982374316559148e-05, "loss": 0.1181, "step": 27007 }, { "epoch": 2.79, "learning_rate": 1.1980406326478962e-05, "loss": 0.1431, "step": 27008 }, { "epoch": 2.79, "learning_rate": 1.1978438457702469e-05, "loss": 0.7816, "step": 27009 }, { "epoch": 2.79, "learning_rate": 1.197647071024291e-05, "loss": 0.1938, "step": 27010 }, { "epoch": 2.79, "learning_rate": 1.1974503084113531e-05, "loss": 0.4463, "step": 27011 }, { "epoch": 2.79, "learning_rate": 1.1972535579327581e-05, "loss": 0.2759, "step": 27012 }, { "epoch": 2.79, "learning_rate": 1.1970568195898313e-05, "loss": 0.5106, "step": 27013 }, { "epoch": 2.79, "learning_rate": 1.1968600933838955e-05, "loss": 0.2526, "step": 27014 }, { "epoch": 2.79, "learning_rate": 1.1966633793162763e-05, "loss": 0.5265, "step": 27015 }, { "epoch": 2.79, "learning_rate": 1.1964666773882961e-05, "loss": 0.1506, "step": 27016 }, { "epoch": 2.79, "learning_rate": 1.1962699876012806e-05, "loss": 0.0683, "step": 27017 }, { "epoch": 2.8, "learning_rate": 1.1960733099565532e-05, "loss": 0.313, "step": 27018 }, { "epoch": 2.8, "learning_rate": 1.1958766444554385e-05, "loss": 0.6637, "step": 27019 }, { "epoch": 2.8, "learning_rate": 1.19567999109926e-05, "loss": 0.6431, "step": 27020 }, { "epoch": 2.8, "learning_rate": 1.1954833498893414e-05, "loss": 0.5549, "step": 27021 }, { "epoch": 2.8, "learning_rate": 1.1952867208270067e-05, "loss": 0.5437, "step": 27022 }, { "epoch": 2.8, "learning_rate": 1.1950901039135806e-05, "loss": 0.1175, "step": 27023 }, { "epoch": 2.8, "learning_rate": 1.1948934991503847e-05, "loss": 0.7222, "step": 27024 }, { "epoch": 2.8, "learning_rate": 1.1946969065387435e-05, "loss": 0.4178, "step": 27025 }, { "epoch": 2.8, "learning_rate": 1.1945003260799815e-05, "loss": 0.6, "step": 27026 }, { "epoch": 2.8, "learning_rate": 1.1943037577754192e-05, "loss": 0.4432, "step": 27027 }, { "epoch": 2.8, "learning_rate": 1.194107201626382e-05, "loss": 0.7198, "step": 27028 }, { "epoch": 2.8, "learning_rate": 1.1939106576341923e-05, "loss": 0.4889, "step": 27029 }, { "epoch": 2.8, "learning_rate": 1.1937141258001736e-05, "loss": 0.1418, "step": 27030 }, { "epoch": 2.8, "learning_rate": 1.1935176061256486e-05, "loss": 0.4162, "step": 27031 }, { "epoch": 2.8, "learning_rate": 1.1933210986119406e-05, "loss": 0.5471, "step": 27032 }, { "epoch": 2.8, "learning_rate": 1.1931246032603732e-05, "loss": 0.4876, "step": 27033 }, { "epoch": 2.8, "learning_rate": 1.1929281200722672e-05, "loss": 0.2142, "step": 27034 }, { "epoch": 2.8, "learning_rate": 1.1927316490489462e-05, "loss": 0.5216, "step": 27035 }, { "epoch": 2.8, "learning_rate": 1.1925351901917328e-05, "loss": 0.3677, "step": 27036 }, { "epoch": 2.8, "learning_rate": 1.1923387435019499e-05, "loss": 0.3149, "step": 27037 }, { "epoch": 2.8, "learning_rate": 1.1921423089809203e-05, "loss": 0.2089, "step": 27038 }, { "epoch": 2.8, "learning_rate": 1.1919458866299647e-05, "loss": 0.5303, "step": 27039 }, { "epoch": 2.8, "learning_rate": 1.1917494764504067e-05, "loss": 0.7113, "step": 27040 }, { "epoch": 2.8, "learning_rate": 1.1915530784435678e-05, "loss": 0.5099, "step": 27041 }, { "epoch": 2.8, "learning_rate": 1.1913566926107702e-05, "loss": 0.2445, "step": 27042 }, { "epoch": 2.8, "learning_rate": 1.1911603189533376e-05, "loss": 0.5226, "step": 27043 }, { "epoch": 2.8, "learning_rate": 1.1909639574725895e-05, "loss": 0.43, "step": 27044 }, { "epoch": 2.8, "learning_rate": 1.1907676081698487e-05, "loss": 0.0465, "step": 27045 }, { "epoch": 2.8, "learning_rate": 1.1905712710464366e-05, "loss": 0.4642, "step": 27046 }, { "epoch": 2.8, "learning_rate": 1.1903749461036758e-05, "loss": 0.2101, "step": 27047 }, { "epoch": 2.8, "learning_rate": 1.1901786333428872e-05, "loss": 0.4996, "step": 27048 }, { "epoch": 2.8, "learning_rate": 1.1899823327653928e-05, "loss": 0.4051, "step": 27049 }, { "epoch": 2.8, "learning_rate": 1.1897860443725135e-05, "loss": 0.7687, "step": 27050 }, { "epoch": 2.8, "learning_rate": 1.1895897681655725e-05, "loss": 0.7047, "step": 27051 }, { "epoch": 2.8, "learning_rate": 1.1893935041458884e-05, "loss": 0.1735, "step": 27052 }, { "epoch": 2.8, "learning_rate": 1.1891972523147842e-05, "loss": 0.1325, "step": 27053 }, { "epoch": 2.8, "learning_rate": 1.1890010126735796e-05, "loss": 0.2184, "step": 27054 }, { "epoch": 2.8, "learning_rate": 1.1888047852235966e-05, "loss": 0.6763, "step": 27055 }, { "epoch": 2.8, "learning_rate": 1.1886085699661561e-05, "loss": 0.1526, "step": 27056 }, { "epoch": 2.8, "learning_rate": 1.1884123669025787e-05, "loss": 0.0849, "step": 27057 }, { "epoch": 2.8, "learning_rate": 1.1882161760341852e-05, "loss": 0.2647, "step": 27058 }, { "epoch": 2.8, "learning_rate": 1.1880199973622966e-05, "loss": 0.2221, "step": 27059 }, { "epoch": 2.8, "learning_rate": 1.1878238308882336e-05, "loss": 0.4007, "step": 27060 }, { "epoch": 2.8, "learning_rate": 1.1876276766133172e-05, "loss": 0.2297, "step": 27061 }, { "epoch": 2.8, "learning_rate": 1.1874315345388667e-05, "loss": 0.1895, "step": 27062 }, { "epoch": 2.8, "learning_rate": 1.1872354046662036e-05, "loss": 0.3559, "step": 27063 }, { "epoch": 2.8, "learning_rate": 1.1870392869966468e-05, "loss": 0.1274, "step": 27064 }, { "epoch": 2.8, "learning_rate": 1.1868431815315171e-05, "loss": 0.3425, "step": 27065 }, { "epoch": 2.8, "learning_rate": 1.1866470882721351e-05, "loss": 0.1168, "step": 27066 }, { "epoch": 2.8, "learning_rate": 1.1864510072198205e-05, "loss": 0.3909, "step": 27067 }, { "epoch": 2.8, "learning_rate": 1.1862549383758935e-05, "loss": 0.0652, "step": 27068 }, { "epoch": 2.8, "learning_rate": 1.1860588817416735e-05, "loss": 0.1545, "step": 27069 }, { "epoch": 2.8, "learning_rate": 1.1858628373184809e-05, "loss": 0.6229, "step": 27070 }, { "epoch": 2.8, "learning_rate": 1.1856668051076365e-05, "loss": 0.1083, "step": 27071 }, { "epoch": 2.8, "learning_rate": 1.1854707851104575e-05, "loss": 0.3648, "step": 27072 }, { "epoch": 2.8, "learning_rate": 1.1852747773282642e-05, "loss": 0.1258, "step": 27073 }, { "epoch": 2.8, "learning_rate": 1.1850787817623768e-05, "loss": 0.4489, "step": 27074 }, { "epoch": 2.8, "learning_rate": 1.1848827984141143e-05, "loss": 0.1844, "step": 27075 }, { "epoch": 2.8, "learning_rate": 1.1846868272847975e-05, "loss": 0.212, "step": 27076 }, { "epoch": 2.8, "learning_rate": 1.1844908683757427e-05, "loss": 0.3499, "step": 27077 }, { "epoch": 2.8, "learning_rate": 1.184294921688271e-05, "loss": 0.4771, "step": 27078 }, { "epoch": 2.8, "learning_rate": 1.1840989872237006e-05, "loss": 0.5831, "step": 27079 }, { "epoch": 2.8, "learning_rate": 1.1839030649833513e-05, "loss": 0.5957, "step": 27080 }, { "epoch": 2.8, "learning_rate": 1.1837071549685426e-05, "loss": 0.7529, "step": 27081 }, { "epoch": 2.8, "learning_rate": 1.1835112571805912e-05, "loss": 0.6445, "step": 27082 }, { "epoch": 2.8, "learning_rate": 1.1833153716208168e-05, "loss": 0.1891, "step": 27083 }, { "epoch": 2.8, "learning_rate": 1.1831194982905389e-05, "loss": 0.2162, "step": 27084 }, { "epoch": 2.8, "learning_rate": 1.1829236371910748e-05, "loss": 0.3603, "step": 27085 }, { "epoch": 2.8, "learning_rate": 1.1827277883237437e-05, "loss": 0.1362, "step": 27086 }, { "epoch": 2.8, "learning_rate": 1.1825319516898643e-05, "loss": 0.1675, "step": 27087 }, { "epoch": 2.8, "learning_rate": 1.182336127290754e-05, "loss": 0.1228, "step": 27088 }, { "epoch": 2.8, "learning_rate": 1.182140315127733e-05, "loss": 0.1033, "step": 27089 }, { "epoch": 2.8, "learning_rate": 1.181944515202117e-05, "loss": 0.2352, "step": 27090 }, { "epoch": 2.8, "learning_rate": 1.1817487275152262e-05, "loss": 0.498, "step": 27091 }, { "epoch": 2.8, "learning_rate": 1.1815529520683762e-05, "loss": 0.1981, "step": 27092 }, { "epoch": 2.8, "learning_rate": 1.1813571888628864e-05, "loss": 0.2739, "step": 27093 }, { "epoch": 2.8, "learning_rate": 1.1811614379000746e-05, "loss": 0.3406, "step": 27094 }, { "epoch": 2.8, "learning_rate": 1.1809656991812585e-05, "loss": 0.7205, "step": 27095 }, { "epoch": 2.8, "learning_rate": 1.1807699727077555e-05, "loss": 0.6448, "step": 27096 }, { "epoch": 2.8, "learning_rate": 1.1805742584808832e-05, "loss": 0.6315, "step": 27097 }, { "epoch": 2.8, "learning_rate": 1.1803785565019597e-05, "loss": 0.0991, "step": 27098 }, { "epoch": 2.8, "learning_rate": 1.1801828667723027e-05, "loss": 0.3313, "step": 27099 }, { "epoch": 2.8, "learning_rate": 1.179987189293228e-05, "loss": 0.0806, "step": 27100 }, { "epoch": 2.8, "learning_rate": 1.1797915240660546e-05, "loss": 0.1896, "step": 27101 }, { "epoch": 2.8, "learning_rate": 1.1795958710920977e-05, "loss": 0.292, "step": 27102 }, { "epoch": 2.8, "learning_rate": 1.1794002303726756e-05, "loss": 0.3559, "step": 27103 }, { "epoch": 2.8, "learning_rate": 1.1792046019091051e-05, "loss": 0.3198, "step": 27104 }, { "epoch": 2.8, "learning_rate": 1.1790089857027034e-05, "loss": 0.1196, "step": 27105 }, { "epoch": 2.8, "learning_rate": 1.1788133817547868e-05, "loss": 0.4495, "step": 27106 }, { "epoch": 2.8, "learning_rate": 1.1786177900666726e-05, "loss": 0.1164, "step": 27107 }, { "epoch": 2.8, "learning_rate": 1.1784222106396769e-05, "loss": 0.1788, "step": 27108 }, { "epoch": 2.8, "learning_rate": 1.1782266434751183e-05, "loss": 0.3897, "step": 27109 }, { "epoch": 2.8, "learning_rate": 1.17803108857431e-05, "loss": 0.1102, "step": 27110 }, { "epoch": 2.8, "learning_rate": 1.1778355459385704e-05, "loss": 0.2929, "step": 27111 }, { "epoch": 2.8, "learning_rate": 1.1776400155692157e-05, "loss": 0.1723, "step": 27112 }, { "epoch": 2.81, "learning_rate": 1.1774444974675619e-05, "loss": 1.635, "step": 27113 }, { "epoch": 2.81, "learning_rate": 1.1772489916349265e-05, "loss": 0.4943, "step": 27114 }, { "epoch": 2.81, "learning_rate": 1.1770534980726232e-05, "loss": 0.1614, "step": 27115 }, { "epoch": 2.81, "learning_rate": 1.1768580167819693e-05, "loss": 0.3061, "step": 27116 }, { "epoch": 2.81, "learning_rate": 1.1766625477642802e-05, "loss": 0.4291, "step": 27117 }, { "epoch": 2.81, "learning_rate": 1.1764670910208726e-05, "loss": 0.3524, "step": 27118 }, { "epoch": 2.81, "learning_rate": 1.1762716465530631e-05, "loss": 0.2924, "step": 27119 }, { "epoch": 2.81, "learning_rate": 1.1760762143621646e-05, "loss": 0.2115, "step": 27120 }, { "epoch": 2.81, "learning_rate": 1.1758807944494948e-05, "loss": 0.5906, "step": 27121 }, { "epoch": 2.81, "learning_rate": 1.1756853868163682e-05, "loss": 0.4708, "step": 27122 }, { "epoch": 2.81, "learning_rate": 1.175489991464101e-05, "loss": 0.1083, "step": 27123 }, { "epoch": 2.81, "learning_rate": 1.1752946083940086e-05, "loss": 0.7943, "step": 27124 }, { "epoch": 2.81, "learning_rate": 1.1750992376074055e-05, "loss": 0.7082, "step": 27125 }, { "epoch": 2.81, "learning_rate": 1.1749038791056083e-05, "loss": 0.5541, "step": 27126 }, { "epoch": 2.81, "learning_rate": 1.1747085328899302e-05, "loss": 0.6976, "step": 27127 }, { "epoch": 2.81, "learning_rate": 1.1745131989616872e-05, "loss": 0.1213, "step": 27128 }, { "epoch": 2.81, "learning_rate": 1.1743178773221955e-05, "loss": 0.7793, "step": 27129 }, { "epoch": 2.81, "learning_rate": 1.1741225679727672e-05, "loss": 0.6359, "step": 27130 }, { "epoch": 2.81, "learning_rate": 1.1739272709147188e-05, "loss": 0.1363, "step": 27131 }, { "epoch": 2.81, "learning_rate": 1.1737319861493648e-05, "loss": 0.7932, "step": 27132 }, { "epoch": 2.81, "learning_rate": 1.1735367136780198e-05, "loss": 0.2513, "step": 27133 }, { "epoch": 2.81, "learning_rate": 1.173341453501998e-05, "loss": 0.364, "step": 27134 }, { "epoch": 2.81, "learning_rate": 1.1731462056226144e-05, "loss": 0.2597, "step": 27135 }, { "epoch": 2.81, "learning_rate": 1.1729509700411828e-05, "loss": 0.1275, "step": 27136 }, { "epoch": 2.81, "learning_rate": 1.1727557467590192e-05, "loss": 0.0721, "step": 27137 }, { "epoch": 2.81, "learning_rate": 1.1725605357774349e-05, "loss": 0.3755, "step": 27138 }, { "epoch": 2.81, "learning_rate": 1.172365337097747e-05, "loss": 0.6861, "step": 27139 }, { "epoch": 2.81, "learning_rate": 1.1721701507212668e-05, "loss": 0.3028, "step": 27140 }, { "epoch": 2.81, "learning_rate": 1.1719749766493094e-05, "loss": 0.4448, "step": 27141 }, { "epoch": 2.81, "learning_rate": 1.1717798148831886e-05, "loss": 0.0706, "step": 27142 }, { "epoch": 2.81, "learning_rate": 1.1715846654242183e-05, "loss": 0.1414, "step": 27143 }, { "epoch": 2.81, "learning_rate": 1.1713895282737126e-05, "loss": 0.0301, "step": 27144 }, { "epoch": 2.81, "learning_rate": 1.1711944034329845e-05, "loss": 0.2722, "step": 27145 }, { "epoch": 2.81, "learning_rate": 1.1709992909033475e-05, "loss": 0.224, "step": 27146 }, { "epoch": 2.81, "learning_rate": 1.1708041906861167e-05, "loss": 0.5891, "step": 27147 }, { "epoch": 2.81, "learning_rate": 1.170609102782603e-05, "loss": 0.2165, "step": 27148 }, { "epoch": 2.81, "learning_rate": 1.170414027194121e-05, "loss": 0.1977, "step": 27149 }, { "epoch": 2.81, "learning_rate": 1.1702189639219835e-05, "loss": 0.3215, "step": 27150 }, { "epoch": 2.81, "learning_rate": 1.1700239129675046e-05, "loss": 0.4896, "step": 27151 }, { "epoch": 2.81, "learning_rate": 1.1698288743319958e-05, "loss": 0.2103, "step": 27152 }, { "epoch": 2.81, "learning_rate": 1.1696338480167705e-05, "loss": 0.4611, "step": 27153 }, { "epoch": 2.81, "learning_rate": 1.1694388340231424e-05, "loss": 0.5688, "step": 27154 }, { "epoch": 2.81, "learning_rate": 1.1692438323524232e-05, "loss": 0.3005, "step": 27155 }, { "epoch": 2.81, "learning_rate": 1.1690488430059263e-05, "loss": 0.292, "step": 27156 }, { "epoch": 2.81, "learning_rate": 1.1688538659849654e-05, "loss": 0.5574, "step": 27157 }, { "epoch": 2.81, "learning_rate": 1.1686589012908507e-05, "loss": 0.6353, "step": 27158 }, { "epoch": 2.81, "learning_rate": 1.1684639489248958e-05, "loss": 0.7578, "step": 27159 }, { "epoch": 2.81, "learning_rate": 1.168269008888413e-05, "loss": 0.4134, "step": 27160 }, { "epoch": 2.81, "learning_rate": 1.1680740811827145e-05, "loss": 0.5953, "step": 27161 }, { "epoch": 2.81, "learning_rate": 1.1678791658091128e-05, "loss": 0.0262, "step": 27162 }, { "epoch": 2.81, "learning_rate": 1.1676842627689196e-05, "loss": 0.3605, "step": 27163 }, { "epoch": 2.81, "learning_rate": 1.1674893720634486e-05, "loss": 0.7432, "step": 27164 }, { "epoch": 2.81, "learning_rate": 1.1672944936940092e-05, "loss": 0.2317, "step": 27165 }, { "epoch": 2.81, "learning_rate": 1.1670996276619144e-05, "loss": 0.1668, "step": 27166 }, { "epoch": 2.81, "learning_rate": 1.1669047739684772e-05, "loss": 0.0925, "step": 27167 }, { "epoch": 2.81, "learning_rate": 1.1667099326150073e-05, "loss": 0.1721, "step": 27168 }, { "epoch": 2.81, "learning_rate": 1.166515103602817e-05, "loss": 0.0846, "step": 27169 }, { "epoch": 2.81, "learning_rate": 1.1663202869332178e-05, "loss": 0.4156, "step": 27170 }, { "epoch": 2.81, "learning_rate": 1.1661254826075216e-05, "loss": 0.21, "step": 27171 }, { "epoch": 2.81, "learning_rate": 1.1659306906270396e-05, "loss": 0.3505, "step": 27172 }, { "epoch": 2.81, "learning_rate": 1.1657359109930827e-05, "loss": 0.1291, "step": 27173 }, { "epoch": 2.81, "learning_rate": 1.165541143706963e-05, "loss": 0.1834, "step": 27174 }, { "epoch": 2.81, "learning_rate": 1.1653463887699919e-05, "loss": 0.622, "step": 27175 }, { "epoch": 2.81, "learning_rate": 1.1651516461834784e-05, "loss": 0.4403, "step": 27176 }, { "epoch": 2.81, "learning_rate": 1.164956915948736e-05, "loss": 0.2628, "step": 27177 }, { "epoch": 2.81, "learning_rate": 1.164762198067073e-05, "loss": 0.1709, "step": 27178 }, { "epoch": 2.81, "learning_rate": 1.1645674925398016e-05, "loss": 0.6235, "step": 27179 }, { "epoch": 2.81, "learning_rate": 1.1643727993682323e-05, "loss": 0.2453, "step": 27180 }, { "epoch": 2.81, "learning_rate": 1.1641781185536762e-05, "loss": 0.231, "step": 27181 }, { "epoch": 2.81, "learning_rate": 1.163983450097443e-05, "loss": 0.3384, "step": 27182 }, { "epoch": 2.81, "learning_rate": 1.1637887940008439e-05, "loss": 0.2133, "step": 27183 }, { "epoch": 2.81, "learning_rate": 1.1635941502651889e-05, "loss": 0.4088, "step": 27184 }, { "epoch": 2.81, "learning_rate": 1.1633995188917892e-05, "loss": 0.145, "step": 27185 }, { "epoch": 2.81, "learning_rate": 1.1632048998819533e-05, "loss": 0.6255, "step": 27186 }, { "epoch": 2.81, "learning_rate": 1.1630102932369925e-05, "loss": 0.0984, "step": 27187 }, { "epoch": 2.81, "learning_rate": 1.1628156989582163e-05, "loss": 0.1004, "step": 27188 }, { "epoch": 2.81, "learning_rate": 1.1626211170469362e-05, "loss": 0.3956, "step": 27189 }, { "epoch": 2.81, "learning_rate": 1.1624265475044597e-05, "loss": 0.1106, "step": 27190 }, { "epoch": 2.81, "learning_rate": 1.1622319903320976e-05, "loss": 0.2664, "step": 27191 }, { "epoch": 2.81, "learning_rate": 1.1620374455311597e-05, "loss": 0.1843, "step": 27192 }, { "epoch": 2.81, "learning_rate": 1.1618429131029561e-05, "loss": 0.229, "step": 27193 }, { "epoch": 2.81, "learning_rate": 1.1616483930487956e-05, "loss": 0.2022, "step": 27194 }, { "epoch": 2.81, "learning_rate": 1.161453885369989e-05, "loss": 0.1076, "step": 27195 }, { "epoch": 2.81, "learning_rate": 1.1612593900678436e-05, "loss": 0.3779, "step": 27196 }, { "epoch": 2.81, "learning_rate": 1.1610649071436698e-05, "loss": 0.6717, "step": 27197 }, { "epoch": 2.81, "learning_rate": 1.160870436598777e-05, "loss": 0.4461, "step": 27198 }, { "epoch": 2.81, "learning_rate": 1.1606759784344737e-05, "loss": 0.349, "step": 27199 }, { "epoch": 2.81, "learning_rate": 1.1604815326520695e-05, "loss": 0.3895, "step": 27200 }, { "epoch": 2.81, "learning_rate": 1.1602870992528732e-05, "loss": 0.4754, "step": 27201 }, { "epoch": 2.81, "learning_rate": 1.160092678238195e-05, "loss": 0.1454, "step": 27202 }, { "epoch": 2.81, "learning_rate": 1.159898269609341e-05, "loss": 0.0559, "step": 27203 }, { "epoch": 2.81, "learning_rate": 1.1597038733676215e-05, "loss": 0.388, "step": 27204 }, { "epoch": 2.81, "learning_rate": 1.1595094895143459e-05, "loss": 0.6033, "step": 27205 }, { "epoch": 2.81, "learning_rate": 1.1593151180508205e-05, "loss": 0.2595, "step": 27206 }, { "epoch": 2.81, "learning_rate": 1.1591207589783552e-05, "loss": 0.3566, "step": 27207 }, { "epoch": 2.81, "learning_rate": 1.1589264122982585e-05, "loss": 0.1943, "step": 27208 }, { "epoch": 2.82, "learning_rate": 1.1587320780118381e-05, "loss": 0.2814, "step": 27209 }, { "epoch": 2.82, "learning_rate": 1.1585377561204029e-05, "loss": 0.173, "step": 27210 }, { "epoch": 2.82, "learning_rate": 1.15834344662526e-05, "loss": 0.0599, "step": 27211 }, { "epoch": 2.82, "learning_rate": 1.1581491495277187e-05, "loss": 0.2083, "step": 27212 }, { "epoch": 2.82, "learning_rate": 1.1579548648290874e-05, "loss": 0.838, "step": 27213 }, { "epoch": 2.82, "learning_rate": 1.1577605925306714e-05, "loss": 0.5247, "step": 27214 }, { "epoch": 2.82, "learning_rate": 1.1575663326337816e-05, "loss": 0.1584, "step": 27215 }, { "epoch": 2.82, "learning_rate": 1.1573720851397227e-05, "loss": 0.0466, "step": 27216 }, { "epoch": 2.82, "learning_rate": 1.1571778500498041e-05, "loss": 0.5147, "step": 27217 }, { "epoch": 2.82, "learning_rate": 1.1569836273653329e-05, "loss": 0.5589, "step": 27218 }, { "epoch": 2.82, "learning_rate": 1.1567894170876165e-05, "loss": 0.4018, "step": 27219 }, { "epoch": 2.82, "learning_rate": 1.1565952192179627e-05, "loss": 0.1857, "step": 27220 }, { "epoch": 2.82, "learning_rate": 1.1564010337576787e-05, "loss": 0.1602, "step": 27221 }, { "epoch": 2.82, "learning_rate": 1.1562068607080712e-05, "loss": 0.2488, "step": 27222 }, { "epoch": 2.82, "learning_rate": 1.156012700070449e-05, "loss": 0.5554, "step": 27223 }, { "epoch": 2.82, "learning_rate": 1.1558185518461167e-05, "loss": 0.2565, "step": 27224 }, { "epoch": 2.82, "learning_rate": 1.1556244160363823e-05, "loss": 0.075, "step": 27225 }, { "epoch": 2.82, "learning_rate": 1.1554302926425531e-05, "loss": 0.2017, "step": 27226 }, { "epoch": 2.82, "learning_rate": 1.1552361816659364e-05, "loss": 0.4265, "step": 27227 }, { "epoch": 2.82, "learning_rate": 1.1550420831078372e-05, "loss": 0.1016, "step": 27228 }, { "epoch": 2.82, "learning_rate": 1.154847996969563e-05, "loss": 0.2825, "step": 27229 }, { "epoch": 2.82, "learning_rate": 1.15465392325242e-05, "loss": 0.4293, "step": 27230 }, { "epoch": 2.82, "learning_rate": 1.1544598619577153e-05, "loss": 0.1086, "step": 27231 }, { "epoch": 2.82, "learning_rate": 1.1542658130867551e-05, "loss": 0.3276, "step": 27232 }, { "epoch": 2.82, "learning_rate": 1.1540717766408464e-05, "loss": 0.3868, "step": 27233 }, { "epoch": 2.82, "learning_rate": 1.1538777526212936e-05, "loss": 0.309, "step": 27234 }, { "epoch": 2.82, "learning_rate": 1.1536837410294036e-05, "loss": 0.3675, "step": 27235 }, { "epoch": 2.82, "learning_rate": 1.1534897418664832e-05, "loss": 0.4227, "step": 27236 }, { "epoch": 2.82, "learning_rate": 1.1532957551338374e-05, "loss": 0.0769, "step": 27237 }, { "epoch": 2.82, "learning_rate": 1.1531017808327726e-05, "loss": 0.1021, "step": 27238 }, { "epoch": 2.82, "learning_rate": 1.1529078189645955e-05, "loss": 0.2554, "step": 27239 }, { "epoch": 2.82, "learning_rate": 1.1527138695306096e-05, "loss": 0.0623, "step": 27240 }, { "epoch": 2.82, "learning_rate": 1.1525199325321218e-05, "loss": 0.0088, "step": 27241 }, { "epoch": 2.82, "learning_rate": 1.1523260079704376e-05, "loss": 0.3933, "step": 27242 }, { "epoch": 2.82, "learning_rate": 1.1521320958468633e-05, "loss": 0.7305, "step": 27243 }, { "epoch": 2.82, "learning_rate": 1.1519381961627025e-05, "loss": 0.1616, "step": 27244 }, { "epoch": 2.82, "learning_rate": 1.1517443089192614e-05, "loss": 0.453, "step": 27245 }, { "epoch": 2.82, "learning_rate": 1.1515504341178448e-05, "loss": 0.2107, "step": 27246 }, { "epoch": 2.82, "learning_rate": 1.1513565717597585e-05, "loss": 0.1434, "step": 27247 }, { "epoch": 2.82, "learning_rate": 1.1511627218463072e-05, "loss": 0.4673, "step": 27248 }, { "epoch": 2.82, "learning_rate": 1.1509688843787955e-05, "loss": 0.6689, "step": 27249 }, { "epoch": 2.82, "learning_rate": 1.150775059358529e-05, "loss": 0.5596, "step": 27250 }, { "epoch": 2.82, "learning_rate": 1.1505812467868132e-05, "loss": 0.0093, "step": 27251 }, { "epoch": 2.82, "learning_rate": 1.1503874466649504e-05, "loss": 0.652, "step": 27252 }, { "epoch": 2.82, "learning_rate": 1.1501936589942477e-05, "loss": 0.0608, "step": 27253 }, { "epoch": 2.82, "learning_rate": 1.1499998837760076e-05, "loss": 0.4527, "step": 27254 }, { "epoch": 2.82, "learning_rate": 1.1498061210115353e-05, "loss": 0.1032, "step": 27255 }, { "epoch": 2.82, "learning_rate": 1.1496123707021353e-05, "loss": 0.2717, "step": 27256 }, { "epoch": 2.82, "learning_rate": 1.1494186328491116e-05, "loss": 0.4136, "step": 27257 }, { "epoch": 2.82, "learning_rate": 1.1492249074537689e-05, "loss": 0.3201, "step": 27258 }, { "epoch": 2.82, "learning_rate": 1.1490311945174109e-05, "loss": 0.7206, "step": 27259 }, { "epoch": 2.82, "learning_rate": 1.148837494041342e-05, "loss": 0.1286, "step": 27260 }, { "epoch": 2.82, "learning_rate": 1.148643806026867e-05, "loss": 0.555, "step": 27261 }, { "epoch": 2.82, "learning_rate": 1.1484501304752877e-05, "loss": 0.2419, "step": 27262 }, { "epoch": 2.82, "learning_rate": 1.1482564673879086e-05, "loss": 0.6479, "step": 27263 }, { "epoch": 2.82, "learning_rate": 1.148062816766035e-05, "loss": 0.5209, "step": 27264 }, { "epoch": 2.82, "learning_rate": 1.1478691786109682e-05, "loss": 0.2693, "step": 27265 }, { "epoch": 2.82, "learning_rate": 1.1476755529240126e-05, "loss": 0.6749, "step": 27266 }, { "epoch": 2.82, "learning_rate": 1.1474819397064718e-05, "loss": 0.1744, "step": 27267 }, { "epoch": 2.82, "learning_rate": 1.1472883389596492e-05, "loss": 0.2157, "step": 27268 }, { "epoch": 2.82, "learning_rate": 1.1470947506848477e-05, "loss": 0.0724, "step": 27269 }, { "epoch": 2.82, "learning_rate": 1.1469011748833712e-05, "loss": 0.6269, "step": 27270 }, { "epoch": 2.82, "learning_rate": 1.1467076115565233e-05, "loss": 0.3695, "step": 27271 }, { "epoch": 2.82, "learning_rate": 1.1465140607056047e-05, "loss": 0.2471, "step": 27272 }, { "epoch": 2.82, "learning_rate": 1.1463205223319202e-05, "loss": 0.367, "step": 27273 }, { "epoch": 2.82, "learning_rate": 1.146126996436772e-05, "loss": 0.3486, "step": 27274 }, { "epoch": 2.82, "learning_rate": 1.1459334830214636e-05, "loss": 0.6582, "step": 27275 }, { "epoch": 2.82, "learning_rate": 1.1457399820872967e-05, "loss": 0.3851, "step": 27276 }, { "epoch": 2.82, "learning_rate": 1.1455464936355754e-05, "loss": 0.0919, "step": 27277 }, { "epoch": 2.82, "learning_rate": 1.1453530176676002e-05, "loss": 0.3561, "step": 27278 }, { "epoch": 2.82, "learning_rate": 1.1451595541846746e-05, "loss": 0.3342, "step": 27279 }, { "epoch": 2.82, "learning_rate": 1.1449661031881009e-05, "loss": 0.5432, "step": 27280 }, { "epoch": 2.82, "learning_rate": 1.1447726646791822e-05, "loss": 0.5481, "step": 27281 }, { "epoch": 2.82, "learning_rate": 1.1445792386592193e-05, "loss": 0.1543, "step": 27282 }, { "epoch": 2.82, "learning_rate": 1.1443858251295145e-05, "loss": 0.6847, "step": 27283 }, { "epoch": 2.82, "learning_rate": 1.1441924240913697e-05, "loss": 0.281, "step": 27284 }, { "epoch": 2.82, "learning_rate": 1.143999035546088e-05, "loss": 0.6883, "step": 27285 }, { "epoch": 2.82, "learning_rate": 1.1438056594949702e-05, "loss": 0.03, "step": 27286 }, { "epoch": 2.82, "learning_rate": 1.1436122959393183e-05, "loss": 0.7286, "step": 27287 }, { "epoch": 2.82, "learning_rate": 1.1434189448804344e-05, "loss": 0.8104, "step": 27288 }, { "epoch": 2.82, "learning_rate": 1.1432256063196204e-05, "loss": 0.7646, "step": 27289 }, { "epoch": 2.82, "learning_rate": 1.1430322802581766e-05, "loss": 0.3679, "step": 27290 }, { "epoch": 2.82, "learning_rate": 1.1428389666974055e-05, "loss": 0.3543, "step": 27291 }, { "epoch": 2.82, "learning_rate": 1.1426456656386073e-05, "loss": 0.7699, "step": 27292 }, { "epoch": 2.82, "learning_rate": 1.1424523770830834e-05, "loss": 0.2351, "step": 27293 }, { "epoch": 2.82, "learning_rate": 1.142259101032136e-05, "loss": 0.1244, "step": 27294 }, { "epoch": 2.82, "learning_rate": 1.1420658374870652e-05, "loss": 0.1166, "step": 27295 }, { "epoch": 2.82, "learning_rate": 1.1418725864491727e-05, "loss": 0.3755, "step": 27296 }, { "epoch": 2.82, "learning_rate": 1.1416793479197595e-05, "loss": 0.4367, "step": 27297 }, { "epoch": 2.82, "learning_rate": 1.1414861219001256e-05, "loss": 0.1952, "step": 27298 }, { "epoch": 2.82, "learning_rate": 1.1412929083915737e-05, "loss": 0.8778, "step": 27299 }, { "epoch": 2.82, "learning_rate": 1.1410997073954018e-05, "loss": 0.4654, "step": 27300 }, { "epoch": 2.82, "learning_rate": 1.1409065189129118e-05, "loss": 0.4849, "step": 27301 }, { "epoch": 2.82, "learning_rate": 1.1407133429454049e-05, "loss": 0.4937, "step": 27302 }, { "epoch": 2.82, "learning_rate": 1.14052017949418e-05, "loss": 0.0481, "step": 27303 }, { "epoch": 2.82, "learning_rate": 1.140327028560538e-05, "loss": 0.9389, "step": 27304 }, { "epoch": 2.83, "learning_rate": 1.1401338901457793e-05, "loss": 0.4084, "step": 27305 }, { "epoch": 2.83, "learning_rate": 1.1399407642512043e-05, "loss": 0.6985, "step": 27306 }, { "epoch": 2.83, "learning_rate": 1.1397476508781127e-05, "loss": 0.5342, "step": 27307 }, { "epoch": 2.83, "learning_rate": 1.1395545500278046e-05, "loss": 0.5101, "step": 27308 }, { "epoch": 2.83, "learning_rate": 1.139361461701581e-05, "loss": 0.4129, "step": 27309 }, { "epoch": 2.83, "learning_rate": 1.1391683859007396e-05, "loss": 0.2904, "step": 27310 }, { "epoch": 2.83, "learning_rate": 1.1389753226265814e-05, "loss": 0.2426, "step": 27311 }, { "epoch": 2.83, "learning_rate": 1.1387822718804058e-05, "loss": 0.3941, "step": 27312 }, { "epoch": 2.83, "learning_rate": 1.1385892336635125e-05, "loss": 0.4212, "step": 27313 }, { "epoch": 2.83, "learning_rate": 1.1383962079772012e-05, "loss": 0.3196, "step": 27314 }, { "epoch": 2.83, "learning_rate": 1.1382031948227716e-05, "loss": 0.7047, "step": 27315 }, { "epoch": 2.83, "learning_rate": 1.1380101942015215e-05, "loss": 0.7004, "step": 27316 }, { "epoch": 2.83, "learning_rate": 1.1378172061147512e-05, "loss": 1.0289, "step": 27317 }, { "epoch": 2.83, "learning_rate": 1.1376242305637602e-05, "loss": 0.371, "step": 27318 }, { "epoch": 2.83, "learning_rate": 1.1374312675498476e-05, "loss": 0.2266, "step": 27319 }, { "epoch": 2.83, "learning_rate": 1.1372383170743107e-05, "loss": 0.5303, "step": 27320 }, { "epoch": 2.83, "learning_rate": 1.1370453791384503e-05, "loss": 0.6174, "step": 27321 }, { "epoch": 2.83, "learning_rate": 1.1368524537435641e-05, "loss": 0.7729, "step": 27322 }, { "epoch": 2.83, "learning_rate": 1.1366595408909511e-05, "loss": 0.4091, "step": 27323 }, { "epoch": 2.83, "learning_rate": 1.1364666405819103e-05, "loss": 0.7552, "step": 27324 }, { "epoch": 2.83, "learning_rate": 1.1362737528177398e-05, "loss": 0.3361, "step": 27325 }, { "epoch": 2.83, "learning_rate": 1.1360808775997386e-05, "loss": 0.2269, "step": 27326 }, { "epoch": 2.83, "learning_rate": 1.1358880149292057e-05, "loss": 0.0978, "step": 27327 }, { "epoch": 2.83, "learning_rate": 1.1356951648074375e-05, "loss": 0.2577, "step": 27328 }, { "epoch": 2.83, "learning_rate": 1.1355023272357344e-05, "loss": 0.5765, "step": 27329 }, { "epoch": 2.83, "learning_rate": 1.1353095022153921e-05, "loss": 0.4198, "step": 27330 }, { "epoch": 2.83, "learning_rate": 1.1351166897477101e-05, "loss": 0.2911, "step": 27331 }, { "epoch": 2.83, "learning_rate": 1.1349238898339863e-05, "loss": 0.3894, "step": 27332 }, { "epoch": 2.83, "learning_rate": 1.1347311024755184e-05, "loss": 0.6741, "step": 27333 }, { "epoch": 2.83, "learning_rate": 1.1345383276736042e-05, "loss": 0.6514, "step": 27334 }, { "epoch": 2.83, "learning_rate": 1.1343455654295416e-05, "loss": 0.4127, "step": 27335 }, { "epoch": 2.83, "learning_rate": 1.1341528157446279e-05, "loss": 0.1322, "step": 27336 }, { "epoch": 2.83, "learning_rate": 1.1339600786201622e-05, "loss": 0.0699, "step": 27337 }, { "epoch": 2.83, "learning_rate": 1.1337673540574392e-05, "loss": 0.1882, "step": 27338 }, { "epoch": 2.83, "learning_rate": 1.133574642057758e-05, "loss": 0.2035, "step": 27339 }, { "epoch": 2.83, "learning_rate": 1.1333819426224165e-05, "loss": 0.5053, "step": 27340 }, { "epoch": 2.83, "learning_rate": 1.1331892557527103e-05, "loss": 0.2786, "step": 27341 }, { "epoch": 2.83, "learning_rate": 1.1329965814499368e-05, "loss": 0.447, "step": 27342 }, { "epoch": 2.83, "learning_rate": 1.1328039197153938e-05, "loss": 0.1241, "step": 27343 }, { "epoch": 2.83, "learning_rate": 1.1326112705503778e-05, "loss": 0.3566, "step": 27344 }, { "epoch": 2.83, "learning_rate": 1.1324186339561857e-05, "loss": 0.458, "step": 27345 }, { "epoch": 2.83, "learning_rate": 1.1322260099341144e-05, "loss": 0.534, "step": 27346 }, { "epoch": 2.83, "learning_rate": 1.1320333984854618e-05, "loss": 0.113, "step": 27347 }, { "epoch": 2.83, "learning_rate": 1.1318407996115222e-05, "loss": 0.3142, "step": 27348 }, { "epoch": 2.83, "learning_rate": 1.1316482133135932e-05, "loss": 0.3596, "step": 27349 }, { "epoch": 2.83, "learning_rate": 1.131455639592971e-05, "loss": 0.1666, "step": 27350 }, { "epoch": 2.83, "learning_rate": 1.1312630784509525e-05, "loss": 0.0687, "step": 27351 }, { "epoch": 2.83, "learning_rate": 1.1310705298888346e-05, "loss": 0.4549, "step": 27352 }, { "epoch": 2.83, "learning_rate": 1.1308779939079118e-05, "loss": 0.6306, "step": 27353 }, { "epoch": 2.83, "learning_rate": 1.1306854705094807e-05, "loss": 0.4016, "step": 27354 }, { "epoch": 2.83, "learning_rate": 1.1304929596948379e-05, "loss": 0.3554, "step": 27355 }, { "epoch": 2.83, "learning_rate": 1.1303004614652791e-05, "loss": 0.5007, "step": 27356 }, { "epoch": 2.83, "learning_rate": 1.130107975822101e-05, "loss": 0.1214, "step": 27357 }, { "epoch": 2.83, "learning_rate": 1.1299155027665973e-05, "loss": 0.6419, "step": 27358 }, { "epoch": 2.83, "learning_rate": 1.1297230423000652e-05, "loss": 0.2426, "step": 27359 }, { "epoch": 2.83, "learning_rate": 1.1295305944237996e-05, "loss": 0.3178, "step": 27360 }, { "epoch": 2.83, "learning_rate": 1.1293381591390968e-05, "loss": 0.3065, "step": 27361 }, { "epoch": 2.83, "learning_rate": 1.1291457364472513e-05, "loss": 0.5206, "step": 27362 }, { "epoch": 2.83, "learning_rate": 1.1289533263495593e-05, "loss": 0.3311, "step": 27363 }, { "epoch": 2.83, "learning_rate": 1.1287609288473157e-05, "loss": 0.0659, "step": 27364 }, { "epoch": 2.83, "learning_rate": 1.1285685439418167e-05, "loss": 0.0766, "step": 27365 }, { "epoch": 2.83, "learning_rate": 1.1283761716343553e-05, "loss": 0.2487, "step": 27366 }, { "epoch": 2.83, "learning_rate": 1.1281838119262288e-05, "loss": 0.1577, "step": 27367 }, { "epoch": 2.83, "learning_rate": 1.1279914648187296e-05, "loss": 0.2397, "step": 27368 }, { "epoch": 2.83, "learning_rate": 1.1277991303131541e-05, "loss": 0.2021, "step": 27369 }, { "epoch": 2.83, "learning_rate": 1.1276068084107966e-05, "loss": 0.8039, "step": 27370 }, { "epoch": 2.83, "learning_rate": 1.127414499112952e-05, "loss": 0.4368, "step": 27371 }, { "epoch": 2.83, "learning_rate": 1.1272222024209149e-05, "loss": 0.152, "step": 27372 }, { "epoch": 2.83, "learning_rate": 1.1270299183359797e-05, "loss": 0.5245, "step": 27373 }, { "epoch": 2.83, "learning_rate": 1.126837646859441e-05, "loss": 0.593, "step": 27374 }, { "epoch": 2.83, "learning_rate": 1.1266453879925938e-05, "loss": 0.2649, "step": 27375 }, { "epoch": 2.83, "learning_rate": 1.1264531417367306e-05, "loss": 0.1013, "step": 27376 }, { "epoch": 2.83, "learning_rate": 1.1262609080931463e-05, "loss": 0.0942, "step": 27377 }, { "epoch": 2.83, "learning_rate": 1.1260686870631365e-05, "loss": 0.3531, "step": 27378 }, { "epoch": 2.83, "learning_rate": 1.1258764786479925e-05, "loss": 0.2238, "step": 27379 }, { "epoch": 2.83, "learning_rate": 1.1256842828490095e-05, "loss": 0.2698, "step": 27380 }, { "epoch": 2.83, "learning_rate": 1.1254920996674818e-05, "loss": 0.8024, "step": 27381 }, { "epoch": 2.83, "learning_rate": 1.1252999291047024e-05, "loss": 0.1692, "step": 27382 }, { "epoch": 2.83, "learning_rate": 1.1251077711619654e-05, "loss": 0.5478, "step": 27383 }, { "epoch": 2.83, "learning_rate": 1.124915625840564e-05, "loss": 0.1944, "step": 27384 }, { "epoch": 2.83, "learning_rate": 1.1247234931417926e-05, "loss": 0.2043, "step": 27385 }, { "epoch": 2.83, "learning_rate": 1.124531373066943e-05, "loss": 0.2655, "step": 27386 }, { "epoch": 2.83, "learning_rate": 1.1243392656173097e-05, "loss": 0.3215, "step": 27387 }, { "epoch": 2.83, "learning_rate": 1.1241471707941855e-05, "loss": 0.6578, "step": 27388 }, { "epoch": 2.83, "learning_rate": 1.1239550885988634e-05, "loss": 0.1159, "step": 27389 }, { "epoch": 2.83, "learning_rate": 1.1237630190326382e-05, "loss": 0.198, "step": 27390 }, { "epoch": 2.83, "learning_rate": 1.1235709620967998e-05, "loss": 0.7081, "step": 27391 }, { "epoch": 2.83, "learning_rate": 1.1233789177926427e-05, "loss": 0.0825, "step": 27392 }, { "epoch": 2.83, "learning_rate": 1.1231868861214599e-05, "loss": 0.6187, "step": 27393 }, { "epoch": 2.83, "learning_rate": 1.1229948670845436e-05, "loss": 0.632, "step": 27394 }, { "epoch": 2.83, "learning_rate": 1.122802860683188e-05, "loss": 0.2976, "step": 27395 }, { "epoch": 2.83, "learning_rate": 1.122610866918683e-05, "loss": 0.416, "step": 27396 }, { "epoch": 2.83, "learning_rate": 1.1224188857923223e-05, "loss": 1.0981, "step": 27397 }, { "epoch": 2.83, "learning_rate": 1.1222269173053986e-05, "loss": 0.4272, "step": 27398 }, { "epoch": 2.83, "learning_rate": 1.1220349614592037e-05, "loss": 0.0952, "step": 27399 }, { "epoch": 2.83, "learning_rate": 1.1218430182550303e-05, "loss": 0.6389, "step": 27400 }, { "epoch": 2.84, "learning_rate": 1.1216510876941702e-05, "loss": 0.4119, "step": 27401 }, { "epoch": 2.84, "learning_rate": 1.1214591697779153e-05, "loss": 0.3042, "step": 27402 }, { "epoch": 2.84, "learning_rate": 1.1212672645075587e-05, "loss": 0.3005, "step": 27403 }, { "epoch": 2.84, "learning_rate": 1.1210753718843905e-05, "loss": 0.0345, "step": 27404 }, { "epoch": 2.84, "learning_rate": 1.1208834919097041e-05, "loss": 0.0075, "step": 27405 }, { "epoch": 2.84, "learning_rate": 1.1206916245847894e-05, "loss": 0.3827, "step": 27406 }, { "epoch": 2.84, "learning_rate": 1.120499769910939e-05, "loss": 0.1797, "step": 27407 }, { "epoch": 2.84, "learning_rate": 1.1203079278894445e-05, "loss": 0.3154, "step": 27408 }, { "epoch": 2.84, "learning_rate": 1.120116098521597e-05, "loss": 0.2864, "step": 27409 }, { "epoch": 2.84, "learning_rate": 1.119924281808688e-05, "loss": 0.2229, "step": 27410 }, { "epoch": 2.84, "learning_rate": 1.119732477752009e-05, "loss": 0.5285, "step": 27411 }, { "epoch": 2.84, "learning_rate": 1.1195406863528512e-05, "loss": 0.1327, "step": 27412 }, { "epoch": 2.84, "learning_rate": 1.1193489076125062e-05, "loss": 0.6554, "step": 27413 }, { "epoch": 2.84, "learning_rate": 1.1191571415322636e-05, "loss": 0.0134, "step": 27414 }, { "epoch": 2.84, "learning_rate": 1.1189653881134155e-05, "loss": 0.5032, "step": 27415 }, { "epoch": 2.84, "learning_rate": 1.1187736473572517e-05, "loss": 0.2224, "step": 27416 }, { "epoch": 2.84, "learning_rate": 1.1185819192650635e-05, "loss": 0.0704, "step": 27417 }, { "epoch": 2.84, "learning_rate": 1.1183902038381418e-05, "loss": 0.435, "step": 27418 }, { "epoch": 2.84, "learning_rate": 1.1181985010777764e-05, "loss": 0.5103, "step": 27419 }, { "epoch": 2.84, "learning_rate": 1.118006810985259e-05, "loss": 0.153, "step": 27420 }, { "epoch": 2.84, "learning_rate": 1.117815133561879e-05, "loss": 0.6978, "step": 27421 }, { "epoch": 2.84, "learning_rate": 1.1176234688089274e-05, "loss": 0.3357, "step": 27422 }, { "epoch": 2.84, "learning_rate": 1.117431816727695e-05, "loss": 0.0461, "step": 27423 }, { "epoch": 2.84, "learning_rate": 1.11724017731947e-05, "loss": 0.176, "step": 27424 }, { "epoch": 2.84, "learning_rate": 1.1170485505855438e-05, "loss": 0.558, "step": 27425 }, { "epoch": 2.84, "learning_rate": 1.1168569365272063e-05, "loss": 0.5281, "step": 27426 }, { "epoch": 2.84, "learning_rate": 1.1166653351457469e-05, "loss": 0.456, "step": 27427 }, { "epoch": 2.84, "learning_rate": 1.116473746442457e-05, "loss": 0.3559, "step": 27428 }, { "epoch": 2.84, "learning_rate": 1.1162821704186243e-05, "loss": 0.7675, "step": 27429 }, { "epoch": 2.84, "learning_rate": 1.1160906070755391e-05, "loss": 0.4655, "step": 27430 }, { "epoch": 2.84, "learning_rate": 1.1158990564144909e-05, "loss": 0.0929, "step": 27431 }, { "epoch": 2.84, "learning_rate": 1.1157075184367699e-05, "loss": 0.3751, "step": 27432 }, { "epoch": 2.84, "learning_rate": 1.1155159931436657e-05, "loss": 0.1444, "step": 27433 }, { "epoch": 2.84, "learning_rate": 1.1153244805364661e-05, "loss": 0.1666, "step": 27434 }, { "epoch": 2.84, "learning_rate": 1.115132980616461e-05, "loss": 0.3837, "step": 27435 }, { "epoch": 2.84, "learning_rate": 1.1149414933849395e-05, "loss": 0.1704, "step": 27436 }, { "epoch": 2.84, "learning_rate": 1.1147500188431908e-05, "loss": 0.3595, "step": 27437 }, { "epoch": 2.84, "learning_rate": 1.114558556992504e-05, "loss": 0.5394, "step": 27438 }, { "epoch": 2.84, "learning_rate": 1.1143671078341678e-05, "loss": 0.3682, "step": 27439 }, { "epoch": 2.84, "learning_rate": 1.114175671369472e-05, "loss": 0.519, "step": 27440 }, { "epoch": 2.84, "learning_rate": 1.1139842475997034e-05, "loss": 0.2122, "step": 27441 }, { "epoch": 2.84, "learning_rate": 1.1137928365261515e-05, "loss": 0.5022, "step": 27442 }, { "epoch": 2.84, "learning_rate": 1.1136014381501059e-05, "loss": 0.3034, "step": 27443 }, { "epoch": 2.84, "learning_rate": 1.1134100524728531e-05, "loss": 0.4354, "step": 27444 }, { "epoch": 2.84, "learning_rate": 1.1132186794956826e-05, "loss": 0.0951, "step": 27445 }, { "epoch": 2.84, "learning_rate": 1.113027319219882e-05, "loss": 0.222, "step": 27446 }, { "epoch": 2.84, "learning_rate": 1.1128359716467404e-05, "loss": 0.6815, "step": 27447 }, { "epoch": 2.84, "learning_rate": 1.1126446367775453e-05, "loss": 0.0196, "step": 27448 }, { "epoch": 2.84, "learning_rate": 1.1124533146135847e-05, "loss": 0.5257, "step": 27449 }, { "epoch": 2.84, "learning_rate": 1.112262005156147e-05, "loss": 0.2859, "step": 27450 }, { "epoch": 2.84, "learning_rate": 1.1120707084065197e-05, "loss": 0.1648, "step": 27451 }, { "epoch": 2.84, "learning_rate": 1.1118794243659916e-05, "loss": 0.2286, "step": 27452 }, { "epoch": 2.84, "learning_rate": 1.1116881530358494e-05, "loss": 0.2556, "step": 27453 }, { "epoch": 2.84, "learning_rate": 1.1114968944173798e-05, "loss": 0.2594, "step": 27454 }, { "epoch": 2.84, "learning_rate": 1.1113056485118715e-05, "loss": 0.2011, "step": 27455 }, { "epoch": 2.84, "learning_rate": 1.1111144153206112e-05, "loss": 0.6033, "step": 27456 }, { "epoch": 2.84, "learning_rate": 1.1109231948448866e-05, "loss": 0.2445, "step": 27457 }, { "epoch": 2.84, "learning_rate": 1.1107319870859855e-05, "loss": 0.155, "step": 27458 }, { "epoch": 2.84, "learning_rate": 1.1105407920451947e-05, "loss": 0.3728, "step": 27459 }, { "epoch": 2.84, "learning_rate": 1.1103496097238005e-05, "loss": 0.292, "step": 27460 }, { "epoch": 2.84, "learning_rate": 1.1101584401230922e-05, "loss": 0.5004, "step": 27461 }, { "epoch": 2.84, "learning_rate": 1.109967283244354e-05, "loss": 0.2072, "step": 27462 }, { "epoch": 2.84, "learning_rate": 1.1097761390888738e-05, "loss": 0.1992, "step": 27463 }, { "epoch": 2.84, "learning_rate": 1.1095850076579383e-05, "loss": 0.1966, "step": 27464 }, { "epoch": 2.84, "learning_rate": 1.1093938889528351e-05, "loss": 0.1703, "step": 27465 }, { "epoch": 2.84, "learning_rate": 1.109202782974849e-05, "loss": 0.2852, "step": 27466 }, { "epoch": 2.84, "learning_rate": 1.1090116897252673e-05, "loss": 0.2721, "step": 27467 }, { "epoch": 2.84, "learning_rate": 1.1088206092053765e-05, "loss": 0.8547, "step": 27468 }, { "epoch": 2.84, "learning_rate": 1.1086295414164626e-05, "loss": 0.6016, "step": 27469 }, { "epoch": 2.84, "learning_rate": 1.1084384863598123e-05, "loss": 0.299, "step": 27470 }, { "epoch": 2.84, "learning_rate": 1.1082474440367125e-05, "loss": 0.1908, "step": 27471 }, { "epoch": 2.84, "learning_rate": 1.1080564144484473e-05, "loss": 0.8419, "step": 27472 }, { "epoch": 2.84, "learning_rate": 1.1078653975963035e-05, "loss": 0.4529, "step": 27473 }, { "epoch": 2.84, "learning_rate": 1.1076743934815674e-05, "loss": 0.4076, "step": 27474 }, { "epoch": 2.84, "learning_rate": 1.1074834021055244e-05, "loss": 0.1379, "step": 27475 }, { "epoch": 2.84, "learning_rate": 1.1072924234694604e-05, "loss": 0.1144, "step": 27476 }, { "epoch": 2.84, "learning_rate": 1.1071014575746607e-05, "loss": 0.5722, "step": 27477 }, { "epoch": 2.84, "learning_rate": 1.1069105044224123e-05, "loss": 0.0937, "step": 27478 }, { "epoch": 2.84, "learning_rate": 1.1067195640139982e-05, "loss": 0.1569, "step": 27479 }, { "epoch": 2.84, "learning_rate": 1.1065286363507051e-05, "loss": 0.19, "step": 27480 }, { "epoch": 2.84, "learning_rate": 1.1063377214338193e-05, "loss": 0.2875, "step": 27481 }, { "epoch": 2.84, "learning_rate": 1.1061468192646241e-05, "loss": 0.133, "step": 27482 }, { "epoch": 2.84, "learning_rate": 1.1059559298444049e-05, "loss": 0.4695, "step": 27483 }, { "epoch": 2.84, "learning_rate": 1.1057650531744474e-05, "loss": 0.4072, "step": 27484 }, { "epoch": 2.84, "learning_rate": 1.1055741892560364e-05, "loss": 0.1696, "step": 27485 }, { "epoch": 2.84, "learning_rate": 1.1053833380904568e-05, "loss": 0.3276, "step": 27486 }, { "epoch": 2.84, "learning_rate": 1.105192499678993e-05, "loss": 0.3524, "step": 27487 }, { "epoch": 2.84, "learning_rate": 1.1050016740229299e-05, "loss": 0.7447, "step": 27488 }, { "epoch": 2.84, "learning_rate": 1.1048108611235524e-05, "loss": 0.4255, "step": 27489 }, { "epoch": 2.84, "learning_rate": 1.1046200609821458e-05, "loss": 0.5595, "step": 27490 }, { "epoch": 2.84, "learning_rate": 1.104429273599993e-05, "loss": 0.5293, "step": 27491 }, { "epoch": 2.84, "learning_rate": 1.1042384989783783e-05, "loss": 0.2077, "step": 27492 }, { "epoch": 2.84, "learning_rate": 1.104047737118586e-05, "loss": 0.3363, "step": 27493 }, { "epoch": 2.84, "learning_rate": 1.1038569880219012e-05, "loss": 0.6638, "step": 27494 }, { "epoch": 2.84, "learning_rate": 1.1036662516896074e-05, "loss": 0.4592, "step": 27495 }, { "epoch": 2.85, "learning_rate": 1.1034755281229885e-05, "loss": 0.6532, "step": 27496 }, { "epoch": 2.85, "learning_rate": 1.1032848173233285e-05, "loss": 0.2898, "step": 27497 }, { "epoch": 2.85, "learning_rate": 1.1030941192919116e-05, "loss": 0.1273, "step": 27498 }, { "epoch": 2.85, "learning_rate": 1.1029034340300222e-05, "loss": 0.0185, "step": 27499 }, { "epoch": 2.85, "learning_rate": 1.1027127615389418e-05, "loss": 0.3941, "step": 27500 }, { "epoch": 2.85, "learning_rate": 1.1025221018199553e-05, "loss": 0.2098, "step": 27501 }, { "epoch": 2.85, "learning_rate": 1.1023314548743461e-05, "loss": 0.2542, "step": 27502 }, { "epoch": 2.85, "learning_rate": 1.1021408207033985e-05, "loss": 0.3735, "step": 27503 }, { "epoch": 2.85, "learning_rate": 1.1019501993083938e-05, "loss": 0.3299, "step": 27504 }, { "epoch": 2.85, "learning_rate": 1.101759590690616e-05, "loss": 0.1341, "step": 27505 }, { "epoch": 2.85, "learning_rate": 1.1015689948513488e-05, "loss": 0.0538, "step": 27506 }, { "epoch": 2.85, "learning_rate": 1.1013784117918749e-05, "loss": 0.4932, "step": 27507 }, { "epoch": 2.85, "learning_rate": 1.1011878415134771e-05, "loss": 0.5348, "step": 27508 }, { "epoch": 2.85, "learning_rate": 1.1009972840174394e-05, "loss": 0.3023, "step": 27509 }, { "epoch": 2.85, "learning_rate": 1.100806739305043e-05, "loss": 0.3248, "step": 27510 }, { "epoch": 2.85, "learning_rate": 1.100616207377571e-05, "loss": 0.0988, "step": 27511 }, { "epoch": 2.85, "learning_rate": 1.1004256882363064e-05, "loss": 0.1298, "step": 27512 }, { "epoch": 2.85, "learning_rate": 1.1002351818825316e-05, "loss": 0.1085, "step": 27513 }, { "epoch": 2.85, "learning_rate": 1.1000446883175289e-05, "loss": 0.3337, "step": 27514 }, { "epoch": 2.85, "learning_rate": 1.0998542075425812e-05, "loss": 0.6607, "step": 27515 }, { "epoch": 2.85, "learning_rate": 1.099663739558971e-05, "loss": 0.4152, "step": 27516 }, { "epoch": 2.85, "learning_rate": 1.0994732843679789e-05, "loss": 0.2609, "step": 27517 }, { "epoch": 2.85, "learning_rate": 1.099282841970888e-05, "loss": 0.0984, "step": 27518 }, { "epoch": 2.85, "learning_rate": 1.0990924123689813e-05, "loss": 0.4097, "step": 27519 }, { "epoch": 2.85, "learning_rate": 1.0989019955635386e-05, "loss": 0.0671, "step": 27520 }, { "epoch": 2.85, "learning_rate": 1.098711591555843e-05, "loss": 0.3081, "step": 27521 }, { "epoch": 2.85, "learning_rate": 1.098521200347176e-05, "loss": 0.2445, "step": 27522 }, { "epoch": 2.85, "learning_rate": 1.0983308219388195e-05, "loss": 0.3194, "step": 27523 }, { "epoch": 2.85, "learning_rate": 1.0981404563320548e-05, "loss": 0.1096, "step": 27524 }, { "epoch": 2.85, "learning_rate": 1.0979501035281638e-05, "loss": 0.2026, "step": 27525 }, { "epoch": 2.85, "learning_rate": 1.0977597635284272e-05, "loss": 0.3105, "step": 27526 }, { "epoch": 2.85, "learning_rate": 1.0975694363341272e-05, "loss": 0.3385, "step": 27527 }, { "epoch": 2.85, "learning_rate": 1.0973791219465453e-05, "loss": 0.1316, "step": 27528 }, { "epoch": 2.85, "learning_rate": 1.097188820366962e-05, "loss": 0.4287, "step": 27529 }, { "epoch": 2.85, "learning_rate": 1.0969985315966578e-05, "loss": 0.5539, "step": 27530 }, { "epoch": 2.85, "learning_rate": 1.096808255636914e-05, "loss": 0.4183, "step": 27531 }, { "epoch": 2.85, "learning_rate": 1.0966179924890114e-05, "loss": 0.0642, "step": 27532 }, { "epoch": 2.85, "learning_rate": 1.0964277421542314e-05, "loss": 0.2562, "step": 27533 }, { "epoch": 2.85, "learning_rate": 1.0962375046338544e-05, "loss": 0.4764, "step": 27534 }, { "epoch": 2.85, "learning_rate": 1.096047279929161e-05, "loss": 0.1271, "step": 27535 }, { "epoch": 2.85, "learning_rate": 1.0958570680414318e-05, "loss": 0.0462, "step": 27536 }, { "epoch": 2.85, "learning_rate": 1.0956668689719473e-05, "loss": 0.3187, "step": 27537 }, { "epoch": 2.85, "learning_rate": 1.0954766827219889e-05, "loss": 0.0541, "step": 27538 }, { "epoch": 2.85, "learning_rate": 1.095286509292835e-05, "loss": 0.1773, "step": 27539 }, { "epoch": 2.85, "learning_rate": 1.0950963486857663e-05, "loss": 0.2333, "step": 27540 }, { "epoch": 2.85, "learning_rate": 1.0949062009020648e-05, "loss": 0.4632, "step": 27541 }, { "epoch": 2.85, "learning_rate": 1.0947160659430077e-05, "loss": 0.3126, "step": 27542 }, { "epoch": 2.85, "learning_rate": 1.0945259438098762e-05, "loss": 0.7422, "step": 27543 }, { "epoch": 2.85, "learning_rate": 1.0943358345039502e-05, "loss": 0.3715, "step": 27544 }, { "epoch": 2.85, "learning_rate": 1.0941457380265094e-05, "loss": 0.0341, "step": 27545 }, { "epoch": 2.85, "learning_rate": 1.0939556543788336e-05, "loss": 0.2863, "step": 27546 }, { "epoch": 2.85, "learning_rate": 1.0937655835622035e-05, "loss": 0.8651, "step": 27547 }, { "epoch": 2.85, "learning_rate": 1.0935755255778962e-05, "loss": 0.2003, "step": 27548 }, { "epoch": 2.85, "learning_rate": 1.0933854804271928e-05, "loss": 0.2952, "step": 27549 }, { "epoch": 2.85, "learning_rate": 1.0931954481113718e-05, "loss": 0.0867, "step": 27550 }, { "epoch": 2.85, "learning_rate": 1.0930054286317131e-05, "loss": 0.7277, "step": 27551 }, { "epoch": 2.85, "learning_rate": 1.0928154219894956e-05, "loss": 0.131, "step": 27552 }, { "epoch": 2.85, "learning_rate": 1.0926254281859998e-05, "loss": 0.8379, "step": 27553 }, { "epoch": 2.85, "learning_rate": 1.092435447222502e-05, "loss": 0.3871, "step": 27554 }, { "epoch": 2.85, "learning_rate": 1.0922454791002824e-05, "loss": 0.1649, "step": 27555 }, { "epoch": 2.85, "learning_rate": 1.0920555238206199e-05, "loss": 0.6618, "step": 27556 }, { "epoch": 2.85, "learning_rate": 1.091865581384794e-05, "loss": 0.1936, "step": 27557 }, { "epoch": 2.85, "learning_rate": 1.091675651794082e-05, "loss": 0.1547, "step": 27558 }, { "epoch": 2.85, "learning_rate": 1.0914857350497628e-05, "loss": 0.5202, "step": 27559 }, { "epoch": 2.85, "learning_rate": 1.0912958311531148e-05, "loss": 0.3629, "step": 27560 }, { "epoch": 2.85, "learning_rate": 1.0911059401054167e-05, "loss": 0.2156, "step": 27561 }, { "epoch": 2.85, "learning_rate": 1.0909160619079472e-05, "loss": 0.6411, "step": 27562 }, { "epoch": 2.85, "learning_rate": 1.0907261965619835e-05, "loss": 0.4782, "step": 27563 }, { "epoch": 2.85, "learning_rate": 1.0905363440688047e-05, "loss": 0.1291, "step": 27564 }, { "epoch": 2.85, "learning_rate": 1.0903465044296883e-05, "loss": 0.5209, "step": 27565 }, { "epoch": 2.85, "learning_rate": 1.0901566776459137e-05, "loss": 0.0643, "step": 27566 }, { "epoch": 2.85, "learning_rate": 1.0899668637187573e-05, "loss": 0.6564, "step": 27567 }, { "epoch": 2.85, "learning_rate": 1.0897770626494959e-05, "loss": 0.2164, "step": 27568 }, { "epoch": 2.85, "learning_rate": 1.0895872744394082e-05, "loss": 0.5117, "step": 27569 }, { "epoch": 2.85, "learning_rate": 1.0893974990897724e-05, "loss": 0.0881, "step": 27570 }, { "epoch": 2.85, "learning_rate": 1.0892077366018655e-05, "loss": 0.2974, "step": 27571 }, { "epoch": 2.85, "learning_rate": 1.0890179869769648e-05, "loss": 0.1334, "step": 27572 }, { "epoch": 2.85, "learning_rate": 1.0888282502163482e-05, "loss": 0.3951, "step": 27573 }, { "epoch": 2.85, "learning_rate": 1.0886385263212926e-05, "loss": 0.5089, "step": 27574 }, { "epoch": 2.85, "learning_rate": 1.0884488152930751e-05, "loss": 0.7676, "step": 27575 }, { "epoch": 2.85, "learning_rate": 1.088259117132974e-05, "loss": 0.1725, "step": 27576 }, { "epoch": 2.85, "learning_rate": 1.088069431842264e-05, "loss": 0.1075, "step": 27577 }, { "epoch": 2.85, "learning_rate": 1.0878797594222232e-05, "loss": 0.2972, "step": 27578 }, { "epoch": 2.85, "learning_rate": 1.0876900998741297e-05, "loss": 0.2779, "step": 27579 }, { "epoch": 2.85, "learning_rate": 1.0875004531992579e-05, "loss": 0.796, "step": 27580 }, { "epoch": 2.85, "learning_rate": 1.0873108193988857e-05, "loss": 0.3007, "step": 27581 }, { "epoch": 2.85, "learning_rate": 1.0871211984742895e-05, "loss": 0.8997, "step": 27582 }, { "epoch": 2.85, "learning_rate": 1.0869315904267457e-05, "loss": 0.2261, "step": 27583 }, { "epoch": 2.85, "learning_rate": 1.0867419952575306e-05, "loss": 0.0407, "step": 27584 }, { "epoch": 2.85, "learning_rate": 1.0865524129679221e-05, "loss": 0.2886, "step": 27585 }, { "epoch": 2.85, "learning_rate": 1.0863628435591938e-05, "loss": 0.2396, "step": 27586 }, { "epoch": 2.85, "learning_rate": 1.0861732870326232e-05, "loss": 0.1471, "step": 27587 }, { "epoch": 2.85, "learning_rate": 1.0859837433894862e-05, "loss": 0.4541, "step": 27588 }, { "epoch": 2.85, "learning_rate": 1.085794212631059e-05, "loss": 0.5184, "step": 27589 }, { "epoch": 2.85, "learning_rate": 1.085604694758617e-05, "loss": 0.3031, "step": 27590 }, { "epoch": 2.85, "learning_rate": 1.085415189773437e-05, "loss": 0.0402, "step": 27591 }, { "epoch": 2.86, "learning_rate": 1.0852256976767936e-05, "loss": 0.0583, "step": 27592 }, { "epoch": 2.86, "learning_rate": 1.0850362184699627e-05, "loss": 0.1524, "step": 27593 }, { "epoch": 2.86, "learning_rate": 1.0848467521542195e-05, "loss": 0.144, "step": 27594 }, { "epoch": 2.86, "learning_rate": 1.0846572987308409e-05, "loss": 0.274, "step": 27595 }, { "epoch": 2.86, "learning_rate": 1.0844678582011003e-05, "loss": 0.7751, "step": 27596 }, { "epoch": 2.86, "learning_rate": 1.084278430566274e-05, "loss": 0.7794, "step": 27597 }, { "epoch": 2.86, "learning_rate": 1.0840890158276372e-05, "loss": 0.4266, "step": 27598 }, { "epoch": 2.86, "learning_rate": 1.0838996139864645e-05, "loss": 0.1911, "step": 27599 }, { "epoch": 2.86, "learning_rate": 1.0837102250440314e-05, "loss": 0.8785, "step": 27600 }, { "epoch": 2.86, "learning_rate": 1.0835208490016127e-05, "loss": 0.1977, "step": 27601 }, { "epoch": 2.86, "learning_rate": 1.0833314858604835e-05, "loss": 0.6211, "step": 27602 }, { "epoch": 2.86, "learning_rate": 1.083142135621918e-05, "loss": 0.1754, "step": 27603 }, { "epoch": 2.86, "learning_rate": 1.082952798287192e-05, "loss": 0.2056, "step": 27604 }, { "epoch": 2.86, "learning_rate": 1.0827634738575793e-05, "loss": 0.163, "step": 27605 }, { "epoch": 2.86, "learning_rate": 1.0825741623343534e-05, "loss": 0.1026, "step": 27606 }, { "epoch": 2.86, "learning_rate": 1.0823848637187896e-05, "loss": 0.7261, "step": 27607 }, { "epoch": 2.86, "learning_rate": 1.0821955780121621e-05, "loss": 0.1269, "step": 27608 }, { "epoch": 2.86, "learning_rate": 1.0820063052157453e-05, "loss": 0.8951, "step": 27609 }, { "epoch": 2.86, "learning_rate": 1.0818170453308132e-05, "loss": 1.4669, "step": 27610 }, { "epoch": 2.86, "learning_rate": 1.08162779835864e-05, "loss": 0.1046, "step": 27611 }, { "epoch": 2.86, "learning_rate": 1.0814385643004997e-05, "loss": 0.9073, "step": 27612 }, { "epoch": 2.86, "learning_rate": 1.081249343157666e-05, "loss": 0.4814, "step": 27613 }, { "epoch": 2.86, "learning_rate": 1.081060134931414e-05, "loss": 0.4252, "step": 27614 }, { "epoch": 2.86, "learning_rate": 1.0808709396230147e-05, "loss": 0.0722, "step": 27615 }, { "epoch": 2.86, "learning_rate": 1.0806817572337446e-05, "loss": 0.2685, "step": 27616 }, { "epoch": 2.86, "learning_rate": 1.0804925877648745e-05, "loss": 0.1589, "step": 27617 }, { "epoch": 2.86, "learning_rate": 1.0803034312176792e-05, "loss": 0.2897, "step": 27618 }, { "epoch": 2.86, "learning_rate": 1.0801142875934324e-05, "loss": 0.2178, "step": 27619 }, { "epoch": 2.86, "learning_rate": 1.0799251568934068e-05, "loss": 0.247, "step": 27620 }, { "epoch": 2.86, "learning_rate": 1.0797360391188755e-05, "loss": 0.1661, "step": 27621 }, { "epoch": 2.86, "learning_rate": 1.0795469342711123e-05, "loss": 0.7726, "step": 27622 }, { "epoch": 2.86, "learning_rate": 1.0793578423513893e-05, "loss": 0.1526, "step": 27623 }, { "epoch": 2.86, "learning_rate": 1.0791687633609813e-05, "loss": 0.7118, "step": 27624 }, { "epoch": 2.86, "learning_rate": 1.0789796973011587e-05, "loss": 0.1818, "step": 27625 }, { "epoch": 2.86, "learning_rate": 1.0787906441731951e-05, "loss": 0.5601, "step": 27626 }, { "epoch": 2.86, "learning_rate": 1.0786016039783637e-05, "loss": 0.3218, "step": 27627 }, { "epoch": 2.86, "learning_rate": 1.0784125767179366e-05, "loss": 0.41, "step": 27628 }, { "epoch": 2.86, "learning_rate": 1.0782235623931874e-05, "loss": 0.1411, "step": 27629 }, { "epoch": 2.86, "learning_rate": 1.0780345610053865e-05, "loss": 0.7919, "step": 27630 }, { "epoch": 2.86, "learning_rate": 1.0778455725558069e-05, "loss": 0.3889, "step": 27631 }, { "epoch": 2.86, "learning_rate": 1.0776565970457217e-05, "loss": 0.2532, "step": 27632 }, { "epoch": 2.86, "learning_rate": 1.0774676344764034e-05, "loss": 0.0302, "step": 27633 }, { "epoch": 2.86, "learning_rate": 1.0772786848491219e-05, "loss": 1.0021, "step": 27634 }, { "epoch": 2.86, "learning_rate": 1.0770897481651507e-05, "loss": 0.1898, "step": 27635 }, { "epoch": 2.86, "learning_rate": 1.076900824425761e-05, "loss": 0.2217, "step": 27636 }, { "epoch": 2.86, "learning_rate": 1.0767119136322251e-05, "loss": 0.2503, "step": 27637 }, { "epoch": 2.86, "learning_rate": 1.076523015785815e-05, "loss": 0.9026, "step": 27638 }, { "epoch": 2.86, "learning_rate": 1.0763341308878014e-05, "loss": 0.3791, "step": 27639 }, { "epoch": 2.86, "learning_rate": 1.0761452589394567e-05, "loss": 0.2895, "step": 27640 }, { "epoch": 2.86, "learning_rate": 1.0759563999420524e-05, "loss": 0.2249, "step": 27641 }, { "epoch": 2.86, "learning_rate": 1.075767553896859e-05, "loss": 0.5122, "step": 27642 }, { "epoch": 2.86, "learning_rate": 1.075578720805149e-05, "loss": 0.7403, "step": 27643 }, { "epoch": 2.86, "learning_rate": 1.0753899006681917e-05, "loss": 0.2328, "step": 27644 }, { "epoch": 2.86, "learning_rate": 1.0752010934872592e-05, "loss": 0.6413, "step": 27645 }, { "epoch": 2.86, "learning_rate": 1.0750122992636228e-05, "loss": 0.2717, "step": 27646 }, { "epoch": 2.86, "learning_rate": 1.0748235179985532e-05, "loss": 0.6648, "step": 27647 }, { "epoch": 2.86, "learning_rate": 1.0746347496933211e-05, "loss": 0.5482, "step": 27648 }, { "epoch": 2.86, "learning_rate": 1.0744459943491971e-05, "loss": 0.3463, "step": 27649 }, { "epoch": 2.86, "learning_rate": 1.0742572519674525e-05, "loss": 0.2325, "step": 27650 }, { "epoch": 2.86, "learning_rate": 1.0740685225493577e-05, "loss": 0.5103, "step": 27651 }, { "epoch": 2.86, "learning_rate": 1.0738798060961834e-05, "loss": 0.844, "step": 27652 }, { "epoch": 2.86, "learning_rate": 1.0736911026091988e-05, "loss": 0.2603, "step": 27653 }, { "epoch": 2.86, "learning_rate": 1.073502412089676e-05, "loss": 0.1027, "step": 27654 }, { "epoch": 2.86, "learning_rate": 1.0733137345388833e-05, "loss": 0.2102, "step": 27655 }, { "epoch": 2.86, "learning_rate": 1.0731250699580916e-05, "loss": 0.2521, "step": 27656 }, { "epoch": 2.86, "learning_rate": 1.0729364183485713e-05, "loss": 0.5005, "step": 27657 }, { "epoch": 2.86, "learning_rate": 1.072747779711592e-05, "loss": 0.5509, "step": 27658 }, { "epoch": 2.86, "learning_rate": 1.0725591540484234e-05, "loss": 0.2272, "step": 27659 }, { "epoch": 2.86, "learning_rate": 1.0723705413603358e-05, "loss": 0.2658, "step": 27660 }, { "epoch": 2.86, "learning_rate": 1.072181941648599e-05, "loss": 0.2125, "step": 27661 }, { "epoch": 2.86, "learning_rate": 1.071993354914483e-05, "loss": 0.2929, "step": 27662 }, { "epoch": 2.86, "learning_rate": 1.0718047811592557e-05, "loss": 0.1304, "step": 27663 }, { "epoch": 2.86, "learning_rate": 1.0716162203841874e-05, "loss": 0.1059, "step": 27664 }, { "epoch": 2.86, "learning_rate": 1.0714276725905473e-05, "loss": 0.085, "step": 27665 }, { "epoch": 2.86, "learning_rate": 1.071239137779606e-05, "loss": 0.1575, "step": 27666 }, { "epoch": 2.86, "learning_rate": 1.071050615952631e-05, "loss": 0.4302, "step": 27667 }, { "epoch": 2.86, "learning_rate": 1.070862107110891e-05, "loss": 0.0317, "step": 27668 }, { "epoch": 2.86, "learning_rate": 1.0706736112556566e-05, "loss": 0.5426, "step": 27669 }, { "epoch": 2.86, "learning_rate": 1.070485128388196e-05, "loss": 0.4561, "step": 27670 }, { "epoch": 2.86, "learning_rate": 1.070296658509779e-05, "loss": 0.0659, "step": 27671 }, { "epoch": 2.86, "learning_rate": 1.0701082016216723e-05, "loss": 0.7033, "step": 27672 }, { "epoch": 2.86, "learning_rate": 1.0699197577251454e-05, "loss": 0.3494, "step": 27673 }, { "epoch": 2.86, "learning_rate": 1.0697313268214674e-05, "loss": 0.6296, "step": 27674 }, { "epoch": 2.86, "learning_rate": 1.069542908911906e-05, "loss": 0.2276, "step": 27675 }, { "epoch": 2.86, "learning_rate": 1.0693545039977306e-05, "loss": 0.534, "step": 27676 }, { "epoch": 2.86, "learning_rate": 1.0691661120802085e-05, "loss": 0.2034, "step": 27677 }, { "epoch": 2.86, "learning_rate": 1.0689777331606088e-05, "loss": 0.3584, "step": 27678 }, { "epoch": 2.86, "learning_rate": 1.0687893672402e-05, "loss": 0.2734, "step": 27679 }, { "epoch": 2.86, "learning_rate": 1.0686010143202483e-05, "loss": 0.2345, "step": 27680 }, { "epoch": 2.86, "learning_rate": 1.0684126744020235e-05, "loss": 0.1246, "step": 27681 }, { "epoch": 2.86, "learning_rate": 1.068224347486792e-05, "loss": 0.6031, "step": 27682 }, { "epoch": 2.86, "learning_rate": 1.0680360335758218e-05, "loss": 0.5978, "step": 27683 }, { "epoch": 2.86, "learning_rate": 1.0678477326703813e-05, "loss": 0.4006, "step": 27684 }, { "epoch": 2.86, "learning_rate": 1.0676594447717374e-05, "loss": 0.2529, "step": 27685 }, { "epoch": 2.86, "learning_rate": 1.0674711698811582e-05, "loss": 0.1888, "step": 27686 }, { "epoch": 2.86, "learning_rate": 1.0672829079999107e-05, "loss": 0.1649, "step": 27687 }, { "epoch": 2.87, "learning_rate": 1.0670946591292627e-05, "loss": 0.8645, "step": 27688 }, { "epoch": 2.87, "learning_rate": 1.0669064232704813e-05, "loss": 0.3529, "step": 27689 }, { "epoch": 2.87, "learning_rate": 1.0667182004248344e-05, "loss": 0.1273, "step": 27690 }, { "epoch": 2.87, "learning_rate": 1.0665299905935874e-05, "loss": 1.0809, "step": 27691 }, { "epoch": 2.87, "learning_rate": 1.0663417937780088e-05, "loss": 0.5159, "step": 27692 }, { "epoch": 2.87, "learning_rate": 1.066153609979364e-05, "loss": 0.2658, "step": 27693 }, { "epoch": 2.87, "learning_rate": 1.0659654391989206e-05, "loss": 0.4843, "step": 27694 }, { "epoch": 2.87, "learning_rate": 1.065777281437945e-05, "loss": 0.239, "step": 27695 }, { "epoch": 2.87, "learning_rate": 1.0655891366977046e-05, "loss": 0.683, "step": 27696 }, { "epoch": 2.87, "learning_rate": 1.0654010049794652e-05, "loss": 0.4376, "step": 27697 }, { "epoch": 2.87, "learning_rate": 1.0652128862844937e-05, "loss": 0.1422, "step": 27698 }, { "epoch": 2.87, "learning_rate": 1.0650247806140564e-05, "loss": 0.2473, "step": 27699 }, { "epoch": 2.87, "learning_rate": 1.0648366879694204e-05, "loss": 0.7598, "step": 27700 }, { "epoch": 2.87, "learning_rate": 1.0646486083518503e-05, "loss": 0.0222, "step": 27701 }, { "epoch": 2.87, "learning_rate": 1.0644605417626129e-05, "loss": 1.3472, "step": 27702 }, { "epoch": 2.87, "learning_rate": 1.0642724882029738e-05, "loss": 0.5493, "step": 27703 }, { "epoch": 2.87, "learning_rate": 1.0640844476742008e-05, "loss": 0.2434, "step": 27704 }, { "epoch": 2.87, "learning_rate": 1.0638964201775573e-05, "loss": 0.1239, "step": 27705 }, { "epoch": 2.87, "learning_rate": 1.06370840571431e-05, "loss": 0.4295, "step": 27706 }, { "epoch": 2.87, "learning_rate": 1.0635204042857245e-05, "loss": 0.3781, "step": 27707 }, { "epoch": 2.87, "learning_rate": 1.063332415893067e-05, "loss": 0.4676, "step": 27708 }, { "epoch": 2.87, "learning_rate": 1.0631444405376031e-05, "loss": 0.3716, "step": 27709 }, { "epoch": 2.87, "learning_rate": 1.0629564782205969e-05, "loss": 0.5947, "step": 27710 }, { "epoch": 2.87, "learning_rate": 1.0627685289433142e-05, "loss": 0.0968, "step": 27711 }, { "epoch": 2.87, "learning_rate": 1.0625805927070206e-05, "loss": 0.2888, "step": 27712 }, { "epoch": 2.87, "learning_rate": 1.0623926695129812e-05, "loss": 0.4461, "step": 27713 }, { "epoch": 2.87, "learning_rate": 1.062204759362461e-05, "loss": 0.506, "step": 27714 }, { "epoch": 2.87, "learning_rate": 1.0620168622567249e-05, "loss": 0.2686, "step": 27715 }, { "epoch": 2.87, "learning_rate": 1.0618289781970376e-05, "loss": 0.1159, "step": 27716 }, { "epoch": 2.87, "learning_rate": 1.0616411071846655e-05, "loss": 0.8016, "step": 27717 }, { "epoch": 2.87, "learning_rate": 1.0614532492208707e-05, "loss": 0.0344, "step": 27718 }, { "epoch": 2.87, "learning_rate": 1.0612654043069201e-05, "loss": 0.5691, "step": 27719 }, { "epoch": 2.87, "learning_rate": 1.0610775724440763e-05, "loss": 0.5386, "step": 27720 }, { "epoch": 2.87, "learning_rate": 1.0608897536336044e-05, "loss": 0.3266, "step": 27721 }, { "epoch": 2.87, "learning_rate": 1.0607019478767693e-05, "loss": 0.2934, "step": 27722 }, { "epoch": 2.87, "learning_rate": 1.0605141551748344e-05, "loss": 0.7165, "step": 27723 }, { "epoch": 2.87, "learning_rate": 1.060326375529065e-05, "loss": 0.1629, "step": 27724 }, { "epoch": 2.87, "learning_rate": 1.060138608940724e-05, "loss": 0.8043, "step": 27725 }, { "epoch": 2.87, "learning_rate": 1.0599508554110764e-05, "loss": 0.0881, "step": 27726 }, { "epoch": 2.87, "learning_rate": 1.0597631149413854e-05, "loss": 0.4085, "step": 27727 }, { "epoch": 2.87, "learning_rate": 1.0595753875329165e-05, "loss": 0.1098, "step": 27728 }, { "epoch": 2.87, "learning_rate": 1.0593876731869319e-05, "loss": 0.3883, "step": 27729 }, { "epoch": 2.87, "learning_rate": 1.0591999719046945e-05, "loss": 0.3374, "step": 27730 }, { "epoch": 2.87, "learning_rate": 1.0590122836874685e-05, "loss": 0.1408, "step": 27731 }, { "epoch": 2.87, "learning_rate": 1.0588246085365179e-05, "loss": 0.5154, "step": 27732 }, { "epoch": 2.87, "learning_rate": 1.058636946453106e-05, "loss": 0.0789, "step": 27733 }, { "epoch": 2.87, "learning_rate": 1.0584492974384954e-05, "loss": 0.1895, "step": 27734 }, { "epoch": 2.87, "learning_rate": 1.0582616614939502e-05, "loss": 0.3733, "step": 27735 }, { "epoch": 2.87, "learning_rate": 1.058074038620733e-05, "loss": 0.2504, "step": 27736 }, { "epoch": 2.87, "learning_rate": 1.0578864288201073e-05, "loss": 0.0535, "step": 27737 }, { "epoch": 2.87, "learning_rate": 1.0576988320933365e-05, "loss": 0.4468, "step": 27738 }, { "epoch": 2.87, "learning_rate": 1.0575112484416815e-05, "loss": 0.1128, "step": 27739 }, { "epoch": 2.87, "learning_rate": 1.0573236778664067e-05, "loss": 0.592, "step": 27740 }, { "epoch": 2.87, "learning_rate": 1.0571361203687741e-05, "loss": 0.2706, "step": 27741 }, { "epoch": 2.87, "learning_rate": 1.0569485759500476e-05, "loss": 0.3578, "step": 27742 }, { "epoch": 2.87, "learning_rate": 1.056761044611488e-05, "loss": 0.7349, "step": 27743 }, { "epoch": 2.87, "learning_rate": 1.0565735263543578e-05, "loss": 0.7148, "step": 27744 }, { "epoch": 2.87, "learning_rate": 1.05638602117992e-05, "loss": 0.6604, "step": 27745 }, { "epoch": 2.87, "learning_rate": 1.0561985290894369e-05, "loss": 0.237, "step": 27746 }, { "epoch": 2.87, "learning_rate": 1.0560110500841706e-05, "loss": 0.3006, "step": 27747 }, { "epoch": 2.87, "learning_rate": 1.055823584165384e-05, "loss": 0.2097, "step": 27748 }, { "epoch": 2.87, "learning_rate": 1.0556361313343369e-05, "loss": 0.7279, "step": 27749 }, { "epoch": 2.87, "learning_rate": 1.0554486915922925e-05, "loss": 0.3538, "step": 27750 }, { "epoch": 2.87, "learning_rate": 1.0552612649405125e-05, "loss": 0.0269, "step": 27751 }, { "epoch": 2.87, "learning_rate": 1.0550738513802589e-05, "loss": 0.5226, "step": 27752 }, { "epoch": 2.87, "learning_rate": 1.054886450912793e-05, "loss": 0.24, "step": 27753 }, { "epoch": 2.87, "learning_rate": 1.0546990635393771e-05, "loss": 0.3134, "step": 27754 }, { "epoch": 2.87, "learning_rate": 1.054511689261271e-05, "loss": 0.4276, "step": 27755 }, { "epoch": 2.87, "learning_rate": 1.0543243280797372e-05, "loss": 0.3245, "step": 27756 }, { "epoch": 2.87, "learning_rate": 1.0541369799960376e-05, "loss": 0.3369, "step": 27757 }, { "epoch": 2.87, "learning_rate": 1.0539496450114317e-05, "loss": 0.2411, "step": 27758 }, { "epoch": 2.87, "learning_rate": 1.0537623231271813e-05, "loss": 0.1838, "step": 27759 }, { "epoch": 2.87, "learning_rate": 1.0535750143445477e-05, "loss": 0.1592, "step": 27760 }, { "epoch": 2.87, "learning_rate": 1.0533877186647914e-05, "loss": 0.1075, "step": 27761 }, { "epoch": 2.87, "learning_rate": 1.0532004360891738e-05, "loss": 0.2047, "step": 27762 }, { "epoch": 2.87, "learning_rate": 1.053013166618955e-05, "loss": 0.3251, "step": 27763 }, { "epoch": 2.87, "learning_rate": 1.0528259102553965e-05, "loss": 0.5185, "step": 27764 }, { "epoch": 2.87, "learning_rate": 1.052638666999758e-05, "loss": 0.3871, "step": 27765 }, { "epoch": 2.87, "learning_rate": 1.0524514368533018e-05, "loss": 0.3427, "step": 27766 }, { "epoch": 2.87, "learning_rate": 1.0522642198172863e-05, "loss": 0.4733, "step": 27767 }, { "epoch": 2.87, "learning_rate": 1.0520770158929714e-05, "loss": 0.685, "step": 27768 }, { "epoch": 2.87, "learning_rate": 1.0518898250816186e-05, "loss": 0.293, "step": 27769 }, { "epoch": 2.87, "learning_rate": 1.0517026473844875e-05, "loss": 0.312, "step": 27770 }, { "epoch": 2.87, "learning_rate": 1.0515154828028382e-05, "loss": 0.1391, "step": 27771 }, { "epoch": 2.87, "learning_rate": 1.0513283313379309e-05, "loss": 0.412, "step": 27772 }, { "epoch": 2.87, "learning_rate": 1.0511411929910253e-05, "loss": 0.4302, "step": 27773 }, { "epoch": 2.87, "learning_rate": 1.050954067763381e-05, "loss": 0.6887, "step": 27774 }, { "epoch": 2.87, "learning_rate": 1.0507669556562583e-05, "loss": 0.754, "step": 27775 }, { "epoch": 2.87, "learning_rate": 1.0505798566709172e-05, "loss": 0.1127, "step": 27776 }, { "epoch": 2.87, "learning_rate": 1.050392770808615e-05, "loss": 0.0488, "step": 27777 }, { "epoch": 2.87, "learning_rate": 1.0502056980706129e-05, "loss": 0.1818, "step": 27778 }, { "epoch": 2.87, "learning_rate": 1.0500186384581698e-05, "loss": 0.1109, "step": 27779 }, { "epoch": 2.87, "learning_rate": 1.0498315919725456e-05, "loss": 0.3019, "step": 27780 }, { "epoch": 2.87, "learning_rate": 1.0496445586149984e-05, "loss": 0.3103, "step": 27781 }, { "epoch": 2.87, "learning_rate": 1.0494575383867872e-05, "loss": 0.3735, "step": 27782 }, { "epoch": 2.87, "learning_rate": 1.0492705312891718e-05, "loss": 0.7555, "step": 27783 }, { "epoch": 2.88, "learning_rate": 1.0490835373234105e-05, "loss": 0.2895, "step": 27784 }, { "epoch": 2.88, "learning_rate": 1.0488965564907623e-05, "loss": 0.3727, "step": 27785 }, { "epoch": 2.88, "learning_rate": 1.0487095887924872e-05, "loss": 1.119, "step": 27786 }, { "epoch": 2.88, "learning_rate": 1.0485226342298415e-05, "loss": 0.7239, "step": 27787 }, { "epoch": 2.88, "learning_rate": 1.0483356928040849e-05, "loss": 0.5015, "step": 27788 }, { "epoch": 2.88, "learning_rate": 1.0481487645164758e-05, "loss": 0.2423, "step": 27789 }, { "epoch": 2.88, "learning_rate": 1.0479618493682721e-05, "loss": 0.3304, "step": 27790 }, { "epoch": 2.88, "learning_rate": 1.0477749473607328e-05, "loss": 0.2951, "step": 27791 }, { "epoch": 2.88, "learning_rate": 1.047588058495117e-05, "loss": 0.4618, "step": 27792 }, { "epoch": 2.88, "learning_rate": 1.04740118277268e-05, "loss": 0.1958, "step": 27793 }, { "epoch": 2.88, "learning_rate": 1.0472143201946817e-05, "loss": 0.5163, "step": 27794 }, { "epoch": 2.88, "learning_rate": 1.0470274707623809e-05, "loss": 0.5315, "step": 27795 }, { "epoch": 2.88, "learning_rate": 1.0468406344770329e-05, "loss": 0.1483, "step": 27796 }, { "epoch": 2.88, "learning_rate": 1.0466538113398966e-05, "loss": 0.1756, "step": 27797 }, { "epoch": 2.88, "learning_rate": 1.0464670013522301e-05, "loss": 0.3507, "step": 27798 }, { "epoch": 2.88, "learning_rate": 1.0462802045152905e-05, "loss": 0.569, "step": 27799 }, { "epoch": 2.88, "learning_rate": 1.0460934208303357e-05, "loss": 0.2164, "step": 27800 }, { "epoch": 2.88, "learning_rate": 1.0459066502986226e-05, "loss": 0.5598, "step": 27801 }, { "epoch": 2.88, "learning_rate": 1.0457198929214086e-05, "loss": 0.2871, "step": 27802 }, { "epoch": 2.88, "learning_rate": 1.045533148699951e-05, "loss": 0.3587, "step": 27803 }, { "epoch": 2.88, "learning_rate": 1.0453464176355081e-05, "loss": 0.4365, "step": 27804 }, { "epoch": 2.88, "learning_rate": 1.0451596997293359e-05, "loss": 0.2035, "step": 27805 }, { "epoch": 2.88, "learning_rate": 1.04497299498269e-05, "loss": 0.2306, "step": 27806 }, { "epoch": 2.88, "learning_rate": 1.0447863033968285e-05, "loss": 0.2427, "step": 27807 }, { "epoch": 2.88, "learning_rate": 1.044599624973008e-05, "loss": 0.3763, "step": 27808 }, { "epoch": 2.88, "learning_rate": 1.0444129597124855e-05, "loss": 0.2151, "step": 27809 }, { "epoch": 2.88, "learning_rate": 1.044226307616517e-05, "loss": 0.6978, "step": 27810 }, { "epoch": 2.88, "learning_rate": 1.0440396686863601e-05, "loss": 0.4688, "step": 27811 }, { "epoch": 2.88, "learning_rate": 1.04385304292327e-05, "loss": 0.044, "step": 27812 }, { "epoch": 2.88, "learning_rate": 1.0436664303285034e-05, "loss": 0.5928, "step": 27813 }, { "epoch": 2.88, "learning_rate": 1.043479830903318e-05, "loss": 0.1204, "step": 27814 }, { "epoch": 2.88, "learning_rate": 1.0432932446489671e-05, "loss": 0.1747, "step": 27815 }, { "epoch": 2.88, "learning_rate": 1.0431066715667088e-05, "loss": 0.4853, "step": 27816 }, { "epoch": 2.88, "learning_rate": 1.0429201116577993e-05, "loss": 0.974, "step": 27817 }, { "epoch": 2.88, "learning_rate": 1.0427335649234926e-05, "loss": 0.4414, "step": 27818 }, { "epoch": 2.88, "learning_rate": 1.0425470313650456e-05, "loss": 0.086, "step": 27819 }, { "epoch": 2.88, "learning_rate": 1.0423605109837139e-05, "loss": 0.216, "step": 27820 }, { "epoch": 2.88, "learning_rate": 1.0421740037807536e-05, "loss": 0.3521, "step": 27821 }, { "epoch": 2.88, "learning_rate": 1.0419875097574192e-05, "loss": 0.4435, "step": 27822 }, { "epoch": 2.88, "learning_rate": 1.041801028914967e-05, "loss": 0.0907, "step": 27823 }, { "epoch": 2.88, "learning_rate": 1.0416145612546533e-05, "loss": 0.3148, "step": 27824 }, { "epoch": 2.88, "learning_rate": 1.041428106777731e-05, "loss": 0.1831, "step": 27825 }, { "epoch": 2.88, "learning_rate": 1.0412416654854564e-05, "loss": 0.7782, "step": 27826 }, { "epoch": 2.88, "learning_rate": 1.0410552373790847e-05, "loss": 0.4537, "step": 27827 }, { "epoch": 2.88, "learning_rate": 1.0408688224598707e-05, "loss": 0.3856, "step": 27828 }, { "epoch": 2.88, "learning_rate": 1.0406824207290694e-05, "loss": 0.142, "step": 27829 }, { "epoch": 2.88, "learning_rate": 1.0404960321879364e-05, "loss": 0.7162, "step": 27830 }, { "epoch": 2.88, "learning_rate": 1.0403096568377249e-05, "loss": 0.5683, "step": 27831 }, { "epoch": 2.88, "learning_rate": 1.0401232946796903e-05, "loss": 0.2534, "step": 27832 }, { "epoch": 2.88, "learning_rate": 1.039936945715087e-05, "loss": 0.2697, "step": 27833 }, { "epoch": 2.88, "learning_rate": 1.0397506099451705e-05, "loss": 0.059, "step": 27834 }, { "epoch": 2.88, "learning_rate": 1.0395642873711933e-05, "loss": 0.2908, "step": 27835 }, { "epoch": 2.88, "learning_rate": 1.0393779779944104e-05, "loss": 0.1032, "step": 27836 }, { "epoch": 2.88, "learning_rate": 1.0391916818160765e-05, "loss": 0.1971, "step": 27837 }, { "epoch": 2.88, "learning_rate": 1.0390053988374452e-05, "loss": 0.2448, "step": 27838 }, { "epoch": 2.88, "learning_rate": 1.0388191290597707e-05, "loss": 0.1975, "step": 27839 }, { "epoch": 2.88, "learning_rate": 1.0386328724843072e-05, "loss": 0.2701, "step": 27840 }, { "epoch": 2.88, "learning_rate": 1.0384466291123083e-05, "loss": 0.4445, "step": 27841 }, { "epoch": 2.88, "learning_rate": 1.0382603989450283e-05, "loss": 0.3748, "step": 27842 }, { "epoch": 2.88, "learning_rate": 1.0380741819837207e-05, "loss": 0.2803, "step": 27843 }, { "epoch": 2.88, "learning_rate": 1.0378879782296375e-05, "loss": 0.34, "step": 27844 }, { "epoch": 2.88, "learning_rate": 1.0377017876840332e-05, "loss": 0.3223, "step": 27845 }, { "epoch": 2.88, "learning_rate": 1.0375156103481615e-05, "loss": 0.6984, "step": 27846 }, { "epoch": 2.88, "learning_rate": 1.0373294462232754e-05, "loss": 0.1681, "step": 27847 }, { "epoch": 2.88, "learning_rate": 1.0371432953106283e-05, "loss": 0.212, "step": 27848 }, { "epoch": 2.88, "learning_rate": 1.0369571576114733e-05, "loss": 0.0133, "step": 27849 }, { "epoch": 2.88, "learning_rate": 1.0367710331270635e-05, "loss": 0.8036, "step": 27850 }, { "epoch": 2.88, "learning_rate": 1.0365849218586516e-05, "loss": 0.181, "step": 27851 }, { "epoch": 2.88, "learning_rate": 1.0363988238074917e-05, "loss": 0.1661, "step": 27852 }, { "epoch": 2.88, "learning_rate": 1.0362127389748346e-05, "loss": 0.4258, "step": 27853 }, { "epoch": 2.88, "learning_rate": 1.0360266673619335e-05, "loss": 0.7346, "step": 27854 }, { "epoch": 2.88, "learning_rate": 1.0358406089700425e-05, "loss": 0.6057, "step": 27855 }, { "epoch": 2.88, "learning_rate": 1.0356545638004118e-05, "loss": 0.6455, "step": 27856 }, { "epoch": 2.88, "learning_rate": 1.035468531854295e-05, "loss": 0.2564, "step": 27857 }, { "epoch": 2.88, "learning_rate": 1.0352825131329445e-05, "loss": 0.3503, "step": 27858 }, { "epoch": 2.88, "learning_rate": 1.0350965076376123e-05, "loss": 0.2954, "step": 27859 }, { "epoch": 2.88, "learning_rate": 1.034910515369551e-05, "loss": 0.2446, "step": 27860 }, { "epoch": 2.88, "learning_rate": 1.034724536330012e-05, "loss": 0.2716, "step": 27861 }, { "epoch": 2.88, "learning_rate": 1.0345385705202488e-05, "loss": 0.2623, "step": 27862 }, { "epoch": 2.88, "learning_rate": 1.0343526179415107e-05, "loss": 0.1312, "step": 27863 }, { "epoch": 2.88, "learning_rate": 1.0341666785950511e-05, "loss": 0.5083, "step": 27864 }, { "epoch": 2.88, "learning_rate": 1.0339807524821216e-05, "loss": 0.2285, "step": 27865 }, { "epoch": 2.88, "learning_rate": 1.0337948396039732e-05, "loss": 0.257, "step": 27866 }, { "epoch": 2.88, "learning_rate": 1.0336089399618592e-05, "loss": 0.1973, "step": 27867 }, { "epoch": 2.88, "learning_rate": 1.0334230535570288e-05, "loss": 0.6128, "step": 27868 }, { "epoch": 2.88, "learning_rate": 1.033237180390734e-05, "loss": 0.2805, "step": 27869 }, { "epoch": 2.88, "learning_rate": 1.0330513204642265e-05, "loss": 0.4438, "step": 27870 }, { "epoch": 2.88, "learning_rate": 1.0328654737787572e-05, "loss": 0.0506, "step": 27871 }, { "epoch": 2.88, "learning_rate": 1.0326796403355785e-05, "loss": 0.2123, "step": 27872 }, { "epoch": 2.88, "learning_rate": 1.0324938201359387e-05, "loss": 0.6653, "step": 27873 }, { "epoch": 2.88, "learning_rate": 1.0323080131810906e-05, "loss": 0.9839, "step": 27874 }, { "epoch": 2.88, "learning_rate": 1.0321222194722843e-05, "loss": 0.1619, "step": 27875 }, { "epoch": 2.88, "learning_rate": 1.0319364390107707e-05, "loss": 0.228, "step": 27876 }, { "epoch": 2.88, "learning_rate": 1.0317506717978007e-05, "loss": 0.6625, "step": 27877 }, { "epoch": 2.88, "learning_rate": 1.0315649178346245e-05, "loss": 0.3594, "step": 27878 }, { "epoch": 2.89, "learning_rate": 1.0313791771224928e-05, "loss": 0.5017, "step": 27879 }, { "epoch": 2.89, "learning_rate": 1.0311934496626569e-05, "loss": 0.1176, "step": 27880 }, { "epoch": 2.89, "learning_rate": 1.0310077354563658e-05, "loss": 0.2931, "step": 27881 }, { "epoch": 2.89, "learning_rate": 1.0308220345048692e-05, "loss": 0.0463, "step": 27882 }, { "epoch": 2.89, "learning_rate": 1.0306363468094177e-05, "loss": 0.5064, "step": 27883 }, { "epoch": 2.89, "learning_rate": 1.0304506723712615e-05, "loss": 0.163, "step": 27884 }, { "epoch": 2.89, "learning_rate": 1.0302650111916505e-05, "loss": 0.5108, "step": 27885 }, { "epoch": 2.89, "learning_rate": 1.0300793632718349e-05, "loss": 0.4896, "step": 27886 }, { "epoch": 2.89, "learning_rate": 1.0298937286130637e-05, "loss": 0.5582, "step": 27887 }, { "epoch": 2.89, "learning_rate": 1.0297081072165872e-05, "loss": 0.4499, "step": 27888 }, { "epoch": 2.89, "learning_rate": 1.0295224990836546e-05, "loss": 0.232, "step": 27889 }, { "epoch": 2.89, "learning_rate": 1.0293369042155167e-05, "loss": 0.2278, "step": 27890 }, { "epoch": 2.89, "learning_rate": 1.0291513226134204e-05, "loss": 0.3299, "step": 27891 }, { "epoch": 2.89, "learning_rate": 1.0289657542786163e-05, "loss": 0.0867, "step": 27892 }, { "epoch": 2.89, "learning_rate": 1.0287801992123548e-05, "loss": 0.3901, "step": 27893 }, { "epoch": 2.89, "learning_rate": 1.0285946574158825e-05, "loss": 0.183, "step": 27894 }, { "epoch": 2.89, "learning_rate": 1.0284091288904499e-05, "loss": 0.5526, "step": 27895 }, { "epoch": 2.89, "learning_rate": 1.0282236136373055e-05, "loss": 0.6535, "step": 27896 }, { "epoch": 2.89, "learning_rate": 1.0280381116576984e-05, "loss": 0.6768, "step": 27897 }, { "epoch": 2.89, "learning_rate": 1.0278526229528776e-05, "loss": 0.3809, "step": 27898 }, { "epoch": 2.89, "learning_rate": 1.0276671475240912e-05, "loss": 0.2959, "step": 27899 }, { "epoch": 2.89, "learning_rate": 1.027481685372589e-05, "loss": 0.6266, "step": 27900 }, { "epoch": 2.89, "learning_rate": 1.0272962364996178e-05, "loss": 0.2901, "step": 27901 }, { "epoch": 2.89, "learning_rate": 1.0271108009064266e-05, "loss": 0.9065, "step": 27902 }, { "epoch": 2.89, "learning_rate": 1.0269253785942642e-05, "loss": 0.2043, "step": 27903 }, { "epoch": 2.89, "learning_rate": 1.026739969564378e-05, "loss": 0.208, "step": 27904 }, { "epoch": 2.89, "learning_rate": 1.0265545738180182e-05, "loss": 0.8784, "step": 27905 }, { "epoch": 2.89, "learning_rate": 1.0263691913564299e-05, "loss": 0.1334, "step": 27906 }, { "epoch": 2.89, "learning_rate": 1.0261838221808623e-05, "loss": 0.2232, "step": 27907 }, { "epoch": 2.89, "learning_rate": 1.0259984662925635e-05, "loss": 0.4616, "step": 27908 }, { "epoch": 2.89, "learning_rate": 1.0258131236927811e-05, "loss": 0.3223, "step": 27909 }, { "epoch": 2.89, "learning_rate": 1.025627794382764e-05, "loss": 0.1752, "step": 27910 }, { "epoch": 2.89, "learning_rate": 1.0254424783637575e-05, "loss": 0.7603, "step": 27911 }, { "epoch": 2.89, "learning_rate": 1.0252571756370103e-05, "loss": 0.079, "step": 27912 }, { "epoch": 2.89, "learning_rate": 1.0250718862037696e-05, "loss": 0.8224, "step": 27913 }, { "epoch": 2.89, "learning_rate": 1.024886610065283e-05, "loss": 0.3485, "step": 27914 }, { "epoch": 2.89, "learning_rate": 1.0247013472227977e-05, "loss": 0.5137, "step": 27915 }, { "epoch": 2.89, "learning_rate": 1.0245160976775603e-05, "loss": 0.4318, "step": 27916 }, { "epoch": 2.89, "learning_rate": 1.0243308614308186e-05, "loss": 0.2811, "step": 27917 }, { "epoch": 2.89, "learning_rate": 1.0241456384838206e-05, "loss": 0.0967, "step": 27918 }, { "epoch": 2.89, "learning_rate": 1.0239604288378105e-05, "loss": 0.4082, "step": 27919 }, { "epoch": 2.89, "learning_rate": 1.0237752324940374e-05, "loss": 0.3118, "step": 27920 }, { "epoch": 2.89, "learning_rate": 1.0235900494537459e-05, "loss": 0.4742, "step": 27921 }, { "epoch": 2.89, "learning_rate": 1.0234048797181842e-05, "loss": 0.3911, "step": 27922 }, { "epoch": 2.89, "learning_rate": 1.023219723288598e-05, "loss": 0.6279, "step": 27923 }, { "epoch": 2.89, "learning_rate": 1.0230345801662341e-05, "loss": 0.1265, "step": 27924 }, { "epoch": 2.89, "learning_rate": 1.022849450352339e-05, "loss": 0.1271, "step": 27925 }, { "epoch": 2.89, "learning_rate": 1.0226643338481586e-05, "loss": 1.0703, "step": 27926 }, { "epoch": 2.89, "learning_rate": 1.0224792306549395e-05, "loss": 0.2508, "step": 27927 }, { "epoch": 2.89, "learning_rate": 1.0222941407739284e-05, "loss": 0.3662, "step": 27928 }, { "epoch": 2.89, "learning_rate": 1.0221090642063692e-05, "loss": 0.4927, "step": 27929 }, { "epoch": 2.89, "learning_rate": 1.0219240009535104e-05, "loss": 0.6994, "step": 27930 }, { "epoch": 2.89, "learning_rate": 1.0217389510165948e-05, "loss": 0.5331, "step": 27931 }, { "epoch": 2.89, "learning_rate": 1.0215539143968703e-05, "loss": 0.2754, "step": 27932 }, { "epoch": 2.89, "learning_rate": 1.0213688910955816e-05, "loss": 0.2241, "step": 27933 }, { "epoch": 2.89, "learning_rate": 1.0211838811139748e-05, "loss": 0.3253, "step": 27934 }, { "epoch": 2.89, "learning_rate": 1.0209988844532947e-05, "loss": 0.1439, "step": 27935 }, { "epoch": 2.89, "learning_rate": 1.0208139011147875e-05, "loss": 0.2669, "step": 27936 }, { "epoch": 2.89, "learning_rate": 1.0206289310996979e-05, "loss": 0.1889, "step": 27937 }, { "epoch": 2.89, "learning_rate": 1.0204439744092722e-05, "loss": 0.4859, "step": 27938 }, { "epoch": 2.89, "learning_rate": 1.0202590310447533e-05, "loss": 0.1385, "step": 27939 }, { "epoch": 2.89, "learning_rate": 1.0200741010073874e-05, "loss": 0.7477, "step": 27940 }, { "epoch": 2.89, "learning_rate": 1.0198891842984196e-05, "loss": 0.6449, "step": 27941 }, { "epoch": 2.89, "learning_rate": 1.0197042809190945e-05, "loss": 0.3719, "step": 27942 }, { "epoch": 2.89, "learning_rate": 1.0195193908706574e-05, "loss": 0.2771, "step": 27943 }, { "epoch": 2.89, "learning_rate": 1.0193345141543518e-05, "loss": 0.7631, "step": 27944 }, { "epoch": 2.89, "learning_rate": 1.019149650771423e-05, "loss": 0.4548, "step": 27945 }, { "epoch": 2.89, "learning_rate": 1.0189648007231149e-05, "loss": 0.5011, "step": 27946 }, { "epoch": 2.89, "learning_rate": 1.0187799640106722e-05, "loss": 0.2086, "step": 27947 }, { "epoch": 2.89, "learning_rate": 1.0185951406353404e-05, "loss": 0.3509, "step": 27948 }, { "epoch": 2.89, "learning_rate": 1.0184103305983611e-05, "loss": 0.3609, "step": 27949 }, { "epoch": 2.89, "learning_rate": 1.0182255339009803e-05, "loss": 0.1992, "step": 27950 }, { "epoch": 2.89, "learning_rate": 1.0180407505444413e-05, "loss": 0.0883, "step": 27951 }, { "epoch": 2.89, "learning_rate": 1.0178559805299884e-05, "loss": 0.373, "step": 27952 }, { "epoch": 2.89, "learning_rate": 1.0176712238588652e-05, "loss": 0.1129, "step": 27953 }, { "epoch": 2.89, "learning_rate": 1.0174864805323152e-05, "loss": 0.7248, "step": 27954 }, { "epoch": 2.89, "learning_rate": 1.0173017505515836e-05, "loss": 0.4684, "step": 27955 }, { "epoch": 2.89, "learning_rate": 1.0171170339179117e-05, "loss": 0.2134, "step": 27956 }, { "epoch": 2.89, "learning_rate": 1.0169323306325438e-05, "loss": 0.4485, "step": 27957 }, { "epoch": 2.89, "learning_rate": 1.0167476406967245e-05, "loss": 0.724, "step": 27958 }, { "epoch": 2.89, "learning_rate": 1.0165629641116952e-05, "loss": 0.3814, "step": 27959 }, { "epoch": 2.89, "learning_rate": 1.0163783008786999e-05, "loss": 0.2768, "step": 27960 }, { "epoch": 2.89, "learning_rate": 1.0161936509989818e-05, "loss": 0.4686, "step": 27961 }, { "epoch": 2.89, "learning_rate": 1.0160090144737838e-05, "loss": 0.2034, "step": 27962 }, { "epoch": 2.89, "learning_rate": 1.0158243913043487e-05, "loss": 0.1726, "step": 27963 }, { "epoch": 2.89, "learning_rate": 1.0156397814919197e-05, "loss": 0.1458, "step": 27964 }, { "epoch": 2.89, "learning_rate": 1.0154551850377393e-05, "loss": 1.1333, "step": 27965 }, { "epoch": 2.89, "learning_rate": 1.0152706019430514e-05, "loss": 0.3827, "step": 27966 }, { "epoch": 2.89, "learning_rate": 1.0150860322090962e-05, "loss": 0.1297, "step": 27967 }, { "epoch": 2.89, "learning_rate": 1.0149014758371187e-05, "loss": 0.454, "step": 27968 }, { "epoch": 2.89, "learning_rate": 1.014716932828359e-05, "loss": 0.0979, "step": 27969 }, { "epoch": 2.89, "learning_rate": 1.0145324031840602e-05, "loss": 0.1832, "step": 27970 }, { "epoch": 2.89, "learning_rate": 1.0143478869054645e-05, "loss": 1.1283, "step": 27971 }, { "epoch": 2.89, "learning_rate": 1.0141633839938145e-05, "loss": 0.6498, "step": 27972 }, { "epoch": 2.89, "learning_rate": 1.013978894450352e-05, "loss": 1.0199, "step": 27973 }, { "epoch": 2.89, "learning_rate": 1.0137944182763186e-05, "loss": 0.461, "step": 27974 }, { "epoch": 2.9, "learning_rate": 1.0136099554729567e-05, "loss": 0.1441, "step": 27975 }, { "epoch": 2.9, "learning_rate": 1.0134255060415088e-05, "loss": 0.3781, "step": 27976 }, { "epoch": 2.9, "learning_rate": 1.0132410699832147e-05, "loss": 0.042, "step": 27977 }, { "epoch": 2.9, "learning_rate": 1.0130566472993163e-05, "loss": 0.7679, "step": 27978 }, { "epoch": 2.9, "learning_rate": 1.012872237991056e-05, "loss": 0.8024, "step": 27979 }, { "epoch": 2.9, "learning_rate": 1.0126878420596749e-05, "loss": 0.4569, "step": 27980 }, { "epoch": 2.9, "learning_rate": 1.012503459506415e-05, "loss": 0.2224, "step": 27981 }, { "epoch": 2.9, "learning_rate": 1.0123190903325162e-05, "loss": 0.2444, "step": 27982 }, { "epoch": 2.9, "learning_rate": 1.01213473453922e-05, "loss": 0.2076, "step": 27983 }, { "epoch": 2.9, "learning_rate": 1.0119503921277677e-05, "loss": 0.1848, "step": 27984 }, { "epoch": 2.9, "learning_rate": 1.0117660630994001e-05, "loss": 0.1326, "step": 27985 }, { "epoch": 2.9, "learning_rate": 1.0115817474553592e-05, "loss": 0.2859, "step": 27986 }, { "epoch": 2.9, "learning_rate": 1.0113974451968834e-05, "loss": 0.0447, "step": 27987 }, { "epoch": 2.9, "learning_rate": 1.0112131563252152e-05, "loss": 0.13, "step": 27988 }, { "epoch": 2.9, "learning_rate": 1.0110288808415945e-05, "loss": 0.3942, "step": 27989 }, { "epoch": 2.9, "learning_rate": 1.0108446187472619e-05, "loss": 0.3069, "step": 27990 }, { "epoch": 2.9, "learning_rate": 1.0106603700434579e-05, "loss": 0.5847, "step": 27991 }, { "epoch": 2.9, "learning_rate": 1.010476134731423e-05, "loss": 0.4861, "step": 27992 }, { "epoch": 2.9, "learning_rate": 1.010291912812398e-05, "loss": 0.3494, "step": 27993 }, { "epoch": 2.9, "learning_rate": 1.0101077042876216e-05, "loss": 0.2742, "step": 27994 }, { "epoch": 2.9, "learning_rate": 1.0099235091583344e-05, "loss": 0.3262, "step": 27995 }, { "epoch": 2.9, "learning_rate": 1.0097393274257774e-05, "loss": 0.8171, "step": 27996 }, { "epoch": 2.9, "learning_rate": 1.0095551590911886e-05, "loss": 0.2928, "step": 27997 }, { "epoch": 2.9, "learning_rate": 1.0093710041558088e-05, "loss": 0.5206, "step": 27998 }, { "epoch": 2.9, "learning_rate": 1.0091868626208776e-05, "loss": 0.1028, "step": 27999 }, { "epoch": 2.9, "learning_rate": 1.0090027344876345e-05, "loss": 0.2951, "step": 28000 }, { "epoch": 2.9, "learning_rate": 1.0088186197573194e-05, "loss": 0.6016, "step": 28001 }, { "epoch": 2.9, "learning_rate": 1.0086345184311713e-05, "loss": 0.9218, "step": 28002 }, { "epoch": 2.9, "learning_rate": 1.0084504305104296e-05, "loss": 0.1891, "step": 28003 }, { "epoch": 2.9, "learning_rate": 1.0082663559963349e-05, "loss": 0.6783, "step": 28004 }, { "epoch": 2.9, "learning_rate": 1.0080822948901238e-05, "loss": 0.1143, "step": 28005 }, { "epoch": 2.9, "learning_rate": 1.0078982471930377e-05, "loss": 0.1819, "step": 28006 }, { "epoch": 2.9, "learning_rate": 1.0077142129063136e-05, "loss": 0.5649, "step": 28007 }, { "epoch": 2.9, "learning_rate": 1.0075301920311908e-05, "loss": 0.298, "step": 28008 }, { "epoch": 2.9, "learning_rate": 1.007346184568909e-05, "loss": 0.3789, "step": 28009 }, { "epoch": 2.9, "learning_rate": 1.0071621905207063e-05, "loss": 0.5094, "step": 28010 }, { "epoch": 2.9, "learning_rate": 1.0069782098878214e-05, "loss": 0.2771, "step": 28011 }, { "epoch": 2.9, "learning_rate": 1.0067942426714932e-05, "loss": 0.1189, "step": 28012 }, { "epoch": 2.9, "learning_rate": 1.0066102888729593e-05, "loss": 0.1957, "step": 28013 }, { "epoch": 2.9, "learning_rate": 1.0064263484934599e-05, "loss": 0.3889, "step": 28014 }, { "epoch": 2.9, "learning_rate": 1.0062424215342305e-05, "loss": 0.3559, "step": 28015 }, { "epoch": 2.9, "learning_rate": 1.0060585079965112e-05, "loss": 0.3797, "step": 28016 }, { "epoch": 2.9, "learning_rate": 1.0058746078815389e-05, "loss": 0.2584, "step": 28017 }, { "epoch": 2.9, "learning_rate": 1.0056907211905534e-05, "loss": 0.2802, "step": 28018 }, { "epoch": 2.9, "learning_rate": 1.0055068479247901e-05, "loss": 0.1847, "step": 28019 }, { "epoch": 2.9, "learning_rate": 1.0053229880854885e-05, "loss": 0.5101, "step": 28020 }, { "epoch": 2.9, "learning_rate": 1.0051391416738854e-05, "loss": 0.6217, "step": 28021 }, { "epoch": 2.9, "learning_rate": 1.0049553086912188e-05, "loss": 0.1083, "step": 28022 }, { "epoch": 2.9, "learning_rate": 1.0047714891387264e-05, "loss": 0.4202, "step": 28023 }, { "epoch": 2.9, "learning_rate": 1.0045876830176466e-05, "loss": 0.3097, "step": 28024 }, { "epoch": 2.9, "learning_rate": 1.0044038903292147e-05, "loss": 0.7354, "step": 28025 }, { "epoch": 2.9, "learning_rate": 1.0042201110746686e-05, "loss": 0.2911, "step": 28026 }, { "epoch": 2.9, "learning_rate": 1.0040363452552458e-05, "loss": 0.1773, "step": 28027 }, { "epoch": 2.9, "learning_rate": 1.0038525928721834e-05, "loss": 0.4854, "step": 28028 }, { "epoch": 2.9, "learning_rate": 1.0036688539267182e-05, "loss": 0.2843, "step": 28029 }, { "epoch": 2.9, "learning_rate": 1.0034851284200876e-05, "loss": 0.389, "step": 28030 }, { "epoch": 2.9, "learning_rate": 1.0033014163535283e-05, "loss": 0.5599, "step": 28031 }, { "epoch": 2.9, "learning_rate": 1.003117717728276e-05, "loss": 0.2056, "step": 28032 }, { "epoch": 2.9, "learning_rate": 1.0029340325455681e-05, "loss": 0.4277, "step": 28033 }, { "epoch": 2.9, "learning_rate": 1.0027503608066418e-05, "loss": 0.1227, "step": 28034 }, { "epoch": 2.9, "learning_rate": 1.0025667025127318e-05, "loss": 0.6936, "step": 28035 }, { "epoch": 2.9, "learning_rate": 1.0023830576650755e-05, "loss": 0.5788, "step": 28036 }, { "epoch": 2.9, "learning_rate": 1.0021994262649088e-05, "loss": 0.612, "step": 28037 }, { "epoch": 2.9, "learning_rate": 1.0020158083134686e-05, "loss": 0.2045, "step": 28038 }, { "epoch": 2.9, "learning_rate": 1.00183220381199e-05, "loss": 0.0755, "step": 28039 }, { "epoch": 2.9, "learning_rate": 1.0016486127617097e-05, "loss": 0.7349, "step": 28040 }, { "epoch": 2.9, "learning_rate": 1.0014650351638632e-05, "loss": 0.2536, "step": 28041 }, { "epoch": 2.9, "learning_rate": 1.0012814710196875e-05, "loss": 0.5269, "step": 28042 }, { "epoch": 2.9, "learning_rate": 1.0010979203304164e-05, "loss": 0.5592, "step": 28043 }, { "epoch": 2.9, "learning_rate": 1.0009143830972875e-05, "loss": 0.487, "step": 28044 }, { "epoch": 2.9, "learning_rate": 1.0007308593215339e-05, "loss": 0.1955, "step": 28045 }, { "epoch": 2.9, "learning_rate": 1.0005473490043927e-05, "loss": 1.0707, "step": 28046 }, { "epoch": 2.9, "learning_rate": 1.0003638521470988e-05, "loss": 0.2736, "step": 28047 }, { "epoch": 2.9, "learning_rate": 1.0001803687508875e-05, "loss": 0.2127, "step": 28048 }, { "epoch": 2.9, "learning_rate": 9.99996898816994e-06, "loss": 0.6035, "step": 28049 }, { "epoch": 2.9, "learning_rate": 9.998134423466532e-06, "loss": 0.1637, "step": 28050 }, { "epoch": 2.9, "learning_rate": 9.996299993411008e-06, "loss": 0.0667, "step": 28051 }, { "epoch": 2.9, "learning_rate": 9.99446569801572e-06, "loss": 0.2283, "step": 28052 }, { "epoch": 2.9, "learning_rate": 9.992631537292997e-06, "loss": 0.2941, "step": 28053 }, { "epoch": 2.9, "learning_rate": 9.990797511255197e-06, "loss": 0.2779, "step": 28054 }, { "epoch": 2.9, "learning_rate": 9.988963619914668e-06, "loss": 0.2665, "step": 28055 }, { "epoch": 2.9, "learning_rate": 9.987129863283763e-06, "loss": 0.1682, "step": 28056 }, { "epoch": 2.9, "learning_rate": 9.985296241374808e-06, "loss": 1.1164, "step": 28057 }, { "epoch": 2.9, "learning_rate": 9.983462754200152e-06, "loss": 0.188, "step": 28058 }, { "epoch": 2.9, "learning_rate": 9.981629401772144e-06, "loss": 0.0763, "step": 28059 }, { "epoch": 2.9, "learning_rate": 9.979796184103124e-06, "loss": 0.3545, "step": 28060 }, { "epoch": 2.9, "learning_rate": 9.97796310120543e-06, "loss": 0.1718, "step": 28061 }, { "epoch": 2.9, "learning_rate": 9.976130153091415e-06, "loss": 0.2516, "step": 28062 }, { "epoch": 2.9, "learning_rate": 9.974297339773395e-06, "loss": 0.5927, "step": 28063 }, { "epoch": 2.9, "learning_rate": 9.97246466126372e-06, "loss": 0.1802, "step": 28064 }, { "epoch": 2.9, "learning_rate": 9.970632117574729e-06, "loss": 0.7319, "step": 28065 }, { "epoch": 2.9, "learning_rate": 9.968799708718754e-06, "loss": 0.4261, "step": 28066 }, { "epoch": 2.9, "learning_rate": 9.966967434708131e-06, "loss": 0.1088, "step": 28067 }, { "epoch": 2.9, "learning_rate": 9.965135295555208e-06, "loss": 0.1945, "step": 28068 }, { "epoch": 2.9, "learning_rate": 9.963303291272296e-06, "loss": 0.5309, "step": 28069 }, { "epoch": 2.9, "learning_rate": 9.961471421871739e-06, "loss": 0.5758, "step": 28070 }, { "epoch": 2.91, "learning_rate": 9.959639687365867e-06, "loss": 0.2749, "step": 28071 }, { "epoch": 2.91, "learning_rate": 9.957808087767021e-06, "loss": 0.5399, "step": 28072 }, { "epoch": 2.91, "learning_rate": 9.955976623087511e-06, "loss": 0.4573, "step": 28073 }, { "epoch": 2.91, "learning_rate": 9.954145293339676e-06, "loss": 0.4215, "step": 28074 }, { "epoch": 2.91, "learning_rate": 9.952314098535847e-06, "loss": 0.3724, "step": 28075 }, { "epoch": 2.91, "learning_rate": 9.950483038688348e-06, "loss": 0.2033, "step": 28076 }, { "epoch": 2.91, "learning_rate": 9.948652113809506e-06, "loss": 0.2044, "step": 28077 }, { "epoch": 2.91, "learning_rate": 9.946821323911645e-06, "loss": 0.3733, "step": 28078 }, { "epoch": 2.91, "learning_rate": 9.944990669007093e-06, "loss": 0.1135, "step": 28079 }, { "epoch": 2.91, "learning_rate": 9.943160149108179e-06, "loss": 0.1242, "step": 28080 }, { "epoch": 2.91, "learning_rate": 9.94132976422721e-06, "loss": 0.475, "step": 28081 }, { "epoch": 2.91, "learning_rate": 9.939499514376523e-06, "loss": 1.2504, "step": 28082 }, { "epoch": 2.91, "learning_rate": 9.937669399568421e-06, "loss": 0.0967, "step": 28083 }, { "epoch": 2.91, "learning_rate": 9.935839419815238e-06, "loss": 0.4786, "step": 28084 }, { "epoch": 2.91, "learning_rate": 9.934009575129288e-06, "loss": 0.5054, "step": 28085 }, { "epoch": 2.91, "learning_rate": 9.932179865522891e-06, "loss": 0.0789, "step": 28086 }, { "epoch": 2.91, "learning_rate": 9.930350291008363e-06, "loss": 0.4413, "step": 28087 }, { "epoch": 2.91, "learning_rate": 9.92852085159802e-06, "loss": 0.6489, "step": 28088 }, { "epoch": 2.91, "learning_rate": 9.926691547304177e-06, "loss": 0.4685, "step": 28089 }, { "epoch": 2.91, "learning_rate": 9.924862378139162e-06, "loss": 0.2015, "step": 28090 }, { "epoch": 2.91, "learning_rate": 9.923033344115268e-06, "loss": 0.2059, "step": 28091 }, { "epoch": 2.91, "learning_rate": 9.921204445244814e-06, "loss": 0.1565, "step": 28092 }, { "epoch": 2.91, "learning_rate": 9.919375681540114e-06, "loss": 0.3074, "step": 28093 }, { "epoch": 2.91, "learning_rate": 9.917547053013486e-06, "loss": 0.377, "step": 28094 }, { "epoch": 2.91, "learning_rate": 9.915718559677223e-06, "loss": 0.648, "step": 28095 }, { "epoch": 2.91, "learning_rate": 9.913890201543647e-06, "loss": 0.2808, "step": 28096 }, { "epoch": 2.91, "learning_rate": 9.91206197862506e-06, "loss": 0.2027, "step": 28097 }, { "epoch": 2.91, "learning_rate": 9.910233890933771e-06, "loss": 0.3558, "step": 28098 }, { "epoch": 2.91, "learning_rate": 9.908405938482087e-06, "loss": 0.3061, "step": 28099 }, { "epoch": 2.91, "learning_rate": 9.906578121282325e-06, "loss": 0.6999, "step": 28100 }, { "epoch": 2.91, "learning_rate": 9.904750439346766e-06, "loss": 0.7972, "step": 28101 }, { "epoch": 2.91, "learning_rate": 9.902922892687723e-06, "loss": 0.6748, "step": 28102 }, { "epoch": 2.91, "learning_rate": 9.901095481317504e-06, "loss": 0.3332, "step": 28103 }, { "epoch": 2.91, "learning_rate": 9.899268205248407e-06, "loss": 0.6003, "step": 28104 }, { "epoch": 2.91, "learning_rate": 9.897441064492731e-06, "loss": 0.0571, "step": 28105 }, { "epoch": 2.91, "learning_rate": 9.895614059062791e-06, "loss": 0.1216, "step": 28106 }, { "epoch": 2.91, "learning_rate": 9.893787188970865e-06, "loss": 0.068, "step": 28107 }, { "epoch": 2.91, "learning_rate": 9.891960454229255e-06, "loss": 0.559, "step": 28108 }, { "epoch": 2.91, "learning_rate": 9.890133854850264e-06, "loss": 0.1124, "step": 28109 }, { "epoch": 2.91, "learning_rate": 9.888307390846193e-06, "loss": 0.3126, "step": 28110 }, { "epoch": 2.91, "learning_rate": 9.886481062229326e-06, "loss": 0.0599, "step": 28111 }, { "epoch": 2.91, "learning_rate": 9.88465486901196e-06, "loss": 0.0432, "step": 28112 }, { "epoch": 2.91, "learning_rate": 9.88282881120639e-06, "loss": 0.5243, "step": 28113 }, { "epoch": 2.91, "learning_rate": 9.88100288882491e-06, "loss": 0.789, "step": 28114 }, { "epoch": 2.91, "learning_rate": 9.879177101879812e-06, "loss": 0.2182, "step": 28115 }, { "epoch": 2.91, "learning_rate": 9.877351450383387e-06, "loss": 0.3481, "step": 28116 }, { "epoch": 2.91, "learning_rate": 9.875525934347921e-06, "loss": 0.325, "step": 28117 }, { "epoch": 2.91, "learning_rate": 9.873700553785718e-06, "loss": 0.2847, "step": 28118 }, { "epoch": 2.91, "learning_rate": 9.87187530870904e-06, "loss": 0.3393, "step": 28119 }, { "epoch": 2.91, "learning_rate": 9.870050199130201e-06, "loss": 0.075, "step": 28120 }, { "epoch": 2.91, "learning_rate": 9.868225225061463e-06, "loss": 0.5589, "step": 28121 }, { "epoch": 2.91, "learning_rate": 9.866400386515123e-06, "loss": 0.7977, "step": 28122 }, { "epoch": 2.91, "learning_rate": 9.864575683503464e-06, "loss": 0.6679, "step": 28123 }, { "epoch": 2.91, "learning_rate": 9.86275111603877e-06, "loss": 0.1678, "step": 28124 }, { "epoch": 2.91, "learning_rate": 9.860926684133326e-06, "loss": 0.0651, "step": 28125 }, { "epoch": 2.91, "learning_rate": 9.859102387799409e-06, "loss": 0.1298, "step": 28126 }, { "epoch": 2.91, "learning_rate": 9.857278227049304e-06, "loss": 0.8119, "step": 28127 }, { "epoch": 2.91, "learning_rate": 9.855454201895295e-06, "loss": 0.6781, "step": 28128 }, { "epoch": 2.91, "learning_rate": 9.853630312349649e-06, "loss": 0.8471, "step": 28129 }, { "epoch": 2.91, "learning_rate": 9.851806558424647e-06, "loss": 0.149, "step": 28130 }, { "epoch": 2.91, "learning_rate": 9.849982940132583e-06, "loss": 0.3197, "step": 28131 }, { "epoch": 2.91, "learning_rate": 9.848159457485703e-06, "loss": 0.4654, "step": 28132 }, { "epoch": 2.91, "learning_rate": 9.846336110496303e-06, "loss": 0.3734, "step": 28133 }, { "epoch": 2.91, "learning_rate": 9.844512899176653e-06, "loss": 0.2554, "step": 28134 }, { "epoch": 2.91, "learning_rate": 9.842689823539023e-06, "loss": 0.1978, "step": 28135 }, { "epoch": 2.91, "learning_rate": 9.840866883595693e-06, "loss": 0.14, "step": 28136 }, { "epoch": 2.91, "learning_rate": 9.839044079358926e-06, "loss": 0.932, "step": 28137 }, { "epoch": 2.91, "learning_rate": 9.837221410841007e-06, "loss": 0.4807, "step": 28138 }, { "epoch": 2.91, "learning_rate": 9.83539887805419e-06, "loss": 0.4187, "step": 28139 }, { "epoch": 2.91, "learning_rate": 9.833576481010747e-06, "loss": 0.1301, "step": 28140 }, { "epoch": 2.91, "learning_rate": 9.831754219722946e-06, "loss": 0.0833, "step": 28141 }, { "epoch": 2.91, "learning_rate": 9.829932094203062e-06, "loss": 0.5744, "step": 28142 }, { "epoch": 2.91, "learning_rate": 9.828110104463351e-06, "loss": 0.3516, "step": 28143 }, { "epoch": 2.91, "learning_rate": 9.826288250516096e-06, "loss": 0.514, "step": 28144 }, { "epoch": 2.91, "learning_rate": 9.824466532373537e-06, "loss": 0.124, "step": 28145 }, { "epoch": 2.91, "learning_rate": 9.82264495004795e-06, "loss": 0.4492, "step": 28146 }, { "epoch": 2.91, "learning_rate": 9.82082350355159e-06, "loss": 0.4641, "step": 28147 }, { "epoch": 2.91, "learning_rate": 9.819002192896742e-06, "loss": 0.1851, "step": 28148 }, { "epoch": 2.91, "learning_rate": 9.817181018095639e-06, "loss": 0.3602, "step": 28149 }, { "epoch": 2.91, "learning_rate": 9.815359979160545e-06, "loss": 0.2318, "step": 28150 }, { "epoch": 2.91, "learning_rate": 9.813539076103732e-06, "loss": 0.08, "step": 28151 }, { "epoch": 2.91, "learning_rate": 9.811718308937446e-06, "loss": 0.9804, "step": 28152 }, { "epoch": 2.91, "learning_rate": 9.80989767767395e-06, "loss": 0.2156, "step": 28153 }, { "epoch": 2.91, "learning_rate": 9.808077182325496e-06, "loss": 0.0986, "step": 28154 }, { "epoch": 2.91, "learning_rate": 9.806256822904346e-06, "loss": 0.4093, "step": 28155 }, { "epoch": 2.91, "learning_rate": 9.804436599422758e-06, "loss": 0.1891, "step": 28156 }, { "epoch": 2.91, "learning_rate": 9.802616511892966e-06, "loss": 0.5419, "step": 28157 }, { "epoch": 2.91, "learning_rate": 9.800796560327248e-06, "loss": 0.5914, "step": 28158 }, { "epoch": 2.91, "learning_rate": 9.798976744737827e-06, "loss": 0.1117, "step": 28159 }, { "epoch": 2.91, "learning_rate": 9.79715706513697e-06, "loss": 0.4053, "step": 28160 }, { "epoch": 2.91, "learning_rate": 9.795337521536928e-06, "loss": 0.7397, "step": 28161 }, { "epoch": 2.91, "learning_rate": 9.793518113949941e-06, "loss": 0.1573, "step": 28162 }, { "epoch": 2.91, "learning_rate": 9.791698842388262e-06, "loss": 0.2232, "step": 28163 }, { "epoch": 2.91, "learning_rate": 9.789879706864143e-06, "loss": 0.6947, "step": 28164 }, { "epoch": 2.91, "learning_rate": 9.78806070738982e-06, "loss": 0.0937, "step": 28165 }, { "epoch": 2.91, "learning_rate": 9.786241843977555e-06, "loss": 0.6971, "step": 28166 }, { "epoch": 2.92, "learning_rate": 9.784423116639571e-06, "loss": 0.3259, "step": 28167 }, { "epoch": 2.92, "learning_rate": 9.782604525388121e-06, "loss": 0.2435, "step": 28168 }, { "epoch": 2.92, "learning_rate": 9.780786070235455e-06, "loss": 0.2061, "step": 28169 }, { "epoch": 2.92, "learning_rate": 9.778967751193797e-06, "loss": 0.4199, "step": 28170 }, { "epoch": 2.92, "learning_rate": 9.777149568275398e-06, "loss": 0.3253, "step": 28171 }, { "epoch": 2.92, "learning_rate": 9.775331521492494e-06, "loss": 0.134, "step": 28172 }, { "epoch": 2.92, "learning_rate": 9.773513610857328e-06, "loss": 0.7654, "step": 28173 }, { "epoch": 2.92, "learning_rate": 9.771695836382134e-06, "loss": 0.5278, "step": 28174 }, { "epoch": 2.92, "learning_rate": 9.76987819807915e-06, "loss": 0.1249, "step": 28175 }, { "epoch": 2.92, "learning_rate": 9.768060695960624e-06, "loss": 0.1955, "step": 28176 }, { "epoch": 2.92, "learning_rate": 9.766243330038769e-06, "loss": 0.4854, "step": 28177 }, { "epoch": 2.92, "learning_rate": 9.76442610032583e-06, "loss": 0.5656, "step": 28178 }, { "epoch": 2.92, "learning_rate": 9.76260900683404e-06, "loss": 0.3926, "step": 28179 }, { "epoch": 2.92, "learning_rate": 9.76079204957563e-06, "loss": 0.6372, "step": 28180 }, { "epoch": 2.92, "learning_rate": 9.758975228562833e-06, "loss": 0.8031, "step": 28181 }, { "epoch": 2.92, "learning_rate": 9.75715854380789e-06, "loss": 0.7008, "step": 28182 }, { "epoch": 2.92, "learning_rate": 9.755341995323009e-06, "loss": 0.3955, "step": 28183 }, { "epoch": 2.92, "learning_rate": 9.753525583120428e-06, "loss": 0.6133, "step": 28184 }, { "epoch": 2.92, "learning_rate": 9.751709307212375e-06, "loss": 0.3228, "step": 28185 }, { "epoch": 2.92, "learning_rate": 9.749893167611087e-06, "loss": 0.3056, "step": 28186 }, { "epoch": 2.92, "learning_rate": 9.748077164328769e-06, "loss": 0.1958, "step": 28187 }, { "epoch": 2.92, "learning_rate": 9.74626129737766e-06, "loss": 0.3715, "step": 28188 }, { "epoch": 2.92, "learning_rate": 9.74444556676998e-06, "loss": 0.7097, "step": 28189 }, { "epoch": 2.92, "learning_rate": 9.742629972517951e-06, "loss": 0.5614, "step": 28190 }, { "epoch": 2.92, "learning_rate": 9.7408145146338e-06, "loss": 0.0684, "step": 28191 }, { "epoch": 2.92, "learning_rate": 9.738999193129743e-06, "loss": 0.5118, "step": 28192 }, { "epoch": 2.92, "learning_rate": 9.737184008018004e-06, "loss": 0.7225, "step": 28193 }, { "epoch": 2.92, "learning_rate": 9.735368959310811e-06, "loss": 0.2859, "step": 28194 }, { "epoch": 2.92, "learning_rate": 9.733554047020363e-06, "loss": 0.0813, "step": 28195 }, { "epoch": 2.92, "learning_rate": 9.731739271158898e-06, "loss": 0.6393, "step": 28196 }, { "epoch": 2.92, "learning_rate": 9.72992463173861e-06, "loss": 0.4786, "step": 28197 }, { "epoch": 2.92, "learning_rate": 9.728110128771727e-06, "loss": 0.0731, "step": 28198 }, { "epoch": 2.92, "learning_rate": 9.726295762270464e-06, "loss": 0.5274, "step": 28199 }, { "epoch": 2.92, "learning_rate": 9.724481532247031e-06, "loss": 0.2307, "step": 28200 }, { "epoch": 2.92, "learning_rate": 9.72266743871365e-06, "loss": 0.1507, "step": 28201 }, { "epoch": 2.92, "learning_rate": 9.720853481682523e-06, "loss": 0.5315, "step": 28202 }, { "epoch": 2.92, "learning_rate": 9.719039661165868e-06, "loss": 0.1803, "step": 28203 }, { "epoch": 2.92, "learning_rate": 9.717225977175901e-06, "loss": 0.5725, "step": 28204 }, { "epoch": 2.92, "learning_rate": 9.715412429724811e-06, "loss": 0.4734, "step": 28205 }, { "epoch": 2.92, "learning_rate": 9.713599018824823e-06, "loss": 0.6937, "step": 28206 }, { "epoch": 2.92, "learning_rate": 9.711785744488145e-06, "loss": 0.2274, "step": 28207 }, { "epoch": 2.92, "learning_rate": 9.70997260672697e-06, "loss": 0.3027, "step": 28208 }, { "epoch": 2.92, "learning_rate": 9.708159605553514e-06, "loss": 0.0327, "step": 28209 }, { "epoch": 2.92, "learning_rate": 9.706346740979978e-06, "loss": 0.3869, "step": 28210 }, { "epoch": 2.92, "learning_rate": 9.704534013018569e-06, "loss": 0.1469, "step": 28211 }, { "epoch": 2.92, "learning_rate": 9.702721421681487e-06, "loss": 1.1562, "step": 28212 }, { "epoch": 2.92, "learning_rate": 9.700908966980935e-06, "loss": 0.3025, "step": 28213 }, { "epoch": 2.92, "learning_rate": 9.699096648929125e-06, "loss": 0.5776, "step": 28214 }, { "epoch": 2.92, "learning_rate": 9.697284467538234e-06, "loss": 0.2124, "step": 28215 }, { "epoch": 2.92, "learning_rate": 9.695472422820476e-06, "loss": 0.4549, "step": 28216 }, { "epoch": 2.92, "learning_rate": 9.693660514788047e-06, "loss": 0.8087, "step": 28217 }, { "epoch": 2.92, "learning_rate": 9.691848743453141e-06, "loss": 0.5179, "step": 28218 }, { "epoch": 2.92, "learning_rate": 9.690037108827969e-06, "loss": 0.759, "step": 28219 }, { "epoch": 2.92, "learning_rate": 9.688225610924705e-06, "loss": 0.088, "step": 28220 }, { "epoch": 2.92, "learning_rate": 9.686414249755555e-06, "loss": 0.5601, "step": 28221 }, { "epoch": 2.92, "learning_rate": 9.684603025332711e-06, "loss": 0.3391, "step": 28222 }, { "epoch": 2.92, "learning_rate": 9.682791937668366e-06, "loss": 0.1746, "step": 28223 }, { "epoch": 2.92, "learning_rate": 9.68098098677472e-06, "loss": 0.1807, "step": 28224 }, { "epoch": 2.92, "learning_rate": 9.679170172663946e-06, "loss": 0.1961, "step": 28225 }, { "epoch": 2.92, "learning_rate": 9.677359495348247e-06, "loss": 0.3149, "step": 28226 }, { "epoch": 2.92, "learning_rate": 9.675548954839806e-06, "loss": 0.0445, "step": 28227 }, { "epoch": 2.92, "learning_rate": 9.673738551150814e-06, "loss": 0.7616, "step": 28228 }, { "epoch": 2.92, "learning_rate": 9.671928284293457e-06, "loss": 0.4117, "step": 28229 }, { "epoch": 2.92, "learning_rate": 9.670118154279924e-06, "loss": 0.0909, "step": 28230 }, { "epoch": 2.92, "learning_rate": 9.668308161122401e-06, "loss": 0.125, "step": 28231 }, { "epoch": 2.92, "learning_rate": 9.666498304833078e-06, "loss": 0.1618, "step": 28232 }, { "epoch": 2.92, "learning_rate": 9.664688585424119e-06, "loss": 0.1142, "step": 28233 }, { "epoch": 2.92, "learning_rate": 9.66287900290773e-06, "loss": 0.3194, "step": 28234 }, { "epoch": 2.92, "learning_rate": 9.661069557296071e-06, "loss": 0.0779, "step": 28235 }, { "epoch": 2.92, "learning_rate": 9.659260248601336e-06, "loss": 0.2182, "step": 28236 }, { "epoch": 2.92, "learning_rate": 9.657451076835697e-06, "loss": 0.1266, "step": 28237 }, { "epoch": 2.92, "learning_rate": 9.65564204201134e-06, "loss": 0.4865, "step": 28238 }, { "epoch": 2.92, "learning_rate": 9.653833144140438e-06, "loss": 0.2013, "step": 28239 }, { "epoch": 2.92, "learning_rate": 9.652024383235174e-06, "loss": 0.2085, "step": 28240 }, { "epoch": 2.92, "learning_rate": 9.650215759307717e-06, "loss": 0.075, "step": 28241 }, { "epoch": 2.92, "learning_rate": 9.64840727237026e-06, "loss": 0.3517, "step": 28242 }, { "epoch": 2.92, "learning_rate": 9.64659892243495e-06, "loss": 0.2725, "step": 28243 }, { "epoch": 2.92, "learning_rate": 9.644790709513986e-06, "loss": 0.144, "step": 28244 }, { "epoch": 2.92, "learning_rate": 9.642982633619516e-06, "loss": 0.1452, "step": 28245 }, { "epoch": 2.92, "learning_rate": 9.641174694763726e-06, "loss": 0.1174, "step": 28246 }, { "epoch": 2.92, "learning_rate": 9.639366892958782e-06, "loss": 0.5477, "step": 28247 }, { "epoch": 2.92, "learning_rate": 9.63755922821686e-06, "loss": 0.1402, "step": 28248 }, { "epoch": 2.92, "learning_rate": 9.63575170055012e-06, "loss": 0.3862, "step": 28249 }, { "epoch": 2.92, "learning_rate": 9.63394430997074e-06, "loss": 0.6739, "step": 28250 }, { "epoch": 2.92, "learning_rate": 9.632137056490879e-06, "loss": 0.2876, "step": 28251 }, { "epoch": 2.92, "learning_rate": 9.630329940122714e-06, "loss": 0.0991, "step": 28252 }, { "epoch": 2.92, "learning_rate": 9.628522960878395e-06, "loss": 0.4164, "step": 28253 }, { "epoch": 2.92, "learning_rate": 9.626716118770088e-06, "loss": 0.5267, "step": 28254 }, { "epoch": 2.92, "learning_rate": 9.624909413809967e-06, "loss": 0.6223, "step": 28255 }, { "epoch": 2.92, "learning_rate": 9.623102846010184e-06, "loss": 0.0729, "step": 28256 }, { "epoch": 2.92, "learning_rate": 9.621296415382916e-06, "loss": 0.2647, "step": 28257 }, { "epoch": 2.92, "learning_rate": 9.619490121940302e-06, "loss": 0.3859, "step": 28258 }, { "epoch": 2.92, "learning_rate": 9.617683965694514e-06, "loss": 0.1223, "step": 28259 }, { "epoch": 2.92, "learning_rate": 9.615877946657707e-06, "loss": 0.0728, "step": 28260 }, { "epoch": 2.92, "learning_rate": 9.61407206484204e-06, "loss": 0.6354, "step": 28261 }, { "epoch": 2.93, "learning_rate": 9.61226632025968e-06, "loss": 0.341, "step": 28262 }, { "epoch": 2.93, "learning_rate": 9.610460712922762e-06, "loss": 0.4139, "step": 28263 }, { "epoch": 2.93, "learning_rate": 9.608655242843455e-06, "loss": 0.1016, "step": 28264 }, { "epoch": 2.93, "learning_rate": 9.606849910033906e-06, "loss": 0.3018, "step": 28265 }, { "epoch": 2.93, "learning_rate": 9.605044714506273e-06, "loss": 0.2277, "step": 28266 }, { "epoch": 2.93, "learning_rate": 9.60323965627271e-06, "loss": 0.1091, "step": 28267 }, { "epoch": 2.93, "learning_rate": 9.601434735345362e-06, "loss": 0.5949, "step": 28268 }, { "epoch": 2.93, "learning_rate": 9.59962995173639e-06, "loss": 0.5785, "step": 28269 }, { "epoch": 2.93, "learning_rate": 9.59782530545793e-06, "loss": 0.0863, "step": 28270 }, { "epoch": 2.93, "learning_rate": 9.596020796522134e-06, "loss": 0.488, "step": 28271 }, { "epoch": 2.93, "learning_rate": 9.594216424941162e-06, "loss": 0.5154, "step": 28272 }, { "epoch": 2.93, "learning_rate": 9.59241219072714e-06, "loss": 0.8329, "step": 28273 }, { "epoch": 2.93, "learning_rate": 9.590608093892226e-06, "loss": 0.2828, "step": 28274 }, { "epoch": 2.93, "learning_rate": 9.588804134448562e-06, "loss": 0.0269, "step": 28275 }, { "epoch": 2.93, "learning_rate": 9.587000312408292e-06, "loss": 0.1317, "step": 28276 }, { "epoch": 2.93, "learning_rate": 9.585196627783563e-06, "loss": 0.1266, "step": 28277 }, { "epoch": 2.93, "learning_rate": 9.583393080586511e-06, "loss": 0.128, "step": 28278 }, { "epoch": 2.93, "learning_rate": 9.581589670829279e-06, "loss": 0.2299, "step": 28279 }, { "epoch": 2.93, "learning_rate": 9.57978639852402e-06, "loss": 0.1866, "step": 28280 }, { "epoch": 2.93, "learning_rate": 9.57798326368285e-06, "loss": 0.4307, "step": 28281 }, { "epoch": 2.93, "learning_rate": 9.576180266317928e-06, "loss": 0.3874, "step": 28282 }, { "epoch": 2.93, "learning_rate": 9.57437740644137e-06, "loss": 0.507, "step": 28283 }, { "epoch": 2.93, "learning_rate": 9.572574684065327e-06, "loss": 0.1101, "step": 28284 }, { "epoch": 2.93, "learning_rate": 9.570772099201929e-06, "loss": 0.0483, "step": 28285 }, { "epoch": 2.93, "learning_rate": 9.568969651863315e-06, "loss": 0.2569, "step": 28286 }, { "epoch": 2.93, "learning_rate": 9.567167342061617e-06, "loss": 0.2261, "step": 28287 }, { "epoch": 2.93, "learning_rate": 9.565365169808967e-06, "loss": 0.2068, "step": 28288 }, { "epoch": 2.93, "learning_rate": 9.563563135117496e-06, "loss": 0.1238, "step": 28289 }, { "epoch": 2.93, "learning_rate": 9.561761237999349e-06, "loss": 0.1789, "step": 28290 }, { "epoch": 2.93, "learning_rate": 9.559959478466631e-06, "loss": 0.109, "step": 28291 }, { "epoch": 2.93, "learning_rate": 9.558157856531483e-06, "loss": 0.1527, "step": 28292 }, { "epoch": 2.93, "learning_rate": 9.556356372206035e-06, "loss": 0.2677, "step": 28293 }, { "epoch": 2.93, "learning_rate": 9.554555025502412e-06, "loss": 0.5216, "step": 28294 }, { "epoch": 2.93, "learning_rate": 9.55275381643275e-06, "loss": 0.2338, "step": 28295 }, { "epoch": 2.93, "learning_rate": 9.550952745009158e-06, "loss": 0.6367, "step": 28296 }, { "epoch": 2.93, "learning_rate": 9.549151811243763e-06, "loss": 0.0349, "step": 28297 }, { "epoch": 2.93, "learning_rate": 9.547351015148695e-06, "loss": 0.5796, "step": 28298 }, { "epoch": 2.93, "learning_rate": 9.545550356736073e-06, "loss": 0.9023, "step": 28299 }, { "epoch": 2.93, "learning_rate": 9.543749836018031e-06, "loss": 0.7264, "step": 28300 }, { "epoch": 2.93, "learning_rate": 9.541949453006671e-06, "loss": 0.3987, "step": 28301 }, { "epoch": 2.93, "learning_rate": 9.540149207714116e-06, "loss": 0.2377, "step": 28302 }, { "epoch": 2.93, "learning_rate": 9.538349100152492e-06, "loss": 0.6967, "step": 28303 }, { "epoch": 2.93, "learning_rate": 9.536549130333913e-06, "loss": 0.1502, "step": 28304 }, { "epoch": 2.93, "learning_rate": 9.5347492982705e-06, "loss": 0.3601, "step": 28305 }, { "epoch": 2.93, "learning_rate": 9.532949603974363e-06, "loss": 0.4309, "step": 28306 }, { "epoch": 2.93, "learning_rate": 9.531150047457632e-06, "loss": 0.3721, "step": 28307 }, { "epoch": 2.93, "learning_rate": 9.529350628732398e-06, "loss": 0.7246, "step": 28308 }, { "epoch": 2.93, "learning_rate": 9.52755134781079e-06, "loss": 0.2308, "step": 28309 }, { "epoch": 2.93, "learning_rate": 9.52575220470492e-06, "loss": 0.1586, "step": 28310 }, { "epoch": 2.93, "learning_rate": 9.523953199426891e-06, "loss": 0.7017, "step": 28311 }, { "epoch": 2.93, "learning_rate": 9.522154331988817e-06, "loss": 0.5228, "step": 28312 }, { "epoch": 2.93, "learning_rate": 9.520355602402805e-06, "loss": 0.428, "step": 28313 }, { "epoch": 2.93, "learning_rate": 9.51855701068097e-06, "loss": 0.7872, "step": 28314 }, { "epoch": 2.93, "learning_rate": 9.516758556835416e-06, "loss": 0.1262, "step": 28315 }, { "epoch": 2.93, "learning_rate": 9.514960240878254e-06, "loss": 0.7649, "step": 28316 }, { "epoch": 2.93, "learning_rate": 9.513162062821581e-06, "loss": 0.2566, "step": 28317 }, { "epoch": 2.93, "learning_rate": 9.51136402267752e-06, "loss": 0.7554, "step": 28318 }, { "epoch": 2.93, "learning_rate": 9.509566120458155e-06, "loss": 0.375, "step": 28319 }, { "epoch": 2.93, "learning_rate": 9.507768356175604e-06, "loss": 0.1816, "step": 28320 }, { "epoch": 2.93, "learning_rate": 9.505970729841953e-06, "loss": 0.6523, "step": 28321 }, { "epoch": 2.93, "learning_rate": 9.504173241469309e-06, "loss": 0.0794, "step": 28322 }, { "epoch": 2.93, "learning_rate": 9.502375891069777e-06, "loss": 0.1245, "step": 28323 }, { "epoch": 2.93, "learning_rate": 9.500578678655454e-06, "loss": 0.5341, "step": 28324 }, { "epoch": 2.93, "learning_rate": 9.498781604238438e-06, "loss": 0.5753, "step": 28325 }, { "epoch": 2.93, "learning_rate": 9.496984667830829e-06, "loss": 0.2652, "step": 28326 }, { "epoch": 2.93, "learning_rate": 9.495187869444722e-06, "loss": 0.1778, "step": 28327 }, { "epoch": 2.93, "learning_rate": 9.493391209092221e-06, "loss": 0.1671, "step": 28328 }, { "epoch": 2.93, "learning_rate": 9.491594686785402e-06, "loss": 0.6861, "step": 28329 }, { "epoch": 2.93, "learning_rate": 9.489798302536374e-06, "loss": 0.2837, "step": 28330 }, { "epoch": 2.93, "learning_rate": 9.488002056357219e-06, "loss": 0.062, "step": 28331 }, { "epoch": 2.93, "learning_rate": 9.486205948260047e-06, "loss": 0.2185, "step": 28332 }, { "epoch": 2.93, "learning_rate": 9.484409978256924e-06, "loss": 0.0964, "step": 28333 }, { "epoch": 2.93, "learning_rate": 9.482614146359956e-06, "loss": 0.4637, "step": 28334 }, { "epoch": 2.93, "learning_rate": 9.480818452581227e-06, "loss": 0.1591, "step": 28335 }, { "epoch": 2.93, "learning_rate": 9.47902289693283e-06, "loss": 0.3346, "step": 28336 }, { "epoch": 2.93, "learning_rate": 9.477227479426847e-06, "loss": 0.2571, "step": 28337 }, { "epoch": 2.93, "learning_rate": 9.475432200075377e-06, "loss": 0.7132, "step": 28338 }, { "epoch": 2.93, "learning_rate": 9.473637058890487e-06, "loss": 0.6929, "step": 28339 }, { "epoch": 2.93, "learning_rate": 9.471842055884267e-06, "loss": 0.1767, "step": 28340 }, { "epoch": 2.93, "learning_rate": 9.470047191068803e-06, "loss": 0.3757, "step": 28341 }, { "epoch": 2.93, "learning_rate": 9.468252464456177e-06, "loss": 0.3302, "step": 28342 }, { "epoch": 2.93, "learning_rate": 9.466457876058472e-06, "loss": 0.6256, "step": 28343 }, { "epoch": 2.93, "learning_rate": 9.46466342588777e-06, "loss": 0.1505, "step": 28344 }, { "epoch": 2.93, "learning_rate": 9.462869113956158e-06, "loss": 0.3741, "step": 28345 }, { "epoch": 2.93, "learning_rate": 9.461074940275695e-06, "loss": 0.4047, "step": 28346 }, { "epoch": 2.93, "learning_rate": 9.459280904858472e-06, "loss": 0.3663, "step": 28347 }, { "epoch": 2.93, "learning_rate": 9.457487007716572e-06, "loss": 0.354, "step": 28348 }, { "epoch": 2.93, "learning_rate": 9.45569324886205e-06, "loss": 0.1806, "step": 28349 }, { "epoch": 2.93, "learning_rate": 9.453899628306997e-06, "loss": 0.5458, "step": 28350 }, { "epoch": 2.93, "learning_rate": 9.452106146063487e-06, "loss": 1.3376, "step": 28351 }, { "epoch": 2.93, "learning_rate": 9.450312802143589e-06, "loss": 0.2218, "step": 28352 }, { "epoch": 2.93, "learning_rate": 9.448519596559374e-06, "loss": 0.1084, "step": 28353 }, { "epoch": 2.93, "learning_rate": 9.446726529322918e-06, "loss": 0.3986, "step": 28354 }, { "epoch": 2.93, "learning_rate": 9.44493360044629e-06, "loss": 0.2756, "step": 28355 }, { "epoch": 2.93, "learning_rate": 9.443140809941573e-06, "loss": 0.2526, "step": 28356 }, { "epoch": 2.93, "learning_rate": 9.441348157820805e-06, "loss": 0.3566, "step": 28357 }, { "epoch": 2.94, "learning_rate": 9.439555644096088e-06, "loss": 0.7546, "step": 28358 }, { "epoch": 2.94, "learning_rate": 9.437763268779456e-06, "loss": 0.4037, "step": 28359 }, { "epoch": 2.94, "learning_rate": 9.435971031882996e-06, "loss": 0.5056, "step": 28360 }, { "epoch": 2.94, "learning_rate": 9.434178933418763e-06, "loss": 0.3361, "step": 28361 }, { "epoch": 2.94, "learning_rate": 9.432386973398826e-06, "loss": 0.1991, "step": 28362 }, { "epoch": 2.94, "learning_rate": 9.430595151835249e-06, "loss": 0.3013, "step": 28363 }, { "epoch": 2.94, "learning_rate": 9.428803468740091e-06, "loss": 0.0838, "step": 28364 }, { "epoch": 2.94, "learning_rate": 9.427011924125419e-06, "loss": 0.2904, "step": 28365 }, { "epoch": 2.94, "learning_rate": 9.425220518003293e-06, "loss": 0.2214, "step": 28366 }, { "epoch": 2.94, "learning_rate": 9.423429250385761e-06, "loss": 0.2736, "step": 28367 }, { "epoch": 2.94, "learning_rate": 9.421638121284888e-06, "loss": 0.3608, "step": 28368 }, { "epoch": 2.94, "learning_rate": 9.419847130712733e-06, "loss": 0.5349, "step": 28369 }, { "epoch": 2.94, "learning_rate": 9.418056278681365e-06, "loss": 0.4637, "step": 28370 }, { "epoch": 2.94, "learning_rate": 9.416265565202811e-06, "loss": 0.1997, "step": 28371 }, { "epoch": 2.94, "learning_rate": 9.414474990289144e-06, "loss": 0.7326, "step": 28372 }, { "epoch": 2.94, "learning_rate": 9.412684553952414e-06, "loss": 0.4597, "step": 28373 }, { "epoch": 2.94, "learning_rate": 9.410894256204673e-06, "loss": 0.2344, "step": 28374 }, { "epoch": 2.94, "learning_rate": 9.409104097057977e-06, "loss": 0.4931, "step": 28375 }, { "epoch": 2.94, "learning_rate": 9.40731407652438e-06, "loss": 0.4902, "step": 28376 }, { "epoch": 2.94, "learning_rate": 9.40552419461592e-06, "loss": 0.2097, "step": 28377 }, { "epoch": 2.94, "learning_rate": 9.403734451344653e-06, "loss": 0.1668, "step": 28378 }, { "epoch": 2.94, "learning_rate": 9.401944846722626e-06, "loss": 0.6345, "step": 28379 }, { "epoch": 2.94, "learning_rate": 9.400155380761885e-06, "loss": 0.218, "step": 28380 }, { "epoch": 2.94, "learning_rate": 9.398366053474479e-06, "loss": 0.4569, "step": 28381 }, { "epoch": 2.94, "learning_rate": 9.396576864872454e-06, "loss": 0.485, "step": 28382 }, { "epoch": 2.94, "learning_rate": 9.39478781496786e-06, "loss": 0.2218, "step": 28383 }, { "epoch": 2.94, "learning_rate": 9.392998903772726e-06, "loss": 0.1031, "step": 28384 }, { "epoch": 2.94, "learning_rate": 9.391210131299099e-06, "loss": 0.2318, "step": 28385 }, { "epoch": 2.94, "learning_rate": 9.389421497559035e-06, "loss": 0.2802, "step": 28386 }, { "epoch": 2.94, "learning_rate": 9.387633002564551e-06, "loss": 0.1984, "step": 28387 }, { "epoch": 2.94, "learning_rate": 9.385844646327699e-06, "loss": 0.3589, "step": 28388 }, { "epoch": 2.94, "learning_rate": 9.38405642886052e-06, "loss": 0.3862, "step": 28389 }, { "epoch": 2.94, "learning_rate": 9.382268350175048e-06, "loss": 0.1867, "step": 28390 }, { "epoch": 2.94, "learning_rate": 9.380480410283322e-06, "loss": 0.2537, "step": 28391 }, { "epoch": 2.94, "learning_rate": 9.378692609197375e-06, "loss": 0.1925, "step": 28392 }, { "epoch": 2.94, "learning_rate": 9.376904946929246e-06, "loss": 0.2352, "step": 28393 }, { "epoch": 2.94, "learning_rate": 9.375117423490977e-06, "loss": 0.5702, "step": 28394 }, { "epoch": 2.94, "learning_rate": 9.373330038894585e-06, "loss": 0.1967, "step": 28395 }, { "epoch": 2.94, "learning_rate": 9.371542793152112e-06, "loss": 0.8514, "step": 28396 }, { "epoch": 2.94, "learning_rate": 9.369755686275582e-06, "loss": 0.0749, "step": 28397 }, { "epoch": 2.94, "learning_rate": 9.367968718277029e-06, "loss": 0.0907, "step": 28398 }, { "epoch": 2.94, "learning_rate": 9.36618188916848e-06, "loss": 0.8321, "step": 28399 }, { "epoch": 2.94, "learning_rate": 9.36439519896197e-06, "loss": 0.2075, "step": 28400 }, { "epoch": 2.94, "learning_rate": 9.362608647669522e-06, "loss": 0.3093, "step": 28401 }, { "epoch": 2.94, "learning_rate": 9.360822235303163e-06, "loss": 0.2088, "step": 28402 }, { "epoch": 2.94, "learning_rate": 9.35903596187492e-06, "loss": 0.2547, "step": 28403 }, { "epoch": 2.94, "learning_rate": 9.35724982739683e-06, "loss": 0.6382, "step": 28404 }, { "epoch": 2.94, "learning_rate": 9.355463831880889e-06, "loss": 0.1883, "step": 28405 }, { "epoch": 2.94, "learning_rate": 9.353677975339139e-06, "loss": 0.331, "step": 28406 }, { "epoch": 2.94, "learning_rate": 9.351892257783594e-06, "loss": 0.2835, "step": 28407 }, { "epoch": 2.94, "learning_rate": 9.350106679226293e-06, "loss": 0.3707, "step": 28408 }, { "epoch": 2.94, "learning_rate": 9.34832123967923e-06, "loss": 0.111, "step": 28409 }, { "epoch": 2.94, "learning_rate": 9.346535939154436e-06, "loss": 0.7197, "step": 28410 }, { "epoch": 2.94, "learning_rate": 9.344750777663931e-06, "loss": 0.4028, "step": 28411 }, { "epoch": 2.94, "learning_rate": 9.342965755219732e-06, "loss": 0.2746, "step": 28412 }, { "epoch": 2.94, "learning_rate": 9.341180871833852e-06, "loss": 0.3299, "step": 28413 }, { "epoch": 2.94, "learning_rate": 9.339396127518318e-06, "loss": 0.4421, "step": 28414 }, { "epoch": 2.94, "learning_rate": 9.337611522285128e-06, "loss": 0.2843, "step": 28415 }, { "epoch": 2.94, "learning_rate": 9.3358270561463e-06, "loss": 0.2645, "step": 28416 }, { "epoch": 2.94, "learning_rate": 9.334042729113852e-06, "loss": 0.547, "step": 28417 }, { "epoch": 2.94, "learning_rate": 9.332258541199791e-06, "loss": 0.3031, "step": 28418 }, { "epoch": 2.94, "learning_rate": 9.33047449241613e-06, "loss": 0.0731, "step": 28419 }, { "epoch": 2.94, "learning_rate": 9.328690582774887e-06, "loss": 0.4955, "step": 28420 }, { "epoch": 2.94, "learning_rate": 9.326906812288052e-06, "loss": 0.164, "step": 28421 }, { "epoch": 2.94, "learning_rate": 9.325123180967645e-06, "loss": 0.5322, "step": 28422 }, { "epoch": 2.94, "learning_rate": 9.323339688825671e-06, "loss": 0.178, "step": 28423 }, { "epoch": 2.94, "learning_rate": 9.321556335874144e-06, "loss": 0.4659, "step": 28424 }, { "epoch": 2.94, "learning_rate": 9.319773122125056e-06, "loss": 0.5711, "step": 28425 }, { "epoch": 2.94, "learning_rate": 9.317990047590413e-06, "loss": 0.4884, "step": 28426 }, { "epoch": 2.94, "learning_rate": 9.316207112282225e-06, "loss": 0.4113, "step": 28427 }, { "epoch": 2.94, "learning_rate": 9.314424316212486e-06, "loss": 0.0875, "step": 28428 }, { "epoch": 2.94, "learning_rate": 9.312641659393206e-06, "loss": 0.4817, "step": 28429 }, { "epoch": 2.94, "learning_rate": 9.31085914183638e-06, "loss": 0.0652, "step": 28430 }, { "epoch": 2.94, "learning_rate": 9.309076763554015e-06, "loss": 0.4767, "step": 28431 }, { "epoch": 2.94, "learning_rate": 9.307294524558107e-06, "loss": 0.1766, "step": 28432 }, { "epoch": 2.94, "learning_rate": 9.305512424860644e-06, "loss": 0.2665, "step": 28433 }, { "epoch": 2.94, "learning_rate": 9.303730464473636e-06, "loss": 0.1115, "step": 28434 }, { "epoch": 2.94, "learning_rate": 9.301948643409065e-06, "loss": 0.888, "step": 28435 }, { "epoch": 2.94, "learning_rate": 9.300166961678934e-06, "loss": 0.8622, "step": 28436 }, { "epoch": 2.94, "learning_rate": 9.298385419295234e-06, "loss": 0.0616, "step": 28437 }, { "epoch": 2.94, "learning_rate": 9.296604016269959e-06, "loss": 0.1892, "step": 28438 }, { "epoch": 2.94, "learning_rate": 9.294822752615102e-06, "loss": 0.7499, "step": 28439 }, { "epoch": 2.94, "learning_rate": 9.293041628342652e-06, "loss": 0.7131, "step": 28440 }, { "epoch": 2.94, "learning_rate": 9.291260643464603e-06, "loss": 0.2861, "step": 28441 }, { "epoch": 2.94, "learning_rate": 9.289479797992952e-06, "loss": 0.1897, "step": 28442 }, { "epoch": 2.94, "learning_rate": 9.287699091939665e-06, "loss": 0.1445, "step": 28443 }, { "epoch": 2.94, "learning_rate": 9.285918525316743e-06, "loss": 0.4679, "step": 28444 }, { "epoch": 2.94, "learning_rate": 9.28413809813618e-06, "loss": 0.4004, "step": 28445 }, { "epoch": 2.94, "learning_rate": 9.282357810409941e-06, "loss": 0.2233, "step": 28446 }, { "epoch": 2.94, "learning_rate": 9.280577662150025e-06, "loss": 0.2174, "step": 28447 }, { "epoch": 2.94, "learning_rate": 9.27879765336841e-06, "loss": 0.3295, "step": 28448 }, { "epoch": 2.94, "learning_rate": 9.277017784077082e-06, "loss": 0.0606, "step": 28449 }, { "epoch": 2.94, "learning_rate": 9.27523805428802e-06, "loss": 0.5423, "step": 28450 }, { "epoch": 2.94, "learning_rate": 9.273458464013207e-06, "loss": 0.4431, "step": 28451 }, { "epoch": 2.94, "learning_rate": 9.271679013264633e-06, "loss": 0.3278, "step": 28452 }, { "epoch": 2.94, "learning_rate": 9.269899702054256e-06, "loss": 0.1316, "step": 28453 }, { "epoch": 2.95, "learning_rate": 9.268120530394061e-06, "loss": 0.2526, "step": 28454 }, { "epoch": 2.95, "learning_rate": 9.26634149829603e-06, "loss": 0.8366, "step": 28455 }, { "epoch": 2.95, "learning_rate": 9.264562605772135e-06, "loss": 0.4411, "step": 28456 }, { "epoch": 2.95, "learning_rate": 9.262783852834356e-06, "loss": 0.2801, "step": 28457 }, { "epoch": 2.95, "learning_rate": 9.261005239494671e-06, "loss": 0.6152, "step": 28458 }, { "epoch": 2.95, "learning_rate": 9.259226765765039e-06, "loss": 0.4348, "step": 28459 }, { "epoch": 2.95, "learning_rate": 9.257448431657438e-06, "loss": 0.4178, "step": 28460 }, { "epoch": 2.95, "learning_rate": 9.25567023718384e-06, "loss": 0.1267, "step": 28461 }, { "epoch": 2.95, "learning_rate": 9.253892182356223e-06, "loss": 0.3417, "step": 28462 }, { "epoch": 2.95, "learning_rate": 9.252114267186544e-06, "loss": 0.3767, "step": 28463 }, { "epoch": 2.95, "learning_rate": 9.250336491686778e-06, "loss": 0.1073, "step": 28464 }, { "epoch": 2.95, "learning_rate": 9.24855885586889e-06, "loss": 0.1928, "step": 28465 }, { "epoch": 2.95, "learning_rate": 9.246781359744848e-06, "loss": 0.3868, "step": 28466 }, { "epoch": 2.95, "learning_rate": 9.245004003326618e-06, "loss": 0.6772, "step": 28467 }, { "epoch": 2.95, "learning_rate": 9.243226786626167e-06, "loss": 0.554, "step": 28468 }, { "epoch": 2.95, "learning_rate": 9.241449709655454e-06, "loss": 0.1841, "step": 28469 }, { "epoch": 2.95, "learning_rate": 9.239672772426454e-06, "loss": 0.7783, "step": 28470 }, { "epoch": 2.95, "learning_rate": 9.23789597495111e-06, "loss": 0.2331, "step": 28471 }, { "epoch": 2.95, "learning_rate": 9.236119317241403e-06, "loss": 0.134, "step": 28472 }, { "epoch": 2.95, "learning_rate": 9.234342799309272e-06, "loss": 0.1725, "step": 28473 }, { "epoch": 2.95, "learning_rate": 9.232566421166686e-06, "loss": 0.4244, "step": 28474 }, { "epoch": 2.95, "learning_rate": 9.230790182825605e-06, "loss": 0.1714, "step": 28475 }, { "epoch": 2.95, "learning_rate": 9.229014084297984e-06, "loss": 0.4793, "step": 28476 }, { "epoch": 2.95, "learning_rate": 9.227238125595781e-06, "loss": 0.4664, "step": 28477 }, { "epoch": 2.95, "learning_rate": 9.22546230673095e-06, "loss": 0.8585, "step": 28478 }, { "epoch": 2.95, "learning_rate": 9.223686627715446e-06, "loss": 0.2605, "step": 28479 }, { "epoch": 2.95, "learning_rate": 9.221911088561232e-06, "loss": 0.3538, "step": 28480 }, { "epoch": 2.95, "learning_rate": 9.220135689280243e-06, "loss": 0.2807, "step": 28481 }, { "epoch": 2.95, "learning_rate": 9.218360429884435e-06, "loss": 0.2848, "step": 28482 }, { "epoch": 2.95, "learning_rate": 9.216585310385774e-06, "loss": 0.8863, "step": 28483 }, { "epoch": 2.95, "learning_rate": 9.21481033079619e-06, "loss": 0.2067, "step": 28484 }, { "epoch": 2.95, "learning_rate": 9.213035491127635e-06, "loss": 0.081, "step": 28485 }, { "epoch": 2.95, "learning_rate": 9.211260791392063e-06, "loss": 0.3948, "step": 28486 }, { "epoch": 2.95, "learning_rate": 9.209486231601417e-06, "loss": 0.6463, "step": 28487 }, { "epoch": 2.95, "learning_rate": 9.207711811767647e-06, "loss": 0.1124, "step": 28488 }, { "epoch": 2.95, "learning_rate": 9.205937531902697e-06, "loss": 0.1355, "step": 28489 }, { "epoch": 2.95, "learning_rate": 9.204163392018521e-06, "loss": 0.3124, "step": 28490 }, { "epoch": 2.95, "learning_rate": 9.202389392127039e-06, "loss": 0.2408, "step": 28491 }, { "epoch": 2.95, "learning_rate": 9.200615532240207e-06, "loss": 0.3447, "step": 28492 }, { "epoch": 2.95, "learning_rate": 9.198841812369965e-06, "loss": 0.1822, "step": 28493 }, { "epoch": 2.95, "learning_rate": 9.19706823252825e-06, "loss": 0.4906, "step": 28494 }, { "epoch": 2.95, "learning_rate": 9.195294792727005e-06, "loss": 0.0927, "step": 28495 }, { "epoch": 2.95, "learning_rate": 9.193521492978181e-06, "loss": 0.2116, "step": 28496 }, { "epoch": 2.95, "learning_rate": 9.191748333293689e-06, "loss": 0.5336, "step": 28497 }, { "epoch": 2.95, "learning_rate": 9.189975313685477e-06, "loss": 0.1806, "step": 28498 }, { "epoch": 2.95, "learning_rate": 9.188202434165485e-06, "loss": 0.1819, "step": 28499 }, { "epoch": 2.95, "learning_rate": 9.186429694745653e-06, "loss": 0.1115, "step": 28500 }, { "epoch": 2.95, "learning_rate": 9.184657095437897e-06, "loss": 0.6679, "step": 28501 }, { "epoch": 2.95, "learning_rate": 9.182884636254158e-06, "loss": 0.1958, "step": 28502 }, { "epoch": 2.95, "learning_rate": 9.18111231720637e-06, "loss": 0.3487, "step": 28503 }, { "epoch": 2.95, "learning_rate": 9.179340138306462e-06, "loss": 0.1429, "step": 28504 }, { "epoch": 2.95, "learning_rate": 9.177568099566364e-06, "loss": 0.1776, "step": 28505 }, { "epoch": 2.95, "learning_rate": 9.175796200998008e-06, "loss": 0.5099, "step": 28506 }, { "epoch": 2.95, "learning_rate": 9.174024442613315e-06, "loss": 0.1986, "step": 28507 }, { "epoch": 2.95, "learning_rate": 9.172252824424232e-06, "loss": 0.3516, "step": 28508 }, { "epoch": 2.95, "learning_rate": 9.170481346442658e-06, "loss": 0.67, "step": 28509 }, { "epoch": 2.95, "learning_rate": 9.168710008680536e-06, "loss": 1.2882, "step": 28510 }, { "epoch": 2.95, "learning_rate": 9.166938811149776e-06, "loss": 0.5983, "step": 28511 }, { "epoch": 2.95, "learning_rate": 9.165167753862311e-06, "loss": 0.1701, "step": 28512 }, { "epoch": 2.95, "learning_rate": 9.163396836830063e-06, "loss": 0.2479, "step": 28513 }, { "epoch": 2.95, "learning_rate": 9.161626060064951e-06, "loss": 0.3807, "step": 28514 }, { "epoch": 2.95, "learning_rate": 9.159855423578895e-06, "loss": 0.283, "step": 28515 }, { "epoch": 2.95, "learning_rate": 9.158084927383817e-06, "loss": 0.2873, "step": 28516 }, { "epoch": 2.95, "learning_rate": 9.156314571491633e-06, "loss": 0.1429, "step": 28517 }, { "epoch": 2.95, "learning_rate": 9.154544355914277e-06, "loss": 0.5263, "step": 28518 }, { "epoch": 2.95, "learning_rate": 9.152774280663636e-06, "loss": 0.4068, "step": 28519 }, { "epoch": 2.95, "learning_rate": 9.15100434575164e-06, "loss": 0.3843, "step": 28520 }, { "epoch": 2.95, "learning_rate": 9.149234551190219e-06, "loss": 0.2496, "step": 28521 }, { "epoch": 2.95, "learning_rate": 9.147464896991259e-06, "loss": 0.2926, "step": 28522 }, { "epoch": 2.95, "learning_rate": 9.145695383166688e-06, "loss": 0.3536, "step": 28523 }, { "epoch": 2.95, "learning_rate": 9.143926009728416e-06, "loss": 0.486, "step": 28524 }, { "epoch": 2.95, "learning_rate": 9.142156776688353e-06, "loss": 0.3226, "step": 28525 }, { "epoch": 2.95, "learning_rate": 9.140387684058412e-06, "loss": 0.4126, "step": 28526 }, { "epoch": 2.95, "learning_rate": 9.138618731850499e-06, "loss": 0.317, "step": 28527 }, { "epoch": 2.95, "learning_rate": 9.136849920076535e-06, "loss": 0.6205, "step": 28528 }, { "epoch": 2.95, "learning_rate": 9.135081248748402e-06, "loss": 0.0995, "step": 28529 }, { "epoch": 2.95, "learning_rate": 9.133312717878023e-06, "loss": 0.3323, "step": 28530 }, { "epoch": 2.95, "learning_rate": 9.131544327477299e-06, "loss": 0.1622, "step": 28531 }, { "epoch": 2.95, "learning_rate": 9.129776077558138e-06, "loss": 0.2591, "step": 28532 }, { "epoch": 2.95, "learning_rate": 9.128007968132452e-06, "loss": 0.1561, "step": 28533 }, { "epoch": 2.95, "learning_rate": 9.126239999212119e-06, "loss": 0.1962, "step": 28534 }, { "epoch": 2.95, "learning_rate": 9.124472170809056e-06, "loss": 0.6488, "step": 28535 }, { "epoch": 2.95, "learning_rate": 9.122704482935163e-06, "loss": 0.1762, "step": 28536 }, { "epoch": 2.95, "learning_rate": 9.120936935602337e-06, "loss": 0.3689, "step": 28537 }, { "epoch": 2.95, "learning_rate": 9.119169528822489e-06, "loss": 0.3791, "step": 28538 }, { "epoch": 2.95, "learning_rate": 9.117402262607492e-06, "loss": 0.2477, "step": 28539 }, { "epoch": 2.95, "learning_rate": 9.115635136969263e-06, "loss": 0.2852, "step": 28540 }, { "epoch": 2.95, "learning_rate": 9.11386815191969e-06, "loss": 0.2914, "step": 28541 }, { "epoch": 2.95, "learning_rate": 9.112101307470668e-06, "loss": 0.7945, "step": 28542 }, { "epoch": 2.95, "learning_rate": 9.110334603634092e-06, "loss": 0.1457, "step": 28543 }, { "epoch": 2.95, "learning_rate": 9.108568040421856e-06, "loss": 0.2333, "step": 28544 }, { "epoch": 2.95, "learning_rate": 9.106801617845856e-06, "loss": 0.4547, "step": 28545 }, { "epoch": 2.95, "learning_rate": 9.105035335917986e-06, "loss": 0.2131, "step": 28546 }, { "epoch": 2.95, "learning_rate": 9.103269194650118e-06, "loss": 0.0342, "step": 28547 }, { "epoch": 2.95, "learning_rate": 9.101503194054161e-06, "loss": 0.0955, "step": 28548 }, { "epoch": 2.95, "learning_rate": 9.099737334141987e-06, "loss": 0.1427, "step": 28549 }, { "epoch": 2.96, "learning_rate": 9.09797161492549e-06, "loss": 0.1353, "step": 28550 }, { "epoch": 2.96, "learning_rate": 9.09620603641656e-06, "loss": 0.114, "step": 28551 }, { "epoch": 2.96, "learning_rate": 9.094440598627078e-06, "loss": 0.6487, "step": 28552 }, { "epoch": 2.96, "learning_rate": 9.092675301568929e-06, "loss": 0.888, "step": 28553 }, { "epoch": 2.96, "learning_rate": 9.090910145254001e-06, "loss": 0.3641, "step": 28554 }, { "epoch": 2.96, "learning_rate": 9.08914512969417e-06, "loss": 0.7795, "step": 28555 }, { "epoch": 2.96, "learning_rate": 9.087380254901332e-06, "loss": 0.4716, "step": 28556 }, { "epoch": 2.96, "learning_rate": 9.085615520887349e-06, "loss": 0.4078, "step": 28557 }, { "epoch": 2.96, "learning_rate": 9.083850927664117e-06, "loss": 0.3748, "step": 28558 }, { "epoch": 2.96, "learning_rate": 9.082086475243499e-06, "loss": 0.5311, "step": 28559 }, { "epoch": 2.96, "learning_rate": 9.080322163637377e-06, "loss": 0.2164, "step": 28560 }, { "epoch": 2.96, "learning_rate": 9.078557992857635e-06, "loss": 0.0537, "step": 28561 }, { "epoch": 2.96, "learning_rate": 9.076793962916141e-06, "loss": 0.6374, "step": 28562 }, { "epoch": 2.96, "learning_rate": 9.07503007382478e-06, "loss": 0.4454, "step": 28563 }, { "epoch": 2.96, "learning_rate": 9.073266325595418e-06, "loss": 0.2471, "step": 28564 }, { "epoch": 2.96, "learning_rate": 9.07150271823993e-06, "loss": 0.5407, "step": 28565 }, { "epoch": 2.96, "learning_rate": 9.069739251770197e-06, "loss": 0.1281, "step": 28566 }, { "epoch": 2.96, "learning_rate": 9.067975926198075e-06, "loss": 0.5886, "step": 28567 }, { "epoch": 2.96, "learning_rate": 9.066212741535438e-06, "loss": 0.4029, "step": 28568 }, { "epoch": 2.96, "learning_rate": 9.064449697794163e-06, "loss": 0.4208, "step": 28569 }, { "epoch": 2.96, "learning_rate": 9.06268679498611e-06, "loss": 0.4061, "step": 28570 }, { "epoch": 2.96, "learning_rate": 9.060924033123166e-06, "loss": 0.8272, "step": 28571 }, { "epoch": 2.96, "learning_rate": 9.059161412217167e-06, "loss": 0.0127, "step": 28572 }, { "epoch": 2.96, "learning_rate": 9.057398932279995e-06, "loss": 0.5784, "step": 28573 }, { "epoch": 2.96, "learning_rate": 9.055636593323513e-06, "loss": 0.4111, "step": 28574 }, { "epoch": 2.96, "learning_rate": 9.053874395359588e-06, "loss": 0.2615, "step": 28575 }, { "epoch": 2.96, "learning_rate": 9.052112338400085e-06, "loss": 0.4024, "step": 28576 }, { "epoch": 2.96, "learning_rate": 9.050350422456851e-06, "loss": 0.2953, "step": 28577 }, { "epoch": 2.96, "learning_rate": 9.04858864754176e-06, "loss": 1.3667, "step": 28578 }, { "epoch": 2.96, "learning_rate": 9.046827013666665e-06, "loss": 0.0683, "step": 28579 }, { "epoch": 2.96, "learning_rate": 9.045065520843427e-06, "loss": 0.5925, "step": 28580 }, { "epoch": 2.96, "learning_rate": 9.043304169083907e-06, "loss": 0.3589, "step": 28581 }, { "epoch": 2.96, "learning_rate": 9.041542958399961e-06, "loss": 0.7136, "step": 28582 }, { "epoch": 2.96, "learning_rate": 9.03978188880344e-06, "loss": 0.12, "step": 28583 }, { "epoch": 2.96, "learning_rate": 9.038020960306215e-06, "loss": 0.1304, "step": 28584 }, { "epoch": 2.96, "learning_rate": 9.036260172920118e-06, "loss": 0.1984, "step": 28585 }, { "epoch": 2.96, "learning_rate": 9.034499526657023e-06, "loss": 0.8932, "step": 28586 }, { "epoch": 2.96, "learning_rate": 9.032739021528761e-06, "loss": 0.2677, "step": 28587 }, { "epoch": 2.96, "learning_rate": 9.030978657547193e-06, "loss": 0.0693, "step": 28588 }, { "epoch": 2.96, "learning_rate": 9.029218434724174e-06, "loss": 0.7502, "step": 28589 }, { "epoch": 2.96, "learning_rate": 9.027458353071547e-06, "loss": 0.4084, "step": 28590 }, { "epoch": 2.96, "learning_rate": 9.025698412601162e-06, "loss": 0.0681, "step": 28591 }, { "epoch": 2.96, "learning_rate": 9.023938613324869e-06, "loss": 0.2201, "step": 28592 }, { "epoch": 2.96, "learning_rate": 9.022178955254515e-06, "loss": 0.3824, "step": 28593 }, { "epoch": 2.96, "learning_rate": 9.02041943840195e-06, "loss": 0.3366, "step": 28594 }, { "epoch": 2.96, "learning_rate": 9.018660062779007e-06, "loss": 0.4089, "step": 28595 }, { "epoch": 2.96, "learning_rate": 9.01690082839754e-06, "loss": 0.1247, "step": 28596 }, { "epoch": 2.96, "learning_rate": 9.01514173526938e-06, "loss": 0.5379, "step": 28597 }, { "epoch": 2.96, "learning_rate": 9.013382783406375e-06, "loss": 0.2326, "step": 28598 }, { "epoch": 2.96, "learning_rate": 9.01162397282037e-06, "loss": 0.7009, "step": 28599 }, { "epoch": 2.96, "learning_rate": 9.009865303523199e-06, "loss": 0.1326, "step": 28600 }, { "epoch": 2.96, "learning_rate": 9.008106775526701e-06, "loss": 0.3751, "step": 28601 }, { "epoch": 2.96, "learning_rate": 9.00634838884272e-06, "loss": 0.134, "step": 28602 }, { "epoch": 2.96, "learning_rate": 9.00459014348309e-06, "loss": 0.7501, "step": 28603 }, { "epoch": 2.96, "learning_rate": 9.002832039459652e-06, "loss": 0.9549, "step": 28604 }, { "epoch": 2.96, "learning_rate": 9.001074076784231e-06, "loss": 0.2785, "step": 28605 }, { "epoch": 2.96, "learning_rate": 8.999316255468661e-06, "loss": 0.3465, "step": 28606 }, { "epoch": 2.96, "learning_rate": 8.997558575524786e-06, "loss": 0.4713, "step": 28607 }, { "epoch": 2.96, "learning_rate": 8.995801036964429e-06, "loss": 0.3822, "step": 28608 }, { "epoch": 2.96, "learning_rate": 8.994043639799433e-06, "loss": 0.1426, "step": 28609 }, { "epoch": 2.96, "learning_rate": 8.992286384041613e-06, "loss": 0.1107, "step": 28610 }, { "epoch": 2.96, "learning_rate": 8.990529269702804e-06, "loss": 0.3457, "step": 28611 }, { "epoch": 2.96, "learning_rate": 8.988772296794838e-06, "loss": 0.1161, "step": 28612 }, { "epoch": 2.96, "learning_rate": 8.98701546532954e-06, "loss": 0.4029, "step": 28613 }, { "epoch": 2.96, "learning_rate": 8.985258775318747e-06, "loss": 0.3539, "step": 28614 }, { "epoch": 2.96, "learning_rate": 8.983502226774264e-06, "loss": 0.5074, "step": 28615 }, { "epoch": 2.96, "learning_rate": 8.98174581970793e-06, "loss": 0.5338, "step": 28616 }, { "epoch": 2.96, "learning_rate": 8.97998955413156e-06, "loss": 0.0758, "step": 28617 }, { "epoch": 2.96, "learning_rate": 8.978233430056988e-06, "loss": 0.6724, "step": 28618 }, { "epoch": 2.96, "learning_rate": 8.976477447496027e-06, "loss": 0.5956, "step": 28619 }, { "epoch": 2.96, "learning_rate": 8.9747216064605e-06, "loss": 0.459, "step": 28620 }, { "epoch": 2.96, "learning_rate": 8.972965906962235e-06, "loss": 0.341, "step": 28621 }, { "epoch": 2.96, "learning_rate": 8.97121034901304e-06, "loss": 0.1915, "step": 28622 }, { "epoch": 2.96, "learning_rate": 8.96945493262473e-06, "loss": 0.2309, "step": 28623 }, { "epoch": 2.96, "learning_rate": 8.96769965780914e-06, "loss": 0.0852, "step": 28624 }, { "epoch": 2.96, "learning_rate": 8.965944524578069e-06, "loss": 0.0946, "step": 28625 }, { "epoch": 2.96, "learning_rate": 8.964189532943335e-06, "loss": 0.1435, "step": 28626 }, { "epoch": 2.96, "learning_rate": 8.962434682916757e-06, "loss": 0.4643, "step": 28627 }, { "epoch": 2.96, "learning_rate": 8.960679974510143e-06, "loss": 0.2502, "step": 28628 }, { "epoch": 2.96, "learning_rate": 8.958925407735311e-06, "loss": 0.0959, "step": 28629 }, { "epoch": 2.96, "learning_rate": 8.957170982604068e-06, "loss": 0.5771, "step": 28630 }, { "epoch": 2.96, "learning_rate": 8.95541669912823e-06, "loss": 0.3743, "step": 28631 }, { "epoch": 2.96, "learning_rate": 8.953662557319608e-06, "loss": 0.055, "step": 28632 }, { "epoch": 2.96, "learning_rate": 8.951908557189999e-06, "loss": 0.7838, "step": 28633 }, { "epoch": 2.96, "learning_rate": 8.950154698751226e-06, "loss": 0.2456, "step": 28634 }, { "epoch": 2.96, "learning_rate": 8.948400982015076e-06, "loss": 0.3058, "step": 28635 }, { "epoch": 2.96, "learning_rate": 8.946647406993362e-06, "loss": 0.3227, "step": 28636 }, { "epoch": 2.96, "learning_rate": 8.944893973697896e-06, "loss": 0.2757, "step": 28637 }, { "epoch": 2.96, "learning_rate": 8.943140682140474e-06, "loss": 0.2198, "step": 28638 }, { "epoch": 2.96, "learning_rate": 8.941387532332906e-06, "loss": 0.699, "step": 28639 }, { "epoch": 2.96, "learning_rate": 8.939634524286985e-06, "loss": 0.2333, "step": 28640 }, { "epoch": 2.96, "learning_rate": 8.93788165801452e-06, "loss": 0.2657, "step": 28641 }, { "epoch": 2.96, "learning_rate": 8.936128933527316e-06, "loss": 0.1708, "step": 28642 }, { "epoch": 2.96, "learning_rate": 8.934376350837154e-06, "loss": 0.7834, "step": 28643 }, { "epoch": 2.96, "learning_rate": 8.932623909955841e-06, "loss": 0.2688, "step": 28644 }, { "epoch": 2.97, "learning_rate": 8.930871610895175e-06, "loss": 0.1723, "step": 28645 }, { "epoch": 2.97, "learning_rate": 8.929119453666961e-06, "loss": 0.3419, "step": 28646 }, { "epoch": 2.97, "learning_rate": 8.927367438282977e-06, "loss": 0.2955, "step": 28647 }, { "epoch": 2.97, "learning_rate": 8.925615564755022e-06, "loss": 0.2365, "step": 28648 }, { "epoch": 2.97, "learning_rate": 8.923863833094891e-06, "loss": 0.835, "step": 28649 }, { "epoch": 2.97, "learning_rate": 8.92211224331438e-06, "loss": 0.1705, "step": 28650 }, { "epoch": 2.97, "learning_rate": 8.920360795425277e-06, "loss": 0.4327, "step": 28651 }, { "epoch": 2.97, "learning_rate": 8.918609489439382e-06, "loss": 0.1653, "step": 28652 }, { "epoch": 2.97, "learning_rate": 8.916858325368465e-06, "loss": 0.3879, "step": 28653 }, { "epoch": 2.97, "learning_rate": 8.915107303224323e-06, "loss": 0.7108, "step": 28654 }, { "epoch": 2.97, "learning_rate": 8.91335642301875e-06, "loss": 0.4509, "step": 28655 }, { "epoch": 2.97, "learning_rate": 8.911605684763525e-06, "loss": 0.1011, "step": 28656 }, { "epoch": 2.97, "learning_rate": 8.909855088470436e-06, "loss": 0.2471, "step": 28657 }, { "epoch": 2.97, "learning_rate": 8.908104634151268e-06, "loss": 0.1361, "step": 28658 }, { "epoch": 2.97, "learning_rate": 8.906354321817814e-06, "loss": 0.4242, "step": 28659 }, { "epoch": 2.97, "learning_rate": 8.904604151481841e-06, "loss": 0.1004, "step": 28660 }, { "epoch": 2.97, "learning_rate": 8.902854123155131e-06, "loss": 0.5512, "step": 28661 }, { "epoch": 2.97, "learning_rate": 8.901104236849485e-06, "loss": 0.8693, "step": 28662 }, { "epoch": 2.97, "learning_rate": 8.899354492576656e-06, "loss": 0.4865, "step": 28663 }, { "epoch": 2.97, "learning_rate": 8.897604890348437e-06, "loss": 0.2126, "step": 28664 }, { "epoch": 2.97, "learning_rate": 8.895855430176603e-06, "loss": 0.3268, "step": 28665 }, { "epoch": 2.97, "learning_rate": 8.894106112072935e-06, "loss": 0.1859, "step": 28666 }, { "epoch": 2.97, "learning_rate": 8.892356936049204e-06, "loss": 0.6752, "step": 28667 }, { "epoch": 2.97, "learning_rate": 8.890607902117189e-06, "loss": 0.4318, "step": 28668 }, { "epoch": 2.97, "learning_rate": 8.888859010288662e-06, "loss": 0.5981, "step": 28669 }, { "epoch": 2.97, "learning_rate": 8.887110260575404e-06, "loss": 0.4441, "step": 28670 }, { "epoch": 2.97, "learning_rate": 8.885361652989171e-06, "loss": 0.5241, "step": 28671 }, { "epoch": 2.97, "learning_rate": 8.883613187541754e-06, "loss": 0.3964, "step": 28672 }, { "epoch": 2.97, "learning_rate": 8.881864864244903e-06, "loss": 0.5572, "step": 28673 }, { "epoch": 2.97, "learning_rate": 8.880116683110392e-06, "loss": 0.6192, "step": 28674 }, { "epoch": 2.97, "learning_rate": 8.878368644149998e-06, "loss": 0.4051, "step": 28675 }, { "epoch": 2.97, "learning_rate": 8.876620747375483e-06, "loss": 0.1783, "step": 28676 }, { "epoch": 2.97, "learning_rate": 8.874872992798612e-06, "loss": 0.3736, "step": 28677 }, { "epoch": 2.97, "learning_rate": 8.873125380431157e-06, "loss": 0.4773, "step": 28678 }, { "epoch": 2.97, "learning_rate": 8.871377910284878e-06, "loss": 0.4197, "step": 28679 }, { "epoch": 2.97, "learning_rate": 8.86963058237155e-06, "loss": 0.6723, "step": 28680 }, { "epoch": 2.97, "learning_rate": 8.867883396702911e-06, "loss": 0.1912, "step": 28681 }, { "epoch": 2.97, "learning_rate": 8.866136353290742e-06, "loss": 0.6669, "step": 28682 }, { "epoch": 2.97, "learning_rate": 8.864389452146791e-06, "loss": 0.9341, "step": 28683 }, { "epoch": 2.97, "learning_rate": 8.862642693282839e-06, "loss": 0.493, "step": 28684 }, { "epoch": 2.97, "learning_rate": 8.860896076710623e-06, "loss": 0.9262, "step": 28685 }, { "epoch": 2.97, "learning_rate": 8.859149602441904e-06, "loss": 0.1375, "step": 28686 }, { "epoch": 2.97, "learning_rate": 8.857403270488446e-06, "loss": 0.2641, "step": 28687 }, { "epoch": 2.97, "learning_rate": 8.855657080862002e-06, "loss": 0.4515, "step": 28688 }, { "epoch": 2.97, "learning_rate": 8.853911033574328e-06, "loss": 0.2498, "step": 28689 }, { "epoch": 2.97, "learning_rate": 8.852165128637185e-06, "loss": 0.8265, "step": 28690 }, { "epoch": 2.97, "learning_rate": 8.850419366062308e-06, "loss": 0.4043, "step": 28691 }, { "epoch": 2.97, "learning_rate": 8.84867374586146e-06, "loss": 0.2676, "step": 28692 }, { "epoch": 2.97, "learning_rate": 8.846928268046392e-06, "loss": 0.5111, "step": 28693 }, { "epoch": 2.97, "learning_rate": 8.845182932628854e-06, "loss": 0.2113, "step": 28694 }, { "epoch": 2.97, "learning_rate": 8.843437739620593e-06, "loss": 0.2685, "step": 28695 }, { "epoch": 2.97, "learning_rate": 8.841692689033357e-06, "loss": 0.8239, "step": 28696 }, { "epoch": 2.97, "learning_rate": 8.839947780878908e-06, "loss": 0.6212, "step": 28697 }, { "epoch": 2.97, "learning_rate": 8.838203015168972e-06, "loss": 0.4381, "step": 28698 }, { "epoch": 2.97, "learning_rate": 8.836458391915298e-06, "loss": 0.2839, "step": 28699 }, { "epoch": 2.97, "learning_rate": 8.834713911129646e-06, "loss": 0.4387, "step": 28700 }, { "epoch": 2.97, "learning_rate": 8.832969572823735e-06, "loss": 0.3361, "step": 28701 }, { "epoch": 2.97, "learning_rate": 8.831225377009323e-06, "loss": 0.1307, "step": 28702 }, { "epoch": 2.97, "learning_rate": 8.829481323698149e-06, "loss": 0.1893, "step": 28703 }, { "epoch": 2.97, "learning_rate": 8.827737412901951e-06, "loss": 0.5347, "step": 28704 }, { "epoch": 2.97, "learning_rate": 8.825993644632472e-06, "loss": 0.0679, "step": 28705 }, { "epoch": 2.97, "learning_rate": 8.82425001890145e-06, "loss": 1.4252, "step": 28706 }, { "epoch": 2.97, "learning_rate": 8.82250653572062e-06, "loss": 0.4211, "step": 28707 }, { "epoch": 2.97, "learning_rate": 8.820763195101731e-06, "loss": 0.1982, "step": 28708 }, { "epoch": 2.97, "learning_rate": 8.819019997056502e-06, "loss": 1.1119, "step": 28709 }, { "epoch": 2.97, "learning_rate": 8.81727694159668e-06, "loss": 0.1856, "step": 28710 }, { "epoch": 2.97, "learning_rate": 8.815534028733985e-06, "loss": 0.7254, "step": 28711 }, { "epoch": 2.97, "learning_rate": 8.81379125848016e-06, "loss": 0.2234, "step": 28712 }, { "epoch": 2.97, "learning_rate": 8.812048630846935e-06, "loss": 0.2276, "step": 28713 }, { "epoch": 2.97, "learning_rate": 8.81030614584604e-06, "loss": 0.7929, "step": 28714 }, { "epoch": 2.97, "learning_rate": 8.808563803489205e-06, "loss": 0.3128, "step": 28715 }, { "epoch": 2.97, "learning_rate": 8.80682160378816e-06, "loss": 0.7231, "step": 28716 }, { "epoch": 2.97, "learning_rate": 8.805079546754637e-06, "loss": 0.3121, "step": 28717 }, { "epoch": 2.97, "learning_rate": 8.803337632400364e-06, "loss": 0.1138, "step": 28718 }, { "epoch": 2.97, "learning_rate": 8.801595860737055e-06, "loss": 0.3588, "step": 28719 }, { "epoch": 2.97, "learning_rate": 8.799854231776442e-06, "loss": 0.2203, "step": 28720 }, { "epoch": 2.97, "learning_rate": 8.798112745530252e-06, "loss": 0.1283, "step": 28721 }, { "epoch": 2.97, "learning_rate": 8.796371402010214e-06, "loss": 0.1094, "step": 28722 }, { "epoch": 2.97, "learning_rate": 8.794630201228033e-06, "loss": 0.2821, "step": 28723 }, { "epoch": 2.97, "learning_rate": 8.792889143195438e-06, "loss": 0.2491, "step": 28724 }, { "epoch": 2.97, "learning_rate": 8.791148227924153e-06, "loss": 0.1109, "step": 28725 }, { "epoch": 2.97, "learning_rate": 8.789407455425897e-06, "loss": 0.5111, "step": 28726 }, { "epoch": 2.97, "learning_rate": 8.787666825712385e-06, "loss": 0.4012, "step": 28727 }, { "epoch": 2.97, "learning_rate": 8.78592633879535e-06, "loss": 0.2647, "step": 28728 }, { "epoch": 2.97, "learning_rate": 8.78418599468648e-06, "loss": 0.3167, "step": 28729 }, { "epoch": 2.97, "learning_rate": 8.782445793397509e-06, "loss": 0.2338, "step": 28730 }, { "epoch": 2.97, "learning_rate": 8.780705734940147e-06, "loss": 0.3682, "step": 28731 }, { "epoch": 2.97, "learning_rate": 8.778965819326111e-06, "loss": 0.5783, "step": 28732 }, { "epoch": 2.97, "learning_rate": 8.77722604656711e-06, "loss": 0.3866, "step": 28733 }, { "epoch": 2.97, "learning_rate": 8.775486416674868e-06, "loss": 0.9437, "step": 28734 }, { "epoch": 2.97, "learning_rate": 8.773746929661075e-06, "loss": 0.5972, "step": 28735 }, { "epoch": 2.97, "learning_rate": 8.772007585537452e-06, "loss": 0.7667, "step": 28736 }, { "epoch": 2.97, "learning_rate": 8.770268384315708e-06, "loss": 0.4165, "step": 28737 }, { "epoch": 2.97, "learning_rate": 8.768529326007555e-06, "loss": 0.2954, "step": 28738 }, { "epoch": 2.97, "learning_rate": 8.76679041062469e-06, "loss": 0.8693, "step": 28739 }, { "epoch": 2.97, "learning_rate": 8.765051638178823e-06, "loss": 0.8065, "step": 28740 }, { "epoch": 2.98, "learning_rate": 8.763313008681658e-06, "loss": 0.3008, "step": 28741 }, { "epoch": 2.98, "learning_rate": 8.761574522144903e-06, "loss": 0.2691, "step": 28742 }, { "epoch": 2.98, "learning_rate": 8.759836178580255e-06, "loss": 0.1866, "step": 28743 }, { "epoch": 2.98, "learning_rate": 8.758097977999425e-06, "loss": 0.2067, "step": 28744 }, { "epoch": 2.98, "learning_rate": 8.756359920414108e-06, "loss": 0.717, "step": 28745 }, { "epoch": 2.98, "learning_rate": 8.754622005836012e-06, "loss": 0.1586, "step": 28746 }, { "epoch": 2.98, "learning_rate": 8.752884234276821e-06, "loss": 0.6939, "step": 28747 }, { "epoch": 2.98, "learning_rate": 8.751146605748254e-06, "loss": 0.4146, "step": 28748 }, { "epoch": 2.98, "learning_rate": 8.749409120261984e-06, "loss": 0.0943, "step": 28749 }, { "epoch": 2.98, "learning_rate": 8.74767177782972e-06, "loss": 0.3947, "step": 28750 }, { "epoch": 2.98, "learning_rate": 8.745934578463159e-06, "loss": 0.3387, "step": 28751 }, { "epoch": 2.98, "learning_rate": 8.74419752217399e-06, "loss": 0.2614, "step": 28752 }, { "epoch": 2.98, "learning_rate": 8.742460608973913e-06, "loss": 0.1447, "step": 28753 }, { "epoch": 2.98, "learning_rate": 8.740723838874616e-06, "loss": 0.854, "step": 28754 }, { "epoch": 2.98, "learning_rate": 8.73898721188779e-06, "loss": 0.4666, "step": 28755 }, { "epoch": 2.98, "learning_rate": 8.737250728025142e-06, "loss": 0.5644, "step": 28756 }, { "epoch": 2.98, "learning_rate": 8.735514387298334e-06, "loss": 0.1968, "step": 28757 }, { "epoch": 2.98, "learning_rate": 8.733778189719069e-06, "loss": 0.2419, "step": 28758 }, { "epoch": 2.98, "learning_rate": 8.732042135299041e-06, "loss": 0.3243, "step": 28759 }, { "epoch": 2.98, "learning_rate": 8.730306224049918e-06, "loss": 0.2754, "step": 28760 }, { "epoch": 2.98, "learning_rate": 8.728570455983398e-06, "loss": 0.4116, "step": 28761 }, { "epoch": 2.98, "learning_rate": 8.72683483111117e-06, "loss": 0.2745, "step": 28762 }, { "epoch": 2.98, "learning_rate": 8.725099349444906e-06, "loss": 0.378, "step": 28763 }, { "epoch": 2.98, "learning_rate": 8.723364010996297e-06, "loss": 0.2297, "step": 28764 }, { "epoch": 2.98, "learning_rate": 8.721628815777024e-06, "loss": 0.4596, "step": 28765 }, { "epoch": 2.98, "learning_rate": 8.719893763798777e-06, "loss": 0.3357, "step": 28766 }, { "epoch": 2.98, "learning_rate": 8.718158855073215e-06, "loss": 0.567, "step": 28767 }, { "epoch": 2.98, "learning_rate": 8.716424089612028e-06, "loss": 0.2657, "step": 28768 }, { "epoch": 2.98, "learning_rate": 8.714689467426893e-06, "loss": 0.2907, "step": 28769 }, { "epoch": 2.98, "learning_rate": 8.712954988529493e-06, "loss": 0.5576, "step": 28770 }, { "epoch": 2.98, "learning_rate": 8.711220652931497e-06, "loss": 0.1926, "step": 28771 }, { "epoch": 2.98, "learning_rate": 8.70948646064459e-06, "loss": 0.344, "step": 28772 }, { "epoch": 2.98, "learning_rate": 8.707752411680427e-06, "loss": 0.3272, "step": 28773 }, { "epoch": 2.98, "learning_rate": 8.706018506050698e-06, "loss": 0.1309, "step": 28774 }, { "epoch": 2.98, "learning_rate": 8.70428474376707e-06, "loss": 0.2202, "step": 28775 }, { "epoch": 2.98, "learning_rate": 8.702551124841218e-06, "loss": 0.6359, "step": 28776 }, { "epoch": 2.98, "learning_rate": 8.700817649284803e-06, "loss": 0.3914, "step": 28777 }, { "epoch": 2.98, "learning_rate": 8.699084317109499e-06, "loss": 0.301, "step": 28778 }, { "epoch": 2.98, "learning_rate": 8.697351128326979e-06, "loss": 0.1452, "step": 28779 }, { "epoch": 2.98, "learning_rate": 8.695618082948904e-06, "loss": 0.4693, "step": 28780 }, { "epoch": 2.98, "learning_rate": 8.693885180986942e-06, "loss": 0.2505, "step": 28781 }, { "epoch": 2.98, "learning_rate": 8.692152422452761e-06, "loss": 0.3196, "step": 28782 }, { "epoch": 2.98, "learning_rate": 8.690419807358025e-06, "loss": 0.2215, "step": 28783 }, { "epoch": 2.98, "learning_rate": 8.688687335714404e-06, "loss": 0.5229, "step": 28784 }, { "epoch": 2.98, "learning_rate": 8.686955007533547e-06, "loss": 0.2882, "step": 28785 }, { "epoch": 2.98, "learning_rate": 8.685222822827129e-06, "loss": 0.3192, "step": 28786 }, { "epoch": 2.98, "learning_rate": 8.683490781606794e-06, "loss": 0.3315, "step": 28787 }, { "epoch": 2.98, "learning_rate": 8.681758883884214e-06, "loss": 0.434, "step": 28788 }, { "epoch": 2.98, "learning_rate": 8.680027129671043e-06, "loss": 0.3901, "step": 28789 }, { "epoch": 2.98, "learning_rate": 8.678295518978941e-06, "loss": 0.1383, "step": 28790 }, { "epoch": 2.98, "learning_rate": 8.676564051819566e-06, "loss": 0.3467, "step": 28791 }, { "epoch": 2.98, "learning_rate": 8.674832728204573e-06, "loss": 0.2035, "step": 28792 }, { "epoch": 2.98, "learning_rate": 8.673101548145613e-06, "loss": 0.0458, "step": 28793 }, { "epoch": 2.98, "learning_rate": 8.671370511654356e-06, "loss": 0.8043, "step": 28794 }, { "epoch": 2.98, "learning_rate": 8.669639618742429e-06, "loss": 0.2245, "step": 28795 }, { "epoch": 2.98, "learning_rate": 8.667908869421501e-06, "loss": 0.1709, "step": 28796 }, { "epoch": 2.98, "learning_rate": 8.666178263703227e-06, "loss": 0.2872, "step": 28797 }, { "epoch": 2.98, "learning_rate": 8.66444780159924e-06, "loss": 0.3703, "step": 28798 }, { "epoch": 2.98, "learning_rate": 8.662717483121197e-06, "loss": 0.0906, "step": 28799 }, { "epoch": 2.98, "learning_rate": 8.66098730828075e-06, "loss": 0.4737, "step": 28800 }, { "epoch": 2.98, "learning_rate": 8.659257277089542e-06, "loss": 0.1462, "step": 28801 }, { "epoch": 2.98, "learning_rate": 8.657527389559223e-06, "loss": 0.2437, "step": 28802 }, { "epoch": 2.98, "learning_rate": 8.655797645701431e-06, "loss": 0.2776, "step": 28803 }, { "epoch": 2.98, "learning_rate": 8.65406804552783e-06, "loss": 0.0971, "step": 28804 }, { "epoch": 2.98, "learning_rate": 8.652338589050035e-06, "loss": 0.6501, "step": 28805 }, { "epoch": 2.98, "learning_rate": 8.650609276279706e-06, "loss": 0.5468, "step": 28806 }, { "epoch": 2.98, "learning_rate": 8.648880107228472e-06, "loss": 0.9213, "step": 28807 }, { "epoch": 2.98, "learning_rate": 8.647151081907988e-06, "loss": 0.0636, "step": 28808 }, { "epoch": 2.98, "learning_rate": 8.645422200329884e-06, "loss": 0.1223, "step": 28809 }, { "epoch": 2.98, "learning_rate": 8.643693462505813e-06, "loss": 0.8004, "step": 28810 }, { "epoch": 2.98, "learning_rate": 8.641964868447391e-06, "loss": 0.845, "step": 28811 }, { "epoch": 2.98, "learning_rate": 8.640236418166262e-06, "loss": 0.2083, "step": 28812 }, { "epoch": 2.98, "learning_rate": 8.638508111674064e-06, "loss": 0.9583, "step": 28813 }, { "epoch": 2.98, "learning_rate": 8.636779948982444e-06, "loss": 0.5002, "step": 28814 }, { "epoch": 2.98, "learning_rate": 8.635051930103009e-06, "loss": 0.4107, "step": 28815 }, { "epoch": 2.98, "learning_rate": 8.633324055047407e-06, "loss": 0.8869, "step": 28816 }, { "epoch": 2.98, "learning_rate": 8.63159632382727e-06, "loss": 0.6505, "step": 28817 }, { "epoch": 2.98, "learning_rate": 8.629868736454223e-06, "loss": 0.1103, "step": 28818 }, { "epoch": 2.98, "learning_rate": 8.6281412929399e-06, "loss": 0.2177, "step": 28819 }, { "epoch": 2.98, "learning_rate": 8.62641399329593e-06, "loss": 0.6732, "step": 28820 }, { "epoch": 2.98, "learning_rate": 8.624686837533944e-06, "loss": 0.294, "step": 28821 }, { "epoch": 2.98, "learning_rate": 8.62295982566557e-06, "loss": 0.3163, "step": 28822 }, { "epoch": 2.98, "learning_rate": 8.621232957702419e-06, "loss": 0.4942, "step": 28823 }, { "epoch": 2.98, "learning_rate": 8.619506233656135e-06, "loss": 0.1179, "step": 28824 }, { "epoch": 2.98, "learning_rate": 8.617779653538323e-06, "loss": 0.3603, "step": 28825 }, { "epoch": 2.98, "learning_rate": 8.61605321736062e-06, "loss": 0.4653, "step": 28826 }, { "epoch": 2.98, "learning_rate": 8.614326925134638e-06, "loss": 0.4614, "step": 28827 }, { "epoch": 2.98, "learning_rate": 8.612600776872005e-06, "loss": 0.1746, "step": 28828 }, { "epoch": 2.98, "learning_rate": 8.610874772584338e-06, "loss": 0.178, "step": 28829 }, { "epoch": 2.98, "learning_rate": 8.60914891228326e-06, "loss": 0.4678, "step": 28830 }, { "epoch": 2.98, "learning_rate": 8.607423195980384e-06, "loss": 1.0606, "step": 28831 }, { "epoch": 2.98, "learning_rate": 8.60569762368734e-06, "loss": 0.1943, "step": 28832 }, { "epoch": 2.98, "learning_rate": 8.603972195415723e-06, "loss": 0.4997, "step": 28833 }, { "epoch": 2.98, "learning_rate": 8.60224691117716e-06, "loss": 0.9226, "step": 28834 }, { "epoch": 2.98, "learning_rate": 8.600521770983276e-06, "loss": 0.2162, "step": 28835 }, { "epoch": 2.98, "learning_rate": 8.598796774845657e-06, "loss": 0.1827, "step": 28836 }, { "epoch": 2.99, "learning_rate": 8.597071922775938e-06, "loss": 0.2856, "step": 28837 }, { "epoch": 2.99, "learning_rate": 8.595347214785716e-06, "loss": 0.2204, "step": 28838 }, { "epoch": 2.99, "learning_rate": 8.593622650886614e-06, "loss": 0.157, "step": 28839 }, { "epoch": 2.99, "learning_rate": 8.59189823109023e-06, "loss": 0.4228, "step": 28840 }, { "epoch": 2.99, "learning_rate": 8.590173955408181e-06, "loss": 0.131, "step": 28841 }, { "epoch": 2.99, "learning_rate": 8.588449823852084e-06, "loss": 1.8296, "step": 28842 }, { "epoch": 2.99, "learning_rate": 8.586725836433518e-06, "loss": 0.19, "step": 28843 }, { "epoch": 2.99, "learning_rate": 8.585001993164107e-06, "loss": 0.2872, "step": 28844 }, { "epoch": 2.99, "learning_rate": 8.583278294055451e-06, "loss": 0.168, "step": 28845 }, { "epoch": 2.99, "learning_rate": 8.581554739119153e-06, "loss": 0.6304, "step": 28846 }, { "epoch": 2.99, "learning_rate": 8.579831328366827e-06, "loss": 0.2094, "step": 28847 }, { "epoch": 2.99, "learning_rate": 8.578108061810054e-06, "loss": 0.8048, "step": 28848 }, { "epoch": 2.99, "learning_rate": 8.576384939460448e-06, "loss": 0.6768, "step": 28849 }, { "epoch": 2.99, "learning_rate": 8.574661961329606e-06, "loss": 0.3073, "step": 28850 }, { "epoch": 2.99, "learning_rate": 8.572939127429125e-06, "loss": 0.134, "step": 28851 }, { "epoch": 2.99, "learning_rate": 8.571216437770614e-06, "loss": 0.385, "step": 28852 }, { "epoch": 2.99, "learning_rate": 8.569493892365652e-06, "loss": 0.4149, "step": 28853 }, { "epoch": 2.99, "learning_rate": 8.56777149122584e-06, "loss": 0.2305, "step": 28854 }, { "epoch": 2.99, "learning_rate": 8.566049234362776e-06, "loss": 0.385, "step": 28855 }, { "epoch": 2.99, "learning_rate": 8.564327121788053e-06, "loss": 0.1845, "step": 28856 }, { "epoch": 2.99, "learning_rate": 8.562605153513267e-06, "loss": 0.188, "step": 28857 }, { "epoch": 2.99, "learning_rate": 8.560883329550006e-06, "loss": 0.4047, "step": 28858 }, { "epoch": 2.99, "learning_rate": 8.559161649909862e-06, "loss": 0.1313, "step": 28859 }, { "epoch": 2.99, "learning_rate": 8.557440114604435e-06, "loss": 0.6436, "step": 28860 }, { "epoch": 2.99, "learning_rate": 8.555718723645293e-06, "loss": 0.1531, "step": 28861 }, { "epoch": 2.99, "learning_rate": 8.553997477044043e-06, "loss": 0.4466, "step": 28862 }, { "epoch": 2.99, "learning_rate": 8.552276374812257e-06, "loss": 0.4605, "step": 28863 }, { "epoch": 2.99, "learning_rate": 8.550555416961527e-06, "loss": 0.2835, "step": 28864 }, { "epoch": 2.99, "learning_rate": 8.548834603503443e-06, "loss": 0.3117, "step": 28865 }, { "epoch": 2.99, "learning_rate": 8.54711393444958e-06, "loss": 0.5349, "step": 28866 }, { "epoch": 2.99, "learning_rate": 8.54539340981153e-06, "loss": 0.3039, "step": 28867 }, { "epoch": 2.99, "learning_rate": 8.54367302960087e-06, "loss": 0.3735, "step": 28868 }, { "epoch": 2.99, "learning_rate": 8.541952793829182e-06, "loss": 0.2102, "step": 28869 }, { "epoch": 2.99, "learning_rate": 8.540232702508058e-06, "loss": 0.4077, "step": 28870 }, { "epoch": 2.99, "learning_rate": 8.538512755649055e-06, "loss": 0.8678, "step": 28871 }, { "epoch": 2.99, "learning_rate": 8.536792953263773e-06, "loss": 0.6143, "step": 28872 }, { "epoch": 2.99, "learning_rate": 8.535073295363772e-06, "loss": 0.2192, "step": 28873 }, { "epoch": 2.99, "learning_rate": 8.533353781960634e-06, "loss": 0.3379, "step": 28874 }, { "epoch": 2.99, "learning_rate": 8.531634413065935e-06, "loss": 0.1986, "step": 28875 }, { "epoch": 2.99, "learning_rate": 8.529915188691248e-06, "loss": 0.4298, "step": 28876 }, { "epoch": 2.99, "learning_rate": 8.528196108848148e-06, "loss": 0.6665, "step": 28877 }, { "epoch": 2.99, "learning_rate": 8.526477173548212e-06, "loss": 0.3466, "step": 28878 }, { "epoch": 2.99, "learning_rate": 8.524758382803003e-06, "loss": 0.3587, "step": 28879 }, { "epoch": 2.99, "learning_rate": 8.523039736624108e-06, "loss": 0.0888, "step": 28880 }, { "epoch": 2.99, "learning_rate": 8.521321235023072e-06, "loss": 0.521, "step": 28881 }, { "epoch": 2.99, "learning_rate": 8.519602878011477e-06, "loss": 0.0811, "step": 28882 }, { "epoch": 2.99, "learning_rate": 8.517884665600894e-06, "loss": 0.1947, "step": 28883 }, { "epoch": 2.99, "learning_rate": 8.516166597802878e-06, "loss": 0.8828, "step": 28884 }, { "epoch": 2.99, "learning_rate": 8.514448674629015e-06, "loss": 0.243, "step": 28885 }, { "epoch": 2.99, "learning_rate": 8.512730896090845e-06, "loss": 0.3163, "step": 28886 }, { "epoch": 2.99, "learning_rate": 8.511013262199944e-06, "loss": 0.838, "step": 28887 }, { "epoch": 2.99, "learning_rate": 8.509295772967873e-06, "loss": 0.4563, "step": 28888 }, { "epoch": 2.99, "learning_rate": 8.507578428406198e-06, "loss": 0.5491, "step": 28889 }, { "epoch": 2.99, "learning_rate": 8.50586122852648e-06, "loss": 0.1765, "step": 28890 }, { "epoch": 2.99, "learning_rate": 8.504144173340269e-06, "loss": 0.203, "step": 28891 }, { "epoch": 2.99, "learning_rate": 8.502427262859129e-06, "loss": 0.2446, "step": 28892 }, { "epoch": 2.99, "learning_rate": 8.500710497094621e-06, "loss": 0.5829, "step": 28893 }, { "epoch": 2.99, "learning_rate": 8.498993876058297e-06, "loss": 0.2115, "step": 28894 }, { "epoch": 2.99, "learning_rate": 8.497277399761715e-06, "loss": 0.1201, "step": 28895 }, { "epoch": 2.99, "learning_rate": 8.49556106821643e-06, "loss": 0.0827, "step": 28896 }, { "epoch": 2.99, "learning_rate": 8.493844881433998e-06, "loss": 0.4522, "step": 28897 }, { "epoch": 2.99, "learning_rate": 8.49212883942598e-06, "loss": 0.8585, "step": 28898 }, { "epoch": 2.99, "learning_rate": 8.490412942203906e-06, "loss": 0.4611, "step": 28899 }, { "epoch": 2.99, "learning_rate": 8.48869718977935e-06, "loss": 0.3366, "step": 28900 }, { "epoch": 2.99, "learning_rate": 8.486981582163842e-06, "loss": 0.1318, "step": 28901 }, { "epoch": 2.99, "learning_rate": 8.485266119368943e-06, "loss": 0.084, "step": 28902 }, { "epoch": 2.99, "learning_rate": 8.483550801406194e-06, "loss": 0.2465, "step": 28903 }, { "epoch": 2.99, "learning_rate": 8.481835628287147e-06, "loss": 0.084, "step": 28904 }, { "epoch": 2.99, "learning_rate": 8.48012060002335e-06, "loss": 0.5136, "step": 28905 }, { "epoch": 2.99, "learning_rate": 8.478405716626347e-06, "loss": 0.2093, "step": 28906 }, { "epoch": 2.99, "learning_rate": 8.476690978107678e-06, "loss": 0.4975, "step": 28907 }, { "epoch": 2.99, "learning_rate": 8.4749763844789e-06, "loss": 0.0921, "step": 28908 }, { "epoch": 2.99, "learning_rate": 8.473261935751535e-06, "loss": 0.4096, "step": 28909 }, { "epoch": 2.99, "learning_rate": 8.471547631937143e-06, "loss": 0.1266, "step": 28910 }, { "epoch": 2.99, "learning_rate": 8.469833473047245e-06, "loss": 0.5366, "step": 28911 }, { "epoch": 2.99, "learning_rate": 8.468119459093394e-06, "loss": 0.5862, "step": 28912 }, { "epoch": 2.99, "learning_rate": 8.46640559008712e-06, "loss": 0.7297, "step": 28913 }, { "epoch": 2.99, "learning_rate": 8.464691866039966e-06, "loss": 0.0876, "step": 28914 }, { "epoch": 2.99, "learning_rate": 8.462978286963472e-06, "loss": 0.1513, "step": 28915 }, { "epoch": 2.99, "learning_rate": 8.461264852869163e-06, "loss": 0.5273, "step": 28916 }, { "epoch": 2.99, "learning_rate": 8.459551563768582e-06, "loss": 0.3318, "step": 28917 }, { "epoch": 2.99, "learning_rate": 8.45783841967327e-06, "loss": 0.1309, "step": 28918 }, { "epoch": 2.99, "learning_rate": 8.456125420594739e-06, "loss": 0.1636, "step": 28919 }, { "epoch": 2.99, "learning_rate": 8.454412566544529e-06, "loss": 0.281, "step": 28920 }, { "epoch": 2.99, "learning_rate": 8.452699857534177e-06, "loss": 0.5071, "step": 28921 }, { "epoch": 2.99, "learning_rate": 8.450987293575203e-06, "loss": 0.3043, "step": 28922 }, { "epoch": 2.99, "learning_rate": 8.449274874679153e-06, "loss": 0.2706, "step": 28923 }, { "epoch": 2.99, "learning_rate": 8.44756260085753e-06, "loss": 0.1371, "step": 28924 }, { "epoch": 2.99, "learning_rate": 8.445850472121877e-06, "loss": 0.717, "step": 28925 }, { "epoch": 2.99, "learning_rate": 8.444138488483711e-06, "loss": 0.1187, "step": 28926 }, { "epoch": 2.99, "learning_rate": 8.442426649954562e-06, "loss": 0.2918, "step": 28927 }, { "epoch": 2.99, "learning_rate": 8.440714956545963e-06, "loss": 0.3589, "step": 28928 }, { "epoch": 2.99, "learning_rate": 8.439003408269416e-06, "loss": 0.659, "step": 28929 }, { "epoch": 2.99, "learning_rate": 8.437292005136455e-06, "loss": 0.3889, "step": 28930 }, { "epoch": 2.99, "learning_rate": 8.435580747158596e-06, "loss": 0.0678, "step": 28931 }, { "epoch": 2.99, "learning_rate": 8.433869634347364e-06, "loss": 0.5063, "step": 28932 }, { "epoch": 3.0, "learning_rate": 8.432158666714278e-06, "loss": 0.5624, "step": 28933 }, { "epoch": 3.0, "learning_rate": 8.430447844270849e-06, "loss": 0.1289, "step": 28934 }, { "epoch": 3.0, "learning_rate": 8.428737167028611e-06, "loss": 0.2378, "step": 28935 }, { "epoch": 3.0, "learning_rate": 8.427026634999054e-06, "loss": 0.226, "step": 28936 }, { "epoch": 3.0, "learning_rate": 8.425316248193707e-06, "loss": 0.0974, "step": 28937 }, { "epoch": 3.0, "learning_rate": 8.423606006624094e-06, "loss": 0.1888, "step": 28938 }, { "epoch": 3.0, "learning_rate": 8.421895910301708e-06, "loss": 0.4909, "step": 28939 }, { "epoch": 3.0, "learning_rate": 8.420185959238067e-06, "loss": 0.4801, "step": 28940 }, { "epoch": 3.0, "learning_rate": 8.418476153444689e-06, "loss": 0.0707, "step": 28941 }, { "epoch": 3.0, "learning_rate": 8.416766492933078e-06, "loss": 0.535, "step": 28942 }, { "epoch": 3.0, "learning_rate": 8.415056977714742e-06, "loss": 0.2136, "step": 28943 }, { "epoch": 3.0, "learning_rate": 8.413347607801197e-06, "loss": 0.2207, "step": 28944 }, { "epoch": 3.0, "learning_rate": 8.41163838320394e-06, "loss": 0.3121, "step": 28945 }, { "epoch": 3.0, "learning_rate": 8.409929303934495e-06, "loss": 1.1242, "step": 28946 }, { "epoch": 3.0, "learning_rate": 8.408220370004342e-06, "loss": 0.328, "step": 28947 }, { "epoch": 3.0, "learning_rate": 8.40651158142501e-06, "loss": 0.4964, "step": 28948 }, { "epoch": 3.0, "learning_rate": 8.404802938207979e-06, "loss": 0.3867, "step": 28949 }, { "epoch": 3.0, "learning_rate": 8.40309444036476e-06, "loss": 0.3791, "step": 28950 }, { "epoch": 3.0, "learning_rate": 8.401386087906855e-06, "loss": 0.2255, "step": 28951 }, { "epoch": 3.0, "learning_rate": 8.399677880845771e-06, "loss": 0.3984, "step": 28952 }, { "epoch": 3.0, "learning_rate": 8.397969819192997e-06, "loss": 0.1562, "step": 28953 }, { "epoch": 3.0, "learning_rate": 8.396261902960037e-06, "loss": 0.4043, "step": 28954 }, { "epoch": 3.0, "learning_rate": 8.394554132158387e-06, "loss": 0.3326, "step": 28955 }, { "epoch": 3.0, "learning_rate": 8.392846506799551e-06, "loss": 0.5321, "step": 28956 }, { "epoch": 3.0, "learning_rate": 8.391139026895009e-06, "loss": 0.716, "step": 28957 }, { "epoch": 3.0, "learning_rate": 8.389431692456264e-06, "loss": 0.0619, "step": 28958 }, { "epoch": 3.0, "learning_rate": 8.387724503494807e-06, "loss": 0.3, "step": 28959 }, { "epoch": 3.0, "learning_rate": 8.38601746002214e-06, "loss": 1.1805, "step": 28960 }, { "epoch": 3.0, "learning_rate": 8.384310562049736e-06, "loss": 0.0736, "step": 28961 }, { "epoch": 3.0, "learning_rate": 8.382603809589098e-06, "loss": 0.5531, "step": 28962 }, { "epoch": 3.0, "learning_rate": 8.380897202651713e-06, "loss": 0.1973, "step": 28963 }, { "epoch": 3.0, "learning_rate": 8.379190741249068e-06, "loss": 0.0681, "step": 28964 }, { "epoch": 3.0, "learning_rate": 8.377484425392657e-06, "loss": 0.5215, "step": 28965 }, { "epoch": 3.0, "learning_rate": 8.375778255093967e-06, "loss": 0.1923, "step": 28966 }, { "epoch": 3.0, "learning_rate": 8.374072230364468e-06, "loss": 0.1598, "step": 28967 }, { "epoch": 3.0, "learning_rate": 8.372366351215654e-06, "loss": 0.3123, "step": 28968 }, { "epoch": 3.0, "learning_rate": 8.370660617659014e-06, "loss": 0.3517, "step": 28969 }, { "epoch": 3.0, "learning_rate": 8.368955029706022e-06, "loss": 0.348, "step": 28970 }, { "epoch": 3.0, "learning_rate": 8.367249587368164e-06, "loss": 0.1987, "step": 28971 }, { "epoch": 3.0, "learning_rate": 8.36554429065692e-06, "loss": 0.3144, "step": 28972 }, { "epoch": 3.0, "learning_rate": 8.36383913958378e-06, "loss": 0.0595, "step": 28973 }, { "epoch": 3.0, "learning_rate": 8.3621341341602e-06, "loss": 0.5679, "step": 28974 }, { "epoch": 3.0, "learning_rate": 8.360429274397673e-06, "loss": 0.7464, "step": 28975 }, { "epoch": 3.0, "learning_rate": 8.358724560307684e-06, "loss": 0.9163, "step": 28976 }, { "epoch": 3.0, "learning_rate": 8.357019991901687e-06, "loss": 0.2411, "step": 28977 }, { "epoch": 3.0, "learning_rate": 8.355315569191164e-06, "loss": 0.1606, "step": 28978 }, { "epoch": 3.0, "learning_rate": 8.353611292187596e-06, "loss": 0.797, "step": 28979 }, { "epoch": 3.0, "learning_rate": 8.35190716090245e-06, "loss": 0.3779, "step": 28980 }, { "epoch": 3.0, "learning_rate": 8.350203175347202e-06, "loss": 0.4446, "step": 28981 }, { "epoch": 3.0, "learning_rate": 8.348499335533319e-06, "loss": 0.3203, "step": 28982 }, { "epoch": 3.0, "learning_rate": 8.346795641472273e-06, "loss": 0.4522, "step": 28983 }, { "epoch": 3.0, "learning_rate": 8.345092093175539e-06, "loss": 0.3336, "step": 28984 }, { "epoch": 3.0, "learning_rate": 8.343388690654573e-06, "loss": 0.0502, "step": 28985 }, { "epoch": 3.0, "learning_rate": 8.341685433920853e-06, "loss": 0.122, "step": 28986 }, { "epoch": 3.0, "learning_rate": 8.339982322985832e-06, "loss": 0.6417, "step": 28987 }, { "epoch": 3.0, "learning_rate": 8.338279357860982e-06, "loss": 0.8223, "step": 28988 }, { "epoch": 3.0, "learning_rate": 8.336576538557763e-06, "loss": 0.2912, "step": 28989 }, { "epoch": 3.0, "learning_rate": 8.334873865087646e-06, "loss": 0.5418, "step": 28990 }, { "epoch": 3.0, "learning_rate": 8.333171337462086e-06, "loss": 0.3716, "step": 28991 }, { "epoch": 3.0, "learning_rate": 8.33146895569255e-06, "loss": 0.2836, "step": 28992 }, { "epoch": 3.0, "learning_rate": 8.329766719790492e-06, "loss": 0.6629, "step": 28993 }, { "epoch": 3.0, "learning_rate": 8.328064629767388e-06, "loss": 0.7067, "step": 28994 }, { "epoch": 3.0, "learning_rate": 8.326362685634668e-06, "loss": 0.2493, "step": 28995 }, { "epoch": 3.0, "learning_rate": 8.324660887403805e-06, "loss": 0.0869, "step": 28996 }, { "epoch": 3.0, "learning_rate": 8.322959235086255e-06, "loss": 0.4057, "step": 28997 }, { "epoch": 3.0, "learning_rate": 8.321257728693476e-06, "loss": 0.082, "step": 28998 }, { "epoch": 3.0, "learning_rate": 8.319556368236912e-06, "loss": 0.6988, "step": 28999 }, { "epoch": 3.0, "learning_rate": 8.31785515372802e-06, "loss": 0.8692, "step": 29000 }, { "epoch": 3.0, "learning_rate": 8.316154085178258e-06, "loss": 0.1487, "step": 29001 }, { "epoch": 3.0, "learning_rate": 8.31445316259907e-06, "loss": 0.2824, "step": 29002 }, { "epoch": 3.0, "learning_rate": 8.312752386001908e-06, "loss": 0.5413, "step": 29003 }, { "epoch": 3.0, "learning_rate": 8.311051755398231e-06, "loss": 0.1587, "step": 29004 }, { "epoch": 3.0, "learning_rate": 8.309351270799472e-06, "loss": 0.4331, "step": 29005 }, { "epoch": 3.0, "learning_rate": 8.307650932217086e-06, "loss": 0.2301, "step": 29006 }, { "epoch": 3.0, "learning_rate": 8.305950739662515e-06, "loss": 0.3754, "step": 29007 }, { "epoch": 3.0, "learning_rate": 8.30425069314721e-06, "loss": 0.0889, "step": 29008 }, { "epoch": 3.0, "learning_rate": 8.302550792682611e-06, "loss": 0.4058, "step": 29009 }, { "epoch": 3.0, "learning_rate": 8.300851038280166e-06, "loss": 0.2694, "step": 29010 }, { "epoch": 3.0, "learning_rate": 8.299151429951323e-06, "loss": 0.3341, "step": 29011 }, { "epoch": 3.0, "learning_rate": 8.297451967707506e-06, "loss": 0.0989, "step": 29012 }, { "epoch": 3.0, "learning_rate": 8.295752651560162e-06, "loss": 0.1516, "step": 29013 }, { "epoch": 3.0, "learning_rate": 8.294053481520745e-06, "loss": 0.2691, "step": 29014 }, { "epoch": 3.0, "learning_rate": 8.29235445760067e-06, "loss": 0.1666, "step": 29015 }, { "epoch": 3.0, "learning_rate": 8.290655579811383e-06, "loss": 0.5097, "step": 29016 }, { "epoch": 3.0, "learning_rate": 8.288956848164329e-06, "loss": 0.24, "step": 29017 }, { "epoch": 3.0, "learning_rate": 8.287258262670932e-06, "loss": 0.5937, "step": 29018 }, { "epoch": 3.0, "learning_rate": 8.285559823342637e-06, "loss": 0.1716, "step": 29019 }, { "epoch": 3.0, "learning_rate": 8.283861530190872e-06, "loss": 0.0745, "step": 29020 }, { "epoch": 3.0, "learning_rate": 8.282163383227066e-06, "loss": 0.5314, "step": 29021 }, { "epoch": 3.0, "learning_rate": 8.280465382462668e-06, "loss": 0.4901, "step": 29022 }, { "epoch": 3.0, "learning_rate": 8.278767527909084e-06, "loss": 0.5476, "step": 29023 }, { "epoch": 3.0, "learning_rate": 8.277069819577761e-06, "loss": 0.0756, "step": 29024 }, { "epoch": 3.0, "learning_rate": 8.275372257480115e-06, "loss": 0.2558, "step": 29025 }, { "epoch": 3.0, "learning_rate": 8.273674841627582e-06, "loss": 0.0536, "step": 29026 }, { "epoch": 3.0, "learning_rate": 8.271977572031583e-06, "loss": 1.2708, "step": 29027 }, { "epoch": 3.01, "learning_rate": 8.270280448703545e-06, "loss": 0.1378, "step": 29028 }, { "epoch": 3.01, "learning_rate": 8.2685834716549e-06, "loss": 0.5768, "step": 29029 }, { "epoch": 3.01, "learning_rate": 8.266886640897065e-06, "loss": 0.3766, "step": 29030 }, { "epoch": 3.01, "learning_rate": 8.265189956441462e-06, "loss": 0.5709, "step": 29031 }, { "epoch": 3.01, "learning_rate": 8.263493418299525e-06, "loss": 0.7256, "step": 29032 }, { "epoch": 3.01, "learning_rate": 8.261797026482652e-06, "loss": 0.3915, "step": 29033 }, { "epoch": 3.01, "learning_rate": 8.26010078100228e-06, "loss": 0.3511, "step": 29034 }, { "epoch": 3.01, "learning_rate": 8.258404681869821e-06, "loss": 0.1024, "step": 29035 }, { "epoch": 3.01, "learning_rate": 8.256708729096703e-06, "loss": 0.4298, "step": 29036 }, { "epoch": 3.01, "learning_rate": 8.255012922694325e-06, "loss": 0.1664, "step": 29037 }, { "epoch": 3.01, "learning_rate": 8.253317262674113e-06, "loss": 0.2672, "step": 29038 }, { "epoch": 3.01, "learning_rate": 8.251621749047481e-06, "loss": 0.3555, "step": 29039 }, { "epoch": 3.01, "learning_rate": 8.24992638182584e-06, "loss": 0.1487, "step": 29040 }, { "epoch": 3.01, "learning_rate": 8.248231161020609e-06, "loss": 0.3391, "step": 29041 }, { "epoch": 3.01, "learning_rate": 8.246536086643205e-06, "loss": 0.2809, "step": 29042 }, { "epoch": 3.01, "learning_rate": 8.244841158705021e-06, "loss": 0.3788, "step": 29043 }, { "epoch": 3.01, "learning_rate": 8.243146377217475e-06, "loss": 0.2448, "step": 29044 }, { "epoch": 3.01, "learning_rate": 8.24145174219198e-06, "loss": 0.512, "step": 29045 }, { "epoch": 3.01, "learning_rate": 8.239757253639936e-06, "loss": 0.4202, "step": 29046 }, { "epoch": 3.01, "learning_rate": 8.238062911572758e-06, "loss": 0.4914, "step": 29047 }, { "epoch": 3.01, "learning_rate": 8.236368716001853e-06, "loss": 0.5377, "step": 29048 }, { "epoch": 3.01, "learning_rate": 8.234674666938618e-06, "loss": 0.4879, "step": 29049 }, { "epoch": 3.01, "learning_rate": 8.232980764394457e-06, "loss": 0.2508, "step": 29050 }, { "epoch": 3.01, "learning_rate": 8.231287008380777e-06, "loss": 0.3398, "step": 29051 }, { "epoch": 3.01, "learning_rate": 8.22959339890899e-06, "loss": 0.3842, "step": 29052 }, { "epoch": 3.01, "learning_rate": 8.227899935990475e-06, "loss": 0.1227, "step": 29053 }, { "epoch": 3.01, "learning_rate": 8.226206619636647e-06, "loss": 0.033, "step": 29054 }, { "epoch": 3.01, "learning_rate": 8.224513449858896e-06, "loss": 0.2082, "step": 29055 }, { "epoch": 3.01, "learning_rate": 8.222820426668632e-06, "loss": 0.226, "step": 29056 }, { "epoch": 3.01, "learning_rate": 8.22112755007724e-06, "loss": 0.357, "step": 29057 }, { "epoch": 3.01, "learning_rate": 8.219434820096123e-06, "loss": 0.5347, "step": 29058 }, { "epoch": 3.01, "learning_rate": 8.217742236736677e-06, "loss": 0.5777, "step": 29059 }, { "epoch": 3.01, "learning_rate": 8.2160498000103e-06, "loss": 1.6517, "step": 29060 }, { "epoch": 3.01, "learning_rate": 8.214357509928369e-06, "loss": 0.4337, "step": 29061 }, { "epoch": 3.01, "learning_rate": 8.212665366502297e-06, "loss": 0.0892, "step": 29062 }, { "epoch": 3.01, "learning_rate": 8.210973369743455e-06, "loss": 0.4374, "step": 29063 }, { "epoch": 3.01, "learning_rate": 8.20928151966324e-06, "loss": 0.2638, "step": 29064 }, { "epoch": 3.01, "learning_rate": 8.207589816273045e-06, "loss": 0.1615, "step": 29065 }, { "epoch": 3.01, "learning_rate": 8.205898259584256e-06, "loss": 0.1932, "step": 29066 }, { "epoch": 3.01, "learning_rate": 8.204206849608261e-06, "loss": 0.5228, "step": 29067 }, { "epoch": 3.01, "learning_rate": 8.202515586356444e-06, "loss": 0.5189, "step": 29068 }, { "epoch": 3.01, "learning_rate": 8.200824469840192e-06, "loss": 0.2437, "step": 29069 }, { "epoch": 3.01, "learning_rate": 8.199133500070901e-06, "loss": 0.5168, "step": 29070 }, { "epoch": 3.01, "learning_rate": 8.197442677059929e-06, "loss": 0.1645, "step": 29071 }, { "epoch": 3.01, "learning_rate": 8.195752000818677e-06, "loss": 0.2576, "step": 29072 }, { "epoch": 3.01, "learning_rate": 8.194061471358524e-06, "loss": 0.4092, "step": 29073 }, { "epoch": 3.01, "learning_rate": 8.19237108869084e-06, "loss": 0.4968, "step": 29074 }, { "epoch": 3.01, "learning_rate": 8.190680852827011e-06, "loss": 0.427, "step": 29075 }, { "epoch": 3.01, "learning_rate": 8.188990763778416e-06, "loss": 0.3886, "step": 29076 }, { "epoch": 3.01, "learning_rate": 8.187300821556433e-06, "loss": 0.1652, "step": 29077 }, { "epoch": 3.01, "learning_rate": 8.185611026172437e-06, "loss": 0.239, "step": 29078 }, { "epoch": 3.01, "learning_rate": 8.1839213776378e-06, "loss": 0.4563, "step": 29079 }, { "epoch": 3.01, "learning_rate": 8.18223187596391e-06, "loss": 0.067, "step": 29080 }, { "epoch": 3.01, "learning_rate": 8.180542521162122e-06, "loss": 0.1307, "step": 29081 }, { "epoch": 3.01, "learning_rate": 8.178853313243821e-06, "loss": 0.3786, "step": 29082 }, { "epoch": 3.01, "learning_rate": 8.177164252220367e-06, "loss": 0.5054, "step": 29083 }, { "epoch": 3.01, "learning_rate": 8.175475338103142e-06, "loss": 0.3285, "step": 29084 }, { "epoch": 3.01, "learning_rate": 8.173786570903506e-06, "loss": 0.2098, "step": 29085 }, { "epoch": 3.01, "learning_rate": 8.172097950632843e-06, "loss": 0.2411, "step": 29086 }, { "epoch": 3.01, "learning_rate": 8.170409477302502e-06, "loss": 0.1017, "step": 29087 }, { "epoch": 3.01, "learning_rate": 8.168721150923854e-06, "loss": 0.1187, "step": 29088 }, { "epoch": 3.01, "learning_rate": 8.167032971508267e-06, "loss": 0.2257, "step": 29089 }, { "epoch": 3.01, "learning_rate": 8.165344939067117e-06, "loss": 0.6455, "step": 29090 }, { "epoch": 3.01, "learning_rate": 8.163657053611743e-06, "loss": 0.3643, "step": 29091 }, { "epoch": 3.01, "learning_rate": 8.161969315153524e-06, "loss": 1.1232, "step": 29092 }, { "epoch": 3.01, "learning_rate": 8.160281723703817e-06, "loss": 0.4329, "step": 29093 }, { "epoch": 3.01, "learning_rate": 8.158594279273982e-06, "loss": 0.6458, "step": 29094 }, { "epoch": 3.01, "learning_rate": 8.156906981875381e-06, "loss": 0.655, "step": 29095 }, { "epoch": 3.01, "learning_rate": 8.155219831519373e-06, "loss": 0.3511, "step": 29096 }, { "epoch": 3.01, "learning_rate": 8.15353282821731e-06, "loss": 0.157, "step": 29097 }, { "epoch": 3.01, "learning_rate": 8.151845971980563e-06, "loss": 0.4863, "step": 29098 }, { "epoch": 3.01, "learning_rate": 8.150159262820471e-06, "loss": 0.3058, "step": 29099 }, { "epoch": 3.01, "learning_rate": 8.148472700748399e-06, "loss": 0.0876, "step": 29100 }, { "epoch": 3.01, "learning_rate": 8.146786285775689e-06, "loss": 0.0816, "step": 29101 }, { "epoch": 3.01, "learning_rate": 8.145100017913701e-06, "loss": 0.2839, "step": 29102 }, { "epoch": 3.01, "learning_rate": 8.143413897173784e-06, "loss": 0.1981, "step": 29103 }, { "epoch": 3.01, "learning_rate": 8.141727923567293e-06, "loss": 0.2943, "step": 29104 }, { "epoch": 3.01, "learning_rate": 8.140042097105571e-06, "loss": 0.2007, "step": 29105 }, { "epoch": 3.01, "learning_rate": 8.138356417799973e-06, "loss": 0.5914, "step": 29106 }, { "epoch": 3.01, "learning_rate": 8.136670885661848e-06, "loss": 0.4443, "step": 29107 }, { "epoch": 3.01, "learning_rate": 8.134985500702544e-06, "loss": 0.3902, "step": 29108 }, { "epoch": 3.01, "learning_rate": 8.133300262933395e-06, "loss": 0.3343, "step": 29109 }, { "epoch": 3.01, "learning_rate": 8.131615172365748e-06, "loss": 0.3515, "step": 29110 }, { "epoch": 3.01, "learning_rate": 8.129930229010965e-06, "loss": 0.1515, "step": 29111 } ], "logging_steps": 1, "max_steps": 38300, "num_train_epochs": 4, "save_steps": 500, "total_flos": 2.2076239138620703e+18, "trial_name": null, "trial_params": null }